Assessing the sustainability of brackish-water aquaculture

systems in the Philippines

Final report on DFID/AFGRP research project R8288

Xavier Irz and James R. Stevenson

University of Reading, Department of Agricultural & Food Economics

Contributors:

Rose-Glenda Alcade
Philippine Council for Aquatic and Marine Research and Development (PCAMRD), Los
Banos, Philippines

Pierre Morissens
Coopération Internationale en Recherche Agronomique pour le Développement
(CIRAD), Equipe Aquaculture, Montpellier, France

Jean Petit
INRA, UMR Sol, Agronomie et Spatialisation de Rennes-Quimper, Rennes, France

Arnold Tanoy

Portia Villarante

Other Collaborators: Rafael D. Guerrero III (PCAMRD-DOST); Jurgenne Primavera

(SEAFDEC-AQD); Clarissa Rubio (UP Diliman)
ACKNOWLEDGMENTS

The authors would like to acknowledge the help of several people in designing and carrying out this project,
including Kristine-Joy Lapitan; James Muir (AFGRP-University of Stirling); Roger Pullin; and Tahir Rehman
(University of Reading). We are also grateful to all the participants at the workshop organized in Los Banõs on
22 April 2005 by the University and Reading and PCAMRD on the theme ‘aquaculture and poverty’ for their
insightful comments and suggestions. Finally, we would like to thank the numerous individuals who helped us in
collecting primary data for this project, including the municipal and barangay officials who allowed us to carry
out this research in their communities, as well as the residents of these communities who accepted to be
interviewed and to openly share their views and experiences with us.

DISCLAIMER

This document is an output from a project funded by the UK Department for International Development (DFID)
for the benefit of developing countries. The views expressed are not necessarily those of DFID.

ii
 TABLE OF CONTENTS

Executive summary 1

Background to the project 8

Project purpose 8

Research Activities 8

Research Outputs 8

Contribution of outputs to development impacts 9

List of outputs 10

Appendices 12

Paper 1 13

Paper 2 50

Paper 3 145

Paper 4 174

Paper 5 231

Project logical framework 261

Project proposal 263

iii
EXECUTIVE SUMMARY

1- The project comprises five separate papers investigating different aspects of brackish water aquaculture
in the Philippines. Papers 1 to 3 form the core of the project and progress towards the establishment of a
ranking of different farming systems in terms of their relative sustainability. Paper 4 represents an
extension of the original project proposal, as agreed with James Muir, which investigates in detail the
role that aquaculture plays in the lives of poor people in the coastal areas of the Philippines. Finally,
paper 5 investigates whether an efficiency case can be made for land reform in brackish water
aquaculture in the Philippines. Hence, it is somewhat peripheral to the project but relates to the socio-
economic analysis presented in paper 2, and in this regard was deemed worth including.

Paper Title Aim

number
1 A typology of brackish-water pond Identify a small number of farming
aquaculture systems in the Philippines systems for comparative analysis of
sustainability
2 Indicators of economic, ecological and Establish the relative performance of
socio-economic performance of farming systems in economic,
aquaculture systems ecological and social dimensions
3 Analysing trade-offs among indicators of Rank the farming systems identified
sustainability: a multi-criteria approach in paper 2 in terms of their relative
sustainability
4 Aquaculture and poverty – A case study Analyse the contributions that
of five coastal communities in the aquaculture makes to the livelihoods
Philippines of poor people in coastal areas
5 Is there an efficiency case for land Test for the presence of an inverse
redistribution in Philippine brackish water relationship between farm size and
aquaculture? efficiency, which would provide an
efficiency rationale for land
redistribution

2- We start by presenting the findings of the project and continue by formulating a number of
recommendations.

AQUACULTURE FARMING SYSTEMS (Paper 1)

3- The construction of typologies of farming systems using multivariate datasets is increasingly perceived in
the agriculture and, to a lesser extent, aquaculture literatures as being an important first step in
comparative analysis.

4- A survey of brackish-water pond aquaculture farms was carried out in Regions III and VI – the two
regions that account for the bulk of brackish-water aquaculture production in the Philippines. This
yielded a net dataset (after removing inconsistent respondents) of 136 farms. Based on eight variables
related to the use of labour, land and intermediate inputs (fertiliser, feed, fry), we identify three
underlying latent variables (components) that explain 68% of the total variance in the original data by
carrying out a principal components analysis. These three components (specialisation, land vs. labour
intensity, and feed intensity) are used in a cluster analysis to establish five farm types.

5- The five farm types are labelled as follows: extensive polyculture (type 1, n=54); semi-intensive prawn
polyculture (type 2, n=15); low-input labour-intensive systems (type 3, n=37); large, milkfish-oriented
systems (type 4, n=11); and semi-intensive milkfish monoculture (type 5, n=19). These represent
different combinations of production factors and characterise the full complexity of production systems
in the Philippines, where traditional practices are mixed with new technologies and techniques.

1
 6- This kind of typology offers greater realism than artificially imposed uni-dimensional classifications. For
example, the concept of “intensity” can only refer to a specific input (e.g. land, labour, feed) even though
the common division into extensive, semi-intensive and intensive systems requires a consideration of
combinations of inputs.

SUSTAINABILITY OF AQUACULTURE FARMING SYSTEMS (Paper 2)

7- Aquaculture in the Philippines provides important livelihood opportunities directly (i.e, on-farm) as well
as indirectly through multiplier employment effects that spread throughout the rural economy. Further, it
represents a potential net source of dietary protein for Filipinos, although it may actually reduce the stock
of protein because of inefficient farm management or a focus on culturing species situated high on the
aquatic food-web. There are also environmental issues associated with brackish-water aquaculture,
particularly nutrient enrichment of surrounding waters, and the appropriation of coastal resources in form
of wild fry and broodstock, and wild-caught feeds. Finally, aquaculture development has sometimes been
accused of increasing social inequity and the deprivation of coastal communities. Altogether, the sector
faces economic, social and environmental problems that may challenge its long-term sustainability and
provide the motivation for this research.

8- Sustainability, as conceived under the common “three pillars” or “weak sustainability” frameworks, is
considered to be achievable through the attainment of economic, social and ecological objectives in an
appropriate balance. In this context, comparative sustainability at the farm level can be analysed using a
number of indicators for specific objectives. We follow this approach to study farm-level sustainability
by comparing indicator results for each of the five farm types established in paper 1.

9- Indicators are developed for six farm-level objectives: profit maximisation, risk minimisation,
maximisation of technical efficiency, maximisation of employment effects, maximisation of net protein
production, and minimisation of nutrient loss.

10- Mean gross margin per hectare per year is used as an indicator of farm-level profit. We find that all farm
types are profitable on average, but with considerable heterogeneity. Semi-intensive prawn-oriented
polyculture (farm type 2) returns the highest level of profit on average at 59,100 pesos per hectare per
year. Extensive polyculture (farm type 1) has the second highest with 51,200 pesos per hectare per year.
The lowest level of profit per unit area is found for very large farms (farm type 4, 27,800 pesos per
hectare per year).

11- Risk is measured by the standard error of the mean gross margin for a given farm type, divided by the
mean gross margin. This measure conflates price and production risks and would ideally be measured
from panel data from different farms over a time-series. However, such data is not available and we are
restricted to considering variability in one time period only. As might be expected, the farm type that
returns the highest level of profit is also the most variable – semi-intensive prawn-oriented polyculture
(farm type 2). It would seem that operators of this kind of farming system attempt to mitigate their
production and price risks by diversifying revenue streams across four species (prawns, milkfish, tilapia
and crab) because revenue diversity, as measured by a Shannon index, is highest for this farm type.

12- Technical efficiency is estimated using a stochastic production frontier analysis. A favoured model is
specified and the technical efficiency effects of particular farm-level characteristics are identified. We
find that the following factors all reduce technical inefficiency: the operator lives on the farm; uses a
traditional lagum-lagum pattern of stock movement as well as probiotics; and tests the pH of the soil. Of
the five farm types, the semi-intensive prawn-oriented polyculture systems (type 2) are the most
inefficient at less than 50%, which means that they could potentially double their output without
increasing inputs. The other four farm types all fall in the 55% to 75% average efficient range.

13- The impact of aquaculture systems on employment is measured using an indicator that combines the on-
farm employment per unit area and the employment in associated industries immediately upstream and
downstream from the farm. This link is quantified using an assumed fixed technical coefficient for labour

2
 in each of these industries, and we use interviews with a number of businesses to generate the required
data. We find that 80% of the employment effect of aquaculture is through on-farm labour use, although
the downstream linkages are underrepresented in the analysis. On average, the total employment effect is
260 person days per hectare per year. However, there is heterogeneity that can be strongly linked to farm
type. Low-input labour-intensive systems (farm type 3) provide the most employment with an average of
410 person days per hectare per year; extensive polyculture systems provide the lowest with 150 person
days per hectare per year.

14- Food security impacts of aquaculture are measured by two indicators – the net dietary protein production
(in kg per hectare per year) and the protein conversion ratio (ratio of inputs to outputs). The latter
indicator shows that a majority of the farms produce more protein than they use as observable feed
inputs, and thus are exploiting the natural productivity of the pond environment. However, there is
considerable within-type heterogeneity for both these indicators, and some farms use considerable
quantities of protein-rich feed to achieve relatively low yields. The highest net protein production per unit
area comes from low-input labour-intensive systems (farm type 3). If we consider both this result and the
high employment effect that these systems have on the local economy, we conclude that they can be
considered relatively socially equitable. There is thus a possible tension between social equity and
economic performance from the sector because low-input labour-intensive systems also perform
relatively poorly when judged by economic indicators.

15- Nutrient loss to receiving waters is measured simply by the difference between the observable nutrient
inputs and the nutrients embodied in the harvest. The fate of these excess nutrients are not known, but
might be expected to correlate with the amount of nitrogen released into the water, hence contributing to
nutrient enrichment of the coastal areas. All farm types present positive values for this indicator and
therefore generate nutrient losses. However, there is considerable heterogeneity across types. Semi-
intensive prawn-oriented polyculture farms lose, on average, 230 kg N per hectare per year - compared to
only 20 kg N per hectare per year for extensive polyculture and 50 kg N per hectare per year for low-
input labour-intensive systems.

16- Pesticides are applied to most farms, in fairly consistent quantities of between two and five kilograms per
hectare per year. However, the kind of pesticides used varies significantly in terms of toxicity. Sodium
cyanide is favoured in region III, whereas Brestan is more commonly used in region VI. Both of these
chemicals are highly toxic and their status with regards to national and municipal laws is confused.

17- The ecological footprint method was applied in a partial manner by focusing on two inputs to the farming
systems: small bivalve and univalve shells used as feeds, and prawn fry. Both of these inputs are
dependent on the same areas that are occupied by the farms themselves – i.e. the inter-tidal coastal areas.
Using secondary data from other ecological studies and data on the use of these inputs from the farm-
level survey, we estimate that approximately nine hectares of inter-tidal areas are indirectly appropriated
to provide small shells and fry to one hectare of semi-intensive prawn-oriented polyculture farm (farm
type 2). The equivalent estimate for extensive polyculture is two hectares per hectare, whereas the areas
for the other farm types are negligible / zero. These are only very partial estimates of the full ecosystem
support required by brackish-water pond aquaculture in the Philippines. However, they are useful to
highlight the paradox that has often been at the heart of criticisms directed at coastal aquaculture: while
one of the rationales for aquaculture is that it can provide a solution to the problem of stagnating or
declining wild fisheries yields, the mobilisation of many of the inputs consumed by aquaculture can add
to the stresses faced by capture fisheries.

ANALYSIS OF TRADE-OFFS AMONG SUSTAINABILITY OBJECTIVES (Paper 3)

18- Farm-level trade-offs among sustainability objectives are analysed using a discrete- choice multi-criteria
decision-making model. Central to the model is the concept of the utopian point – where all objectives
are maximised simultaneously. Of course, such a point is exactly that, utopian, and thus we must be
content with minimising the distant to such a point. Thus, the utopian concept of sustainability is made
operational for modelling purposes.

3
 19- While we find that there are important trade-offs between some pairs of indicators (e.g. profit and net
protein production; profit and employment), there are other pairs of indicators for which a “nearly
optimal” solution in one provides a good level of achievement in the other.

20- The measures of six indicators for five farm types are used as the basis of the analysis. The relative
importance of each indicator in determining the distance to the utopian point is determined locally, and
four case-studies in two barangays are used in applying the model. These weights are elicited from focus
group exercises using the technique of matrix scoring.

21- For most sets of weights chosen by the groups, extensive polyculture (farm type 1) is the preferred farm
type. It scores highly (but not maximally) for profit and has a relatively weak score only for employment
effect. Large milkfish-oriented systems (farm type 4) perform poorly in all scenarios, and are the least
preferred of the five farm types. The other three farm types are intermediate between these two.

AQUACULTURE AND THE LIVELIHOODS OF THE POOR (Paper 4)

22- The analysis of the social effects of aquaculture is extended by investigating the relationship between
aquaculture and poverty based on a household-level survey of five coastal communities (or ‘barangays’)
in the Philippines.

23- A review of the literature reveals that, at a conceptual level, the relationship between aquaculture and
poverty is relatively straightforward. The arguments supporting the pro-poor nature of aquaculture rely
on the income stream, employment and nutritional benefits that it can potentially generate. However, the
empirical evidence regarding the ability of aquaculture to reduce poverty is rather limited and, altogether,
mixed at best. Hence, there is room for additional empirical enquiry, which this study aims to fill.

24- Five communities were chosen in Regions III and VI. These communities were purposefully selected in
order to capture a variety of contexts in terms of remoteness, relative importance of fishing and
aquaculture as economic activities, geographic location and proportion of land area covered by
mangroves. Altogether, qualitative and quantitative data on the socio-economic conditions and
perceptions of 148 randomly selected households in the five communities was collected.

25- The results reveal a high, but also variable, incidence of poverty in the five studied communities. The
quantitative analysis establishes that 59% of the surveyed households fall below the official poverty line,
while 43% do not even reach the ‘food threshold’, which we interpret as an absolute poverty line.
Meanwhile, a self-assessment exercise indicates that more than two thirds of respondents consider
themselves to be poor. These figures compare unfavourably to a Head Count index of 37% for the
country as a whole, confirming the received wisdom that coastal communities tend to be particularly
deprived – and, hence, leaving room for a potentially large poverty alleviating effect of aquaculture.

26- There is clear evidence that both poor and non-poor households benefit from aquaculture, but that the
poor benefit more in relative terms. The quantitative analysis indicates that a poor household derives, on
average, PhP23,863, or 44% of its income, from aquaculture, as compared to PhP30,809, or 23%, for a
non-poor household. Further, the extreme poor in the selected communities derive, on average, more than
half their income from aquaculture. Aquaculture can therefore be considered pro-poor in the sense that its
relative importance in income generation is larger for the poor than the non-poor. Confirming the
analysis of income data, a very large majority of respondents expressed the view that aquaculture is
mutually beneficial to the poor and non-poor.

27- A Gini decomposition exercise was carried out to investigate the influence of aquaculture on income
inequality. The results establish unambiguously that aquaculture represents an inequality-reducing source
of income in the studied communities. In particular, while aquaculture accounts for an average 30% of
household income, it contributes to only 8% of overall inequality as measured by the Gini coefficient.

4
 The inequality-reducing effect is even more pronounced at the margin, with a one percent increase in
aquacultural income estimated to reduce the Gini coefficient by 0.08. This result is explained primarily
by the lack of correlation between household income and aquacultural income - in other words, by the
fact that the non-poor are not much more likely to derive large incomes from aquaculture than poor
households. Further, the finding that aquaculture has an inequality-reducing effect appears robust in the
sense that it applies to both the whole sample and the individual communities taken individually.

28- A minority of respondents (30%) believe that aquaculture may have a negative impact on the poor,
mainly because it generates externalities (water pollution) from which poor fisher-folks suffer. However,
these negative effects are clearly not thought to exceed the benefits from the industry, because an
overwhelming majority (95%) expressed an overall positive view of aquaculture. Hence, it is clear that in
these communities, aquaculture is seen as a potential solution to the poverty problem – and certainly not
its cause.

29- The main benefit from aquaculture to the poor lies with the employment that it generates in labour-
surplus communities where the lack of jobs is perceived as the main cause of poverty. The supervision of
ponds (‘caretaking’) provides stable jobs to almost a quarter of households in the sample, which is
usually sufficient to keep them out of poverty. The hiring of daily labourers and harvesters by fishpond
operators is also important because it allows for a broad sharing of the value added generated by the
sector. Hence, almost half of the households derive some income from the sale of wage labour to
aquaculture farms, while one in five participates in the harvesting of fishponds. Additional employment
is created through linkages, although its importance appears limited, as mentioned previously with
reference to paper 2. Finally, in the two barangays where aquaculture is the most developed, only six
households out of 62 do not derive any income from fish farming-related activities.

30- The analysis establishes that the poor themselves do not operate fish farms, although the causality of this
relationship appears difficult to establish. On the one hand, and contrary to expectations, more than half
of the respondents expressed the view that some poor people manage to establish themselves as
aquaculture producers. Hence, there appears to be some degree of social mobility within the studied
communities, and aquaculture itself may be regarded as an instrument of that mobility. On the other
hand, respondents acknowledged that there are important barriers preventing entry into the sector by the
poor, with access to credit representing the main one.

31- Although difficult to quantify, there appears to be important nutritional benefits to the studied
communities from aquaculture, which materialize through a variety of non-market mechanisms. For
instance, the practice of allowing the collection of ‘free fish’ from ponds, i.e. any residual fish or
crustaceans left after the main harvest, represents a source of fish for a large number of households.

AQUACULTURE AND LAND DISTRIBUTION (Paper 5)

32- The tidally inundated brackish-water land supporting fish ponds has so far been exempted from the
Comprehensive Agrarian Reform Laws in spite of being unequally distributed, as indicated by a large
Gini coefficient of pond ownership (0.72), as well as the existence of very large farms (with sizes in
excess of 100 hectares).

33- Beyond obvious equity issues, this unequal distribution raises the possibility that brackish water land
may be used inefficiently if large farms are less productive than small farms, as is usually believed to be
the case in the agricultural sector of developing countries. We investigate this question formally by
testing for the presence of an inverse relationship between farm size and technical efficiency.

34- The econometric results provide evidence of an inverse relationship of only limited strength. Hence, it is
estimated that, on average, a 10 percent increase in farm size raises the level of technical efficiency by
only 1.4 percent. Further, farm size explains only 13 percent of the variation in production not accounted
for by inputs.

5
 35- The analysis also reveals that brackish water fish farms are very inefficient, confirming the conclusions
reached in paper 2 with a different methodology. However, it is unlikely that land reallocation holds the
key to unlocking the productive potential of brackish water aquaculture in the Philippines.

CONCLUSIONS AND RECOMMENDATIONS

36- Two factors suggest that extensive aquaculture is sustainable in the Philippines. Firstly, it would appear
to be resilient in the face of a number of perturbations, such as disease outbreaks and the eruption of
Mount Pinatubo. Secondly, from the results in the multi-criteria decision-making model in paper 3, we
find that extensive polyculture and low-input labour-intensive systems are those preferred under sets of
weights chosen by stakeholder groups in our study regions. Thus, these feed-extensive systems perform
the best relative to the other farm types in the regions.

37- However, there are problem areas away from farm-level issues. Prime amongst these must be the reliance
of the sector on wild broodstock and/or fry as well as on wild-harvested live feeds (small bivalve and
univalve shells); and the conversion of mangrove areas to make way for more fishponds. All of these
issues should be considered relevant areas for further research and possible government action.

38- An overall ranking of farming systems from the analysis in paper 3 is: extensive polyculture – top; low-
input labour-intensive systems, semi-intensive milkfish monoculture, semi-intensive prawn-oriented
polyculture – middle rankings; large milkfish-oriented systems – bottom.

39- The poor performance of a small number of very large farms that occupy almost half of the area of our
sample farms provides motivation for including brackish-water fishponds under the Comprehensive
Agrarian Reform Law. Powerful operators in the sector have thus far successfully lobbied for exemptions
from the law, but this would appear to have no support in economic, social or ecological terms from the
findings in this project.

40- The evidence presented here suggests that intensification of either prawn-oriented polyculture or milkfish
monoculture would result in: a higher level of profit on average; a more variable level of profit; reduced
nutrient and protein efficiency; reduced technical efficiency; and would likely have little effect on
employment. In terms of choices of output, polycultures would appear to be preferred to monocultures on
every level.

41- The ability of the poor to partake in aquaculture production should be facilitated by reducing barriers to
entry into the sector, which means primarily improving their access to financial capital. If that is not
possible, ensuring that caretakers do obtain a share of total revenue or profit, as stipulated in their verbal
contracts, would improve entry into the sector by relatively poor households.

42- The research (papers 2 and 5) establishes that fish farmers exhibit high levels of technical inefficiency,
which simply means that they could, on average, increase output by a considerable amount with the
available technology and the input quantities currently used. While it is recognised that brackish-water
farms operate in a difficult and changing environment, it appears that best practices are not shared among
farmers, leading to a high level of heterogeneity in terms of technical performance. This suggests that
there is considerable room for improvement in farm management, which could be brought about by
public R&D investments specifically targeted at brackish-water aquaculture. Such investments are almost
non-existent currently, which reinforces the idea that they might deliver high returns.

43- Further, it is clear from the results of paper 4 that, altogether, aquaculture makes a positive contribution
to the livelihoods of the poor. Hence, the afore-mentioned investments appear desirable from both
efficiency and equity points of view, because aquaculture development is likely to generate large benefits
to the poor. At that level, the research found little support for the contention that aquaculture increases
the deprivation of rural coastal communities.

6
44- However, our results also imply that technological progress in the sector does not automatically yield
benefits to the poor, who are unlikely to use those technology themselves. It is really the employment-
generating potential of new technologies that should be considered when assessing their contributions to
poverty alleviation, rather than their potential to increase yield or profitability (which, of course, are
important as well in ensuring that the new technology is adopted by farm operators).

7
BACKGROUND TO THE PROJECT

The Philippines is a country in which a laissez-faire government policy to aquaculture development over the
previous two decades has been replaced with an official commitment to the principles of sustainability in its
National Policy on Fisheries, embedded in the Philippines Fisheries Code of 1998. The crash of the intensive
prawn industry in the early 1990s and public anxiety after numerous fish kills reported in the press have acted as
drivers to bring about this change. However, in developing a strategy for sustainable development of the
aquaculture sector, the Philippine government is constrained by the lack of an inter-disciplinary study on the
specific farm-level impacts of different types of production systems. A suitable framework is required in which
the trade-offs between the triple objectives of maximising economic performance, maximising social
acceptability and maximising ecological sustainability, can be made explicit.

PROJECT PURPOSE

The primary purpose of the project is to improve knowledge of the interactions among the economic, social and
ecological properties of pond aquaculture systems in the Philippines. This should in turn increase the coherence
of the policies aimed at addressing the problems of the sector and thereby increase its sustainability. The project
also intends to demonstrate the utility of a systems-thinking approach to the analysis of aquaculture.

In addition, following discussions with the program manager, it was decided to extend the scope of the project as
defined in the original proposal to analyse the role that aquaculture plays in the livelihoods of the poor in the
coastal areas of the Philippines.

RESEARCH ACTIVITIES

• Review of the literature on typologies, sustainability indicators, and multi-criteria decision making
(MCDM) methods.
• Farm-level survey of a sample of 150 operators in regions III and VI of the Philippines.
• Principal component and cluster analyses of the farm-level data.
• Selection of sustainability indicators in economic, social and ecological dimensions.
• Collection of the primary and secondary data necessary to measure the selected indicators.
• Elicitation of preference weights for MCDM modelling through focus group discussions with
stakeholders.
• Construction of a variety of MCDM models.

• Workshop on the theme ‘Aquaculture and poverty’ organized in the Philippines.

• Household survey of five coastal communities in the Philippines.
• Poverty and inequality analysis of the household survey.

• Analysis of the relationship between the technical efficiency of farms and their size.

RESEARCH OUTPUTS

• The literature reviewed is weaved into the five papers included in this report.
• A farm-level data set that represents an important and unique source of information on brackish water
aquaculture in the Philippines.
• A typology of brackish-water farms that forms a concise summary of the diversity of production
systems in Philippine aquaculture. The analysis identifies five farm types: extensive polyculture; semi-
intensive prawn production; low-input labour-intensive systems; large, milkfish-oriented systems; and
semi-intensive milkfish monoculture (paper 1).

8
 • A comparative analysis of the five farm types identified in the typology based on the following
sustainability indicators: farm profit; production and price risk; technical efficiency; employment
generation; net protein production; nutrient loss; and (partial) ecological footprint (paper 2).
• A trade-off analysis of a selection of sustainability indicators (paper 3).
• A ranking of the five farm types derived from MCDM models. In most cases, extensive polyculture
represents the preferred farm type. It scores highly (but not maximally) for profit and has a relatively
weak score only for employment effect. By contrast, large milkfish-oriented systems perform poorly in
all scenarios, and are the least preferred of the five farm types. The other three farm types are
intermediate between these two (paper 3).

• A characterisation of the poor in the selected communities with regard to their relationship to
aquaculture. We establish that aquaculture is clearly ‘pro-poor’ in the sense that the poor derive a
relatively larger share of their income from aquaculture than the non-poor. We also find that
aquaculture is perceived very positively by the poor and non-poor alike (paper 4).
• A Gini decomposition exercise on household income establishes that aquaculture reduces inequality in
the selected communities (paper 4).
• An explanation for the pro-poor nature of aquaculture in the Philippines that lies with the employment
generated by the sector. Aquaculture provides jobs to a large number of unskilled workers in
communities characterized by surplus labour (paper 4).

• Evidence of an inverse relationship between farm size and efficiency but of only limited strength.
Hence, it is unlikely that land reform is the key to unlocking the productivity potential of brackish
water aquaculture in the Philippines (paper 5).

CONTRIBUTION OF OUTPUTS TO DEVELOPMENTAL IMPACTS

Several policy implications, recommendations and methodological insights can be drawn from the research:

• From a sustainability point of view, polyculture and (relatively) extensive systems tend to perform
better than intensive and monoculture systems. This simple statement can help select options available
to the government to encourage the development of the sector.

• There are clear trade-offs among the economic, social and ecological properties of aquaculture systems.
Policy options aimed at orientating the development of the sector should therefore not be evaluated
based on a single criterion, such as yield or profit per unit area.

• The poor performance of a small number of very large farms that occupy almost half of the area of our
sample farms provides motivation for including brackish-water fishponds under the Comprehensive
Agrarian Reform Law.

• Although the methodology appears ill-suited to evaluate the sustainability of aquaculture systems in
absolute (rather than relative) terms, the research identifies the key areas of concern as: the reliance of
the sector on wild broodstock and/or fry as well as on wild-harvested live feeds (small bivalve and
univalve shells); and the conversion of mangrove areas to make way for more fishponds.

• Public R&D investments aimed at improving farm management in brackish water areas are likely to
deliver high returns.

• Aquaculture development appears desirable from both efficiency and equity points of view because, in
addition to high returns, it generates large benefits to the poor. The research found little support for the
contention that aquaculture increases the deprivation of rural coastal communities.

9
 • Improving access to financial capital by the poor would remove the main barrier to entry by the poor
into aquaculture. If it is possible to bring about such an improvement, ensuring that caretakers obtain a
share of total revenue or profit would improve entry into the sector by relatively poor households.

• If concerned with the contribution of aquaculture development to poverty alleviation, it is primarily the
employment-generating potential of new technologies that should be considered in the Philippines.

The development impacts of the project can only be indirect through the adoption of some of the
recommendations detailed above by policy makers and researchers involved in aquaculture development. At that
level, it might be worth pointing out that the project was developed in collaboration with institutions in the
Philippines – for instance, the poverty workshop was co-organized by PCAMRD and the University of Reading –
and that the project report, when finalised, will be disseminated to all partner institutions.

LIST OF OUTPUTS

WORKSHOPS

Aquaculture and Poverty, co-convened by the University of Reading and PCAMRD, organized in Los Banos,
Philippines on 22 April 2004.

List of speakers:
• Edwards, P., Emeritus Professor, AIT, Thailand
• Guerrero, R., Director, PCAMRD, Philippines
• Hejdova, E., Doctoral student, CIRAD, France
• Irz, X., Lecturer, The University of Reading, UK
• Morrisens, P., Scientist, CIRAD, France
• Primavera, J. H., Senior Scientist, SEAFDEC, Philippines
• Stevenson, J., Research Officer, The University of Reading, UK

Other participants:
• Barrios, E., University of the Philippines in Diliman
• Ong, D., BFAR Region VI
• Edra, R., PCAMRD
• Darvin, L.C. , PCAMRD
• Bondoc, L.V., PCAMRD
• Piscano, J., Independent
• Natividad, J., University of the Philippines in Diliman
• Pullin, R., Independent
• Rubio, C., University of the Philippines in Diliman
• Tanoy, A., Independent
• Villarante, P., Independent

PUBLICATIONS

Stevenson, J. R. (2005). Sustainability of brackish-water pond aquaculture systems: A farm level analysis of
economic, social and ecological dimensions in the Philippines. PhD dissertation (first version submitted July
2005), University of Reading, Department of Agricultural and Food Economics.

CAB International (forthcoming, 2005). Aquaculture and poverty – A case study of five coastal communities in
the Philippines [original text by Irz, X. and Stevenson, J. R.]. In: Aquaculture Compendium. Wallingford, UK:
CAB International.

10
CAB International (forthcoming, 2005). Typology of farming systems in brackish water ponds, Philippines
[original text by Stevenson, J. R. and Irz, X.]. In: Aquaculture Compendium. Wallingford, UK: CAB
International.

Stevenson, J. R., X. Irz, J. H. Primavera, G. Sepulveda (2003). Costal aquaculture system in the Philippines:
social equity, property rights and disregarded duties. Proceedings of the conference on rights and duties in the
coastal zone, 12-14 June, Stockholm, Sweden.

ARTICLES UNDER REVIEW

Irz, X. and J. R. Stevenson. Is there an efficiency case for land redistribution in Philippine brackishwater
aquaculture? Analysis in a ray production framework, submitted to Applied Economics.

PAPERS UNDER PREPARATION FOR SUBMISSION

Stevenson, J., X. Irz, R. Alcalde, J. Petit and P. Morrisens, A typology of brackish-water aquaculture systems in
the Philippines.

Irz, X., J. Stevenson, A. Tanoy and P. Villarante. Aquaculture and poverty – A case study of five coastal
communities in the Philippines, working paper posted on project website
http://www.dfid.stir.ac.uk/Afgrp/report14.htm#R8288.

CONFERENCES AND SEMINARS

Irz, X. (2004). Déterminants des différents systèmes de production aquacoles aux Philippines – Mesure,
déterminants et implications, invited seminar at the French Agricultural Academy, 07 April 2004, paper available
at http://www.academie-agriculture.fr/seances/?dir=current.

Stevenson, J.R. (2004) Indicators of Sustainability for Philippine Aquaculture. Invited presentation to a
roundtable organised by INRA-IFREMER as part of Bordeaux Aquaculture 2004, Bordeaux Exhibition Centre,
Bordeaux, France, 22nd -24th Sept 2004.

Stevenson, J.R. (2003) Coastal Aquaculture Systems in the Philippines: Social Equity, Property Rights and
Disregarded Duties. Presentation to the multidisciplinary scientific conference on Rights and Duties in the
Coastal Zone, Royal Swedish Academy of Sciences, Stockholm, Sweden, 12-14 June 2003.

OTHER PRESENTATIONS

Stevenson, J.R. (2003, November). Invited presentation at a one-day workshop on sustainable aquaculture in the
Philippines and Brazil organized by INRA-ENSAR, Rennes, France.

Stevenson, J.R. (2003, November). Presentation at the Development Studies Association workshop for
postgraduate research students, Overseas Development Institute, London.

Stenvenson, J. R. (2003, March). Project presentation at the University of the Philippines, Department of
Sociology.

Stevenson, J. R. (2003, April). Sustainability as compromise: brackish water pond systems in the Philippines.
Presentation at SEAFDEC, Department of Aquaculture, Ioilo, Philippines.

11
APPENDICES

12
 Appendix

PROJECT LOGICAL FRAMEWORK

Hierarchy of Objectives Objectively verifiable Means of Verification Important
indicators Assumptions
Goal
Growth of the sector and National statistics Political will to use
Improved policies for contribution to gross value results of the
development of aquaculture added at national, regional research as an input
in an economically viable, and local levels. into policy making.
socially acceptable and
environmentally sustainable Increased productivity. National and farm-level Political stability
way. statistics.
Reduced opposition to End-user links can
aquaculture development. Local surveys, local news be maintained over
the course of the
Creation of formal livelihood project
opportunities, in particular Local surveys
for the rural poor. Co-operation of
PCMARD and
SEAFDEC

Purposes

Improved knowledge of the

economic, social and See Outputs See outputs See outputs
ecological properties of
existing farming systems.

Improved understanding of
the necessary trade-offs
among the economic, social
and environmental impacts
of aquaculture.

Tools for planning and

management developed
through inter-disciplinary
research.

Outputs
Description of farming Farm-level data set from a Extended literature review
systems in terms of their representative sample of on the economic, social and Co-operation of
technical and managerial brackish water farms ecological impacts of farmers
characteristics. aquaculture.
Typology of brackish-water Quality and
Measurement of the pond aquaculture systems Working papers for each consistency of data
economic, social and stage of the research:
environmental impacts of Data set of sustainability - WP1: typology of farming Local institutions
main farming systems. indicators for brackish- systems maintain co-
water pond aquaculture - WP2: sustainability operation
Improved understanding of systems indicators and trade-off
the trade-offs among analysis
economic, social ad Trade-off analysis of - WP3: MCDM modelling
environmental impacts of indicator data. exercise
aquaculture.
Multi-Criteria Decision Final report will summarise
Preferences (weights) of Making (MCDM) model of the findings and draw
stakeholders for economic, brackish-water pond conclusions on the
social and environmental aquaculture systems and usefulness of the approach.
properties of aquaculture. associated rankings.
Final workshop will

261
 Appendix

Ranking of production disseminate outputs to local

systems. partners/stakeholders.

Relevance of MCDM for Peer reviewed publications.

policy making in aquaculture
assessed. Stakeholders consultation
and participation.
Activities Cost (% of total budget)

Literature review
5%
Farm-level survey of technical, physical and attitudinal characteristics
20%
Factor and Cluster analysis of farm-level technical and managerial data
10%
Developing sustainability indicators for brackish-water pond aquaculture
systems
10%

Data collection on Sustainability Indicators

15%
Stakeholder Analysis with policy-makers and community groups
10%
MCDM modelling
20%
Workshops
10%

262
 Appendix

RESEARCH PROPOSAL

SECTION A: KEY INFORMATION

1. Project Title
Assessing the sustainability of brackish-water pond aquaculture systems in the Philippines: A
multi-criteria approach

Abbreviated Title:
Sustainability of brackish-water aquaculture

2. Is the research strategic? (delete as appropriate)

3. Project Summary (maximum 100 words)

Aquaculture in the Philippines has experienced fast growth in recent decades but recent problems
have brought this growth to a halt. If the sector is to play an important role in the development of the
country, there is a need to devise new ways of expanding the sector in an economically viable, socially
acceptable and environmentally sustainable manner. The project seeks to contribute towards that aim
by establishing the economic, social and environmental properties of existing aquaculture systems and
by shedding light on the trade-offs among these properties that constrains the choice of a strategy for
the sector. After establishing a typology of brackish water farms, sustainability indicators will be
measured for each farm type and a multi-criteria decision making model developed to produce a
ranking of farming systems. The project will make policy recommendations and determine the
relevance of a multi-criteria approach to policy making in aquaculture.

4. Keywords (including subject, species, countries etc.)

Sustainability, Philippines, Brackish-water, Aquaculture, Penaeus sp., Chanos chanos, Equity, Multi-
Criteria Decision Making

5. RNRRS Programme 6. RNRRS Production System

Aquaculture and Fish Genetics Coastal Aquaculture
Research

7. Project Goal (include RNRRS Programme Purpose where appropriate)

Practical policy options for the development of aquaculture in the Philippines in an economically viable,
socially acceptable and environmentally sustainable way.

8. Geographic Focus 9. Commodity Base

Philippines / S.E. Asia Fish and crustaceans

263
 Appendix

SECTION A: KEY INFORMATION Continued

10. Applicant’s full name, title, post held and department

Dr Xavier Irz, Lecturer, Department of Agricultural and Food Economics

Name, address, telephone and fax of applicant’s institution

Department of Agricultural and Food Economics
University of Reading
PO Box 237
Reading
RG6 6AR

Phone: 0118 931 8970

Fax: 0118 975 6467
Email: [email protected]

11. Name and address of any collaborators

James Stevenson, Department of Agricultural and Food Economics, University of Reading, Address as
above

Dr Rafael Guerrero III, Philippine Council for Aquatic and Marine Research and Development
(PCMARD), Los Banos, Philippines

Pierre Morissens, Centre de Cooperation Internationale en Recherche Agronomique pour le

Developpement (CIRAD), Visiting Scientist in PCMARD, Los Banos, Philippines

Dr Tahir Rehman, Department of Agriculture, University of Reading, Address as above

Dr Jurgenne Primavera, South-East Asian Fisheries Development Centre (SEAFDEC), Tigbauan,

Iloilo, Philippines

Dr Clarissa Rubio, Department of Sociology, University of the Philippines, Diliman, Quezon City,
Philippines

12. Project Location

University of Reading

Republic of the Philippines: Pangasinan, Pampanga and other coastal fieldwork sites.

13. If the project is located overseas or if there is an overseas collaborator, has the approval
of the overseas government been obtained? If so, provide details.
The Philippine Council for Aquatic and Marine Research and Development (PCMARD) are an agency
of the Department of Science and Technology (DOST) and have significant policy influence in the
areas of aquaculture and fisheries in the Philippine government. The head of PCMARD, Dr Raphael
Guerrero is a collaborator in the project, as is Pierre Morissens, a visiting scientist at PCMARD. A letter
of support from Dr Guerrero is attached.

14. Starting and finishing dates

1st April 2003 31st March 2005

264
 Appendix

SECTION B: DEMAND, UPTAKE AND GEOGRAPHICAL FOCUS

15a. What is the project’s purpose (maximum 50 words)?

Knowledge of the interactions between economic, social and ecological properties of pond aquaculture
systems increased. Encouraging coherence in addressing the problems of the sector. The utility of a
systems-thinking approach to aquaculture promoted among Philippine institutions.

15b. What developmental problems or needs is the project aimed at?

The project aims to address the multiple environmental and social impacts of aquaculture that have not
been subjected to systematic treatment in relation to the performance of specific aquaculture systems.
The Philippines is a country in which a laissez-faire government policy to aquaculture development
over the previous two decades has been replaced with an official commitment to the principles of the
sustainability in its National Policy on Fisheries, embedded in the Philippines Fisheries Code of 1998
(Guerrero, 2001). The crash of the intensive prawn industry in the early 1990s and public anxiety after
numerous fish kills reported in the press have acted as drivers to bring about this change.

However, in developing a strategy for sustainable development of the aquaculture sector, the
Philippine government is constrained by the lack of an inter-disciplinary study on the specific farm-level
impacts of different types of production system. A suitable framework is required in which the trade-offs
between the triple objectives of maximising economic performance, maximising social acceptability and
maximising ecological sustainability, can be made explicit.

15c. What is the evidence for the demand for the research?
There is policy-driven demand for the research as evidenced by the government policy on sustainable
aquaculture (The National Policy on Fisheries, including aquaculture, embedded in the Philippines
Fisheries Code 1998) which states:
the policy of the state is “to ensure the rational and sustainable development, management and
conservation of the fisheries and aquatic resources in Philippine waters including the Exclusive
Economic Zone and in the adjacent high seas consistent with the primordial objective of maintaining a
sound ecological balance, protecting and enhancing the quality of the environment.” Moreover, the
objective of the fisheries sector is the “improvement of productivity of aquaculture within ecological
limits”.

The main body for delivering initiatives towards achievement of this policy is the Philippine Council for
Aquatic and Marine Research and Development (PCAMRD) whose mission statement is:
“Enhance the productivity, global competitiveness and ecological sustainability of Philippine aquatic
and marine resources through appropriate, strategic and people-oriented R & D programs”.

Equally important, there is bottom-up demand from aquaculture operators as evidenced from
participatory rural appraisal (PRA) exercises carried out in April/May 2002 in the province of Pampanga
(James Stevenson, pers. comm.). Aquaculturists see the need for government intervention to address
the problems of water quality and disease that affect the performance of their farms.

15d. What will the project contribute to resolving these problems or needs and over what time-
scale?
PCMARD researchers will collaborate in the project over the course of the two years, becoming
familiar with the methods proposed by the team. It is expected that the findings will have significant
policy influence in the Department of Science and Technology and the Bureau of Fisheries and Aquatic
Resources (BFAR). By highlighting the trade-offs between ecological, social and economic properties
of the systems, the project will help target interventions in areas of particular need.

265
 Appendix

The timescale for positive impacts from the implementation of the project findings are likely to be of the
order of 2 -10 years.

15e. What is the geographical focus of the project?

- The literature review will attempt to draw on experience from across South-East Asia.
- The coastal areas of the Republic of the Philippines will be the focus of the fieldwork.
- The findings will be of interest in all countries in South-East Asia with significant areas of brackish-
water pond production.
- The methodological aspects of the project (application of MCDM techniques for the evaluation of
strategies and projects in aquaculture) will have general relevance.

15f. Which are the identified target institutions?

The primary target institutions are the Philippine Council for Marine and Aquatic Research and
Development (PCMARD) and the Bureau of Fisheries and Aquatic Resources (BFAR). The Executive
Director of PCAMRD is a member of the research team and will serve in an advisory role to ensure the
policy relevance of the project at all stages.

The secondary target institutions are research institutions in the Philippines and in the region, namely
the University of the Philippines (UP), South-East Asian Fisheries Development Centre (SEAFDEC),
International Centre for Living Aquatic Resource Management (ICLARM) and the Asian Institute for
Technology (AIT).

15g. What are the proposed promotion pathways for the uptake of the project outputs?
Promotion of the results in these specifically targeted institutions will be carried out throughout the
length of the project and beyond. Workshops with interested parties will take place at an early stage to
ensure that there is the opportunity for input to influence the research process.

The secondary target institutions will not necessarily feel a sense of ownership of the results but, in the
case of SEAFDEC and UP, have members on the research team who can promote the findings from
within the organisations.

15h. Who will the beneficiaries be and are there any groups who will be disadvantaged by the
application of the research findings?
The indirect beneficiaries of the project will be rural coastal communities through improved
management of aquaculture and it is hoped that the rural poor will benefit disproportionately. The
project aims to ease the decision-making process and will have an impact if the research findings are
taken into account in policy making. Consequently, as with any policy exercise, there is the possibility
of some agents being made worse off as the result of a particular policy being implemented. However,
part of the project is aimed at quantifying the social impacts of alternative forms of aquaculture so that
the costs to any group of agents will be explicit. In this context, a group of agents will only suffer from
the policies derived from the project if that is consistent with the dominant value judgements expressed
through participatory research by policy-makers and other stakeholders.

16. Is this proposal a continuation or extension of work already funded by DFID?

This proposal is not a direct extension of any work funded by DFID but it will build on past research
funded as part of DFID’s Renewable Natural Resource Strategy. In aquaculture, the most relevant
project is entitled ‘Coastal Aquaculture and Environment: Strategies for Sustainability’ (Reference
Number - R6011) that investigates alternative strategies for sustainable aquaculture management in
developing countries with a particular emphasis on shrimp farming in Thailand. That project described
some of the economic, social and environmental impacts of aquaculture but did not attempt to
formalise their interactions nor to provide a comprehensive analytical framework to support the policy
making process. Furthermore, our proposal will not be limited to the analysis of a single production

266
 Appendix

(shrimps) but will consider a much wider range of options since there are important alternatives to
shrimp farming in the Philippines.
The second DFID project that is worth mentioning is entitled ‘The effects of policy and institutional
environment on natural resource management and investment by farmers and rural households in east
and southern Africa’ (Reference Number – R7076CA). The project, although focusing on agriculture
rather than aquaculture, presents many methodological similarities with our proposal. In particular, it
presents the literature on sustainability indicators and discusses its relevance for applied research in
developing countries (see for instance the working paper series e.g. “A Review of Indicators of
Agricultural and Rural Livelihood Sustainability”).

267
 Appendix

SECTION C: SCIENTIFIC BACKGROUND

17. What work has previously been done or is currently being pursued towards the purpose,
outputs and activities of the project? ( A review of literature should be attached)

Literature review is attached separately.

268
 Appendix

SECTION D: OUTPUTS AND ACTIVITIES

18a. What are the outputs of the project?

1- An appreciation of the diversity of production systems in Philippine aquaculture and a
characterisation of these systems in terms of their technical and managerial characteristics.
2- Increased knowledge of the economic, social and environmental impacts of brackish water
aquaculture in the Philippines.
3- Improved understanding of the necessary trade-offs between economic efficiency, social
acceptability and ecological sustainability that are implicit in the choice of production systems
in aquaculture.
4- A set of explicit weights that are attributed to the economic, social and ecological impacts of
aquaculture by policy makers and other stakeholders in the Philippines.
5- Given this set of weights, an explicit ranking of the different production systems
6- An evaluation of the usefulness of MCDM methods in informing the policy making process in
developing country aquaculture

18b. What are the objective verifiable indicators for the outputs?
1) A set of farm-level data on technical and managerial characteristics from a representative
sample of brackish water farms.
2) From these data, a typology of brackish-water pond aquaculture systems established through
cluster analysis and other types of multivariate analysis.
3) A data set of sustainability indicators for brackish-water pond aquaculture systems. These
indicators will measure the economic, social and ecological impacts of the different production
systems.
4) A trade-off analysis of indicator data.
5) A Multi-Criteria Decision Making (MCDM) model of brackish-water pond aquaculture systems
and the rankings obtained as an output of the model.

18c. What are the means of verification of the outputs?

1- An extended literature review will place the research in the context of the existing scientific
knowledge on the economic, social and ecological impacts of aquaculture.
2- Working papers will be produced at each stage of the research:
- WP1 will present the typology of farming systems
- WP2 will present the sustainability indicators and the trade-off analysis
- WP3 will present the result of the MCDM modelling exercise
3- The final report will summarise the findings and draw conclusions on the usefulness of the
approach.
4- The final workshop will provide a way of disseminating these outputs to the local partners and
stakeholders.
5- The relevance of the approach and validity of the results will be assessed by submission of articles
for publication in peer reviewed journals.

18d. What are the expected environmental impacts? (beneficial, harmful, neutral)
i) Direct
None.

ii) Indirect
Implementation of the project findings should have a significant beneficial impact on the environment of
the coastal zone. Improvements could take the form of targeted promotion of environmental
innovations, greater understanding of environmental impacts in the government bureaus and the
possibility of local community resource management in the medium-term. However, the overall impact

269
 Appendix

of the project will depend on its influence on the policy process in the aquaculture sector.
The project will contribute to environmental sustainability by:
- Generating new knowledge about the ecological impacts of alternative forms of
aquaculture in the Philippines.
- Identifying the production systems that have the least negative impacts on the
environment for similar levels of economic efficiency and social acceptability.
- Making explicit the trade-off between environmental quality and socio-economic
performance inherent in the policy-making process.
- Providing a new tool (MCDM) for the assessment of projects and strategies in the
aquaculture sector that takes explicit account of environmental impacts

19a. Describe the project activities

The flow chart below shows the interaction of the various project activities.

270
 Appendix

= Philippine Fieldwork

Sustainability Definition

3 Public Objectives

Maximise Ecological Maximise Economic Maximise Social

Sustainability Performance Acceptability

Indicators Indicators Indicators

Conceptual Model
Stakeholder Weights
Analysis

Detailed data collection on representative

farms

MCDM Trade-Off
Expert Consultation modelling Analysis
and Verification

Brackish-water Pond Aquaculture

System Typology Training Community
Workshops Workshops

Cluster Analysis Presentations

Factor Analysis

Primary / Secondary Primary Farm Study of farmers’

Physical Farm Data Technical Data objectives

271
 Appendix

Literature review (April - August 2003)

This task involves building on existing experience. The literature review will focus on issues
specific to the Philippines as well as wider issues in aquaculture in South-East Asia. There will be
a strong methodological component in order to situate the use of sustainability indicators and
MCDM modelling in a wider academic tradition. The peer-reviewed and “grey literature” will
receive equal attention and we also intend to carry out extensive email and personal networking
with the main researchers in the field. This review will lead to an interim review report.

Farm-level survey of technical, physical and attitudinal characteristics (April – June 2003)
and Factor and Cluster analysis of farm-level technical and managerial data (July – Aug
2003)
This farm-level survey is the gathering of data in order to produce a typology of farms according to
their structure and function, and an attitudinal survey of farmers, to judge which objectives are
important to them in carrying out aquaculture. Such a survey will help identify relationships
between types of farmers and their farms and highlight differences across farm types. It is a
necessary prior step in carrying out an appraisal of the performance of representative farms. This
is a continuation of work due to start in the field in February 2003, with initial pilots being carried
out by James Stevenson, prior to the start of the project.

Factor analysis will be used to reduce the dimensionality of a multivariate data set comprising data
such as farm size, species cultured, feeding rates, owner or tenant managed and the objectives of
the manager. Cluster Analysis is a multivariate data analysis method that groups individuals from
a population according to their degree of similarity. In this context, ‘similarity’ can take a number of
specific quantitative definitions but most widespread is the use of distance measures (technically a
measure of dissimilarity). The clusters emerging from this analysis are considered specific “types”
and representative farms from each cluster selected for re-visiting for data collection on the
sustainability indicators.

Developing sustainability indicators for brackish-water pond aquaculture systems (April –

June ’03)
Sustainability indicators are measurable properties of systems that are used to compare between
systems in the current time period (state indicators) and to analyse the performance of a system
over time (rate indicators). Given the absence of time-series data sets in the current context, a
state indicator method is proposed.

Indicators from the three spheres of sustainability (economic, ecological and social) will be
developed for the purposes of assessing the trade-offs at the farm-level (see the attached
literature review for a more extensive treatment). The final list for measuring on representative
farms will be produced after consultation with a number of researchers in the disciplines of
aquaculture, ecology, sociology and agricultural/aquaculture economics and guided by leading
researchers in the use of sustainability indicators and MCDM modelling.

Data collection on Sustainability Indicators (November ’03 – May ’04)

This survey will be on representative farms (and in their neighbouring villages and communities)
from each of the clusters identified in the cluster analysis. The indicators will be measured over the
course of a number of days visit to the area of the farm, allowing a detailed picture of the network
of livelihoods around the farm to build up. Participatory methods used at this stage include
seasonal calendars, matrix scoring with a number of stakeholder groups and flow and venn
diagramming. Ecological impacts will be largely inferred from the on-farm activities rather than
measured directly (e.g. nitrogen budgets calculated using the patterns of inputs and outputs).
Economic properties will be gathered from farmer testament relating to profits and risks.

272
 Appendix

Stakeholder Analysis (August – Sept ’04)

Workshops will be held with different stakeholder groups and with policy-makers in the competent
government departments, to establish what weighting they place on the different objectives. How
important are the social impacts of aquaculture in comparison to economic advantages and
ecological costs? How do different groups place different weights on different objectives and why?
How committed are the policy-makers to the principles of sustainability in aquaculture? All of these
issues will be explored in detail which, as well as a constituting an end in itself, is a requirement of
the kinds of modelling that will be carried out in the later stages.

Multi-Criteria Decision Making (MCDM) Modelling (November ’03 – July ’04)

Explicit weights from the stakeholder analysis and the results from the typology and indicator
measurement are to be combined in a model to show the trade-offs at the farm-level between
economic performance, social acceptability and ecological sustainability. The methods of Multi-
Objective Programming and Compromise Programming will most likely be used in the analysis of
these trade-offs and in determining rankings of the farming system types.

Sensitivity Analysis (December ’04 – February ’05)

The sensitivity of rankings to changes in the weights will be tested, producing alternative rankings
according to different sets of weights. This can be used to show how technical changes in different
regions will affect different stakeholder groups. This will involve manipulating the MCDM model
with different sets of weights from different stakeholder workshops.

Workshops, Training and Presentations (January ’05 – February ’05)

The dissemination of the findings will be via the mechanisms of workshops, training days at
research institutions and presentations. Workshops will be held with members of all stakeholder
groups. Training days at target institutions (PCAMRD, UP, SEAFDEC) will begin by introducing
systems thinking and sustainability and following on to the specific elements of the research.
Presentations will be made at relevant conferences. Project working papers will be posted on the
Internet and appropriate publication in development and scientific journals will be pursued.

19b. What factors could prevent the attainment of:

i) Planned activities
Security concerns in the Philippines have been heightened following recent attacks on Western
targets. If the security situation deteriorates, some members of the research team may be forced to
evacuate. This is not anticipated but a contingency plan of activity whereby Filipino enumerators are
trained to carry out fieldwork is available and can be implemented at any stage.

Fall in the value of the UK pound (or, conversely, a surge in strength of the Philippine peso) erodes the
budget.

Lack of co-operation and participation by various stakeholders in fieldwork and in workshops.

ii) Outputs
The quality of the data and the consistency of the responses are reliant on the concentration of the
farmers in responding to a series of questions. Data consistency checks will be built into the project
design where appropriate (e.g. to ensure transitivity conditions are not violated). The surveys will be
carried out on the farm to allow for visual inspection and qualitative data collection to support
quantitative replies.

There are several points at which consultation between the project team and wider academic and
governmental institutions will be sought, in order to verify that the findings are consistent with
established local expert opinion. Reaching a consensus during this period cannot be guaranteed and
where discrepancies and disagreements emerge, they will be documented carefully.

273
 Appendix

iii) Project Purpose

If the partner institutions do not develop or maintain an enabling environment for innovative extension
or policy changes in promoting the outputs from the project, then the project purpose will not be
achieved.

iv) Project Goal

N/A

274
 Appendix

20. Complete a bar chart or attach milestone charts over the life of the project.
Final Report

Working Paper
on MCDM
Working Paper
on Indicators
Working Paper
on Typology
Literature
Review
Training at
SEAFDEC,
UP
Workshops

Sensitivity
Analysis
Stakeholder
Analysis
MCDM
Programming
Data Coding

Indicator
Measurement
2nd Pilot Study

Verification of
Typology
Sample
Definition
Factor and
Cluster
Analysis
Pilot Study
(indicators)
Indicator
Definitions
Typology Data
Collection
Jan Feb Mar April May Jun July Aug Sept Oct Nov Dec Jan Feb Mar April May Jun Jul Aug Sept Oct Nov Dec Jan Feb Mar
‘03 ‘03 ‘03 ‘03 ‘03 ‘03 ‘03 ‘03 ‘03 ‘03 ‘03 ‘03 ‘04 ‘04 ‘04 ‘04 ‘04 ‘04 ‘04 ‘04 ‘04 ‘04 ‘04 ‘04 ‘05 ‘05 ‘05

275
Appendix

276
 Paper 1/5

Research project R8288: Assessing the sustainability

of brackish-water aquaculture systems in the
Philippines – Paper 1/5

A typology of brackish-water pond aquaculture

systems in the Philippines

Abstract: Aquaculture in the Philippines is a long-standing activity but has witnessed

relatively recent, rapid, technical change with the introduction of hatchery technology
and commercial feed-mills changing the production possibilities for a fishpond
operator. We are confronted with a diversity of aquaculture practices in the coastal
areas of the Philippines, with new technologies being incorporated into more
traditional systems. As a first step to understanding the sector, we therefore present a
typology of farming systems with the motivation of generating domains (farm
“types”) over which we can compare performance on a number of indicators. Our
typology, restricted to brackish-water pond systems, is constructed using multivariate
methods (principal components analysis, cluster analysis). Eight variables are used
relating to the management of the farm across all the major factors of production. A
stratified net sample of 136 observations provides the data for the analysis, from a
farm-level survey carried out between January and June 2003 in the two main
brackish-water production regions in the Philippines. We define five farm types from
this analysis. In later work we will show how the use of this typology can be used for
comparative study of economic, social and ecological performance at the farm-level.

13
 Paper 1/5

Contents

1. Introduction 15

2. Highlighting the problem – Measuring production intensity 18

3. A multivariate approach to classification 19

3.1 Factor analysis 19
3.2. Cluster analysis 19
3.3 Previous applications to agriculture and aquaculture 20

4. Farm-level survey 21
4.1 Sampling 21
4.2 Motivations for choosing the variable set 21
4.3 Factor Model - Testing for appropriateness 22

5. Results 24
5.1 Factor extraction 24
5.2 Interpreting the component loadings 25
5.3 Cluster analysis 27
5.4 Cluster solution x Principal Component Scores: “Farm Types” 28
5.5 Identifying the farm types 31

6. Geographic distribution 34

7. Conclusion 35

Appendices 38

14
 Paper 1/5

1. Introduction
The central proposition underlying work on typologies of farming systems is that
farms are similar or identical to other farms of the same type (according to particular
characteristics of interest) and that they are dissimilar to farms of other types. This
follows from Derek Byerlee’s idea (Byerlee et al, 1980) of “recommendation
domains”, such that policy statements made regarding one member of a domain are
likely to hold for the other members of the domain. While interesting as an end in
itself, a typology increases the likelihood that analysis of, for example, productivity
(within domain) or comparative study (between domains) will be conducted properly
(Shang, 1981).

For example, we might expect technical efficiency (or any other indicator of
performance) “within domain” to be explained in the most part by managerial
competence. In contrast, differences in the level of technical efficiency “across
domains” might be accounted for by underlying differences in the nature of the
technology. The conclusions for policy of determining differences within domain, as
opposed to differences across domains, are very different. A lack of managerial
competence might be addressed by increased investment in government extension
services with respect to particular production techniques within domain. Alternatively,
apparent differences in technical efficiency that are actually driven by inherent
differences (e.g. in production risk) in the nature of the techniques used across
domains.

There are numerous sets of criteria that can be used to classify aquaculture systems,
drawn from technical, geographical, economic, ecological and social perspectives.
Shang (1981) identifies ten criteria by which aquaculture systems could be
conceptually divided. Some of Shang’s criteria are nominal (e.g. nature of enclosure).
Highlighted in bold are the categories of specific nominal criteria that serve to define
the scope of the present study; i.e. only aquaculture systems producing human food
products in ponds filled with standing brackish-water of a tropical temperature are
included in our survey. In the Philippines however, this apparently already narrow
sub-set of aquaculture systems contains a great heterogeneity in Shang’s other
categories.

We argue here that some of Shang’s other criteria are not qualitative but quantitative.
Shang has produced quasi-qualitative criteria by either imposing categories on a
quantitative scale (e.g. monoculture/polyculture, water temperature) or by simplifying
a complex combination of quantitative data in different dimensions (e.g. level of
management intensity 1). In these cases there are not the same fundamental shifts
between categories as for truly qualitative criteria.

The objective of the typology work here is to tackle this heterogeneity using
multivariate statistical methods. Most fundamentally however, we are interested in
representing the current range of techniques employed in the Philippines. In later
work we will consider the “farm types” established here as possible options for the
use of the brackish-water land area and use them in the construction of models of the
farm-level trade-offs between different policy objectives. Examples of these
1
We use the term production intensity from hereon, in the same way Shang uses management intensity.
This is to avoid confusion with the level of managerial or supervisional input the farm requires, which
might also be considered “management intensity”.

15
 Paper 1/5

objectives are to maximise economic efficiency, maximise social acceptability and to

minimise ecological impact.

The structure of this paper is as follows: in section 2 we further highlight the need for
clarity with respect to production techniques under appraisal and why a typology is
necessary; in section 3 we outline the methods we will employ in analysing our data,
and refer to the literature on related studies; in section 4 we describe our data and the
analyses; section 5 has our results and interpretations; section 6 examines the
geographical distribution of farm types within our study regions; section 7 concludes.

16
 Paper 1/5

Table 1 – List of criteria by which aquaculture grow-out systems might be classified

(from Shang, 1981).
Criteria Category
Purpose of Culture Human food
Improvement of natural stock
Sports and recreation
Ornamental fish
Bait
Industrial products

Nature of enclosure Pond culture

Cage and pen culture
Raceway culture
Raft culture
Closed high-density culture
Sea ranching

Sources of fry Natural waters

Captured gravid females
Hatching

Level of management intensity Extensive

Semi-intensive
Intensive

Number of species stocked Monoculture (single species)

Polyculture (more than one species)

Water salinity Fresh water

Brackish water
Marine water

Water movement Running water

Standing water

Water temperature Cold water

Warm water

Food habit Herbivorous species culture

Omnivorous species culture
Carnivorous species culture

Combination with agriculture production Rice-fish farming

Poultry-fish farming
Pig-fish farming

17
 Paper 1/5

2. Highlighting the Problem – Measuring Production Intensity

“’Intensity’ relates to resource utilization (land, water, capital, labour,

seed, feed, fertilizer and fuel) and different systems may be more or less
intensive depending on which resource is considered. It is important to
understand the use of all these resources if a thorough assessment of the
sustainability of different kinds of shrimp culture is to be made.” (World
Bank, 2002)

There are many studies that classify aquaculture systems using a measure of
production intensity. However, the choice of variable or combination of variables with
which to represent the concept of production intensity is not a trivial issue. The most
important variables are perhaps stocking density, feeding rate and fertilizer
application rate.

For monoculture of tiger prawn Penaeus monodon, Clay suggests stocking densities
of greater than 200,000 pieces/ha for intensive systems, 25,000 – 200,000 pieces/ha
for semi-intensive systems and less than 25,000 pieces/ha for extensive systems (Clay
1997 - cited in World Bank, 2002). However, the use of stocking densities alone
makes comparison between polyculture systems, or between monocultures of
different species, problematic. This is because different species have different habits
of feeding, different body sizes at stocking and at maturity, different survival rates and
patterns of natural behaviours. Given that the brackish-water pond production in the
Philippines is a long-standing activity (Primavera 1995) that has evolved from
traditional systems of polyculture with natural recruitment, there is a need to move
away from reliance on the use of stocking density to characterize systems.

Using the feeding rate to describe the systems has distinct advantages as Ravagnan
(1981) advocates:
“If we accept the concept that it is not the degree of crowding, but
rather the energy derivation from the feed, i.e. the feed regime,
which distinguishes the various methods of farming (Ravagnan,
1980), we consider: extensive the farm that takes its feed
exclusively from the environment; intensive the one that instead
takes it exclusively from outside sources; and semi-intensive the
one that exploits the environment but integrates it with feed coming
from outside sources. The production technologies available to us
range around these three methods”. Ravagnan (1981, p.65).

This is an ecological classification and is an appealing approach from theoretical and

practical viewpoints. The overall energy balance of the system gives important
insights for those examining the food security implications of aquaculture and for
ecologists studying the effect of nutrient enrichment on coastal waters. From a
practical viewpoint, the problem of equivalence across species is lost and so feeding
rate can be used for monoculture or polyculture systems.

In generating a measure of “farming” (or “production”) intensity, there are both

definitional and measurement issues. The most accurate definition probably comes
from an economic perspective and argues that intensity is the use of variable inputs
(e.g. fry, feeds, fertilizers) in relation to land. However, as with economic measures of

18
 Paper 1/5

partial productivity, the inputs can be substituted for one another to some extent, so
that measuring one input cannot be totally satisfactory. It is therefore important to
employ a multivariate approach to classification, which allows us to look at the
particular sets of combinations of inputs that currently define production practices.

3. A Multivariate Approach to Classification

3.1 Factor Analysis
Factor analysis belongs to a group of models known as latent variable models. A
latent variable is a concept that cannot be directly measured such as human
“intelligence” or farming “intensity”, that theory suggests is correlated to a number of
specific tests (in psychology) or measurable features of systems (in farming systems
research). Factor analysis is actually a group name for a range of specific statistical
techniques used on multivariate datasets to explain either patterns of covariance or
correlation in a set of observed variables, as function of a limited number of
underlying factors.

In factor analysis there are no dependent and independent variables. Rather a matrix
of covariance scores (for the specific method of “Principal Axis Factoring”) or
correlation coefficients (for “Principal Components Analysis”) is used as the basis to
explore how all the variables are related to underlying factors. Factor analysis yields
interesting information as an end in itself – it gives us information on variables that
are not directly observable. However, its purpose is also commonly to transform a
multivariate dataset with many variables and correlations between the variables
(multi-colinearity), to a new smaller set of underlying factors. It is considered a “data
reduction” technique in this regard. Useful properties of these factors are that they are
orthogonal to each other in factor space and are linear combinations of the original
variables 2.

If the data set is well suited to factor analysis, and the original variables are well
chosen, the construction of factors will occur without significant loss of information.
Following this, factor scores for the individual observations (farms), showing the
position of the observation in factor space (which can be positive or negative) can be
calculated. Factor scores should then be used as a new dataset for carrying out cluster
analysis. It would be inappropriate to carry out cluster analysis on the original dataset
due to multi-colinearity in the original variables, which would effectively create
weights in any clustering algorithm. Factors are orthogonal and therefore this problem
is resolved. We can use cluster analysis to calculate the degree of similarity or
dissimilarity between individual farms, in terms of their scores on the underlying
factors.

3.2 Cluster Analysis

Cluster analysis is another exploratory technique where we do not have a priori
hypotheses (in this case, with respect to the number or properties of the farm types)
but are examining the structure of the data. It is a multivariate technique for
classifying observations into groups (clusters). In the case of farming systems,
ultimately each farm is unique in some respect. However, by using the factor scores

2
Detailed treatments of this technique can be found in Comrey and Lee (1992). A more accessible
introduction is provided by Kline (1994).

19
 Paper 1/5

we obtain in carrying out principal components analysis, we can see how similar each
farm system is to the other individual farming systems in our sample.

The objective of cluster analysis in the present case is to generate solutions that
maximize similarity within groups of farms so that we might label each group as a
particular farm “type”. The term cluster analysis actually covers a number of different
algorithms and measures of distance that can be employed in generating a typology.
We employ Ward’s method (Ward, 1963) and use squared Euclidean distance in
multidimensional space.

The geometric distance in multidimensional space (in this case 3 components – see
section 5.1 below) is known as Euclidean distance. However, we employ squared
Euclidean distance so that progressively greater weight will be placed on farming
systems that are further apart. Squared Euclidean distance is computed as:

distance(x,y) = Σi (xi - yi)2

where x and y are two observations (farms) and i relates to each of the dimensions.

Ward’s method of cluster analysis is one of a family of specific algorithms known as

hierarchical (agglomerative) cluster analysis. These algorithms start with a set of
individuals 3 (in our case farms), and begins by attempting to identify the two
individuals with most similar scores. At this step in the analysis, these two individuals
are merged to form a cluster. This cluster is treated as a single individual in the next
step in the analysis, thereby reducing the number of individuals to be considered in
the second step of the analysis by one. This process continues, with the number of
individuals decreasing, and the number of clusters increasing. Therefore, as the
analysis proceeds, the statistical software merges clusters rather than just individuals
at each step (Stevenson, 1989).

Algorithms differ in the way in which the clusters are formed at each stage. Ward’s
method operates by testing which merger at each stage produces the least reduction in
the overall within cluster squared distances. The output from any hierarchical cluster
analysis is in the form of a dendrogram, showing the history of the cluster mergers
from all individuals in the sample (i.e. all farms unique), up to a single cluster (i.e. all
farms of the same “type”). The point at which we ‘cut’ the dendrogram will determine
the number of clusters (i.e. farm “types”) produced.

