See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/289506093

Sugar, Fat, or Protein: Are All Food Insecure

Households Eating the Same? The Case of Rice
Producers in Peru

Chapter · January 2016

CITATIONS READS

0 48

1 author:

Pilar Useche
University of Florida
38 PUBLICATIONS 136 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Valuing Ecological Benefits in Coastal Ecuador View project

Ecological Benefits and Gender Roles in Cacao Agroforestry Systems in Ecuador View project

All content following this page was uploaded by Pilar Useche on 06 January 2016.

The user has requested enhancement of the downloaded file.

 1

Volume 16, Chapter 7

Sugar, Fat, or Protein: Are All Food Insecure Households Eating the Same?

The Case of Small Rice Producers in Peru

Pilar Useche* and Jennifer Twyman**

*University of Florida and Center for Latin American Studies

**International Center for Tropical Agriculture (CIAT-CGIAR)

 Volume 16, Chapter 7, page 1

Sugar, Fat, or Protein: Are All Food Insecure Households Eating the Same? The Case of

Small Rice Producers in Peru

Abstract:

As world food security concerns evolve and poor diet quality emerges as a major worry, several

studies propose new food security indicators. However, aggregation of multiple food security

dimensions in one index results in loss of information of potentially diverse food intake patterns

across heterogeneously constrained households. In particular, dietary composition may change

with the severity of constraints faced by the household. In this study, we examine the joint

demand for components of a household’s diet diversity and its relationship with household and

regional characteristics embodying diverse household constraints. We study food intake patterns

of rice producers in Peru, one of the countries with increasing climate change-related food

insecurity. We find diverse food intake patterns for households, depending on their wealth,

education, demographic structure, market access, and geographic location, as well as the shocks

faced in the past. We also find that several obesogenic foods are consumption complements and

milk is their strong substitute. Particularly vulnerable are female-headed households (too low

consumption of micronutrient rich foods) and of households with more children (too high

consumption of beverages with added sugars). Poor diet quality is also shown to be highly

associated with climate shocks.

Key Words: food security, diet quality, diet patterns, nutrition and agriculture.
 Volume 16, Chapter 7, page 2

Introduction

In consonance with changes in world food constraints, the concept of food security has evolved

from one primarily focused on food supply problems to considering food access, diversity, and

quality. The prevailing definition of food security, agreed upon at the 1996 World Food Summit

and reconfirmed in 2002, holds that food security exists when all people, at all times, “secure

access to sufficient amounts of safe and nutritious food for normal growth and development and

an active, healthy life” (Food and Agriculture Organization of the United Nations [FAO], 2001,

p. 49).

This definition rests on the three core and hierarchical pillars of food availability, access,

and adequate utilization, of which the last two components have proven to be the hardest to pin

down in terms of conceptualization and measurement (Barrett, 2010; Webb, Coates, Frongilio,

Rogers, Swindale, & Bilinsky, 2006). In a way, the hierarchical nature of these three goals

anticipates the increasing complexity in determining the factors that influence their achievement.

While access depends on the many socioeconomic factors that influence the range of choices that

are effectively available to a person (e.g., income, prevailing prices, distance to markets), the

adequate use of food depends not only on having access to nutritional food, but also on the

knowledge of what is adequate or nutritional and on the individual or household preferences over

this food. As such, carefully examining this last component of food security provides a rich

picture of the food insecurity conundrum, since it relates simultaneously to access and

availability. It is also influenced by risk, a factor that cuts across all components of food

insecurity.

Most studies focus on single proxies or on summary measures of adequate food

utilization (FAO, 2002; Radimer, Olson, & Campbell, 1990). This may be problematic as
 Volume 16, Chapter 7, page 3

different proxies influence prioritization among food security interventions differently, since

they capture or neglect different aspects of the concept (Barrett, 2010). As an example,

widespread indicators of food insecurity in the context of developing countries tend to capture

scarce food access (e.g., total caloric intake and anthropometric measures for under nutrition),

but they neglect concerns related to excess intake of certain nutrients and foods which increase

day to day due to shifting food consumption patterns in the developing world. Similarly,

aggregation of multiple dimensions in the form of indicators may result in important loss of

information and even erroneous policy recommendations if the base components are essentially

different in kind.

In this chapter, we examine food utilization by analysing the joint demand of households

for different dietary components. Rather than focusing on summary measures of the multi-

dimensional concept of dietary diversity, we analyse its multiple dimensions simultaneously,

preventing the loss of information that occurs by aggregating these components into an index,

yet acknowledging the interrelated nature of their consumption (e.g., due to preferences). Our

analysis captures the diverse food intake patterns of heterogeneously constrained households and

relates them to differences in access to and availability of food.

One of the most difficult issues inherent to the analysis of food utilization is that it

involves non-pecuniary comparisons across food groups, as well as subjective values of the

individuals consuming the food (preference and culture dependent). This implies that the

analysis of food security may have not only different data requirements, but also different

methods for the analysis of demand. For example, one of the most popular assumptions about

consumption groups in applied demand analysis is that preferences are additively separable

across groups. This simplifies econometric estimations of demand parameters, but implies that
 Volume 16, Chapter 7, page 4

no group occupies any special position in consumers overall preferences. This assumes that the

researcher can arbitrarily create new groups by combining any others, neglecting the existence of

any particular relationships between any pairs of groups. This assumption often leads to ruling

out the possibility of inferior goods, and other times to only allowing for goods that are

substitutes. However, in the case of food groups, consumers may see some of these as

complements (e.g., milk and coffee, or coffee and sugar) or, from a nutritional perspective, some

goods may be considered inferior (e.g., sugars and fats). Thus, we use an econometric approach

which accounts for potential complementarities (and substitutability) across food groups

consumed and estimate the strength and significance of these complementarities in consumer

preferences.

We apply this framework to the study of small rice farmers in Peru who face

heterogeneous constraints to access food markets and are affected by very diverse climatic

events. We address specific questions: Are the main access constraints to the consumption of

fruits and vegetables by poor households related to lack of education and knowledge about their

nutritional qualities, or rather lack of affordability? How do household diets change when the

household members face limited food access? Do different types of weather shocks affect food

consumption patterns distinctively? Do relatively wealthier households have a better diet? How

does access to markets influence consumption patterns?

In the following sections, we first elaborate on what is adequate food utilization and how

it is typically operationalized and measured. Second, we discuss potential determinants of food

access and utilization as described in the previous literature. Next, we describe the context of the

study setting in terms of agricultural production systems, climate and the trade environment. We
 Volume 16, Chapter 7, page 5

go on to discuss the data and method of empirical analysis. Finally, we present the results and

conclude with a brief discussion.

Adequate Food Utilization

Different elements of globalization and urbanization have been associated with negative changes

in dietary patterns: increases in sedentary lifestyles, increases in women entering the workforce,

changes in food systems such as patenting of seeds, increases in the food supply that is sold

through supermarkets, and the reduction of the influence of local culture and food traditions on

local diets. These economic development trends have an impact on diets that are recognized to be

heterogeneous and complex. Two important phenomena have been identified to be of global

relevance: dietary convergence and dietary adaptation. Convergence refers to an increased

reliance on a narrow base of staple grains; increased consumption of meat and meat products,

dairy products, edible oils, salt, and sugar; and a lower intake of dietary fiber. Adaptation

consists of an increased consumption of processed and standardized store bought foods and

eating behaviors driven by the appeal of new foods available in the market (Hawkes, 2006).

The degree to which convergence and adaptation have occurred in developing countries

has not been well studied, yet it has been established that these changes have been consistent

with increased consumption of macronutrients such as carbohydrates, fat, and protein; increases

in micronutrient deficiencies (vitamins and minerals); and increases in the incidence of diet-

related non-communicative (chronic) diseases such as obesity, coronary heart disease, diabetes,

and hypertension. These diseases foster persistent poverty and reinforce the consequences of

food insecurity (Barrett, 2010).

 Volume 16, Chapter 7, page 6

Dietary guidelines in developed countries now promote the consumption of fresh fruits

and vegetables, as well as the reduced intake of foods such as edible oils, animal foods, and

caloric sweeteners (Mendez & Popkin, 2004). Fresh fruits, vegetables, seeds, nuts, meats, and

dairy products contain the highest amounts of micronutrients which are more available in fresh

products. These recommended foods unfortunately tend to be relatively less available or more

expensive in supermarkets which exacerbate the prevalence of micronutrient and fiber

deficiencies. These deficiencies are often overlooked by basic measures of malnutrition since

they cannot be observed or diagnosed by simply measuring weight. In other words, underweight,

normal weight, and overweight individuals may all suffer from such deficiencies.