3.3 Previous Applications to Agriculture and Aquaculture

Kobrich, Rehman and Kahn (2003) report on two applications of multivariate
techniques to the problem of classifying agricultural system types; one for Chilean
peasant farming systems and one for wheat-rice farms in Pakistani Punjab. Both of
these typologies were based on primary data collection surveys in their study regions.

Within aquaculture research, Coche (1982) and Muir (1995) present uni-dimensional
guides for classification of different aquaculture systems, using production intensity –
a variable that is itself multidimensional as we have outlined. Therefore, Michielsens
et al (2002) represents the first application of factor analysis and cluster analysis to
3
Cluster analysis, as well as factor analysis, both have strong traditions in the psychometric literature.
It is only relatively recently that these methods have been adopted by the farming systems research
community.

20
 Paper 1/5

yield an empirical typology of aquaculture systems. In classifying carp farming

systems across Asia, Michielsens et al (2002) used the following 12 variables: Area of
the aquaculture facility (ha); Ratio of aquaculture facility to total farm area (%);
Water added during the culture period (cm / month); Purchased inorganic fertiliser (kg
/ ha /yr); Total organic fertiliser (103 kg / ha / yr); Ratio of organic fertiliser collected
(on or off-farm) to total organic fertiliser used (%); Total feed added (103 kg /ha /yr);
Number of fish species cultivated; Stocking density (104 fishes / ha); Total labour (102
days / ha /yr); Ratio of family labour to total labour (%). Data came from a
NACA/ADB farm-level survey.

Michelsens et al’s analysis suggests six “types” of carp farming system in Asia, which
they name: Super-intensive, Intensive, Specialised semi-intensive, Specialised semi-
extensive, Integrated semi-intensive, Integrated semi-extensive. The typology is put to
use to examine the resource-use efficiencies of different farm types. We follow the
lead shown by Michelsens et al by applying the multivariate techniques factor
analysis and cluster analysis to data collected from a farm-level survey in the
Philippines.

4. Farm-level Survey
4.1 Sampling
The two top regions for brackish water pond aquaculture production, regions 3
(Pampanga, Bulacan, Bataan and Zambales) and 6 (Iloilo, Capiz, Negros Occidental
and Aklan), were chosen as study areas. The sample was stratified by farm size and by
province, based on census data from 1997 provided by the Philippine Bureau of
Agricultural Statistics (BAS). A breakdown of the net samples used in the analysis is
given in appendices 1a and 1b.

Interviews with fishpond “operators” (those who invest capital, take the financial risks
and gain the profits) and “caretakers” (salaried supervisors for those farms where the
operator does not live on the farm) 4 were carried out on representative samples of fish
farms from regions 3 and 6 between January and May 2003.

A net dataset comprising 11 variables and 137 observations (farms) was initially
compiled after processing the farm-level interview data was complete. One farm
observation was subsequently dropped from the dataset after test runs with principal
components analysis and subsequent cluster analysis found it to be an outlier 5. 136
farms were retained in the final dataset.

4.2 Motivations for choosing the variable set

The typology is based on technical aspects of the farming systems. The choice of the
final list of variables, from a large dataset compiled during the farm-level survey, was
motivated by experience gained in carrying out the interviews. Most of the farms are
polyculture systems, but different priorities dominate on different farms. Operators
are generally either orientated towards prawn production or towards milkfish

4
For a treatment of the nature of this relationship, see Stevenson et al (2003)
5
The farm represented a unique cluster with anomalously large Euclidean distance in the cluster solution from all
other farms. This is likely to be due to the fact that it is a small farm (0.17 ha) and so any accuracy problems in the
data for that farm are magnified when units/ha/yr are calculated.

21
 Paper 1/5

production. This is supported by farmer testimonies and by Pierre Morissens, a

researcher with CIRAD (Cooperation Internationale en Recherche Agronomique pour
le Developpmente) with more than six years experience of working with farmers in
the study area. Crabs and tilapia will sometimes be added as secondary species, for
the purposes of aeration, or opportunisitically if the market and environmental (i.e.
salinity) conditions are good, but these are not of fundamental importance to the
operators and are not given “special treatment” on the farm.

None of the farms in the sample operated at a sufficiently high level of intensity to
require mechanised aeration, in the form of paddlewheels, and in general the level of
capital investment on the farm is relatively low. Therefore, the main differences
between farms in terms of the use of factors of production (land, labour, capital)
would seem to be the relative importance of land and labour. A degree of
substitutability between these two factors might be expected in the production
function for these farms.

With these key ideas in mind, the following 8 variables were chosen for principal
components analysis:

farmsize Farm size (ha)

inorg Total inorganic fertiliser applied (kg/ha/yr)
organic Total organic fertiliser applied (kg/ha/yr)
totlabor Total labour input (man days/ha/yr)
commfeed Ratio of commercial feeds to total feeds added (%)
sugposd Stocking density of tiger prawn (Fry/m2)
bangussd Stocking density of milkfish (Fry/m2)
totfeeds Total feeds added (kg/ha/yr)

The raw data reported by farmers were in a variety of local units and over different
periods. The emphasis in the data collection was placed on getting credible data,
rather than on convenience for analysis. Therefore a lengthy process of sorting and
coding the data was required.

4.3 Factor Model - Testing for appropriateness

The underlying assumption of a factor model (the existence of a few factors that
underlie variability in the data) in carrying out a Principal Components Analysis
(PCA) may be more or less appropriate depending on the nature of the data. Two
widely used statistics to determine the validity of using a factor model on a data are
the Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy and Bartlett’s test of
sphericity. The results of both of these tests for the current data set are given in table 2
below.

22
 Paper 1/5

Table 2 – KMO and Bartlett’s statistics for the factor model data set
Kaiser-Meyer-Olkin .671
Measure of Sampling
Adequacy.
Bartlett's Test of Sphericity Approx. Chi-Square 118.223
df 28
Sig. .000

Bartlett’s test is used to test the null hypothesis that the variables are uncorrelated in
the population (Hair et al, 1998). For this test, the population correlation matrix is an
identity matrix. Bartlett’s test uses a chi-square transformation of the determinant of
the correlation matrix and, with the current dataset the null hypothesis can be rejected,
thereby giving no reason to question the validity of using a factor model on the data.

The KMO statistic compares the magnitudes of the observed correlation coefficients
to the magnitudes of the partial correlation coefficients. While there is no absolute
cut-off or statistical tests for the value of the KMO statistic, a value of 0.7 and above
is desirable, but values of 0.5 and above are tolerable. Essentially, a small value for
the KMO would suggest that the correlations between pairs of variables cannot be
explained by other variables (Sharma, 1996).

In addition to statistical tests outlined above, the inspection of the correlation matrix
for the dataset should show a significant number of correlated pairs of variables of
around 0.3. This final rule of thumb for the appropriateness of the dataset to factoring
is less satisfactorily resolved in this case than the formal statistical tests. There are
many pairs of correlations but usually in the range 0.2 to 0.25. It is therefore
important that all the variables show high communality after the extraction (the
amount of variance in each variable explained by the factor model) and that the
components are easily interpretable. The correlation matrix for the 8 variables is given
in appendix 2.

23
 Paper 1/5

5. Results
5.1 Factor Extraction
A PCA extraction was carried out on the dataset in SPSS using the correlation matrix
and standardised variables (i.e. with mean = 0 and standard deviation = 1).
Communalities for the 8 variables are sufficient (the extraction accounts for at least
half of the original variance for most of the variables) and are shown in table 4 below.

There is a degree of subjectivity with regard to the number of factors that should be
extracted. Common stopping rules are: to stop when eigenvalues go below 1 (see table
3 below); and the scree test (to extract at a noticeable step change in the scree plot –
see appendix 3). Of these, the scree test is inconclusive (3, 4 or 5 components could
be defended) and so the eigenvalue rule is used, which suggests that 3 components
should be extracted.

Table 3 – Variance explained by the 3 principal components

Component Eigenvalue % of Variance Cumulative %
1 2.235 27.941 27.941
2 1.328 16.606 44.546
3 1.067 13.341 57.887
4 .810 10.121 68.008
5 .781 9.759 77.768
6 .659 8.233 86.001
7 .632 7.897 93.898
8 .488 6.102 100.000

Table 4 – Communalities, the % variance each of the 8 original variables explained

by the 3 component extraction
Initial Extraction
Zscore(FARMSIZE) 1.000 .585
Zscore(INORG) 1.000 .436
Zscore(ORGANIC) 1.000 .557
Zscore(TOTLABOR) 1.000 .598
Zscore(COMMFEED) 1.000 .459
Zscore(SUGPOSD) 1.000 .599
Zscore(BANGUSSD) 1.000 .630
Zscore(TOTFEEDS) 1.000 .767

24
 Paper 1/5

Table 5 – Component solution matrix

Component
1 2 3
Zscore(FARMSIZE) .250 -.668 .275
Zscore(INORG) .617 .187 -.144
Zscore(ORGANIC) .690 -8.104E-02 .273
Zscore(TOTLABOR) -.198 .668 -.334
Zscore(COMMFEED) .568 .326 .174
Zscore(SUGPOSD) -.638 3.538E-02 .437
Zscore(BANGUSSD) .642 .352 .307
Zscore(TOTFEEDS) -.367 .403 .685

5.2 Interpreting the Component Loadings

The correlations between the observed variables and the underlying factor model are
called the component loadings. Any given variable should only load significantly on
one or possibly two components, but any given component may have several variables
that load on it. The components should be interpretable in line with hypotheses held
prior to the analysis. Table 6 below (from Hair et al, 1998) gives the guidelines for
identifying significant component loadings based on sample size. In this analysis,
significance is based on: α=0.05, a power level of 80 percent, and standard errors that
are assumed to be twice those of conventional correlation coefficients. With our
sample size of 136, we can consider factor loadings of around 0.48 and above to be
significant.

Factor Loading Sample Size Table 6 - Guidelines from Hair et al (1998)

Needed for for identifying significant factor loadings
Significance based on sample size.
30 350
.35 250
.40 200
.45 150
.50 120
.55 100
.60 85
.65 70
.70 60
.75 50

Referring to the factor loadings in the solution in table 5, we can interpret and name
the components according to those variables that load significantly on that
component 6. Statistically significant loadings are marked in bold.

5.2.1 Component 1 – “Specialisation”

This component describes the orientation of the production system towards either of
the two main crops: prawns or milkfish. The stocking density of milkfish fry
(BANGUSSD) is positively loaded on this component. In addition, however,

6
The solution given is unrotated. Rotations, such as Kaiser’s varimax rotation (Kaiser, 1958), are
sometimes used in interpreting component loadings. However, rotations represent a simplification of
the underlying component structure that in our case is not required – the results are interpretable as they
are.

25
 Paper 1/5

variables for both kinds of fertilizer (INORG and ORGANIC) load significantly
positively on this component. This is because inorganic and organic fertilizers are
used to culture natural food in the grow-out ponds (“lab-lab” 7 and “lumot” 8) and
milkfish can feed on this natural productivity throughout their life-cycle. In the cases
where the operator wishes to fatten the fish prior to harvest (“finishing”) or increase
their growth rate due to low water temperature or low density of lab-lab, then
commercial feeds are added. This explains the significant loading for the ratio of
commercial feeds to total feed added in the year (COMMFEED), because only high
quality commercial feeds are used to supplement the natural productivity of the pond
ecosystem.

Commercial feeds are used much more sparingly in prawn polyculture systems.
Commercial feeds are expensive, and increasingly so. For an operator to use
commercial formulated prawn feeds in polyculture would make little economic sense.
The feed would tend to be eaten by the secondary species in the pond, the prices of
which attract only a fraction of the price of the prawn.

Commercial feeds are often employed by prawn polyculture operators during the
critical nursing period (usually up to a month in duration), where the recently hatched
prawn fry are separated from the rest of the species on the farm to avoid mortality
from predation. However, the quantities of commercial feeds used in this period (“Fry
Mash”) are small, due to the minute size of the fry.

Additionally, milkfish experience relatively low rates of mortality once they have
brought through their nursing period, whereas prawns tend to be much more prone to
shocks in the pond ecosystem. They are more sensitive to salinity changes,
temperature changes and viral outbreaks (Kautsky, Ronnback et al. 2000). Operators
can make the decision to add commercial feeds for milkfish during later periods of
grow-out, relatively safe in the knowledge that the production risk is low. In many
cases, the important component of risk in milkfish monoculture is the fluctuation in
price due to the multiple sources of milkfish in the market (i.e. pond culture, pen and
cage culture). Operators will often keep the fish at market size until they have agreed
a good price in the market before harvesting.

As a corollary, prawn stocking density (SUGPOSD) loads significantly negatively on

this component, suggesting that farms that are specialized in prawn production do not
stock milkfish at high densities or use large quantities or commercial feeds or
fertilizers. Farms with a neutral score for this component are those with no
specialization for either milkfish or prawns and are likely to be particularly extensive
(i.e. with low stocking densities).

Overall, the component “Specialisation” accounts for 27.9% of the original variance
in the set of eight variables.

5.2.2 Component 2 – “Labour vs Land Intensity”

The second component to be extracted has farm size (FARMSIZE) as a significant
negative loading and total labour (TOTLABOUR) as a significant positive loading.
7
Lab-lab is the Filipino term for a dense mat of microbenthic organism communities, composed of
algae and diatoms, that rests on the pond floor (Sumagaysay-Chavoso & San Diego-McGlone, 2003).
8
Lumot is the Filipino term for filamentous algae.

26
 Paper 1/5

The component has been named the “Land vs Labour Intensity” because conceptually,
farms may lie anywhere on a continuum where land is the major factor of production
at one end (those with a negative component score), and labour is used in attempted
compensation for lack of land at the other end (farms with a positive component
score). This would suggest the possibility of some degree of input substitution
between labour and land in the production function.

Overall, the component “Labour vs Land Intensity” accounts for 16.6% of the original
variance in the set of eight variables.

5.2.3 Component 3 – “Feed Intensity”

Only one variable loads positively on this component: the total feeding rate (kg feeds /
ha /yr “TOTFEEDS”). Prawn stocking density loads positively on this component, but
at a level that is not significant at the level α = 5%. The interviews carried out during
the survey showed that some farms, particularly prawn-oriented systems, used large
quantities of low-quality feeds, particularly small molluscs collected from nearby
riverbeds (“gasang”, “suso” “agiis”; Cruz, 1997). The major explanation for the
evolution of this method of production is the converse of the details given above in
relation to milkfish culture. By using low quality feeds of low cost (P1-2 / kg in
comparison to P12-20 / kg for commercial feeds) it is possible to minimize the costs
of rearing prawns as a strategy to minimize the production risk associated with a
given production cycle. The lower growth rates associated with low quality feeds,
thereby lengthening the production cycle and extending the possibility of exposure to
disease and other shocks, would appear to be more than compensated for by the
reduced loss in the case of high mortality.

In these systems, it would appear that a survival rate to adult size of only 2 or 3%
would result in a positive gross margin. 5% survival would result in a healthy profit,
thus making the polyculture of prawn based on low quality feeds a very economically
resilient system in the short to medium term. Longer term, there could be a problem
with excessive harvesting of shells from the riverbed. If the cost of these natural
resource-based supplemental feeds were to rise significantly, the economic feasibility
of these systems would be under threat, in the same way milkfish intensification is
under threat from rising commercial feed prices.

The component “Feed Intensity” accounts for 13.3% of the original variance in the set
of eight variables.

5.3 Cluster Analysis

The principal components analysis has given us 3 dimensions along which the farms
are distributed according to their technical and management characteristics. We can
visualise the distribution of the 136 farms in 3 dimensions but the picture is somewhat
complex. By carrying out a cluster analysis, we can identify groups of farms that
similar to each other and different from farms of other groups. Ward’s method (Ward,
1963) allows us to work with distance functions in any number of dimensions.

A cluster analysis was carried out using the factor scores from the 3 principal
components over the 136 farms using Ward’s algorithm in SPSS. The dendrogram
showing a possible cutting line is shown in appendix 4. The number of clusters to
choose (i.e. at which point to “cut”) depends largely on the purposes of the exercises.

27
 Paper 1/5

In the absence of any a priori expectations as to a “true” number of farm types, we

choose the 5 cluster solution. We think that this gives a range of solutions with
sufficient level of disaggregation to illustrate the complexity in how fishponds
operate. Too much disaggregation, and the farm types would become too numerous,
too complicated to understand and too difficult to communicate effectively.

5.4 Cluster solution x Principal Component Scores: “Farm Types”

The next stage is to interpret the results of the cluster analysis. To do this, we cross
the cluster solution with the factor scores to see which factors are important in
defining each “farm type”. Figure 1 below shows the 5 cluster solution in a 3-
dimensional scatterplot where the axes are each of the three principal components.

Illustrated in figures 2a-c, are factor scores for the 3 principal components for each of
the clusters. This allows us to examine the factors that characterise the farm types.
Statistically significant differences between the mean factor scores for each farm type
are tested in appendix 5. In addition, we can examine the five farm types according to
the original set of 8 variables in shown in table 7 below. From this we can identify the
features that define each farm type and name them. This will be the focus of the next
section.

28
 Paper 1/5

Figure 1 – Distribution of farm types in the three principal components.

s
p
6
e
c
i 4
a
l 2
i
s 0 farm types
a
t 5.00
-2
i
4.00
o
n 6
4
6 3.00
2 4
0 2
-2 0
land vs labour -4 -2 feed intensity 2.00

1.00

Figure 2a – 5-cluster solution with scores for “specialisation” by farm type

3

2
Specialisation

-1

-2
N= 54 15 37 11 19

1 2 3 4 5

5 farm types

Figure 2b – 5-cluster solution with scores for “land vs labour” by farm type

29
 Paper 1/5

1
Land vs Labour

-1

-2

-3
N= 54 15 37 11 19

1 2 3 4 5

5 farm types

Figure 2c – 5-cluster solution with scores for “feed intensity” by farm type
3

2
Feed Intensity

-1

-2
N= 54 15 37 11 19

1 2 3 4 5

5 farm types

30
 Paper 1/5

Table 7 – Details of a 5-cluster solution, giving rise to 5 farm types

FARM INORG ORGANIC TOTAL COMM. PRAWN MILKFISH TOTAL
2
SIZE FERT FERT LABOUR FEED S.D. S.D. (fry/m ) FEEDS
2
(ha) (kg/ha/yr) (kg/ha/yr) (person (%) (fry/m ) (kg/ha/yr)
days/ha/yr)
1 N = 54
Mean 9.24 35.65 52.65 235.52 1.28 4.47 0.09 1908.55
SE Mean 1.09 11.71 24.24 20.20 0.77 0.45 0.01 308.14
2 N = 15
Mean 2.88 31.40 0.00 370.41 6.71 9.42 0.18 16248.34
SE Mean 0.77 14.08 0.00 35.85 5.69 1.76 0.11 3429.52
3 N = 37
Mean 3.97 189.33 214.80 501.00 8.44 1.20 0.16 559.96
SE Mean 0.85 35.55 69.33 65.28 4.05 0.36 0.02 270.91
4 N = 11
Mean 63.73 136.03 2750.48 119.81 11.31 1.63 0.39 1251.60
SE Mean 13.66 40.39 713.00 38.84 9.14 0.82 0.12 705.15
5 N = 19
Mean 6.13 407.08 1113.36 272.05 76.93 0.00 0.71 1858.59
SE Mean 1.46 115.60 379.81 35.28 8.53 0.00 0.12 872.66
All N = 136
Mean 11.08 137.00 457.35 318.37 15.20 3.27 0.23 3063.13
SE Mean 1.80 22.23 102.67 23.13 2.86 0.37 0.03 577.81

5.5 Identifying the Farm Types

One-way Analysis of Variance (ANOVA) were carried out to determine statistically
significant differences between the farm types. Important results from Scheffe’s Post-
Hoc tests are reported in the descriptions below, with full details given in appendix 6.

5.5.1 Type 1 – “Generalists” (n = 54, or 39.7% of sample)

Roughly neutral to all factors; Average size, no specialisation.

When carrying out a typology exercise, it is normal for a certain proportion of the
sample to be average with regards to the variables of interest. By describing farms of
this type as “Generalists”, we observe that they show no degree of specialization in
terms of production practices for either milkfish or prawn, and may therefore flexibly
switch between these two species depending on supply of fry or market conditions.

Given that these are farms with no distinguishing features with regard to the dataset
used in this exercise, it is possible that other variables are more important in
identifying their characteristics. An example is the level of economic diversification
on the farm between prawn, milkfish, crab and tilapia which are the four crops found
to be cultured in our sample. Diversification of the farm between these crops has the
benefit of spreading production risk and possible disease control via biological
interactions between the crops.

Reporting the number of species cultured on the farm is one way of describing the
diversity of the farm. However, this does not give information regarding the relative
importance of each species to the revenue stream of the farm. Metzel and Ateng
(1993), and Irz and Fatch (2004) use Simpson’s Index to report crop diversity on
farms in Bangladesh and Malawi respectively. Here we use the closely related
Shannon Index, a more widely used index that has the attractive property of being

31
 Paper 1/5

bound between 0 (no diversification - the farm produces only one crop) and 1 (perfect
diversification – the farm produces all possible crops, in our case 4, in equal
proportions according to revenue).

The formula for the Shannon Index (H’) is:

H’ = -Σ pi * ln(pi)

where p is the proportion of total revenue from species i.

Hmax = log(S)

where S is the number of species. Equitability (E), the measure reported here is given
by H’/Hmax and produces the index bound between 0 and 1 (Shaw, 2003, p.34).

From the results shown graphically in figure 3, we can observe that farm types 1 and 2
are economically diverse, whereas 3, 4 and 5 are less so. The differences between
farms 1 and 2 in comparison with 3, 4 and 5 are statistically significant at the 5% level
(see table 8 below).

Figure 3 – Revenue diversity indices by farm type

.7

.6

.5
Diversity Index

.4

.3

.2

.1

0.0

-.1
N= 53 12 35 11 19

1.00 2.00 3.00 4.00 5.00

Farm Types

Table 8 – One-way ANOVA with Scheffe’s Post-Hoc test for differences in mean
diversity index between farm types

(J) Farm Std.

(I) Farm Type Type Mean Difference (I-J) Error Sig.
1 2 -0.052 0.073 0.974
3 0.179 0.050 0.015

32
 Paper 1/5

4 0.259 0.076 0.024

5 0.355 0.061 0.000

2 1 0.052 0.073 0.974

3 0.231 0.077 0.066
4 0.311 0.096 0.037
5 0.406 0.085 0.000

Therefore, we can conclude that “Generalist” is a good term to describe farm type 1,
not just in terms of the production practices outlined in the typology, but when
referenced to the level of economic diversification across possible crops.

5.5.2 Type 2 – “Prawn-Oriented Polyculture” (n = 15, or 11.0%)

Farms of this type have negative scores for factor 1, and are therefore oriented
towards the production of prawns. Relative to the sample, labour is more important
than land as a factor of production, as can be seen from small farm size and above-
average labour intensity.

Farms of this type also have high total feeding rate, but with a low percentage of
commercial feeds in the diet (6.7% by weight). From carrying out a one-way
ANOVA, we can see that feeding rates are higher than those of all other farm types
(significant at the 1% level) and that there are higher stocking densities of prawn fry
on these farms than on all other farm types (significant at the 1% level). We therefore
name this type as “Prawn-oriented polyculture”.

By referring to the results of the diversity index shown in figure 3, we can see that
these farms, whilst oriented in their production towards prawns, actually are most
successful in spreading their revenue most evenly across the candidate species for
culture. This would suggest that the farmers attempt to ameliorate the risk associated
with the prawns themselves, in the knowledge of how variable returns from prawn
production can be. This means that while these farms are specialised in prawn
production relative to the sample as a whole, they are not so in absolute terms (i.e.
they are not prawn monoculture systems).

5.5.3 Type 3 – “Low Input, Labour Intensive Farms” (n = 37, or 27.2%)

Farms of this type are neutral to factor 1, and are therefore not specialized with
respect to either prawn or milkfish production. They have positive scores for factor 2
and so labour is much more important as a factor of production than land. From the
results of a one-way ANOVA, we observe that these farms have higher labour use
than farm types 1 and 4 (significant at the 1% level) and farm type 5 (significant at the
5% level). They have negative scores for factor 3 and are very feed-extensive. They
have lower feed-intensities than all other farm types.

The amount of fertiliser used in these systems is moderate and low stocking densities
are used. These really are very extensive farms and there could be interesting issues to
do with access to credit for poor fish farmers for these systems. Alternatively, it may
be that these farms are under-utilised because the motivations of the operator do not
relate to maximising levels of production or profit but to some other objective (e.g. to
have control of land in the area; for leisure as a break from another activity). These
are hypotheses that can be explored.

33
 Paper 1/5

5.5.4 Type 4 – “Large, Milkfish-Oriented Systems” (n = 11, or 8.1%)

Farms of this type are positive for factor 1 and are therefore specialized in milkfish
production. They have negative scores for factor 2 and therefore land is more
important than labour as a factor of production. Feeding rates are approximately
average for the sample.

From the results of a one-way ANOVA, we observe that these farms are larger in size
than farms belonging to all other farm types (significant at the 1% level). It is their
size that defines them primarily, but there is a tendency for milkfish-orientation on
these farms. Large ponds, fertilized with large amounts of manure are used on this
farm type. Labour intensity is lower than for any other farm type and stocking
densities are low.

These farms tend to be operated by local elites and the income from the fish farms,
despite their size, is unlikely to be the main one for the operator. As mentioned
previously, the motivation for entering fish farming may not be production-related
and control of the land may be important for local political objectives held by the
operator. There are a number of social issues relating to these farm types, notably the
potential for land reform.

5.5.5 Type 5 – “Small Milkfish Monoculture Farms” (n = 19, or 14.0%)

Positive scores for factor 1 suggest that these farms are milkfish-specialised. They are
positive for factor 2 and therefore labour is more important than land as a factor of
production. From the results of a number of one-way ANOVA test, we find that these
farms have higher use of commercial feeds than all other farm types (significant at the
1%); have higher stocking densities of milkfish fry than all other farm types
(significant at the 5% level); and have higher use of inorganic fertilizers than all other
farm types (significant at at least the 5% level).

No other species are stocked in these systems and therefore they are entirely
dependent on milkfish for their income. There has been a large increase in the
production of milkfish from non-pond aquaculture (e.g. pens, cages) recently in the
Philippines and so these systems are suffering from lower prices and their margins are
being squeezed. It will be interesting to follow whether diversification into the other
candidate species will follow (along the lines of the “Generalist” farm type) or
whether these operators, whose production practices are entirely set up for milkfish
culture, will try and stick it out in the hope of an increase again in the market price.

6. Geographic Distribution

We can observe that there is a geographically biased distribution for farm types in our
sample. As previously outlined in section 4.1, sampling effort was split almost equally
between two regions – 49% in Central Luzon (region 3) and 51% in Western Visayas
(region 6). As table 9 and figure 4 show below, most prawn-oriented farms are found
in region 3 (80%). Generalists are also somewhat concentrated in region 3 (74%). On
the other hand, low-input high-labour systems are particularly strongly concentrated
in region 6.

34
 Paper 1/5

Table 9 – Values for regional dummy variables

Farm Type % in Region 6
1 – Generalist
0.26
2 – Prawn-oriented polyculture
0.20
3 – Low input systems, labour intensive
0.81
4 – Large, milkfish oriented-systems
0.64
5 – Small milkfish monoculture farms
0.63
Total sample (all farm types)
0.49

Figure 4 – Geographic distribution by farm type

50

40

30

20

10
Reg. 3
Count

0 Reg. 6
1 2 3 4 5

Farm type

7. Conclusion

The results of the principal components analysis are easily interpreted and are in line
with expectations held ex-ante with regard to possible latent variable structure. One
possible cluster solution containing five farm types has been outlined and their
defining characteristics highlighted.

What we have effectively generated with this analysis is a survey of the current
farming systems in the Philippines. While this is useful as an end in itself, we see it

35
 Paper 1/5

primarily as a means to aid comparative study in policy. For instance, in the case of a
single objective (such as increase food security), we can then study protein budgets
for each of the clusters to identify the ‘best’ production system for this objective.
However, we believe that policy regarding sustainable development of the aquaculture
sector requires a multiple-criteria approach to appraisal, as the following two quotes
should help illustrate:

“The super-intensive cage farms are inefficient in nutrient and labour

use….but provide very high returns to land and capital investment.
Clearly there are trade-offs between the use efficiencies of different
resources, and local demand for these resources has implications for the
relative merits of alternative systems.” Michielsens et al (2002, p. 412).

“The extensive farming method finds its validity but also its limits in its
link with natural productivity. It is characterized by a favourable energy
balance, but by rather low production per surface unit.” Ravagnan (1981,
p.66)

In complex cases where there are a number of competing objectives (e.g. maximize
economic performance, minimize ecological impact, maximize social equity) then we
can explore the trade-offs at the farm-level between ecological, social and economic
properties of the farming systems. In addition, we can examine the determinants of
geographical biases in the adoption of particular farming systems. These topics will be
the subject of subsequent papers in this report.

36
 Paper 1/5

References:

Byerlee, D., Collinson, M., Perrin, R., Winkelmann, D., Biggs, S., Moscardi, E.,
Martinez, J. C., Harrington, L. & Benjamin, A. (1980) Planning Technologies
Appropriate to Farmers – Concepts and Procedures, CIMMYT, Mexico City

Clay, J.W. (1997) Toward sustainable shrimp culture, World Aquaculture, 28 (3) 32-
37.

Coche, A. F. (1982) Culture of tilapias in cages. In: “Biology and Culture of

Tilapias”, Pullin, R. S. V. & Lowe-McConnell, R. H. (Eds.), ICLARM, Manila.

Comrey, A. L. & H. B. Lee (1992) A first course in factor analysis, 2nd Edition,
Erlbaum, Hillsdale, N. J., USA, 430pp.

Cruz, P. (1997) Aquaculture feed resource atlas of the Philippines. FAO Fisheries
Technical Paper No. 366. Rome, FAO. 253p.

Hair, J. F., Jr., Anderson, R. E. & Tatham, R. L. (1998) Multivariate Data Analysis,
Prentice Hall, Upper Saddle River, N.J., USA.

Irz, X. T. & Fatch, T. (2004) Does output diversification increase farm income? A test
for Malawi, In Manuscript.

Kaiser, H. F. (1958) The varimax criterion for analytic rotation in factor analysis,
Psychometrika, 23, 3, 187-200

Kautsky, N., P. Ronnback, et al. (2000). Ecosystem perspectives on management of

disease in shrimp pond farming, Aquaculture, 191, 1-3, 145-161.

Kline, P. (1994) An easy guide to factor analysis, Routledge, London, 194 pp.

Kobrich, C., Rehman, T. & Khan, M. (2003) Typification of farming systems for
constructing representative farm models: two illustrations of the application of multi-
variate analyses in Chile and Pakistan, Agricultural Systems, 76, 141-157

Metzel, J. & B. Ateng (1993) Constraints to diversification in Bangladesh: A survey

of farmers’ views, The Bangladesh Development Studies, Vol XXI, 3, 39-71

Michielsens, C. G. J., Lorenzen, K., Phillips, M. J. & Gaulthier, R. (2002) Asian carp
farming systems: towards a typology and increased resource use efficiency,
Aquaculture Research, 33(6), 403-413

Muir, J. F. (1995) Aquaculture development trends: perspectives for food security.

Paper presented at the “International Conference on Sustainable Contribution of
Fisheries to Food Security, Kyoto, Japan, 4-9 Dec 1995”. FAO, Rome.

Primavera, J. H. (1995). Mangroves and brackishwater pond culture in the

Philippines, Hydrobiologia, 295(1-3): 303-309.

37
 Paper 1/5

Ravagnan, G. (1981) Development prospects for coastal aquaculture in the

Mediterranean region: Energy potential of the natural environment. Section in:
“Realism in Aquaculture: Achievements, constraints, perspectives – review papers,
World Conference on Aquaculture, Venice, Italy, 21-25 September 1981” Bilio, M.,
Rosenthal, H. & Sindermann, C.J. (Eds.) European Aquaculture Society, Bredene,
Belgium.

Shang, Y. C. (1981) Aquaculture Economics: Basic Concepts and Methods of

Analysis, Croom Helm, London

Sharma, S. (1996) Applied Multivariate Techniques, Wiley, Chichester

Shaw, P. (2003) Multivariate statistics for the environmental sciences, Arnold,

London.

Stevenson, J. E. (1989) Multivariate statistics II – Using different types of

multivariate statistics to identify types of psychopathology, Nord Psykiatr Tidsskr, 43,
325-332.

Stevenson, J. R., X. T. Irz, J. H. Primavera & G. Sepulveda (2003) Coastal

aquaculture systems in the Philippines: Social equity, property rights and disregarded
duties. Rights and Duties in the Coastal Zone: Multidisciplinary scientific conference
on coastal management, Beijer Institute, Stockholm. CD-Rom available from the
Beijer Institute, or contact the authors for the PDF.

Sumagaysay-Chavoso, N. S. and M. L. San Diego-McGlone (2003) Water quality and

holding capacity of intensive and semi- intensive milkfish (Chanos chanos) ponds,
Aquaculture, 219(1-4): 413-429

Ward, J.H. (1963), Hierarchical Grouping to Optimize an Objective

Function, Journal of the American Statistical Association, 58, 236-244

World Bank (2002) Shrimp Farming and the Environment: Can shrimp farming be
undertaken sustainably? A discussion paper designed to assist in the development of
sustainable shrimp aquaculture, World Bank, Washington D.C.

38
 Paper 1/5

Appendix 1a – Stratification of a net sample size of 70 farms for region 3

Total Farms in
Sample Bataan Bulacan Pampanga Zambales
# Farms > 10 ha 2 (2) 4 (4) 9 (8) 0 (1)
# Farms 2 – 10 ha 6 (5) 7 (7) 21 (19) 0 (3)
# Farms < 2 ha 3 (3) 5 (3) 11 (10) 2 (4)
Total 11 (10) 16 (14) 41 (38) 2 (8)

The table above shows the stratification of the net sample of 70 farms in region 3. The
figures in bold represent the stratification of the actual net sample collected during
fieldwork. The figures in parentheses represent the number that would be completely
representative of the region, according to the Bureau of Agricultural Statistics (BAS)
inventory from 1997. Zambales is deliberately under-represented because, in carrying
out work in that province it was found that most of the farms were actually only
nursery systems (“kawagan”) that supply fingerlings to the grow-out systems in
Bataan, Bulacan and Pampanga. They are not directly comparable in our analysis as
they are not aquaculture grow-out systems and so have been omitted from the net
sample.

Appendix 1b - Stratification of a net sample size of 67 farms for region 6

Total Farms in Negros
Sample Aklan Antique Capiz Guimaras Iloilo Occ.
# Farms > 10 ha 2 (2) 0 (0) 5 (5) 0 (0) 5 (5) 7 (3)
# Farms 2 - 10 ha 8 (7) 0 (1) 10 (10) 0 (1) 4 (5) 7 (8)
# Farms < 2 ha 7 (7) 0 (1) 3 (5) 0 (1) 4 (4) 5 (4)
Total 17 (16) 0 (2) 18 (19) 0 (2) 13 (14) 19 (16)

The table above shows the stratification of the net sample of 67 farms in region 6.
Antique and Guimaras provinces have limited suitable area for fishponds and are only
marginal in the regional production, so were not included in the fieldwork. As before,
the figures in bold represent the stratification of the actual net sample collected during
fieldwork. The figures in parentheses represent the number that would be completely
representative of the region, according to the BAS inventory from 1997.

39
 Paper 1/5

Appendix 2 – Correlation matrix for the 8 variables

a) Correlation Matrix
FARMSIZE INORG ORGANIC TOTAL SUGPO BANGUS TOTAL COMM.
LABOUR S.D. S.D. FEEDS FEED
FARMSIZE 1.000 .001 .182 -.248 -.102 .054 -.120 -.018
INORG .001 1.000 .330 -.015 -.292 .245 -.138 .241
ORGANIC .182 .330 1.000 -.166 -.233 .363 -.134 .245
TOTLABOR -.248 -.015 -.166 1.000 .031 .027 .106 -.031
SUGPOSD -.102 -.292 -.233 .031 1.000 -.295 .312 -.196
BANGUSSD .054 .245 .363 .027 -.295 1.000 .025 .360
TOTFEEDS -.120 -.138 -.134 .106 .312 .025 1.000 -.074
COMMFEED -.018 .241 .245 -.031 -.196 .360 -.074 1.000

b) Significance of correlations
FARMSIZE INORG ORGANIC TOTAL SUGPO BANGUS TOTAL COMM.
LABOUR S.D. S.D. FEEDS FEED
FARMSIZE .497 .017 .002 .118 .267 .081 .417
INORG .497 .000 .432 .000 .002 .054 .002
ORGANIC .017 .000 .026 .003 .000 .060 .002
TOTLABOR .002 .432 .026 .360 .379 .109 .361
SUGPOSD .118 .000 .003 .360 .000 .000 .011
BANGUSSD .267 .002 .000 .379 .000 .385 .000
TOTFEEDS .081 .054 .060 .109 .000 .385 .198
COMMFEED .417 .002 .002 .361 .011 .000 .198

Appendix 3 – Scree Plot showing extraction of principal components

Scree Plot
2.5

2.0

1.5

1.0
Eigenvalue

.5

0.0
1 2 3 4 5 6 7 8

Component Number

40
 Paper 1/5

Appendix 4 – Dendrogram of Cluster Analysis output

Rescaled Distance Cluster Combine
C A S E 0 5 10 15 20 25
Label Num +---------+---------+---------+---------+---------+
86
99
109
87
105 5 4 3
107
108
112
113
117
98
102
44
62
103
73
74
79
110
80
90
85
127
130
131
132
38
126
122
128
93
16
28
18
41
83
106
15
21
13
52
39
104
22
7
89
53
88
17
25
42
51
40
49
9
55
4
10
23
43
5
34
37
26
31
6
27

41
 Paper 1/5

8
14
134
3
1
30
63
2
54
57
33
97
135
68
118
29
114
101
124
92
96
115
116
75
48
56
32
11
36
19
35
58
84
12
111
20
77
47
24
61
72
81
94
69
59
125
70
64
71
82
65
67
95
46
60
66
129
123
78
119
121
133
50
76
91
100
120
45
136

42
 Paper 1/5

Appendix 5 – Analysis of variance (ANOVA)of the differences in

mean factor scores for five farm types using Scheffe's Post-Hoc Test
Mean
95%
Dependent Difference Std. Confidence
Variable Grouping Variables (I-J) Error Sig. Interval
(I) Farm (J) Farm Lower Upper
Type Type Bound Bound
Factor 1 -
Specialism 1 2 0.656 0.165 0.005* 0.140 1.172
3 -0.457 0.121 0.008* -0.834 -0.079
4 -1.658 0.187 0.000* -2.243 -1.073
5 -2.035 0.151 0.000* -2.507 -1.563
2 1 -0.656 0.165 0.005* -1.172 -0.140
3 -1.113 0.173 0.000* -1.654 -0.571
4 -2.314 0.225 0.000* -3.016 -1.612
5 -2.691 0.196 0.000* -3.302 -2.080
3 1 0.457 0.121 0.008* 0.079 0.834
2 1.113 0.173 0.000* 0.571 1.654
4 -1.201 0.194 0.000* -1.809 -0.594
5 -1.578 0.160 0.000* -2.077 -1.079
4 1 1.658 0.187 0.000* 1.073 2.243
2 2.314 0.225 0.000* 1.612 3.016
3 1.201 0.194 0.000* 0.594 1.809
5 -0.377 0.214 0.545 -1.047 0.293
5 1 2.035 0.151 0.000* 1.563 2.507
2 2.691 0.196 0.000* 2.080 3.302
3 1.578 0.160 0.000* 1.079 2.077
4 0.377 0.214 0.545 -0.293 1.047
Factor 2 -
Land vs
Labour 1 2 -1.191 0.204 0.000* -1.830 -0.552
3 -0.726 0.149 0.000* -1.193 -0.259
4 1.346 0.232 0.000* 0.623 2.070
5 -1.296 0.187 0.000* -1.880 -0.713
2 1 1.191 0.204 0.000* 0.552 1.830
3 0.465 0.214 0.324 -0.205 1.135
4 2.537 0.278 0.000* 1.669 3.406
5 -0.105 0.242 0.996 -0.861 0.650
3 1 0.726 0.149 0.000* 0.259 1.193
2 -0.465 0.214 0.324 -1.135 0.205
4 2.072 0.240 0.000* 1.321 2.824
5 -0.570 0.198 0.087*** -1.188 0.047
4 1 -1.346 0.232 0.000* -2.070 -0.623
2 -2.537 0.278 0.000* -3.406 -1.669
3 -2.072 0.240 0.000* -2.824 -1.321
5 -2.643 0.265 0.000* -3.472 -1.814
5 1 1.296 0.187 0.000* 0.713 1.880
2 0.105 0.242 0.996 -0.650 0.861
3 0.570 0.198 0.087*** -0.047 1.188
4 2.643 0.265 0.000* 1.814 3.472

43
 Paper 1/5

95% C.I. 95% C.I.

Dependent Difference Std. Lower Upper
Variable (I) (J) (I-J) Error Sig. Bound Bound
Factor 3 -
Feed
Intensity 1 2 -1.696 0.183 0.000* -2.269 -1.123
3 0.760 0.134 0.000* 0.341 1.178
4 -1.290 0.208 0.000* -1.940 -0.641
5 -0.413 0.168 0.200 -0.937 0.110
2 1 1.696 0.183 0.000* 1.123 2.269
3 2.456 0.192 0.000* 1.855 3.057
4 0.406 0.249 0.619 -0.373 1.185
5 1.283 0.217 0.000* 0.605 1.961
3 1 -0.760 0.134 0.000* -1.178 -0.341
2 -2.456 0.192 0.000* -3.057 -1.855
4 -2.050 0.216 0.000* -2.724 -1.376
5 -1.173 0.177 0.000* -1.727 -0.619
4 1 1.290 0.208 0.000* 0.641 1.940
2 -0.406 0.249 0.619 -1.185 0.373
3 2.050 0.216 0.000* 1.376 2.724
5 0.877 0.238 0.011** 0.133 1.621
5 1 0.413 0.168 0.200 -0.110 0.937
2 -1.283 0.217 0.000* -1.961 -0.605
3 1.173 0.177 0.000* 0.619 1.727
4 -0.877 0.238 0.011** -1.621 -0.133

where (*) denotes significant at the 1% level

where (**) denotes significant at the 5% level
where (***) denotes significant at the 10% level

44
 Paper 1/5

Appendix 6 - Analysis of variance (ANOVA) of the differences in mean values for the
original set of 8 variables for five farm types using Scheffe's Post-Hoc Test
95% Confidence Interval
(J) Mean
(I) Farm Farm Difference Std. Lower Upper
Type Type (I-J) Error Sig. Bound Bound
FARMSIZE 1 2 6.366 4.094 0.660 -6.425 19.157
3 5.271 2.993 0.543 -4.082 14.625
4 -54.485 4.640 0.000 -68.982 -39.987
5 3.116 3.741 0.952 -8.574 14.806
2 1 -6.366 4.094 0.660 -19.157 6.425
3 -1.095 4.293 0.999 -14.509 12.320
4 -60.851 5.568 0.000 -78.248 -43.453
5 -3.250 4.845 0.978 -18.387 11.888
3 1 -5.271 2.993 0.543 -14.625 4.082
2 1.095 4.293 0.999 -12.320 14.509
4 -59.756 4.817 0.000 -74.807 -44.705
5 -2.155 3.959 0.990 -14.525 10.214
4 1 54.485 4.640 0.000 39.987 68.982
2 60.851 5.568 0.000 43.453 78.248
3 59.756 4.817 0.000 44.705 74.807
5 57.601 5.314 0.000 40.997 74.205
5 1 -3.116 3.741 0.952 -14.806 8.574
2 3.250 4.845 0.978 -11.888 18.387
3 2.155 3.959 0.990 -10.214 14.525
4 -57.601 5.314 0.000 -74.205 -40.997
INORG 1 2 4.248 66.822 1.000 -204.545 213.041
3 -153.676 48.861 0.048 -306.347 -1.005
4 -100.380 75.736 0.780 -337.024 136.265
5 -371.430 61.069 0.000 -562.249 -180.612
2 1 -4.248 66.822 1.000 -213.041 204.545
3 -157.924 70.080 0.285 -376.896 61.048
4 -104.628 90.883 0.856 -388.601 179.346
5 -375.678 79.078 0.000 -622.766 -128.591
3 1 153.676 48.861 0.048 1.005 306.347
2 157.924 70.080 0.285 -61.048 376.896
4 53.296 78.625 0.977 -192.377 298.969
5 -217.755 64.618 0.027 -419.661 -15.848
4 1 100.380 75.736 0.780 -136.265 337.024
2 104.628 90.883 0.856 -179.346 388.601
3 -53.296 78.625 0.977 -298.969 192.377
5 -271.051 86.741 0.050 -542.083 -0.019
5 1 371.430 61.069 0.000 180.612 562.249
2 375.678 79.078 0.000 128.591 622.766
3 217.755 64.618 0.027 15.848 419.661
4 271.051 86.741 0.050 0.019 542.083
ORGANIC 1 2 52.650 271.480 1.000 -795.624 900.923
3 -162.153 198.509 0.955 -782.417 458.111
4 -2697.827 307.695 0.000 -3659.256 -1736.398
5 -1060.715 248.110 0.002 -1835.964 -285.465
2 1 -52.650 271.480 1.000 -900.923 795.624

45
 Paper 1/5

3 -214.802 284.716 0.966 -1104.431 674.827

4 -2750.477 369.234 0.000 -3904.191 -1596.762
5 -1113.364 321.273 0.021 -2117.219 -109.510
3 1 162.153 198.509 0.955 -458.111 782.417
2 214.802 284.716 0.966 -674.827 1104.431
4 -2535.674 319.433 0.000 -3533.781 -1537.567
5 -898.562 262.526 0.023 -1718.857 -78.267
4 1 2697.827 307.695 0.000 1736.398 3659.256
2 2750.477 369.234 0.000 1596.762 3904.191
3 2535.674 319.433 0.000 1537.567 3533.781
5 1637.112 352.407 0.000 535.976 2738.249
5 1 1060.715 248.110 0.002 285.465 1835.964
2 1113.364 321.273 0.021 109.510 2117.219
3 898.562 262.526 0.023 78.267 1718.857
4 -1637.112 352.407 0.000 -2738.249 -535.976
TOTLABOR 1 2 -134.881 70.787 0.462 -356.063 86.301
3 -265.478 51.760 0.000 -427.208 -103.748
4 115.713 80.230 0.721 -134.974 366.400
5 -36.528 64.693 0.988 -238.670 165.614
2 1 134.881 70.787 0.462 -86.301 356.063
3 -130.597 74.238 0.544 -362.563 101.368
4 250.594 96.275 0.155 -50.230 551.418
5 98.353 83.770 0.847 -163.396 360.102
3 1 265.478 51.760 0.000 103.748 427.208
2 130.597 74.238 0.544 -101.368 362.563
4 381.191 83.290 0.001 120.941 641.442
5 228.950 68.452 0.029 15.063 442.837
4 1 -115.713 80.230 0.721 -366.400 134.974
2 -250.594 96.275 0.155 -551.418 50.230
3 -381.191 83.290 0.001 -641.442 -120.941
5 -152.241 91.888 0.603 -439.356 134.874
5 1 36.528 64.693 0.988 -165.614 238.670
2 -98.353 83.770 0.847 -360.102 163.396
3 -228.950 68.452 0.029 -442.837 -15.063
4 152.241 91.888 0.603 -134.874 439.356
COMMFEED 1 2 -5.431 6.469 0.950 -25.644 14.781
3 -7.160 4.730 0.683 -21.939 7.620
4 -10.033 7.332 0.759 -32.941 12.876
5 -75.651 5.912 0.000 -94.123 -57.179
2 1 5.431 6.469 0.950 -14.781 25.644
3 -1.728 6.784 0.999 -22.926 19.469
4 -4.601 8.798 0.991 -32.092 22.889
5 -70.220 7.655 0.000 -94.139 -46.300
3 1 7.160 4.730 0.683 -7.620 21.939
2 1.728 6.784 0.999 -19.469 22.926
4 -2.873 7.611 0.998 -26.656 20.909
5 -68.491 6.255 0.000 -88.037 -48.946
4 1 10.033 7.332 0.759 -12.876 32.941
2 4.601 8.798 0.991 -22.889 32.092
3 2.873 7.611 0.998 -20.909 26.656
5 -65.618 8.397 0.000 -91.855 -39.381
5 1 75.651 5.912 0.000 57.179 94.123

46
 Paper 1/5

2 70.220 7.655 0.000 46.300 94.139

3 68.491 6.255 0.000 48.946 88.037
4 65.618 8.397 0.000 39.381 91.855
SUGPOSD 1 2 -4.946 0.976 0.000 -7.997 -1.895
3 3.270 0.714 0.001 1.039 5.501
4 2.838 1.107 0.167 -0.619 6.296
5 4.471 0.892 0.000 1.683 7.259
2 1 4.946 0.976 0.000 1.895 7.997
3 8.216 1.024 0.000 5.017 11.416
4 7.785 1.328 0.000 3.635 11.934
5 9.417 1.155 0.000 5.807 13.028
3 1 -3.270 0.714 0.001 -5.501 -1.039
2 -8.216 1.024 0.000 -11.416 -5.017
4 -0.431 1.149 0.998 -4.021 3.158
5 1.201 0.944 0.805 -1.749 4.151
4 1 -2.838 1.107 0.167 -6.296 0.619
2 -7.785 1.328 0.000 -11.934 -3.635
3 0.431 1.149 0.998 -3.158 4.021
5 1.633 1.267 0.798 -2.328 5.593
5 1 -4.471 0.892 0.000 -7.259 -1.683
2 -9.417 1.155 0.000 -13.028 -5.807
3 -1.201 0.944 0.805 -4.151 1.749
4 -1.633 1.267 0.798 -5.593 2.328
BANGUSSD 1 2 -0.092 0.080 0.862 -0.343 0.160
3 -0.075 0.059 0.805 -0.259 0.109
4 -0.296 0.091 0.038 -0.581 -0.011
5 -0.622 0.074 0.000 -0.851 -0.392
2 1 0.092 0.080 0.862 -0.160 0.343
3 0.017 0.084 1.000 -0.247 0.280
4 -0.204 0.109 0.484 -0.546 0.138
5 -0.530 0.095 0.000 -0.828 -0.233
3 1 0.075 0.059 0.805 -0.109 0.259
2 -0.017 0.084 1.000 -0.280 0.247
4 -0.221 0.095 0.252 -0.517 0.075
5 -0.547 0.078 0.000 -0.790 -0.304
4 1 0.296 0.091 0.038 0.011 0.581
2 0.204 0.109 0.484 -0.138 0.546
3 0.221 0.095 0.252 -0.075 0.517
5 -0.326 0.104 0.050 -0.652 0.000
5 1 0.622 0.074 0.000 0.392 0.851
2 0.530 0.095 0.000 0.233 0.828
3 0.547 0.078 0.000 0.304 0.790
4 0.326 0.104 0.050 0.000 0.652
TOTFEEDS 1 2 -14339.789 1432.247 0.000 -18815.02 -9864.563
3 1348.594 1047.270 0.798 -1923.727 4620.915
4 656.947 1623.302 0.997 -4415.253 5729.148
5 49.955 1308.951 1.000 -4040.021 4139.930
2 1 14339.789 1432.247 0.000 9864.563 18815.015
3 15688.383 1502.073 0.000 10994.977 20381.789
4 14996.736 1947.962 0.000 8910.096 21083.376
5 14389.743 1694.934 0.000 9093.719 19685.768
3 1 -1348.594 1047.270 0.798 -4620.915 1923.727

47
 Paper 1/5

2 -15688.383 1502.073 0.000 -20381.79 -10994.977

4 -691.647 1685.230 0.997 -5957.350 4574.056
5 -1298.640 1385.008 0.927 -5626.262 3028.982
4 1 -656.947 1623.302 0.997 -5729.148 4415.253
2 -14996.736 1947.962 0.000 -21083.38 -8910.096
3 691.647 1685.230 0.997 -4574.056 5957.350
5 -606.993 1859.188 0.999 -6416.248 5202.262
5 1 -49.955 1308.951 1.000 -4139.930 4040.021
2 -14389.743 1694.934 0.000 -19685.77 -9093.719
3 1298.640 1385.008 0.927 -3028.982 5626.262
4 606.993 1859.188 0.999 -5202.262 6416.248

48
 Paper 2/5

Research project R8288: Assessing the sustainability

of brackish-water aquaculture systems in the
Philippines – Paper 2/5

Sustainability indicators for brackish-water pond

aquaculture in the Philippines: Economic, social and
ecological dimensions

Abstract
The concept of sustainable development is commonly associated with a balance
between economic viability, social equity, and ecological sustainability. In this paper
we critically review these issues with respect to aquaculture in the Philippines and
introduce specific objectives to characterise these properties: maximise profit;
minimise risk; maximise technical efficiency; minimise nutrient loss; maximise net
dietary protein production; and maximise employment.

We describe methodologies for generating specific indicators of these objectives for

aquaculture at the farm level and present results for five farming system types in
brackish-water ponds in the Philippines. We also consider some other related
analyses (e.g. ecological footprints, mortality by different species, pesticide
application) that help us build a picture of the complexity of the farming systems.

Overall, we find that semi-intensive prawn-oriented polyculture has the highest level
of profit (gross margin per unit area) but at the cost of low achievement in other
objectives – most notably nutrient loss, technical efficiency and risk. Extensive
polyculture performs well with regard to most indicators except for employment
effect, the highest values for which are seen in low-input labour-intensive systems.
Semi-intensive milkfish monoculture systems and very large milkfish-oriented systems
would seem to perform averagely or poorly for most indicators. The results for all
indicators are summarised, but analysis of the implications is reserved for paper 3.

50
 Paper 2/5

Contents
1. Introduction 52

2. Review of sustainability indicators 52

3. Brackish-water aquaculture in the Philippines: A critical review of the

economic, social and ecological issues 55
3.1 Economic viability 55
3.2 Social equity 58
3.3 Environmental sustainability 60

4. Issues of scale 63

5. Sustainability: Operationalising a utopian concept 66

5.1 Economic objectives 66
5.2 Socio-economic / social objectives 67
5.3 Ecological objectives 67
5.4 Sustainability indicators 68

6. Economic indicators 70
6.1 Profit 70
6.2 Risk 75
6.3 Technical efficiency and its determinants 80
6.4 Summary 93

7. Socio-economic / social indicators 96

7.1 Introduction 96
7.2 Employment generation 97
7.3 Aquaculture and food security 110
7.4 Summary 116

8. Ecological indicators 117

8.1 Introduction 117
8.2 Ecological footprint 117
8.3 Nitrogen balance 126
8.4 Pesticide application 128
8.5 Summary 129

9. Conclusion 130

Appendix 142

51
 Paper 2/5

1. Introduction
Indicators are measurable properties of systems and are the most widely-used means
through which broad notions of “social good” (following Milon, 1987) can be used
for comparative study. Given the particular focus of this study, our choice of a set of
indicators is more specifically grounded in the literature on sustainable development.
There follows a partial review of studies that have used sustainability indicators in the
analysis of farming systems, and, subsequently, a brief discussion of our own choice
of indicators. The selected indicators are then presented in details and used for
comparative analysis of the five farming systems identified in paper 1.

2. Review of sustainability indicators

Indicators are, in a general sense, used to provide information about a complex system
(e.g., agroecosystem), or an unmeasurable concept, such as that of sustainability
(Bockstaller and Girardin, 2003). Their use in the analysis of sustainability issues is
justified by Bell and Morse (1999) as a way of giving a degree of rigour to the study
of what are, ultimately, values held by people. However, it should be recognised from
the outset that devising sustainability indicators is not an exact science, but, instead,
represents a synthesis of different perspectives on a problem to give insight to
decision-making. Hence, the approach is a paradigm in the Kuhnian sense – the
paradigm of sustainability – combining the objective measurement of a subjective
choice of indicators. Ultimately, selecting sustainability indicators is not inferior to
more rigorous approaches in science, but merely different, and the subjectivity of the
exercise is laid bare.

In practice, the process of developing indicators is itself a compromise integrating the

scientific knowledge of the moment, the need for conciseness, simplicity of use, and
the availability of data (Girardin et al. 1999). Conceptually, in the case of agriculture
and related natural resource systems, there are two broad categories of indicators,
which can be characterised as means-based and effect-based (van der Werf and Petit,
2002). Means-based measures relate to production practices, with an assumed link to
the outcome of this practice. An example would be “amount of nitrogen fertiliser
applied” with an assumed, but indirect, link to an objective that is to be minimised
(e.g. eutrophication of a river). Effect-based measures relate to the direct outcomes of
production practices and are to be preferred. An example would be “amount of nitrate
lost to a river”, which gives much more relevant and precise information to the
decision-maker concerned with the river ecosystem than the equivalent means-based
measure. However, effects-based measures are more data intensive, and the marginal
value of the information gained over a means-based indicator has to be weighed up
against the costs in extra data requirements.

Any given application of the sustainability indicator paradigm may invoke a single
global indicator or, more commonly, a number of different indicators. We focus our
review on applications that use a multiple-indicator approach, owing to the conceptual
difficulties associated with single, global indicators. Such difficulties are primarily
centred on the problem of incommensurability of values, both in technical terms (i.e.
adding apples and oranges together) and in social terms (i.e. what is important in the
decision problem).

52
 Paper 2/5

Van der Werf and Petit (2002) review twelve applications of sustainability indicators
to different agricultural systems. At a technical level, they find great diversity in the
number of indicators selected (from 2 to 13), the manner with which interactions
between indicators are treated, and whether scores (assigned by the researcher) or
measured values are used. In terms of their broader aims, studies using sustainability
indicators may assess environmental objectives only, or may include indicators for
social and economic objectives as well. This choice of scope relates to whether or not
the research team are interested in “environmental” or “ecological” sustainability as in
the former case, or in “sustainability” or “sustainable development” as in the latter
case. Beyond these general considerations, sustainability indicators are used to
investigate a wide range of issues, as we now attempt to illustrate.

Nutrient disequilibrium is a common theme to many case-studies, and nutrient

budgets then represent the preferred methodology for calculating a related indicator.
Funge-Smith and Briggs (1998), for instance, compute complete nutrient budgets for
intensive prawn aquaculture systems in Thailand. The limitation of this kind of total
accounting of nutrient flows is, however, that it can usually only be carried out on a
few experimental farms owing to heavy data requirements. In an attempt to address
this issue, Islam et al. (2004) and Wijnhoud et al. (2003) provide methods for a less
comprehensive accounting, which offers the possibility of application to a greater
number of farms. Whatever the method used, nutrient loss, expressed in kilogram per
hectare or unit of product, is the related indicator for these studies.

Energy-use is another objective for which indicators have been developed, and may
be expressed as total energy used in the system, or energy used per unit of product or
land (Bailey et al., 2003). For instance, expression of an indicator per unit area of
farming system is particularly useful in comparing agricultural practices. On the other
hand, calculating energy requirements per unit of product is similar to what is done in
Life-Cycle Assessment (LCA), where energy flows from the production, use, and
disposal of a product are considered, and the focus is thus more on the impact of
consumption trends (Heller and Keoleian, 2003) rather than that of different
production practices. It is worth pointing out that LCA has seen recent application to
commercial feeds for aquaculture (Papatryphon et al., 2004), and that this type of
analysis implicitly assumes technical homogeneity (i.e., that the impact of the use of
one unit of feed is the same on all farms), which represents an obvious simplification
of reality.

In an attempt to measure the sustainability impact of aquaculture, Prein et al. (1998)

use four indicators to characterise farming systems before and after the integration of
aquaculture into them. Their indicators are: net income, expressed as a monetary
value; recycling and diversity, both expressed as scores; and production capacity,
expressed as a quantity of product per unit area. The results are displayed in a kite
diagram to visually demonstrate the Pareto dominance of integration. Along similar
lines, another research team (Dalsgaard et al., 1995; Dalsgaard and Oficial, 1997)
base their approach on four key ecological properties of sustainable systems and
suggest four related indicators: diversity, cycling, stability and capacity. Their
analysis is connected to the ECOPATH II mass-balance modelling software for
modelling ecosystems (Christensen and Pauly, 1992). ECOPATH models represent
trophically linked biomass „pools‟ and are parameterised by satisfying two master
equations describing production and consumption within an ecosystem (or farm).

53
 Paper 2/5

ECOPATH-based models have now seen hundreds of applications in the literature,

mostly in capture fisheries. An important recent paper uses ECOPATH in association
with a series of objective functions in economic, ecological and social dimensions
(Christensen and Walters, 2004), to examine the consequences of alternative fisheries
management policies. The authors find that the trade-offs between objectives are more
significant than might have been expected, thereby showing the relevance of a multi-
objective approach.

Particularly relevant to our own study, Rasul and Thapa (2003) compare ecological
and conventional agricultural systems in Bangladesh by measuring sustainability
indicators derived from a large farm-level survey. Their work is of interest because of
its focus on a developing country and its coverage of economic, ecological and social
objectives, as detailed in Table 1. The table shows the selection of profitability, an
index of stability across farms, equity (in the form of employment per hectare) and
food security – all important indicators for which specific methodologies are
developed in this paper. In the end, the authors establish that the two farming systems
differ significantly with respect to some indicators but not others, and that, overall,
ecological agriculture is relatively more sustainable.

Table 1 – Objectives, indicators and specific methods of analysis used by Rasul and
Thapa (2003)
Broad objective Indicator Specific method
Ecological Land-use pattern Proportion of land under field crops,
sustainability homestead and orchard
Cropping pattern Cropping intensity, crop
diversification, mixed cropping
Soil fertility management Proportion of farmers using
inorganic and organic fertilizers, and
area covered by each type
Pest and disease management Proportion of farmers using
biological, mechanical, and chemical
methods
Soil fertility status Chemical analysis of soil samples
collected from both kinds of farms
Economic Land productivity Crop yields
sustainability
Yield stability Index of yield trend across farms
Profitability Financial return, economic return
and value added per unit of land
Social Input self-sufficiency Ratio of local inputs cost to total
acceptability inputs cost
Equity Employment per ha
Food security Adequacy of food grain produced
Risks and uncertainties Index

Altogether, the generation of sustainability indicators has become an industry in itself

(King et al., 2000) so that the above review is far from exhaustive. However, and
perhaps surprisingly, there is little consensus on the proper way in which indicators
should be selected and used. This can be explained by the fact that sustainability is a
multifaceted concept - something which has unfortunately not always been recognised
in the literature. As a result, the definition and measurement of sustainability
indicators draw on different literatures when used to compare different kinds of
systems (e.g. an economy, a farm) at different scales (e.g. country-level, farm-level),

54
 Paper 2/5

and there has been little by way of a consistent underlying conceptual framework. The
objective of the following section is to define the issues relevant to the selection of
indicators for aquaculture in the Philippines.

3. Brackish-water aquaculture in the Philippines: A critical review of the

economic, social and economic issues
This section reviews the economic, social and environmental issues that aquaculture
in the Philippines is currently facing, based on the existing literature and, in some
cases, data collected whilst carrying out a farm-level survey between January and
June 20031. In addition, semi-structured interviews, direct observation, personal
participation in on-farm and off-farm activities, and life-history discussions with key
informants are all used to develop a greater understanding of the issues. The emphasis
is placed primarily on description but linkages to theories in the relevant literature are
made wherever possible. Our choice as over-riding themes of “economic viability”,
“social equity” and “environmental sustainability” is influenced by the sustainability
literature, in particular Adger, et al. (2003) and Brown et al. (2001).

3.1 Economic viability

The most comprehensive study of aquaculture economics is that based on the survey
carried out by the Asian Development Bank (ADB) and the Network of Aquaculture
Centres in Asia-Pacific (NACA). The related dataset is enormous, covering the ten
prawn producing countries in the region (Thailand, Indonesia, China, Vietnam,
Bangladesh, Philippines, India, Taiwan, Malaysia, Sri Lanka) and has led to a number
of papers. Most important among these is that by Shang et al. (1998) on the
comparative economics of prawn production and marketing in the ten Asian countries
listed above, which we now attempt to review.

3.1.1 Relative economic performance of Philippine aquaculture

Using data from 1994, Shang et al. (1998) describe the cost and revenue structures of
extensive, semi-intensive and intensive prawn grow-out operations in Asian countries,
as summarized in Table 2 for extensive farms. In terms of profitability, the Philippines
top the set of countries studied with a profit rate of US$4.67 per kilo for extensive
systems. Semi-intensive farms also return a healthy level of profit of US$ 2.54 per
kilo, which is at the upper end of the range for all Asian countries. However, for
intensive farming systems, the level of profit in the Philippines is low, at US$ 0.29 per
kilo, which ranks the country second bottom among the ten Asian countries. Hence,
the results suggest that the Philippines are internationally competitive in prawn
production for relatively extensive systems only. The evolution of the industry since
Shang et al. (1998)‟s study was published confirms that view as virtually all the farms
currently in commercial operation in the Philippines can be described as semi-
intensive or extensive (according to the definition adopted by these authors).

We now try to explain the profitability levels described above. Table 2 shows that, for
extensive systems, the Philippines are characterized by a relatively low level of land
productivity, an average total cost of production but the highest level of profit per unit
output. These apparently paradoxical findings can be reconciled when output prices
are considered, because the Philippines have the highest farm-gate price for
extensively-raised prawn at US$ 7.28 per kilo. This relatively high price is likely due

55
 Paper 2/5

to the fact that extensive farms in the Philippines raise relatively large prawns,
although Shang et al. (1998) do not report the average individual size for each
country. The other production systems in the Philippines do not enjoy the same level
of price advantage when compared to other countries. Hence, Philippine semi-
intensive prawn farms obtain an average farm-gate price of US$ 6.55 per kilo and
intensive farms achieve US$ 7.10 per kilo, which ranks fifth and fourth respectively
among eight Asian countries.