While there has been little change in the amount of fruits and vegetables consumed, very

large and rapid increases in the consumption of obesogenic energy-dense foods have been

observed. Increased sugar and sweetener use has been linked with the proliferation of processed

foods and beverages that are salt-laden and/or sugar-laden such as tea, coffee, cocoa, and soft

drinks (Popkin & Nielsen, 2003). Thus, micronutrient deficiencies and obesity have been linked

to the expansion of food value chains in developed and developing countries (Gomez et al.,

2013). In developing countries, women and children are disproportionately impacted by these

problems (Mendez & Popkin, 2004), and there are wide differences within countries between

wealthy and poor, and rural and urban areas.

The breadth of issues and complexities related to food insecurity yield widely used

measures ill-suited to capture the full panorama of the problem (Coates, 2013; Webb et al.,

2006). Until recently, measuring caloric intake or the adequacy of household food availability

has been the “benchmark” measure for food security (Chung, Haddad, Ramakrishna, & Riely,

1997; Maxwell, 1999). Developments in dietary patterns have prompted the creation of new
 Volume 16, Chapter 7, page 7

measurement tools in developed countries, based on evolving dietary guidelines regarding, for

example, the proportion of energy from fat; the ratio of polyunsaturated to saturated fat; and

intake of dietary fiber, cholesterol, and alcohol (Ruel, 2003). However, these new tools are not

available to analyse food security in developing countries where the necessary data are not

regularly collected and few or no local dietary guidelines exist.

One of the most popular approaches to examine food access and utilization in developing

countries has been to calculate indices of dietary diversity which count the number of foods or

food groups consumed by a household or individual over a given time period. Such indices

measure both access and utilization (as they are proxies for nutrient adequacy). Several studies

have shown the validity of this type of index (i.e., food consumption score and household dietary

diversity score), finding them to be significantly correlated with caloric intake and nutrient

adequacy (Rathnayake, Madushani, & Silva, 2012; Ruel, 2003). The nutrition literature has

found that increases in this dietary diversity measure are associated with socio-economic status

and household food security; adequate micronutrient density of complementary foods for infants

and young children; and macro- and micronutrient adequacy of the diet for non-breastfed

children, adolescents, and adults (Foote, Murphy, Wilkens, Basiotis, & Carlson, 2004; Hatloy,

Halland, Diarra, & Oshaug, 2000; Kennedy, Pedro, Seghieri, Nantel, & Brouwer, 2007; Steyn,

Nell, Nantel, Kennedy, & Labadarios, 2006).

In this chapter, we make use of this type of data collected for rice growers in Peru, yet we

decompose the index to explain food demand for separate but jointly determined consumption of

food groups. In this way, different dietary patterns of heterogeneous households are revealed, as

well as the complementarities among food groups for a specific context.

 Volume 16, Chapter 7, page 8

Determinants of Access and Utilization

Economic demand studies have typically focused on pecuniary measures of availability and

access and have often overlooked or oversimplified the facet of utilization. As Behrman and

Deolalikar (1987) recognize, even when access exists or increases, (e.g., income rises), it does

not necessarily imply that nutritional status improves for a specific population. Several studies

have shown that food expenditure-income elasticities are much higher than nutrition-income

elasticities, as increases in foods purchased may consist of non-nutritious foods. Also, at low

levels of household income a strong demand for tasty and palatable foodstuffs as opposed to

calories per se may exist, as evidenced by Greer and Thorbecke (1984) for Kenya. They used this

evidence to explain the existence of significantly higher food expenditure elasticities than calorie

elasticities in poor populations. Thus, while income may influence dietary patterns, there are

other factors, that shape food consumption and nutrition.

Some authors have illustrated the importance of heterogeneous tastes and preferences in

explaining the demand for different types of food. For example, Behrman and Wolfe (1984) have

argued that high food income elasticities may be consistent with low nutrient income elasticities.

In this sense, as observed by Shah (1983), even at low income levels, the weight placed on

nutritious foods may be considerably lower than the weight placed on taste when making food

demand decisions.

In agricultural settings, farm income and food prices play a role in access, yet the type of

production on the farm may have an important impact on utilization as well. Monoculture in

commercial farms makes farmers rely on cash income and market availability to obtain a

balanced diet, while farmers that intercrop their main products with other types of plants and

trees have a broader variety of foods directly available. In addition, intercropped plants often
 Volume 16, Chapter 7, page 9

provide for organic matter and nutrients that are absorbed by crops, increasing their nutritional

content (Davids, 2004).