Table 2 – Farm-level economics of extensive prawn production in Asia, from Shang

et al. (1998)
FARM-LEVEL
Thailand Indonesia Philippines Vietnam India Bangladesh Sri Lanka China
CHARACTERISTICS

Total Number of
farms 2002 24000 3029 22374 45040 6500 80 2681
Sample size
7 1024 68 296 744 163 17 117
% of total farms
0 4 2 1 2 3 21 4
Average farm size
(ha) 12.2 5 10.8 10.3 1.2 16.6 5 39.5
Stocking density
(PL/m2) 0 3.1 1.7 0.3 3.7 1.5 14.9 7.9
Feed Conversion
Ratio 0 0.3 0.4 0.1 1.2 0.4 0.3 1.4
Number of crops / yr
2.4 2 1.7 1.3 1.3 1.2 1.7 1
Production (kg/ha/yr)
394 162 260 79 696 216 2944 421

COST STRUCTURE Thailand Indonesia Philippines Vietnam India Bangladesh Sri Lanka China
($/kg)
Fixed Costs 0.9 1.2 1.01 2.01 1.33 1.34 1.92 0.74
Overhead 0.68 0.4 0.18 1.07 0.28 0.99 1.33 0.13
Depreciation 0.22 0.8 0.83 0.93 0.99 0.35 0.59 0.54
Interest 0 0 0 0.01 0.07 0 0 0.06
Variable costs 0.84 2.66 1.6 1.04 3.08 2.73 1.52 0.88
Feed 0 0.22 0.49 0.17 1.39 0.13 0.31 0.29
Seed 0 1.26 0.53 0.42 0.99 1.77 0.76 0.13
Power 0.46 0.01 0.03 0.04 0.27 0.04 0.12 0.1
Labour 0.08 0.65 0.37 0.24 0.26 0.56 0.31 0.13
Other 0.3 0.52 0.19 0.16 0.18 0.23 0.02 0.23
Total Cost 1.74 3.86 2.61 3.04 4.42 4.07 3.45 1.62
Farm-gate price 3.63 6.84 7.28 2.73 7.19 6.9 7.05 3.05
Profit (US$ /kg) 1.89 2.98 4.67 -0.31 2.77 2.83 3.6 1.43

COST STRUCTURE Thailand Indonesia Philippines Vietnam India Bangladesh Sri Lanka China
(% Total Cost)
Fixed Costs 51.8 31.1 38.5 66 30.2 32.9 55.8 45.4
Overhead 39.2 10.3 6.9 35.1 6.3 24.3 38.6 7.8
Depreciation 12.6 20.7 31.7 30.6 22.4 8.5 17.2 33.7
Interest 0 0.1 0 0.2 1.5 0 0 3.9
Variable costs 48.2 68.9 61.5 34 69.8 67.1 44.2 54.6
Feed 0 5.8 18.7 5.6 31.4 3.3 8.9 17.8
Seed 0 32.7 20.2 13.9 22.4 43.5 22.2 8.2
Power 26.3 0.2 1 1.3 6 1 3.5 6.2
Labour 4.6 16.7 14.4 8 6 13.8 9.1 8.1
Other 17.3 13.4 7.2 5.2 4.1 5.6 0.5 14.4

56
 Paper 2/5

For intensive systems, the cost of production per kilo is second highest in the
Philippines at US$ 6.81. and relates to the relatively high costs of variable inputs.
Hence, Shang et al. (1998) establish that the costs of feed (US$ 2.61 per kilo of
output), seed (US$ 1.27 per kilo) and labour (US$ 0.43 per kilo) are higher in the
Philippines than in other countries for intensive farms, which is explained by high
variable input prices, as well as, possibly, low levels of technical efficiency of the
farms. This cost structure is responsible for the low level of profitability of intensive
farms in the Philippines. Given the level of financial and production risk associated
with intensive systems, this level of profit can be considered economically non-viable,
as has been suggested by the continuing pattern of de-intensification or abandonment
of formerly intensive shrimp farms in the Philippines.

The other production systems present more satisfactory cost structures. For semi-
intensive farms, the Philippines have a middle-ranking production cost of US$ 4.01
per kilo. Feeds represent the main cost for all countries for these systems, but the
Philippines report the highest cost share for this item (over 55%). For extensive
systems, the Philippines produces at US$ 2.61 per kilo and seed is the most important
variable cost. One can hypothesize that these more satisfactory cost structures, when
compared to that of intensive systems, relates to the fact that non-intensive farms
manage to substitute relatively expensive commercial feeds with other cheap sources
of protein, such as small shells (as described in paper 1).

Altogether, we conclude from this review that the Republic of the Philippines appears
internationally competitive in extensive and semi-intensive prawn production.
However, the high costs of commercial feeds in the Philippines mean that the
competitive position of the country is dependent on the availability of „natural feeds‟,,
so that one could argue that the continued success of exports from the Philippines
depends on the health of the coastal ecosystem.

3.1.2 Within-Country Heterogeneity

There is a long history of brackish water aquaculture in the Philippines, and the
techniques currently used represent combinations of traditional practices and modern
technologies that are often not familiar to aquaculture technicians. As a rule, farmers
adopt only partially, and to different degrees, the bundle of technical innovations that
have been developed in the past few decades. These innovations, in terms of inputs,
include commercial pellet feeds, hatchery-bred prawn fry, commercial inorganic
fertilizers, hatchery-bred tilapia fry, and saline-tolerant tilapia fry. In addition,
techniques for the management of pond conditions and stock growth come from
innovations introduced by operators and from technicians at national and international
research institutions. These include the use of nursery ponds (operator innovation),
modular systems of production (improved traditional), salinometers to measure
salinity of source water (private sector innovation), laboratory testing of fry for the
presence of disease agents (innovation from Negros Prawn Producers Cooperative),
green-water technology (innovation of SEAFDEC-UPV), closed-recycling systems
(operator innovation and SEAFDEC tested). Aquaculture operators draw on these
technologies and techniques in various combinations according to the constraints that
they face (e.g. working capital, market access, environmental conditions).

Thus, we have to recognize the complex combinations of strategies that fishpond

operators employ to remain profitable. Intensive aquaculture is usually fairly

57
 Paper 2/5

standardized, but this is not the case for extensive and semi-intensive systems. This
heterogeneity is partly historical (some current systems were adapted from traditional
milkfish ponds, some newly built on mangrove acid-sulphate soils or on mudflats),
and partly as a result of differences in the specific objectives and constraints faced by
the operators.

3.2 Social Equity

3.2.1 Distribution of fishpond landholdings
The distribution of farm size is perhaps the most important factor in determining
social equity for the general agricultural sector in developing countries and appears
relevant in assessing the social impact of aquaculture in the Philippines. Using data
from the Bureau of Agricultural Statistics (BAS) inventory of fishponds from 1997,
we compare the distribution of land ownership in regions 3 and 6 by plotting Lorenz
curves and calculating Gini coefficients (Figure 1)2.

Figure 1 – Lorenz curves for regions 3 (top) and 6 (bottom)

Cumulative share of fishpond area

1.2

1
The Gini coefficient is the ratio
of the area enclosed by the
0.8
Lorenz curve and the diagonal
0.6
(G), and the area under the
0.4 G diagonal (i.e. Gini coefficient =
0.2 G/0.5 = 2G). The greater the
0 inequality in land distribution,
0

1
0.5
0.06
0.12
0.19
0.25
0.31
0.37
0.44

0.56
0.62
0.68
0.75
0.81
0.87
0.93

the greater the value of the Gini

Cumulative share of population coefficient up to a maximum of 1
where all the land is owned by
1.2 one member of the population.
Cumulative share of fishpond area

0.8

0.6

0.4
G
0.2

0
0

0.3

0.6

0.9
0.06
0.12
0.18
0.24

0.36
0.42
0.48
0.54

0.66
0.72
0.78
0.84

0.96

Cum ulative share of fishpond population

The two regions are almost identical in their distribution of fishpond unit sizes, as
0.7192bfor– Lorenz
summarized by Gini coefficients of 0.715 for region 3 andFigure region 6. Thisfor
curve
compares unfavourably with Thailand which had a Gini coefficient
region 6. of 0.42 in 1988
for the distribution of land holdings in general. Further, it should be noted that the
analysis is based on the total number of brackish water fishpond units, and so these
results are an indication of the concentration of pond ownership only among those
engaged in the industry. If data on family landholdings for the entire household
population in these regions were available, the Gini coefficient for the distribution of
all landholdings in general would approach unity for both regions. Altogether, it is

58
 Paper 2/5

clear that the land supporting brackish water pond aquaculture in the Philippines is
very unequally distributed.

3.2.2 Absenteeism
It is well known that many operators do not live close to their fishponds, and this
phenomenon has implications both for management and for social equity. The need
for the operator to have a trusted caretaker can often take precedence over the desire
for the caretaker to be technically competent. The central importance of trust in the
caretaker-operator relationship means that operators that live outside the area where
the fishpond is situated often bring their own employees from outside to be caretakers.
This can be a cause for tension between absentee operators and the local communities.

Hence, while the rationale for aquaculture as a rural development strategy is often
cited, the predominance of absentee operators lessens the potential benefits to the
rural areas. Profits are spent outside the region, thus eliminating any potentially
beneficial multiplier effects through consumption. Indeed, in the case of prawn
aquaculture in Honduras, Stanley found that the farms operated as an enclave,
separate from the community and the local economy, rather than as a stimulus to
growth in the coastal regions (Stanley, 2003).

3.2.3 Impacts on the livelihoods of fishermen

Aquaculture has both positive and negative impacts on the livelihoods of the people
dependent on the rivers and seas for their livelihood. The negative impacts are organic
pollution and thus a deterioration in environmental quality; a loss of habitat for
commercially important species; and loss of an open-access multi-functional resource
in the case of conversion of mangroves to new fishponds. Positive impacts are
primarily through employment in the aquaculture sector both directly (as caretakers,
labourers) and indirectly (supplying inputs such as small shells, fry and by-catch for
fishmeal).

The fisherfolk lobby in the Philippines is large in number but weak in representation
and voice at high levels of governance. A number of non-governmental organizations
(NGOs) represent the interests of fisherfolk with the rationale that they are among the
poorest members of society. In addition, the activity of municipal fishing is seen by
many as culturally important and worthy of support.

3.2.4 Impacts on food security

There is an equivocal literature on the benefits of aquaculture to food security. Prawns
and crabs are recognised as serving no food security purpose at the national level, as
they are luxury products and may actually reduce global protein supply significantly
(Naylor, Goldburg et al., 2000). Conversely, milkfish and tilapia are an important part
of the daily diet for most social classes in the Philippines. When prices of these fish
fall during periods of local market saturation (harvests from fishponds often occur in
clusters around the full moon where the tidal range is greatest) they may even be
affordable to the poor. In addition, immediately following harvests, people are often
allowed to enter the drained pond to collect any fish left behind by the harvest
labourers. This represents a possibility for free fish for the poor, although it merely
mimics the de facto harvesting-rights for local people that would operate under open-
access to mangroves and other inter-tidal habitats.

59
 Paper 2/5

3.2.5 Increased vulnerability of coastal villages to typhoons and

flooding
Fishpond areas do not provide protection to coastal villages from strong winds, tidal
waves and tsunamis. The shearing effect of a mangrove in the same area dissipates the
force of incoming water. Villagers in Lat-Asan, Capiz, Philippines experienced this
first hand and shared their experiences during key-informant interviews in April 2004.
Mangroves were the only structures to survive in their village when a tidal wave
struck in the 1980s. Most of the residents that survived were those that had managed
to climb the mangrove and coconut trees in sufficient time.

In the aftermath of the tsunami that engulfed areas of Indonesia, India, Thailand and
Sri Lanka on 26th December 2004, questions are being raised as to whether the
development of the coastline in these areas contributed to the extent of the
devastation. Aquaculture is not the only driver of change in land-use in the coastal
areas of these countries, but it is certainly a significant one. The issue of coastal
resilience to such events is likely to now become more widely understood and hold
higher priority than was previously the case in countries pursuing aquaculture
development.

3.3 Environmental Sustainability

3.3.1 Initial land clearance for the ponds
Built on inter-tidal lands usually cleared of its cover of saline-adapted vegetation (e.g.
mangroves, nipa palms), fishponds alter the functioning of the coastal ecosystem. A
typical tropical inter-tidal zone covered in mangroves provides a number of
extractable goods (e.g. firewood, crabs that can be collected from the mud, honey,
tannins - Kaplowitz, 2001), but it is the ecological and hydrological services that
represent their most important sources of value (Gilbert and Janssen, 1998; Nickerson,
1999; Barbier, 2000; Janssen, Gilbert et al., 2000; Ronnback and Primavera, 2000;
Janssen, Gilbert et al., 2001). When these habitats are destroyed to make way for
fishponds, these services are lost. Therefore the mere presence of fishponds in an area
is disruptive and has a negative impact on the functioning of the coastal ecosystem.
The resultant loss of mangroves from the Philippines represents a feature of
aquaculture development over the past 50 years that is alarming to some (Primavera,
1995).

3.3.2 Environmental externalities during operation

3.3.2.1 Exogenous to aquaculture
(i.e. from fishpond to environment in general)
The release of organic pollution from the farms, and subsequent nutrient enrichment
of surrounding waters has been documented in a number of cases (Funge-Smith and
Briggs, 1998; Tovar et al., 2000; Islam et al., 2004). Nutrient enrichment occurs when
the nutrient input to the pond (in the form of feeds and fertilisers primarily) exceeds
the rate at which the fish assimilate the nutrients (directly through feed or indirectly
through fertilisers, via primary production in the pond). Therefore, when water
exchange occurs, the nutrients are released into the surrounding water bodies and can
cause excessive primary production.

It is not easy to assess complete nutrient flows, particularly in low-data environments

(Janssen, 1999). However, in an example of a complete nitrogen budget in
aquaculture, Funge-Smith and Briggs (1998) analyse experimental pond systems for

60
 Paper 2/5

production of shrimp in Thailand. Their results show that in intensive shrimp ponds,
the majority of the total nitrogen input (78%) to the farming system is introduced in
the form of pelleted feed. Erosion of the pond bottom provides 16 per cent of N
input, with negligible inputs from all other sources.

The composition of nitrogen outflows is more diverse. Nitrogen embodied in

harvested shrimps accounts for only 18 percent of all outputs. The fate of the
remaining nitogen produced by these systems can be summarized as follows:
denitrification and ammonia volatilisation (30%); sedimentation (24%); loss to
surrounding waters during water exchange (17%); and loss to surrounding waters
during harvest drainage (10%). Nitrogen therefore becomes loaded in the effluent
(27% of the total), which can potentially seriously enrich the receiving waters
(Funge-Smith and Briggs, 1998). This conclusion seems relevant in the context of the
study areas as it is supported by casual observation of local management practices. In
particular, the “pulse” of effluent emission when the harvest takes place is well-
known to fish farmers in the Philippines, who respond by avoiding as much as
possible water exchanges in the days following a neighbours‟ harvest. Of course, we
acknowledge that the farms in our sample do not operate at the same level of
intensity as those investigated by Funge-Smith and Briggs, but they still have high
levels of external inputs, which gives relevance to the study of nutrient flows
presented later in this paper.

In an attempt to address the problems created by nutrient disequilibrium in

aquaculture, new technologies have recently been developed. For instance, in a
closed-recycling water system, intensive aquaculture can proceed without the daily
water exchange that causes chronic environmental stress. However, in an
experimental study (Thakur and Lin, 2003), the water drained from ponds (during
harvest) of this kind still contained between 14 and 28 percent of total N input. The
“pulse” effect of this emission to the environment remains a problem, although
closed-recycling clearly delivers some environmental benefits and should help
decrease inter-farm transmission of diseases.

We now turn to the evidence regarding the link between nutrient enrichment and
environmental damage. Figure 2 illustrates the relationship between Phosphorus input
and biodiversity that Merrington et al. (2002) established for freshwater systems. If,
as suggested by the literature, this relationship is a general one that applies to brackish
waters as well, it must have important consequences for the health of the ecosystems
supporting the fish ponds in the Philippines.

61
 Paper 2/5

Figure 2 – Relationship between nutrient enrichment, biodiversity, dissolved oxygen

and primary production (redrawn from Merrington, Winder et al., 2002).

100 Primary production

Index P-input

50

Biodiversity
0 Dissolved oxygen
Nutrient Enrichment
Oligotrophic Mesotrophic Eutrophic

Ranges for the 3 states are as follows:

Oligotrophic – Total P<0.01 mg l-1
Total N<0.2 mg l-1
Hypolimnetic dissolved oxygen (% saturation) > 80
Phytoplankton production (g C m-2d-1) 7-25

Mesotrophic – Total P = 0.01 – 0.02 mg l-1

Total N = 0.2 – 0.5 mg l-1
Hypolimnetic dissolved oxygen (% saturation) =10-80
Phytoplankton production (g C m-2d-1) = 75-250

Eutrophic – Total P > 0.02 mg l-1

Total N > 500 mg l-1
Hypolimnetic dissolved oxygen (% saturation) <10
Phytoplankton production (g C m-2d-1) 350-700

The final externality generated by aquaculture relates to its use of pesticides, which
represents a source of pollution to the surrounding waters with dramatic short-term
consequences. The toxicity of sodium cyanide (a widely used piscicide) in particular
represents a significant human health risk.

3.3.2.2 Endogenous to aquaculture

(i.e. fishpond to environment to fishpond)
Aquaculture is dependent on the coastal environment. In the language of systems
ecology we can consider fishponds as open systems within the coastal ecosystem.
What makes aquaculture such a unique natural resource system is that the success of
individual fishponds is dependent on the conditions of the immediate natural
environment (for clean, disease-free water with good oxygen content) as well as on
the coastal environment at a higher spatial scale. At the level of the region,

62
 Paper 2/5

aquaculture is dependent on the creeks, rivers and lakes for supplying small shells to
feed to the prawns. At the national level, the fishponds are reliant on the availability
of broodstock for prawn and milkfish for hatching, both of which have the coastal
environment as an important habitat in their life-cycle.

It follows that there are negative feedbacks to the fishponds that result from either
over-intensification on individual farms or from too many fishponds being
constructed. These may be immediate and dramatic such as when over-crowding of
ponds in an area results in wide scale mortality from oxygen depletion or disease
(Holmer et al., 2003). That explains why the concept of carrying capacity is seen as a
central component to planning for sustainable aquaculture (World Bank, 2002) but it
is a concept that is very difficult to operationalise owing to large data requirements, as
explained below.

Carrying capacity represents the maximum number of fishponds that could be

supported by an ecosystem, in terms of the environmental support services provided
by that ecosystem and the use of these services at any one time by the fishponds.
When the carrying capacity is exceeded, there is an emergent environmental event
such as a phytoplankton bloom (as occurred in Bolinao in February 2002 that led to a
fish kill in over 500 fish cages in municipal waters - Siar, 2002) or a disease outbreak
(as occurred frequently in the province of Negros in the mid-1990s, leading to the
abandonment of numerous intensive prawn farms - Yap et al., 2003). The difficulty is
that carrying capacity is difficult to determine prior to an environmental event.
Predictive work to this end is currently being undertaken (e.g. TROPECA project
headed by John Hambrey).

More long-term feedbacks are also undeniably present owing to the pressure put on
the finite wild fishery on which aquaculture is dependent for inputs, by sometimes
demanding more than the environment can supply (Ahmed et al., 2001) and by
indirect damage to the stocks of these fisheries. Indirect damage to the fishery may
occur through: disruption to the life-cycle of wild prawn and milkfish through loss of
coastal habitat – a habitat that is used for spawning and juvenile stages of both species
(Bagarinao, 1994; Primavera, 1998); harvesting of wild fry for grow-out that reduces
wild adult / juvenile populations as well as resulting in extremely high levels of by-
catch (Ahmed et al., 2001; Frankenberger, 2002); genetic interactions between wild
and cultured stocks of the same or similar species, that may result in a loss of
evolutionary fitness in the wild population through loss of adaptive traits (Xu, et al.,
2001); and through the introduction of non-native pathogens or escapee populations
(Bartley et al., 2000). As can be seen from this list there are a significant number of
endogenous environmental problems with coastal aquaculture.

4. Issues of Scale
The multiple problems facing the aquaculture industry are manifest at different spatial
scales: farm-level, regional, and national. The current study is both modest and
ambitious. It is modest as it deals mostly with the farm-level effects of aquaculture, in
recognition that the task of addressing multiple scales in analysis often entails
complexity beyond our means (see Clayton and Radcliffe, 1996 and Giampietro, 2004
for examples of first attempts at multi-scale modelling for sustainability). However,
this study remains ambitious in scope in that we attempt a degree of holism at the
farm level by attempting to measure indicators of economic, ecological and social

63
 Paper 2/5

performance. The interactions between the farms and the economy, society and
environment are outlined in figure 3. Fishponds exert pressure on the environment and
cause undesirable social problems (marked with „-„ signs) but have positive economic
benefits and some social benefits (marked with „+‟ signs). The fishponds also rely on
the economy, community and ecosystems for inputs.

Changes that are marked as exogenous in this scheme are not accounted for, and may
change the opportunity costs and trade-offs in ways that are not taken into account in
our model. However, the positive and negative interactions between economy,
environment, community and the fishponds will now be analysed using indicators.

64
 Paper 2/5

Figure 3 – Positive and negative relationships between fishponds and the regional
economy, environment and community
Exogenous
Changes in factor
prices
Changes in supply
/ demand

Economy

+ Profit + Capital
+ Fish, Shrimp, Crab
+ Manufactured inputs
to Market

+ Labour - Eutrophication,
Fishponds
- Uprising, Violence, - Loss of habitat /
Theft of stock mangrove functions
- Escapees, Diseases

Community Environment

+ Free Fish,
Livelihood + Feeds, Fry,
Opportunities Broodstock
- Displacement / - Diseases, Stresses
Removal of resource on stock (salinity,
access rights temp) Exogenous
Exogenous Industrial
Demographic pollution
change Natural
Livelihood disasters /
opportunities climatic events
in other sectors

65
 Paper 2/5

5. Sustainability - Operationalising a utopian concept

The central concept employed in this paper is that the sustainability of a farming
system is found in an appropriate balance between economic, ecological and social
objectives. In this respect, we can consider a farming system that maximises
economic performance, but with poor ecological and/or social performance, to be
unsustainable. Conversely, a farm that achieves a sufficient level of performance
across all three objectives can be thought of as sustainable.

In this regard, the concept of sustainability is fundamentally a rhetorical device, and

one that a number of people are unwilling to lose from their vocabulary because,
although it is used in a number of different ways, the term has a substantive and
important meaning. In moving from the rhetoric of sustainability, to actually
measuring its constituent elements (i.e. sustainability indicators), one has to pass
through theory. In the case of the sustainability of aquaculture systems in the
Philippines, the relevant literatures are those related to production economics,
agroecology, socio-economics, and sociology.

Underlying these literatures are a number of principle normative ideas or concepts.

That is, we can formulate maximisation or minimisation statements that, ceteris
paribus, would be uncontroversial. Of course, the central point underlying our thesis
is that the ceteris paribus clause in that statement doesn‟t usually apply – thus, there
are trade-offs to be made among objectives, defined in part by the technology.

It then follows that the choice of indicators becomes a matter of choosing specific,
measurable, units of account for these objectives. It is this process that we describe as
the operationalisation of sustainability. As Pannell and Glenn state, criteria for
choosing between indicators might be:

“High uncertainty about the level of the indicator to be monitored (desirable)

Low uncertainty about links between the indicator, management practices and
production (highly desirable)
The indicator can be measured reliably and accurately (desirable)
Low cost of monitoring the indicator over the necessary scale (desirable)”
Pannell and Glenn (2000, p.148)

Thus, the principles governing the operationalisation phase of developing

sustainability indicators is not one that is determined solely by theoretical or
methodological dogma. Rather, it becomes a real-world problem, constrained by data
availability, time and expertise, and with the intention of being outcome-oriented and
influential over policy. Hence, the approach is primarily a pragmatic one.

Following this review, there are a number of objectives for which indicators are
relevant, as discussed next.

5.1 Economic Objectives

For aquaculture to contribute to economic growth, it should give a sufficient return on
investment in order to attract economy-wide resources, notably capital, into the sector.
However, the variability of returns (i.e. the riskiness of the investment) is also

66
 Paper 2/5

important because agents are usually thought to be risk averse. Intuitively, variability
of returns to investment in aquaculture allows for the possibility of bankruptcies,
abandonment of ponds, and, ultimately, capital flight from the sector. In this context,
it is reasonable to posit that, ceteris paribus, an aquaculture system should, maximise
profit; and, with regards to the second idea, that, ceteris paribus, an aquaculture
system should minimise the risk associated with investment, or at least be resilient
in the face of risks.

The general principle of efficiency, (i.e. of minimising the throughput associated with
production, that is, producing on the production possibility frontier), also finds
validity in the sustainability concept. In particular, the principle of maximising
technical efficiency, such that a given level of output is achieved with the minimum
possible use of inputs3, is important. It prevents waste of resources and allows for a
higher level of profit than when systems are inefficient.

5.2 Socio-economic / Social Objectives

Given the aquatic resource-dependence of the economy in coastal areas of the
Philippines, aquaculture is associated with social equity issues. Reducing poverty is
always a central theme of government rhetoric and the link between aquaculture and
poverty, explored further in paper 4, should therefore be a relevant area of study. Put
simply, it is important to establish whether aquaculture benefits the poor.

In terms of livelihood benefits, in the ideal case poor people would be able to become
operators, and a suitable objective would be to maximise accessibility of the sector
to the poor. An alternative would be to maximise the employment possibilities
generated by the sector. Further, given the importance of fish in the diet of nearly all
Filipinos, an indirect benefit of aquaculture might be achieved by maximising food
security objectives relating to the farming system.

5.3 Ecological Objectives

Brackish-water aquaculture is generally sited in the ecologically sensitive inter-tidal
zone. The ponds appropriate aquatic natural resources (e.g. water, wild-harvested
feeds, wild-caught broodstock for hatching to fry, wild-caught fry) and the coastal
waters receive pollution in the form of effluent and undigested feeds from the ponds.
Ceteris paribus, aquaculture should therefore aim to minimise natural resource
appropriation and to minimise pollution of receiving waters.

The normative component to this statement cuts across time and spatial scales. There
are intra-industry issues relating to externalities of production, with costs for
environmental damage being passed from pond to pond; inter-industry issues relating
to aquaculture‟s effects on the capture fisheries sector - a social equity concern, given
the poverty of municipal fishermen; and long-term ecological issues regarding the loss
of natural capital and its impacts on future generations.

We thus consider a number of objectives as representing components of sustainable

aquaculture, as summarized in Table 3. Were these objectives to be studied in
isolation, without data constraints, we might consider a particular approach as
described in the „ideal-world‟ column of the table. In a comparative study where all
these objectives are measured for the same systems, and with limitations regarding
data collection and availability, we consider a practical approach that generates

67
 Paper 2/5

information on each objective under these constraints, as described in the „feasible‟

column of the table.

Table 3 – “Ideal-world” and “feasible” approaches to studying the objectives

associated with sustainable aquaculture
Objective Ideal-world Feasible
Maximise profit Full economic profit Gross margin

Minimise risk Variation in output over time Variation in output on similar

(panel data) farms (cross-section)

Maximise technical Stochastic multi-output distance Stochastic frontier with single

efficiency function output (i.e. revenue)

Maximise food Dietary protein balance Dietary protein balance

security

Maximise regional Social accounting matrix or full Employment multiplier analysis

employment input-output analysis

Minimise coastal Full ecological footprint with all Partial ecological footprint with
resource primary data for inputs and some primary and some
appropriation outputs secondary data for inputs only

Minimise Full nutrient balance Partial nutrient balance

eutrophication
potential

5.4 Sustainability Indicators

In light of the previous discussion, there follows a report of the ten selected
sustainability indicators for brackish-water pond aquaculture systems in the
Philippines. Each indicator is treated identically to cover:

The rationale behind its choice

The theory behind its definition and measurement
The methods used to measure it
The empirical results

The objects of the analysis are the five farm types that were described in the typology
in paper 1. A diagram summarising this introduction is given in figure 4.

68
 Paper 2/5

Figure 4 – Summary of the framework for sustainability indicators

RHETORIC Maximise Economic Performance Maximise Ecological Performance Maximise Social Performance

THEORY Max Profitability Min Resource Appropriation Max Employment

Min Risk / Max Resilience Min Pollution Max Food Security
Max Efficiency

OPERATIONALISATION

Max Min Max Min Min Min Max Max

Mean (SE of Technical Ecological Nitrogen Pesticide Employment Net Protein
Gross Mean Eff. Footprint balance Applic. Production
Margin GM /
Mean)
Pesos % Ha Kg N Kg/ha/yr Person days/ Kg/ha/yr
/ha/yr Ratio resources/h /ha/yr ha/yr
a farm

69
 Paper 2/5

6. Economic Indicators
In this section, we examine how the five farm types (extensive polyculture; semi-
intensive prawn polyculture; labour-intensive low-input systems; large, extensive
milkfish-oriented farms; and semi-intensive milkfish monoculture) differ according to
three main measures of economic performance: profit; risk; and technical efficiency.
Explanations for the results and their implications are then explored.

6.1 Profit
There are a number of definitions and measures of profit. Full economic profit would
take into account all production factors, including land and family labour, while gross
margin considers only the variable costs of production. Calculating full economic
profit in our case would require much more comprehensive data than would be
available from caretakers. Thus, either the sample would be restricted to those
operators that live on the farm and are available for interview, or the sampling process
would be significantly lengthened to incorporate visits to the homes or other
businesses of absentee operators, solutions that were both deemed unsatisfactory.
Besides, it is believed that gross margins best capture differences in production
practices across farming systems. Formally, they are calculated using the following
formula:

GM = (TR – CFeed – CFert – CSeed - CVarLab)/FS (1)

where GM is gross margin in PhP per hectare per year, TR is total revenue from all
sales at mean prices for one year, CFeed is the total cost of feeds in one year, CFert is
the cost of fertilisers for one year, CSeed is the cost of fry and fingerlings for one year,
CVarLab is the cost of seasonal or variable labour and FS is the farm size, expressed in
hectares. We now review the components of the gross margin for each of the five
farming systems under study.

6.1.1 Cost profiles

The absolute level of variable costs per unit area, shown in the final column of table 4
as well as in figure 5, vary greatly across farm types. As expected, the semi-intensive
systems (types 2 and 5) have relatively higher variable costs than the other farm types.
Looking at the composition of variable cost, it is also evident that there is a great deal
of heterogeneity in the relative importance of inputs across farm types. However,
focusing on the average values across all farm types (last row of table 4), it is evident
that seeds represent a very important cost item, while feeds do not, which is indicative
of the relatively low degree on intensity of the farming systems under study. Seeds
account for a particularly large share of variable cost item in the case of both kinds of
prawn production (types 1 and 2), which can be explained as follows. The low
individual unit cost of fry (roughly PhP 0.1 per piece) is as a result of the fry often
having been rejected by capital-intensive operators in other parts of the Philippines, or
from the fry not having been tested for diseases at all. Given the informational
problems in the transaction, fry prices are low, but survival is also extremely low.
This implies that the effective cost of fry is high because operators, expecting high
mortalities, stock at high densities to compensate.

The importance of feeds in variable costs varies greatly across farming systems, from
a minimum of 4.4% for the low-input system (type 3) to almost a third for semi-

70
 Paper 2/5

intensive milkfish monoculture (type 4). Table 4 also reveals that aquaculture
production is relatively labour intensive as hired workers account, across all farm
types, for more than a third of total variable cost. Finally, fertiliser accounts for a
small share of variable cost in all but the most milkfish-oriented systems (types 4 and
5), which make heavy use of imported inorganic fertilizers.

Table 4 – Total variable cost and its composition (by farm type)
TOTAL
feed
fertiliser seed share hired labour Variable costs
Farm type share
share (%) (%) share (%) (PhP/ha/yr)
(%)
1 - Extensive 11.2 6.7 49.2 32.8 33440
polyculture

2 - Semi- 27.0 2.4 54.3 16.3 87136

intensive Prawn-
oriented
polyculture

3 - Low-input, 4.4 12.3 27.9 55.3 16103

labour intensive

4 - Large, 14.5 26.4 25.8 33.3 25055

milkfish oriented

5 - Semi- 30.0 17.6 25.8 26.6 51536

intensive
Milkfish
monoculture

All types 14.0 11.1 38.7 36.2 35567

71
 Paper 2/5

Figure 5 – Mean variable costs by farm type (PhP/ha/yr)

140000

120000

100000

80000

60000

40000

20000

-20000
N = 53 11 33 11 19

1 - Ext. Polyculture 3 - Labour-intensive 5 - Milkfish mono

2 - Semi-int prawn 4 - Large farms

Farm type

6.1.2 Revenue shares

Revenue from aquaculture was calculated by using farm-level production and price
data collected through a survey. During the survey, however, caretakers and operators
could usually not state the “average” price achieved for each species in the last
harvest, but were much more confident about the “highest” price and the “lowest”
price achieved in that year. Therefore, we simply took the average of these two values
when computing total revenue.

Table 5 reveals that the main species in terms of revenue are prawns for types 1
(extensive polyculture) and 2 (semi-intensive prawn-oriented polyculture), and
milkfish for the others. Not surprisingly, overall revenue per unit area reflects the
species orientation of each farming system as well as its degree of intensification.
Semi-intensive prawn polyculture farms (type 2) have the highest (and most variable)
revenue per unit area, with twice the revenue of extensive prawn farms (type 1) and
semi-intensive milkfish monoculture (type 5). Low-input labour intensive farms (type
3) and large farms (type 4) have the lowest revenue per unit area. Analysis of the
degree of revenue diversification across species is examined further in a later section,
in the context of price risks.

72
 Paper 2/5

Table 5 – Composition of total revenue by farm type

TOTAL
Farm type Prawns Milkfish Crabs Tilapia Revenue
(PhP/ha/yr)

1 – Extensive 53.7 25.1 15.4 5.8 86854

polyculture

2 – Semi- 57.0 14.6 22.6 5.8 168091

intensive
prawn-oriented
polyculture
3 – Low-input, 28.6 60.9 8.3 2.2 51430
labour
intensive
4 – Large 20.0 79.3 0.0 0.7 55053
farms

5 – Semi- 0.0 90.0 3.7 6.4 90889

intensive
milkfish
monoculture
Mean 36.5 47.8 11.1 4.5 82535

Figure 6 – Mean revenue by farm type (PhP/ha/yr)

300000

200000

100000

-100000
N = 53 11 33 11 19

1 - Ext. Polyculture 3 - Labour-intensive 5 - Milkfish monoc

2 - Semi-int prawn 4 - Large farms

Farm type

73
 Paper 2/5

6.1.3 Mean gross margin by farm type

The gross margin, as defined previously, is a crude measure of profit, excluding
payments made to credit, land rents, and the opportunity cost of family labour.
However, it can be directly related to the factors that define farm “types” because of
its link to inputs and associated production techniques. In economic terms, gross
margins can be interpreted as the returns to land, family labour and other fixed farm
assets. The distribution of gross margins for the sample is given in Figure 7 below,
while Figure 8 presents a plot of the means by farm type. Clearly, farming systems
integrating prawn production (types 1 and 2) are noticeably more profitable than the
other farm types. Type 5 farms achieve intermediate gross margins, while types 3 and
4 stand out by their low levels of profitability. Further, Figure 8 indicates that the 95
percent confidence intervals of the mean are spread widely, most notably for type 2
farms. This intra-group variability will be discussed in the following section in
relation to the concept of risk.

Figure 7 – Distribution of mean gross margin (PhP/ha/yr)

50

40

30

20

10
Std. Dev = 69104.36
Mean = 43822
0 N = 127.00
10

18

26

30
-6

-2

14

22
20

60
00

00

00

00

00

00
00

00
00

00
0

00

00

00

00
00

00
0

0
0

Mean Gross Margin (PhP / ha / yr)

74
 Paper 2/5

Figure 8 – Mean gross margin by farm type (PhP/ ha/yr)

200000

100000

-100000
N = 53 11 33 11 19

1 - Generalists 3 - Labour intensive 5 - Milkfish monoc

2 - Prawn oriented 4 - Large farms

Farm type

6.2 Risk
6.2.1 Theory
Variation in output from a single farm over time is usually taken as a measure of the
production risk (Just and Pope, 2001). However, time-series data are not available in
our case so that a related measure of risk cannot be derived. There is some evidence
from Figure 8 that the level of within-type variability (shown in the graph by the 95%
confidence intervals of the mean) differs across farm types. We use this observation to
define the ratio of the standard error of the mean gross margin to the mean gross
margin itself as an indicator of overall risk for each farming system type. The
underlying assumption is that variability in one growing year between farms of the
same type gives us similar information to measuring trends over time on a single
farm. It is therefore only a snapshot, and, unfortunately, the variability in gross margin
may be attributable in part to farm-level characteristics that are not used in the
typology exercise (e.g. farmer‟s experience), rather than the underlying risk profile of
the techniques common to each farm type. The results for these calculations are given
in the following section, followed by a decomposition into production risk and price
risk.

6.2.2 Overall Risk – Results

Table 6 and Figure 8 (above) show that semi-intensive prawn-oriented polyculture
(type 2) returns the highest mean gross margin, but with the highest level of risk. This
stands in contrast to the situation for extensive polyculture (type 1), which returns the
second highest mean gross margin but with the lowest level of risk. Hence, in this
case, there exists a clear trade-off between overall profitability and riskiness of
production systems, although that conclusion cannot be generalised to all five farming

75
 Paper 2/5

systems. Indeed, in terms of these two dimensions (profit and risk), farm types 3, 4
and 5 are „Pareto dominated‟ by farm type 1. We now attempt to explain overall profit
variability by dividing it into several sub-components.

Table 6 – Profit and risk indicators by farm type

Average GM
(pesos / ha /yr) Std. Error of SE Mean / Mean
Farm type N
PROFIT Mean RISK INDICATOR
INDICATOR
1 51220 53 8647 0.169
2 59065 12 32235 0.546
3 29769 35 10214 0.343
4 27845 11 13943 0.501
5 43252 19 19826 0.458
Average 43026 130

6.2.3 Production Risk – Mortality by species and farm type

Comparing mortality rates in prawn production across farm types shows a surprising
pattern (Table 7 and Figure 9). The small, labour-intensive, low-input system (farm
type 3) displays a significantly lower rate of prawn mortality than the other farming
systems that also stock prawns (types 1 and 2). This would suggest that a high use of
labour per unit area can result in lower mortality rates, although the large standard
deviations in Table 7 make this interpretation somewhat tentative. However, it is also
intuitively appealing that a large input of supervisory labour should improve water
and disease management.

That said, interpreting Table 7 in terms of the relative production riskiness of farming
systems is not straightforward because the absolute levels of mortality rates might not
be as informative as their variability from year to year. In this light, it may be that
prawn production in extensive polyculture systems (type 1) is not considered risky in
spite of extremely large mortality rates because operators simply expect one or two
per cent survival of the prawn fry. If that is the case, then the low-input farming
system can be characterised as being high risk, as indicated by the large standard
deviation of the mortality rate (26.8). This could be explained by the fact that those
farms lack access to risk-reducing inputs (apart from labour). The next section
considers another source of risk associated with unanticipated variations in output
prices.

76
 Paper 2/5

Table 7 - Mortality rates by species and farm type

Farm Prawn (% Milkfish (% Tilapia (% Crab (%
type mortality) mortality) mortality) mortality)
1 Mean 95.3 39.3 75.5 54.1
N 45 45 18 28
Std. 6.1 35.0 24.0 20.3
Deviation
2 Mean 96.3 43.0 90.6 55.7
N 12 9 2 8
Std. 4.9 34.2 9.0 17.5
Deviation
3 Mean 80.6 39.9 95.0 61.6
N 20 31 1 8
Std. 26.8 30.9 18.8
Deviation
4 Mean 96.4 42.0 48.8
N 4 10 1
Std. 2.2 27.8
Deviation
5 Mean 40.8 49.5 20.0
N 18 1 1
Std. 32.6
Deviation
Total Mean 91.8 40.2 75.4 55.0
N 81 113 23 45
Std. 15.4 32.3 23.4 19.9
Deviation

Figure 9 – Mortality rates of prawns by farm type (%)

110

100

90

80

70

60
N = 45 12 20 4

1 2 3 4

Farm type

77
 Paper 2/5

6.2.4 Price Risk

We define a measure of price risk for a particular farm type using the following
formula:

Price risk = (TRMHP – TRMLP) / TRMLP (2)

where TR is total revenue, and the subscripts MHP and MLP refer to Mean High
Price and Mean Low Price respectively. MHP (MLP) is the mean of the highest
(lowest) prices achieved in the year of study across all farms of a given type. Prawns
and crabs, while being the high value species, are also the most variable in price. This
is explained by volatility in the export market for prawns, as well as seasonality in the
domestic demand for crabs, a luxury food consumed primarily during festivals and
holidays. One could argue that as long as these price variations are anticipated, they
do not generate risk. However, while the seasonality in domestic demand can be
predicted to some extent (prices rise at Easter and Christmas), the random component
to export markets for prawns is more difficult to predict. In order to pursue the
analysis of price risk further, the price risk index defined above is calculated for each
farm type and reported in Table 8 and Figure 10.

Figure 10 – Price risk index by farm type

2.5

2.0

1.5

1.0
Price risk index

.5

0.0
N = 53 11 33 11 19

1 2 3 4 5

Farm type

We observe that farm types 1 and 2, which have a strong orientation towards prawn
production, face the highest level of price risk. Rather surprisingly, the pattern across
farm types in Figure 10 shares a great deal of similarity with that of the index of
revenue diversity, detailed in paper 1 (section 5.5.1), and also reported in Table 8. In
particular, we note that farm types 1 and 2 have diverse revenue sources, whereas
types 3, 4 and 5 are more reliant on a single crop - milkfish. Further, Table 9
establishes that the differences between farms 1 and 2 in comparison with types 3, 4

78
 Paper 2/5

and 5 are statistically significant. Hence, the mean score for equitability for farm types
1 and 2 are not significantly different from each other but are statistically different
from the other farm types at the ten percent level. The mean of the price risk index
follows a roughly similar pattern, although the high value of the standard error for the
mean for farm type 2 results in low statistical power of the test.

The finding that relatively more specialised farming systems are exposed to less risk
may seem counter-intuitive because diversification across enterprises is usually
presented as a mechanism for reducing risk (Zenger and Schurle, 1981; Bhende and
Venkataram, 1994). However, in the case where one enterprise is significantly more
risky than others, the degree of risk simply reflects the importance of that enterprise in
the production plan of the farm. Thus, operators producing prawns in combination
will always face greater price risks than operators producing only milkfish. This
interpretation is confirmed by the absence of prawn monoculture in the study area,
probably because prawn production is simply too risky.

Table 8 – Revenue diversity and price risk by farm type

Equitability (H’) - REVENUE

PRICE RISK INDICATOR
DIVERSITY INDICATOR

1 – Generalists 0.397 1.263

2 – Prawn-oriented 0.449 1.173

3 – Low-input, labour intensive 0.218 0.549

4 – Large farms 0.138 0.499

5 – Milkfish monoculture 0.043 0.478

Mean 0.280 0.886

Table 9 – Comparison of the mean level of equitability and price risk index using
ANOVA and Scheffe‟s post-hoc test.
Dependent Variable (I) Farm type (J) Farm Mean Difference Std. Error Sig.
type (I-J)

Equitability 1 2 -0.052 0.073 0.974

3 0.179 0.050 0.015
4 0.259 0.076 0.024
5 0.355 0.061 0.000
2 1 0.052 0.073 0.974
3 0.231 0.077 0.066
4 0.311 0.096 0.037
5 0.406 0.085 0.000
Price risk 1 2 0.090 0.259 0.998
3 0.714 0.173 0.003
4 0.765 0.259 0.075
5 0.785 0.209 0.009
2 1 -0.090 0.259 0.998
3 0.624 0.272 0.268
4 0.675 0.333 0.398
5 0.695 0.296 0.246

79
Paper 2/5

80
 Paper 2/5

6.3 Technical efficiency and its determinants

6.3.1 Rationale
As Kumbhakar and Lovell (2003) note, not all producers are successful in solving the
optimisation problems that underlie microeconomic theory. In particular, all producers
do not succeed in utilising the minimum inputs required to produce the output levels
that they choose, given the technology at their disposal, and, as a result, they exhibit
some degree of technical inefficiency. Further even technically efficient producers
may not succeed in allocating inputs in a cost-effective manner, in which case they are
allocatively inefficient. Finally, allocative inefficiency can also occur if producers
choose output combinations that do not maximise revenue. These various sources of
inefficiency represent formidable informational and managerial challenges for
fishpond operators, and, in this regard, assuming profit maximisation might seem
somewhat heroic. Here, we limit ourselves to the analysis of technical efficiency

81
 Paper 2/5

because it requires the weakest behavioural assumption in a context where it is likely

that producers pursue complex livelihoods strategies.

More formally, technical efficiency (TE) is defined as the ability of a firm to produce
on the frontier isoquant (Farrell, 1957), as is the case for points B and C in Figure 11.
Perhaps more intuitively, with an output orientation, technical efficiency corresponds
to the ability to achieve maximum possible output from a given set of inputs; or, with
an input orientation, to the ability to use least inputs to achieve a given level of output.
Farrell showed how overall economic efficiency is the product of technical efficiency
and allocative efficiency, where input-orientated allocative efficiency is defined as
the ability to produce at a given level of output using the cost-minimising input ratios
(Bravo-Ureta and Pinheiro, 1997). Technical inefficiency therefore defines the extent
to which producers fail to produce on the frontier isoquant whereas allocative
inefficiency relates to the extent to which producers do not choose the combination of
inputs that would minimise their costs, or the combination of outputs that would
maximise their revenue (in an output-based approach). Although efficiency analysis
does not make any assumption regarding substitutability among inputs, it becomes
rather meaningless if producers are unable to modify input proportions, in which case
allocative efficiency is always achieved.

82
 Paper 2/5

Figure 11 – A graphical illustration of different efficiency concepts. Technical

efficiency is defined by the ratio OB/OA; allocative efficiency is defined by OD/OB;
and overall economic efficiency is the product of these two ratios OD/OA. Adapted
from Sharma and Leung (2003).

TE = OB/OA
Input 2 (X2) AE = OD/OB
EE = TE x AE = OD/OA

B
W Isocost line

Isoquant
C

W‟
O

Alternatively, technical efficiency can be measured by the distance to a best

production frontier rather than by distance to an isoquant as illustrated in figure 12
below.

83
 Paper 2/5

Figure 12 – Production frontier approach to technical efficiency estimation (after

Kumbhakar and Lovell, 2003, p. 47).

φAyA

yA

x
A A A
θ x x

The production frontier f(x) illustrates the technically efficient relationship between
an input (x) and a single output (y). A producer using x A to produce yA is technically
inefficient as it operates beneath the frontier f(x). In an input-oriented approach,
TE(yA,xA) measures the maximum contraction of xA that enables continued production
of a level of output (i.e. yA), where TE(yA,xA) = θA < 1. In an output oriented
approach, TE(xA,yA) measures the reciprocal of the maximum expansion of yA that is
possible using xA inputs, and TE(xA,yA)=(φA)-1<1 (Kumbhakar and Lovell, 2003).

There are two families of approaches to the estimation of technical efficiency –

parametric and non-parametric. In parametric approaches, a functional form for the
technology is assumed (e.g. Cobb-Douglas). However, it is possible to introduce a
stochastic term such that deviations below the frontier can be decomposed into
random errors and technical inefficiency. In non-parametric approaches, no functional
form for the technology is assumed but all deviations are assumed to be attributable to
technical inefficiency. This latter assumption is particularly problematic for
agriculture and aquaculture owing to the numerous weather, environmental and
ecological effects that can influence production, and thus parametric approaches, such
as the stochastic frontier production function method discussed in the next section,
tend to be favoured over non-parametric approaches (such as data envelopment
analysis – DEA).

Before developing the theory underlying efficiency measurement, it is worth

emphasising the central importance of technical efficiency in a development strategy
for aquaculture, as argued by Bravo-Ureta and Pinheiro (1997, p. 49):

84
 Paper 2/5

“Many researchers and policymakers have focused their attention

on the impact that the adoption of new technologies can have on
increasing farm productivity and income (Hayami and Ruttan
1985; Kuznets 1966; Schultz 1964; Seligson 1982). However….
major technological gains stemming from the green revolution
seem to have been largely exhausted across the developing world.
This suggests that attention to productivity gains arising from a
more efficient use of existing technology is justified (Bravo-Ureta
and Pinheiro 1993; Squires and Tabor 1991).”

In the case of aquaculture, new technologies, such as fry hatcheries (in particular for
prawns), and the development of commercial feed pellets with optimal combinations
of nutrients for growing fish, have shifted the production frontier up and to the left.
However, these changes have occurred at a very fast rate, as they did with the Green
Revolution. As a result, institutional innovations such as environmental regulation,
land reform, and government extension provision, which would have allowed
aquaculture to adjust to changes in the technical arena more comfortably, did not
occur fast enough. Therefore, a number of questions remain regarding how well the
sector has adapted to technical innovation. Diagnosing the extent of technical
inefficiencies in the sector and examining their determinants is a useful tool in
addressing some of these questions.

6.3.2 Stochastic production frontier models – General theory

The measurement of technical efficiency across all farms in the sample is based on
their distance to a single stochastic production frontier that characterises “current best
practice” in the sample. This kind of parametric methodology is now dominant in the
production economics literature, owing to the problems of noisy data that make DEA
less attractive.

In the original framework developed by Farrell (1957) and extended by Kopp (1981),
we first consider a deterministic production frontier given by:
Yi = f(Xi ; β) - ui (2)

where Yi is the maximum achievable output of the ith farm, Xi is the input vector that it
uses, and β is a vector of unknown parameters. This deterministic approach suffers
from the limiting assumption that all deviations from the frontier are attributed to
technical inefficiency. However, efficiency measures obtained from these models are
affected by statistical noise (Schmidt, 1985-86). As a way of addressing this problem,
a stochastic production frontier can be rewritten as:
Yi = f(Xi ; β) + ε (3)

where ε is a “composed” error term (Aigner et al., 1977) which can be written as:
εi = vi – ui (4)

The term vi is a two-sided (-∞ < v < ∞) normally distributed random error that, it is
assumed, captures exogenous factors outside of the control of the producer. These
might include weather shocks, natural disasters and good luck. It should be noted in
passing that the suitability of this specification for aquaculture (of prawns in

85
 Paper 2/5

particular) has been questioned (Brennan et al., 2000), given the number of downward
shocks that can be experienced (e.g. disease, flooding, water pollution) compared to
the possibility for beneficial stochastic events (i.e. good luck).

The term ui is a one-sided (u ≥ 0) random component that captures the technical

inefficiency of the farmer. This is essentially the distance (i.e. the shortfall in output)
of a given farm below the stochastic frontier given by f(Xi ; β) + vi. The distribution of
u can take several forms, including the half normal, truncated normal, exponential and
gamma distributions. In what follows, we use the model developed by Battese and
Coelli (1995), which assumes that the inefficiency term is a normal random variable
N(mi ; ζu2) truncated at zero, where mi = zi and zi is a vector of firm-specific variables
which may influence the firms' efficiency

The most commonly used method for estimating stochastic production frontiers is
maximum likelihood (Battese and Coelli, 1995; Coelli et al. 1998), which can easily
be carried out using the program FRONTIER 4.1 (Coelli, 1996). The software
operates the numerical maximisation of the likelihood function that is expressed in
terms of two variance parameters σ2 = σu2 + σv2 and γ = σu2 / (σu2 + σv2) (Battese and
Corra, 1977). It is clear that parameter γ lies between zero and unit, with zero values
indicating that deviations from the frontier are due entirely to noise, while values of
one indicate that all deviations are due to technical inefficiencies. FRONTIER also
predicts individual farm technical efficiencies by using the expression of the
expected value of uj conditional on the estimated value of εj, as first established by
Jondrow et al. (1982).

6.3.3 Stochastic production frontier models –

Previous applications to aquaculture
Applications of stochastic frontier production models were late to emerge in the
aquaculture literature in comparison to agricultural economics. This initial lack of
frontier studies in aquaculture can be attributed to the inherent bio-economic
complexity in these systems, and the concomitant problems of obtaining appropriate
production data (Sharma and Leung, 1998). However, after this slow start, there is
currently a burgeoning literature on efficiency measurement in aquaculture which is
summarized by Sharma and Leung (2003). The problems of bio-economic complexity
remain, but it would seem that authors have perhaps become braver in asserting the
assumptions necessary to carry out the analysis.

Sharma and Leung (1998) and Sharma (1999) present applications to carp polyculture
systems in Nepal and Pakistan respectively. Owing to the similarities in value of the
various carp species, these papers use fish production per unit area (kg / carps / ha)
with all factor inputs also expressed per unit area. The overall level of technical
efficiency in Nepal was 77 percent, with intensive farms more efficient that extensive
farms. Lower estimates were derived for Pakistan, but the pattern of semi-
intensive/intensive systems having higher technical efficiency than extensive systems
held (67 percent for semi-intensive/intensive farms, versus 56 percent for extensive
farms).

In terms of the technical efficiency effects investigated in the model, both papers
specify three simple indices related to fish management, water management and feed
management. These indices take a value between zero (no recommended management

86
 Paper 2/5

practices used in that area) and unity (all of the recommended management practices
are used), with decimal values representing intermediate cases where some practices
are employed and others are not. These indices, while crude, establish a clear link
between the model and actual production practices and, hence, can help generate
useful insights for, among others, extensionists..

With data from Nepal (Sharma and Leung, 1998), the t-ratios for the slope
coefficients of these indices were insignificant individually under maximum
likelihood estimation. However, under generalised likelihood ratio tests, both
production intensity and the grouped indices for these regular management practices
had a significant and positive effect on technical efficiency. With Pakistani data
(Sharma, 1999), the water and feed management indices had significant and positive
effects on technical efficiency for both semi-intensive/intensive and extensive
production systems. Neither study finds a significant relationship between farm size
and technical efficiency. Altogether, those two studies demonstrate that fish farmers
often do not use the available technology in a technically optimal way, and that
inefficiencies relate to a large extent from the failure of adopting recommended
production practices.

In another context, Dey et al. (2000) analysed technical efficiency in freshwater

tilapia aquaculture in the Philippines and reached very different conclusions. They
found that experience and farm size both had a significant, positive effect on technical
efficiency and that the overall level of technical efficiency in their sample was high
(mean TE = 83%). On the basis of these findings, the authors recommend the
introduction of new technologies to raise productivity, owing to the nearly fully
technically efficient exploitation of the current technology. It should be noted that
freshwater tilapia pond farming is a relatively new activity and tilapia is an exotic
species to the Philippines. Culture techniques do not appear to differ from farm to
farm as much as in brackish water ponds. There is also virtually no polyculture in
fresh-water systems in the Philippines.

With regards to the species of interest in this study, two recent papers are particularly
relevant. Irz and McKenzie (2003) compare the technical efficiency of intensive
freshwater tilapia production systems with extensive brackish-water polyculture in
Pampanga, Philippines. They find that technical efficiency in the brackish-water
systems is low (53% on average), whereas the technical efficiency in intensive tilapia
production is high (83% on average). From these results, the authors conclude that
extension services for brackish-water systems would be a cost-effective way of
increasing productivity, but, in accord with Dey et al. (2000), that technological
change would be required to raise productivity in freshwater systems. These findings
may provide helpful insights to the government agencies in the Philippines in their
effort to coordinate research and development activities for aquaculture.

Chiang et al. (2004) analyse technical efficiency in milkfish production in Taiwan.

Their study presents methodological problems associated with polyculture systems
and these authors prefer to define output only as milkfish yield, leaving aside other
productions, rather than aggregating across species. This is surely an unsatisfactory
assumption as secondary species can represent an important percentage of the total
yield – up to 40 per cent of total revenue in the case of farms in their sample.

87
 Paper 2/5

The literature on technical efficiency in aquaculture has yet to see its potential
realised, in terms of use in informing policy and improving management. The models
presented in the literature thus far have mostly not laid out explicit mechanisms
through which a (in some cases) low level of technical efficiency can be related to
specific failures at policy or farm-management level. In addition, the econometrics
behind the method makes the results difficult to communicate to policy-makers.

6.3.4 Model specification

We choose the translog functional form for the parametric represention of the
production technology. Output is given by total annual revenue from aquaculture
because alternatives, such as total weight of production or total revenue in average
prices, are inappropriate. The difference in the nature of the four possible outputs
makes summing them by weight meaningless, and substituting individual prices for
average prices would not reflect any premium that the operator would achieve from
producing higher quality products (e.g. prawns of an individual size large enough to
be exported, larger milkfish for fillets etc).

Five inputs are used in the analysis: X1 is land area of the farm in hectares, X2 is
labour in person-days per year, X3 is feeds cost per year (in PhP), X4 is fertilisers cost
per year (in PhP), and X5 is fry and fingerlings cost per year (in PhP). Data is taken
for one year in a cross-section of 127 farms. To ease interpretation of the estimated
parameters, each logged variable is mean-centred prior to estimation to ensure that the
output elasticities at the sample mean are simply equal to the first-order coefficients.

We use the model of Battese and Coelli (1995) because it allows for the simultaneous
estimation of technical efficiency levels and their determinants. Further, the
appropriateness of alternative specifications can easily be investigated in this
framework by carrying out a likelihood ratio test. The test compares a null hypothesis
for restrictions on parameters (H0) against an unrestricted alternative (H1), based on
the following ρ statistic:
ρ = -2{L(H0) – L(H1)} (5)

where L(.) denotes the value of the log-likelihood function. The test statistic is then
compared to a critical value obtained from a chi-squared distribution. This procedure
makes it easy to test, for instance, the hypothesized influence of a number of farm-
specific variables on technical (in)efficiency, or whether the technology exhibits
increasing, constant, or decreasing returns to scale.

We initially consider three model specifications. Model 1 does not allow for technical
inefficiencies by imposing the parametric restriction γ = 0. In this context, all
deviations from the frontier are due to statistical noise and the frontier reduces to the
average response function which is estimated by ordinary least squares (OLS). Model
2 allows for inefficiencies, but assumes that those are not explained by farm-specific
variables (or technical efficiency effects). This second specification is referred to as
the error components model. Model 3 adopts the most general specification, which
allows for inefficiencies and introduces 15 technical efficiency effects (δ 1-15) for
which data are available. These relate to technical, managerial and socio-economic
characteristics of the farm and are all dummy variables, taking the value one when a
particular technique is used or when a particular management structure is in force, and

88
 Paper 2/5

zero otherwise. In addition, there are dummy variables for each of the farm types from
the typology4 and a regional dummy variable to allow for differences in natural
resource endowment. Finally, a dummy variable is introduced that takes the value one
when the operator reported an exogenous shock to their farming system. This follows
from the critique of Brennan (2000) regarding the one-sided, downwards nature of the
shocks experienced by prawn aquaculture, in conflict with the assumed normal
distribution of the v term in the error that is introduced with the aim of capturing
stochasticity. A preferred specification is now selected based on the results of likely-
hood ratio tests.

Table 10 – Values of the log-likelihood function of alternative specifications

MODEL 1 – No MODEL 3 –
MODEL 2 – Error
inefficiencies (ols Technical
components
regression) efficiency effects
Log-likelihood function -157.43 -139.99 -126.40

Model 1 (H0) vs Model 3 (H1). The calculated value of the test statistic is ρ = 62.06,
which is compared to a critical value of 40.11, itself derived from a mixed chi-squared
distribution5 at α=0.001 with 17 restrictions (i.e. γ=δ0=δ1= δ2 ... δ15=0). In this case,
the null hypothesis of absence of inefficiencies is strongly rejected, and the analysis
supports the view that many producers are not using the available technology in a
technically optimal manner.

Model 2 (H0) vs Model 3 (H1). The test statistic ρ = 27.18 is compared to a critical
value of 25.00 for the chi-squared distribution at α = 0.05 with 15 restrictions (i.e.
δ1=δ2…δ15=0). Here again, the null hypothesis is rejected, implying that the farm-
specific variables introduced in the model have some power in explaining
inefficiencies..

Finally, a similar procedure was carried out to determine the appropriate functional
form of the production function. More specifically, the trans-log specification was
tested again the Cobb-Douglas by setting the final 15 cross-product terms equal to
zero. This null hypothesis was again rejected. We therefore conclude from this series
of tests that the full translog stochastic frontier with technical efficiency effects is the
most appropriate specification of the model. The results derived from this preferred
specification are now analysed.

6.3.5 Estimation Results

The coefficients in the production function, reported in Table 11, are somewhat
disappointing with regards to expectations held ex-ante. Only feeds and fry return
statistically significant and positive output elasticities. The results indicate that fry is
the more important input, which is consistent with the extensive nature of the
production process – mean output is therefore most constrained by a lack of fry input
to the system. The output elasticities for land and fertilizers are positive, as would be
expected, but also statistically insignificant. The most puzzling result is for labour
input, which returns a negative output elasticity, although with a low degree of
statistical significance.

However, there are a number of important factors to consider before dismissing these
results as inappropriately representing the technology. Of course, we might expect
labour input for activities such as monitoring the stock and ensuring optimal feeding

89
 Paper 2/5

patterns to be productive. However, these types of labour use are responsible for only
a small percentage of the total labour demand in aquaculture. The majority of the
labour input is used for maintenance of the dykes, harvesting the fish, and removing
weeds from the pond – all important activities that are, however, not directly related
to output. These activities are carried out with the aim of avoiding negative
consequences, or what is referred to in the agricultural literature as „damage control‟.
It is therefore perhaps more appropriate to consider the labour input used by these
activities as being similar in essence to pesticides – an input that is not normally
included in production functions as it is not directly productive. In fact, one could
argue that use of large quantities of labour in maintaining the dykes and removing
weeds is indicative of sub-optimal construction or positioning of the farm, which are
characteristics for which data are not available. This issue is taken up in the later
section entitled “Technical efficiency, stochasticity and the labour input problem”.

Table 11 – Maximum likelihood estimates of model 3. Only the first five terms of the
full trans-log specification are given here, for convenience.
Inputs Name Coefficient Std Error t-ratio
β0 Intercept -0.636 0.590 -1.078
β1 Land 0.123 0.135 0.914
β2 Labour -0.135 0.112 -1.213
β3 Feeds ** 0.086 0.025 3.409
β4 Ferts 0.018 0.017 1.075
β5 Fry ** 0.680 0.062 10.897
2
σ 2.753 0.993 2.774
ρ 0.941 0.030 31.665
Log-likelihood -126.40
function
Mean % Technical 62.8%
efficiency
Technical
Name Coefficient Std. Error t-ratio
Efficiency Effects
δ0 Intercept 0.195 0.965 0.202
δ1 Region -0.433 0.818 -0.529
δ2 Live on farm * -2.683 1.416 -1.895
δ3 Main Income -0.333 0.799 -0.417
δ4 Exog. Shock ** 2.749 1.289 2.132
δ5 Lagum-Lagum * -3.320 1.693 -1.961
δ6 Continuous -0.911 0.830 -1.098
δ7 Record Keeping -0.965 0.843 -1.145
δ8 Probiotics * -2.561 1.479 -1.732
δ9 Salinometer -1.102 0.878 -1.255
δ10 pH Test * -3.678 2.013 -1.827
δ11 Farm type 1 * -1.678 0.924 -1.815
δ12 Farm type 2 ** 2.893 1.311 2.206
δ13 Farm type 3 0.529 0.874 0.605
δ14 Farm type 4 ** -4.997 2.429 -2.057
δ15 Owns land -0.506 0.801 -0.632

Table 11 further establishes that farms in the sample exhibit high levels of
inefficiency, with a mean of 63%. This value indicates that, on average, brackish
water farms produce less than two thirds of the maximum achievable output, given the

90
 Paper 2/5

available technology and the recorded input uses. The result suggests that productivity
of brackish water aquaculture in the Philippines could therefore increase substantially
from improvements in farm management, which is consistent with the finding of Irz
and McKenzie (2003). Further, the model allows for the identification of possible
inefficiency effects (i.e, factors which tend to increase/decrease inefficiency). Any
positive value for a parameter in Table 11 indicates that the related variable
increases inefficiency. Only those determinants with a high level of statistical
significance are now discussed in detail.

The dummy variable for membership of farm type 2 (semi-intensive prawn

production) significantly increases technical inefficiency, as is confirmed by the mean
efficiency scores plotted by farm type in Figure 13. With regard to the other farm
types, the estimation results demonstrate that types 1 (polyculture) and type 4 (large
farms) are relatively more technically efficient than the others, while types 3 and 5
(the reference) are comparable in terms of technical efficiency. Hence, our typology
does capture differences across farms that have a significant influence on their
technical performances.

Beyond this general statement, however, it is difficult to provide a definite

explanation as to why a particular farming system might be more technically efficient
than another, but the results might suggest that intensification reduces technical
efficiency. This hypothesis is tested by a one-way ANOVA comparison of the means
between types 1 and 2 (systems with prawn accounting for more than 50 percent of
total revenue) and between types 4 and 5 (systems with milkfish accounting for more
than 50 percent of total revenue). These paired comparisons are justified by the fact
that farm types considered present large similarities apart from their degrees of
intensification. Hence, the stocking density of prawns and feeding rate for type 2
farms is significantly greater than for type 1 farms, while the species revenue shares
are similar; and the stocking density of milkfish and feeding rate for type 5 farms is
significantly greater than for type 4, but both farming systems are oriented towards
milkfish production.

Let us consider first the two polyculture systems, where type 1 farms can be
considered extensive and type 2 farms semi-intensive. Table 12 shows that type 1 has
a higher mean technical efficiency than type 2, and that the difference is significant at
the one percent level. Similarly, for milkfish systems, the extensive farms (type 4)
achieve, on average a level of efficiency which is sixteen percent greater than that of
the intensive farms, although the difference is not statistically significant. Hence, this
analysis provides support for the idea that intensification tends to reduces technical
efficiency, and particularly so in the case of prawn polyculture. This might be
explained by the fact that the strength of the polyculture system is that it enables
species to exploit different trophic niches in the pond, but this effect is likely to be
diminished with the increased densities associated with intensification.

91
 Paper 2/5

Table 12 - Scheffe‟s Post-hoc tests on a one-way analysis of variance comparing

mean technical efficiency between farm types 1 and 2, and between types 4 and 5.

(J) Farm Mean Difference Std.

(I) Farm type Significance
type (I-J) Error

1 2 .2569 .06854 .009

4 5 .1619 .07838 .376

The analysis of inefficiency effects pursues by noting in Table 11 that, as intuitively

expected, the dummy variable for negative exogenous shocks (related to parameter
δ4) significantly increases inefficiency. Finally, parameters δ2, δ5, δ8 and δ10 all take
statistically significant negative coefficients, indicating that the related variables have
a positive influence on efficiency. In particular, an operator living on his farm (δ2),
employing the traditional lagum-lagum system of production (δ5), using probiotics
(δ8) and testing the pH of the soil (δ10) achieves higher levels of technical efficiency
than otherwise.

These results, in part, support the traditional wisdom related to aquaculture. Hence, a
saying reported by a number of operators asserts that the best inputs to the farming
system are “the shadow of the operator across the pond and his footmarks on the
dyke”. In this regard, a producer living on his farm is likely to perform better than an
absentee operator, as is confirmed by the econometric results . In addition, the lagum-
lagum system of stock movements is a traditional one, dating back to a period when
the construction of concrete sluice gates was not possible, and the much-researched
modular (or continuous) systems of operation were therefore unavailable.

However, the use of probiotics, a relatively new technical innovation for

strengthening the resilience of the culture medium with respect to bacterial diseases
(Corre, Janeo et al., 1999), also improves technical efficiency. This conclusion, while
intuitive, should be treated with caution because of the small proportion (7%) of farms
that have adopted this innovation. Further, it would perhaps be more accurate to state
that those farms using probiotics are more efficient that those that do not, because the
use of probiotics may be indicative of unobserved heterogeneity among operators,
possibly explained by characteristics such as a higher education level. The same
cautious interpretation should be applied to the finding that pH testing of the soil
(δ10) increases technical efficiency. We now return to a detailed discussion of why
the labour elasticity of the estimated production function might be negative.

92
 Paper 2/5

Figure 13 – Mean and 95 per cent confidence intervals of technical efficiency by farm
type.
.9

.8
(model 3)

.7

.6

.5

.4

.3

.2
N = 53 11 33 11 19

1 - Generalists 3 - Low-input 5 - Milkfish mono

2 - Prawn oriented 4 - Large farms

Farm type

6.3.6 Technical efficiency, stochasticity and the labour input problem

The previous approach to efficiency measurement implicitly assumes that the firms
have access to the same technology. However, it is unlikely that firms face exactly the
same natural resource conditions, or other exogenous factors, such as weather, that
can shock or bias production (Just and Pope, 2001). Therefore, Just and Pope (2001)
emphasise the need for greater biological and physical realism in representing
agricultural technologies. For instance, they stress the timing of input use as an
important variable in determining output, a view that is strongly supported by the
empirical literature in aquaculture. An illustration relates to temporary over-feeding,
which damages the culture environment through the degradation of unused feeds
(Sumagaysay-Chavoso and San Diego-McGlone, 2003).

In this context, with timing as an important variable, Just and Pope (2001) consider
that production response dependent on local weather is represented by:

y f ( x1 ,.....x m , k , ) (6)

where k represents all relevant capital inputs, ε is a vector of weather occurrences,

and, for now, the issue of timing of input use is suppressed for convenience. Then,
adding temporal details, the representation becomes:

y { f * ( f1 ( x1, y0 , z1, 1 ),..... f m ( xm , ym 1, z m , m

)) z i k} (7)

93
 Paper 2/5

describing an m-stage technically efficient input-output relationship using the smooth

function f, where εi represents local weather events occurring during stage ti of the
production process, xm is harvest inputs applied at harvest time tm, and y0 represents
the initial conditions. Just and Pope (2001, p. 646) suggest that the possibility for
weather events to cause significant variation in final or stage output is large:

“Weather can cause certain operations (stages) to be largely

ineffective or consume excessive resources unless choices of
timing are altered. For example, trying to cultivate a field that is
too wet can cause tillage to be ineffective or consume excessive
labour. Or trying to plant a crop before adequate rain can result
in an inadequate stand of seedlings. The associated consequences
for output can be dramatic….An important result following from
the lags [in equation 5.6] is that realised output may not be
monotonically increasing in input variables. For example, bad
weather (pest infestations) can reduce yields while motivating
managers to use more labour (pesticides). Thus a regression of
output ytn on some total input vector x = Σixi may suggest a
negative association for some variables even though δEi(ytn) / δxi
is positive, where Ei is the expectation of ytn taken at time ti (using
information available at time ti). This has led some economists to
model particular inputs as controlling the damage to normal
growth (Feder, 1979).”

While the idea that pesticides are a damage control input is familiar and well-
established, the idea that labour should be modelled as such may be more
controversial. However, there would seem to be a definite validity for considering this
possibility in the case of aquaculture. This is because the weather and other
exogenous events have a direct impact on the quality of the medium of culture
(water), as explained below.

Excessive water temperature, salinity, nutrient content, and, thus, the presence of
disease agents are largely out of the control of the operator but can be linked to
weather events. Storm surges, flooding, organic pollution from unregulated industry
and organic pollution from other fishponds are all environmental events that the
fishpond operator has no control over. The main labour input into aquaculture ponds
is supervision (i.e. caretaking). The importance of this role increases when conditions
on the farm are bad, owing to poor water quality, and a careful watch on the stock is
then needed. The second most significant use of labour involves reinforcing the
dykes, which is also important when conditions are bad (e.g. there is the threat of a
coming storm). Hence, in these two examples, it is conceivable that labour input has
primarily a damage control function, which can explain the negative elasticity of
labour in Table 11.

Kumbhakar (2002) takes a different, but related, approach to modelling the productive
use of labour in aquaculture by formulating an alternative to the seminal stochastic
frontier specification proposed by Aigner et al. (1977). His simultaneous estimation of
production risks and technical inefficiency using data for salmon farms in Norway,

94
 Paper 2/5

establishes that “production risks are increasing with feed and decreasing with labour
and capital” (Kumbhakar, 2002, p.20). He argues thus:

“Labour plays an important role in production risk management.

Farm workers’ main tasks are monitoring of the live fish in the
pens, biophysical variables (sea temperature, salinity, oxygen
concentration, algae concentrations etc) and the condition of the
physical production equipment (pens, nets, feeding equipment
etc). Thus workers ability to detect and diagnose abnormal fish
behaviour, detect changes in biophysical variables and make
prognoses on future development is crucial to mitigate adverse
production conditions” (Kumbhakar, 2002, p.16).

Kumbhakar uses this logic to support his findings that the quantity of labour decreases
production risks. However, the argument put forward in the above passage seems to
imply that the ability/quality of the supervisor may be more important to the
production process than the quantity of supervision.

The previous discussion can be summarised as follows. First, labour input in

aquaculture may be production risk-reducing (Kumbhakar, 2002). Second, labour may
be considered a damage control input rather than a yield increasing input. This implies
that, in the absence of a temporal specification of the technology, negative output
elasticities for labour may result, owing to exogenous events that absorb labour and,
simultaneously, reduce output (Just and Pope, 2001).In light of our finding of a
negative output elasticity for labour (model 3), Just and Pope‟s argument seems
attractive.

6.4 Summary

The results of the economic component of the analysis in this paper are summarized
in a simplified manner in Table 13. At a first level, we find that there is considerable
heterogeneity with regards to the economic performance of the five farming systems
under study, as described below.

Extensive polyculture farms (type 1) achieve, on average, relatively average levels of

profitability. One particular advantage of this farming system lies in its resilience.
Although price risks are high, owing to the importance of prawns, the overall riskiness
of this farming system is relatively low because revenue diversification across four
species acts as a form of insurance. Finally, farms adopting this production system
achieve relatively high levels of technical efficiency, in spite of very large mortality
rates for prawns (in excess of 95%).

95
 Paper 2/5

Table 13 – Summary of analysis of economic indicators

Type 2 –
Type 3 – Type 4 – Type 5 –
Semi-
Type 1 – Low-input Large Semi-
intensive
Extensive labour- milkfish- intensive
prawn-
polyculture intensive oriented milkfish
oriented
systems systems monoculture
polyculture
Costs Average High Low Average Average

Revenues Average High Low Low Average

Gross Average High Low Low Average

margin

Price Risk High High Low Low Low

Overall Risk Low High Average High High

Revenue High High Average Low Low

Diversity
Prawn High High Low High N/A
Mortality
Technical High Low Average High Average
Efficiency

The strength of semi-intensive prawn-oriented polyculture systems (type 2) lies with

their high profitability. The large production costs involved are more than
compensated by the yield increases permitted by intensification, resulting in relatively
large gross margins. However, the catch for these high economic returns is a high
level of risk as well, due primarily to the importance of prawns in these systems. In
fact, the high degree of revenue diversification in this group of farms demonstrates the
importance of secondary species in mitigating risks. Average technical efficiency of
type 2 farms is low (40%).

Not surprisingly, low-input labour-intensive farms (type 3) only achieve low

profitability because of the restrictions on yields that this production system imposes.
On a more positive note, price risks are also low, possibly owing to marketing
strategies that do not depend on the export market. The analysis also suggests that
type 3 farms perform reasonably at a technical level, as indicated, for instance, by
prawn mortality rates that are low compared to the rest of the sample (although, with a
mean of around 80 percent, there is still considerable room for improvement).
Technical efficiency is average at just below 60 percent for this group of farms.

Relatively low profitability characterises large milkfish-oriented systems (type 4),

which is explained by average costs but low revenues. Price risks are low but these
production systems are nonetheless risky, because of the compound effect of high
prawn mortalities and a low level of revenue diversification due to a strong orientation
towards milkfish. However, technical efficiency is estimated to be higher than average
for this farm type at around 80 percent.

Finally, the indicators reported in Table 13 suggest that semi-intensive milkfish

monoculture systems (type 5) are fairly average in terms of their economic
performance. In particular, compared to the other farm types, farms in this group

96
 Paper 2/5

exhibit average profitability and technical efficiency. Further, a relatively low price
risk index is as expected for systems that do not produce prawns, but the overall level
of risk is high, owing to the specialisation of these systems.

Altogether, it is clear that the five production systems investigated here achieve vastly
different economic performances, although it is not possible to single out one farm
type as being economically superior to the others. This is in part due to the apparent
trade-off existing between profitability and riskiness of farming systems, which
operators can influence primarily by adjusting the output mix and level of
intensification of their production plans.

97
 Paper 2/5

7. Social / socioeconomic indicators

7.1 Introduction
At a broad level, the Philippines can be characterised as a labour-surplus economy,
and the only source of livelihood available to many poor Filipinos is often their
labour. Thus, farming systems that require labour directly or indirectly can be
considered to make a positive contribution to social equity by providing jobs to poor
people. However, the social effect of aquaculture may be more complex than
suggested above. As pointed out by Funge-Smith and Briggs (1998) for instance,
social transformations accompanying the development of coastal areas for aquaculture
“can be both positive and negative; the increased income into traditionally poor
coastal areas must be balanced against loss of job diversity, loss of independence,
rising prices and growing inequity between farmers and non-farmers ” (Funge-Smith
and Briggs, 1998 p.118). It is also important to consider the possibility that
aquaculture development may make some people strictly worse off, and that these are
likely to be among the poorest. Hence, if there are employment benefits that can
compensate for the inevitable losses in municipal fisheries brought about by fishpond
development, these have also to be balanced against the loss of independence and,
possibly, of diversity that result from such a change. These changes are made all the
more perilous and fragile by the quasi-irreversibility involved in aquaculture
development, whereby the landscape of an entire region can be rapidly transformed.

The distribution of farm size is also a characteristic of aquaculture that impacts on

social equity. The highly inequitable distribution of farm size within brackish-water
aquaculture in the Philippines, and the exemption of brackish-water areas from
agrarian reform legislation, have both served to concentrate the benefits of
aquaculture. In addition to the distribution of farm size and employment linkages, we
might consider a third level at which aquaculture affects social equity through its
impact on nutrition. The Philippines, like many developing countries, pursues an
agricultural strategy in which the objectives of price stabilisation and self-sufficiency
figure prominently. As Coxhead‟s general equilibrium analysis of the effects of these
interventions shows, there can be negative impacts on the environment (through
incentives to cultivate marginal land), as well as a rise in food prices as a result of
food security related policies. Thus, great caution is needed when making social
equity claims made in connection with food security (Coxhead, 2000). For certain, it
can be said that achievement of food security as an objective can come at the cost of
other objectives, and that it is an important and often-used term in the policies of the
Philippine Department of Agriculture (DA).

The following sections review the employment and food security impacts of
aquaculture. We introduce the specific methodologies and present results for each
farm type. The social equity impact of an inequitable distribution of farm size within
aquaculture is not given treatment here as this is not a farm-level impact. Rather, it is
a sector-level phenomenon. Evidence appraising an efficiency case for agrarian
reform within aquaculture is given in paper 5.

98
 Paper 2/5

7.2 Employment Generation

7.2.1 Background and methodological approach
Unemployment in the Philippines, particularly in the rural areas, is a persistent social
problem. At the national level, the rate of unemployment rose between the years 1996
and 2003 from 7.4 percent to 10.1 percent (Anon, 2005). Further, unrelenting
population growth in the Philippines, constant at the high rate of two to three percent
per annum, suggests that the problem will not easily be resolved and makes
employment creation in rural areas a policy imperative. In this context, aquaculture
has been proposed as a means to increase rural employment possibilities, and prevent
urban drift.

On-farm employment is part of the contribution of aquaculture to total employment.

For example, in Thailand, direct employment in aquaculture in 1995 was estimated at
80,000 (ADB/NACA, 1995 – cited in World Bank, 2002). However, the impacts of
aquaculture on employment are also felt through the multiplier effect, where
aquaculture consumes inputs from other sectors (upstream) and provides outputs for
processing and marketing sectors (downstream).

Multiplier effects in terms of economic activity in general, and employment in

particular, follow Walrasian principles of general equilibrium and were originally
developed by Leontief (1936, 1966). The main methodology, called input-output
analysis, models the interactions among different sectors in the economy such that the
impact of changes in one sector can be predicted. Although useful, input-output
economic analysis is very data-intensive, and usually relies on government-generated
“input-output” or “inter-industry transaction” tables that can often take years to
compile. Hence, the most recent input-output dataset available for the Philippines is
only for year 1994 (Anon, 2005).

In our case, although a full input-output analysis can not be carried out, we employ
the underlying principles of the approach at a more local scale. Our aim is to
distinguish between employment effects of different types of farming systems within
the general agricultural sub-sector of aquaculture. This requires the following primary
data:

On-farm Employment
Input „consumption‟ by different farm types
Labour use of input supply industries
Output produced by different farm types
Labour use per unit of output in processing and marketing industries

Data covering on-farm employment, input use and output are available from the farm-
level survey. Data on labour use in downstream and upstream industries were
collected from a number of interviews with business owners, employees and own-
account individuals working in the input industries and output marketing sector.

7.2.2 On-farm Employment

The use of labour on-farm is important as an indicator of the possibilities for direct
employment generation in aquaculture grow-out. It is clear that many people in
coastal barangays in the Philippines rely on seasonal or full-time employment on
fishponds for their livelihood (see paper 4), but quantifying the effect is difficult.

99
 Paper 2/5

However, a socio-economic survey of the aquaculture industry in the Philippines,

carried out in 1979 (Aspuria and Fabro, 1979, cited in Bailey, 1982), found that total
labour demand, exclusive of guarding and supervising, to be less than 18 man days
per hectare per year for milkfish monoculture, and 45 man days per hectare per year
for polyculture prawn and milkfish. Another survey, (ADB/NACA, 1995), which
covers 13 Asian countries, reports that the Philippines has the lowest labour use per
unit area on extensive shrimp farms. It is calculated from the survey that on-farm
labour use is only 1.5 person months per hectare per year for the Philippines, as
compared to an average of 6.4 person months per hectare per year across all 13
countries. The previous estimate for the Philippines converts into roughly 45 person
days per hectare per year, and so there would seem to have been little change in
labour intensity over the period between these two studies. Further, these studies
suggest that the employment generation potential of brackish water aquaculture in the
Philippines is unlikely to be fully exploited.

By contrast, the average on-farm labour use across all farm types in our sample is
approximately 200 person days per hectare per year, as detailed in Table 17. Unlike
the studies mentioned previously, this figure includes the labour input of the caretaker
and includes both family and hired labour. Our estimate of 200 person days per
hectare per year is roughly equivalent to the Asian average in the ADB/NACA survey
and calls into question the validity of the figure of 1.5 person months per hectare per
year stated calculated from that same survey. The way in which on-farm labour
demand varies across production systems will be discussed later.

7.2.3 Employment Effects

The main assumption of the input-output technique is that of a fixed relationship
between the level of output of an industry and its use of inputs. Thus, by increasing
demand for the output of an industry (e.g. feeds from feed mills), the use of particular
inputs to that industry (e.g. labour) will increase in fixed proportions. The empirical
approach starts by calculating a coefficient that measures the labour intensity in an
industry that supplies inputs to aquaculture:

Backward link labour use coefficient = Labour use / Unit output (8)

In a similar fashion we calculate a coefficient that measures the labour intensity in an

industry that demands the products of aquaculture (e.g. marketing and processing
industries):

Forward link labour use coefficient = Labour use / Unit Input from aquaculture (9)

In the case of backward linkages from aquaculture, when calculating the coefficients
one has to assume that there is zero final demand from all sectors other than
aquaculture. We therefore restrict the analysis to commodities produced by businesses
that state (during interview) that aquaculture represents near total demand for their
output. Additional difficulties concerning forward linkages from aquaculture and the
problem of multiple outputs will be considered in a later section.

100
 Paper 2/5

7.2.4 Input (Backward) linkages

7.2.4.1 Employment in input-supply industries
Some of the industries delivering inputs to aquaculture also supply agriculture and
capture fisheries. Hence, ice plants, transport service providers, artisanal gasoline
stations, fertiliser/feed dealers and fishing supply stores all supply goods and services
to the aquaculture sector, but not exclusively. Given the assumption underlying the
input-output methodology that total demand from non-aquaculture activities should be
zero, we omit certain items and sectors from the analysis. Those include: fertilisers
(mostly imported and also used by agriculture); pesticides (mostly imported, mostly
illegal); transport, gasoline and ice (all with significant demand from households and
fisheries); and „recycled‟ or joint products from other industries such as old bread and
manure.

Figure 14 below sketches the supply chains for each input, and identifies, in bold, the
inputs that are included in the employment effect analysis. The labour use coefficients
for each of the included inputs are given in Table 14 and details of the calculations
and assumptions underlying them are given in appendix 1. The coefficients for
commercial feeds (1.17 person days per kilo) and small shell feeds (1.04 person days
per kilo) are similar, which might suggest that the choice of feeding strategy has a
broadly neutral impact on employment in the input-supply sector. However, this
would be a false conclusion. Small shells are required in much greater quantities
compared to commercial feeds, owing to their inferior nutrient content. As a result,
adopting a feeding strategy based on small shells is likely to have a much greater
positive impact on employment.

The three seed options included in the analysis (hatchery-reared prawn fry, wild
milkfish fry or milkfish fingerlings) have very different coefficients. Prawn hatcheries
produce fry by the million with only a handful of staff, whereas wild collection of
milkfish fry is labour intensive (Ahmed et al., 2001; Frankenberger, 2002). Milkfish
fingerlings have a significantly greater coefficient because these fingerlings are
cultured from fry size in specialised aquaculture ponds by operators that purchase
wild fry during peak season and smooth the annual supply by stunting them in ponds
before selling them on for grow-out. Each fingerling has thus been harvested from the
wild (with a coefficient of 1.05 person days per thousand), and labour has been
employed in tending to and feeding them, as well as maintaining the facilities in
which they have been reared. Mortality of some the fry during this period increases
the value of the coefficient because, obviously, only fry that survive the nursing
period can be sold to operators.

101
 Paper 2/5

Table 14 – Coefficients for the employment effects of input use

Input Coefficient Units Source
-1
Commercial Feeds 1.17 Person days * tonne 1 Interview and site-visit to a

feed production plant

-1
Small shells 1.04 Person days * tonne 4 Interviews and direct

observation in the field

-3
Milkfish Wild Fry 1.05 Person days * fry Ahmed, Magnayon-Umali et

al., 2001
-3
Prawn Hatchery Fry 0.045 Person days * fry 1 Interview and site-visit to

San Felipe, Zambales

-3
Milkfish Fingerlings 9.23 Person days *fry 3 Interviews and site-visits in

Zambales

102
 Paper 2/5

Feed mills
(Commercial
pellets) Figure 14 – The input supply chain of fishponds. Highlighted in bold are
activities that are included in the employment effect analysis.

Feed collectors
(Small shells) Dealers

Feeds and
fertilisers
Recycled feeds
(e.g. by-products FISHPONDS
from other Fry
industries)

Specialist
suppliers (e.g. Fingerlings
nets, bamboo)

Labour
Fertiliser and
pond treatment
dealers (e.g. Fry agents “Kawagan” ponds “Bantay” “Pakyaw”
urea, lime) nursing fingerlings
to resell
Organic fertiliser
dealers (e.g. “Kavgado” “Arawan”
agricultural Hatcheries (prawn, crab, Fry gatherers
wastes, manure) tilapia, some milkfish) (milkfish only)

103
 Paper 2/5

7.2.4.2 Input consumption by farm type

Table 15 presents the intensity of use, for each farm type, of the inputs included in the
employment effect analysis. There is considerable heterogeneity, with large
consumption of small shells and prawn fry by farm type 2 (semi-intensive prawn
polyculture), and of commercial feeds and milkfish fry by farm type 5 (semi-intensive
milkfish monoculture). Farm types 3 (low-input, labour intensive systems) and 4
(large milkfish-oriented systems) are extensive with respect to most of these inputs,
while farm type 1 (extensive polyculture) represents an intermediate case. Although
not directly interpretable, these figures suggest that there are important linkages of
aquaculture to the rest of the local economy, and that the intensity of this linkage is
likely to vary significantly by farm type.

Table 15 – Use of inputs included in the employment effect analysis, by farm type
Commercial Milkfish
Farm Small shells Wild milkfish Prawn fry
feed fingerlings
type (kg/ha/yr) fry (#/ha/yr) (#/ha/yr)
(kg/ha/yr) (#/ha/yr)

1 Mean 2,057 62 1,518 88 102,293

N 53 53 53 53 53
Std. Dev. 4,498 386 1,857 431 88,470
2 Mean 10,780 164 4,737 0 253,526
N 13 13 13 13 13
Std. Dev. 13,637 402 11,211 0 207,352
3 Mean 395 21 2,202 611 21,700
N 35 35 35 35 35
Std. Dev. 1,412 52 2,409 1,554 38,385
4 Mean 234 283 3,198 1,724 38,970
N 11 11 11 11 11
Std. Dev. 521 672 4,796 3,536 77,190
5 Mean 0 1,772 3,050 7,084 0
N 19 19 19 19 19
Std. Dev. 0 3,809 4,101 10,935 0
Total Mean 2,027 328 2,383 1,371 75,615
N 131 131 131 131 131
Std. Dev. 5,921 1,575 4,426 4,901 114,588

104
 Paper 2/5

7.2.5 Output (Forward) linkages

7.2.5.1 Employment generated by the marketing and
processing of aquaculture products
Employment is generated in the sorting, handling, dealing, transportation, processing
(if any) and marketing of the output produced by aquaculture farming systems. The
marketing chain for aquaculture products is described in further details in Figure 15.
Though this supply chain appears reasonably simple, linking the output of aquaculture
farms to the level of employment generated in each of the output-related industries is
fraught with difficulties. One particular problem relates to the fact that fish dealers are
very heterogeneous in terms of size of operation, clients (fish farmers and/or
fishermen), and markets served. In view of this problem, we limit our quantitative
exercise to the analysis of an activity known as cha-cha, which corresponds roughly
to the sorting of harvest that takes place on the dockside. This activity involves men
and women who wait by the consignacions (buying stations) to help sort the harvests
by species and approximate size classes. They work in groups, and sit or crouch in a
circle around the area where the fish are docked. The wages for the cha-cha are paid
by the fishpond operator, through the consignacion, at a fixed rate per weight of each
species.

In Sasmuan, Pampanga, we interviewed six consignacion owners. The terms under

which the cha-cha operated in each one were almost identical. They are typically paid
in nature or in cash at the following rates (for the group in total):

Prawn = 0.5 kg for every 40 kilo processed = PhP150 per 40.5 kilo6

Milkfish / Tilapia = PhP100 per 40 kilo processed = PhP100 per 40 kilo

Crabs = PhP2 per piece

Following interviews with consignacion owners and with cha-cha worker, we found
that each individual earns around PhP100 per day. While the majority of the work is
finished by early afternoon, the workers start early in the morning, and the job can
thus be considered a full-time activity. In fact, few cha-cha workers had other sources
of livelihood. The nature of the work is such that when there is a harvest, there is a
pressure, from both the fishpond operator (who wishes for the sorting to be carried out
as quickly as possible) and from the unemployed cha-cha workers on the dock, for
more people to enter the group until the quantity of fish processed is such that each
worker takes home around PhP100. It is therefore legitimate to calculate the
employment contribution per unit of aquacultural product via the peso metric, as
follows:

Prawns = 1.5 person day per 40.5 kilo

Milkfish / Tilapia = 1 person day per 40 kilo

Crabs = 1/50 person day per piece

7.2.5.2 Output produced by farm type

The sorting process that takes place in the cha-cha and in marketing is dependent, to a
large extent, on the production from aquaculture. Hence, the more productive the

105
 Paper 2/5

fishponds are, the more work is available through these downstream linkages. Table
16 below gives the annual mean production of each species per unit area for each farm
type. The mean values for farm types 1 and 2 are evenly spread across the four
species, although with considerable variability across farms (indicated by the large
standard deviations). Farm type 2, while characterized by an orientation towards
prawn production (paper1), has mean productions of tilapia and milkfish that are
higher, in weight, than that of prawn. These secondary species, although of relatively
low value, generate large amounts of fish for the local markets, and, consequently, a
lot of work for the cha-cha.

Farm types 3 and 4 have a low level of production for all species but milkfish, and
farm type 5 produces milkfish and/or tilapia in large quantities. The levels of output
for the semi-intensive systems (farm types 2 and 5) are approximately double that of
the average for the whole sample.

Table 16 – Output produced by farm type

TOTAL
Tilapia
Farm Prawn Milkfish Crabs PRODUCTION
produced
type (kg/ha/yr) (kg/ha/yr) (kg/ha/yr) (All species
(kg/ha/yr)
kg/ha/yr)
1 Mean 117 194 88 169 568
N 53 53 53 53
Std. 121 308 160 272
Deviation
2 Mean 254 521 219 744 1738
N 13 13 13 13
Std. 378 1199 272 2560
Deviation
3 Mean 50 320 60 20 450
N 35 35 35 35
Std. 122 398 183 95
Deviation
4 Mean 33 793 0 23 849
N 11 11 11 11
Std. 70 1247 0 77
Deviation
5 Mean 0 1030 35 673 1738
N 19 19 19 19
Std. 0 805 153 2896
Deviation
Total Mean 89 431 78 247 845

N 131 131 131 131

Std. 168 716 179 1368
Deviation

106
 Paper 2/5

“Cha-cha”
Figure 15 - Output supply chains for fish, crabs and prawns
produced in brackish-water fishponds. Highlighted in bold are Local markets
activities that are carried out primarily by the poor.

Labour (sorting)

Fish, crabs, “Consignacions” Fish dealers

export-rejected
prawns National markets

FISHPONDS Ice

Ice Plant
Ice Ice Dealers

Exportable Export International

Prawns companies markets

107
 Paper 2/5

Table 17 – Employment generation by farming system and its break down by activity.
Input-use
Hired labour Family Labour Total employment
Farm type employment “Cha cha”
On-farm On-farm effect
effect
average All 29.1 112.6 90.3 26.5 258.5
(person days / ha /yr)
1 20.9 79.4 30.9 19.5 150.7
2 66.5 84.7 44.4 58.2 253.8
3 20.9 211.5 163.3 15.6 411.2
4 31.3 53.0 50.2 22.2 156.7
5 39.9 82.2 180.3 46.0 348.3
% total employment All 11.3 43.6 34.9 10.2 100.0

1 13.9 52.7 20.5 13.0 100.0

2 26.2 33.4 17.5 22.9 100.0
3 5.1 51.4 39.7 3.8 100.0
4 20.0 33.9 32.0 14.2 100.0
5 11.4 23.6 51.7 13.2 100.0

108
 Paper 2/5

7.2.6 Employment generation by farm type

Table 17 synthesizes the results of our partial input-output analysis. It establishes that
on-farm activities generate the most employment from aquaculture. Hence, looking at
the whole sample, on-farm labour demand accounts for over three-quarters of the
estimated labour demand from the sector, the rest relating in almost equal measure to
backward and forward linkages. However, Table 17 also reveals that there is some
variability in the extent to which farming systems generate linkages and employment.
In particular, farm types 2 and 5 are semi-intensive systems and, as a result, generate
more off-farm employment through both backward and forward linkages than
extensive systems. For example, a ten-hectare type-2 farm (semi-intensive prawn
polyculture) generates 665 person days of employment in the sector supplying inputs
to the farms and 582 person days of employment through the cha-cha. The generated
employment corresponds to approximately four full-time equivalents (FTEs) and
matches the labour-use on the farm (also approximately 4 FTEs, divided between
hired and family labour). By contrast, the employment effect of type 3 farms (low-
input labour-intensive systems) is predominantly on-farm.

We now turn to a comparison of the overall labour intensity of the five farming
systems. Although type 3 farms (low-input labour-intensive farms) are only loosely
linked to the broader economy, they generate the most employment (411 person days
per hectare and year) due to their heavy on-farm labour requirements. By contrasts, it
is the strong employment linkages for types 2 (semi-intensive prawn polyculture) and
5 (semi-intensive milkfish monoculture) that allow them to have intermediate
employment impacts. Finally, types 1 (extensive polyculture) and 4 (large milkfish
farms) stand out by the fact that they generate the least employment. Altogether, our
empirical results establish that there is considerable heterogeneity in the extent to
which farming systems create employment, with type 3 farms generating almost three
times more labour demand than type 1 farms.

109
 Paper 2/5

Figure 16 – Total employment generation from different farming systems

600

500

400

300

200

100

0
N = 53 13 34 11 19

1 - Ext. polyculture 3 - Labour-intensive 5 - Milkfish mono

2 - S.I. Prawn poly 4 - Large farms

Farm type

7.2.7 Indirect effects of aquaculture on the regional economy

The previous analysis is now extended by considering the broader effects of
aquaculture development on the local economy. Barrow & Hall (1995, cited in
Stanley, 2003) suggest that regional economic multiplier effects can be broken down
into different categories, presented in Table 18 below. They describe “Growth Poles”
and “Enclaves” as extremes of a continuum of various modes of export-driven
development. Their application is to multinational corporations, whereas Stanley
(2003) applies their framework to the case of shrimp aquaculture in Honduras.
Growth-poles stimulate the local economy in the ways outlined in Table 18, whereas
enclave-type development is vertically integrated with the economic benefits accruing
primarily to investors outside the region.

110
 Paper 2/5

Table 18 - Assessment of business impact on a local economy (Barrow and Hall,

1995, cited in Stanley, 2003)
Growth pole multipliers Enclave-type development
(1)Primary linkages: Large backward linkages (local Small backward linkages
- Backward linkages input use) (high imported
and the purchase of Large forward linkages (local input content)
input supplies processing) Small forward linkages
- Forward linkages and (export or consume
the composition of primary product)
demand
(2)Fiscal links Sector pays substantial taxes to Sector exonerated from
local and national government taxes or exhibits tax
avoidance

(3)Consumption links Sector payroll is well-distributed Sector payroll is

to allow spending by lower- concentrated among high-
skilled employees on locally skilled, professional
made goods employees who purchase
imported consumer goods
(4)Investment spillovers Majority of shareholders local, Investment funds raised
majority of profits/dividends internationally, with profits
reinvested locally remitted outside the region

(5)Employment skill A range of employment Skilled labor imported to the

spillovers opportunities available with region and most
extensive training in skills employment opportunities
transferable to other sectors offered in unskilled, or highly
specialized positions
(6)Secondary links Sector interested in the broad Sector has little interest in
development of the host community; staff establish
economy; political participation few residency/political ties
and assistance to community in and provide few donations
education, health efforts etc to community development
efforts

The primary linkages outlined in the first row of Barrow and Hall‟s description have
been given quantitative treatment, with regards to different farm types. However, the
linkages numbered (2) – (6) in the table above are community, region or country-level
phenomena that emerge at higher scales. Fiscal links are likely to be very low, given
the widespread failure to tax in the Philippines. At the local level, the “municipal
ticket” is charged when operators make sales through the consignacions. While this
tax is collected, the rate in Sasmuan, Pampanga in April 2004 was 0.2 percent of
sales, hence generating minimal revenue for the municipally funded institutions.

Consumption linkages for aquaculture in the Philippines fall more into the growth-
pole column of Stanley‟s typology, with local labour hired seasonally and their wages
spent in local markets. This is consistent with the findings in Irz et al. (2001) who
review the relationship between agriculture and poverty, and stress the importance of
consumption linkages to the poverty alleviating effect of agricultural growth.

Investment spill-overs are dependent to a great extent on where the operator lives with
the positive impact decreasing with the distance between the operator‟s residence and
his fishponds. Employment skill spill-overs are likely to be favourable, and paper 4
provides evidence from household data that suggest that the experience gained as
caretakers and labourers does allow some people to become fishpond operators in

111
 Paper 2/5

their own right later in life. Philanthropic and community-minded initiatives, such as
gifting secondary species from prawn ponds to the local community, that Stanley
terms secondary links, are entirely dependent on the attitudes and motivations of the
operators.

The potential for aquaculture to contribute indirectly to the regional economies of the
Philippines lies in a more effective and systematic collection of revenue from
aquaculture operators. In general, we can say that aquaculture is currently operating as
something of a growth-pole in the Philippines, especially in comparison to other
countries where industrial prawn aquaculture provides the archetypal enclave model
in Stanley‟s typology. However, there is considerable potential for redistributive
benefits through innovations in governance.

7.2.8 Aquaculture for rural development?

The impact of growth in the aquaculture sector is dependent on the technology used
on the farms. Biased technical change that is labour-saving while raising productivity
changes the balance of importance between on-farm employment and off-farm
employment (in the input-supply and marketing sectors relating to aquaculture). Less
labour is used on the farm per unit of output, and even possibly per unit area if the
process of intensification involves some level of mechanization. However, the greater
throughput of intensive systems results in more use of inputs per unit area, which
creates employment in the input-supply sector, while greater output creates
employment in sectors downstream from the farm. In view of the above, it is clear
that the relationship between employment generation and technology/intensification
can only be determined empirically.

In Philippine brackish water aquaculture, our results suggest that, due to the
technologies currently used in the sector, employment generation occurs primarily
through on-farm labour use. Further, although the forward linkages are clearly under-
estimated in our analysis, this is unlikely to affect the quantitative results significantly
owing to the short marketing chain and the lack of processing of aquaculture products.
The result suggests that on-farm labour intensity of new technologies should therefore
be of prime importance when assessing their social impact. However, marketing and
processing activities offer livelihood opportunities primarily to women and, as such,
are important in a qualitatively different manner. Hence, increased economic activity
that adds value in forward linkages is likely to have a positive gender bias in favour of
women (Siar and Caneba, 1998).

7.3 Aquaculture and food Security

7.3.1 Background
The net impact of aquaculture on protein production is ambiguous because most pond
aquaculture farming systems use protein-rich feed inputs of some kind, although it is
possible to raise stock up to a certain density without the need for supplemental
feeding. Further, the overall protein productivity of the farm is difficult to establish as
it depends on the managerial expertise of the operator (in avoiding mortality and
achieving good stock growth), the species under culture, and the farming intensity.
However, it is well established that more intensive farms (and managerially inefficient
farms) will require higher levels of “throughput” (Giampietro, 1997), in the form of

112
 Paper 2/5

undigested feeds, in comparison to non- or low-feed systems. This will result in lower
feed conversion efficiency and net protein production.

Hence, for farming systems that are both managerially inefficient and intensive, or
those that grow species with high protein requirements, it is highly likely that the farm
generates a net loss of global dietary protein (as famously discussed in Naylor, et al.,
2000). This is in contrast to the view that aquaculture has a significant role to play in
providing food security in developing countries (Williams, 1997), as it has the
potential to be an efficient provider of protein to developing nations. While rice and
corn remain the staples of the Filipino, fish for protein is a crucial component to the
nutritional health of the nation.

The previous observation is important in assessing the food security impact of

aquaculture in the coastal areas of the Philippines because the tendency of the Filipino
to “eat nearly every species of fish, even the very smallest fish, and nearly every part
of every fish, including fins, heads and most of the entrails” (Bell and Canterbery,
1976) is mentioned in the literature and can be observed in all local food markets and
homes at mealtimes7. In this light, the use of what the aquaculture industry refers to as
“trash fish” as an input for feeding prawns becomes a social equity issue. In
particular, it is possible that aquaculture development, while lowering the price of
good quality fish for comfortably-off families, may also serve to increase the price of
the poorest quality fish (i.e. feed manufacturers, or even the farms directly, add
demand for poor quality fish alongside consumers), that has long been the staple
protein of the coastal poor.

Casual observation of production practices in Philippine brackish water aquaculture

reveals, however, that the potential for negative food security impact is likely to vary
across farming systems. For instance, many farms culture filamentous algae (“lumot”)
and/or a cyanobacterial mat (“lab-lab”) in their ponds that fish can use as feed. The
protocol for this involves drying the soil until it cracks; allowing a little water into the
pond (a few centimetres depth); and then filling the pond completely once plant
growth has started. Farms using this practice well can raise fish (mainly milkfish)
with little supplementary feeding and may, therefore, generate a positive net protein
production. This stands in sharp contrast with the situation of intensive or semi-
intensive prawn farms that use vast amounts of protein-rich feeds to produce only a
small quantity of output. We now pursue this analysis further by calculating net
protein production for each of the five farming systems. As a last point, however, it is
worth emphasising that most forms of aquaculture compare favourably with livestock
production in terms of protein balance, although there is considerable heterogeneity
within the sector (Naylor et al. (2000) provide a critical appraisal).

7.3.2 Methods
Data on feed types and feeding rates were collected in the farm-level survey.
Subsequently, the feeds used were categorised into three types: Wild Fisheries
Products; Agricultural/Food Industry By-Products; and Commercial Feeds. Table 19
below details the composition of these feed categories as well as the number of farms
in the sample on which each individual type of feed is used. The table clearly
illustrates the diversity of feeds which are used in brackish water aquaculture. Further,
data on the proximate analysis of each feed type were collected from secondary
sources where available. For the small bivalve shells (“gasang” and “agiis”) and the

113
 Paper 2/5

small univalve shell (“suso”), samples were donated for analysis by fishermen
collecting them in the estuaries and lakes in Pampanga province. The fresh samples
were roughly crushed, sun-dried and crushed again. These dry samples were then
taken to the Central Analytical Laboratory of the South-East Asian Fisheries
Development Center – Aquaculture Department (SEAFDEC - AQD). The crude
protein content and nitrogen (N) content of all the feeds are given in Table 19.

Using this information, net dietary protein production (NDPP) for each farm was
calculated as follows:

Net dietary protein production (kg/ha/yr) = Σ Xi – Σ Zj (10)

where Xi denotes protein output per hectare and year of item i, while Zj measures
protein input per hectare and year corresponding to feed type j. Note that inputs
include the fry or fingerlings, although the bulk of the protein input is in the form of
feed. An alternative way of investigating the food security impact of aquaculture is
through the calculation of a protein conversion ratio (PCR), defined simply as the
ratio of feed inputs to fish production:

PCR = Total protein in feed / (protein in output – protein in fry) (11)

Thus, PCR takes a value of one when there is no overall loss of dietary protein as a
result of the activity of a farming system, a value greater than one when there is a loss
of dietary protein, and a value between zero and one for systems that use less protein
in external inputs they produce. These latter systems thus generate a surplus of protein
by using the natural productivity of the pond ecosystem.

114
 Paper 2/5

Table 19 – Proximate analysis of the sample farms‟ feeds, fertilisers and outputs

Number of
farms in the
% Crude sample using
Item %N Source for proximate analysis
protein item (or
producing
output)
WILD FISHERY
PRODUCTS
Seaweed 10.64 1.70 Gerpacio & Castillo, 1979 1
"Alamang" (small 67.23 10.76 Cruz 1997 21
shrimps, capture fishery)
Boiled Fish (usually low- 32.18 5.15 Gerpacio & Castillo, 1979 22
value small fish or “trash
fish”, capture fishery)
"Suso" (univalve shell) 9.34 1.49 SEAFDEC-AQD 11
"Gasang" (small bivalve) 0.69 0.11 SEAFDEC-AQD 46
"Tahong" (mussel meat) 61.14 9.78 Cruz 1997 2
"Agiis" (small bivalve) 0.69 0.11 5
Assume the same as Gasang
AGRICULTURAL / FOOD
INDUSTRY
BYPRODUCTS
Old Bread 7.64 1.22 USDA (2003) 13
Cracked Corn 8.76 1.40 Gerpacio & Castillo, 1979 3
Rice Bran 11.40 1.82 Gerpacio & Castillo, 1979 7
Expired Noodles 9.92 1.59 Gerpacio & Castillo, 1979 1
Egg Yolk 12.40 1.98 Gerpacio & Castillo, 1979 6
Boiled Squash 0.91 0.15 USDA (2003) 1
Chiz Curls (expired 6.67 1.07 Packet 6
cheese-flavoured crisps)
Carabao skin No data 1
Mudpress (from sugar 2.86 0.46 Gerpacio & Castillo, 1979 2
fields)
COMMERCIAL FEEDS
Fry Mash 35.20 5.63 Cruz 1997 10
Starter 33.44 5.35 Cruz 1997 22
Grower (prawn) 32.56 5.21 Cruz 1997 2
Grower (milkfish) 24.36 3.90 Cruz 1997 4
Crumble 33.44 5.35 Cruz 1997 2
Juvenile 24.36 3.90 Cruz 1997 1
Fattener 23.49 3.76 Cruz 1997 2
Finisher 23.49 3.76 Cruz 1997 10
Hog Mash No data 1
FERTILISERS
Carabao Manure - 0.30 Cruz 1997 4
Chicken Manure - 3.80 Cruz 1997 31
Commercial organic - 2.00 Cruz 1997 1
fertiliser
Ammonium sulphate - 14.00 Cruz 1997 1
OUTPUTS
Milkfish 20.53 3.28 USDA (2003) 115
Prawn 23.08 3.69 82
Tilapia 17.60 2.82 G&C, p. 6 26
Crab 18.50 2.96 USDA for Queen Crab 46

115
 Paper 2/5

The NDPP indicator is useful as it accounts for the protein productivity per unit area
of the farming system under study, but is difficult to interpret. The PCR indicator is
more easily interpretable, but is a ratio and thus does not account for land
productivity.

7.3.3 Results
7.3.3.1 Net dietary protein production
Five outliers were removed from the sample of 131 farms. There is still considerable
within-type variability in these results but we can observe that farm types 1 (extensive
polyculture) and 3 (low-input labour intensive) have marginally positive mean NDPP,
suggesting that they make a positive contribution to net protein production. The other
farm types have wide 95% confidence intervals so that it is difficult to reach definitive
conclusions regarding their net impact on protein production. However, we note that,
as expected, the more intensive farm types (2 and 5) have negative NDPP means with
similar values and ranges, suggesting that these systems consume more proteins than
they produce.

Figure 17 – Mean and 95% confidence intervals for NDDP by farm type
400

200

-200

-400

-600
N= 51 11 34 11 19

1 2 3 4 5

Farm type

7.3.3.2 Protein conversion ratio

The sample mean of the protein conversion ratio, at 0.89, indicates that the sample
farms, on average, produce more dietary protein than they use. However, Figures 18
(scores by farm type) and 19 (entire distribution) also establish that there is
heterogeneity within the sample. As expected, the semi-intensive systems (farm types
2 and 5) have mean PCR values greater than unity, owing to the higher level of
throughput in these systems. Operators of these farming systems cannot use the
natural productivity of the water efficiently, due to the crowding of stock in the ponds,
and therefore have to use external feeds.

116
 Paper 2/5

Very high values (in excess of three) are explained by inefficiency in the feeding
strategies of caretakers. Often these strategies are based on rules of thumb given to
them by the operators (e.g. two sacks per pond per day), and thus are not adaptive in
the sense that they do not respond to changes in the behaviour of the prawns/fish, the
presence of undigested feeds, or an increase in stock mortality.

Figure 18 – Mean and 95% confidence intervals of PCR by farm type.

2.5

2.0

1.5

1.0

.5

0.0

-.5
N = 51 11 34 11 19

1 2 3 4 5

Farm type

Figure 19 – Frequency chart of PCR

70

60

50

40

30
Number of farms

20

Std. Dev = 1.52

10
Mean = .89
0 N = 126.00
0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00
.50 1.50 2.50 3.50 4.50 5.50 6.50 7.50

Protein conversion ratio (PCR)

117
 Paper 2/5

7.4 Summary
This section has established that the employment generating potential of aquaculture
in the study area varies considerably by farm type – both in the composition of
employment created (on or off the farm) and its volume. Altogether, we find that on-
farm activities dominate the employment effect of aquaculture, and, consequently,
that farm types that are directly labour-intensive are more likely to increase regional
employment. It is worth noting, however, that in some cases, the on-farm labour for
peak periods to carry out tasks such as harvesting or dyke maintenance is not provided
by the local communities. Instead, it is brought in by absentee operators keen to use
people they can trust. This serves to limit the positive regional benefits of aquaculture
to rural coastal areas. With regard to food security, the results should be treated with
caution as there is great variability across the sample that prevents us from reaching
definite conclusions. However, it appears that aquaculture, altogether, makes a
positive contribution to dietary protein production in the study area, and that extensive
systems tend to outperform semi-intensive ones in this regard.

Although the results demonstrate that there is heterogeneity in various social

indicators across farming systems, farm type 3 (low-input labour-intensive systems)
stands out by performing well in several respects. For this group of farms, the average
size is small (four hectares) and labour is the most important factor of production.
While this farming system generates the smallest employment effect, the large on-
farm labour use more than compensates so that type 3 farms generate the most
employment. This farm type also accounts for the largest net surplus of dietary
protein, owing to its limited use of external protein-rich feeds.

The favourable conclusions that we reach for this model of low-input aquaculture
contrast with the results of the economic analysis in the previous section. There, it
was shown that farm types 1 (extensive polyculture) and 2 (semi-intensive prawn-
oriented polyculture) were achieving the highest levels of economic performance.
Thus, there is a possible tension between the objectives of economic efficiency and
social equity, which will be explored further in paper 3 by deriving compromise
solutions. We now turn to the analysis of the ecological and environmental properties
of the five farming systems under investigation.

118
 Paper 2/5

8. Ecological indicators
8.1 Introduction
There is some evidence from the highly aggregated green accounting methods used by
the United Nations statistical division that the Philippine economy grew between
1988 and 1994 in a manner that was environmentally sustainable (Bartelmus, 1999).
Indicators that monitor an entire economy are helpful for monitoring general progress
towards the ideal of sustainable development, and particularly when used to mitigate
optimistic development diagnoses based on GDP. However, such indicators do not
reveal the problems of specific sectors and, thus, can be of limited practical value in
guiding policy.

In this context, this section introduces indicators of the three areas of environmental
and ecological concern that can be linked to the management of fishponds in the
Philippines. These correspond to: the issue of coastal natural resources appropriation,
which we propose studying with ecological footprint analysis; the problem of
eutrophication, which we approach by an analysis of the nutrient balance of the farms;
and the issue of pesticide use. This section introduces each indicator in turn and
includes the following: an introduction to the theory that motivates the choice and use
of the indicator; a review of other studies that have employed the indicator (where
appropriate); a detailed explanation of how the indicators are calculated; and, finally,
the results, which are analysed across farming systems.

8.2 Ecological Footprint

8.2.1 Introduction
The ecological footprint is a biophysical accounting method of the appropriation of
natural resources by different human enterprises that was initially introduced by Rees
and Wackernagel (1992). The concept has huge intuitive appeal, largely through the
strength of the imagery associated with the footprint metaphor, the size of which
determines the degree of pressure that human beings impose on their environment.

The ecological footprint has found particularly wide usage in the study of aquaculture
systems for a variety of reasons. Firstly, the use of natural resources and generation of
wastes by aquaculture, particularly of prawn and salmon, have concerned ecologists
for many years (Bilio, et al., 1981; Brown, 1989; Primavera, et al., 1993; Primavera,
1995; Roberts and Muir, 1997; Naylor et al., 2000). Secondly, the natural resources
used by aquaculture are generally appropriated from highly sensitive ecosystems in
the coastal zone, where changes as a result of human pressure can be observed within
short time horizons. This is in comparison to other human activities, such as car use,
for which the impacts have longer gestation periods spread over wide areas, and, thus,
entail greater degrees of complexity and uncertainty in their measurement. Thirdly,
one of the rationales for aquaculture is that it can help ameliorate the plateau or
decline in wild fisheries yields. However, many of the inputs required by, and
undesirable joint products from, aquaculture can add to the stresses faced by the
capture fisheries. The ecological footprint has been seen as a useful tool to highlight
this paradox.

119
 Paper 2/5

8.2.2 The ecological footprint method

The ecological footprint method begins with the identification of all the major direct
and indirect material and energy inputs and outputs used and produced by a system of
interest. These input and output flows are then quantified where possible, and the
most complex stage of the method then involves their subsequent conversion into a
commensurate unit: the ecosystem area. In particular, this last step requires ecological
study to inform the conversion, based on average productivities for the inputs of
interest from their source ecosystems (e.g. marine coastal shelf, agricultural field,
mangrove stand). The subsequent summing of ecosystems areas gives an overall
indicator of total ecosystem support to the system. Thus, the total supporting
ecosystem area to a system needed to maintain ecological equilibrium is known as the
ecological footprint. A number of important limitations and assumptions apply to the
conversion and summing stages, which we discuss in the next section.

Ultimately, the ecological footprint can be expressed in two ways. Most studies have
used the ratio of shadow ecosystem area to physical area of system (Larsson et al.,
1994; Wackernagel and Rees, 1996; Kautsky et al., 1997; Folke, et al., 1998;
Ronnback, et al., 2003). This places emphasis on the technologies involved in
production. However, in the case of production systems for a single product, Bunting
(2001) argues that shadow ecosystem area per unit of product is the most helpful
measure for comparative analysis. This latter formulation creates similarities with
life-cycle assessment (LCA), a broad methodology that treats the product as the object
of analysis, and aims to incorporate the many different kinds of environmental impact
involved in the production and consumption of the product „from the cradle to the
grave‟ (Heller and Keoleian, 2003; Papatryphon et al., 2004).

Expressing the ecological footprint using the Bunting formulation allows for high
levels of resource appropriation to be off-set by high levels of productivity. A farming
system with low impact and low productivity would give similar results to a system
with high impact and high productivity. Therefore the hectare per kilo formulation
can be considered an indicator of ecological efficiency – a concept that is
fundamentally different from that conveyed by the original ecological footprint
(Stoorvogel, et al., 2004 represents a recent example of a paper that measures
ecological efficiency rather than absolute impact). Given the fact that we have
accounted for the productivity of the production systems elsewhere, and, given the
multiple-output nature of brackish water aquaculture, we favour the original
formulation – thus ecological footprints are expressed as hectares of coastal
ecosystem per hectare of fishpond.

In an ideal case, all inputs and waste products would be convertible into a
corresponding ecosystem area, allowing full accounting. However, in practice, some
are difficult, if not impossible, to incorporate. For example, mineral resources are
conceptually difficult to footprint as are many types of wastes (Tyedmers, 2000). A
careful examination of the limitations of the method is therefore in order.

8.2.3 Limitations and assumptions underlying the ecological footprint

Van den Bergh and Verbruggen (1999) offer a rigorous critique of the method. First,
regarding the usefulness of the method, they object to the exaggerated claims made on
behalf of the ecological footprint (EF) that it can reveal (un)sustainability on its own.

120
 Paper 2/5

As they point out, “a single aggregate indicator like the EF does not allow for trade-
offs among the three central dimensions of ecological economics‟ evaluation, i.e.
efficiency, equity and sustainability” (Van den Bergh and Verbruggen, 1999, p. 64).
At a more practical level, they suggest that the hypothetical nature of the calculated
land areas makes communicating the results of the analysis to the users difficult.

Second, implementation of the EF is fraught with practical and conceptual difficulties.

Hence, the aggregation of different kinds of land-use inherent to the method is
problematic, and often relies on weightings that are often not made explicit. Van den
Bergh and Verbruggen (1999) also object to the treatment of energy in most
applications of the EF (land areas are calculated to sequester carbon, in a manner
completely incompatible with marginal thinking in economics), and the arbitrariness
of the scales at which the EF is calculated (e.g. global, regional, national, cities).
Finally, they argue that trade can, in principle, spatially distribute the environmental
burden among the least sensitive natural systems and that the EF thus “has an anti-
trade bias” (Van den Bergh and Verbruggen, 1999, p.67).

We share the concerns expressed above and propose the following in an attempt to
limit their significance. The ecological footprint is only one indicator in our study,
alongside economic, social and other ecological indicators, and does not support on its
own any claims regarding the sustainability of alternative farming systems. Further,
only one broad category of land is used in the analysis (mudflats / seabeds /
mangroves), which, thus, justifies the implicit equal weighting of each hectare of land
input in the calculations. Finally, what we attempt is only a partial calculation, but,
hopefully, one that is free from many of the problems associated with more ambitious
applications of the EF (e.g. Tyedmers, 2000).

8.2.4 Applications of the ecological footprint method to aquaculture –

a review
The ecological footprint and its use for sustainability assessment are reviewed by
Deutsch et al., (2000), and, with a specific focus on seafood production, by Folke, et
al. (1998). A study of particular relevance for this project is Larsson et al. (1994) who
calculate the ecological footprint for the resources required to sustain production in
semi-intensive shrimp farms in Columbia. Their findings are highlighted in Figure 20
below, which shows that the mangrove nursery function for the production of post-
larvae creates the largest dependence of shrimp production on natural ecosystem.
Illustrating the uncertainty and difficulties that surround the method, the authors give
wide estimates corresponding to the upper and lower bounds for the abundance of
gravid spawners and post-larvae, their dependence on the mangrove, and the efficacy
of their capture by collectors. Establishing a functional relationship between
mangroves and other inter-tidal habitats as nursery areas for fisheries appears
particularly difficult, but critical to the results of the analysis. This challenge was
tackled by many other authors who subsequently included the results in various kinds
of ecological and economic appraisal (Primavera, 1998; Ley, et al., 1999; Ronnback
et al., 1999; Barbier, 2000; Stevenson, 2002).

121
 Paper 2/5

CO2 sequestering
Agricultural forest 0.8-2.5 m2
ecosystems
0.5 m2

Shrimp
pond 1 m2
Marine
ecosystems
14.5 m2

Postlarval
mangrove nursery
Mangrove 9.6-160 m2
detritus 4.2 m2 Mangrove
lagoon 7.2 m2

Figure 20 – Ecosystem areas appropriated for each square meter of semi-intensive

shrimp farm in a coastal mangrove area, Bay of Barbacoas, Columbia (Larsson, et
al., 1994, redrawn in Folke et al., 1998). Where the values are expressed as a range,
the smaller white squares represent lower bounds for the estimated area, and the
larger grey squares the upper bounds.

A related study establishes, based on a case-study in India, that each prawn hatchery
is dependent on an estimated 455 hectares of mangroves for the supply of spawners
(Ronnback et al., 2003). The paradox is, of course, that the demand for spawners by
the hatcheries is driven by coastal prawn aquaculture ponds that occupy the same
mangrove areas on which the supply is dependent.

In a different context, Kautsky et al. (1997) compare two distinct farming systems for
the production of tilapia in Lake Kariba, Zimbabwe: intensive cage culture and semi-
intensive pond culture. Their calculations suggest that cage culture in the lake is
appropriating resources at a much greater rate than pond culture in the areas around
the lake, and suggest that the stress placed on the lake ecosystem by tilapia cage
culture violates minimum conditions for sustainability.

Finally, Tyedmers (2000) employs the Bunting formulation of the ecological footprint
to compare salmon production from aquaculture and wild fisheries. The results
indicate that both farmed Atlantic and Chinook salmons have a larger total ecological
footprint than any of the five commercially harvested species considered in the study.
For instance, at over 12.5 and 16 hectares per tonne respectively, farmed Atlantic and
Chinook salmons impose a much heavier pressure on the environment than wild

122
 Paper 2/5

caught Sockeye, Chub and Pink salmons – the footprints of which are all under 6
hectares per tonne (Tyedmers, 2000). Demonstrating the usefulness of the approach,
the author simply concludes that the biophysical costs of production are greater for
farmed salmon than for wild caught salmon.

Finally, this literature review establishes that there are currently no published studies
on the ecological footprint of polyculture production systems – a gap that the current
study aims to fill.

8.2.5 Footprint calculations

This section calculates the area of resource appropriation for prawn fry and for shell
collections, in hectares of coastal ecosystem per hectare of farming system. Hence,
this is only a partial calculation of the total ecological footprint of the farm, owing to
the difficulty in obtaining data for footprinting other inputs, such as milkfish fry, crab
fry, other feed inputs and fertilisers. However, this partial calculation is relevant as it
indicates appropriation of coastal resources in the areas around the ponds. This local
footprint is in contrast to some of the global calculations that have been carried out for
aquaculture when impacts as diverse as greenhouse gas emissions and energy use in
plastics, building materials, and other indirect inputs have been included in the
calculation (Tyedmers, 2000). Conversion of all these diverse impacts into non-
equivalent ecosystem areas and their subsequent aggregation make the interpretation
of ecological footprint results difficult.

In addition, the local nature of the resources that are included in this analysis mean
that the results can be interpreted as being strongly linked to local environmental
carrying capacity – a concept that is becoming much more widely used in aquaculture
as production sites proliferate (Kautsky et al., 2000; Ronnback et al., 2003). As an
informal indication of the relevance of this concept of local environmental capacity,
one interview with a Municipal Agrarian in one of our study areas revealed that
fishermen often complain about the effects of trawling for shells8 on their fishing. The
trawling has two main ecologically disruptive impacts: it directly disturbs the benthic
ecosystem on the floor of the rivers and estuaries, and it generates large quantities of
bycatch. A particularly important component of the bycatch is the fry of commercially
important, low-value species on which the municipal fisheries are dependent. Hence,
there is competition for alternative uses of the ecosystem, which creates social
tensions. In fact, fishpond operators, being so dependent on the shells, often
intimidate the fishermen and flout municipal ordinances banning the trawling for
shells in particular waters, in a context where the Department of the Environment and
Natural Resources (DENR) does not have the resources to enforce the ordinances. We
now turn to the the analysis of the empirical results.

8.2.5.1 Prawn fry

We use the values calculated in the study by Ronnback et al. (2003) for the
productivity of the P. monodon spawner collection industry in mangrove areas in
Bangladesh, and for the productivity of hatcheries in producing viable postlarvae fry
from the spawners. Their estimates are expressed as ranges between lower and upper
bounds and are presented in Table 20 below.

123
 Paper 2/5

Table 20 – Footprint of P. monodon postlarvae.

Hectares of
Spawners per hectare of Spawners required by mangrove required
mangrove hatcheries to produce 1 per 1 million post-
million post-larvae larvae
lower 0.9 6.7 6.00
estimate
upper 1.8 14.1 25.4
estimate
Source: Ronnback et al. (2003); Ecological field study and interviews with hatchery
operators.

8.2.5.2 Small shells

Following Tyedmers (2000), we employ the methods of Pauly and Christensen (1995)
to explore the relationship between primary productivity and fisheries yield. The unit
of account for primary productivity is carbon, and the primary productivity required P
to support the marine organisms consumed as feeds in aquaculture is given by:

P = (M/W)*10(T-1) (12)

where P is expressed in grams of carbon fixed; M denotes the wet weight mass (g); W
is the ratio between wet weight and carbon content; and T is the trophic level at which
the organism feeds (where autotrophs are assigned a value of unity). This equation
assumes an average transfer efficiency between trophic levels of ten percent

In the case of the small live shells fed to tiger prawns, all the organisms9 are
herbivorous and therefore have a trophic level of two. The percentage carbon for
bivalve species is in the region of 40 to 50 percent of total weight, based on which we
assume a value of two for W.

The shells feed on phytoplankton suspended in the water column, and reside on the
bottom of the river, where they can be harvested using trawl gears. The area of
ecosystem support required is calculated by dividing the value of P by an estimate of
the average rate at which carbon is fixed (i.e. the net primary productivity - NPP) in
the supporting coastal ecosystem (Tyedmers, 2000, p. 47). Following Tyedmers
(2000), such an estimate is derived from Longhurst, et al. (1995) who list NPP
estimates for 57 discrete biogeochemical provinces encompassing the oceans of the
world. Complicating the analysis, the Philippines lie at the boundary of three of
Longhurst‟s provinces, referred to as China Seas Coastal (CHIN); Western Pacific
Warm Pool (WARM); and the Kuroshio current (KURO) that has its origin to the east
of Luzon and flows up towards Japan. The primary production rates for these three
provinces are presented in Table 21 below.

124
 Paper 2/5

Table 21 – Primary productivity in oceanic provinces bordering the Philippines.

Primary production rate Primary production rate
Domain Province
-2 -1 -2 -1
(g carbon m day ) (g carbon m yr )

Coastal CHIN 1.70 619

Westerlies KURO 0.53 193

Trades WARM 0.22 82

Source: Longhurst et al. (1995)

We use data for the CHIN province because the fisheries that supply the aquaculture
sector with feeds are coastal or estuarine. There is uncertainty in identifying
boundaries between provinces in the coastal domain because “the processes that force
or constrain algal blooms are more diverse than in the open oceans: river plums,
bathymetric features, bottom roughness, tidal fronts, tidal mixing, local (mountain
gap) extreme wind stress, shelf break and coastal upwelling and downwelling. Even
this list does not exhaust the possibilities.” (Longhurst et al, 1995, p.1258). We are
sure that the coastal waters of the Philippines are some of the most productive in the
world, and therefore justify using the CHIN province data. This province is ranked 6th
most productive out of 57 global provinces.

Thus, the ecosystem area to support the production of one kilo of small shells is
calculated as:

Shells footprint = Productivity required (P) / Primary Production rate (CHIN)

= (1000 / 2 * 10(2-1)) / 619
= 8.08 m2 kg-1 (13)

8.2.6 Results
The results of the ecological footprint calculations by farm type are given in Figure
21. They should be interpreted as conservative estimates, based on the use of
generous figures for the primary production rate (i.e. the highest primary productivity
of three possible coastal provinces was used in calculating the footprint for the
harvested small shells), and the productivity of spawners from mangroves (i.e. the
lower estimate from the paper by Ronnback et al. 2003).

125
 Paper 2/5

Figure 21 – Ecological footprint estimates by farm type for two inputs: small shells
and prawn fry
20

10

-10
N = 53 15 37 11 19

1 2 3 4 5

Farm type

In central Luzon, the extent of brackish-water aquaculture development has led to a

situation in which there are virtually no unexploited semi-natural areas in the inter-
tidal zone. As Figure 21 shows, farm type 2 (semi-intensive prawn polyculture),
depends on ecosystem support from these areas at the average rate of 8.9 hectares per
hectare of farming system for feed and prawn fry inputs alone, with the ratio for some
farms reaching 50. Thus farming systems of this type are forced to appropriate
resources from areas outside of the region, owing to overexploitation in the region
where the ponds are situated. This result seems plausible and has important, as well as
practical implications, as illustrated by an anecdotal example. The recent development
by a prominent local aquaculture entrepreneur of a business that “imports” shells from
outside the region has recently been observed. These shells, harvested in the large
freshwater lake Laguna de Bay and elsewhere, are transported by truck to Pampanga
province, and finally distributed to caretakers working for the operator or sold to other
fish farms. Thus, these systems are appropriating resources from other stressed
ecosystems at low cost to the prawn aquaculture operators through vertical integration
of the supply chain.

At another level, and although not directly related to our footprint analysis, there
would also appear to be a critical shortage of gravid spawners in the wild, in part due
to the appropriation of their nursery areas by aquaculture ponds. This endogenous
feedback within the broad aquaculture system is perhaps obvious to an ecologist, but
it has not been recognised to any significant extent by fish farmers themselves, or by
the Philippine government. Further, the problem is unlikely to disappear in the future
as complete domestication of the P. monodon life-cycle is not imminent.

126
 Paper 2/5

Going back to the ecological footprint analysis summarised in Figure 21, the low /
zero values of the indicator for farm types three to five indicate that fry and wild-
caught feed requirements for these systems are met exogenously to the ecosystems
where the ponds are situated. Thus, these production systems do not create the same
endogeneity problem where the success of the ponds is dependent on the health of
ecosystems that are destroyed to produce them. This is an important conclusion in
planning new sites for aquaculture, although it does not mean that these apparently
more environmentally friendly farming systems do not appropriate coastal or marine
resources – rather, they do so in a diffuse manner (i.e, from seas across the nation,
and, for the case of imported fish meal and milkfish fry, from other nations).

8.2.7 Summary
Owing to data constraints and its land-based orientation, the ecological footprint
concept is difficult to operationalise for comparative study. In the current case this is
particularly problematic for inputs relating to milkfish fry. Milkfish aquaculture
depends on the same kinds of coastal resources as prawn-oriented systems, but
calculating the ecological footprint for milkfish fry is difficult because the milkfish
are not mangrove-dependent to the same extent as prawns, and thus cannot be
intercepted for ecological study. Therefore, any comparison of farming systems with
different milkfish and prawn orientations is likely to be biased against the prawn-
oriented systems owing to the impossibility of calculating ecosystem support areas for
milkfish production. However, the ecological footprint has been useful in the current
study in highlighting the farming systems (particularly farm type 2) that have
intensified to the point where their resource needs cannot be met by the local
ecosystems.

8.3 Nitrogen Balance

8.3.1 Rationale
The nutrient balance of the farm is proposed here as a means-based measure of the
likely contribution of the farming system to eutrophication in the coastal zone. In
systems terms, nutrient balance represents an indicator of the pressure applied to the
coastal ecosystem by the farming subsystem. However, it should be recognised at the
outset that this indicator is likely to be particularly site-specific in its relevance and
importance, because the effect of the release of nitrogen by a farm depends on the
local context. That is why, as Sumagaysay-Chavoso and San Diego-McGlone (2003)
note, discharge limits should not only be set per unit of output, but, instead, on the
total amount of aquaculture production activity in a given area: “Even if stocking
density is low for one farm, an increasing number of farms may exceed the
environmental capacity of certain receiving water” (Sumagaysay-Chavoso and San
Diego-McGlone, 2003, p.416).

Regarding the causes of nitrogen emissions from fish farms, Handy and Poxton
(1993) suggest that excessive pollution may arise from poor farm management rather
than the volume of metabolic output (i.e. faeces production). Poor farm management,
in this instance, refers to an inability to adjust the diet regime in order to ameliorate
the effects of post-feeding excretion . Applying this to the Philippine case, given the
simple rules of thumb that constitute the management regime instructions passed on to
the caretakers from the operators, it might be unrealistic to expect anything but “poor
management” in the case of excretion and nutrient waste management. When
combined with the degree of crowding of the ponds (there is little else but fishponds

127
 Paper 2/5

covering certain municipalities), and the numerous other sources of organic discharge
(e.g. human houses, sugar and alcohol industry processing), nutrient enrichment
(eutrophication) is likely to be a problem. Its consequences are described by
Merrington et al. (2002, p.29) as follows:

Changes in nutrient levels may affect the species composition of algal

communities, with knock-on effects to other animals in the aquatic/marine
food chain;
Decomposing algae and weeds can deoxygenate water causing fish mortality;
Algal blooms are often responsible for taints and problems with filtration in
surface-drawn public water supplies, thereby increasing the cost of
purification;
Surface algal blooms detract from the appearance of waters and impair their
amenity value;
Some algal species produce ecologically-disruptive toxins which can also be
hazardous to humans.

Phosphorus is the other nutrient, alongside nitrogen, that may be instrumental in the
eutrophication of coastal waters as it is the stoichiometry (i.e. the ratio of abundances
of the two chemical species) in particular that is the reason for changes in the algal
communities. However, nitrogen is often used as the metric to measure the ecosystem
disturbance resulting from agricultural and aquacultural practices (Martin, et al.,
1998; Janssen, 1999; Wijnhoud, et al., 2003).

8.3.2 Method
Data on the proximate analysis of all feed inputs to and outputs from the farms in the
sample were collected when possible from secondary sources, and were derived from
primary analysis when necessary (for two species of small shells – see proximate
analysis results in Table 19). For each farm in the sample, the following formula was
then used to calculate the nutrient balance, which is self-explanatory:

(Nfeeds + Nferts + Nfry – Noutputs) / Farm size = Nutrient balance (kg N/ha/yr) (14)

8.3.3 Results
The results of the analysis of the controllable nitrogen flows by farm type are given in
Tables 22 and 23 as well as Figure 22. These results show that feed is the main source
of nitrogen to the ponds for all farm types, with the exception of farm type 1, for
which it is fertiliser. Further, a comparison of the means of the N-loss using ANOVA
shows that there are statistically significant differences across the farm types.
However, one-way ANOVA using Scheffe‟s post-hoc test does not yield statistically
significant differences in multiple comparisons, except between farm types 1 and 2 (at
the 10% level).

Semi-intensive prawn-oriented polyculture (farm type 2) shows particularly wide

within-class variation, yielding a large 95 percent confidence intervals for the mean
(Figure 22). Some individual farms even have zero or negative nutrient balances (i.e.
they actually remove nutrients from the inflowing water), although the means are
positive for all farm types. The within-type variability for farm types 1, 3, 4 and 5 is
quite low, therefore suggesting a strong association between farming system and

128
 Paper 2/5

nutrient efficiency. Farm types 1 and 3 have the best results for this indicator with
mean values just above zero, whereas farm type 2 (semi-intensive prawn polyculture)
has the highest mean value but is highly variable within-type.

Table 22 - Descriptive statistics for the indicator of total nitrogen loss (kg N per ha
per yr)
95% C.I.
Farm type N Mean Std. dev. Std. error Min Max
for mean
Lower Upper

bound bound

1 53 20.0 66.9 9.2 1.6 38.5 -49.9 365.0

2 14 231.9 732.00 195.6 -190.8 654.6 -74.5 2760.7

3 35 49.9 54.9 9.3 31.0 68.8 -28.7 185.4

4 11 121.1 84.9 25.6 64.1 178.2 5.1 272.5

5 19 167.5 139.6 32.0 100.2 234.8 -37.4 483.4

Total 132 80.1 253.7 22.1 36.4 123.8 -74.5 2760.7

ANOVA
Sum of Mean
df F Sig.
Squares Square
Between 709417.1 4 177354.3 2.916 .024

Groups

Within 7724105.7 127 60819.7

Groups
Total 8433522.7 131

Table 23 – Breakdown of the results of the nutrient balance indicators by farm type.
total N loss (kg N /
N N
N N content ha / yr)
Farm content content
content of output NITROGEN
type feeds ferts
fry kg/yr kg/yr BALANCE
kg/yr kg/yr
INDICATOR
1 Mean 98.55 199.58 8.18 124.33 20.04
Std. Error of 27.44 76.28 1.34 18.68 9.18
Mean
2 Mean 492.66 32.02 5.03 70.34 231.90
Std. Error of 311.71 24.55 1.44 24.78 195.64
Mean
3 Mean 18.15 378.15 4.68 74.36 49.91
Std. Error of 6.85 110.09 1.92 22.49 9.28
Mean
4 Mean 1703.09 4251.59 66.32 1089.95 121.13
Std. Error of 1147.57 1140.70 40.18 293.14 25.61
Mean
5 Mean 509.47 1264.76 15.57 407.83 167.53
Std. Error of 282.33 528.97 8.24 161.94 32.03
Mean
Total Mean 311.89 720.15 12.83 226.63 80.09
Std. Error of 112.76 159.03 3.78 42.09 22.08

129
 Paper 2/5

Mean

130
 Paper 2/5

Figure 22 – Nitrogen balances for each farm type.

800

600

400

200

-200

-400
N = 53 14 35 11 19

1 2 3 4 5

Farm type

8.4 Pesticide Application

A number of different pesticides are used in aquaculture, ranging from extremely
toxic chemical compounds, such as sodium cyanide or Brestan, to less toxic, semi-
natural, plant-based compounds, such as tobacco dust. Figure 23 summarises the
application rates derived from the farm-level survey. There would seem to be no clear
technological basis for pesticide application, in the sense that the application rate is
not explained by farm type, as indicated by wide confidence intervals and little
differences in mean values across farm types.

This somewhat surprising result may be explained by a methodological problem.

There is a major issue of commensurability with the calculated indicator because the
products used on the farms differ greatly in terms of their toxicity. Using some kind of
index of toxicity would provide a more satisfactory way of aggregating pesticides (as
suggested in Maud et al. (2001)).

Although our analysis is limited, it is clear that there are potentially serious human
health and environmental issues associated with the more toxic pesticides, which
should be a source of concern because their legal status under municipal and national
laws remains ambiguous. For instance, sodium cyanide application would seem to
have been banned for use on fishponds but is freely available in local markets
(imported from Taiwan or South Korea) and employed on 90 percent of farms in our
Central Luzon sample. Brestan is also an illegal piscicide, but operators were open
about its use, particularly in Western Visayas. It would seem as though laws designed
to protect the marine environment and the workers on fishponds are not enforced in
any meaningful way.

131
 Paper 2/5

Figure 23 – Mean and 95% confidence intervals for pesticide application rate
(kg/ha/yr) by farm type
12

10

-2
N = 54 15 36 11 19

1 2 3 4 5

Farm type

8.5 Summary
The ecological footprint methodology was applied to two inputs to the farms: prawn
fry and the small shells used as feed. We find that a conservative estimate of the
coastal resource appropriation by semi-intensive prawn polyculture farms falls in the
range of ten hectares of coastal ecosystem per hectare of fishpond. This highlights the
pressures that these kinds of aquaculture systems place on the environment in other
areas indirectly, and, often, some distance from the ponds. However, the use of the
ecological footprint as a comparative indicator proved difficult owing to the
conceptual challenge of footprinting milkfish fry inputs.

Pesticide use does not vary significantly by farming system, in contrast with the
situation in agriculture where the rate of pesticide application is usually correlated to
the choice of variety (high-yielding versus traditional) and overall intensification
level.. Thus, pesticide application as calculated in this section is not a useful indicator
for comparative analysis of sustainability in the current study.

It therefore seems that the nutrient loss to the environment, measured from partial
nitrogen budgets, represents the most useful indicator for comparing the ecological
impacts of different fish farming systems, as well as the trade-offs that might exist
between ecological impact and socio-economic properties of these system.
Applications examining these trade-offs, considered in paper 3, will thus only use this
indicator to measure ecological performance.

From the results of the nutrient loss indicator, we find that farm types 1 (extensive
polyculture) and 3 (low-input, labour-intensive systems) are the most environmentally
sustainable.

132
 Paper 2/5

9. Conclusion
We have built up a picture of the complexity of the brackish-water farming
aquaculture systems defined in paper 1 through the use of indicators that provide
information on specific economic, social and ecological objectives. Our findings are
summarised in Table 24 below, that demonstrates that a number of trade-offs appear
to exist at the farm-level among these objectives.
Table 24 – Summary table of results from all indicators
Type 2 –
Type 3 – Type 4 – Type 5 –
Semi-
Type 1 – Low-input Large Semi-
intensive
Extensive labour- milkfish- intensive
prawn-
polyculture intensive oriented milkfish
oriented
systems systems monoculture
polyculture
Costs Average High Low Average Average

Revenues Average High Low Low Average

Gross margin Average High Low Low Average

Price Risk High High Low Low Low

Overall Risk Low High Average High High

Revenue High High Average Low Low

Diversity
Prawn High High Low High N/A
Mortality
Technical High Low Average High Average
Efficiency
Total Low Average High Low High
Employment
Effect
Net Dietary Positive Negative Positive Zero Negative
Protein
Production
Protein <1 >1 <1 <1 Approx. = 1
Conversion
Ratio
(input:output)
Partial Approx. = Approx. = 9 <1 <1 Zero
Ecological 2
Footprint
(ha
ecosystem /
ha fishpond)
Nitrogen loss Low High Low Average High
Pesticide Average Average Average Average Low
Application

Such a trade-off can be illustrated by the properties of farming system 2 (semi-

intensive prawn-oriented polyculture). These farms return high levels of profit on
average (as indicated by high gross margins), and have a positive but fairly average
potential on local employment. However, they perform poorly on a number of other
indicators, including overall and price risks (high), technical efficiency (low), net

133
 Paper 2/5

dietary protein production (negative), ecological footprint (large), and rate of nutrient
loss (high).

Another example corresponds to the low-input labour-intensive system (farm type 3).
These farms have high positive employment effect, a positive dietary protein
production, and low ecological footprints and nutrient loss, but return only low levels
of profit. Altogether, it is not possible to identify a preferred farming system, which
justifies the trade-off analysis developed in the next paper.

134
 Paper 2/5

References:
Adger, W. N., K. Brown, J. Fairbrass, A. Jordan, J. Paavola, S. Rosendo and G.
Seyfang (2003). "Governance for sustainability: Towards a 'thick' analysis of
environmental decisionmaking." Environment and Planning A 35(6): 1095-
1110.

Ahmed, M., G. A. Magnayon-Umali, R. A. Valmonte-Santos, J. Toledo, N. Lopez and

F. Torres Jr. (2001). Bangus fry resource assessment in the philippines.
Manila, International Center for Living Aquatic Resource Management
(ICLARM).

Aigner, D. J., C. A. K. Lovell and P. Schmidt (1977). "Formulation and specification

of stochastic frontier production function models." Journal of Econometrics
6(1): 21-37.

Anon (2005). Philippine minimum national social data set (mnsds). Manila, National
Statistics Coordination Board (NSCB): www.nscb.gov.ph accessed on 3rd
March 2005.

Aspuria, T. G. and R. M. Fabro (1979). A socio-economic survey of the aquaculture

industry in bicol. Socio-economic survey of the aquaculture industry in the
Philippines, Research paper series no. 20. Iloilo City, Philippines, Southeast
Asian Fisheries Development Center, and the Philippine Council for
Agriculture and Resources Research, Los Banos, Philippines: 60.

Bagarinao, T. (1994). "Systematics, distribution, genetics and life-history of milkfish,

chanos-chanos." Environmental Biology of Fishes 39(1): 23-41.

Bailey, A. P., W. D. Basford, N. Penlington, J. R. Park, J. D. H. Keatinge, T. Rehman,

R. B. Tranter and C. M. Yates (2003). "A comparison of energy use in
conventional and integrated arable farming systems in the uk." Agriculture,
Ecosystems and Environment 97: 241-253.

Bailey, C. (1982). Small-scale fisherise of san miguel bay, philippines: Occupational

and geographic mobility. ICLARM Technical reports 10. Manila, Philippines,
Institute of Fisheries Development and Research, College of Fisheries,
University of the Philippines in the Visayas, Quezon City, Philippines;
International Center for Living Aquatic Resources Management, Manila,
Philippines; and United Nations University, Tokyo, Japan: 57.

Barbier, E. B. (2000). "Valuing the environment as input: Review of applications to

mangrove-fishery linkages." Ecological Economics 35: 47-61.

Barrow, M. and M. Hall (1995). "The impact of a large multinational organization on

a small local economy." 29(7): 635-653.

Bartelmus, P. (1999). "Green accounting for a sustainable economy: Policy use and
analysis of environmental accounts in the philippines." Ecological Economics
29: 155-170.

135
 Paper 2/5

Bartley, D. M., K. Rana and A. J. Immink (2000). "The use of inter-specific hybrids
in aquaculture and fisheries." Reviews in Fish Biology and Fisheries 10(3):
325-337.

Battese, G. E. and T. Coelli (1995). "A model of technical efficiency effects in a

stochastic frontier production function for panel data." Empirical Economics
20: 325-32.

Battese, G. E. and G. S. Corra (1977). "Estimation of a production frontier model:

With application to the pastoral zone of eastern australia." Australian Journal
of Agricultural Economics 21: 169-179.

Bell, S. and S. Morse (1999). Sustainability indicators: Measuring the immeasurable.

London, Earthscan.

Bell, F. W. and E. R. Canterbery (1976). Aquaculture for the developing countries: A

feasibility study. Cambridge, Massachusetts, Ballinger Publishing Co.

Bhende, M. J. and J. V. Venkataram (1994). "Impact of diversification on household

income and risk - a whole-farm modeling approach." Agricultural Systems
44(3): 301-312.

Bilio, M., H. Rosenthal and C. J. Sindermann, Eds. (1981). Realism in aquaculture:

Achievements, constraints, perspectives. Bredene, Belgium, European
Aquaculture Society.

Bockstaller, C. and P. Girardin (2003). "How to validate environmental indicators."

Agricultural Systems 76: 639-653.

Bravo-Ureta, B. E. and A. E. Pinheiro (1997). "Technical, economic and allocative

efficiency in peasant farming: Evidence from the domincan republic." The
Developing Economies XXXV(1): 48-67.

Brennan, D., H. Clayton and T. T. Be (2000). "Economic characteristics of extensive

shrimp farms in the mekong delta." Aquaculture Economics & Management
4(3/4): 127-139.

Brown, J. H. (1989). "Antibiotics: Their use and abuse in aquaculture." World

Aquaculture 20(2).

Brown, K., W. N. Adger, E. Tompkins, P. Bacon, D. Shim and K. Young (2001).

"Trade-off analysis for marine protected area management." Ecological
Economics 37: 417-434.

Bunting, S. W. (2001). "Appropriation of environmental goods and services by

aquaculture: A reassessment employing the ecological footprint methodology
and implications for horizontal integration." Aquaculture Research 32(7): 605-
609.

136
 Paper 2/5

Chiang, F.-S., C.-H. Sun and J.-M. Yu (2004). "Technical efficiency analysis of
milkfish (chanos chanos) production in taiwan - an application of the
stochastic frontier production function." Aquaculture 230: 99-116.

Christensen, V. and D. Pauly (1992). "Ecopath ii - a software for balancing steady-

state ecosystem models and calculating network characteristics." Ecological
Modelling 61: 169-185.

Clayton, A. M. H. and N. J. Radcliffe (1996). Sustainability: A systems approach.

London, Earthscan Publications Ltd.

Coelli, T. (1996). A guide to frontier version 4.1: A computer program for stochastic
production and cost function estimation. Armidale, Australia, Centre for
Efficiency and Productivity Analysis: 34.

Coelli, T., G. E. Battese and D. S. Prasada Rao (1998). An introduction to efficiency

and productivity analysis. Boston, Kluwer Academic Publishers.

Corre, V. L. J., R. L. Janeo, C. M. A. Caipang and A. T. Calpe (1999). Sustainable

shrimp culture techniques: Use of probiotics and reservoirs with "green
water". Los Banos, Philippine Council for Aquatic and Marine Research and
Development: 32.

Coxhead, I. (2000). "Consequences of a food security strategy for economic welfare,

income distribution and land degradation: The philippine case." World
Development 28(1): 111-128.

Dalsgaard, J. P. T., C. Lightfoot and V. Christensen (1995). "Towards quantification

of ecological sustainability in farming systems-analysis." Ecological
Engineering 4(3): 181-189.

Dalsgaard, J. P. T. and R. T. Oficial (1997). "A quantitative approach for assessing

the productive performance and ecological contributions of smallholder
farms." Agricultural Systems 55(4): 503-533.

Deutsch, L., A. Jansson, M. Troell, P. Ronnback, C. Folke and N. Kautsky (2000).

"The 'ecological footprint': Communicating human dependence on nature's
work." 32(3): 351-355.

Dey, M. M., F. J. Paraguas, G. B. Bimbao and P. B. Regaspi (2000). "Technical

efficiency of tilapia growout pond operations in the philippines." Aquaculture
Economics & Management 4(1/2): 33-47.

Farrell, M. (1957). "The measurement of productivity efficiency." Journal of the royal

statistics society Series A, 120(3): 253-290.

Feder, G. (1979). "Pesticides, information and pest management under uncertainty."

American Journal of Agricultural Economics 61: 97-103.

137
 Paper 2/5

Folke, C., N. Kautsky, H. Berg, A. Jansson and M. Troell (1998). "The ecological
footprint concept for sustainable seafood production: A review." Ecological
Applications 8(1): S63-S71.

Frankenberger, T. R. (2002). A livelihood analysis of shrimp fry collectors in

bangladesh: Future prospects in relation to a wild fry collection ban. Dhaka,
Bangladesh, Department for International Development: 36 + xv.

Funge-Smith, S. and M. R. P. Briggs (1998). "Nutrient budgets in intensive shrimp

ponds: Implications for sustainability." Aquaculture 164: 117-133.

Giampietro, M. (1997). "Socioeconomic pressure, demographic pressure,

environmental loading and technological changes in agriculture." Agriculture,
Ecosystems and Environment 65(3): 201-229.

Giampietro, M., (2004). "Multi-Scale Integrated Analysis of Agroecosystems: a

complex system approach". CRC Press, Boca Raton, FL; London.

Gilbert, A. J. and R. Janssen (1998). "Use of environmental functions to communicate

the values of a mangrove ecosystem under different management regimes."
Ecological Economics 25(3): 323-346.

Girardin, P., C. Bockstaller and H. Van der Werf (1999). "Indicators: Tools to
evaluate the environmental impacts of farming systems." Journal of
Sustainable Agriculture 13(4): 5-21.

Handy, R. D. and M. G. Poxton (1993). "Nitrogen pollution in mariculture: Toxicity

and excretion of nitrogenous compounds by marine fish." Reviews in Fish
Biology and Fisheries 3: 205-241.

Hayami, Y. and Ruttan, V. (1985) "Agricultural development: an international

perspective." Baltimore, MD: John Hopkins University Press, 2nd Edition, 506
p.

Heller, M. C. and G. A. Keoleian (2003). "Assessing the sustainability of the us food

system: A life cycle perspective." Agricultural Systems 76(3): 1007-1041.

Holmer, M., C. M. Duarte, A. Heilskov, B. Olesen and J. Terrados (2003).

"Biogeochemical conditions in sediments enriched by organic matter from net-
pen fish farms in the bolinao area, philippines." Marine Pollution Bulletin 46:
1470-1479.

Irz, X., L. Lin, C. Thirtle and S. Wiggins (2001). "Agricultural productivity growth
and poverty alleviation." Development Policy Review 19(4): 449-466.

Irz, X. and V. McKenzie (2003). "Profitability and technical efficiency of aquaculture

production systems in pampanga, philippines." Aquaculture Economics &
Management 7: 195-211.

138
 Paper 2/5

Islam, M. S., M. J. Sarker, T. Yamamoto, M. A. Wahab and M. Tanaka (2004).

"Water and sediment quality, partial mass budget and effluent n loading in
coastal brackishwater shrimp farms in bangladesh." Marine Pollution Bulletin
48(5-6): 471-485.

Janssen, B. H. (1999). Basics of budgets, buffers and balances of nutrients in relation

to sustainability of agroecosystems. Nutrient disequilibria in agroecosystems:
Concepts and case studies. E. M. A. Smaling, O. Ooenema and L. O. Fresco.
Wallingford, Oxon, CABI Publishing.

Janssen, R., A. Gilbert and J. Padilla (2000). "Illuminating the need for ecological
knowledge in economic valuation of mangroves under different management
regimes - a critique - reply to ronnbach and primavera." Ecological Economics
35(2): 141-143.

Janssen, R., A. Gilbert and J. Padilla (2001). "Use of environmental functions to

communicate the values of a mangrove ecosystem under different
management regimes. Response to a critique (vol 35, pg 141, 2000)."
Ecological Economics 36(2): 359-359.

Just, R. E. and R. D. Pope (2001). The agricultural producer: Theory and statistical
measurement. Handbook of agricultural economics. B. Gardner and G.
Rausser. Amsterdam, Elsevier. 1A.

Kaplowitz, M. D. (2001). "Assessing mangrove products and services at the local

level: The use of focus groups and individual interviews." Landscape and
Urban Planning 56(1-2): 53-60.

Kautsky, N., H. Berg, C. Folke, J. Larsson and M. Troell (1997). "Ecological footprint
for assessment of resource use and development limitations in shrimp and
tilapia aquaculture." Aquaculture Research 28: 753-766.

Kautsky, N., P. Ronnback, M. Tedengren and M. Troell (2000). "Ecosystem

perspectives on management of disease in shrimp pond farming." Aquaculture
191(1-3): 145-161.

King, C., J. Gunton, D. Freebairn, J. Coutts and I. Webb (2000). "The sustainability
indicator industry: Where to from here? A focus strudy to explore the potential
of farmer participation in the development of indicators." Australian Journal of
Experimental Agriculture 40: 631-642.

Kodde, D. A. and F. C. Palm (1986). "Wald criteria for jointly testing equality and
inequality restrictions." Econometrica 54(5): 1243-1248.

Kopp, R. J. (1981). "The measurement of productive efficiency: A reconsideration."

The Quarterly Journal of Economics: 477-503.

Kumbhakar, S. C. (2002). "Specification and estimation of production risk, risk

preferences and technical efficiency." American Journal of Agricultural
Economics 84(1): 8-22.

139
 Paper 2/5

Kumbhakar, S. C. and C. A. K. Lovell (2003). Stochastic frontier analysis.

Cambridge, Cambridge University Press.

Kuznets, S. (1966). Modern Economic Growth: Rate Structure and Spread. New
Haven CT, Yale University Press.

Larsson, J., C. Folke and N. Kautsky (1994). "Ecological limitations and

appropriation of ecosystem support by shrimp farming in columbia."
Environmental Management 18(5): 663-676.

Leontief, W. (1936). "Quantitative input-output relations in the economic system of

the united states." Review of Economics and Statistics 18(3).

Leontief, W. (1966). Input-output economics. New York, Oxford University Press.

Ley, J. A., C. C. Melvor and C. L. Montague (1999). "Fishes in mangrove prop-root

habitats of northeastern florida bay: Distinct assemblages across an estuarine
gradient." Estuarine, Coastal and Shelf Science 48: 701-723.

Longhurst, A., S. Sathyendranath, T. Platt and C. Caverhill (1995). "An estimate of

global primary production in the ocean from satellite radiometer data." Journal
of Plankton Research 17(6): 1245-1271.

Martin, J. L. M., Y. Veran, O. Guerlorget and D. Pham (1998). "Shrimp rearing:

Stocking density, growth, impact on sediment, waste output and their
relationships studied through the nitrogen budget in rearing ponds."
Aquaculture 164: 135-149.

Maud, J., G. Edwards-Jones and F. Quin (2001). "Comparative evaluation of pesticide

risk indices for policy development and assessment in the united kingdom."
Agriculture Ecosystems & Environment 86(1): 59-73.

Merrington, G., L. Winder, R. Parkinson and M. Redman (2002). Agricultural

pollution: Environmental problems and practical solutions. London, Spon
Press.

Milon, J. W. (1987). The science and art of efficiency analysis: The role of other
performance criteria. Economic efficiency in agricultural and food marketing.
R. L. Kilmer and W. J. Armbruster. Ames, Iowa State University Press.

Naylor, R. L., R. J. Goldburg, J. H. Primavera, N. Kautsky, M. C. M. Beveridge, J.

Clay, C. Folke, J. Lubchenco, H. Mooney and M. Troell (2000). "Effect of
aquaculture on world fish supplies." Nature 405(6790): 1017-1024.

Nickerson, D. J. (1999). "Trade-offs of mangrove area development in the

philippines." Ecological Economics 28(2): 279-298.

140
 Paper 2/5

Pannell, D. J. and N. A. Glenn (2000). "A framework for the economic evaluation and
selection of sustainability indicators in agriculture." Ecological Economics 33:
135-149.

Papatryphon, E., J. Petit, S. J. Kaushik and H. M. G. van der Werf (2004).

"Environmental impact assessment of salmonid feeds using life cycle
assessment (lca)." Ambio 33(6): 316-323.

Pauly, D. and V. Christensen (1995). "Primary production required to sustain global

fisheries." Nature 374: 255-257.

Prein, M., C. Lightfoot and R. S. V. Pullin (1998). Iclarm's approach to the integration
of aquaculture into sustainable farming systems. Aquaculture sustainability
and environment: Report on a regional study and workshop on aquaculture
sustainability and environment. ADB/NACA. Bangkok, Thailand, Asian
Development Bank and Network of Aquaculture Centres in Asia-Pacific: 257.

Primavera, J. H. (1995). "Mangroves and brackishwater pond culture in the

philippines." Hydrobiologia 295(1-3): 303-309.

Primavera, J. H. (1998). "Mangroves as nurseries: Shrimp populations in mangrove

and non-mangrove habitats." Estuarine, Coastal and Shelf Science 46: 457-
464.

Primavera, J. H., C. R. Lavillapitogo, J. M. Ladja and M. R. Delapena (1993). "A

survey of chemical and biological products used in intensive prawn farms in
the philippines." Marine Pollution Bulletin 26(1): 35-40.

Rasul, G. and G. B. Thapa (2003). "Sustainability analysis of ecological and

conventional agricultural systems in bangladesh." World Development 31(10):
1721-1741.

Rees, W. and M. Wackernagel (1992). Ecological footprints and appropriated

carrying capacity: Measuring the natural capital requirements of the human
economy. Second Meeting of the International Society for Ecological
Economics, Stockholm, Sweden.

Roberts, R. J. and J. F. Muir (1997). 25 years of world aquaculture: Sustainability, a

global problem. Sustainable aquaculture. J. E. Bardach. New York, John
Wiley & Sons, Inc.: 167 - 181.

Ronnback, P. and J. H. Primavera (2000). "Illuminating the need for ecological

knowledge in economic valuation of mangroves under different management
regimes - a critique." Ecological Economics 35(2): 135-141.

Ronnback, P., M. Troell, N. Kautsky and J. H. Primavera (1999). "Distribution pattern

of shrimps and fish among avicennia and rhizophora microhabitats in the
pagbilao mangroves, philippines." Estuarine, Coastal and Shelf Science 48:
223-234.

141
 Paper 2/5

Ronnback, P., M. Troell, T. Zetterstrom and D. E. Babu (2003). "Mangrove

dependence and socio-economic concerns in shrimp hatcheries of andhra
pradesh, india." 30(4): 344-352.

Schmidt, P. (1985-86). "Frontier production functions." Econometric reviews 4(2):

289-328.

Schultz, T. W. (1964) Transforming Traditional Agriculture. New Haven CT, Yale

University Press.

Seligson, M. A. (1982). Peasant Participation in Costa Rica‟s Agrarian Reform: A

View from Below. Ithaca, N.Y.: Rural Development Committee, Cornell
University Center for International Studies.

Shang, Y. C., P. Leung and B.-H. Ling (1998). "Comparative economics of shrimp
farming in asia." Aquaculture 164: 183-200.

Sharma, K. R. (1999). "Technical efficiency of carp production in pakistan."

Aquaculture Economics & Management 3(2): 131-141.

Sharma, K. R. and P. Leung (1998). "Technical efficiency of carp production in

nepal: An application of stochastic frontier production function approach."
Aquaculture Economics & Management 2(3): 129-140.

Sharma, K. R. and P. Leung (2003). "A review of production frontier analysis for
aquaculture management." Aquaculture Economics & Management 7(1/2):
15-34.

Siar, S. B. (2002). Whose water? Whose space?: Issues on resource use in

aquaculture. Socioeconomic and policy issues: The Aquaculture sector,
Makati, Manila, Philippines.

Siar, S. V. and L. M. Caneba (1998). "Women and the question of sustainable

development in a philippine fishing village." International Journal of
Sustainable Development and World Ecology 5(1): 51-58.

Springsteen, F. J. and F. M. Leobrera (1986). Shells of the Philippines. Manila, Carfel

Seashell Museum.

Squires, D. and S. Tabor. 1991. “Technical Efficiency and Future Production Gains in
Indonesian Agriculture.” Developing Economies 24 (3): 258–70.

Stanley, D. L. (2003). "The economic impact of mariculture on a small regional

economy." World Development 31(1): 191-210.

Stevenson, J. R. (2002). The benefits to fisheries of uk intertidal salt marsh areas. R &
D Technical Report E2-061. London, Environment Agency: 35.

Stevenson, R.E., (1980). "Likelihood Functions for Generalized Stochastic Frontier

Estimation", Journal of Econometrics, 13: 343-366.

142
 Paper 2/5

Stoorvogel, J. J., J. M. Antle, C. C. Crissman and W. Bowen (2004). "The trade-off

analysis model: Integrated bio-physical and economic modelling of
agricultural production systems." Agricultural Systems 80: 43-66.

Sumagaysay-Chavoso, N. S. and M. L. San Diego-McGlone (2003). "Water quality

and holding capacity of intensive and semi- intensive milkfish (chanos chanos)
ponds." Aquaculture 219(1-4): 413-429.

Thakur, D. P. and C. K. Lin (2003). "Water quality and nutrient budget in closed
shrimp (penaeus monodon) culture systems." Aquacultural Engineering 27(3):
159-176.

Tovar, A., C. Moreno, M. P. Manuel-Vez and M. Garcia-Vargas (2000).

"Environmental implications of intensive marine aquaculture in earthen
ponds." Marine Pollution Bulletin 40(11): 981-988.

Tyedmers, P. H. (2000). Salmon and sustainability: The biophysical cost of producing

salmon through the commercial salmon fishery and the intensive salmon
culture industry. Resource Management and Environmental Studies.
Vancouver, British Columbia, Canada, University of British Columbia: 255.

van den Bergh, J. and H. Verbruggen (1999). "Spatial sustainability, trade and
indicators: An evaluation of the 'ecological footprint'." Ecological Economics
29(1): 61-72.

van der Werf, H. M. G. and J. Petit (2002). "Evaluation of the environmental impact
of agriculture at the farm level: A comparison and analysis of 12 indicator-
based methods." Agriculture, Ecosystems and Environment 93: 131-145.

Wackernagel, M. and W. Rees (1996). Our ecological footprint: Reducing human

impact on the earth. Gabriola, British Columbia, New Society Publishers.

Wijnhoud, J. D., Y. Konboon and R. D. B. Lefroy (2003). "Nutrient budgets:

Sustainability assessment of rainfed lowland rice-based systems in northeast
thailand." Agriculture Ecosystems & Environment 100(2-3): 119-127.

Williams, M. J. (1997). "Aquaculture and sustainable food security in the developing

world". Sustainable aquaculture. J. E. Bardach. New York, John Wiley & Sons
Inc.

World Bank (2002). "Shrimp farming and the environment: Can shrimp farming be
undertaken sustainably? A discussion paper designed to assist in the
development of sustainable shrimp aquaculture". Washington D.C., World
Bank.

Xu, Z. K., J. H. Primavera, L. D. de la Pena, P. Pettit, J. Belak and A. Alcivar-Warren

(2001). "Genetic diversity of wild and cultured black tiger shrimp (penaeus
monodon) in the philippines using microsatellites." Aquaculture 199(1-2): 13-
40.

143
 Paper 2/5

Yap, W. G., A. C. Villaluz and R. F. Agbayani (2003). Feasibility of new shrimp

farming practices in the philippines. Makati, Manila, Philippines, Board of
Investments, The Republic of the Philippines: 41.

Zenger, S. and B. Schurle (1981). "The impact of diversification on farm risk."

American Journal of Agricultural Economics 63(5): 1043-1043.

144
 Paper 2/5

Appendix 1

a) Wild-fry gatherers
Calculated using data from Ahmed et al (2001) from five gathering sites and from
interviews with 42 gatherers. Thus n=42. The fry season is split into two periods –
Peak season for 4 months March-June, and off-peak season for 8 months July-
February.
Daily production
Number of Hrs gathering (fry per gatherer
gatherers / day per day) % survival Live fry per hr
6 6.8 950 94 131.3
10 5.5 960 98 171.1
Peak 16 5.5 1136 96 198.3
1 3 150 95 47.5
9 6.3 1744 98 271.3
6 1.2 68 99 56.1
10 1.2 67 99 55.3
Off-peak 16 1 109 98 106.8
9 1 90 98 88.2

Peak season mean productivity = 194 fry/hr

Off-peak season mean productivity = 83 fry/hr
Assuming an 8-hour working day, and allowing for the seasonal variation in
productivity, we calculate (194*8*17/52)+(83*8*35/52) = 954 fry / person day

b) Small shells
From the results of interviews with four teams of collectors of small shells in
Sasmuan, Pampanga, we estimated 40 “coolers” of shells per two-person team in a 5-
hour collecting period. Each “cooler” (a large polystyrene container) contains 30 kilos
of shells. Thus, an average of 600 kilos are produced per person per 5-hour working
day, which is 960 kilos per 8-hour day, corresponding to 1.04 person days per metric
tonne of shells.

145
 Paper 2/5

c) Commercial feeds
A visit to a well-known commercial feed-mill in region 3 gathered the following data.
The feed-mill operates 24 hrs, with workers on 3-shifts of 8 hours each. The number
of people employed at the factory is 155, all full-time. Output is 5.5 tonnes per hour,
therefore, we estimate:
155 / (5.5*24) = 1.17 person days / tonne

d) Hatchery fry
A visit to one of the hatcheries in Zambales province that provides fry to fishponds in
region 3, gathered the data. The hatchery has 4 “runs” per year and each run produces
an average of 10 million prawn fry. There are 6 full-time employees. Thus we
estimate:
6 / 40 = 0.15 FTE / million fry.
Assuming 300 person days / year for FTE workers, we therefore estimate:
0.15*300/1000 = 0.45 person days per thousand fry.

e) Fingerlings
The coefficient for fingerlings were calculated by adding an additional labour-input to
the hatchery fry estimates, to account for the period when the fry is nursed through the
first few weeks of its life-cycle, and correcting for mortality. On the basis of three
interviews with fingerling producers, we obtained the estimate of 7.1 person days /
thousand fingerlings. However, this estimate has a high degree of error associated
with it, owing to vastly different coefficients for the three operators.

146
 Paper 2/5

Endnotes

1
The details of the sampling procedure used in this survey are given in paper 1, but we draw on some
summary statistics from this survey here.
2
The Gini coefficient is the summary statistic most commonly reported to measure inequality at the
country level (Stiglitz, 1993). Gini coefficients are bound between 1 and 0, where 1 represents perfect
inequality (in this case, all the land is operated by one person) and 0 represents perfect equality (all
farms are the same size).
3
Alternatively, the technical efficiency relation can be described with respect to output, i.e. achieve
the maximum possible output for a given level of inputs.
4
Except for farm type 5, which is necessary in order to avoid singularity problems. Hence, farm type
5 is chosen as a reference to which efficiency levels of other farm types are compared.
5
From Kodde and Palm (1986).
6
This is an imputed value based on market prices.
7
Personal observation; Jurgenne Primavera, personal communication.
8
As should be clear from previous discussion, the collected shells are used as feed in aquaculture.
9
Possible identifications include Dosinia (Phacosoma) troschel or Katelysia hiatina for “Gasang”;
Telescopium telescopium or Clypeomorus coralium for “Suso” (Springsteen and Leobrera, 1986)

147
 Paper 3/5

Research project R8288: Assessing the sustainability

of brackish water aquaculture systems in the
Philippines – Paper 3/5

Analysing trade-offs among indicators of

sustainability: a multi-criteria approach

Abstract: A number of recent papers in agricultural economics and related literatures

have considered the trade-offs associated with particular development projects or
programmes. We briefly review this literature and propose a methodological
framework for analysing sustainability at the farm-level using subjective, local
weights and objective, positive sustainability indicators. More specifically, this paper
uses the measures of six sustainability indicators to examine the relative merits of five
aquaculture pond farming systems in the Philippines. We employ a discrete-choice
multi-criteria decision-making model that calculates the distance of alternative
farming systems to a utopian point – an unfeasible but approachable point where all
objective functions are maximised. We apply our analysis to a number of scenarios
using sets of weights from two barangays elicited during a series of focus group
discussions (FGDs) with particular stakeholder groups.

We find that extensive polyculture is the preferred farming system under most
conditions, and that very large milkfish-oriented systems perform poorly in our
analyses. We interpret these findings with reference to qualitative data gathered
during FGDs and examine the implications with regard to the inclusion of
aquaculture under existing agrarian reform legislation – from which aquaculture has
been exempted for over 15 years.

145
 Paper 3/5

Contents

1. Introduction 147

2. Trade-off analysis 147

2.1 General framework and previous applications to agriculture,
natural resource management and aquaculture 147
2.2 Discrete choice pay-off matrix 148
2.3 Trade-off analysis between profit and other objectives 149

3. Compromise programming models 150

3.1 Introduction 150
3.2 The general vector-optimisation model 151
3.3 Non-dominance filter 152
3.4. Compromise model 154
3.5. Utopian and anti-ideal points of brackish-water aquaculture

4. Elicitation of the weights in the compromise programming model 155

4.1 Presentation of the communities 155
4.2 Eliciting importance coefficients from focus group discussions 157
4.3 Weights from the Barangay Council, Nanding Lopez 158
4.4 Weights from the Barangay Council, Sapang Kawayan 159
4.5 Weights from a group of men in Sapang Kawayan 162
4.6 Weights from a group of women in Sapang Kawayan 163

5. Compromise model results 166

5.1 All weights equal 166
5.2 Weights from the barangay council of Nanding Lopez 167
5.3 Weights from the barangay council of Sapang Kawayan 167
5.4 Weights from the group of men in Sapand Kawayan 168
5.5 Weights from the group of women in Sapang Kawayan 169

6. Overall findings 169

6.1 Implications – Local level 170
6.2. Implications – National level 171

146
 Paper 3/5

1. Introduction
We have thus far approached the analysis of sustainability, in papers 1 and 2, by
defining specific objectives in economic, social and ecological realms. We have
established that no single farming system maximizes all of these objectives. Thus, the
normative problem of choosing one system over another necessarily involves trade-
offs among objectives and this paper will present various ways of analysing such
trade-offs in a discrete choice setting. We start by reviewing applications of trade-off
analysis and multi-criteria decision making frameworks in agriculture, paying
particular attention to a technique called compromise programming. We go on to
describe participatory exercises in two case-study barangays, used to elicit the
weights, or “importance coefficients” (following Munda, 2004) that are necessary
for compromising among objectives. Rankings for the five farming systems are then
established using each set of weights and the implications of the results are finally
discussed.

2. Trade-off Analysis
2.1 General framework and previous applications to agriculture, natural resource
management and aquaculture

The idea of trade-off analysis is simple and relates to the negotiation of competing
objectives in a socio/political-economic manner. However, its translation into a
practical methodology takes a variety of forms. For instance, some authors consider
multiple values (in the societal sense), associated with achievement of each of the
objectives (e.g. Brown et al., 2001, Munda, 2004). That is, there is a recognition that
different groups of people have different priorities. Other authors, whose work still
falls under the general heading of trade-off analysis, present a purely technical
analysis of the opportunity costs (to the other objectives in the problem) of achieving
maximisation in any one objective (e.g. Bouman et al., 1998; Lu and van Ittersum,
2004; Stoorvogel et al., 2004). Another distinction relates to the general dichotomy
in the literature between what can be described as “algorithmic” and “soft”
approaches to conflict resolution (hereon compromising).

The algorithmic approach considers preferences to be modelled by the use of real-

valued weights (that sum to unity) and are founded on substantive rationality. That is,
the quality of the decision is the most important factor in determining the validity of
the methodology. Applications of this method can be found in Kobrich (1997) for
peasant farming systems in Chile, and El-Gayar and Leung (2001) who applied the
method to aquaculture in Northern Egypt. This latter case is of particular relevance,
and the authors considered three objectives in their model: total protein, foreign
exchange, and employment generated by aquaculture. The authors found that
freshwater supply and ricebran were limiting factors to the development of
aquaculture in the region, although policy implications for their findings are unclear.
The choice of an appropriate policy instrument to realise changes in the development
of the industry requires further study to include a firm-level model.

Softer approaches to compromising are based on a procedural rationality where the

quality of the process under which the decision is made is the most important factor by
which a methodology should be judged (Martinez-Alier et al., 1998; Roth, 1999). This
approach is typically founded on a discursive ethics (O'Hara, 1996) where deliberation
is held to be the ultimate democratic principle to which policies should abide (O'Neill,

147
 Paper 3/5

1997). Applications of this softer form of multi-criteria approach include Brown et al.
(2001) who sought compromises among stakeholders in the context of the
establishment of a marine protected area, and Gregory and Wellman (2001) who
consider alternative management plans for an estuary in Oregon, USA. In both of these
successful applications, we find that a dual emphasis on deliberation and analysis
provides significant benefits including generating a sense of local ownership of the
results.

Our analysis in this paper is quantitative, and there is an overall positivist epistemology
underlying it. However, in using focus groups and matrix scoring exercises in our case-
study barangays, there is a degree of geographic specificity to the analysis. The results
are thus not generalisable to the country-level, but in an important sense, this is the
kind of philosophy required to usefully carry out work with the concept of
sustainability. Local imperatives are important and we believe that the subjective
assignment of weights to a positivist set of sustainability indicators combines the
strengths of both algorithmic and soft approaches to trade-off analysis.

2.2 Discrete choice pay-off matrix

We now examine five models of aquaculture farming systems, based on the typology
exercise presented in paper 1. The decision problem takes each of these farm types as
discrete “options” for land-use in the brackish-water zones. Thus, we overlook the
determinants of within-type heterogeneity and examine the differences in mean
scores for each of six indicators 1 : profit, nutrient balance, employment, technical
efficiency, risk and net protein production. Our analysis then aims at establishing the
relative performance of these farming systems, as opposed to their absolute
performance as would be measured with reference to standards or thresholds.

The first step of the analyses involves the definition of a pay-off matrix, which, in
our discrete choice setting, is simply a 6x6 matrix of levels of the six indicators
mentioned above. When a given objective is maximised (across the five options) the
pay-off is the corresponding achievement of the other indicators for that farming
system. Reading the matrix, in Table 1 below, each of the row elements are
maximized in turn and the column elements correspond to the percentage
achievement of each objective for that farm type. For example, reading the first row,
when profit is maximized, nutrient balance achieves only 9 per cent of its maximum,
employment 62 per cent and so on. Further, from the presence of ones on non-
diagional positions in the table, it can be inferred that the nutrient balance and risk
objectives are maximised for the same farming system. The same applies to the
pairing of protein production and employment.

148
 Paper 3/5

Table 1 – Pay-off matrix for six objectives

Pay-off when z is maximised (%)
Nut Employ- Protein
Objective to maximise (z)

Profit Balance ment T.E. Risk Production

Profit 1.00 0.09 0.62 0.58 0.31 0.16
Nut balance 0.87 1.00 0.37 0.93 1.00 0.66
Employment 0.50 0.40 1.00 0.77 0.49 1.00
T.E. 0.47 0.17 0.38 1.00 0.34 0.48
Risk 0.87 1.00 0.37 0.93 1.00 0.66
Protein production 0.50 0.40 1.00 0.77 0.49 1.00

Thus, for nutrient balance versus risk (in row 2) and protein production versus
employment (in row 6), there is no trade-off. However, for other objectives, it is clear
that significant trade-offs exist. Of particular note is the relationship between profit
(the major motivating factor of operators) and the other objectives (which are
important to different stakeholder groups). This is the subject of the following
section.

2.3 Trade-off analysis between profit and other objectives

The pay-off matrix suggests that there are significant opportunity costs (i.e, in terms
of other objectives) to the maximisation of profit. In order to gain further insight into
the relationship between profit and these other objectives, we gradually relax (in 5%
intervals) the profit objective, from its maximum, as a constraint in the individual
maximization of the other objectives. Figure 1 presents the results of these
calculations.
Figure 1 – Trade-off analysis between profit and other indicators

Trade-off between profit and other indicators

120
% Achievement in other indicators

100
Nutrient
balance
80
Employment

60 Net protein
production
40 Risk

20 Technical
efficiency

0
100
90
80
70
60
50
40
30
20
10
0

% Maximum Profit

149
 Paper 3/5

As we move from left to right along the horizontal axis, acceptable profit, expressed
as a percentage of maximum profit, decreases. For each five percent decrease in
profit, the largest level of the other indicator, expressed as a percentage of its own
maximum, is calculated. Thus, at 100 percent profit, the values for the other
indicators in Figure 1 are the same as those reported in the top row of the pay-off
matrix, which corresponds to the results for farm type 2 (semi-intensive prawn-
oriented polyculture). As we reach 85 percent of maximum profit, there is a
significant increase in the feasible achievement of several other objectives (risk,
technical efficiency, nutrient balance, and protein production). These results are
achieved by farm type 1 (extensive polyculture). Thus we find that a relatively
marginal decrease in profit, from its maximum level, generates potentially significant
benefits in terms of the other objectives.

The finding of relatively large non-economic benefits of “nearly optimal” economic

solutions is consistent with that of Lu and van Ittersum (2004). Their application (to
policy objectives on the Loess plateau, China) is based on a continuous choice
problem, but the methodology is essentially the same – the discrete choice problem
can be considered a special case of their continuous choice example. Figure 1 also
shows that some indicators trade-off more significantly against profit than others –
employment and net protein production achieve their maximum feasible values on a
farming system (type 3, low-input labour-intensive systems) that has only 50 per cent
of maximum profit.

These findings demonstrate the relevance of employing multiple criteria in a decision

problem, as it shows the opportunity costs of a narrow focus on optimisation of a
single criterion. Using compromise programming models as a means to reconcile the
conflicting objectives, we can show how the relative merits of particular production
systems are dependent on the weights placed on each of these objectives.

3. Compromise Programming Models

3.1 Introduction
Compromise Programming refers to a family of methods that seek compromise
solutions to problems with multiple objectives. The central motif is an upper or lower
reference point, which is unfeasible (owing to conflicts among objectives) but
approachable. From this family of methods, the nomenclature of which lacks
standardisation, we use a weighted distance function method, where alternatives are
assessed according to their distance to an ideal or “utopian” point, where all
objectives are maximised. The choice of a utopian point based method lies on the
rationale that sustainable development is often considered a utopian concept, where
economic, social and ecological objectives are not conflicting.

Central to the approach is Zeleny’s axiom of choice which states that:

“Alternatives that are closer to the ideal are preferred to those that are farther away.
To be as close as possible to the perceived ideal is the rationale of human choice.”
(Zeleny, 1982, p. 156). This axiom can be considered an alternative to the basic
traditional structure underlying choice problems in economics, which can be
described in a simplified setting by:

150
 Paper 3/5

Max Z ( x1 , x2 ) (1)
( x1, x 2 )
subject to T(x1,x2) = k

where (x1,x2) represents a vector of choice variables for the decision-maker. This can
be a basket of commodities in consumer theory, or a vector of outputs in a joint
production problem (Ballestero and Romero, 1998). Z(x1,x2) is the utility function for
the decision-maker, and T ≤ k represents the feasible set. Thus, economic rationality
is usually defined in terms of maximising a consistent and transitive function such as
Z(x1,x2) subject to the satisfaction of the feasible set. The information required to
carry out such an analysis is formidable for an individual, and, as a result, economists
rarely deal with empirically elicited utility functions, and, even less frequently, with
empirical social utility functions (Ballestero and Romero, 1998). The connections
between the traditional utility paradigm and compromise programming are explored
in Ballestero and Romero (1991, 1994).

The exposition of the basic compromise model that we now present is based on
Glaser (2002). Note that the following two sections include continuous citations of
selected passages from this work between pages 23 and 73.

3.2 The general vector-optimisation model

The decision-maker considers K ≥ 2 objectives (criteria), with each individual
objective function zk having its own dimension [dk] (∀k ∈ K:={1,…..,K}), where K is
a so-called index set which contains the indices of all K objectives. Together they
form the vector-value objective function z(x) which is to be maximized. Thus, K
objectives are to be maximised simultaneously and we have the general vector
optimisation model (VOM):
max{z ( x) x ∈ X }
x

⎛ z1 ( x) ⎞
N K
⎜ ⎟
with z: ℜ → ℜ ; x a z(x) : = ⎜M ⎟ (2)
⎜ z ( x) ⎟
⎝ K ⎠
The objective function now returns a vector for each alternative (x) from the set of
feasible alternatives under consideration (X), and the feasible region of the objective
space of the VOM,
Z(VOM):={z(x) x ∈ X } ⊂ ℜ K (3)
which is a K-dimensional subset of ℜ K . Whereas for each individual objective
function zk (∀k ∈ K) a total order of X is still given, and thus, the individually
optimal objective-function value or individual optimum
⊗
zk := max{zk(x) x ∈ X } (∀k ∈ K) (4)
as well as the set of individually optimal solutions
⊗ ⊗
Xk :={x ∈ X z k (x) = zk }≠∅ (∀k ∈ K) (5)
exist for every objective, the same is generally not true for z, for it only implies a
partial order of X on ℜ K . The definition of an ordered set is a binary relation which
has certain properties (reflexivity, transitivity, anti-symmetry). In the case of a
partially ordered set, there may be elements that are incomparable, which is the case
with Z(VOM) . Only in the rare case that at least one alternative xperf exists which

151
 Paper 3/5

maximizes all objective functions at once 2 , can the decision problem (VOM) be
solved at this stage. All such alternatives
K
⊗
xperf ∈ Xperf:= ∩ Xk (6)
k =1

are called perfect solutions, since they attain the ideal objective vector or utopian
point
⊗ ⊗ ⊗
z :=(z1 ,…, zK )T (7)

Objectives partaking of this property are called complementary, because attaining the
optimum of one objective does not diminish the objective function values of the
others, i.e. there is no trade-off between objectives. However, the entire basis of the
MCDM paradigm is that problems usually exhibit conflicting objectives, or, in other
words, that they do not have any perfect solution (Xperf = ∅). In that situation, Glaser
(2002) recommends the following two-stage procedure to derive a solution :

• Filter – A check is made for non-dominance among the alternatives. Any

dominated alternative is simply discarded at this stage.

• Compromise – Reconcile the conflict in (VOM) by means of a compromise,

with the aim of finding the best compromise solution among the alternatives.

These two steps are explained in details and implemented to analyse the properties of
our fish farming systems in the following sections.

3.3 Non-dominance filter

Some alternatives in the compromise programming model may be Pareto-dominated
because the model does not feature any inherent efficiency component. Thus, in a
first step, it is necessary to carry out an initial check to eliminate the alternatives that
are dominated. In a discrete choice setting such as ours, this task is straightforward
and best performed visually by drawing radar diagrams, as presented in Figure 2. The
diagrams establish immediately that none of our alternatives is Pareto dominated by
others. Hence, there is no obviously ‘best’ and ‘worst’ farming system among the
five alternatives considered here. We note, in particular, that although type 5 (semi-
intensive milkfish monoculture) provides no maximal element, it could potentially
become an optimal choice under a particular combination of weights. Altogether,
application of the non-dominance filter does not allow us to eliminate any of the
alternatives, and Figure 2 makes clear that the choice of a particular farming system
for the brackish waters of the Philippines involves some difficult trade-offs, which
we analyse further through the use of a compromise model.

152
 Paper 3/5

Figure 2 - Radar plots of the six sustainability indicators by farming system. Each of
the six points measures one of the indicators, expressed as the percentage of the
individual optimum for this indicator. The following notation is used: 1=Profit,
2=Nutrient loss, 3=Employment, 4=Technical efficiency, 5=Risk, 6=Net protein
production

Type 1 - Extensive polyculture Type 2 - Semi-intensive prawn polyculture

1 1
1 1

6 0.5 2 6 0.5 2

0 0

5 3 5 3

4 4

Type 3 - Low-input, labour-intensive Type 4 - Large, milkfish oriented farms

1 1
1 1
6 0.5 2 6 0.5 2

0 0

5 3 5 3

4 4

Type 5 - Semi-intensive milkfish monoculture

1
1
6 0.5 2

5 3

153
 Paper 3/5

3.4 Compromise model

A compromise model (CM) can simply be described by the following mathematical
problem:
max{ψ ( z ( x)) x ∈ X } (8)
x

The transformation of the objective vector in (8) is referred to as scalarisation. The

additional preference information with regard to the vector of objective functions z,
as chosen by the decision maker, finds its expression in the scalar real-valued
compromise functional, which is a function of functions:
ψ: ℜ K → ℜ; z ( x) a ψ(z(x)) (9)

This compromise function is usually parametised. Our study uses methods that
require the choice of a vector of weights w and parameter p to complete the
specification of the distance function, as explained in details below. In that case, the
specific compromise function (ψ) can be either a weighted objective function
(otherwise known as weighted sum), or a weighted distance function, which is
actually a family of specific functions that use different p metrics. If the compromise
optimumψ ∗ := max{ψ ( z ( x)) x ∈ X } exists, then an optimal solution of (CM)

x∗ ∈ arg max{ψ ( z ( x)) x ∈ X } is called the best compromise solution with respect to

(CM). Further, z*:=z(x*) is a compromise-optimal objective-function vector.

In the current study, we employ a distance function that minimizes the distance of
objective-function vectors in Z(VOM) to the utopian point z ⊗ .

Next, we define a distance function d ∇ that fulfils the axioms of non-negativity,

identity, symmetry and sub-additivity. The Lp-norm with parameter p ≥ 1 is
employed to measure the distance of a vector zk to the utopian point z⊗. It takes the
following specific form:
⎧ K
(∑ wk (( z k − z k ( x)) /( z k − z k• )) p )1 / p
⊗ ⊗
⎪ 1≤ p < ∞
∇ ⊗
d p , w ( z , z ( x)) := ⎨ k =1 (10)
{ ⎪max w (( z ⊗ − z ( x)) /( z ⊗ − z • ))}
⎩ k k k k k k p=∞

Expression zk• defines the opposite of the utopian point, or anti-ideal point, which is
a vector of minimal elements. While some compromise methods use this anti-ideal
point as a Lower Reference Point (LRP), and aim to maximise the surplus distance
achieved above the LRP, here the argument ( zk⊗ − zk• ) is merely used as a way of
normalising distances. This is the normalisation procedure suggested in Ballestero
and Romero (1998, p.20) and the result is a dimensionless distance that is not
dependent on the unit of measurement. The method therefore overcomes the problem
of incommensurability of individual attributes (Zeleny, 1982) through the use of
relative, rather than absolute, deviations. Thus, apples and oranges are happily
considered in the same problem.

154
 Paper 3/5

When p=1, d ∇ gives the weighted absolute differences (sometimes referred to as the
Manhattan metric as it calculates distance in the same manner as someone navigating
city blocks). When p=2, d ∇ gives the familiar Euclidean distance. When p = ∞,
d ∇ gives the Tchebycheff norm, which is applied to the weighted absolute
differences. In both forms of function d ∇ in equation (10), w ≥ 0 and 1Tw = 1, and
these weights w are used to express the importance of each objective in the decision
problem.

3.5 Utopian and anti-ideal points for brackish-water aquaculture

The decision problem consists of x=5 alternatives (farming systems) and k=6
objectives (sustainability indicators), and the corresponding data can therefore be
conveniently summarized in a 5x6 matrix. The weights (wk) applied to each indicator
come from participatory exercises carried out in two barangays in the Philippines
between June and August, 2004. The process of eliciting these weights is detailed in
the following section and a number of scenarios for (CM) are then modelled, based
on the information gathered in these exercises.

The utopian and anti-ideal points are given in Table 2 below. The objectives which
are to be minimised (i.e. N loss and Risk index) are used in their inverse form in the
compromise model. Thus all objectives in the compromise model are to be
maximised and all distances to the ideal point are to be minimised. As we can see,
the utopian and anti-ideal points differ substantially, which gives meaning to the
analysis.

Table 2– Utopian and anti-ideal points in the compromise model

Total
Employment Net
Mean N loss effect in protein Technical
Profit in in person production Risk efficiency
PhP/ha/yr kg/ha/yr days/ha/yr in kg/ha/yr index estimate
(max) (min) (max) (max) (min) (max)
Utopian 59065 20.0 411.2 53.2 0.169 0.748
Anti-ideal 27845 231.9 150.7 8.6 0.546 0.437

4. Elicitation of the weights in the compromise programming model

4.1 Presentation of the communities
The Philippines has a highly devolved structure of governance. The first step on the
governance ladder is that of the barangay, which is derived from a pre-Hispanic
social unit of 30 to 100 families owing allegiance to one chief (McCoy and de Jesus,
1982). In contemporary Philippines, the barangay remains an important social
construction and source of identity. Each barangay has an elected captain and
council that meet to discuss issues such as policing, peace-keeping, and applications
for infrastructure projects from the municipality and higher levels of government.
Owing to population growth, this social entity is now often as large as 500 families, a
register of which is kept in some form by the barangay secretary. When a barangay
becomes unmanageably large in population terms, it may be split in two.

The barangay was chosen as the unit of analysis for the current study and two were
selected for detailed study, one from region 3 (Barangay Sapang Kawayan, Masantol,

155
 Paper 3/5

Pampanga) and one from region 6 (Barangay Nanding Lopez, Dumangas, Iloilo).
Both barangays are remote and their landscapes are dominated by fishponds. Brief
descriptions of both barangays follow (see more details in paper 4):

• Barangay Sapang Kawayan (SK) is a rural community under the jurisdiction

of the municipality of Masantol in the Pampanga province of Region 3.
Masantol is classified as ‘partially urban’ and has 32,464 registered voters
(National Statistical Coordination Board, 2004). The community of SK is
located to the South of Masantol proper and is only accessible by boat (one
hour from Masantol and 20 minutes from another town, Haganoy, in the
nearby province of Bulacan). It is surrounded by fishponds and river systems
and has a total land area of 265 hectares (Provincial Agriculturalist’s Office),
which supports a population of 2,676 individuals in 559 households (2000
National Statistics Census).

• Barangay Nanding Lopez (NL) is located in the municipality of Dumangas,

province of Iloilo (Region 6). Three major rivers traverse the area and the
barangay is almost bounded by water except for a strip of land that connects
it to the rest of the municipality. About 97 per cent of the 797 hectares of land
in the barangay are occupied by fish ponds and discussions with key
informants suggested that aquaculture and fishing represented the two main
sources of livelihood for the population of 1,359 individuals (as of 2003).

In paper 4, we estimate the extent of poverty in five coastal barangays in the

Philippines, including those studied here. We find the extent of poverty as measured
by the common head-count index to be high in both Sapang Kawayan (59%) and
Nanding Lopez (44%) (see Table 3 below). The “poverty gap” measure, which
considers the severity of poverty by taking into account the distance of each
household to the poverty line, further indicates that the depth of poverty is fairly
similar in the two communities. Finally, the head count index measured at the food
threshold reveal a high incidence of extreme poverty in both barangays (35% in SK
and 36% in NL).

Table 3 – Aggregate income per capita and poverty incidence in case-study

barangays
Income per capita (PhP/yr) Poverty Incidence
Poverty Head
Head
Barangay Mean Median S.D. Gap Count
Count
(Food)
SK 17,214 14,633 12,134 0.59 0.26 0.35
NL 18,381 17,974 11,639 0.44 0.21 0.36

156
 Paper 3/5

4.2 Eliciting importance coefficients from focus group discussions

Members of the two barangays were invited as participants in a series of focus group
discussions (FGDs). This was arranged informally through two research assistants
staying in the barangays for an extended period (at least one month): Portia
Villarante (with help from Maggie Babay) in Sapang Kawayan, and Arnold Tanoy in
Nanding Lopez. It was possible for the research assistants to become familiar with
the community and its members as a result of their stay and to gather additional
information based on the anthropological nexus of immersion, participation and
recording. Three groups were identified as being representative of a cross-section of
the interests in each barangay:

• Barangay council
• A group of men, with various livelihoods
• A group of women, with various livelihoods

The barangay council are an elected body and thus have a remit to represent the
interests of the community. However, the positions on the council may be
disproportionately comprised of those from higher social classes, which provides the
rationale for considering separate groups of men and women with diverse interests
and no formal access to power. The genders were separated to allow for gendered
perspectives to emerge and to prevent domination of the discussion by the men in
the group. For Nanding Lopez, only the results of the barangay council FGD are
available for inclusion in this analysis owing to a problem with the research assistant
working in that community.

Each FGD was comprised of between 8 and 12 members, and had the following
structure:
1. Introductions
2. Background information on the research project given to the group
3. Discussion of the positive and negative aspects of aquaculture in the area
4. Description of indicators by the research assistant to the group
5. Choice by the group of the 3 most important indicators
6. Matrix scoring of the chosen indicators by nominated representatives
7. Justification for the choice
8. Appeals / Changes by other members
9. Acceptance / Vote of the final result by the group
10. Snack and Soft Drinks
11. Thanks

The discussion in step (3) and throughout the course of the events were recorded and
transcribed by the research assistants. Interesting quotes are used in the following
sections to illustrate the group’s rationale behind the choice of weights. The matrix
scoring problem presented to the group in (5) and (6) consisted of first choosing
which three indicators were most important, and then allocating 20 stones between 3
indicators to reflect their relative importance. This process is similar in spirit to the
rapid rural appraisal (RRA) techniques pioneered by Robert Chambers (e.g.
Chambers, 1997).

157
 Paper 3/5

RRA is essentially applied common-sense in data collection, with the imperative of

expression by the people of the full complexity of their situation. RRA is extractive,
in the sense that an external research team are trying to understand a local situation
and take the findings ‘away’ for analysis. This is in comparison to the recent
development in the literature of what is described as a “participation imperative”
(Sumberg et al., 2003) whereby the expectation is for research groups to empower
and facilitate action by local groups and individuals in directly improving their
situtations. Thus RRA has evolved into PRA – participatory rural appraisal.
Whatever their divergence may be in terms of epistemology, there is a shared,
implicit code of conduct for research groups in carrying out what might be
potentially divisive, intrusive activities. As Chambers writes:

“Empirically…the recurrent finding with PRA has been that if the initial behaviour
and attitudes of outsiders are relaxed and right, and if the process can start, the
methods of PRA themselves foster further rapport. Early actions by outsiders can
include transparent honesty about who they are and what they are doing; and
participation in local activities, especially being taught and performing local tasks.
Personal demeanour counts, showing humility, respect, patience and interest in what
people have to say and show; wandering around and not rushing; and paying
attention, listening, watching and not interupting. Having confidence that ‘they can
do it’ and transmitting that confidence, again and again enables local people to get
started with activities like participatory mapping, diagramming or matrix scoring.
Then they quickly lose themselves in the activity and are often pleased and proud of
what they find they know and can do.” (p.134 Chambers, 1997).

In most cases, the group immediately decided on the first one or two indicators that
should be chosen, with some debate between members about the final choice.
Different members were able to reallocate the stones to the three indicators if there
was disagreement on the weights, and in some cases the issue was put to a vote.
Given that 1Tw=1 and wk∈[0,1], then each stone can be interpreted as representing a
relative weight of 5 per cent.

4.3 Weights from the Barangay Council, Nanding Lopez

4.3.1 Focus Group Discussion
The group consisted of eight members of the barangay council: the barangay
captain, six councillors and the barangay secretary. The main sources of livelihood
for the participants, alongside stipends and privileges associated with their position
on the council, were: the farming of a quarter hectare of ricefield; the operation of
one hectare of fishpond; river fishing; the sale of medical supplies; tricycle driving
(own vehicle); and ‘sari-sari’ storekeeping (two instances). The group also included
the housewife of a fisherman.

The initial discussion provided the following main insights:

• Absentee operators bring in their own labourers from the city (Iloilo) so that
the benefits from aquaculture in those cases do not accrue locally. The group
would like more family-oriented aquaculture in which the operator lives on
the farm and uses family and local labour.

158
 Paper 3/5

• Milkfish-oriented farming systems are favoured over prawn production

because the group perceive more work to be available on milkfish ponds than
on prawn-oriented ponds. They also know that it is possible to enter the
milkfish ponds at the end of the harvest and gather stranded fish for their own
consumption. Milkfish are left over after a harvest from a prawn polyculture
pond, but in lesser numbers.

• Smaller landholdings would allow for local people to move into aquaculture,
although the group believe that a minimum size of 10 hectares would be
required to ensure the economically viability of a fish farm.

4.3.2 Matrix scoring

The group, in consensus, chose the following indicators: employment multiplier,
technical efficiency, and ecological footprint. Employment was chosen immediately
by all members. Technical efficiency was proposed by one member and the others
agreed that to be technically competent was important, with the underlying
motivation being that “outsiders” do not understand the local ecology in the same
way and that absentee operators do not supervise their workers well. Ecological
footprint was the last to be selected but was agreed upon, with reference to the local
ecology.

Twenty stones were given to the group as markers to rank the three selected
indicators. After the markers were placed by one nominated member, the group was
asked again if they were satisfied with the ranking. The markers were re-allocated by
another member, which changed the relative weight for each category but not the
ranking. The rest of the group, however, did not agree and opted to return the
markers to their original position. The member that changed the weights conceded
and deferred to the “will of the group”. The final distribution was as follows:

• employment multiplier 45%

• ecological footprint 30%
• technical efficiency 25%

In the modelling, the ecological footprint indicator is not used because it does not
model the environmental impacts of milkfish farming owing to the problems in
establishing appropriated areas (as described in paper 2). However, the group were
reflecting environmental concern and the issues that they mentioned in justifying
their choice of indicator centred around the impacts of the aquaculture systems on
fishing. Pesticide application was perceived to be an issue, but none of the indicators
address this issue as it has been shown that applications do not vary by farming
system. Fish kills during periods when harvests occur were mentioned and these are
likely to be due to oxygen depletion through nutrient enrichment from the effluent.
Therefore, the nutrient balance is used in place of ecological footprint in this case,
despite the group overlooking that indicator – most likely through not fully
understanding the implications of nutrient enrichment and its links to ecological
health. This is clearly a problem with participatory research and we recognise that the
solution presented here is highly imperfect.

4.4 Weights from the Barangay Council, Sapang Kawayan

4.4.1 Focus Group Discussion

159
 Paper 3/5

The group consisted of eleven members: the barangay captain, the youth council
chairman, seven barangay councillors, the barangay treasurer and its secretary.
Mostly composed of men, the group have diverse interests, with some of the
members operatoring fish ponds, and others working as caretakers, labourers and
fishermen.

The initial discussion yielded the following main insights:

• Family involvement in the operations of the fishponds is seen as very
desirable. Some absentee operators do not trust local people to work as
caretakers or even as hired labourers.
• The culture of milkfish is associated with the large farms in the area. The use
of commercial feeds for the culture of milkfish is viewed negatively and is
associated with deterioration in the local water quality. The following quotes
illustrate this point, and are copied verbatim, with each group member
identified by a number, rather than by name.

Group Member 1
Iyong mga noong araw, noong mga unang panahon natural feeding lang. Mga
nilalagay lang lablab at lumot. Pero ngayon talagang gumagamit na ng feeds o
feeding. Ang epekto naman ng feeding sa ilog papangit ito, ginagamit pa rin ito ng
mga malalaking palaisdaan. Pinapatay talaga ng feeds ang mga isda na nasa labas
ng fishpond, tulad ng mga nasa dagat. Naaapektuhan naman ang kabuhayan ng mga
tao.
In the past, we have been using natural feeds for our farm, like the ‘lablab’ and
‘lumot’. At present, [fishponds] use commercial feeds. Although it is a fact that
commercial feeds destroy the river, big fish farms continue to use them. As the
polluted water, the effect of commercial feeds, descends to the nearby bodies of
water, it affects our fish farms and kills fishes in the river and sea as well.

Group Member 2
Nakaksira talaga sa kalikasan.
It destroys the environment.

Group Member 3
Oo, sinisira talaga niyan kasi pinapatay ang mga isda ng tubig na mula sa labas –
kahit nga pa pati ang dagat apektado na rin! Nababawasan ang kabuhayan namin.
Yes, it is true. The contaminated water wipes out the fishes – it even affects the sea!
There is hardly any source of livelihood left for us.

Group Member 4
Apektado kami ng katas ng feeds supply.
We are affected by the overload of feeds supply.

and later…

Group Member 1
Malaki nga ang epekto ng feeds. Kaya nga kinausap ako ng Kaliwa noong araw na
kailangan naming magtulong-tulong para diyan para matanggal ang feeds.
The use of commercial feeds has an enormous effect on the environment. There was
a time that a Leftist group talked to me about abolishing commercial feeds in fish

160
 Paper 3/5

farming. They told me that everybody had to help to stop the use of commercial
feeds.

Group Member 5
Hindi na dapat nagpifeeds.
Totally stop the use of commercial feeds.

Group Member 6
Kasi sa panahon ngayon ay wala ng lablab. Lahat na lang nakadepende sa feeds.
Kaya ang epekto sa ilog ay mas tumitindi dahil sa dami ng pakain.
At present, there isn’t much lablab. Everybody depends on commercial feeds for
their fishponds. The river suffers because of excessive use of feeds.

Group Member 7
Mas maraming feeds kasi mas maraming naghahanapbuhay parang mga
mangingisda, pero ang dagat naman ang benta.
Most of us here depend on fishing for our livelihood, so more people use commercial
feeds. But with this, we are taking the sea [nature] for granted.

• Rather surprisingly, technical knowledge, in the sense of formal education in

fisheries or a related science, is seen as being unnecessary or even damaging:

Group Member 3
I want to discuss the technical knowledge of the fishpond operator. We experienced
that some [operators] hired technicians to work in the fishponds years ago. Their
income improved at first, but when the excessive use of pesticides caused pollution
and eventually affected the quality of soil and water in [Sapang Kawayan], [we
realized] that it’s better for the fishpond to have caretakers instead of technicians. We
have qualified caretakers [here]; they have years of experience and exposure to fish
farming, which make them more qualified than technicians. Caretakers give the
fishpond more income as well. The technicians know too much about pesticides,
chemicals and commercial feeds, they apply it to fishponds. The quality of our soil
was ruined because of too much use of chemicals. That is our experience here. The
caretakers have intensive training on fish farming, they are in the fishpond come day
and night; they are knowledgeable with the current of the water; they know the life
cycle of the fishes. The water current [in this part] is different than the one in
Batangas [a province in Southern Philippines] where it is more predictable. If it rains,
water here become less salty, when high tide comes, salinity increases. The
technicians are not well aware of that; they don’t know that the salinity of water
changes like that.

4.4.2 Matrix scoring

The group chose profit and employment immediately. The choice of the third
indicator was between risk and ecological footprint. Risk was eventually chosen by
the group following the rationale that variability in income affected the availability of
work on the fishponds. The concerns regarding environmental and technical aspects
of the farming system that were brought up in the discussion are not reflected in the
choice of weights. The final distribution was as follows:
• profit 40%
• employment 25%

161
 Paper 3/5

• risk index 35%

4.5 Weights from a group of men in Sapang Kawayan

4.5.1 Focus Group Discussion
The group consisted of six members of between 34 and 56 years of age: two fishpond
operators, one man who collects small shells from the river to sell to fishpond
operators and three labourers (whose work is mainly on fishponds). The initial
discussion can be summarised as follows:

• The group believe that the amount of work available to them per hectare is
the same regardless of the size of the pond.
• Prawns are favoured over milkfish because there is the perception that a lot of
land is required to produce milkfish and that prawns offer the only possibility
for smallholdings to profit.
• Fishpond operators and labourers alike seem to be reliant on the environment,
particularly on the quality of the water in the rivers, for their livelihood. They
recognise their dependency on this resource.

The following passages from the transcript are copied verbatim:

Group Member 1
Ako, maganda din ang #4, pero sa aming mga mahihirap, na umaasa na lang ng
ikinabubuhay sa kalikasan, mas mahalaga ang #8. Kasi marami ang umaasa diyan,
kung maganda ang lagay ng kalikasan, marami din ang magkakaroon ng
pagkakakitaan.
The importance of #4 (technical efficiency) cannot be denied. But for poor people
like us who depend on the environment for our livelihood, I think #8 (ecological
footprint) is the most important of all. For me, the environment is really important
because it is where many of us depend for our source of income.

Group Member 2
Sang-ayon na din ako diyan. Tutol din kasi ako sa mga ganyan na nakakasira sa
kalikasan.
I agree with that. I strongly detest any [activity] that can harm the environment.

and later…

Group Member 3
Magbigay din sila ng puhunan. Kung kami ay may mauutangan, kahit ang mga
maliliit na propitaryo, hindi naman sa lahat ng pagkakataon ay may kita. Kailangan
na bigyan tayo ng puhunan ng gobyerno.
They should lend us money for capital. If we have [an institution] that will grant loan
to small scale fishpond operators there will be more of us who will benefit from it.
The government must provide us money for investment.

Group Member 1
Ang mga technical na kaalaman at magbigay ng atensiyon ang gobyerno sa #8
(kalikasan).
The government must help us increase our technical knowledge. They should give
more attention to the effects of fish farming in our environment as well.

162
 Paper 3/5

Group Member 4
Tama iyon. Dapat ay magbigay din sila ng mga seminar para mamulat ang mga tao
dito sa epekto ng mga nilalagay sa fishpond sa ating kalikasan.
That’s right. They should give seminars so people would be more aware of the
effects on the environment of what they put in the fishpond.

Group Member 3
Bigyan ito ng atensiyon ng gobyerno, pagtuunang pansin ang pollution. Lalo na ang
feeds sa mga bangus.
The government must give attention to water pollution, especially those which is
caused by commercial feeds used in milkfish farming.

and later….
Group Member 3
Kasi ang #8 kapag hindi mo binigyan ng atensiyon at magpatuloy ang paggamit ng
mga feeds at fertilizer masisira talaga ang kalikasan. Kawawa naman ang mga
maliliit na namamalakaya na umaasa sa ilog.
If we give less importance to preserving the environment [as in #8 – ecological
footprint] and continue to use commercial feeds and fertilizers, the environment will
be destroyed and the fishermen who depend on the river for their livelihood will
suffer.

Group Member 1
Lahat naman apektado diyan sa huli, hindi lang ang mga maliliit pati na rin ang mga
may-ari ng malalaking palaisdaan apektado kung masira ang kalikasan dahil sa mga
ginagamit nilang feeds.
All of us, not only those who rely on the river but also the big fishpond operators,
will suffer if the pollution from commercial feeds destroys the environment.

4.5.2 Matrix scoring

The group, in consensus, chose profit, employment and ecological footprint. The
allocation of stones to ecological footprint increased as the discussion continued to a
final distribution as follows:
• profit 35%
• employment 15%
• ecological footprint 50%

As before, the issues discussed in relation to the ecological footprint were all on the
subject of water quality and organic pollution – which are not covered by our
analysis using the footprint indicator, but the translation of the word “ecological” was
what the group were reacting to. The nutrient loss into the river was too technical a
concept for them, but their concerns are in this area, rather than in resource
appropriation (i.e. sink problems, rather than source problems). Therefore, nutrient
loss is used for the scenario modelling.

4.6 Weights from a group of women in Sapang Kawayan

4.6.1 Focus Group Discussion
The group consisted of 10 women of between 26 and 50 years of age. All the women
were only informally employed and some declared their husband’s occupations
(hired labourer on the fishponds or “Arawan”; caretaker on a fishpond; barangay

163
 Paper 3/5

captain). The initial discussions did not go well, as the women seemed unwilling or
unable to abstract from the point of view of the operator to a wider societal
perspective when it came to discuss feelings about the specific characteristics of
farming systems (e.g. family orientation vs. absenteeism; milkfish vs. prawn
orientation; large vs. small farm sizes). Most knew at least one fishpond operator, as
part of their social network, and spoke on their behalf in relaying their experiences of
these various characteristics.

However, the following insights were gained:

• Release of effluent as a result of high feeding rates was cited as a reason for
the decline in fisheries yields from the rivers around the barangay.
• A negative feeling towards technical assistance from hired technicians from
fisheries programmes predominates in the group, owing to the poor results
achieved when a number of operators hired technicians (that is, fisheries
graduates and / or consultants from SEAFDEC). Many operators hiring
technicians went bankrupt, and the presence of technicians likely reflects an
attempt to intensify production by the operators. It would seem as though the
hiring of technicians coincided with the national crash in prawn production
from Vibrio sp. diseases in 1994, and so the timing may have been
unfortunate, but this is not recognised by the group. There is therefore a deep
suspicion of anything other than local knowledge and experience in managing
the ponds.
• Pesticides are perceived as being necessary and important in the functioning
of the ponds. No mention is made of any health-related impacts of pesticides.
• Poor water quality is widely recognised within the group, with members of
the group attributing it variously to fishpond effluents, factory pollution, and
other upstream sources from neighbouring municipalities.

Regarding this last point, the following passage is copied verbatim from the
transcript of the FGD:
Facilitator
Ano kaya ang pinakapangunahing dahilan ng water pollution? Ang factory ba na
sinasabi nila na nakakaapekto sa mga tubig dito? O ang bangus feeds na nakakasira
din?
What could be the major agent of pollution? The wastes from the factories that flows
to the river or the feeds that the fishpond operators use?

Group Member 1
Yung feeds.
The feeds.

Group Member 2
Sabi ng iba yung factory.
Others say it’s the factory.

Group Member 1
Malayo naman sa atin yun. Kinakabahan nga ako sa polusyon. Basta hindi dito
nagmumula ang polusyon.
The factory is far from here. I am worried with the pollution. It isn’t from here.

164
 Paper 3/5

Group Member 3
Sa Apalit, doon nagmumula. Sa Hagonoy doon nagmumula.
Usually dirty water comes from Apalit or Haqonoy [neighbouring municipalities].

Facilitator
Pero nadoon karamihan ang masyadong nagfefeeds dito ng bangus?
Is that so because more fishpond operators there feed their milkfish? [These
municipalities have a lot of semi-intensive milkfish monoculture in ponds and pens].

Group Member 1
Iyan siguro, dumadating na lang ang tubig dito na sa Sapang Kawayan.
That can be a reason - the water flows to Sapang Kawayan from there.

Facilitator
Dumadaloy lamang.
It flows from somewhere else.

Group Member 4
Hindi dito nagmumula ang polusyon.
The polluted water did not come from Sapang Kawayan.

Group Member 1
Dumadaloy.
It flowed from somewhere else.

Group Member 3
Kapag amoy dagat ay malinis.
If the water smells like salt, it’s from the sea and it’s clean.

Group Member 5
Naaamoy namin iyon.
We can smell that.

4.6.2 Matrix scoring

Profit and employment were considered the most important indicators. Ecological
footprint was chosen by the group but their rationale was again based on nutrient
balance issues, reflecting a misunderstanding of the meanings of the terms described.
This situation is similar to that previously described in scenario 2. The markers were
spread evenly across the 3 indicators, with 7 stones each to profit and employment
and 6 stones to the environmental impact. Thus the distribution of weights is as
follows:
• profit 35%
• employment 35%
• ecological footprint 30%

165
 Paper 3/5

5. Compromise model results

5.1 All weights equal
The rationale behind using equal weights is that it is possible to establish a
benchmark, where all criteria are considered. With the inclusion of profit, risk, and
technical efficiency in the final list of six indicators, economic issues dominate in
this scenario. The metric p = ∞ results in almost equal scores for all indicators as the
maximisation operator considers the number of maximal elements (type 3 has two
maximal elements, types 1, 2 and 4 have one each) and in the cases where there is
one maximal element, the distance collapses to the weight attached to each indicator.

In this scenario, where six dimensions are considered, results for all three metrics are
reported. Thus, in the scenarios when p =1 and p = 2, the rankings are the same, with
farm type 1 (extensive polyculture) ranked highest by some distance. Farm type 3
(labour-intensive, low-input systems) ranks second, followed by semi-intensive
milkfish monoculture, large milkfish oriented systems and, ranking last, semi-
intensive prawn polyculture. As an overall balance of the six indicators used in the
modelling exercises, farm type 1 would appear to be superior by some distance, and
farm types 2 and 4 perform poorly.

Table 4 - Compromise model results when all indicators are given equal weights. The
results under p values of the Lp metric of 1, 2 and ∞ are reported.
Indicator Weight
Profit 0.167
N loss 0.167
Employment 0.167
Protein production 0.167
Risk index 0.167
TE 0.167

p=1 p=2 p=∞

Farm type 1 (distance) 0.304 0.105 0.167
Farm type 2 (distance) 0.767 0.364 0.167
Farm type 3 (distance) 0.482 0.180 0.156
Farm type 4 (distance) 0.745 0.341 0.167
Farm type 5 (distance) 0.617 0.222 0.161

Farm type 1 (rank) 1 1 3

Farm type 2 (rank) 5 5 3
Farm type 3 (rank) 2 2 1
Farm type 4 (rank) 4 4 3
Farm type 5 (rank) 3 3 2

As can be seen from these results, the distances (and subsequent rankings of farming
systems) are sensitive to the p value in the Lp metric. In six dimensions, it is necessary
to explore the range of possible metrics, as the properties of distances are not intuitive
when the number of dimensions is greater than three. In the remaining four scenarios,
only three indicators are chosen by each group, thus defining sets of weights in 3-
dimensional space. The most intuitive p metric to use in this case is thus p=2, which

166
 Paper 3/5

defines the Euclidean distance (i.e. a straight line). Thus, in the remaining scenarios,
only the results of the p=2 metric are reported.

5.2 Weights from the barangay council of Nanding Lopez

Farm type 3 (low-input, labour intensive systems) ranks highest. Smallholdings are
not abundant in the areas around Nanding Lopez, where large landowning families
dominate production. These large farms (farm type 4) rank poorly with the group.
Employment is the most important consideration for the council when they think
about aquaculture, as they know that for most of their constituents, the major benefit
is being able to sell their labour to the pond operators. The issue of workers being
brought in from outside the barangay by absentee operators is clearly one that upsets
them. These results suggest that aquaculture as currently practiced in the area is not
satisfactory to the group, and that other farming system types would be preferred.

Table 5 – Compromise model results from the weight set given by the barangay
council, Nanding Lopez. Parameter p = 2.
Indicator Weight
Profit 0
N loss 0.3
Employment 0.45
Protein production 0
Risk index 0
TE 0.25

Distance to utopian point Rank

Farm type 1 0.229 3
Farm type 2 0.357 5
Farm type 3 0.104 1
Farm type 4 0.340 4
Farm type 5 0.186 2

5.3 Weights from the barangay council of Sapang Kawayan

Farm type 1 (extensive polyculture) is optimal and is considerably closer to the
utopian point than any of the other alternatives. Of particular note is the very poor
performance of farm type 4 (very large milkfish-oriented farms). In particular, the
issue of absentee operators not trusting the local community to work as caretakers or
labourers and instead bringing in their own workers “from outside”, was again
prominent in the focus group discussion.

167
 Paper 3/5

Table 6 – Compromise model results for the weight set given by the barangay
council in Sapang Kawayan. Parameter p = 2.
Indicator Weight
Profit 0.4
N loss 0
Employment 0.25
Protein production 0
Risk index 0.35
TE 0

Distance to utopian point Rank

Farm type 1 0.138 1
Farm type 2 0.221 3
Farm type 3 0.271 4
Farm type 4 0.481 5
Farm type 5 0.205 2

5.4 Weights from the group of men in Sapang Kawayan

As in the previous case, Table 7 shows that farm type 1 (extensive polyculture) is
optimal and very close to the utopian point (distance is only 0.086). Farm type 4
farms (large farms) are those that dominate the local area (there is a 130 hectare
fishpond next to Sapang Kawayan) and these perform very poorly under this
scenario. Environmental concerns are paramount in this group, as they are very
aware of the impacts of intensification on the water quality locally and their
dependence on the inter-tidal areas being in a fit state to support aquatic life. The
group’s expression of preference for prawn over milkfish as “the only possibility for
smallholdings to profit” is also borne out in the rankings in Table 7.

Table 7 – Compromise model results for the weight set given by the focus group of
men in Sapang Kawayan. Parameter p = 2.
Indicator Weight
Profit 0.35
N loss 0.5
Employment 0.15
Protein production 0
Risk index 0
TE 0

Distance Rank
Farm type 1 0.086 1
Farm type 2 0.277 3
Farm type 3 0.261 2
Farm type 4 0.455 5
Farm type 5 0.281 4

168
 Paper 3/5

5.5 Weights from the group of women in Sapang Kawayan

While farm type 1 is ranked highest of the five farm types, farm types 1, 2, 3 and 5 all
return distances to the utopian point in the range 0.18 – 0.22. Only farm type 4 (Very
large milkfish-oriented farms) performs poorly.

Table 8 – Compromise model results for the weight set given by a group of women
in Sapang Kawayan. Parameter p = 2.
Indicator Weight
Profit 0.35
N loss 0.3
Employment 0.35
Protein production 0
Risk index 0
TE 0

Distance Rank
Farm type 1 0.186 1
Farm type 2 0.214 3
Farm type 3 0.218 4
Farm type 4 0.467 5
Farm type 5 0.194 2

6. Overall findings

Fairly consistent patterns have emerged from the analysis. Either farm type 1
(extensive polyculture) or farm type 3 (low-input, labour intensive systems) are
optimal under the weight sets considered here, (as modelled in tables 4 to 8, and as
shown by the summary in table 9). These are both extensive farming systems and
these findings suggest that for the communities that live in the area, relatively low
external input aquaculture is the most appropriate. These farming systems are not
significantly different from practices that fishpond operators in the Philippines would
consider traditional. Thus, alongside the results of the compromise model, we might
also consider these systems to be highly socially acceptable. They do not represent a
recent break with tradition (as represented by farm types 2 and 5), nor do they
dominate the landscape owing to their large size (farm type 4). Preferences for
extensive polyculture and low-input labour-intensive systems are thus likely to have a
significant socio-historical determinant.

Looking at the pattern of weights chosen in the four focus groups, we can see that
each group fairly consistently chose one economic indicator, one socio-economic
indicator and one environmental indicator. Net protein production was never chosen
and risk and technical efficiency only once each. If we pool these weights to form a
composite “average weight”, then we can obtain an overall distance and overall
ranking. Pooling the weights and then computing the distances, or computing the
distances using each set of weights and then calculating the average distance gives the
same results.

169
 Paper 3/5

Table 9 - Different weights from each FGD

Scenario Scenario Scenario Scenario 5 AVERAGE
2 (NL 3 (SK 4 (SK (SK WEIGHT
council) council) men) women)
Profit 0 0.4 0.35 0.35 0.275
N loss 0.3 0 0.5 0.3 0.275
Employment 0.45 0.25 0.15 0.35 0.3
Protein production 0 0 0 0 0
Risk index 0 0.35 0 0 0.0875
TE 0.25 0 0 0 0.0625

Table 10 - Distances using these “average” weights and p=2

Distance using average
Rank
weights, p=2
Farm type 1 0.1596 1
Farm type 2 0.2673 4
Farm type 3 0.2136 2
Farm type 4 0.4357 5
Farm type 5 0.2167 3

Taking the mean distance across all four focus group-based scenarios, we find that,
overall farm type 1 (extensive polyculture) performs the best in terms of local
imperatives. It is the closest to the utopian point by some distance. Farm types 2, 3
and 5 are middle-ranking, with farm type 4 (very large milkfish-oriented farms) very
far behind.

6.1 Implications – Local level

In terms of the multiple perspectives that are relevant when we consider the case of
barangay Sapang Kawayan, we find that the barangay council, a group of randomly
selected men, and a group of randomly selected women all choose sets of indicators
and weights that suggest farm type 1 (extensive polyculture) to be the preferred
model. This unexpected level of agreement implies that there are not significant
social tensions within the community regarding aquaculture (i.e. there is social
commensurability of values). While this may be true, there would seem to be an
important issue with regards to the relationship between the communities and the
operators of the very large farms that dominate much of the land-use in both Sapang
Kawayan and Nanding Lopez.

Some members of one of the focus groups mentioned community agitation by leftist
activists, in addition to the widely-held knowledge that the New People’s Army are
active in extorting money with the threat of violence from large-scale operators. This
suggests that land distribution is an important local political issue. Added to this, we
find that the results of the compromise model for farm type 4 (large milkfish-oriented
systems) are poor – it ranks last out of the five farm types in both Nanding Lopez and
Sapang Kawayan – suggesting that these systems are performing relatively poorly in
relation to the concerns of the communities. There is thus significant potential for
social unrest and conflict.

170
 Paper 3/5

6.2 Implications – National level

The top two ranking farm types (1 and 3) are also the most abundant in our sample.
Of a total of 136 farms in the sample for the typology (paper 1), 54 were type 1 farms
and 37 were type 3 farms. Thus, the highest ranking two categories represent 91 out
of a sample of 136 (66.9%). This is encouraging – it suggests that a majority of the
farming systems in regions 3 and 6 are performing well with regard to the
sustainability indicators presented here.

The picture is different if we consider the individual farm sizes. If our sample
stratification is a good representation of the actual distribution (and a significant
amount of energy was put into the process of ensuring that it is), then we can
examine the performance of the sector per unit area. Thus, farm types 1 and 3
represent only 42.9% of the total area in regions 3 and 6, whereas the very large
farms that perform so poorly in the sustainability scenarios occupy 46.5%. There are
implications here for the agrarian reform of the sector – almost half of the brackish-
water pond area in regions 3 and 6 is occupied by very large farms that perform
poorly relative to the other types.

Thus, it would seem that there is considerable potential for all-round improvement in
the farm-level performance of the sector by including aquaculture under agrarian
reform law. The rationale behind this statement is that agrarian reform would reduce
the number of farms performing poorly (farm type 4) and, with appropriate
institutional innovations, promote aquaculture farming systems of the kind that are
performing well. Evidence of a possible win-win situation with regard to efficiency
and equity objectives of such a change are considered in paper 4. The overall
impression from the results presented in this paper is that, were the brackish-water
areas in Central Luzon and Western Visayas covered in fishponds of a kind similar to
farm types 1 and 3 (i.e. relatively low-input, ≤ 10 hectares), then there would be
benefits to the sustainability of the sector.

Our approach to sustainability has shown the trade-offs at the farm level between the
three objectives of maximising economic, social and ecological performance. These
objectives may vary in their importance in different geographic localities, and among
different groups of people in each geo-political unit. Our approach has helped answer
questions regarding who benefits from aquculture as it is currently practiced, and
where areas of tension or prospects for development may lie. Somewhere in the
negotiation between objectives and values, we find sustainability.

171
 Paper 3/5

References:

Ballestero, E. and C. Romero (1991). "A theorem connecting utility function

optimization and compromise programming." Operations Research Letters
10(7): 421-427.

Ballestero, E. and C. Romero (1994). "Utility optimization when the utility function is
virtually unknown." Theory and Decision 37(2): 233-243.

Ballestero, E. and C. Romero (1998). Multiple criteria decision making and its
application to economic problems. Dordrecht, Kluwer Academic Publishers.

Bouman, B. A. M., R. A. Schipper, A. Nieuwenhuyse, H. Hengsdijk and H. G. P.

Jansen (1998). "Quantifying economic and biophysical sustainability trade-
offs in land use exploration at the regional level: A case study for the
northern atlantic zone of costa rica." Ecological Modelling 114: 95-109.

Brown, K., W. N. Adger, E. Tompkins, P. Bacon, D. Shim and K. Young (2001).

"Trade-off analysis for marine protected area management." Ecological
Economics 37: 417-434.

Chambers, R. (1997). Whose reality counts? Putting the last first. London,
Intermediate Technology Publications.

El-Gayar, O. F. and P. Leung (2001). "A multiple criteria decision making framework
for regional aquaculture development." European Journal of Operational
Research 133(3): 462-482.

Glaser, B. (2002). Efficiency versus sustainability in dynamic decision making:

Advances in intertemporal compromising. Lecture Notes in Economics and
Mathematical Systems, Springer. Berlin: 252.

Gregory, R. and K. Wellman (2001). "Bringing stakeholder values into environmental

policy choices: A community-based estuary case study." Ecological
Economics 39: 37-52.

Kobrich, C. J. (1997). The construction and use of compromise programming models

to measure the impact of development policies on the sustainability of
peasant farming systems in central Chile. Department of Agriculture.
Reading, University of Reading.

Lu, C. H. and M. K. van Ittersum (2004). "A trade-off analysis of policy objectives
for ansai, the loess plateau of china." Agriculture Ecosystems &
Environment 102: 235-246.

Martinez-Alier, J., G. Munda and J. O'Neill (1998). "Weak comparability of values as

a foundation for ecological economics." Ecological Economics 26(3): 277-
286.

172
 Paper 3/5

McCoy, A. W. and E. C. de Jesus (1982). Philippine social history: Global trade and
local transformations. Honolulu, Hawaii University Press.

Munda, G. (2004). "Social multi-criteria evaluation: Methodological foundations and

operational consequences." European Journal of Operational Research
158(3): 662-677.

O'Hara, S. U. (1996). "Discursive methods in ecosystem valuation and environmental

policy." Ecological Economics 16(2): 95-108.

O'Neill, J. (1997). "Managing without prices: The monetary valuation of

biodiversity." Ambio 26(8): 546-550.

Roth, T. P. (1999). Ethics, economics and freedom: The failure of consequentialist

social welfare theory. Aldershot, Ashgate Publishing Ltd.

Stoorvogel, J. J., J. M. Antle, C. C. Crissman and W. Bowen (2004). "The trade-off

analysis model: Integrated bio-physical and economic modelling of
agricultural production systems." Agricultural Systems 80: 43-66.

Sumberg, J., C. Okali and D. Reece (2003). "Agricultural research in the face of
diversity, local knowledge and the participation imperative: Theoretical
considerations." Agricultural Systems 76: 739-753.

Zeleny, M. (1982). Multiple criteria decision making. New York, McGraw-Hill Book
Company.

1
Described in detail in paper 2.
⊗
2
That is, the intersection of all individually optimal solutions is nonempty (∩ k =1 Xk ≠∅).
K

173
 Paper 4/5

Research project R8288: Assessing the sustainability

of brackish water aquaculture systems in the
Philippines – Paper 4/5

Aquaculture & Poverty – A Case Study of Five Coastal

Communities in the Philippines

Abstract: After reviewing the rather thin literature on the subject, we investigate the
relationship between aquaculture and poverty based on a case study of five coastal
communities in the Philippines. The analysis relies on a data set collated through a
questionnaire survey of 148 households randomly selected in these five communities.
The methodological approach combines the qualitative analysis of how this
relationship is perceived by the surveyed households and a quantitative analysis of
the levels and determinants of poverty and inequality in these communities. There is
overwhelming evidence that aquaculture benefits the poor in important ways and that
it is perceived very positively by the poor and non-poor alike. In particular, the poor
derive a relatively larger share of their income from aquaculture than the rich, and a
lowering of the poverty line only reinforces this result. Further, a Gini decomposition
exercise shows unambiguously that aquaculture represents an inequality-reducing
source of income. We believe that the pro-poor character of brackish water
aquaculture in the study areas is explained by the fact that the sector provides
employment to a large number of unskilled workers in communities characterized by
large surpluses of labour. Our results also suggest that the analysis of the
relationship between aquaculture and poverty should not focus exclusively on the
socio-economic status of the farm operator/owner, as has often been the case in the
past.

174
 Paper 4/5

Contents

1. Introduction 176

2. Aquaculture and poverty: the state of knowledge 177

3. Methodology & Data 179

3.1 Methodological Overview 179
3.2. The study area and data collection 179

4. Poverty & aquaculture: perceptions 184

4.1 Defining the context: general perceptions of poverty 184
4.2 Aquaculture & poverty: the views from five communities 187

5. Levels and Determinants of Poverty and Inequality in the Selected

Communities 191
5.1 Approach to Poverty Measurement 191
5.2 Levels of Poverty 192
5.3 Is Aquaculture Pro-poor? 194
5.4 Measuring and Explaining Income Inequality 196
5.5 Discussion 200

6. Conclusion 201

Appendix 213

175
 Paper 4/5

1. Introduction
Poverty in the Philippines remains a major problem, which represents a formidable
barrier to the country’s development. As will be discussed in the methodological
section, there exists multiple ways of measuring poverty, which can explain some of
the discrepancies found in the literature regarding its incidence in the Philippines as
elsewhere. However, there is little doubt that poverty levels in the Philippines are high
both in absolute and relative terms. According to the latest set of World Development
Indicators (Table 1), more than one third (37%) of the Philippine population lives
under the national poverty line, while 15% finds itself in absolute poverty as defined
by the $1 a day criterion of the World Bank. Furthermore, these figures compare
unfavourably to those of other countries of the South-East Asian region such as
Indonesia, Malaysia and Thailand. This is explained by a relatively low level of
income per capita (at least when compared with Malaysia and Thailand) and relatively
large income inequalities (when compared with Indonesia). As a consequence, the
incidence of poverty in the Philippines is as acute as in countries with much lower
income levels, such as Cambodia. Its reduction ranks high on the agenda of the
government and other international agencies.

Table 1: Incidence of Poverty in South-East Asia

Country National Poverty Line International Poverty Line GNI per
capita ($)
National Rural Urban Population Population
% % % below $1/day below
$2/day
Cambodia 36.1 40.1 21.1 NA NA 310
Indonesia 27.1 NA NA 7.5 52.4 810
Malaysia 15.5 NA NA <2 <0.5 3,780
Philippines 36.8 50.7 21.5 14.6 46.4 1,080
Thailand 13.1 15.5 10.2 <2 <0.5 2,190
Source: World Bank Development Indicators, 2003

How can poverty be reduced in the Philippines? The starting point to answer this
question is a realisation that the international consensus on the achievement of
poverty alleviation has changed over the last two decades. The view that economic
growth represents a sufficient condition for poverty alleviation has proven wrong, and
it is now clear that a satisfactory rate of poverty reduction cannot automatically be
achieved through the mere trickle-down effect of growth to the poor. Instead, decision
makers need to design policies with a clear pro-poor focus, i.e. policies that benefit
the poor disproportionately. This thinking has also modified the way in which
development agencies function, in particular with respect to the funding of research
in/for developing countries. It is no longer sufficient for researchers to claim that their
activities will lead to productivity gains and economic growth; instead, demonstrating
the effect on the poor of particular projects has become paramount. This general
statement is reflected in DFID’s Renewable Natural Resources Research Strategy
(RNRRS), which no longer focuses only on the generation of new knowledge in
natural and social sciences, but also on the promotion of the use of this knowledge to
improve the livelihoods of poor people. In this context, investment in aquaculture
R&D for developing countries can only be justified to the extent that the resulting
knowledge and technologies make a positive contribution to the livelihoods of the
poor. This paper aims at testing whether this is actually the case based on a

176
 Paper 4/5

community-level analysis of poverty in several coastal areas of the Philippines where

brackish water aquaculture is present. We start with a brief literature review of the
relationship between aquaculture and poverty; move on to present an overview of the
methodology and data collection; and, finally, present the qualitative and quantitative
analysis of the relationship between aquaculture and poverty.

2. Aquaculture and Poverty: The State of Knowledge

At a conceptual level, the potential contributions of aquaculture to poverty reduction
are relatively well understood (Edwards, 1999; Muir, 1999). Several opportunities can
arise for the poor from the improved use of aquatic resources that aquaculture
development permits. There are, first, the direct effects generated by this
development, i.e. effects that can be directly related to the farm’s activities.
Aquaculture growth generates new income, calculated as production sales minus
variable costs, which accrues to the owner(s) of the fixed factors of the fish farm
(mainly the pond/land, family labour, management and other necessary equipment
such as boats and nets). The impact on poverty of this additional income flow depends
on the socio-economic status of the farm operator/owner and will only be significant
if the poor themselves participate in aquaculture. Obstacles to this participation are
potentially numerous and include the capital and skill intensity of the activity as well
as its riskiness. At this level, it is usually thought that extensive or semi-intensive
forms of aquaculture are relatively more pro-poor than intensive systems, due to the
fact that the poor usually lack access to credit, which prevents them from purchasing
the intermediate inputs used in large amounts in intensive systems.

Aquaculture development can also generate employment on the farm, either on a full-
time basis when a ‘caretaker’ is responsible for the day-to-day farm operations, or on
a more occasional basis for seasonal tasks, such as harvest. This is likely to benefit the
poor in countries with large labour surpluses, such as the Philippines, because a poor
person’s labour often represents his main asset and, by the same token, his main
source of livelihood. When comparing different forms of aquaculture, it is also likely
that their relative labour intensities have an important bearing on their relative
potentials for poverty reduction.

However, small direct effects would not necessarily imply that aquaculture is not
‘pro-poor’. It is possible that the additional income stream and employment generated
by aquaculture development trickles down to the poor through a series of linkages
within rural communities. These include production links, both ‘upstream’ from the
farm in demand for inputs and services for aquaculture, as well as ‘downstream’ from
the farm in the demand for processing, storage, and transport of production. There are
also consumption links as fish farmers and farm labourers spend their increased
incomes on goods and services that are provided outside of aquaculture. While
conceptually simple, these growth linkages are difficult to measure but, in agriculture
at least, most empirical studies have estimated large multipliers, explained primarily
by the strength of the consumption linkages (Irz et al., 2001) 1. This implies that our
study should not focus exclusively on the farm but, instead, should take a broader
view of the relationship between aquaculture and poverty.

1
These ideas have been formalised in so-called agriculture-led industrialisation strategies that stress the
importance of agriculture in creating a market for industrial products (Adelman, 1995).

177
 Paper 4/5

Many authors also argue that aquaculture development can have a positive nutritional
effect on the poor as a supplier of high quality animal proteins and essential nutrients
(Prein and Ahmed, 2000). If the poor adopt aquaculture, own-consumption of fish by
the farm household can increase its nutrition and food security. In addition,
aquaculture growth increases the supply of fish, which reduces its price and makes it
more affordable to the poor. Note, however, that the argument depends on the size of
the market where production is sold as well as on the nature of the fish produced. If
aquaculture production is sold locally on small and poorly integrated markets, it is
likely that the price decrease resulting from additional supply will be large; on the
other hand, if production is exported, the nutritional benefits to the country’s poor will
be non-existent. In a similar vein, these benefits will only materialise if the poor,
either locally or nationally, do indeed consume the species produced by aquaculture
and, in the case where only high-value species are farmed, no such nutritional benefits
can be claimed. Sometimes, it can also be argued that aquaculture improves the
nutrition of the poor through other channels. For instance, caretakers are sometimes
allowed to catch fish in the ponds that they supervise to satisfy their family’s
consumption needs. A common practice in the Philippines is also to allow poor people
to catch any residual production after the main harvest has taken place in large
brackish water ponds.

Altogether, the arguments supporting the pro-poor nature of aquaculture rely on the
income stream, employment and nutritional benefits that it can potentially generate.
To some extent, these arguments apply to any agricultural enterprise as well as
fishing, but aquaculture presents some advantages over these activities. First, it often
represents the only option to farm land under saline conditions, which is precisely the
case in the large areas of brackish water of the Philippines. Second, the productivity
of fisheries is often limited by its open-access nature, which results in the well-known
‘tragedy of the commons’ that some identify as a cause of poverty (Hardin, 1968). By
contrast, aquaculture development involves the creation of well-defined property
rights that form, arguably, a pre-condition for productivity growth and represents an
important developmental option for many coastal communities characterized by high
levels of poverty. Finally, fish is a nutrient efficient protein source, in comparison to
livestock, so there is an underlying biological reason for claiming that aquaculture
represents a particularly attractive way of producing cheap proteins for the poor and
the malnourished. In fact, so intuitive is the previous set of arguments that aquaculture
generated massive enthusiasm in the last two decades, with some viewing its
development as a ‘blue revolution’ with tremendous potential for fostering food
security, generating economic growth in rural areas and alleviating poverty.

However, the empirical evidence regarding the ability of aquaculture to reduce

poverty is mixed at best (Edwards and Demaine, 1997). The assessment of this
potential is made all the more difficult that while there are many studies of poverty in
farming communities and among the urban poor, few empirical studies have focused
specifically on aquaculture (FAO, 2003). Yet, there is a general view in the literature
that the promotion of aquaculture in Africa and Latin America has largely been
unsuccessful (Edwards and Demaine, 1997). In Asia, while commercial scale
aquaculture has recently experienced a spectacular expansion, it is households with
better resource bases rather than the poor who have benefited. As a result, donor
support for aquaculture development has declined in the past ten years (Hlwart et al.,
2003). However, there is also anecdotal evidence that coastal aquaculture can

178
 Paper 4/5

represent an important source of employment for the rural poor, through the demand
for labour input, seed and feed (Edwards, 1999; Tacon, 2001). On the other hand,
several case studies have documented the fact that aquaculture development can, in
some cases, have a detrimental effect on the poor due to its environmental impact or
its role in triggering social conflicts. The shrimp industry, in particular, has been
blamed for a whole series of problems that, it is argued, have sometimes made the
poor strictly worse off (Stonich et al., 1997).

3. Methodology & Data

3.1 Methodological Overview
The literature review on the relationship between aquaculture and poverty reveals the
need for further empirical inquiry. Several methodological approaches could be used
to support this inquiry and these were discussed during a workshop organized by
PCAMRD and the University of Reading on 22 April 2004 in Los Banõs, Philippines.
Following the workshop that generated invaluable insights, the following
methodological choices were made by the project leaders:

• The unit of analysis for this study should be the whole ‘community’. It was
felt that focusing solely on fish farms would be too restrictive in the sense
that it would limit our understanding of poverty in the coastal areas of the
Philippines. In particular, a farm-level analysis would make it very
difficult to investigate how important aquaculture really is for the
livelihoods of poor people in these communities, or to put into light any
negative impact of aquaculture on the poor. By contrast, focusing on a few
communities allows us to gain in depth understanding of the economic,
institutional and social characteristics of these communities that are
essential to investigate rural livelihoods and rural poverty in a holistic
manner (Bebbington, 1999).

• The analysis adopts primarily a ‘traditional’ approach to poverty

measurement and evaluation. By that, we simply mean that the
identification of the poor relies on quantitative consumption and income
data that are collected through a survey. Although this approach has
undoubtedly some shortcomings, there is little evidence that the more
qualitative alternatives that have been proposed in recent years are superior
(Ravallion, 1996, p. 124). In short, the methodology aims at defining
profiles of poor and non-poor households in the chosen communities,
hence establishing how the two groups differ in terms of their involvement
in aquaculture (or aquaculture-related activities).

• However, it is also clear that the acceptability and effectiveness of

development and poverty policies depends in large part on the perceptions
by the stakeholders themselves of poverty and poverty reducing measures.
Hence, we also decided to investigate the subjective notion of poverty
through participatory methods (see Hentschel and Waters (2003) for a
recent application of that approach).

3.2. The Study Area and Data Collection

3.2.1. Regional Context

179
 Paper 4/5

Brackish water aquaculture in the Philippines represents a traditional activity which

has grown over the last few decades. Although uncertain, the total surface area of
fishponds is large and has increased until recently through the conversion of
mangroves and swamps. Yap (1999) suggests that there are 239,323 hectares of
brackishwater fishponds in the Philippines, while the electronic data that we obtained
from the Bureau of Agricultural Statistics (BAS) records a total harvested area of
415,272 hectares in year 2000, although this latter figure might be accounted for by
joint or multiple owners of a single fishpond, each having featured in the survey
separately. The ponds are distributed unequally across the country, with Region 3
covering the central part of the Northern island of Luzon and Region 6 in the Western
Visayas (central Philippines) dominating the industry. Hence, Region 3 accounts for
28% of the total area of fish ponds and 39% of the national production in value terms,
while the corresponding figures for Region 6 are 24% and 14% (BAS). We therefore
decided to concentrate the study on these two regions where the economic importance
of brackish water aquaculture is largest.

The situation with regard to poverty appears significantly different in the two selected
regions. Figure 1 reveals that Central Luzon is characterized by the lowest incidence
of regional poverty in the country at less than 20%. This can be related to the relative
economic prosperity of the region due to the proximity of the national capital, Manila,
and its many industries. By contrast, poverty incidence is high in the Western
Visayas, with close to half of the population classified as poor ; only regions in Bicol
and Mindanao fare worse in terms of poverty nationally.

The literature establishes that poverty is really a multi-dimensional concept and that it
should therefore be evaluated from a variety of angles. One such angle is the
prevalence of malnutrition among young children, presented in the second map
(Figure 2). The indicator of malnutrition selected measures the proportion of children
under the age of 59 months with a weight for age ratio smaller than the population
average by at least two standard deviations. The national picture of poverty that
emerges from this second map differs slightly from that described in Figure 1.
Consistent with the previous results, the incidence of child malnutrition in Central
Luzon appears relatively low, but in the Western Visayas region, more than one third
of the children are seriously underweight. The ability of aquaculture to improve
nutrition, as postulated in the literature, seems therefore particularly relevant in
Region 6.

180
 Paper 4/5

Region 3

Region 6

% Population under
poverty line
< 20
21 - 30
31 - 40
41 - 50
> 50
Missing Data

Figure 1: Poverty Map of the Philippines, 1997

Source: 1997 Philippine Poverty Estimates, NSCB 1998

181
 Paper 4/5

Region 3

Region 6

% Children with
weight for age < -2SD
26 – 30
31 – 35
> 36

Figure 2: Proportion of Underweight Children (0-59 Months), 1998

Source: National Nutrition Survey of 1998, FNRI-DOST

3.2.2 Data collection and overview of selected communities

A survey collected household-level data based on a questionnaire presented in the
Appendix. Its core was inspired by the questionnaire developed by the World Bank’s
Living Standards Measurement Survey (LSMS) team (Grosh & Glewwe, 2000) but
we also had to make important adjustments because of specific objectives as well as
time and financial constraints. The questionnaire is divided into ten sections that give
a fairly comprehensive overview of a household’s socio-economic situation
(household composition, education, employment, land-based activities (aquaculture &
agriculture), fishing activities, other sources of income (transfers, remittances, rental
earnings), consumption, asset ownership, housing, access to healthcare and credit). In
addition, it contains a whole section investigating how the respondent perceives
poverty and its relationship to aquaculture. The questionnaire was piloted by the
research team in May 2004, which led to major revisions, and the survey proper was
carried out from June to October 2004 in the two selected regions. The data was
collected by enumerators during two face-to-face interviews taking place at a week’s
interval, with all the recall data on consumption being collected during the second
visit.

An important step in implementing the methodology involves specifying precisely

what is understood by the term ‘community’. For our purposes, we chose the smallest
administrative unit in the Phillipines, called a barangay, which corresponds roughly to

182
 Paper 4/5

the borough of a municipality. Residents appear to have a real sense of belonging to

their barangay, which has its own institutions (in particular, a barangay council and
captain, who are elected) and social events (in particular, the annual fiesta which is
organized on the day of the barangay’s saint). Participants at the workshop confirmed
that barangays represent appropriate communities for our study.

Sampling followed a three-stage strategy. The first stage, discussed previously,

selected the two regions with the most developed aquaculture sectors (Regions 3 & 6).
The second stage selected a few barangays in each region. Our initial intent was to
select those randomly but discussions at the workshop made clear that a purposive
strategy would in fact be preferable. There were several suggestions at the workshop
that the impact of aquaculture on the poor probably depends on the type of
community considered. Relevant characteristics include the remoteness of the
community, its distance to the town/village centre, the level of urbanization of the
barangay and its surroundings, the importance of fishing as an economic activity in
that community, and the presence of mangroves. A brief description of the five
selected communities follows:

• San Antonio (SA) is a barangay of the ‘partially urban’ municipality of

Sasmuan, in the Pampanga province of Region 3 (NSCB Philippine Standard
Geographic Codes, July 2004). It is part of the town centre, close to the
market, municipal hall and other basic institutions of the municipality.
However, its total land area of more than 800 hectares is occupied primarily
by fishponds. The barangay has a total population of 1,603 individuals spread
across 286 households (2000 National Statistics Census).

• Barangay Sapang Kawayan (SK) is a rural community under the jurisdiction

of the municipality of Masantol in the Pampanga province of Region 3.
Masantaol is classified as ‘partially urban’ and has 32,464 registered voters
(National Statistical Coordination Board, 2004). The community of SK is
located to the South of Masantol proper and is only accessible by boat (one
hour from Masantol and 20 minutes from another town, Haganoy, in the
nearby province of Bulacan). It is surrounded by fishponds and river systems
and has a total land area of 265 hectares (Provincial Agriculturalist’s Office),
which supports a population of 2,676 individuals in 559 households (2000
National Statistics Census).

• Barangay Nandin Lopez (NL) is located in the province of Iloilo (Region 6),
municipality of Dumangas. Three major rivers traverse the area and the
barangay is almost bounded by water except for a strip of land that connects it
to the rest of the municipality. About 97% of the 797 hectares of land in the
barangay are occupied by fish ponds and discussions with key informants
suggested that aquaculture and fishing represented the two main sources of
livelihood for the population of 1,359 individuals (as of 2003).

• Barangay Lat-Asan (LA) is located in the province of Capiz (Region 6),

municipality of Pan-ay. It is a small island (46 hectares) which can only be
reached by boat from the barangay of Pawa. Most of the land (30 ha) is
occupied by aquaculture ponds but there are also substantial mangrove areas

183
 Paper 4/5

in the barangay (15 ha). With a population of 680 spread in 139 households
(1999 survey), this is by far the smallest community in our sample.

• Barangay New Buswang (NB) belongs to the municipality of Kalibo, Province

of Aklan, Region 6. It differs from the other four barangays in the sense that
the area has no major waterways (except for small creeks & man-made canals)
and no fishponds. The ponds that used to border the barangay have been
converted to residential lots for the most part following the collapse of the
aquaculture sector due to diseases and lack of access to markets. There are,
however, fishponds remaining in bordering barangays (Old Buswang in
particular). The second particularity of the barangay is that it contains a 20
hectare area of natural and replanted mangroves, which is part of the
Bhakawan project of SEAFDEC. The total population of 8,127 is spread over
223 hectares of land.

Altogether, these five barangays differ in terms of their geographical location,

importance of aquaculture and fishing as economic activities, remoteness, level of
urbanization, and share of area occupied by mangroves. Barangay NB serves as a
reference to get some insights into the poverty situation of coastal communities with
no fish farming industry that make alternative uses of the land (in particular,
developing mangrove-related activities).

The last step of the sampling strategy selected households randomly from an
exhaustive list compiled in each barangay from official records, which sometimes had
to be amended to include recently settled households and households living in remote
parts of the barangay (e.g., migrant caretaker families living in houses built on the
dykes of fish ponds). In the end, 36 households were surveyed in SA, 37 in SK and 25
in each of the three remaining barangays (NL, LA, ND) , for a total of 148. If we
exclude NB, which has no aquaculture, our sample accounts for roughly 10% of the
population of households in barangays SA, SK, NL and LA on which most of our
analysis focuses.

4. Poverty & aquaculture: perceptions

4.1 Defining the context: general perceptions of poverty
Our study reveals that a vast majority of households consider themselves to be poor,
although the data indicates some regional differences. In the province of Pampanga,
around two thirds of the respondents see themselves as poor (75% in SA and 62% in
SK), while in the Visayas region this proportion is even larger (92% in NL and LA,
100% in NB). These regional differences are consistent with the poverty maps
presented above, but the very high levels of perceived poverty incidence observed in
both regions suggest that the coastal communities that we study are also much poorer
than the regions to which they belong. Clearly, poverty is perceived as a major
problem for all the coastal communities under scrutiny, which gives relevance to our
investigation.

As the recent literature on poverty and livelihoods has moved away from single
indicators of poverty to emphasize its multi-dimensional nature, the study attempted

184
 Paper 4/5

to identify how the respondents themselves would define the concept 2 and Table 2
summarizes the results 3. The large number of answers to the question confirms that
poverty is interpreted in very different ways by individuals, although it could also
reflect the open-ended nature of the question. More than half of the respondents
identified having an insufficient income as the key characteristic of poverty, followed
by the inability to purchase enough food to meet the household’s needs, the lack of a
regular income, and, finally, the fact that one is unemployed. The emphasis on food
consumption indicates that people in these communities face absolute, rather than
relative, poverty. It also gives some relevance to the analysis of the potential
nutritional benefits that the literature suggests can be associated with the development
of aquaculture. In a similar vein, the fact that unemployment is identified as a key
characteristic of the poor implies that the potential of aquaculture to reduce poverty is
closely linked to its ability to create jobs in the communities studied here. The results
also indicate, in line with the literature, that poverty should be analysed in a dynamic
rather than static framework, because poor households are identified by their
vulnerability to external shocks. Hence, income poverty is not only related to the low
level of household earnings but also to their variability and the poor are identified by
a substantial number of respondents as being unable to secure a stable/permanent job.

Table 2 suggests further that asset poverty, though present, does not represent the
main dimension of poverty in these three communities. A few respondents associated
household poverty with the inability to own a house, land or durable goods, and the
issue of indebtedness of the poor was also mentioned. Similarly, the lack of social
status of the poor, as indicated by their lack of influence in these communities and
their dependence on others for their livelihoods, was rarely identified as a key
characteristic of poverty. Altogether, the data indicates that in the coastal
communities investigated here, it is the income and consumption dimensions of
poverty that tend to dominate, as opposed to the concepts of asset and social poverty
that have become very popular in the recent livelihoods literature. We also note that
the perceived meaning of poverty seems fairly homogenous across communities.

Table 3 explores the vulnerability of the surveyed households. Weather shocks were
reported by a substantial number of respondents as a cause of crisis but, given that
some of these communities can be flooded for several weeks a year during the rainy
season, it is rather surprising that typhoons and floods were not mentioned more
frequently. It can therefore be postulated that these events are regarded as fairly
normal by many in these communities, and that, as such, proper coping strategies are
in place to overcome them. By contrast, illness or the loss of a job within the
household are identified as much more important sources of vulnerability. This
finding is consistent with the results reported previously in Table 2 and it appears that
a key characteristic of the poor is their heavy dependence on their ability to work,
even in the short term. However, the increased vulnerability resulting from the loss of
a job is clearly more pronounced in Region 3 than Region 6. This could reflect the
fact that salaried employment in region 3 is simply more widespread than in Region 6;
alternatively, it could also be the case that, because wages are notoriously lower in
Region 6 than in Region 3 (roughly PhP 100/day in Region 6 compared to PhP
150/day in Region 3), employment might not guarantee immunity from poverty in
2
The exact question was: In your opinion, what does it mean for a household to be poor? What is the
main characteristic of poverty?
3
From hereon, all tables are presented at the end.

185
 Paper 4/5

that region. The vulnerability to illnesses is easily understood as resulting from the
combined effect of a loss of income due to the inability to work and the additional
drain on household income associated with the cost of medicine and health care,
which was repeatedly mentioned by respondents as an important problem.

The perceived causes of poverty are presented in Table 4 and, once again, confirm
that the key problem for the poor is one of securing a stable job. Note also that low
salaries are perceived as a much more important determinant of poverty in the three
barangays of Region 6 than in SA and SK, which reflects the regional differences in
wages mentioned previously. However, our results also indicate that almost a fifth of
all respondents identified individual behaviour as an important explanatory factor of
poverty, and that for more than half, these respondents considered themselves to be
poor. The situation is therefore not one where only the rich perceive the poor as
deserving their predicament. Personal indolence, laziness and flawed ‘personal
character’ were mentioned most frequently as leading causes of poverty, and there
were repeated suggestions that ‘vices’, meaning drinking and gambling, were
rampant in all five communities. Hence several respondents indicated that expenditure
on alcohol and gambling was often taking priority over the satisfaction of the
household’s basic needs. Overpopulation is regarded as an important cause of poverty
only in SK and NB, which are also the barangays with the highest population
densities. The respondents provided many other possible explanations for poverty,
including the scarcity of fish in the wild and the lack of education of the population,
but these explanations were not mentioned very frequently.

In the face of important shocks, households have developed a whole range of coping
strategies that are summarized in Table 5. In all five communities, the main response
to a crisis is to seek help from a large support network, corresponding primarily to the
extended family, but which can also include friends, neighbours, local politicians and
employers. Borrowing money forms the second most important coping strategy and
was mentioned by more than 40% of respondents. This strategy relies almost
exclusively on the informal credit market, with most of the loans originating from
local money lenders, local stores, employers, and relatives. However, it is also worth
noticing that close to a third of respondents simply do not have any coping
mechanism, and are left to endure shocks by reducing consumption, which is certain
to have a large negative effect on the welfare of their households. Again, this is
interpreted as indicating the presence of absolute poverty in these communities. A
substantial number of households also rely on their own industriousness to cope with
crises, by simply working more, or, rather surprisingly, starting a small family
business (such as a small food or retail store).

The respondents were also asked to identify potential means of reducing poverty
within their communities, and the results are reported in Table 6. Rather disturbingly,
almost a fifth of the respondents consider that such means simply do not exist and that
poverty within their communities is all but inevitable. This fatalistic attitude is often
supported by the argument that the situation of these communities has not improved
in the recent past, with some respondents indicating that the younger generation is not
better educated than the older one, that fish is becoming scarcer, while families are
getting larger. This was interpreted by many as indicating that past measures taken to
tackle poverty, such as investments in family planning and education, have not
worked, and that there is little reason to assume that it will be any different in the

186
 Paper 4/5

future. On a more positive note, however, 39% of respondents consider that

improving individual behaviour represents an important avenue to reduce poverty,
which suggests that some of the solutions exist within these communities. In line with
the discussion on the causes of poverty, ‘improved behaviour’ includes the reduction
of vices, an increase in the number of hours worked, or more careful budgeting at
household level (that would allow the accumulation of savings). Creating more job
opportunities also ranks high as a method of reducing poverty within these
communities, as might have been expected from answers to the other questions. It is
interesting to note that the government is not perceived as being central to poverty
reduction, although 14% of respondents mentioned its role. This could be interpreted
in several ways, but might suggest that most in these communities consider the central
state to be either weak or corrupt.

Altogether, we conclude that the communities under scrutiny consider poverty to be

widespread and of major importance. Poverty, which is defined primarily in terms of
its income and consumption dimensions by the respondents, is perceived as relating
primarily to the issue of unemployment, but individual behaviour is also recognized
as a major contributing factor. The main sources of vulnerability (illness & loss of
job) are dealt with by mobilizing an extensive support network and borrowing from
the informal credit market, although a substantial number of households simply have
to reduce consumption. A majority of respondents believe that poverty can be
reduced, primarily through job creation within the community, as well as improved
individual behaviour. All of these characteristics define the context in which we now
analyse the relationship between aquaculture and poverty.

4.2 Aquaculture & poverty: the views from five communities

Aquaculture development has often been criticised for the inequities that it apparently
generates (Alauddin and Tisdell, 1998; Coull, 1993) and has even been blamed for the
marginalisation of and increased unemployment in some coastal communities
(Primavera, 1997). It therefore seems relevant to identify how aquaculture is
perceived not only by those directly involved in it (i.e., the fish farmers) but by the
entire community. The survey first asked respondents to evaluate whether, in their
view, aquaculture benefited the rich and/or the poor in their community and the
results are summarized in Table 7. A very large majority considers that aquaculture is
mutually beneficial to the poor and the rich, but there is also a small minority (23%),
particularly in Region 6, that believe that only the rich benefit. The regional
difference could be explained by the fact that land ownership in the Visayas region is
typically more inequitably distributed than in Central Luzon, and that the salaries
offered to caretakers and daily workers are also noticeably lower in Region 6 (see
section 5.5).

We then investigated what form the benefits from aquaculture to the poor might take
by asking whether the poor themselves practiced fish farming. Informal discussions
with aquaculture experts and local officials seemed to indicate that fish farming in the
brackish water areas of the Philippines was mainly a rich-man activity, but this
contention is not supported by the results of our survey. Indeed, Table 8 reveals that
more than half of respondents consider that some poor people operate fish farms. The
interviewees repeatedly mentioned examples of individuals who, having started as
caretakers, managed to acquire small fishponds and, from thereon, developed
profitable aquaculture operations. There is therefore some level of social mobility

187
 Paper 4/5

within the studied communities, and aquaculture might be regarded as an instrument

of that mobility. Further, it seems that escape from poverty via aquaculture is only
possible when the caretaker earns a substantial share of the farm’s profit, as is often
the case in Region 3, but much more rare in Region 6.

However, there are also some clear barriers to entry into the sector, as indicated by the
fact that 82% of respondents consider that it would be impossible for them to start a
fish farming operation (Table 9), although the problem seems more severe in Region
3 than in Region 6. When probed further about why the poor might not be able to start
a fish farming operation, the third of respondents who had expressed that view
overwhelmingly mentioned the lack of access to credit and financial capital as the key
hurdle, while a few (8) also thought that access to land was a problem. The lack of
access to credit is also by far the main reason given by the respondents who felt that
they would be unable to start their own fish farm, although lack of knowledge or land
were also mentioned. Hence, it appears that financial capital represents the scarce
factor in these communities that limits entry into the fish farming industry. This is
explained by the fact that the type of aquaculture practiced in these communities
requires the purchase of large amounts of intermediate inputs (prawn fry and
fingerlings in particular) and also by the level of risk involved. Flooding of fish ponds
is a frequent occurrence, as is mass mortality (particularly of prawns), so that the
returns to invested capital are highly uncertain (many respondents made a parallel
between fish farming and gambling). As a consequence, only those in a strong
financial position are willing to take the risk associated with brackish water
aquaculture.

Because part of the literature suggests that aquaculture can be detrimental to some
particularly vulnerable social groups, we did explicitly ask respondents whether, in
their opinion, fish farming might have a negative impact on the poor (Table 10). More
than two thirds of the interviewees thought that it was not the case. The large minority
(30%) that disagreed usually believed that aquaculture had a negative impact on
fishing, which represents an important source of livelihood for the poor. The blame
was put primarily on the feeds used in large quantities to grow milkfish, and a few
individuals, particularly in barangay Lat-Asan, also suggested that the chemicals
employed to fight diseases as well as the pesticides used between cycles were
responsible for the observed decline in wild fish stocks. Note, however, that
perceptions of these negative impacts vary widely across communities, in a way that
is consistent with the importance of fishing as an economic activity. In particular,
aquaculture is perceived as being most detrimental to the poor in the two barangays
supporting a large number of fishermen (SK and LA). Further, barangay SK was
purposely selected on the outer edge of the Pampanga estuary, hence downstream
from most fishponds and relatively more exposed to the negative externalities
generated by aquaculture than the other barangays. This may explain why a majority
of respondents in SK thought that aquaculture had a negative impact on the poor.

In view of the previous set of results, it is not surprising that aquaculture is perceived
positively by an overwhelming majority (95%) within these five communities (Table
11). The number one perceived benefit is the creation of jobs that are crucially needed
in these communities. More details about the employment generated by aquaculture
will be presented in section 5.5. The second major perceived benefit from aquaculture
for these communities is the provision of fish for human consumption. In particular,

188
 Paper 4/5

the practice of allowing the collection of ‘free fish’ from fish ponds, i.e. any residual
fish or crustaceans left after the main harvest, was mentioned as an important benefit
from aquaculture by a large number of respondents in Region 3. As the gains from
such an activity are rather limited, it is mainly the young and the poor who partake in
it, which is likely to substantially improve nutrition in their households. Many
respondents also emphasized the importance of payments in fish that are made for
various tasks (harvest for instance), as well as the traditional practice by farm
operators to give fish as gifts at the time of harvest to neighbours, friends and family.
Hence, there appears to be important nutritional benefits to these communities from
aquaculture, which materialize through a variety of non-market mechanisms that
make fish available to residents. Some other indirect benefits from aquaculture to
these communities were mentioned but only infrequently, including the generation of
local tax revenues, or the provision of credit by farm operators.

The last section of the questionnaire investigating perceptions of the relationship

between aquaculture and poverty deepened the analysis by attempting to get a sense
of the magnitude of the perceived benefits from aquaculture as well as related
problems. This was achieved by asking respondents to choose a step on a ladder,
presented to them on a piece of paper, after explaining that the first step was
describing the worst possible situation and the highest step the best possible situation
(see Questionnaire in Appendix) 4. The ladder presenting ten steps, admissible answers
range from 1 to 10, and Table 12 presents the mean scores together with their standard
deviations for a series of question.

The first two questions were aimed at evaluating the potential nutritional benefits that
aquaculture could generate in these communities. Respondents were asked to evaluate
their overall food security situation by choosing a step on the ladder, where the first
step was described as a situation where all household members barely have anything
to eat, and the highest step as a situation where every member of the household eats
three nutritious meals daily. The average score (5.05) indicates clearly that a majority
of households feel far from fully food secure, and the table also indicates that the
situation is worse in Region 6, as was expected from the poverty maps (Figure 2).
Worryingly, the results also suggest that the food security situation of these three
communities has slightly worsened over the last five years, but respondents expect
this trend to be reversed in the near future. The second question 5 allows us to quantify
the importance of fish in the diet of these communities and the relatively high average
score (5.80) confirms that fish is an important food item in most households. Further,
there is little indication that availability of fish in these communities has changed in
the last five years, and respondents do not anticipate major changes at that level in the
short future either. Altogether, we interpret this set of results as indicating that
aquaculture could potentially make an important contribution to the improvement of
nutrition in these communities, since most households are not fully food secure, while
fish plays an important role in their diets.

4
We are thankful to Dr Peter Edwards for suggesting the use of this tool.
5
The exact question is: ‘The first step on the ladder shows a situation where fish/seafood does not form
part of your household’s diet at all, even if your household members wanted to eat fish. The highest
step on the ladder represents a situation where your household members can eat as much fish as they
want.’

189
 Paper 4/5

Given the crucial role that employment plays for the poor, it is interesting to analyse
how respondents judge the importance of aquaculture as a source of jobs for their
household 6. The average score of 3.88 establishes that aquaculture represents an
important source of employment in these communities, although the situation varies
across barangays. In region 3, households are much more dependent on aquaculture
for employment in SK (average score of 7.53) than in SA (average score of 3.75). We
believe that this simply reflects the relative remoteness of barangay SK compared to
SA, which is located close to the city centre where various forms of employment are
potentially available. In Region 6, aquaculture is a significant source of employment
in barangay NL (average score of 5.96) but appears unimportant in the two other
barangays. This was expected in NB, which was selected as a reference community
with no fish farming industry, but not in LA, where we thought that the community’s
economy would be equally divided between fishing and aquaculture.

The data also reveals that in barangays SK and NL, very few households (6 out of 62)
replied that none of their employment was related to aquaculture, while almost half of
respondents gave that answer in SA. Hence, in three of the barangays under study,
involvement of the household in aquaculture represents the rule rather than the
exception. Finally, the last question was aimed at measuring the overall importance of
aquaculture as a source of household cash income and the results closely mirror those
obtained with regard to employment. This gives a final confirmation that in these
communities, income and employment are closely related to each other.

Altogether, this part of the survey draws a fairly clear picture of poverty in these
communities and how it relates to aquaculture. There is little doubt that most
households are poor, a situation that is defined primarily by low and variable income
and consumption, as well as some degree of food insecurity. The main coping
mechanism during crises consists in requesting help from an extensive support
network and taking loans from the informal credit market. Although unemployment is
identified as the main cause of poverty, many respondents also consider that its
reduction requires improvement in individual behaviour. In this context, aquaculture
is perceived overwhelmingly as being beneficial to the entire community and, more
specifically, to the poor. Although ownership of fish farms might be very
concentrated in the communities under scrutiny, most households derive some
employment from aquaculture and the sector is identified as a key source of income.
In addition, distribution of ‘free fish’, which is a by-product of the farming activity, is
important to a large number of respondents and one can logically suggest that this
benefit is particularly pro-poor. Contradicting a large volume of recent literature, few
negative effects of aquaculture on these communities are identified by their residents.
There are suggestions that the industry imposes some negative externalities on the
fishing industry, but these costs are judged to be largely outweighed by the benefits of
increased employment and increased availability of fish from aquaculture. However,
there are important barriers to the adoption of aquaculture by the poor, most notably
the lack of access to credit, but the study also reveals some level of social mobility
within these communities, sometimes facilitated by aquaculture itself.

6
The exact question is: ‘The first step on the ladder describes a situation where no household member
earns from any aquaculture-related activity. The highest step on the ladder represents a situation where
members of your household obtain all their earnings from various aquaculture-related activities
(production, marketing, harvesting, processing, input supply etc.).’

190
 Paper 4/5

5. Levels and Determinants of Poverty and Inequality in the Selected

Communities
5.1 Approach to Poverty Measurement
The steps involved in measuring and explaining poverty are well understood. In a first
step, an indicator of household welfare or well-being is constructed, and that indicator
can then be compared to a poverty line to identify the poor (Ravallion, 1996). Once
the poor and the non-poor are distinguished, the correlates of poverty are analysed in
an attempt to explain why some households are poor, and draw policy conclusions on
how to facilitate escape from poverty.

Although the underlying logic is straightforward, each step of the methodology is

unfortunately fraught with problems. A fundamental issue arises from the fact that
well-being is not directly observable, so that a proxy needs to be defined to measure
household welfare. Total current income is a popular choice in poverty studies, but it
is also recognised that it suffers from important theoretical shortcomings (Balisacan,
1999). Most importantly, current income does not constrain consumption when
households can borrow or use savings, which is the rule even in low-income countries
where financial markets fail. In fact, the previous section identified borrowing as a
key coping mechanism of households in the study areas and Fafchamps and Gubert
(2003) showed how most informal loans within social networks in the Philippines
attract no interest rate, thus making a debt trap impossible. It follows that
consumption is theoretically a better welfare indicator than income. A more
fundamental critique of the standard approach to poverty measurement, which was
first formulated by Sen (1985), considers that income and consumption indicators are
too limited as concepts of welfare, and that it would therefore be more appropriate to
replace them with various social indicators, such as life expectancy, literacy and
infant mortality. In a similar vein, the recent livelihoods literature emphasizes the
multi-dimensional nature of poverty and the need to analyse it at a disaggregated
level.

For this particular study, we note the previous set of limitations and choose to adopt a
traditional approach to poverty measurement that we complement with additional
indicators of household welfare, such as access to non-market goods, as suggested by
Ravallion (1996). Our initial intention was to derive a consumption-based index of
welfare, but the pilot survey revealed numerous difficulties in collecting the necessary
data: respondents were often unable to recall what they had consumed or were getting
impatient/annoyed with the dozens of questions on consumption of individual items.
By contrast, and contrary to what the literature suggests (Ravallion and Chen, 1997;
Lipton, 1997), it proved much easier to obtain information on household income,
which most respondents were happy to share. We therefore rely for this particular
study on an income-based poverty measure. Although practical considerations drove
this methodological choice, we feel warranted in our approach by the results in
Section 4.1, which established that income represents the main dimension of poverty
in the eyes of a large majority of respondents in all five communities.

Our income measure includes earnings accruing to all household members from
salaried employment, self-employment in activities such as fishing or retailing, rents
of physical assets (land, houses, tri-cycles, boats etc.) as well as transfers from the
state and individuals. A common practice in the study areas consists of making
payments in nature (principally in the form of rice or fish), and their values were

191
 Paper 4/5

imputed on the basis of the prevailing market prices for these commodities. Further, it
is clear that a household’s needs depend on its demographic composition so that total
household income should be adjusted accordingly. Conceptually, the construction of
so-called equivalence scales is relatively simple as it should mainly take into account
two key factors: first, the number of children in the household, because a child’s
needs are likely to be substantially less than an adult’s 7; and second, overall
household size, because the presence of fixed household costs is likely to cause
economies of scale in consumption. Empirically, however, the problem is complex
and has generated an extensive literature (Banks & Johnson, 1994; Coulter et al.,
1992) from which no consensus emerges (Balisacan, 1999). In this context, we simply
define our welfare measure as household income per capita, which is clearly
preferable to the use of unadjusted household income.

The next step in the analysis consists in comparing the income-based welfare
indicator to a reference level in order to identify poor households, but, here again, the
construction of a poverty line raises a number of additional issues8. A common
practice involves setting the poverty line as a constant proportion of the mean income,
but, because the analysis then loses meaning in terms of absolute standards of living,
it is unlikely to be of much relevance to anti-poverty policies (Ravallion, 1996).
Clearly, in a country such as the Philippines, it is absolute poverty that matters, and it
is preferable to build a poverty line interpretable in terms of the subsistence needs of
the population. In the present context, we simply rely on the official poverty line,
reported online by the Philippines National Statistics Office (PNSO) for individual
regions in year 2000 9, which we adjust for inflation by using the national Consumer
Price Index (CPI) also reported on the PNSO website. The poverty line is defined by
the National Statistical Coordination Board (NSCB) 10 as the annual per capita food
threshold plus the cost of other basic non-food requirements, and hence clearly relates
to absolute poverty. For Central Luzon, the poverty line in year 2000 was PhP14,200,
and the CPI increased from 152.1 in July 2000 to 183.3 in July 04, when the survey
took place. The adjusted poverty line for Region 3 is therefore set at PhP17,113, or
$305 at the current exchange rate. The corresponding value in Region 6 is PhP14,703,
or $262 at the current exchange rate. The PNSO also reports a food threshold, defined
by the NSCB as the annual per capita cost of basic food requirements which meet
100% adequacy of the recommended dietary allowance for protein and energy and
80% of all other nutrients. This measure of survival needs was PhP9,183 nationally in
2000, which translates to PhP11,067 in 2004 after taking into account the effect of
inflation.

5.2 Levels of Poverty

Table 13 presents summary measures of income levels in the five communities under
scrutiny. Mean household income in our sample, expressed on an annual and per
capita basis, is equal to PhP18,889, while the median amounts to PhP12,925, and the
7
Balisacan (1992) establishes for the Philippines that a rural couple with one child needs 20% more
income than a childless household to achieve the same welfare level. However, it is also true that food
need per unit of body weight is higher for children than adults; and that children need food of higher
quality (in terms of energy density and protein balance) than adults. We are grateful to Dr Peter
Edwards for that last point.
8
In fact, some authors consider that the exercise introduces so much arbitrariness into the analysis that
the poverty line should simply be set to plus infinity (Deaton, 1996).
9
The address is www.census.gov.ph/data/sectordata/2000/ie00pftx.html.
10
http://www.nscb.gov.ph/ru8/default.asp

192
 Paper 4/5

standard deviation to PhP21,892. These figures do not describe the entire distribution
of income very well but, nonetheless, suggest, when compared to the poverty lines
calculated for each region, that poverty levels in the studied communities are
substantial and that inequalities within these communities are relatively large as well.
The table also reveals some differences across barangays. While the three
communities SK, SA and NL appear relatively similar in terms of income levels,
barangay LA is much poorer (mean income of only PhP9,379) and NB stands out
from its particularly unequal distribution of income, as revealed by a relatively large
mean (PhP23,673) but a very low median (PhP11,742) and an extremely large
standard deviation (PhP40,344).

Going further in our assessment requires the choice of an aggregate poverty measure
and the most popular one is the headcount index, defined as the percentage of the
population below the poverty line. The index is used mainly because of its simplicity
of interpretation but suffers from severe theoretical shortcomings (Ravallion, 1996),
the main issue relating to its invariance with respect to a change in the distribution of
income among the poor. For instance, a clearly undesirable evolution where all poor
people become poorer while the non-poor maintain their income levels does not result
in any change in the headcount index. The issue has also relevance when evaluating
policies because, for instance, if a project manages to raise the income of the
extremely poor, but not enough to lift them above the poverty line, success will not be
reflected in the head-count measure of poverty.

These problems led researchers to develop alternative ‘distribution-sensitive’ poverty

indicators, such as the poverty gap, which is the mean distance separating the
population from the poverty line, with the non-poor being given a distance of zero. It
measures the poverty deficit or depth of poverty of the population, i.e. the resources
that would be needed to lift all the poor out of poverty through perfectly targeted cash
transfers. Mathematically, it is defined as:
1 q ⎡ z − yi ⎤
PG = ∑
n i =1 ⎢⎣ z ⎥⎦
(1)

where n denotes the size of the population, yi the income of individual i, q the number
of poor people and z the poverty line. Even though this measure represents an
improvement over the headcount index, it has some restrictive properties, in particular
the fact that, when judging the impact of an increase in income of a poor person, the
distance of that poor person’s income from the poverty line is not taken into account
(i.e., it does not matter how poor that person is). If it is believed that society or policy
makers should place a larger weight on the welfare of extremely poor people, the
squared poverty gap, which is often described as a measure of the severity of poverty,
might be preferable. It is defined mathematically as:
2
1 q ⎡ z − yi ⎤
P2 = ∑
n i =1 ⎢⎣ z ⎥⎦
(2)

Hence, it can simply be interpreted as a weighted poverty gap, where the weights are
calculated as the distance of each poor person’s income to the poverty line. We note
in passing that all three aggregate measures of poverty relate to each other in the sense

193
 Paper 4/5

that they represent special cases of a class of indicators first proposed by Foster et al.
(1984) defined as 11:
α
1 q ⎡ z − yi ⎤
P (α ) = ∑ ⎢ (3)
n i =1 ⎣ z ⎥⎦
The aggregate poverty measures are reported in Table 13, which indicates that 59% of
the sample population falls below the poverty line. Hence, the incidence of poverty in
the five communities is clearly high and exceeds by far the national average, as the
headcount reported in Table 1 for the whole of the rural areas of the Philippines is
only 40.1%. It was expected to find a higher poverty incidence in Region 6 than in
Region 3, but that is not the case. In fact, barangay NL has the lowest poverty
headcount, while barangay LA has the highest one, and both are located in Region 6.
More relevant than regional differences might be the remoteness of the communities,
because barangays SK and LA, which are both relatively difficult to access, have a
relatively high poverty incidence. Table 13 also presents estimates of the poverty gap
and squared poverty gap for each community and the whole sample. It is reassuring to
find that the ranking of the five barangays does not depend on the choice of aggregate
poverty measure . Poverty is the least prevalent in barangay NL, followed by
barangay SA, barangay SK, barangay NB, and, finally, barangay LA. The poverty gap
for the whole sample (0.28) indicates that the depth of poverty is relatively large
among our sample households as it means that eliminating poverty completely in
these communities would require cash transfers amounting to 28% of the poverty line
for every individual. Finally, Table 13 presents a measure of extreme poverty, which
is simply the headcount index calculated not with respect to the poverty line, but,
instead, the food threshold. The sample average of 0.43% confirms that absolute
poverty represents a major problem for these coastal communities. Comparison of the
five barangays once again confirms that poverty is worst in LA, but the ranking is
slightly modified for the three communities with the lowest poverty incidence (SK,
SA and NL).

In conclusion, these aggregate figures indicate clearly that all five communities are
poor, with a high incidence of extreme poverty defined as a situation where
households are unable to satisfy even their most basic needs for food. Barangays LA
and NB stand out as being particularly poor, while no clear regional differences
regarding the incidence of poverty can be established from our sample.

5.3 Is Aquaculture Pro-poor?

Our analysis of the relationship between aquaculture and poverty starts by evaluating
the quantitative importance of fish-farming related activities in generating income in
the three barangays (SK, SA, and NL) where the industry is present 12. Table 14
reports total income from aquaculture for each barangay and the whole sample,
expressed per household and per capita, as well as the aggregate and average shares of

11
The head count index is obtained for α=0, the poverty gap for α=1, and the squared poverty gap for
α=2.
12
Barangay Lat-asan was selected to represent a community where only part of the land was occupied
by fishponds. However, we found that only a negligible income stream was attributable to aquaculture
in that community, in part due to the small area of ponds, and could therefore not include the barangay
in this analysis.

194
 Paper 4/5

aquaculture in household income 13. It is clear that aquaculture represents a

quantitatively important economic activity in all three communities, as the table
reveals that aquaculture generates an income of, on average, PhP27,194 per household
or PhP5,038 per capita. This income stream represents 29% of total income accruing
to the sample households, while the average income share from aquaculture amounts
to 39%. On the other hand, these figures indicate that households in these
communities are also able to diversify their sources of income, and the previous
percentages can be compared to an average non-farm share of rural household income
of 42% in the whole of the Philippines (FAO, 1998). Hence, aquaculture is
economically significant in these coastal communities, but it is one of many important
activities, including fishing, retailing and construction. The table further reveals that
fish-farming is much more important in barangay SK than in the other two areas. This
is explained by the fact that barangay SA is close to the relatively urban centre of
Sasmuan, which offers job opportunities, while workers from barangay NL can find
jobs in agriculture as the barangay is connected to the mainland. By contrast,
barangay SK is relatively remote so that fishing and aquaculture represent the main
economic activities. Indeed, the share of income from aquaculture in SK is on average
58%, so that the activity is obviously an essential source of livelihood in that
community. These results also confirm that the economic environments of the three
barangays differ substantially, which was deemed desirable when setting up the
survey and suggests that our purposive sampling strategy was, from that point of
view, relatively successful.

Next, we investigate the pro-poor nature of aquaculture by distinguishing the income

generated by aquaculture that accrues to the poor and the non-poor. Table 14
establishes clearly that both groups benefit substantially from the activity: a poor
household derives, on average, an income of PhP23,863 from aquaculture, which
translates into PhP3,951 per capita, or roughly a quarter of the poverty line. A non-
poor household benefits even more, with an average income of PhP30,809, or
PhP6,552 in per capita terms. These results are consistent with the opinions expressed
by respondents about the distributional properties of aquaculture (see section 4.2):
clearly, both the poor and the rich 14 benefit from the activity. Furthermore, while the
poor benefit less from fish-farming in absolute terms, they benefit a lot more in
relative terms. Thus, Table 14 reveals that aquaculture accounts for 44% of income
for the poor, but only half as much (23%) for the rich. This is a key result of our
analysis that gives strong empirical support to the idea that brackish water aquaculture
is indeed pro-poor in the Philippines. Further, this conclusion appears robust to the
choice of community and poverty line and can therefore be stated with confidence.
First, with regard to the choice of barangay, the aggregate shares of income from
aquaculture for the poor and non-poor are 34% and 11% respectively in SA, 31% and
12% in NL, while both are equal to 57% in SK. Then, the pro-poor character of
aquaculture becomes even more evident when focusing on the subset of extremely
poor households, i.e. those with an income below the food threshold. For the whole
sample, these extremely poor households derive more than half their income (54%)
from aquaculture, as opposed to only 25% for the remaining households and, here

13
The aggregate share is the sum of aquacultural incomes divided by the sum of household incomes
for the group of households considered. The average share is the arithmetic average of the ratio
aquaculture income/household income for each household in the group.
14
We take ‘rich’ as simply meaning non-poor. Most households in that category are in fact far from
being rich by any standards.

195
 Paper 4/5

again, the same pattern emerges within each barangay. In particular, in SK where
aquaculture benefits the poor and the rich equally in relative terms, the activity
accounts for a massive 71% of income of the extremely poor, as opposed to only 53%
for the remaining households. The corresponding percentages in SA are 43% versus
13%, while they are 42% and 12% in NL.

We pursue this investigation by carrying out a simple experiment that assesses by

how much poverty would increase if the sample households were not receiving any
income from aquaculture. That is, we reproduce the poverty evaluation of section 5.2
by replacing total household income by non-aquacultural income and the results are
presented in aggregate form in Table 15. Concentrating on the poverty measures, it is
clear that poverty would increase substantially to reach very high levels in all three
communities, and that this conclusion does not hinge on the choice of index or
poverty line. The headcount is simulated to rise from 54% to a massive 70%,
indicating that more than two thirds of households would be poor in the absence of
aquaculture. The poverty gap almost doubles from 24% to 47%, which means that
eliminating poverty in the absence of aquaculture would require perfectly targeted
cash transfers amounting to almost half of the poverty line for every member of these
three communities. Finally, the squared poverty gap would almost triple from 0.14 to
0.39. The fact that the relative increase in the squared poverty gap exceeds that of the
poverty gap, which is itself larger than that of the headcount index, reveals that
eliminating aquacultural income would represent a particularly regressive change, or,
in other words, that it would have a particularly detrimental impact on the extremely
poor. This interpretation is confirmed by the observation that the increase in the head
count index is larger when calculated at the food threshold (19%) than when
calculated at the poverty line (16%). Of course, the above simulation represents an
over-simplification of reality because, if aquaculture was to disappear from a
particular community, individuals deriving income from the sector would be able to
reallocate labour and assets to other sectors to generate alternative income. Hence, the
above figures represent upper bounds of the likely impact of the disappearance of
aquaculture on poverty. Yet, we believe that in reality there would be major obstacles
to such a reallocation of household resources, as the Philippines are usually described
as a ‘labour surplus economy’ where unemployment and under-employment represent
important problems. This view was also shared by most of the respondents, as they
identified the lack of jobs as the main cause of poverty in their communities. Further,
the previous simulation reinforces the conclusion that aquaculture benefits the poor
and the extremely poor disproportionately.

5.4 Measuring and Explaining Income Inequality

The extent of poverty in a particular group of households is simply a function of mean
income and the distribution of income within that group. Hence, there is an obvious
relationship between income inequality and poverty, and the pro-poor nature of
aquaculture depends in large part on how the income generated by the sector is
distributed among households. This motivates our investigation of inequality in the
three study areas and how it relates to aquaculture. For this purpose, we rely on a
large literature that has developed adequate measures of inequality as well as ways of
attributing total inequality to different income sources. The analysis starts with the
Lorenz curve, which plots the cumulative percentage of the population to the
cumulative share of income they control and hence provides a complete geometric
characterization of the distribution of income (Sadoulet & de Janvry, 1995). For the

196
 Paper 4/5

98 households in barangays SA, SK and NL, the Lorenz curve is depicted by the blue
line in Figure 3. The Gini coefficient, which represents the most popular summary
measure of income inequality, is then defined as the area between the Lorenz curve
and the first diagonal (pink line in Figure 3), expressed as a percentage of the total
area below the first diagonal. Admissible values therefore range from zero to unity,
with larger values of the coefficient indicating larger levels of income inequality.

100
Cumulative Household

80
Income (%)

60

40

20

0
0 20 40 60 80 100
Cumulative Households (%)

Figure 3: Lorenz Curve

The Gini coefficient takes different mathematical expressions and we opt for that
proposed by Pyatt, Chen and Fei (1980):
2
G= _
cov( y , r ) (4)
ny
where yi and ri denote the income and rank of household i when the population is
_
ordered by increasing income, n is the total number of households, y is the mean
population income, and cov(.) denotes the covariance operator. In the case of our 98
households, the Gini coefficient takes a value of 0.36, which is indicative of
significant but modest inequalities. However, when this level of inequality is applied
to a mean income that is marginally larger than the poverty line (Table 13), it results
in a large number of households being poor.

A very useful property of the Gini coefficient is that it can be broken down according
to each particular source of income, which can then be compared in equity terms. The
approach was pioneered by Stark, Taylor and Yitzhaki (1986), who investigated the
effect of remittances on inequality in two Mexican villages, and derived the following
expressions (Sadoulet and de Janvry, 1995, p. 22):
G = ∑ S s Rs G s (5)
s
where Ss denotes the share of source s in total income, Gs is the Gini coefficient of the
sth source of income, and Rs denotes the Gini correlation coefficient between income
source s and total income expressed as 15:

15
By analogy with the previous notations, rs is the rank of household i when the population is ordered
by increasing income from source s, denoted ys.

197
 Paper 4/5

cov( y s , r )
Rs = (6)
cov( y s , r s )
Expression (5) is fairly intuitive as it states that the effect of income source s on the
level of inequality is a function of three factors:

• The relative importance of income source s in total income Ss. Clearly,

a source of income accounting for a very small share of total income
can only have a minor impact on overall inequality.

• The distribution of income from source s among all households, as

measured by Gs. If only a few households derive a large income from
source s (large value of Gs), that source of income will tend to increase
overall inequality in the community.

• The correlation between income from source s and total income across
households. A low level of correlation indicates that households
deriving a relatively large (small) income from source s are not
necessarily rich (poor), which therefore tends to reduce overall
inequality.

The analysis can be pursued to investigate whether a particular source of income

increases or decreases inequality in a group of households. First, it is important to
notice that quantities Ss and Gs are both positive and smaller than unity, while Rs can
take values in the -1 to +1 range. It is therefore clear from equation (5) that a negative
Gini correlation coefficient Rs implies that income source s unambiguously reduces
inequality. To determine the overall effect of income source s on inequality when Rs
is positive, it is useful to re-write the Gini decomposition as (Sadoulet & de Janvry,
1995):
Rs G s
∑ Ss G
=1 (7)
s
This expression can be interpreted as a weighted average of each source of income’s
‘concentration coefficient’ Rs G s / G and indicates that a source of income is inequality
increasing (decreasing) if and only if this coefficient is greater (smaller) than unity.

The above formulae help characterize the overall inequality effect of a particular
income source, but it is also interesting to determine the marginal effect, i.e. whether a
small change in income source s would increase or decrease inequality. This is
motivated by the observation that most policies aim at changing the magnitude of an
income source rather than removing it completely or creating it where it did not
previously exist. Stark, Taylor and Yitzhaki (1986) derived the change in the Gini
coefficient as a result of one percent increase in income from source s as:
∂G
= S s ( Rs G s − G ) (8)
∂ ln y s
or, in relative terms:
∂ ln G R G
= S s ( s s − 1) (9)
∂ ln y s G

198
 Paper 4/5

These expressions imply that, at the margin, income source s is inequality increasing
(decreasing) if and only if the concentration coefficient for that income source is
greater (smaller) than unity.

The Gini decomposition was used to investigate the impact of aquacultural income on
inequality in each barangay as well as the whole sample (Table 16). Focusing on the
aggregate results first, while aquaculture represents almost a third of household
income, it accounts for less than 3% of the total Gini coefficient of 36%, with non-
aquaculture therefore accounting for more than 33%. In other words, only 8% of
overall inequality is attributable to aquaculture, while 92% is attributable to other
income sources. This limited impact of aquaculture on inequality occurs in spite of the
fact that aquacultural income is, on the whole, relatively unequally distributed: its
Gini coefficient is 66%, as compared to 57% for income unrelated to fish farming.
However, consistent with expression (5), the result is explained primarily by the fact
that the Gini correlation coefficient for aquaculture is positive but very small, at 5%,
as compared to 90% for non-aquaculture.

This decomposition is therefore extremely useful in understanding the impact of

aquaculture in these coastal communities. First, the relatively large Gini coefficient
for aquaculture simply reflects the fact that a substantial number of households derive
no income from this activity. However, the key result relates to the Gini correlation
coefficient, which indicates that there is little relation between total household income
and aquacultural income. This means that relatively rich (poor) households are not
much more likely to derive large (small) incomes from aquaculture than poor (rich)
households. We are therefore left, once again, with the conclusion that both poor and
rich people benefit substantially from aquaculture in these communities.

From the previous set of results, it should come as no surprise that the overall effect
of aquaculture is to decrease inequality in these communities, as indicated by a
concentration coefficient of 0.11, which is clearly smaller than unity (see equation
(7)). Hence, aquaculture is a more equitable source of income than the available
alternatives taken together. The inequality reducing nature of aquaculture is also
apparent at the margin: a one percent increase in aquacultural income in these
communities results in a decrease in the Gini coefficient of 0.08, or more than 21%
when expressed in relative terms. This supports the view that aquacultural growth has
a strong levelling effect on the distribution of income in these communities.

The previous set of conclusions applies broadly to each barangay taken individually,
although some interesting nuances are also evident. Most remarkably, in all three
communities the decomposition establishes that aquaculture reduces inequality both
overall and at the margin, which is an indication of the robustness of our results.
Hence, the three concentration coefficients are smaller than unity at 0.11 in SA, 0.81
in SK and -0.41 in NL. In all three communities, the inequality reducing impact of
aquaculture is explained primarily by the low or negative correlation between total
household income and fish-farming income. Barangay NL stands out in this regard
because of its negative Gini correlation coefficient for aquaculture, which indicates
that, on average, the poorer a household in that community, the larger the income that
it derives from fish-farming activities. It follows that aquaculture is particularly pro-
poor in NL, which is consistent with the observation reported in Table 16 that a one
percent change in fish-farming income in that community decreases the Gini

199
 Paper 4/5

coefficient by a massive 9% (the corresponding values in SA and SK are 6% and 3%

respectively). Finally, the table also reveals that the distribution of aquacultural
income varies substantially in the three communities: the Gini coefficient for that
source of income is larger than that for other sources of income in SA and NL but
smaller in SK. We believe that this simply confirms our interpretation, stated above,
that the main determinant of the Gini coefficient for aquacultural income is the
number of households deriving no income from the sector. In SA and NL, this number
is relatively large, hence justifying the large Gini coefficients.

5.5 Discussion
The quantitative analysis presented in this section suggests that aquaculture in the
study areas represents a source of income which is both pro-poor and inequality-
reducing. This result deserves an explanation which we now seek through an
investigation of how the sample households derive income from aquaculture. The data
reveals first that very few owners (or operators) of fish farms are represented in our
random sample and we interpret this finding as indicating that the industry is
concentrated in the hands of a few individuals who tend to live away from the
communities where production takes place. In spite of this feature, aquaculture in the
coastal areas of the Philippines represents an important source of employment through
the direct and indirect demand for labour that it generates. Hence, more than half of
the households in barangays SK, SA and NL are involved in at least one aquaculture-
related activity. Table 17 gives additional details about the income and employment
generated by aquaculture in the three barangay and shows that a large number of
economic activities are related to the operation of fish farms:

• Labourers are hired on a daily basis for a wage of approximately PhP150 in

Region 3 and PhP100 in Region 6. They usually carry out maintenance tasks
related to the fishponds, and most importantly the consolidation of dykes,
which involves taking mud by hand from the bottom of the pond and applying
it where the dykes need reinforcement 16. This operation appears particularly
labour intensive in Region 3, where the dykes need to be high and strong to
resist tides and seasonal floods. Hired workers in Region 3 are also used in
large numbers to remove an invasive weed (local name ‘digman’), which is
thought to be detrimental to the survival of prawns. Table 17 indicates that
46% of the sample households derive some income from the sale of wage
labour, which accounts for more than a quarter of the total aquacultural
income accruing to the three barangays. Hence, the demand for wage labour
generated by aquaculture appears particularly pro-poor because most
households can apparently benefit from it.

• Harvests, which take place two to three times a year, also mobilize a large
number of workers (usually ten to twenty for a ten-hectare pond) for a full
day, at a wage rate of approximately PhP250 in Region 3 and PhP150 in
Region 6, which is, sometime, supplemented by a small quantity a fish.
Almost one in five households in our sample participates in that activity, but
the related income is limited (PhP4,499) because harvests provide at best a
few weeks of employment annually for any given individual.

16
In recent years, mechanical diggers have been introduced but their use remains rare.

200
 Paper 4/5

• ‘Caretaking’, which corresponds to the supervision of aquaculture ponds,

provides employment to 23% of households in our sample. The task is usually
carried out by a permanent employee who lives on the dykes of the fish pond
together with his family. The remuneration of caretakers has several
components, including a base monthly salary (around P4,000 in Region 3 and
P3,000 in Region 6), incentive payments in the form of a percentage of the
harvest, as well as payments in nature (provision of free housing, rice and fish)
but arrangements vary from farm to farm and region to region. For instance,
incentive payments range from zero to 20% of the harvest and are a lot more
frequent in Region 3 than in Region 6. Caretaking is important for the three
studied communities because it provides permanent employment to a
significant number of workers, a rule of thumb being that one caretaker is
usually hired to manage ten hectares of fish ponds. The mean income of
participating households is relatively large (PhP51,354) and the activity
accounts for 44% of total aquacultural income in the three communities.
Notice however that a family of four earning the mean caretaking income
would still fall below the poverty line.

• The data reveals that the collection of shells and molluscs which are used as
feeds in fish ponds represents another important activity directly related to
aquaculture, in which 14% of households partake. It is usually carried out as
an own-account activity but large farms sometime hire full-time workers
solely for the purpose of collecting these ‘natural feeds’. Further, the activity
appears relatively lucrative with the mean income of participating households
amounting to PhP26,621. The importance of this type of feeds is a reflection
of the polyculture and extensive nature of the production systems considered
here, which often makes it uneconomical to use high quality feeds to, say,
grow prawns, when most of the feeds are actually consumed by other species.

• The collection of ‘free fish’, i.e. left-overs after the harvest, appears
qualitatively unimportant as a source of income but it might generate
substantial nutritional benefits to poor households.

• The survey finally reveals that there is a whole range of other activities related
to aquaculture, directly or indirectly, which provide income and employment
to the coastal areas of the Philippines, although they are not listed individually
in Table 17. They include the marketing of feeds, seeds (‘fingerling agents’),
fish, prawns and crabs; the collection of wild fry and fingerlings; boat
transportation of workers, inputs and outputs; and even the construction and
maintenance of boats used in the operation of fish ponds. It is likely that Table
17 underestimates the income stream generated by aquaculture through these
activities because it is often difficult to attribute a particular activity, such as
transportation, solely to aquaculture.

6. Conclusion
This paper has investigated the relationship between aquaculture and poverty from
several perspectives based on a household survey of five costal communities. We
started by examining how this relationship was perceived by the residents of these
communities themselves and pursued with a quantitative assessment of the
determinants of poverty and inequality. The findings can be summarized as follows:

201
 Paper 4/5

• Aquaculture in the studied communities generates a substantial income stream

that benefits both the poor and the rich. This quantitative result confirms the
view expressed by a large majority of households, and the perception of
aquaculture in the five barangays is overwhelmingly positive.

• While the rich benefit slightly more in absolute terms, the poor benefit a lot
more in relative terms and it is in that sense that aquaculture can be considered
pro-poor. Further, the pro-poor nature of aquaculture is evident in different
economic environments and only increases when the poverty line is lowered.
In particular, the group of extremely poor households that struggles to meet
even its basic food needs derives more than half of its income from
aquaculture.

• A Gini decomposition exercise established unambiguously that aquaculture

represents an inequality-reducing source of income in these communities. This
is explained primarily by a low level of correlation between total income and
aquaculture income. In other words, it is not necessarily richer households that
derive large incomes from fish farming in these communities.

• Our study does not support the view, present in the literature, that aquaculture
contributes to the marginalisation of the poor. Some residents in coastal
communities are aware of possible negative impacts of aquaculture, most
notably on fishing, but consider them to be more than offset by job creation in
aquaculture.

This set of results might seem surprising at first as brackish water aquaculture in the
Philippines is usually considered a rich-man activity. We believe that it is explained
primarily by the fact that, while the industry remains relatively concentrated in the
hands of rich owners/operators, it is still generating a large demand for relatively
unskilled labour. In the context of communities where the primary cause of poverty is
the lack of employment opportunities, the jobs directly or indirectly related to fish
farming represent an essential source of livelihood for the poor. This also means that
policy makers concerned with developing the sector, if aiming to have an impact on
poverty, should pay attention to the employment effects of new policies and
technologies. While intuitive, this recommendation contrasts with the emphasis that is
usually put on production and land productivity growth in the debate about
aquaculture development in developing countries.

202
 Paper 4/5

Table 2: Meaning of Poverty - Number of occurrences of each answer and related percentages

No access To be
Unable to No
No No to To be indebted/
Insufficient Irregular No purchase No own To be No influence Too many
Barangay permanent durable education/ dependent have to
Income Income Job enough house landless Savings in the children
Job goods health on others borrow to
food community
care get by

SK 17 13 3 8 15 0 3 2 6 1 0 1 2 3
46% 35% 8% 22% 41% 0% 8% 5% 16% 3% 0% 3% 5% 8%
SA 18 5 11 4 14 5 2 6 1 2 1 2 1 0
50% 14% 31% 11% 39% 14% 6% 17% 3% 6% 3% 6% 3% 0%
NL 15 7 1 3 6 2 0 1 0 1 1 2 0 1
60% 28% 4% 12% 24% 8% 0% 4% 0% 4% 4% 8% 0% 4%
LA 17 7 0 1 3 1 0 0 1 0 0 0 0 0
68% 28% 0% 4% 12% 4% 0% 0% 4% 0% 0% 0% 0% 0%
NB 12 3 3 6 3 1 1 2 1 2 0 1 0 0
48% 12% 12% 24% 12% 4% 4% 8% 4% 8% 0% 4% 0% 0%

Total 79 35 18 22 41 9 6 11 9 6 2 6 3 4
53% 24% 12% 15% 28% 6% 4% 7% 6% 4% 1% 4% 2% 3%

203
 Paper 4/5

Table 3: Sources of Vulnerability

Flood Drought Typhon Illness Death Job loss Other

SK 3 1 1 17 1 17 5
8% 3% 3% 46% 3% 46% 14%
SA 2 0 4 10 1 11 2
6% 0% 11% 28% 3% 31% 6%
NL 5 0 3 11 1 0 1
20% 0% 12% 44% 4% 0% 4%
LA 4 0 1 8 1 0 0
16% 0% 4% 32% 4% 0% 0%
NB 5 0 0 9 4 0 0
20% 0% 0% 36% 16% 0% 0%

Total 19 1 9 55 8 28 8
13% 1% 6% 37% 5% 19% 5%

Table 4: Perceived causes of poverty

Over- Vices/
Hig Prices Lack of Dependence population/ Laziness/
Low Unemployment/ of Lack of Physical Scarcity of on Fishing/ Too Many Irresponsible
Barangay Salaries No stable jobs Necessities Education Assets Fish Farming Children Behaviour

SK 5 24 6 4 3 3 1 13 12
14% 65% 16% 11% 8% 8% 3% 35% 32%
SA 0 25 1 5 3 7 7 1 6
0% 69% 3% 14% 8% 19% 19% 3% 17%
NL 5 19 0 0 2 0 2 1 5
20% 76% 0% 0% 8% 0% 8% 4% 20%
LA 4 11 2 0 8 9 14 1 0
16% 44% 8% 0% 32% 36% 56% 4% 0%
NB 10 15 2 5 1 0 0 5 5
40% 60% 8% 20% 4% 0% 0% 20% 20%

Total 24 94 11 14 17 19 24 21 28
16% 64% 7% 9% 11% 13% 16% 14% 19%

204
 Paper 4/5

Table 5: Coping Strategies

Reduce
Help from Consumption/ Rely on Start
Work Sell Support Help from no coping Own Family
Barangay More Borrow Assets Network Government mechanism Savings Business

SK 5 22 7 24 0 6 2 4
14% 59% 19% 65% 0% 16% 5% 11%
SA 5 15 1 18 3 6 1 0
14% 42% 3% 50% 8% 3% 3% 0%
NL 1 6 1 12 2 12 0 0
4% 24% 4% 48% 8% 8% 0% 0%
LA 1 7 2 9 2 11 0 0
4% 28% 8% 36% 8% 44% 0% 0%
NB 0 10 2 12 2 13 0 0
0% 40% 8% 48% 8% 52% 0% 0%

Total 12 60 13 75 9 48 3 4
8% 41% 9% 51% 6% 32% 2% 3%

Table 6: How Can Poverty Be Reduced?

Create
Create More Livelihood Improve More Improve Improve
Job Opportunities Individual Investment/ Government Access to Family Develop
Barangay Opportunities for Women Behaviour credit Help Education Planning Aquaculture None

SK 10 4 11 0 2 0 1 4 14
27% 11% 30% 0% 5% 0% 3% 11% 38%
SA 11 0 10 7 2 3 3 0 9
31% 0% 28% 19% 6% 8% 8% 0% 25%
NL 6 0 16 2 6 0 0 1 0
24% 0% 64% 8% 24% 0% 0% 4% 0%
LA 4 0 7 4 6 0 0 0 5
16% 0% 28% 16% 24% 0% 0% 0% 20%
NB 9 0 14 2 4 1 1 0 0
36% 0% 56% 8% 16% 4% 4% 0% 0%

Total 40 4 58 15 20 4 5 5 28
27% 3% 39% 10% 14% 3% 3% 3% 19%

205
 Paper 4/5

Table 7: Does Aquaculture Benefit the Poor

and/or the Rich?

The The
Rich, Poor,
Not the not the
Barangay Neither poor Rich Both

SK 0 0 0 37
0% 0% 0% 100%
SA 0 1 4 31
0% 1% 11% 86%
NL 0 12 0 13
0% 48% 0% 52%
LA 1 14 1 8
4% 56% 4% 32%
NB 0 7 1 16
0% 28% 4% 64%

Total 1 34 6 105
1% 23% 4% 71%

Table 8: Do the poor Table 9: Could You Start

practice fish farming? Fish Farming?

Barangay Yes No Barangay Yes No

SK 31 6 SK 2 35
84% 16% 5% 95%
SA 18 18 SA 5 31
50% 50% 14% 86%
NL 16 9 NL 10 15
64% 36% 40% 60%
LA 10 15 LA 4 21
40% 60% 16% 84%
NB 6 18 NB 6 19
24% 72% 24% 76%

Total 81 66 Total 27 121

55% 45% 18% 82%

206
 Paper 4/5

Table 10: Does Fish Farming Have Any negative Impact on the Poor?

No Yes
Irregular
Barangay Total Negative Impact on fishing Exploitation of the Poor Arduous work Income
SK 12 25 24 1 1 0
32% 68% 65% 3% 3% 0%
SA 27 9 3 1 0 4
75% 25% 8% 3% 0% 11%
NL 23 2 1 1 1 0
92% 8% 4% 4% 4% 0%
LA 18 7 7 0 0 0
72% 28% 28% 0% 0% 0%
NB 24 1 0 1 0 0
96% 4% 0% 4% 0% 0%

Total 104 44 35 4 2 4
70% 30% 24% 3% 1% 3%

Table 11: Overall, Is Aquaculture a Good Thing in Your

Community?

No Yes
Main benefit from aquaculture
Barangay Total Employment Income Fish/Food Indirect
SK 4 33 26 4 14 1
11% 89% 70% 11% 38% 3%
SA 0 36 32 4 17 3
0% 100% 89% 11% 47% 8%
NL 0 25 19 14 5 2
0% 100% 76% 56% 20% 8%
LA 3 22 3 10 8 1
12% 88% 12% 40% 32% 4%
NB 0 25 16 2 15 0
0% 100% 64% 8% 60% 0%

Total 7 141 96 34 59 7
5% 95% 65% 23% 40% 5%

207
 Paper 4/5

Table 12: Ladder Diagram

Barangay Food Fish Employment Cash

-5 years Now + 5 years -5 years Now + 5 years
SK Average 6.55 6.09 6.46 7.88 7.26 7.00 7.53 7.69
SD 2.11 1.91 1.84 2.19 2.26 2.22 3.14 3.11
SA Average 7.11 6.44 7.45 7.19 7.28 7.91 3.75 3.72
SD 1.75 1.81 2.06 2.07 1.86 2.13 3.59 3.57
NL Average 4.46 4.17 4.71 5.21 4.42 4.71 5.96 5.42
SD 2.15 1.24 1.55 1.74 1.47 1.52 3.24 3.08
LA Average 3.60 3.44 3.24 4.16 4.16 3.52 0.60 0.52
SD 1.29 1.23 1.20 1.52 1.37 1.16 2.08 1.71
NB Average 4.12 4.08 4.48 4.72 4.64 4.72 0.40 0.32
SD 1.76 1.66 1.83 1.51 1.47 1.51 2.00 1.60

Total Average 5.40 5.05 5.43 6.06 5.80 5.74 3.88 3.80
SD 2.32 2.03 2.31 2.36 2.27 2.43 4.05 3.98

Table 13: Aggregate Income & Poverty in the Five Communities

Barangay Income Poverty
Mean Median SD HC PG P2 HC (Food)
SK 17,214 14,633 12,134 0.59 0.26 0.15 0.35
SA 24,242 17,182 21,555 0.50 0.22 0.12 0.31
NL 18,383 17,974 11,638 0.44 0.21 0.12 0.36
LA 9,379 5,600 10,413 0.84 0.50 0.33 0.76
NB 23,673 11,742 40,344 0.60 0.28 0.17 0.44

Total 18,889 12,925 21,892 0.59 0.28 0.17 0.43

208
 Paper 4/5

Table 14: Aquaculture and Income Generation

Income From Aquaculture (PhP)
Average Aggregate
Barangay Households Share of Share of
Per Per household Household
Household Capita Income Income
SK All 43,927 8,167 0.58 0.57
Poor 32,535 5,342 0.58 0.57
Non-Poor 60,636 13,993 0.58 0.57
Extremely Poor 30,872 4,778 0.65 0.71
Non-Extremely Poor 50,999 10,643 0.54 0.53

SA All 19,282 3,403 0.23 0.17

Poor 18,877 2,763 0.31 0.34
Non-Poor 19,686 3,730 0.16 0.11
Extremely Poor 21,072 3,175 0.35 0.43
Non-Extremely Poor 18,494 4,203 0.18 0.13

NL All 13,823 2,743 0.33 0.17

Poor 14,676 2,484 0.41 0.31
Non-Poor 13,153 3,019 0.26 0.12
Extremely Poor 16,056 2,779 0.48 0.42
Non-Extremely Poor 12,567 2,717 0.24 0.12

All All 27,194 5,038 0.39 0.29

Poor 23,863 3,951 0.45 0.44
Non-Poor 30,809 6,552 0.32 0.23
Extremely Poor 23,564 3,721 0.51 0.54
Non-Extremely Poor 29,037 5,898 0.33 0.25

Table 15: Simulated Poverty Levels (No Aquaculture)

Barangay Income Poverty
Mean Median SD HC PG P2 HC (Food)
SK 7,869 2,833 10,778 0.86 0.64 0.54 0.70
SA 20,698 13,555 22,878 0.58 0.35 0.27 0.44
NL 12,631 11,667 11,522 0.64 0.41 0.34 0.48

Total 13,796 8,179 17,212 0.70 0.47 0.39 0.55

209
 Paper 4/5

Table 16: Gini Decomposition

Barangay Source of Income s
Non-
Aquaculture Aquaculture Total
SA Gini Decomposition
Income share (Ss) 0.166 0.834 1.000
Gini coefficient (Gs) 0.928 0.519 0.397
Gini correlation coefficient (Rs) 0.047 0.900 1.000
Overall contribution to Gini coefficient (SsGsRs) 0.007 0.389 0.397
Share of Gini coefficient (SsGsRs/G) 0.018 0.982 1.000
Concentration coefficient (GsRs/G) 0.111 1.178 1.000
Response of Gini coefficient to a change in income source s
Absolute change (Ss(RsGs-G)) -0.059 0.059 0.000
Percentage change (Ss(GsRs/G-1)) -0.148 0.148 0.000

SK Gini Decomposition
Income share (Ss) 0.570 0.430 1.000
Gini coefficient (Gs) 0.402 0.516 0.235
Gini correlation coefficient (Rs) 0.475 0.569 1.000
Overall contribution to Gini coefficient (SsGsRs) 0.109 0.126 0.235
Share of Gini coefficient (SsGsRs/G) 0.463 0.537 1.000
Concentration coefficient (GsRs/G) 0.812 1.249 1.000
Response of Gini coefficient to a change in income source s
Absolute change (Ss(RsGs-G)) -0.025 0.025 0.000
Percentage change (Ss(GsRs/G-1)) -0.107 0.107 0.000

NL Gini Decomposition
Income share (Ss) 0.172 0.828 1.000
Gini coefficient (Gs) 0.606 0.491 0.357
Gini correlation coefficient (Rs) -0.244 0.941 1.000
Overall contribution to Gini coefficient (SsGsRs) -0.025 0.383 0.357
Share of Gini coefficient (SsGsRs/G) -0.071 1.071 1.000
Concentration coefficient (GsRs/G) -0.414 1.293 1.000
Response of Gini coefficient to a change in income source s
Absolute change (Ss(RsGs-G)) -0.087 0.087 0.000
Percentage change (Ss(GsRs/G-1)) -0.243 0.243 0.000

All Gini Decomposition

Income share (Ss) 0.295 0.705 1.000
Gini coefficient (Gs) 0.662 0.574 0.359
Gini correlation coefficient (Rs) 0.149 0.816 1.000
Overall contribution to Gini coefficient (SsGsRs) 0.029 0.330 0.359
Share of Gini coefficient (SsGsRs/G) 0.081 0.919 1.000
Concentration coefficient (GsRs/G) 0.275 1.303 1.000
Response of Gini coefficient to a change in income source s
Absolute change (Ss(RsGs-G)) -0.077 0.077 0.000
Percentage change (Ss(GsRs/G-1)) -0.214 0.214 0.000

210
 Paper 4/5

Table 17: Decomposition of Aquacultural Income

Type of Activity
Wage 'Free fish' Shell
Barangay Labourer Harvester Caretaker Collector Collector Other
SK Share of HHs participating 0.57 0.03 0.27 0.16 0.30 0.11
Mean Income (participating HHs) 18,109 3,900 64,571 4,850 32,779 51,583
Mean Income (all HHs) 10,278 105 17,452 786 9,745 5,577
Share of Aquacultural Income 0.23 0.00 0.40 0.02 0.22 0.13

SA Share of HHs participating 0.28 0.14 0.19 0.03 0.00 0.14

Mean Income (participating HHs) 18,579 10,410 50,914 1,800 0 19,970
Mean Income (all HHs) 5,161 1,446 9,900 50 0 2,774
Share of Aquacultural Income 0.27 0.07 0.51 0.00 0.00 0.14

NL Share of HHs participating 0.56 0.52 0.24 0 0.12 0.12

Mean Income (participating HHs) 8,345 2,272 29,839 0 4,043 2,900
Mean Income (all HHs) 4,673 1,181 7,161 0 485 348
Share of Aquacultural Income 0.34 0.09 0.52 0.00 0.04 0.03

All Share of HHs participating 0.46 0.19 0.23 0.07 0.14 0.12
Mean Income (participating HHs) 15,175 4,499 51,354 4,414 26,621 26,240
Mean Income (all HHs) 6,968 872 12,053 315 3,803 3,213
Share of Aquacultural Income 0.26 0.03 0.44 0.01 0.14 0.12

References:

Alauddin, M. and C. Tisdell (1998). Bangladesh’s shrimp industry and sustainable

development: resource-use conflict and the environment, Asian Fisheries Science, 11: 97-
120.
Balisacan, A. (1992). Equivalence Scale and Poverty Assessment in a Poor Country, Journal
of Philippine Development, 34(XIX-1): 81-96.
Balisacan, A. (1999). What Do We Really Know – or Don’t Know – About Economic
Inequality and Poverty in the Philippines?, in Balisacan, A.M. and Fujisaki, S. Edts
(1999), Causes of Poverty: Myths, Facts & Policies – A Philippine Study, University of
the Philippines Press, Diliman, Quezon City.
Banks, J. & P. Johnson (1994). Equivalence Scales Relativities Revisited, Economic Journal,
104: 883-890.
Bebbington, A. (1999). Capitals and capabilities: a framework for analysing peasant viability,
rural livelihoods and poverty. World Development, 27(12), 2021-44
Coull, J. R. (1993). Will the blue revolution follow the green revolution? The modern upsurge
of aquaculture, Area, 25(4): 350-357.
Coulter, F.A.E., Cowell, F.A. & Jenkins, S.P. (1992). Equivalence Scale Relativities and the
Extent of Inequality and Poverty, Economic Journal, 102: 1067-1082.
Deaton, A. (1996). The Analysis of Household Surveys, Johns Hopkins University Press.
Edwards, P. and Demaine, H. (1997). Rural Aquaculture: Overview and Framework for
Country Reviews, RAP Publication 1997/36, Regional Office for Asia and the Pacific,
FAO, Bangkok, Thailand.
Edwards, P. (1999). Aquaculture and Poverty: Past, Present and Future Prospects of Impact.
A discussion paper prepared for the fifth fisheries development donor consultation,
Rome, Italy, 22-24 February 1999 available at
http://www.sifar.org/Presentation%20Documents/aqua-poverty.html.

211
 Paper 4/5

Estudillo, J.P. & Otsuka, K. (1999) ‘Green Revolution, Human Capital, and Off-farm
Employment: Changing Sources of Income Among Farm Households in Central Luzon,
1966-1994’, Economic Development and Cultural Change 47: 497-523.
Fafchamps, M. & Gubert, F. (2003) Contingent loan repayment in the Philippines, Working
paper, Department of Economics, University of Oxford, 51 pp.
FAO (1998). The State of Food and Agriculture 1998, FAP, Rome.
FAO (2003). The State of the World Fisheries and Aquaculture 2002, FAO, Rome.
Foster, J., Greer, J., & Thorbecke, E. (1984). A class of decomposable poverty measures,
Econometrica, 52(3): 761-766.
Grosh, M. & Glewwe, P. (2000). Designing Household Survey Questionnaires for Developing
Countries: Lessons from 15 Years of the Living Standards Measurement Study,
Washington, The World Bank.
Halwart, M., Funge-Smith, S. and Moehl, J. (2003). The Role of Aquaculture in Rural
Development, in FAO (2003), The State of the World Fisheries and Aquaculture 2002,
FAO, Rome.
Hardin, G. (1968). The Tragedy of the Commons, Science, 162: 1243-1248.
Irz, X., Lin, L., Thirtle, C. & Wiggins, S. (2001). Agricultural growth and poverty alleviation,
Development Policy Review, 19(4): 449-466.
Lipton, M. (1997). Poverty – Are There Holes in the Consensus? World Development, 25(7):
1003-1007.
Muir, J. (1999). Aquaculture and Poverty: Full Baskets or Empty Promises? Perspectives
from the DfID Aquaculture Research Program, Contributed paper, Fifth Fisheries
Development Donor Consultation.
Pyatt, G., Chen, C.-H. & Fei, J. (1980). The Distribution of Income by Factor Components,
Quarterly Journal of Economics, 95: 451-73.
Prein, M. and Ahmed, M. (2000). Integration of Aquaculture into Smallholder Farming
Systems for Improved Food Security and Household Nutrition. Food Nutrition Bulletin,
21: 466-471.
Primavera, J. H. (1997). Socio-economic impacts of shrimp culture, Aquaculture Research,
28(10): 815-827.
Ravallion, M. (1996). Issues in Measuring and Modelling Poverty, Economic Journal 106:
1328-1343.
Ravallion, M. and Chen, S. (1997) What Can New Survey Data Tell Us About Recent
Changes in Distribution of Poverty?, World Bank Economic Review, 11: 282-357.
Sadoulet, E. & de Janvry, A. (1995). Quantitative Development Policy Analysis, Johns
Hopkins University Press, Baltimore & London.
Sen, A. (1985). Commodities and Capabilities. Amsterdam, North-Holland.
Stark, O., Taylor, J.E. & Yitzhaki, S. (1986). Remittances and Inequality, Economic Journal,
96:722-740.
Stonich, S.C., Bort, J.R. and Ovares, L.L. (1997). Globalization of the Shrimp Mariculture:
The Impact on Social Justice and Environmental Quality in Central America, Society and
Natural Resources, 10(2): 161-179.
Tacon, A.G.J. (2001). Increasing the Contribution of Aquaculture for Food Security and
Poverty Alleviation. In Subasinghe, R.P., Bueno, P.B., Phillips, M.J., Hough, C.,
McGladdery, S.E. and Arthur, J.R. (eds), Aquaculture in the Third Millenium, Bangkok,
Thailand, 20-25 February 2000, pp. 63-72. NACA, Bangkok, Thailand, and FAO, Rome,
Italy.
Yap, W. G. (1999). Rural aquaculture in the Philippines, FAO, Rome.

212
 Paper 4/5

Appendix: Questionnaire Survey

SURVEY OF HOUSEHOLDS IN THE COASTAL AREAS OF THE

PHILIPPINES

Municipality: Barangay:

Date of first visit:

Date of second visit:

213
 Paper 4/5

Part 1. HOUSEHOLD ROSTER

PERSON INTERVIEWED: PREFERABLY THE HEAD OF THE HOUSEHOLD.

IF HE/SHE IS NOT AVAILABLE, FIND A “PRINCIPAL RESPONDENT” TO
ANSWER THE QUESTIONS IN HIS/HER PLACE. THE PERSON SELECTED
MUST BE A MEMBER OF THE HOUSEHOLD WHO IS ABLE TO GIVE
INFORMATION ON THE OTHER HOUSEHOLD MEMBERS.

A person is considered part of the household:

- If he/she normally lives and eats his/her meals in the household’s dwelling
- If he/she is not away from the household more than 9 months a year

1. 2. 3. 4. 5. 6.

MAKE A SEX RELATIONSHIP TO HEAD: AGE What is the Religion

ID CODE

COMPLETE LIST HEAD.................1 present

OF ALL WIFE/HUSBAND.........2 marital status CATHOLIC..1
CONCERNED CHILD/ADOPTED CHILD..3 of [NAME]? PROTESTANT...2
BEFORE GOING GRANDCHILD...........4 MUSLIM….3
TO QUESTIONS 2- NIECE/NEPHEW.........5 MARRIED..1 OTHER….4
6 FATHER/MOTHER........6
SISTER/BROTHER.......7 DIVORCED.2
SON/DAUGHT.-IN-LAW...8
BROTHER/SISTER-IN- SEPARATED.3
LAW..................9
GRANDFATHER/MOTHER..10 WIDOW OR
FATHER/MOTHER-IN- WIDOWER..4
LAW.11
OTHER RELATIVE......12 NEVER
SERVANT OR SERVANT'S MARRIED..5
RELATIVE...........13
TENANT OR TENANT'S
RELATIVE...........14
OTHER(SPECIFY_____).15

MALE.1

NAME FEMALE 2 YEARS

10

11

12

214
 Paper 4/5

PART 2. EDUCATION

1. 2. 3. 4. 5. 6.
Are you What is the In what Is the How far How do you go to
ID CODE

currently highest level you grade school you away from school?
enrolled in have completed in are you are your home is
school? school? currently currently the school WALK.1
enrolled enrolled in you have BICYCLE.....2
in public or been CAR...3
Codes: school? private? attending in Public
Elementary: Use the last 12 Transport...…4
1-6 same PUBLIC.1 months? BOAT..5
High school: codes as PRIVATE OTHER (SPE-
7-10 for Q2. SECU- CIFY).6
College: 11- LAR..2
16 PRIVATE
Vocational: RELIG-
V1-V3 IOUS..3

YES..1(»3- >>next
7) person
NO..0 DISTANCE

10

11

12

Do you face any major problem to send your children to school?

No……………………………………………………………………………..1
Yes, fees are too high........................................................................................2
Yes, costs (textbooks, uniforms, etc.) other than fees are too high……….…...3
Yes, transportation to school…………………………………………………..4
Other (Specify)………………...………………………………………………5

215
 Paper 4/5

PART 3: EMPLOYMENT
I would like to ask each household member questions about the work that he/she did the last 12 months, whether work on a farm, on his/her own
account, in a household business enterprise, or for someone else.
1. 2. 3. 4. 5. 6. 7.
For what type of business were you For how During How much was If you received If you received a
What did you do? working? many these your last fish as a form non-monetary
ID CODE

Household/own account weeks in weeks, payment? of payment, payment for this

enterprise…………....... 1
the last how If respondent how much did work other than
Business owned by a family
member outside the household .2 12 many has not been you receive? fish, what was the
Business owned by non- months hours per paid, ask: What Over what time value of this
household/non-family member……3 did you week did payment do interval? payment?
do this you you expect?
work? usually do What period of
this work? time did this
payment
cover?

WRITTEN DESCRIPTION WEEKS/ HOURS/ Pesos Kg Pesos

YEAR WEEK (Specify time (Specify (Specify time
period) time period) period)

For own-account fish-farming, farming and fishing, do not ask questions on payments (i.e., stop at column 5 (included)).

216
 Paper 4/5

PART 4. LAND-BASED ACTIVITIES

4.1. ACCESS TO LAND

1. 2. 3. 4. 5.
PLOT CODE

Please tell me What is What kind of land is Is this land owned How did your
about each plot of the this? and used by your household
land owned, area of Brackish water household, acquire this land?
rented or leased the Fish Pond..1 rented/leased in or INHERITED...1
by your plot? Fresh water rented out? CLEARED.....2
household? fishpond….2 Owned and PURCHASED...3
Non-irrigated used…….1 (SPECIFY__).4
agricultural Owned and
land…….....3 rented out……….2
Irrigated Leased under
agricultural FLA…..3
land…………………4 Rented in from
OTHER(SPE- private
CIFY_)....5 owner…………….4

NAME OF PLOT (Ha)

10

7. Is it difficult to have access to a fish pond to practice aquaculture?

Very difficult…………….1
Rather difficult…………..2
Relatively easy…………...3
Very easy ………………...4
If difficult, why? ____________________________________________________

217
 Paper 4/5

Sections 4.2 and 4.3 should only be completed if the household includes a fish pond operator or a farm operator
4.2. Agricultural productions
4.2.1 Outputs
Cropping Name of crop: Name of crop: Name of crop: Name of crop:

Production Farm gate Production Farm gate Production Farm gate Production Farm gate How many [animals]
(Kg) price (Kg) price (Kg) price (Kg) price does your household
(Pesos) (Pesos) (Pesos) (Pesos) currently own?
1st Crop ANIMAL NUMBER
1 Pigs
2nd Crop
2 Poultry
3rd Crop
3 Goats
Annual total
4 Water Buffalo
5 Cows
4.2.2 Operating expenses 6 Others (specify)
Cropping Seeds Fertilisers Land Rent Hired labour Other
Expenses
Quantity Price Quantity Price Area Price Quantity Price (Pesos)

1st Crop

2nd Crop

3rd Crop

Annual total

4.2.3 Livestock

218
 Paper 4/5

4.3. Aquaculture productions

4.3.1 Output
Harvests Name of fish: Name of fish: Name of fish: Name of fish:

Production Farm gate Production Farm gate Production Farm gate Production Farm gate
(Kg) price (Kg) price (Kg) price (Kg) price
(Pesos) (Pesos) (Pesos) (Pesos)
1st Harvest

2nd Harvest

3rd Harvest

Annual total

4.3.2 Operating expenses

Cropping Fry/fingerlings Feeds Fertilsers Land Fertilisers Land Rent Hired labour Other Expenses

Quantity Price Quantity Price Quantity Price Quantity Price Quantity Price Area Price Quantity Price (Pesos)

1st Crop

2nd Crop

3rd Crop

Annual total

4.3.3 Extension
1. How many times during the last 12 months did members of your household visit an aquacultural extension agent or an agricultural extension
center to discuss fish/prawn/crab production? Times

2. How many times in the last 12 months did any aquacultural cultural extension agent visit? Times

219
 Paper 4/5

Part 5. FISHING

1. How many months a year can do you fish (as a professional activity reported in
Part3)? (Note: answer should be consistent with answer in the ‘employment’ table)

2. If there are months when you cannot fish, why is that? (No fish, bad weather etc.)

3. In the months when you fish, how much do you typically catch in a week/day?

4. What is the sale price of each product?

5. What are the costs involved in fishing? Specify time period over which those apply.

- Boat rental

- Boat maintenance

- Fuel cost

- Fishing gear replacement

- Hired labour

- Other (specify)

220
 Paper 4/5

Part 6. OTHER SOURCES OF INCOME

1. During the past 12 months has your household or any of its members received any money or goods
from persons who are not members of your household? For example for relatives living elsewhere,
child support or alimony, or from friends or neighbours?
Yes=1, No=0
2. If answer to Q.1 was positive, fill out the table below:
1. 2. 3. 4.
What is the How much Where
relationship of money have does the
DONOR

donor to this members of donor earn

household? the his/her
1=spouse household money?
2=sibling received
3=son/daughter from[DONOR]
4=parent/s in the past 12
5=other months?
Pesos Place
1

10

11

12

221
 Paper 4/5

3. Other sources of income over the last 12 months

How much did your household receive in the last 12
months from [SOURCE] including the value of any
payment in the form of goods?
SOURCE PESOS

1 Transfers from the state

Pension
Illness/disability Payment
Social assistance payment
Other (specify)
2 Rental income:
Land
House
Car
Boat
Tricycle
Other (specify)
3 Revenue from sale of assets:
Source 1(Specify)
Source 2
Source 3
Etc..
4 Other Income:
Private pension
Source 2
Source 3
Etc.
5 Other: (Specify)

222
 Paper 4/5

7. Perception of poverty and coping strategies:

1- Do you consider that your household is poor?

a. Yes……………………………………….1
b. No……….………………………………..0
2- In your opinion, what does it mean for a household to be poor/what is the main
characteristic of poverty? (To be landless? To be unable to purchase enough food for
the household? To have no influence within the community? Etc.)

3- What type of crisis have you experienced in the last 12 months?

Crisis Yes=1
Flood
Drought
Typhoon
Illness in the family
Death of a household
member
Loss of job
Eviction
Others (specify)
4- How did you cope with the crisis? (Loan, sale of livestock/land/other assets, support
network, migration etc.)

5- In your opinion, what is the main cause of poverty in your community?

6- In your opinion, could poverty be reduced in your community? If so, how?

Aquaculture and poverty:

7- Does fish farming benefit the poor and/or the rich?

8- Do the poor practice fish farming? If not, why? (lack of information; financial
resources; suitable land etc.)

9- Could you yourself start fish farming if you wished to do so? If not, why?

10- Does fish farming have any negative impact on the poor? If so, which ones?

11- In your opinion, is fish farming altogether a good thing for this community and why?

12- Should the government do more in your area to develop fish farming?

223
 Paper 4/5

I will now show you a ladder diagram. Please choose a step on the ladder that realistically
describes your situation. The first step on the ladder describes the worst possible situation. As
the step goes higher, the situation gets better. Thus, the highest step on the ladder represents
the best situation that you can have.

Ladder Diagram

10
9
8
7
6
5
4
3
2
Step 1

13- Overall food consumption.

The first step on the ladder represents a situation where all household members barely
have anything to eat. You hardly know where the next meal is coming from. The highest
step indicates a situation where every member of your household eats three nutritious
meals daily.
Five years ago ______ At present_____ Five years from now_____

13- Fish consumption

The first step on the ladder shows a situation where fish/seafood does not form part of
you household’s diet at all, even if your household members wanted to eat fish. The
highest step on the ladder represents a situation where your household members can eat as
much fish as they want.
Five years ago ______ At present_____ Five years from now_____

14- Employment
The first step on the ladder describes a situation where no household member earns from
any aquaculture-related activity. The highest step on the ladder represents a situation
where members of your household obtain all their earnings from various aquaculture-
related activities (production, marketing, harvesting, processing, input supply etc.).
At present_____

15- Cash income from aquaculture

The first step shows no cash income from aquaculture for your household. The highest
step on the ladder shows cash income from aquaculture-related activities that is more than
adequate to provide for you household’s needs
At present_____

224
 Paper 4/5

PART 8: CONSUMPTION
8.1 FOOD CONSUMPTION EXCLUDING FISH - MOST
KNOWLEDGEABLE MEMBER
Number of days since last visit:

In the following questions, I want to ask about all purchases made for your household,
regardless of which person made them. Q1. Has your household consumed [FOOD] since
my last visit? Please exclude from your answer any [FOOD] purchased for processing or
resale in a household enterprise.

HOME
PURCHASE SINCE LAST VISIT PRODUCTION GIFTS
1. 2. 3. 4. 5.

How much of What was Since my last visit, Since my

[FOOD] have you or the total how much [FOOD] last visit,
PUT AN X (X) IN any other member of value of did your household how much
THE the household your consume that you [FOOD]
APPROPRIATE bought since my last purchase? grew or produced did your
BOX FOR EACH visit, that is since []? at home? household
FOOD ITEM. IF consume
THE ANSWER that you
TO Q.1 IS YES, received
ASK Q.2-5. as a gift?

Quantity Value Quantity Quantity

(Specify (Specify
NO YES (Specify unit) (Pesos) Unit) Unit)

1 Rice

2 Other cereals

3 Roots/tubers

4 Fruits

5 Vegetables

6 Poultry

7 Pork

8 Beef
Other meats and
9 meat products
Dairy products
10 and eggs

11 Coffee/cocoa/tea
Non-alcoholic
12 beverages

13 Other foods
Alcoholic
14 beverages

225
 Paper 4/5

8.2. FISH CONSUMPTION - MOST KNOWLEDGEABLE MEMBER

In the following questions, I want to ask about all purchases made for your household,
regardless of which person made them. Q1. Has your household consumed [FISH] since
my last visit? Please exclude from your answer any [FISH] purchased for processing or
resale in a household enterprise.
PURCHASE SINCE LAST VISIT AQUACULTURE FISHING GIFTS
1. 2. 3. 4. 5. 6.

How much of [FISH] What was Since my last Since my last Since my
have you or any the total visit, how much visit, how last visit,
PUT AN X (X) other member of the value of [FISH] did your much [FISH] how much
IN THE household bought your household did your [FISH] did
APPROPRIATE since my last visit, purchase? consume that household your
BOX FOR that is since? you produced ‘at consume that household
EACH FOOD home’ through was obtained consume
ITEM. IF THE aquaculture? by fishing? that you
ANSWER TO received
Q.1 IS YES, as a gift?
ASK Q.2-5.

Quantity Value Quantity Quantity Quantity

(Specify (Specify (Specify
NO YES (Specify unit) (Pesos) Unit) unit) Unit)

1 Tilapia

2 Bangus

3 Crabs

4 Shrimps

5 Prawns

6 Round scad

10

11

12

13

14

226
 Paper 4/5

8.3. OTHER
1. Total consumption – Please refer back to the ladder diagram.
The first step on the ladder represents the (unrealistic) situation where your household
would not spend any of its cash-income on food (including fish). The highest step
indicates a situation, unrealistic as well, where all cash income would be spent on food.
Please characterize your current situation
At present_____

2. Over the last 12 months, in what month was it most difficult to provide
adequate food for your household? Why?

8.4. CONSUMER DURABLES – MOST KNOWLEDGEABLE PERSON

1. 2. 3.
Do you
LIST OF ITEMS own
[ITEM]

Yes..1
DESCRIPTION No…0

1
Stove
2
Refrigerator
3
Washing Machine
4
Sewing/knitting machine
5
Fan
6
Television
7
Radio
8
Tape player/CD player
9
Camera, video camera
10
Bicycle
11
Motorcycle/scooter
12
Tricycle
13
Car
14
Truck
15
Water pump/tube well
16
Fish pond equipment
17
Farm equipment
18
Non-motorized boat
19
Motorized boat
20
Phone/cell phone

227
 Paper 4/5

Part 9. HOUSING MODULE – SHORT -- PART A. DESCRIPTION OF THE

DWELLING 4. How many rooms do the members of your household occupy, including
bedrooms, living rooms and rooms used for household enterprises?
Now I would like to ask you about your housing conditions. I mean by housing all
the rooms and all separate buildings used by your household members. What DO NOT COUNT BATHROOMS, KITCHENS, BALCONIES AND CORRIDORS
buildings or rooms do the members of your household occupy?
NUMBER OF ROOMS
1. WHAT IS THE MAJOR CONSTRUCTION MATERIAL OF THE EXTERNAL WALLS?
5. What is the space of your dwelling including living and accessory rooms?
BRICK .........................................................................1 SQUARE METERS
CONCRETE BLOCKS ................................................2
WOOD, LOGS.............................................................3 6.. In approximately what year was this dwelling built?
TIN, ZINC, GI SHEETING...........................................4 ASK THE RESPONDENT TO PROVIDE AN ESTIMATE IF UNSURE
BAMBOO ....................................................................5 OF THE EXACT YEAR
SACS/PLASTIC ………………………………………….6 YEAR BUILT
OTHER (SPECIFY____________________).............7
7. Do you own or rent this house?
2. WHAT IS THE MAJOR MATERIAL OF THE ROOF?
OWN ........................................................................... 1
CONCRETE ................................................................1 RENT .......................................................................... 2
METAL SHEETS.........................................................2 OTHER (SPECIFY)..................................................... 3
WOOD.........................................................................3
NIPA ...........................................................................4 PART B. HOUSING SERVICES
OTHER (SPECIFY ___________________)............ 5
1 What is the main source of water for drinking and cooking for your household?
3. WHAT IS THE PRIMARY MATERIAL OF THE FLOOR?
PRIVATE CONNECTION TO PIPELINE ............... 1
WOOD.........................................................................1 PRIVATE WELL..................................................... 2
TILE.............................................................................2 PUBLIC TAPS/ STANDPIPE ................................. 3
CONCRETE ................................................................3 PUBLIC WELL ....................................................... 4
CLAY/EARTHEN FLOOR ...........................................4 NEIGHBORS ......................................................... 5
BAMBOO ....................................................................5 WATER VENDOR ................................................. 6
LINOLEUM..................................................................6 OTHER (SPECIFY _____________________) .... 7
OTHERS(SPECIFY ___________________) ............7
2. How do you usually treat your drinking water?

NO TREATMENT………………………………………………1
BOIL IT............................................................................... 2
FILTER IT (ONLY) ....................................................... .. . 3
OTHER (SPECIFY)…………………………………………….4

3. What is the main source of water for bathing and washing for your household?

228
 Paper 4/5

PRIVATE CONNECTION TO PIPELINE ............... 1

PRIVATE WELL ..................................................... 2
PUBLIC TAPS/ STANDPIPE ................................. 3
PUBLIC WELL ....................................................... 4
NEIGHBORS.......................................................... 5
WATER VENDOR ................................................. 6
RIVER, SEA, POND…………………………………..7
OTHER (SPECIFY _____________________)…….8

4. What is the type of toilet that is used in your household?

FLUSH TOILET ............................................................1

WATER SEALED…………………………………………..2
OUT HOUSE OVER PIT ...............................................3
OUT HOUSE CONNECTED TO SEWAGE ..................4
OTHER (SPECIFY ______________________)..........5
NO TOILET ...................................................................6
5. What is the main source of lighting in your dwelling ?

ELECTRICITY.............................................................1
KEROSENE, OIL OR GAS LAMPS ............................2
CANDLES OR BATTERY FLASHLIGHTS..................3
OTHER (SPECIFY____________________).............4

6. What fuel do you use most often for cooking?

GAS.............................................................................1
ELECTRICITY.............................................................2
WOOD.........................................................................3
CHARCOAL ................................................................4
KEROSENE ................................................................5
OTHER (SPECIFY _____________________)..........7

229
 Paper 4/5

Part 10. HEALTH

1. During the last four weeks, how many days of primary activity did you miss due to poor health?
ID Code Number of days

2. How far away is the closest surgery/hospital/health station? Km

3. Is medical care easily available to your family members?

Yes……. ............................................................ 1
No....................................................................... 2

4. If not, why?

Part 11. CREDIT

1. Can you borrow money if you need it?
Yes…….. ........................................................... 1
No....................................................................... 2
2. If your answer is positive, from which sources?
Source of credit Yes=1
No=0
Cooperative
Agricultural
Bank
Other Bank
Money lender
Savers’ group
Viillage Fund
Family
Friends
Other (specify)

3. How far is the closest bank? km

230
 Paper 5/5

Research project R8288: Assessing the sustainability

of brackish-water aquaculture systems in the
Philippines – Paper 5/5

Is there an efficiency case for land redistribution in

Philippine brackish-water aquaculture? Analysis in a
ray production frontier framework

Abstract: We investigate the possible existence of an inverse relationship (IR)

between farm size and productivity in Philippine brackishwater pond aquaculture.
The study is motivated by the fact that fish ponds have so far been exempted from the
Comprehensive Agrarian Reform Laws and suggestions in the literature of inefficient
management of fish farms. The analysis of technical efficiency is based on the
estimation of a multi-product ray production function estimated in a stochastic
frontier framework. There is some evidence of an IR but of only limited strength.
Hence, it is unlikely that agrarian reform is the key to unlocking the productivity
potential of brackishwater aquaculture in the Philippines.

231
 Paper 5/5

Contents

1. Introduction 233

2. Measuring the productivity/efficiency of polyculture farms 235

2.1 A review of alternative approaches 235
2.2 Implementation issues 239

3. Estimation strategy 240

4. Data and estimable model 243

5. Empirical Results 245

5. 1 Partial productivity indicators 245
5. 2 Specification tests and the structure of the technology 246
5.3 Inefficiencies and the inverse relationship 248

6. Discussion and conclusion 250

Appendix 259

232
 Paper 5/5

1. Introduction
While global production of capture fisheries stagnated over the last decade, output
from aquaculture expanded steadily.1 The FAO (2002) reports that global catch from
capture fisheries barely returned in year 2000 to the level observed in the early 1990s
at roughly 78 million tonnes. Meanwhile, production growth in aquaculture took place
at an average annual rate of 7.1% in the 1980s and 5.1% in the 1990s, which makes
aquaculture one of the fastest growing food-producing sub-sectors (Ahmed and
Lorica, 2002). This spectacular development has sometimes been described as a ‘blue
revolution’, with the underlying idea that aquaculture could potentially solve some
aspects of the world’s chronic hunger and malnutrition problems (Coull, 1993). While
there is no arguing with the increase in aquaculture production, it is however
necessary to acknowledge that this development has generated a number of social,
environmental and economic problems. Hence, questions have been raised about the
ecological impact of aquaculture, in particular with regard to biodiversity (Jana and
Webster, 2003; Tisdell, 2003) and mangrove destruction (Primavera, 2000); about the
equity of its development (Primavera, 1997; Alauddin and Tisdell, 1998; Coull, 1993)
and about its food security benefits (Naylor et al., 2000; Primavera, 1997).

The Philippines conform to these general trends. Yap (1999) reports that aquaculture
output in the country has grown at the average annual rate of 5.4% in the 1990s and
that its share of total fisheries production keeps increasing. Yet, its development has
had a detrimental effect on mangroves, resulted in the salinisation of previously
productive agricultural land, generated conflicts over the use of natural resources
(Yap, 1999) and some have even argued that it has been responsible for the
marginalisation of some coastal communities and an increase in the rate of
unemployment (Primavera, 1997). Against this background, the aim of this article
consists in addressing one equity aspect of aquaculture development in the Philippines
that relates to the distribution of fishpond holdings 2 . We investigate whether there is
any evidence of an inverse relationship (IR) between farm size and productivity in
brackishwater aquaculture in order to evaluate the case, on efficiency ground, for
reform of the existing tenurial system, land redistribution, or other policies aimed at
improving the functioning of the land market.

The study is motivated first by a common perception that the vast areas of Philippine
brackishwaters 3 represent a valuable resource that is not exploited optimally and is
not contributing fully to the development process of coastal areas. We believe that it
will make a contribution to an important and ongoing policy debate that emerges from
the fact that, while the Philippines adopted several land reform laws in the late 1980s,
aquaculture ponds have so far been exempted 4 . As a result, the distribution of
holdings in brackishwater aquaculture remains very unequal as indicated by a Gini
coefficient of 0.72 for the two regions that form the focus of our study 5 and it is well-
known that fish farms of more than a hundred hectares are not uncommon. Naturally,
1
As there is increasing doubts regarding the validity of Chinese fisheries statistics, our statements refer
to the world excluding China. See FAO (2002) for a discussion of this issue.
2
Although it is not always specified, our study relates only to brackishwater pond aquaculture.
3
Yap (1999) reports that there are 239,323 hectares of brackiswater fishponds in the Philippines. The
electronic data that we obtained from the Bureau of Agricultural Statistics gives a total harvested area
of 415,272 hectares in year 2000.
4
The most recent one is the Comprehensive Agrarian Reform Law (CARL) of 1988 that imposes land
redistribution with a five hectare retention limit set on all agricultural land.
5
Source of data: Bureau of Agricultural Statistics’ inventory of fishponds from 1997.

233
 Paper 5/5

large fishpond owners and leaseholders believe that agrarian reform would, if
anything, only worsen the severe problems of poverty and inequality in the
communities where fish farming represents an important activity. Yap (1999) cites a
telling extract from the newsletter of Negros Prawn Producers and Marketing
Cooperative:

“The implementation of the (land reform) law is liable to cause widespread strife
among the landowners…. There is no showing that land reform will enliven the
plight of the poor. Without undermining their capabilities, it is also doubtful
whether they (the farmers) can put up the necessary capital to maximize land use.
Having been used to having a landlord on whom to call in times of need, this
plunge to independence may have a crippling effect.”

This view stands in sharp contrast with the common belief in agriculture that small
farmers tend to achieve higher productivity and efficiency levels than large farmers,
i.e. that there usually is an IR, as demonstrated in Sen’s seminal paper (Sen, 1962) 6 .
Besides, the experience of Thailand, where the extremely dynamic prawn industry is
supported by relatively small farmers (Yap, 1999), suggests that there is no particular
impediment to the development of a competitive aquaculture sector based on
smallholders. We therefore believe that testing the IR in Philippine brackishwater
aquaculture will generate important policy insights; in particular, a strong IR would
suggest that institutional changes leading to a more equal size distribution of holdings
could increase both equity and efficiency.

Our analysis is based on the analysis of a sample of 127 farms in two of the three
main regions for brackishwater aquaculture in the Philippines and investigates the
level and determinants, including farm size, of their technical efficiency. From a
methodological standpoint, we believe that the article makes three contributions to the
agricultural economics literature. First, we represent the technology by a ray
production function first proposed by Löthgren (1997), which, to the best of our
knowledge, has not been previously attempted on farming data. This approach
presents several advantages that we explain in the next two sections. Second, we
explore the properties of the output ray function which arise from its duality with both
the minimum cost and maximum revenue functions. This is important in interpreting
our estimations results, and could be useful in the future to use the model to analyse
issues of allocative efficiency. Finally, we propose two approaches to quantify the
explanatory power of the inefficiency effect variables in the Battese and Coelli (1995)
stochastic frontier model. This is extremely useful in the empirical section to measure,
in a way that is entirely consistent with the underlying frontier model, the strength of
the IR relationship.

The paper is organised as follows. The section presents the different approaches to the
measurement of efficiency in polyculture systems, insisting on the advantages and
properties of the ray production function. Section three presents the estimation
strategy and proposes an approach to quantify the explanatory power of the
inefficiency effect variables of the econometric model. The remaining sections

6
A recent review of the IR literature is Fan and Chan-Kang (2003). It concludes to the lack of
consensus on the validity of the IR hypothesis.

234
 Paper 5/5

discuss the data and empirical model, present the empirical results, and offer
conclusions.

2. Measuring the productivity/efficiency of polyculture farms

2.1 A review of alternative approaches
The IR literature started with the simple observation that yields, defined as output per
unit of surface area, differed according to the size of farms. In this context, output is
measured either in quantity or value terms and the negative relationship seemingly
implies that reallocating land from large to small farms would result in a net increase
in production. However, output per hectare is only a partial productivity indicator
which cannot satisfactorily measure overall farm productivity (Jha, Chitkara and
Gupta, 2000). Hence, farm A can achieve a higher yield than farm B not because it is
more efficient in transforming inputs into outputs but simply because it uses more
fertilisers or pesticides. Furthermore, any introductory economics textbook makes it
clear that economic optimality usually differs from yield maximisation. To address
this concern, albeit only partially, one can investigate the relationship between gross
margin and farm size, but this again fails to account for the input of primary factors,
such as labour and machinery services, in comparing farm performance. There is also
a concern that gross margins do not only reflect the productive ability of farm
managers but also the price environment in which they operate (Coelli, Rahman and
Thirtle, 2002). Hence, a farm can generate a high gross margin per hectare due to its
proximity to a particular market rather than its productive performance. There is
therefore a strong case to investigate the IR relationship within the confines of
production economics, which can accommodate the multi-dimensional aspect of farm
production.

Aquaculture production in the study area involves the polyculture of prawns, fish
(tilapia and/or milkfish) and crabs, which are produced simultaneously in the same
ponds. This raises a number of interesting modelling issues that we now attempt to
tackle. Obviously, all of the inputs, with the exception of the fry and fingerlings, are
largely non-allocable, i.e. it is not possible to determine the amount of each input used
in the production process of each individual output. Hence, land cannot be allocated
to different productions, and neither can the feeds or the labour input used to
exchange the pond water or maintain the mud dykes. This introduces a first linkage
among the different outputs of the aquaculture farm. Second, it is necessary to
recognize the possible jointness of production as it is likely that the different species
interact with each other in the aquaculture pond. For instance, biologists and
aquaculture experts often consider that the association prawn/tilapia in ponds tends to
reduce the rate of prawn mortality because tilapias, through their filtering activity and
consumption of organic matter lying at the bottom of the pond, improve the
bacteriological quality of the pond water (Corre et al., 1999). If that is so, output of
any single species depends not only on the inputs used in the production process of
that species but also on the quantities of other species grown simultaneously in the
pond. We therefore conclude that the production process relies on a truly multiple-
output technology, and that it is not possible to specify different production functions
for each output.

There exist several possible avenues to measure efficiency in a multiple-output

context. Maybe the most common approach involves the estimation of dual cost,
revenue or profit functions (Löthgren, 2000). However, this group of methods relies

235
 Paper 5/5

on relatively restrictive behavioural assumptions of economic optimization, such as

that of profit maximisation, which might not be expected to hold in developing
country aquaculture as farmers are likely to adopt complex livelihood strategies in the
face of multiple market failures. For instance, in addition to expected returns, it is
likely that the riskiness of alternative enterprises is taken into account by farm
operators when formulating production plans because perfect insurance markets
simply do not exist. Furthermore, estimation of these dual functions requires data on
prices of inputs and/or outputs that present sufficient variability to allow the use of
regression techniques. However, such variability is unfortunately often not present in
cross-sectional data because input and output markets are relatively well integrated
within regions. We therefore believe that a primal approach is better suited to the
analysis of efficiency and productivity for this particular study.

In a primal setting, a straightforward method of productivity/efficiency measurement

consists in aggregating all outputs into a single index, which can in turn be used in the
estimation of a production function/frontier. The method, suggested by Mundlak
(1963) in the agricultural economics literature, has frequently been used in the study
of efficiency in aquaculture (e.g., Karagianis, Katranidis & Tzouvelekas (2002)
analyse seabass and sea bream production in Greece; Irz and McKenzie (2003) study
polyculture systems in the Philippines; and Sharma and Leung (1998) as well as
Sharma (1999) investigate the efficiency of carp polyculture systems in Nepal and
Pakistan respectively). The single output index is usually obtained as the total weight
of production, which, although not theoretically sound, seems acceptable for
relatively similar products 7 .

When the products are not close substitutes, however, it is necessary to use output
prices to aggregate them. The simplest method consists in expressing output in value
terms, but there is a concern that in that case the resulting index reflects not only
output quantities but also the prices at which the farm products are sold. This general
index number problem is partially circumvented by the use of superlative output
indices, such as the Fischer index or Tornqvist-Theil index, first proposed by Caves,
Christensen and Diewert (1982) and presented in details in Coelli, Rao and Battese
(1998) (chapters 4 and 5). However, even superlative indices measure output in a
satisfactorily manner only under a number of restrictive assumptions, most notably
that output markets are perfectly competitive. Also, estimating an aggregate
production function implicitly imposes restrictions on the form of the underlying
multi-product technology. The very existence of an aggregate output index that can be
built from output quantities and prices depends on the technology being separable in
outputs and inputs (Orea, Alvarez and Morrison Paul, 2002). Hence, we believe that,
given the complexity of the input-output relationship in aquaculture, it is desirable to
use a framework of analysis that offers more flexibility in the representation of the
multi-product technology.

Fortunately, such a framework has recently become available with different ways of
representing the technology. An intuitive idea consists in re-writing the
transformation function so as to express one particular output as a function of the
input vector and the quantities of all other outputs. There are two difficulties with this

7
That is how all of the studies mentioned above proceeded, with the exception of Irz and McKenzie
(2003).

236
 Paper 5/5

approach. First, the choice of output that is used as dependent variable in the
regression analysis is arbitrary and this introduces an artificial asymmetry in the
method. Second, and most problematic, is the fact that not only the technology
parameters but also the efficiency scores depend on the particular output that is
chosen as dependent variable. The efficiency scores are therefore output specific and
there is no guarantee that the rankings obtained from alternative formulations of the
model be consistent with each other, as is easily demonstrated in Figure 1 for the two-
output case. The technology is represented by the production possibility frontier PP’.
Farm A is clearly closer to the frontier than farm B when efficiency is measured
according to a ‘fish’ orientation, but the reverse is true when a ‘prawn’ orientation to
efficiency measurement is adopted. Furthermore, when a farm is not producing the
output used as dependent variable in the regression, the interpretation of the efficiency
scores becomes difficult 8 .

For these reasons, the efficiency literature has moved away from the estimation of
transformation functions. A first alternative that has become popular in recent years
corresponds to the estimation of input or output distance functions (Coelli and
Perelman, 2000; Morrison Paul, Johnston and Frengley, 2000; Brümmer, Glauben and
Thijssen, 2002; Irz and Hadley, 2003). The output distance function introduced by
Shephard (1970) is defined formally from the output set P(x) by:
Do (x,y)=Min{θ>0: y/θ ∈ P(x)} (1)
It measures the fraction of maximum achievable output y/θ that the firm produces,
given a vector of inputs x and the technology, and assuming that any increase in
production would involve a proportional increase in all individual outputs. For any
input-output combination (x,y) belonging to the technology set, the distance function
takes a value no larger than unity, with a value of unity indicating technical
efficiency. For instance, in Figure 1, farm C is clearly inefficient as its output vector
does not lie on the border of the output set, and the resulting value of the distance
function is equal to ratio OC/OCd. The output distance function gives directly the
well-known Farrell (1957) output-based index of technical efficiency (Brummer,
Glauben and Thijssen, 2002). The output distance function is always homogenous of
degree one in outputs and inherits properties from the parent technology as detailed in
Färe and Primont (1995). 9

The last set of techniques available to investigate the efficiency of multi-output firms
relies on the estimation of a ray frontier production function, first proposed by
Lothgren (1997). His basic insight consists in expressing the output vector in polar
coordinates, which makes it possible to represent the technology by a function
relating the Euclidian norm of the output vector to the inputs and output mix,
represented by the output polar coordinates. Formally, the output vector y of
dimension M is expressed as:
y = y m(θ ( y )) (2)

8
This is a problem for the empirical application presented here as there is great heterogeneity within
our sample with regard to the subset of the four species actually produced on the farm, as discussed in
detail in the data section.
9
In particular, as described in Lovell et al. (1994), the output distance function is non-decreasing,
positively linearly homogeneous and convex in y, and decreasing in x.

237
 Paper 5/5

M
where y denotes the Euclidian norm of vector y ( y = ∑y
i =1
2
i ), θ(y) is an (M-1)

vector of polar coordinate angles of the output vector y, and the M functions mi: [0,
π/2]M-1 → [0,1] define the coordinates of the normalized output vector. This is
illustrated in the two-output case in Figure 1. The output vector of farm C is expressed
in terms of its norm, OC/OCr, and a single angle θc measuring the relative proportions
of fish and prawn outputs, i.e. the output mix. The two functions mf and mp of the
polar-coordinate angle θc simply define the (regular) coordinates of the normalized
output vector OCr obtained by radial projection of vector OC on the circle of radius 1.
Formally, the (M-1) polar coordinate angles are obtained by applying recursively the
following formulae (Löthgren, 1997):
y
θ i ( y ) = cos −1 ( i −1 i ), i = 1,...., M − 1 (3)
y ∏ sin θ j ( y )
j =0

where sin θ 0 = cos θ M = 1 . Hence, the first angle θ1 is equal to cos −1 ( y1 / y ) ; the
second angle θ2 is equal to cos −1 ( y 2 / y sin θ1 ) and so on. Note that all (M-1)
functions θ(y) are homogenous of degree zero in outputs, which simply reflects that
they capture only the proportions of outputs in vector y. The coordinates of the
normalized output vector are also easily recovered as:
i −1
mi (θ ( y )) = cos θ i ( y )∏ sin θ j ( y ) (4)
j =0

This set up allows us to represent any technology by a multi-output ray production

function f(x,θ(y)) as follows:
f ( x, θ ( y )) = max{ρ > 0 : ρ .m(θ ( y )) ∈ P ( x )} (5)
This function gives the maximum norm of the output vector that the firm can produce,
given a vector of inputs x and the existing technology, and assuming that any increase
in production would involve a proportional increase in all individual outputs. Hence,
any technology feasible input-output combination (x, y) is defined by the
inequality f ( x,θ ( y )) ≥ y . In terms of Figure 1, the value of the ray production
function is simply equal for farm C to the ratio OCd/OCr. Under the assumption of
strong input disposability, the ray function is positively monotonic in inputs
(Löthgren, 2000).

In order to understand how the ray production function can be used to measure
efficiency, it suffices to recognize that the ray production and output distance
functions are closely related to each other. It follows from equation (5) that, for any
observed output vector y, the radial frontier output vector is simply defined by
f(x,θ(y)).m(θ(y)) so that the distance function is recovered as:
y
Do ( x , y ) = (6)
f ( x, θ ( y ))
This is indeed observed in our graphical example, as ratio OC/OCd is obviously equal
to OC/OCr divided by OCd/OCr. This relationship is most important because we know
that virtually all the properties of a multi-output technology can be recovered from the
distance function. For instance, Brummer, Glauben and Thijssen (2002) and Irz and
Hadley (2003) use it to characterize technological change and productivity growth,

238
 Paper 5/5

while Kim (2000) derives measures of output substitutability from it. Equation (6)
therefore implies that the same can be done from the ray production function. For our
purpose, it is sufficient to recognize that output elasticities are easily derived from the
ray production function as:
∂ ln y ∂f ( x;θ ( y )) xj
ε y,x j = = . (7)
∂ ln x j ∂x j f ( x, θ ( y ))
This expression gives the percentage change in all outputs resulting from a one
percent change in input j and is expected to take a positive value (Fousekis, 2000).
Alternatively, the appendix demonstrates that because the ray production function
entertains some duality with both the maximum revenue and minimum cost functions,
this elasticity can be interpreted as the revenue elasticity or the scale-adjusted cost
share of input j. The scale elasticity follows immediately (Löthgren, 2000):
∇ f ( x;θ ( y )).x
ε scale = x (8)
f ( x, θ ( y ))
This elasticity should be compared to unity to establish whether the firm operates
under decreasing, constant or increasing returns to scale. Finally, the derivatives with
respect to the coordinate angles reflect the change in output norm when the output
mix is changed along the production frontier. They therefore relate to the degree of
substitutability of the different outputs, but in a rather indirect way as demonstrated in
the appendix, where the expression for the marginal rate of transformation between
any two outputs i and j is derived as:
M −1
∂ ln f ∂ ln θ m
y j (∑ − 1)
pj m =1 ∂ ln θ m ∂ ln y j
= (9)
M −1
∂ ln f ∂ ln θ m
yi ( ∑
pi
− 1)
m =1 ∂ ln θ m ∂ ln y i

We note, however, that these derivatives are simultaneously equal to zero if and only
if the PPF is, at the point of approximation, a perfect sphere in output space.
Furthermore, this expression could easily be used to derive to Morishima-like
elasticities of output substitution.

We conclude from this analysis that, from a theoretical point of view, the ray
production function and the output distance function are superior, as a basis for
efficiency measurement of aquaculture farms, to the alternatives presented in the
literature (dual functions, aggregate production functions and transformation
functions).

2.2 Implementation issues

While from a theoretical point of view the output distance function and ray production
function appear equally satisfactory for our purpose of measuring efficiency in
aquaculture, the same does not hold from an empirical/econometric point of view 10 . A
first issue relates to the fact that the distance function is linear homogenous in outputs.
Imposing this property globally requires the use of a logarithmic functional form that
cannot accommodate zero values on either inputs or outputs. In fact, all of the
10
Note that distance functions can be used in a non-parametric setting in order to measure technical
efficiency, as is the case for instance in Data Envelopment Analysis. Because these techniques do not
account for noise, and production shocks seem important in aquaculture as indicated for instance by the
levels and variability of mortality rates of prawns, we do not think that they are suitable for our
purpose. We therefore focus our discussion on parametric techniques.

239
 Paper 5/5

published papers on distance functions of which we are aware use a transcendental

logarithmic functional form, in order to impose homogeneity while conferring
sufficient flexibility to the parametric function. A common practice consists then in
replacing zero values by ‘small numbers’ (see Morrison-Paul, Johnston and Frengley
(2000) and Fousekis (2002) for two recent examples) but this seems highly
unsatisfactory as the logarithmic function goes asymptotically to minus infinity at
zero. In fact, Battese (1997) explores this problem in the context of a Cobb-Douglas
production function to conclude that replacement of zero values by small numbers can
seriously bias the parameter estimates. Given that most farms in our sample do not
produce all four outputs, this problem represents a major obstacle to the estimation of
an output distance function on our data.

The second issue arises from the fact that the value of the distance function is
unobservable so that an expression of the form D=f(x,y) is not estimable directly by
standard regression techniques. Following Lovell et al. (1994), this problem is usually
circumvented by modifying the regression equation based on the homogeneity
properties of the distance function. However, it is feared that this clever
transformation of the estimable equation might lead to possible endogeneity of the
regressors (Grosskopf et al., 1997; Löthgren, 2000).

By contrast, no homogeneity restriction needs be imposed on the ray production

function, which can therefore be represented by non-logarithmic functional forms and
hence accommodate zero values. Furthermore, it is also believed that the endogeneity
problem highlighted above for the distance function does not apply to the ray
production function (Löthgren, 2000). Hence, we choose to pursue our investigation
of efficiency of aquaculture farms in the Philippines based on the estimation of a ray
production function.

3. Estimation Strategy
The estimation of firm-level efficiency scores from a ray production function follows
the stochastic frontier methodology initially proposed by Aigner, Lovell and Schmidt
(1977). Accordingly, a scalar-valued composed error term is introduced in the
empirical ray production function 11 :
y = f ( x,θ ( y ); β ) + v − u (10)
where β is a vector of parameters to be estimated; v is a symmetric random variable
that is independently and identically distributed across individuals; and u is a non-
negative random variable. This specification recognizes the fact that production is
first affected by random shocks and measurement errors, which are captured by the
disturbance term v. However, the productive performance of farms is also determined
by the quality of managerial decisions and it is likely that some farmers make
mistakes, i.e., that they are technically inefficient. This is formally captured by the
random variable u that describes the deviation of the norm of the observed output
vector y from the maximum achievable norm f ( x, θ ( y ); β )e , which is conditional on
v

the exogenous shock v.

Given a parameterisation of the ray production function and distributional

assumptions on the random terms, equation (10) can be estimated by the maximum
11
Notice that the error term is introduced in an additive rather than multiplicative way because, as
explained earlier, we do not want to use a logarithmic functional form due to the ‘zero value’ problem.

240
 Paper 5/5

likelihood methods that have now become commonplace in the stochastic frontier
literature 12 . All models consider that the random error term v follows a normal
distribution N(0,σ2v) but differ with respect to the distribution of inefficiencies u. A
first generation of models considers that this term is identically and independently
distributed, following a half-normal distribution (Aigner, Lovell and Schmidt, 1977),
truncated normal (Stevenson, 1980) or gamma (Greene, 1990) distribution. However,
by assuming that inefficiencies are identically distributed, all of these models
implicitly assume that there is no relationship between efficiency and farm-specific
characteristics, as was first noted by Kumbhakar, Gush and McGuckin (1991).
Consequently, they are obviously ill-suited to the analysis of the inverse relationship.

Fortunately, several models have been developed to simultaneously measure

inefficiencies and identify their farm-level determinants. We adopt the formulation of
Battese and Coelli (1995) who relax the assumption of identically distributed
inefficiency terms by considering that ui is obtained by truncation at zero of a normal
variable N(μi; σu2) where: 13
μ i = ziδ (11)
The term zi denotes a vector of potential determinants of inefficiencies, including farm
size, while δ is a vector of parameters to be estimated. Note that because the
inefficiency effects enter the model in a highly non-linear way, there is no
identification problem when using the same variable in the specification of the ray
production function and as an inefficiency effect 14 . The likelihood function is derived
algebraically as in Battese in Coelli (1993) and it can then be maximised numerically
to produce estimates of both the ray production function and the vector of parameters
δ. Further, while the individual inefficiency levels are not directly observable, the
method allows for calculation of their predictors by applying the procedure first
proposed by Jondrow et al. (1982). As the expressions for these predictors are
presented in Battese and Coelli (1993) only for the multiplicative model, while ours is
additive, they are worth reporting here. First, the conditional expectation of the
inefficiency term u given a total residual e=v-u is derived from the expression of the
conditional density function of u given e derived in full in Battese and Coelli (1993):
φ (μ* / σ * )
E (u e) = μ * + σ * (12)
Φ( μ * / σ * )
where:
σ 2 zδ − σ u2 e
μ* = v 2
σ u + σ v2
(13 a & b)
σvσu
2 2
σ *2 =
σ u2 + σ v2
These expressions express mathematically that the random variable u, conditional on
e, is simply obtained by truncation at zero of the normal variable N(μ*,σ*2). The Farell
output-oriented efficiency score follows immediately:
12
See Coelli, Rao and Battese (1998) for an introductory presentation of this literature and Kumbhakar
and Lovell (2000) for a more detailed and technical one.
13
The individual subscript i was ignored up to this point for notational clarity.
14
An example of a stochastic production frontier where land appears both as an input and as an
inefficiency effect is Ngwenya, Battese and Fleming (1997), cited on page 212 of Coelli, Rao and
Battese (1998). The issue of identification is also discussed in Battese and Coelli (1995) where a time
trend is used to capture both technological change and inefficiency change over time.

241
 Paper 5/5

∧ ∧

∧
y − E (u e)
TE = ∧
(14)
y
∧ ∧ ∧
where y denotes the fitted output norm f ( x,θ ( y ); β ) and e is the estimated residual.

Next we turn to the issue of quantifying the explanatory power of the inefficiency
effects introduced in vector z, which is motivated by our primary aim of exploring the
robustness of any potential IR in Philippine aquaculture. This problem has been
largely ignored in the literature, as the only attempt at tackling it of which we are
aware is Pascoe and Coglan (2002) who, on a fisheries model, develop a procedure to
isolate the variation in inefficiencies attributable to the inefficient effect variables.
The procedure simply involves regressing the estimated technical efficiency scores
against the variables introduced as inefficiency effects by ordinary least squares. This
approach is ad hoc and seems unsatisfactory because it fails to recognize the highly
non-linear way in which the inefficiency effects enter our model. From equation (11),
it is evident that the mean of the normal variable truncated at zero to model
inefficiencies is a linear function of the z variables but this implies that the
relationship between predicted efficiencies (14) and the z variables takes a complex
non-linear form. We therefore prefer to investigate this question differently.

A first approach compares the full specification of the model to a restricted one where
this variable is dropped from vector z in equation (11). The comparison is based on
the decomposition of the total variance term e into its random shock and inefficiency
components u and v. Coelli (1995) establishes that the relative contribution of
inefficiency to the variance of the error is given by:
γ*=γ/[γ+(1-γ)π/(π-2)] (15)
where parameter γ=σu /(σu +σv ). This quantity captures the variation in production
2 2 2

not accounted for by physical factors that is attributed to inefficiencies rather than
random shocks. Hence, the difference between this quantity for the full model and the
restricted model gives us directly a measure, in percentage terms, of the explanatory
power of the inefficiency effect zi.

We would also like to be able to measure the strength of the relationship between any
zi variable and technical efficiency by calculating a standard elasticity but once again,
the literature seems to have ignored this issue. From equations (12) and (13), one can
derive the responsiveness of the conditional predictor of u to a change in any
inefficiency effect variable:
∂E (u e) z i δ i (1 − γ ) μ*
= [ Φ ( μ *
/ σ *
)( Φ ( μ *
/ σ *
) − φ ( μ * / σ * )) − φ 2 ( μ * / σ * )]
∂ ln z i (Φ ( μ / σ ))
* * 2
σ *

(16)
Using this expression in equation (14) defining the efficiency score, one obtains:
∂TE z δ (1 − γ ) μ*
=−∧ i i [Φ ( μ * / σ * )(Φ( μ * / σ * ) − * φ ( μ * / σ * )) − φ 2 ( μ * / σ * )]
∂ ln z i y (Φ ( μ * / σ * )) 2 σ
(17)
This elasticity gives the percentage change in efficiency resulting from a unit
percentage change in the inefficiency effect variable zi. Note that it depends not only

242
 Paper 5/5

on the parameter estimates but also on the data so that it can be estimated at any
sample point or at the sample mean. The empirical section of the paper uses this
expression to derive what we call the technical efficiency elasticity of farm size.
Alternatively, Kumbhakar and Lovell (2002) propose to use the mode of the
distribution of u given e as predictor of the inefficiency variable, which gives for our
model:
⎧μ * if μ * ≥ 0
M (u e) = ⎨ (18)
⎩0 otherwise
The resulting elasticity of technical efficiency with respect to any inefficiency effect
variable follows immediately:
⎧ z i δ i (1 − γ )
∂TE ⎪− ∧
if μ * ≥ 0
=⎨ y (19)
∂ ln z i ⎪
⎩0 otherwise

4. Data and estimable model

The two main regions of the Philippines for brackishwater pond aquaculture were
selected for this particular study. Region three covers the central part of the northern
island of Luzon and has brackishwater fish ponds in the four provinces of Pampanga,
Bulacan, Bataan and Zambales. Region six is located in the Western Visayas, central
Philippines, and includes the provinces of Iloilo, Capiz, Negros Occidental and Aklan.
The sample was stratified by farm size and by province, based on census data from
1997 provided by the Bureau of Agricultural Statistics. Production and socio-
economic data were then collected by interviews with farm operators and caretakers
(salaried supervisors). A total of more than 150 farms were initially surveyed but
several observations were dropped because of inconsistencies and/or missing values,
so that our analysis is based on a sample of 127 individuals.

Table 1 presents the summary statistics of the production variables. Starting with the
land input, it is apparent from our data that the farms in the study area are relatively
large, with an average surface area of more than eleven hectares 15 . Further, land is
also unequally distributed, ranging from one tenth of a hectare for the smallest farm to
130 hectares for the largest, and the Gini coefficient of land concentration for our
sample is equal to 0.67. This heterogeneity in farm size gives some relevance to the
investigation of the IR pursued in this paper. Table 1 also informs us about the type of
aquaculture practiced by the farms in the sample. First, in monetary terms, the main
intermediate input corresponds to the seeds (“fry” for prawns, “juveniles” for crabs
and “fingerlings” for milkfish and tilapia), followed by the feeds and, finally,
fertilisers. This simple fact reflects the extensive nature of brackishwater aquaculture
in the Philippines, as even in semi-intensive production systems, the feeds account for
the major share of cash costs. Also, the substantial cost of fertilisers reveals that farm
operators attempt to bolster the natural productivity of aquaculture ponds, while the
production process in intensive aquaculture relies solely on the provision of feeds
from an external source for the growth of the cultivated species. Finally, the summary
statistics also suggest that labour represents an important cost of production, as the
wage rate for farm labour is approximately 150 PhP/day in region 3 and 100 PhP/day

15
This can be compared to an average size of prawn farms in Thailand of only 2.16 hectares (Yap,
1999).

243
 Paper 5/5

in region 6, so that the total wage bills exceeds, on average, the cost of any individual
intermediate input.

With respect to outputs, milkfish is the dominant production in volume. This is not
surprising as the polyculture production system described here represents a recent
evolution of the traditional milkfish monoculture system (Chong et al., 1984). The
average milkfish yield of less than 500kg per hectare confirms the extensive nature of
the production process. The volumes produced of the other species appear relatively
small compared to that of milkfish but the relative importance of the species is
different in value terms. Given that prawns fetch a price nearly ten times as high as
that of milkfish per weight unit, they actually represent the dominant production in
terms of revenue share 16 . This price differential is explained in part by the fact that
milkfish and tilapia are consumed domestically, while an important proportion of the
prawns are exported to the high-income markets of Japan and the United States.
However, notice that crabs, which are also exclusively sold on domestic markets, also
receive high prices and are therefore important productions in economic terms.

Finally, Table 1 also suggests that the farms in the sample choose different
associations of species, as average output for each species differs whether it is
computed on the whole sample of farms or on the sub-sample with a strictly positive
output for that particular species. Figure 2 represents the distribution of the number of
species grown on the farm and brings two valuable insights. First, a large majority of
farms do indeed practice the polyculture of at least two species, hence justifying our
earlier discussion on multi-product technologies. And second, the association of all
four species is only adopted by a relatively small fraction of the sample farms,
implying that there is a large number of zero output values in the sample.

We choose a quadratic functional form as a first step in estimating the output ray
function defined in equation (7):
M + K −1 M + K −1
y = α0 + ∑α
j =1
jwj + ∑
j≤k
∑β
k =1
jk w j wk + D − u + v (20)

where the vector w includes each of the (M-1) polar coordinate angles θ(y) and the K
inputs, and D is a regional dummy taking a value of unity for the farms located in
region 6. 17 The quadratic production function is a flexible functional form in the sense
that it can serve as a local second-order approximation to any unknown production
function. This specification therefore gives flexibility to the model which can
accommodate zero values on both inputs and outputs.

The empirical specification includes the three following inputs: land and labour,
defined as in Table 1 as the total surface area of the aquaculture farm and the number
of man days of labour used on the farm; and intermediate inputs, expressed in value
terms, and hence representing an aggregate of the feed, fry/fingerling and fertiliser
inputs. On the output side, all four productions were used to define the three polar
coordinate angles for tilapia, crabs and prawns. The last step in specifying the model
consists of choosing the inefficiency effects that enter equation (11). The literature

16
The average prices per kilogram for our sample are 45PhP for milkfish, 31PhP for tilapia, 412PhP
for prawns and 210 PhP for crabs.
17
We introduce the regional dummy because the preliminary OLS regressions discussed below suggest
that there might be technological differences between the two regions.

244
 Paper 5/5

suggests at that level that one should choose variables susceptible of influencing the
adoption of particular management practices or the determinants of their adoption (Irz
and McKenzie, 2003). Given the focus of this paper on the IR relationship, farm size
is included as it is our aim to establish whether small and large farms adopt different
management practices that lead to differences in efficiency. It is also possible that
management differs across regions, and we therefore include the regional dummy as
well as inefficiency effect. Other variables, such as training and experience of the
operator, probably have an influence on efficiency but our data unfortunately does not
allow for their inclusion.

5. Empirical Results
5. 1 Partial productivity indicators
We start our analysis of the IR relationship by investigating the relationship between
farm size and land productivity. Although imperfect as outlined earlier, partial
productivity indicators have played an important role in the development of the IR
literature and are therefore worth reporting. Table 2 presents the results of three OLS
regressions relating a measure of land productivity to farm size and, in order to
account for possible regional effects, a regional dummy taking a value of unity for the
farms located in region 6. The first regression uses the crudest possible measure of
land productivity, i.e. harvest weight per hectare, and the results seemingly indicate a
significant and positive relationship between farm size and productivity.

However, it makes little sense to add weights of species that fetch widely different
prices and the second regression tackles this problem by measuring land productivity
in terms of revenue per hectare. The regression, which represents the equivalent of
those presented in the influential paper of Berry and Cline (1979), has a surprisingly
large explanatory power, as indicated by a R-square value of 0.42. Further, it reveals a
significant and negative relationship between farm size and revenue per hectare. The
coefficient of the farm size variable is easily interpreted as an elasticity and indicates
that a 10% increase in farm size results in a 2.2% decrease in revenue per hectare.
Finally, the coefficient of the regional dummy is also negative and significant,
indicating that farms tend to be substantially less productive in region 6 than in region
3.

The last regression accounts for differences in use of intermediate inputs when
comparing farms as it measures land productivity by gross margin per hectare 18 . It
does confirm to some extent the results of the previous regression but in a much
weaker way. In particular, the negative relationship land productivity-farm size
persists, with an elasticity of -0.18, but with only a modest level of statistical
significance; the explanatory power of the regression declines to a mere 13%; and it
transpires once again that land productivity in region 3 is significantly higher than in
region 6.

The difference in results between the two first regressions imply that, on a per hectare
basis, larger farms tend to produce more in weight but less in value terms than smaller
ones. Hence, it is likely that larger farms tend to choose output combinations with
greater emphasis on lower value species (tilapia, milkfish). The difference in results
18
Note that for this regression, the dependent variable is the level of the gross margin and not its
logarithm. This is so because some farms have negative gross margins which prohibits the use of a log-
log functional form for this regression.

245
 Paper 5/5

between the two last regressions indicates that the higher revenue per hectare
achieved by smaller farms, is, to a large extent, explained by a more intensive use of
intermediate inputs. All in all, we conclude from this simple analysis that there is
clear evidence of a negative relationship between intensity of land use and farm size
but only weak evidence of an inverse relationship between land productivity and farm
size.

5.2 Specification tests and the structure of the technology

The general specification of the stochastic frontier described above was tested against
a number of simpler alternatives in order to gain some insights into the structure of
the technology and inefficiencies. A second objective is to define a more
parsimonious specification as the full model requires estimation of a relatively large
number of parameters given the sample size 19 . The results of likelihood ratio tests are
presented in Table 3 20 .

First, we test the composed error specification against the hypothesis of absence of
inefficiencies by comparing the log-likelihood of our model against that obtained by
standard OLS regression. The likelihood ratio statistic of 39.2 exceeds by far its
critical value and we therefore conclude to the presence of substantial inefficiencies
across our sample farms 21 . From a methodological angle, this result implies that the
modelling of the technological relationship between inputs and outputs as a stochastic
ray production function rather than a deterministic one is strongly supported by our
data.

The second test investigates the explanatory power of the two variables introduced as
inefficiency effects in our specification. It is also strongly rejected, implying that the
regional dummy and farm size variables have, jointly, a statistically significant
influence on efficiency. The third test considers the null hypothesis that the regional
effects, introduced into the model through the regional dummy in the ray production
function and in the inefficiency effect component of the model (11), are inexistent.
The hypothesis is accepted as dropping these two variables from the model results in a
decrease in the likelihood function of only 0.8, which is marginal. This absence of
regional effects in the ray production function stands in sharp contrast to the results
obtained earlier based on partial productivity indicators. There is no inconsistency
here, however, because the ray production function can accommodate possible
differences in output mix across regions, while partial productivity indicators fail to
do so 22 . Next, the explanatory power of farm size on inefficiencies is tested and the
null hypothesis of no farm-size effect is strongly rejected. We therefore conclude
from these four tests that the regional dummy variable can be dropped from the
specification of the model, while farm size as an inefficient effect should be retained.

19
The total number of parameters in specification (20) is equal to 34, for a sample size of 127.
20
The test statistic is LR=-2*{ln(L(H0))-ln(L(H1))}, where L(H0) and L(H1) denote the values of the
likelihood function under the null and alternative hypotheses (Battese and Coelli, 1998).
21
Note that the null hypothesis includes the restriction σu=0. As this parameter is necessarily positive,
the test statistic follows a mixed chi-square distribution, the critical values of which are found in Kodde
and Palm (1986).
22
In terms of figure 1, the efficiency of farm C is measured radially, which means that this farm is
implicitly compared to farms with a similar output mix. By contrast, gross margin or revenue per
hectare measures fail to account for possible differences in output combinations when comparing
farms.

246
 Paper 5/5

The last two tests investigate the structure of the technology. Most interesting is the
question of whether inputs and outputs are separable, which is tested by comparing
our model to a restricted version where the parameters of all cross-terms between
inputs and polar coordinates angles in (20) are set equal to zero. The null hypothesis is
rejected at any sensible level of significance, which implies that it would not be
possible to aggregate consistently the four outputs into a single index. This is why the
ray production frontier is used rather than a frontier production function, which
requires output aggregation prior to estimation. Finally, the last test considers the null
hypothesis that the parameters associated with all the cross-terms among inputs and
among polar coordinate angles are equal to zero. It is strongly rejected.

Altogether, we conclude from this series of tests that there are substantial
inefficiencies among the sample farms, which are partially explained by farm size.
However, regional effects are not present, and the regional dummy is therefore
dropped from the model’s specification. Further simplification of the specification is
not possible as the tests indicate that the technology is truly multi-product and the
relationship among inputs and outputs is a complex one.

The results of the maximum likelihood estimation for our preferred specification are
presented in Table 4. We note that many of the coefficients present relatively low
levels of statistical significance but this should be expected as there is a high level of
collinearity among the covariates 23 . The individual parameters of the technology are
not directly interpretable and we therefore compute in Table 5 the elasticities of the
production ray function at the sample mean, together with their standard errors. Most
straightforward to interpret are the input elasticities described in equation (7). First,
there is a significant and positive relationship between land input and production, as a
one percent increase in farm size results in a 0.58% increase in all outputs. Hence,
land stands out as a key production factor which can be explained by the extensive
nature of the technology. Second, the elasticity with respect to intermediate inputs is
also highly significant, with a one percent increase in that aggregate resulting in a
0.36% increase in production. Finally, the elasticity with respect to labour is very
small, negative and not statistically significant, which means that the model fails to
capture a positive relationship between labour input and production. There are several
possible explanations for this negative result. One relates to the difficulty of
measuring labour input properly, in particular as far as farm operators are concerned.
We had to make sometimes crude assumptions in building the labour variable 24 ,
which might explain in part this statistically insignificant elasticity. Second, the
labour variable presents a high degree of collinearity with the other inputs, which can
be explained by the fact that most farm operators seem to adhere to the rule of thumb
‘one care taker for ten hectares’. Finally, we note that the finding of a negative and/or
insignificant labour elasticity, although paradoxical, represents an empirical regularity
(Whiteman, 1999). The scale elasticity (8) is obtained by summation of all three input
elasticities to give a value of 0.92, with a standard error of 0.11. Hence, the

23
This is not unusual when using flexible functional forms. For instance, in the full translog
specification of his model, Löthgren (2000) reports only five significant coefficients (5%) from a total
of 21 in the specification of the technology.
24
For instance, we had to assume that the operator was either working full time or half-time on the
farm, which probably does not reflect the heterogeneity of situations regarding the labour contribution
of the operator.

247
 Paper 5/5

technology exhibits slightly decreasing returns to scale at the sample mean but the
hypothesis of constant returns to scale cannot be rejected.

On the output side, the elasticities of the ray function with respect to the polar
coordinate angles are more difficult to interpret. We note, however, that the last
elasticity is large and strongly significant, which implies that the PPF, at the sample
mean, differs significantly from a perfect sphere. We conclude that the representation
of the technology that we obtain appears reasonably consistent with theoretical
expectations and provides an ex-post justification for the ray production function
approach that was chosen over the standard estimation of an aggregate production
function.

5.3 Inefficiencies and the inverse relationship

The likelihood ratio tests established that inefficiencies in our sample were
statistically significant, and that is confirmed by the fact that parameter γ in Table 4
has a large t-ratio. The mean efficiency score for the sample is equal to 0.37, 25 which
is very low and implies that the sample farms could potentially increase production
2.7 times without any increase in inputs or change in technology. This finding
confirms that of Irz and McKenzie (2003) and suggests that there is considerable
room for managerial improvement of the farms in the study area. It represents an
empirical validation of Yap’s contention that many brackishwater ponds are
‘underdeveloped and under-productive’ (Yap, 1999). It can also be explained by the
fact that extensive production systems have not been the focus of much research and
extension activity in the Philippines, which stands in sharp contrast with the situation
of the semi-intensive systems of tilapia production in fresh water that have benefited
from large R&D investment. The interviews carried out with farmers confirmed that
formal extension services are simply not regarded as an important source of technical
information by the operators of extensive farms. Finally, it is also necessary to
recognize that the extensive production systems considered here are intrinsically
complex and offer numerous opportunities for farmers to make mistakes. This is so
because these systems are open, due to the frequent exchange of the pond’s water,
which limits the farmer’s control of the production process. Furthermore, the
production process depends on the natural productivity of the pond, which itself
relates to the populations of various plankton and filamentous algae species that are
difficult to manage and sensitive to temperature, salinity, soil conditions and the
chemical and nutrient composition of the culture water (Arfi and Guiral, 1994). 26 The
situation is very different in intensive production where the growth of the target
species depends primarily on the feeds brought from outside of the farm and the pond
has little biological function beyond the provision of oxygen to the fish/crustaceans
(Kautsky et al., 2000) 27 .

Figure 3 presents the frequency distribution of efficiency scores and indicates a high
level of heterogeneity within the sample. The distribution is very flat, as reflected by a

25
In the additive model presented here, the predicted efficiency scores can take negative values, which
is theoretically impossible. We therefore replaced negative values by zeros when that occurred (in only
a few cases) prior to calculating this average.
26
We are thankful to Pierre Morrisens for this idea.
27
An analogy with agriculture might be useful here. Extensive aquaculture, like organic farming,
seems to be management intensive while intensive aquaculture, like conventional farming, tends to rely
on the application of standard technological packages that leave little initiative to the farmer.

248
 Paper 5/5

standard error of 0.26, and is spread over the whole possible range, from a minimum
of zero to a maximum of 0.97. We now turn to the direct analysis of the IR
relationship by investigating whether these large variations in technical efficiency
scores are related to farm size. The likelihood ratio tests already demonstrated the
existence of a significant relationship between farm size and efficiency, which can
also be seen in Table 4 by the large t-ratio of parameter δa. Furthermore, note that this
parameter takes a strictly positive sign, indicating that larger farms in our sample are
less efficient than smaller ones. Hence, we conclude to the existence of a statistically
significant IR in Philippine brackishwater aquaculture. We would like, however, to go
further in identifying the strength of this relationship, which cannot be established
directly from the parameter estimates and we now implement for that purpose the two
approaches discussed in the methodological section.

Our first contention was that a natural way of investigating the explanatory power of
farm size as an inefficiency effect consists of comparing the full model to a restricted
version where farm size is dropped as an inefficiency effect. We find that for the full
specification, inefficiencies account for 86% of the total variance term, implying that
the bulk of the variation in production not accounted for by physical factors is
attributed to inefficiencies rather than random shocks (i.e., what Pascoe and Coglan
(2002) referred to as luck). For the restricted model, where farm size is dropped from
the z vector, inefficiencies account for only 73% of the total variance term. It is
therefore logical to conclude that variations in production not accounted for by inputs
are attributable to random shocks for 14%; farm size for 13%; and unexplained
inefficiencies for 73%. This implies that the IR, although statistically significant,
appears to be of only limited quantitative importance.

Next, we compute the efficiency elasticity of farm size corresponding to equation (17)
and obtain a value of -0.137 at the sample mean. This indicates that a 10% increase in
farm size decreases the level of farm-level efficiency by a modest 1.4% for the
average farm and confirms the previous result of an IR of only limited strength. When
farm-level efficiency is predicted by the mode of the distribution of u given e, as in
equation (18), the efficiency elasticity at the sample mean takes the same value at the
three-digit level. The results are therefore robust to the choice of predictor used to
infer farm-level efficiency scores.

From a methodological point of view, it is also interesting to compare our results to

those obtained by application of the procedure suggested by Pascoe and Coglan
(2002) to quantify the explanatory power of the inefficiency effect variables. When
regressing by OLS the efficiency scores against the logarithm of farm size, we obtain
results that are simply inconsistent with the first-stage maximum-likelihood
estimation. The estimated efficiency elasticity of farm size at the sample mean is 0.24,
with a t-ratio of 3.52, and the R-square for this regression is only 9%. Clearly, the sign
of the elasticity is inconsistent with that of parameter δa in Table 4. Furthermore,
these results suggest that farm size explains only 7.7% (=0.09*0.86) of the variation
in outputs not accounted for by physical inputs, while we find a value almost twice as
large. Hence, we conclude that this procedure, which is not consistent with the
underlying model of efficiency measurement, can lead to erroneous conclusions
regarding both the direction and the strength of the relationship between inefficiency
effect variables and efficiency scores. We therefore believe that our methodological

249
 Paper 5/5

contribution is important in deriving the policy implications of the popular Battese

and Coelli (1995) model.

6. Discussion and conclusion

This paper uses a stochastic ray production function in order to investigate a potential
inverse relationship in Philippines brackishwater aquaculture, based on a cross-section
of 127 farms. The novelty of our approach is threefold: at a theoretical level, we
derive the dual properties of the ray production function, which are useful in
interpreting the parameter estimates; at an econometric level, we offer two different
approaches to quantity the explanatory power of the inefficiency effects, including
farm size, that are introduced in the model; and at an empirical level, ours is the first
attempt to model a farming technology by a ray production function.

The estimated multi-product technology is not separable in inputs and outputs,

implying that our approach is superior to the estimation of a stochastic production
function, which requires the aggregation of outputs into a single index. Returns to
scale are slightly decreasing at the sample mean but the CRS hypothesis cannot be
rejected. The distribution of efficiency score is spread out over the whole possible
range with an average value of 0.37, which is extremely low. Large potential
productivity gains are therefore achievable in the study area, without any change in
the technology, output mix or input combination. Is land redistribution or an
improvement in the functioning of the land market a key to achieving these efficiency
gains? Our analysis reveals that it is probably not the case. We find that there clearly
exists a significant inverse relationship between farm size and productivity, but that
the strength of this relationship is limited. Farm size explains only 13% of the
variability in outputs not accounted for by physical inputs, against 73% for
unidentified factors, and 14% for random shocks. The elasticities that we derive
indicate that when farm size doubles, efficiency decreases only by a modest 14%. It is
therefore likely that application of the land reform laws to brackishwater fish ponds,
which so far have secured exemptions via intense political lobbying by the pond
owners and lease holders, does not constitute a panacea to unlock the productive
potential of these areas. There might be legitimate reasons, on equity grounds, to call
for the removal of these exemptions, but the efficiency case for this policy carries
only limited weight. We know that the cost of implementing land redistribution
programs is always high, and that is likely to be particularly so in the Philippines
where issues of corruption, weak law enforcement and slow-moving bureaucracy in
coastal areas are well documented (Primavera, 2000).

Although this is an important result for policy formulation, it is unfortunately a

negative one as we are left with the conclusion that variations in efficiency relate to
unexplained factors. The best we can do at this level is therefore to speculate on the
underlying reasons leading to the poor average technical performance of the farms.
Here, we believe that the lack of R&D investment in brackishwater aquaculture is a
key constraint to the production and productivity growth of the sector. Even
aquaculture specialists recognize the difficulty to manage these systems, and it
therefore seems that there is a need to generate knowledge before even considering
investment in extension services. It remains to be shown that such investments are
economically desirable, but our results suggest that the potential gains from improved
farm management are very large.

250
 Paper 5/5

Figure 1: Alternative representations of a multi-output technology.

Prawn Output

Bp
P • •
Bf
• •
B •
Cd
• Ap

1 Cr •C
mp(θ ) c
•
A • • Af
θc
O mf(θc) 1
P’ Fish Output

Figure 2: Distribution of Number of Species Grown on the Farm

60 45 40
Frequency

40 31

20 11

0
1 2 3 4
Number of Species Farmed

251
 Paper 5/5

Figure 3: Frequency distribution of efficiency scores

25

20

15
Number of Farms

10

0
0-0.1 0.1-0.2 0.2-0.3 0.3-0.4 0.4-0.5 0.5-0.6 0.6-0.7 0.7-0.8 0.8-0.9 0.9-1
Efficiency Score

Table 1: Summary statistics*

Mean Std Dev.
Variable Mean Std Dev. Minimum Maximum
producers** Producers**

OUTPUTS
Milkfish (Kg) 4,356 5,075 1,098 13,339 0 80,000
Tilapia (Kg) 674 2,950 230 4,886 0 25,600
Prawns (Kg) 691 1,111 202 2,832 0 22,240
Crabs (Kg) 311 878 79 1,330 0 8,000

INPUTS
Land (ha) 11.5 - 1.9 - 0.1 130.0
Labour (man days) 1,160 - 220 - 187 26,312
Feeds (Pesos) 95,259 - 39,617 - 0 4,420,893
Fert (Pesos) 33,578 - 5,867 - 0 403,260
Fry/fingerlings
(Pesos) 183,770 - 46,518 - 0 4,140,000
*
All variables are expressed on a per year basis.
**
Mean and standard deviation computed over the sub-sample of farms producing a positive quantity
of the product.

252
 Paper 5/5

Table 2: Inverse relationship based on partial productivity measures

Dependent Variable
Log(harvest weight Log(harvest value per Gross Margin per
per hectare) hectare) hectare
Regressors:

Constant 6.52 11.60 90,940

(37.32) (71.78) (6.78)

Log(Farm size) 0.87 -0.22 -8,953

(11.91) (-3.30) (-1.60)

Regional
Dummy -0.85 -1.69 -63,528
(-4.04) (-8.69) (-3.93)

R2 0.55 0.42 0.13

Table 3: Specification tests

Log-likelihood LR statistic Critical Value Outcome

Null Hypothesis 5% 1%

1 No inefficiencies -113.0 39.2 8.8 12.5 Reject

2 No inefficiency effects -104.3 21.9 6.0 9.2 Reject
3 No regional effects -94.1 1.6 6.0 9.2 Accept
4 No farm size effect -100.0 13.4 3.8 6.6 Reject
5 Input-output separability -195.2 203.8 16.9 21.7 Reject
6 No cross-terms -136.2 85.7 12.6 16.8 Reject

253
 Paper 5/5

Table 4: estimation results of ray production frontier

Parameter Estimate t-ratio

Ray frontier
α0 0.569 0.68
αa 0.018 0.01
αl 0.965 0.83
αi -0.890 -1.40
αθt -0.685 -0.64
αθc 0.417 0.36
αθp -0.858 -1.87
βaa 0.047 1.27
βll 0.001 0.08
βii -0.092 -13.76
βθtθt 0.140 0.20
βθcθc -0.320 -0.48
βθpθp 0.448 2.00
βal 0.034 0.98
βai 0.403 11.72
βaθt -0.689 -0.69
βaθc 0.588 0.34
βaθp 0.998 1.88
βli -0.218 -5.16
βlθt 0.103 0.11
βlθc -0.920 -0.86
βlθp -0.031 -0.09
βiθt 0.843 1.53
βiθc 0.580 0.69
βiθp 0.368 1.14
βθtθc 0.120 0.22
βθtθp 0.084 0.27
βθcθp 0.066 0.24

Inefficiency Model
δ0 -2.356 -2.26
δa 2.292 4.43

Variance Parameters
σ2=σu2+σv2 1.052 3.02
γ=σu2/σ2 0.944 32.46

Log-likelihood -94.147
Subscript notations: a=land input, l=labour inputs, i=intermediate inputs, (θt, θc, θp)=three polar
coordinate angles corresponding to tilapia, crabs and prawns respectively; α0 and δo are the constant
parameters.

254
 Paper 5/5

Table 5: Elasticities of estimated ray production function (at sample mean)

Elasticity w.r.t. Estimate t-ratio

Land 0.58 4.51

Labour -0.03 -0.43
Intermediate
0.36 7.90
Inputs
θt 0.03 0.12

θc 0.08 0.34

θp 0.61 7.37

References:
Aigner D., K. Lovell and P. Schmidt. (1977), Formulation and Estimation of
Stochastic Frontier Models, Journal of Econometrics, 6, 21-37.

Ahmed, M. and M.H. Lorica (2002). Improving developing country food security
through aquaculture development – Lessons from Asia, Food Policy, 27, 125-141.

Alauddin, M. and C. Tisdell (1998). Bangladesh’s shrimp industry and sustainable

development: resource-use conflict and the environment, Asian Fisheries Science, 11:
97-120.

Arfi, R. and D. Guiral (1994). Chlorophyll budget in a productive tropical pond - algal
production, sedimentation, and grazing by microzooplankton and rotifers,
Hydrobiologia 272(1-3): 239-249.

Battese, G. E. (1997). A note on the estimation of Cobb-Douglas production functions

when some explanatory variables have zero values, Journal of Agricultural
Economics, 48(2): 250-253.

Battese, G. and T. Coelli (1993). A stochastic frontier production function

incorporating a model for technical inefficiency effects, Working Paper in Economics
and Statistics, No 69, University of New England, Department of Econometrics.

Battese, G. and T. Coelli (1995). A Model for Technical Efficiency Effects in a

Stochastic Frontier Production Function for Panel Data, Empirical Economics, 20,
325-32

Berry, A. and W. Cline (1979). Agrarian structure and productivity in developing

countries, John Hopkins University Press, Baltimore.

Brümmer, B., T. Glauben and G. Thijssen (2002). Decomposition of Productivity

Growth Using Distance Functions: The Case of Dairy Farms in Three European
Countries, American Journal of Agricultural Economics, 84(3): 628-44
Caves, D. W., L. R. Christensen and W. E. Diewert (1082). Multilateral comparisons
of output, input and productivity using superlative index numbers, Economic Journal,
46: 256-273.

255
 Paper 5/5

Chong, K. C., M. S. Lizarondo, Z. S. de la Cruz, C. V. Guerrero and I. R. Smith

(1984). Milkfish production dualism in the Philippines: A multidisciplinary
perspective on continuous low yields and constraints to aquaculture development,
ICLARM technical report 15, manila.

Coelli, T. (1995). Estimators and hypothesis tests for a stochastic production function:
a Monte Carlo analysis, Journal of Productivity Analysis, 6, 247-268.

Coelli, T. and S. Perelman (2000). Technical Efficiency of European railways: a

Distance Function Approach, Applied Economics, 32, pp. 1967-1976.

Coelli, T., S. Rahman and C. Thirtle (2002). Technical, allocative, cost and scale
efficiencies in Bangladesh rice cultivation: a non-parametric approach. Journal of
Agricultural Economics, 53(3): 607-626.

Coelli, T., D.S.P. Rao and G.E. Battese (1998). An introduction to efficiency and
productivity analysis, Kluwer Academic Publishers, Boston.

Coull, J. R. (1993). Will the blue revolution follow the green revolution? The modern
upsurge of aquaculture, Area, 25(4): 350-357.

Fan, S. and C. Chan-Kang (2003), Is small beautiful? Farm size, productivity and
poverty in Asian agriculture, Proceedings of the 25th conference of the International
Association of Agricultural Economists, pp. 112-123, Durban, South Africa, 16-22
August 2003.

FAO (2002). The state of world fisheries and aquaculture 2002, FAO, Rome.

Farrell, M.J. (1957). The Measurement of Technical Efficiency, Journal of the Royal
Statistical Society, 120:253-90.

Fousekis, P. (2002). Distance versus ray function: an application to the inshore fishery
of Greece, Marine Resource Economics, 17: 251-267.

Greene, W.H. (1990). A gamma-distributed stochastic frontier model, Journal of

Econometrics, 13, 101-116.

Irz, X. and D. Hadley (2003). Dual technological development in Botswana

agriculture: A stochastic input distance function approach. Proceedings of the 25th
conference of the International Association of Agricultural Economists, pp. 844-858,
Durban, South Africa, 16-22 August 2003.

Irz, X. and V. McKenzie (2003). Profitability and technical efficiency of aquaculture

systems in Pampanga, Philippines, Aquaculture Economics and Management, 7
(3&4): 195-212.

Jana, B. B. and C. D. Webster (2003). The potential and sustainability of aquaculture

in India, in Sustainable aquaculture: global perspectives, B. B. Jana and C. D. editors,
Food Products Press, Binghamton.

256
 Paper 5/5

Jha, R., P. Chitkara and S. Gupta (2000). Productivity, technical and allocative
efficiency and farm size in wheat farming in India: a DEA approach. Applied
Economics Letters, Vol. 7: 1-5.

Jondrow, J., K. Lovell, I. Matterov and P. Schmidt (1982). On the estimation of

technical ineffiency in the stochastic frontier production model, Journal of
Econometrics, August 1982, pp. 233-38.

Karagianis, G., S. D. Katranidis and V. Tzouvelekas (2002). Measuring and

attributing technical inefficiencies of seabass and seabream production in Greece,
Applied Economics Letters, 9: 519-522.

Kautsky, N., P. Ronnback, et al. (2000). Ecosystem perspectives on management of

disease in shrimp pond farming. Aquaculture 191(1-3): 145-161.

Kim, H.Y. (2000). The Antonelli Versus Hicks Elasticity of Complementarity and
Inverse Input Demand Systems, Australian Economics Papers, 39(2): 245-61

Kodde, D. A. and F.C. Palm (1986). Wald criteria for jointly testing equality and
inequality restrictions, Econometrica, 54, 1243-48.

Kumbhakar, S.C., S. Gosh and J.T. McGuckin (1991). A generalized production

frontier approach for estimating determinants of inefficiency in US dairy farms,
Journal of Business and Economic Statistics, 9, 279-286.

Kumbhakar, S.C. and C.A.K Lovell (2000). Stochastic Frontier Analysis, Cambridge
University Press, Cambridge.

Lothgren, M. (1997). Generalized stochastic frontier production models, Economics

Letters, 57(3): 255-259.

Lovell, C.A.K., S. Richardson, P. Travers and L.L. Wood (1994). Resources and
Functionings: A New View of Inequality in Australia, W. Eichlorn (ed.), Models and
Measurement of Welfare and Inequality, Berlin, Springer-Verlag.

Lothgren, M. (2000). Specification and estimation of stochastic multiple-output

production and technical inefficiency, Applied Economics, 32: 1533,1540.

Morrison Paul, C.J., W.E. Johnston and G.A.G. Frengley (2000). Efficiency in New
Zealand Sheep and Beef Farming: The Impacts of Regulatory Reform. The Review of
Economics and Statistics, 82(2): 325-37

Mundlak, Y. (1963). Specification and estimation of multiproduct production

functions, Journal of Farm Economics, 45: 433-443.

Naylor, R. L., R. J. Goldburg, J. H. Primavera, N. Kautsky, M. C. M. Beveridge, J.

Clay, C. Folke, J. Lubchenco, H. Mooney & M. Troell (2000). Effect of aquaculture
on world fish supplies, Nature, 405: 1017-1024.

257
 Paper 5/5

Orea, L., A. Alvarez and C.J. Morisson Paul (2002). Modeling and measuring
production processes for a multi-species fishery: Alternative technical efficiency
estimates for the Northern Spain hake fishery, Efficiency paper series, Department of
Economics, University of Oviedo.

Pascoe, S. and L. Coglan (2002). The contribution of unmeasurable inputs to fisheries

production: an analysis of technical efficiency of fishing vessels in the English
Channel, American Journal of Agricultural Economics, 84(3): 585-597.

Primavera, J. H. (1997). Socio-economic impacts of shrimp culture, Aquaculture

Research, 28(10): 815-827.

Primavera, J. H. (2000). Development and conservation of Philippine mangroves:

institutional issues, Ecological Economics, 35(1): 91-106.

Sen, A. (1962) An aspect of Indian agriculture, Economic and Political Weekly,

Annual Number, 2, 23-36.

Sharma, K.R. (1999). Technical efficiency of carp production in Pakistan,

Aquaculture Economics and Management, Vol. 3 (2): 131-41.

Sharma K. R., and P. Leung (1998). Technical efficiency of carp production in Nepal:
an application of stochastic frontier production function approach, Aquaculture
Economics and Management, Vol. 2 (3): 129-40.

Shephard, R. W. (1970). Theory of Cost and Production Functions. Princeton:

Princeton University Press.

Stevenson, R.E. (1980). Likelihood functions for generalized stochastic frontier

estimation, Journal of Econometrics, 13, 57-66.

Tisdell, C. (2003). Aquaculture’s potential impacts on conservation of wild stocks and

biodiversity, Aquaculture Economics and Management, 7(1/2): 155-165.

Whiteman, J.L. (1999). The measurement of efficiency where there are multiple
outputs, General paper G-134, Centre of Policy Studies, Monash University.

Yap, W. G. (1999). Rural aquaculture in the Philippines, FAO, Rome.

258
 Paper 5/5

Appendix to paper 5: Dual properties of the ray production function

The revenue maximisation problem can be written in terms of the ray function as:
R( p, x) = Max( py ) : f ( x, θ ( y )) ≥ y (A.1)
y

This constrained optimisation problem is solved by introducing the following

Lagrangian:
L( y, λ ) = py + λ[ f ( x, θ ( y )) − y ] (A.2)
The first order conditions are:
∂L M −1
∂f ∂θ m yj
= p j + λ[( ∑ )− 0] = 0 (A.3)
∂y j m =1 ∂θ m ∂y j f
∂L
= f ( x, θ ( y )) − y = 0 (A.4)
∂λ
which are more conveniently written as:
M −1
∂f ∂θ m yj
p j = −λ[( ∑ )− ] (A.5)
m =1 ∂θ m ∂y j f
f ( x,θ ( y)) = y (A.6)
Multiplying (A.5) by yj and summing over all outputs, one obtains:
M M M −1
∂f ∂θ m y 2j
∑j =1
p j y j = −λ ∑ [( ∑
j =1 m =1 ∂θ m ∂y j
)− ]
f
(A.7)
M

M M M −1
∂f ∂θ m
∑y 2
j

∑p y j = −λ[∑ ∑
j =1
yj − ] (A.8)
j =1 m =1 ∂θ m ∂y j
j
j =1 f

M −1
∂f M
∂θ m
The first term of this sum can be re-written as ∑ ∑ ∂y y j , but since the
m =1 ∂θ m j =1 j
M
∂θ m
function θm is homogenous of degree zero in y, ∑ y j = 0 . Further, using (A.4),
j =1 ∂y j

equation (A.8) reduces to:

M
R ( p, x ) = ∑ p j y j = λ . f ⇒ λ = R ( p , x ) / f (A.9)
j =1

This expression means that the Lagrange multiplier is simply the unit value of the
norm. Applying the envelop theorem to the original problem (A.2) therefore gives us:
∂R ∂f R ∂f ∂ ln R ∂ ln f
=λ = ⇒ = (A.10)
∂x k ∂x x f ∂x x ∂ ln x k ∂ ln x k
Hence, the elastictities of the revenue function and output ray function with respect to
any input k are equal and are expected to be positive. We also use (A.9) to rewrite
(A.3):
R M −1 ∂f ∂θ m yj
p j = − [( ∑ )− ] (A.11)
f m =1 ∂θ m ∂y j f
It follows that the marginal rate of transformation between two outputs is:

259
 Paper 5/5

M −1
∂f ∂θ m yj
(∑ )−
pj m =1 ∂θ m ∂y j f
= M −1 (A.12)
∂f ∂θ m yi
(∑
pi
)−
m =1 ∂θ m ∂y i f
It can also be written in terms of log derivatives as:
M −1
∂ ln f ∂ ln θ m
y j (∑ − 1)
pj m =1 ∂ ln θ m ∂ ln y j
= (A.13)
M −1
∂ ln f ∂ ln θ m
yi ( ∑
pi
− 1)
m =1 ∂ ln θ m ∂ ln y i

Suppose all the derivatives of the ray function with respect to the angles are equal to
0. This implies that the marginal of transformation becomes:
pj yj
= (A.14)
pi yi
The previous expression can only be valid if the PPF is perfectly approximated in the
plane (yi;yj) by a circle. Hence, the restriction that all derivatives of the ray production
function with respect to the (M-1) angles are equal to zero means that the PPF is a
perfect sphere of dimension M.

The ray function also shares some dual properties with the minimum cost function.
We proceed as before to rewrite the Lagrangian of the cost minimisation problem:
L( x, λ ) = − wx + λ[ f ( x,θ ( y )) − y ] (A.15)
The FOCs are:
∂L ∂f
= − wk + λ =0 (A.16)
∂x k ∂x k
∂L
= f ( x, θ ( y )) − y = 0 (A.17)
∂λ
Multiplying (A.15) by xk and summing over all inputs gives:
C
λ= (A.18)
f .ε scale
It follows from (A.16) that:
wk x k ∂ ln f 1
= Sk = (A.18)
C ∂ ln x k ε scale
The elasticity of the ray function with respect to any input xk is therefore interpreted
as the scale-adjusted (optimal) cost share of that input.

260