Agricultural and food export and import policies also affect the availability of specific

foods by changing their relative prices and affordability (Kadiyala, Harris, Headey, Yosef, &

Gillespie, 2014). An additional pathway through which agriculture influences nutrition is related

to the gender division of labor; the way women use time often influences intra-household

allocations of food, health, and care. For example, a mother’s ability to provide adequate

nutrition may be influenced by her engagement in agriculture. Similar conclusions on

determinants of food security have been found by a number of studies (Hoddinott, 2012; World

Bank, 2007). Gomez et al. (2013) review these studies and summarize their views on the main

links of the food system to human nutrition. These are grouped in the following categories: food

production for own consumption, food availability through markets, incomes, food prices,

gender-specific time allocation and educated or informed consumer behavior.

Thus, beyond income and prices, it can be seen that preferences, tastes, gendered intra-

household responsibilities and rural versus urban food system contexts may play an important

role in the way households combine different types of food in their consumption basket.

Agricultural Context and Study Region

We study small rice producers in northern Peru. Rice plays a vital role in food security as one of

the most important staple foods in the country. Rice consumption in Peru is the third largest in

South America, where the average Peruvian consumes 48.7 kilograms (kg) of rice per year. As

an income source, rice is also important; it represents roughly 9.4% of the gross national product
 Volume 16, Chapter 7, page 10

and occupies almost 20% of cultivated area. In rural areas, it is a source of employment and the

livelihoods of many farmers depend on its cultivation.1

Rice cultivation is located mainly in the northern part of the country, in the valleys, forest

rim, and forest region. Our study focuses on the departments of Amazonas, Cajamarca, and San

Martin, in the forest region, and La Libertad, Lambayeque, and Piura, in the coastal area, which

comprise 70% of the national production according to Escobal and Agreda (1994) (see the

distribution of coastal versus forest areas in Table 1, third column).

The study area is characterized by high climatic heterogeneity which influences food

production systems. The coastal climate is characterized by very little rainfall. Mean temperature

is 18.1 Celsius, with a wide range of daily fluctuation between 8 and 30 Celsius. In the forest

area, mean temperature is higher and rainfall varies between 1,500 and 3,000 millimeters (mm)

or more per year distributed over a period of 9 to 11 months (Vera, 2006). In spite of the arid

nature of the soil in the coastal area, more market accessibility in this region has provided for

greater adoption of mechanization and improved varieties. As such, rice production in the coastal

area tends to be greater and more predictable than in the forest area. Furthermore, a large

proportion of fruit exports and commercial livestock farms are found in the coastal area. In 2006,

the main agroexport products of the region were green coffee, asparagus (fresh and preserved),

paprika, artichoke, fresh mangos, wood, evaporated milk, capsicum, and fresh grapes (La

Superintendencia Nacional de Aduanas y de Administración Tributaria [SUNAT], 2011). In

terms of livestock production activities, these are more intensive and specialized in the coastal

area than in the forest area, often requiring higher levels of organization and education.

While rain-fed forest rice production puts less stress on water resources, it tends to be less

profitable, partially due to the lower availability of improved varieties.2 Similarly, animal and
2
Data obtained from http://www.ricepedia.org. Accessed 04-20-2015.
 Volume 16, Chapter 7, page 11

livestock production diverges from the coastal area in that it is for subsistence or semi-

commercial purposes, and has lower levels of technological development. Traditional products

also vary by region of the country. Rice, cotton, maize, and beans are more characteristic of the

coastal area; potatoes, barley, corn, wheat, and fava beans of the high forest area; and coffee,

oranges corn, plantains, and cassava of the lower forest areas.

[Take in Table 7.1]

Beyond naturally determined differences in food production, food access related to trade

may shape a household’s dietary patterns. Peru has a negative food trade balance with its

northern neighbors, where imports coming from Ecuador and Colombia have been increasing

through the Andean Community of Nations.3 Border integration with Ecuador has enabled trade

to increase (e.g., by 200% from 2005 to 2012).4 The main Peruvian imports for 2011 by value

included corn, wheat, cake of soybeans, soybean oil, and palm oil, prepared foods (including

both crop and livestock products),5 and prepared/preserved fruit, refined sugar, apples, and

barley.

Methodology

To explain the food demand of Peruvian rice producers for jointly determined consumption of

food groups, we first calculate the household dietary diversity index (HDDI) for the sampled

households. The number of food groups, yj, consumed by a household, h, during any 24-hour
2
http://agroaldia.minag.gob.pe/biblioteca/download/pdf/agroeconomia/agroeconomiaarroz3.pdf.
3
SUNAT (2011) at http://www.sunat.gob.pe/estad-comExt/modelo_web/web_estadistica.htm.
4
http://www.andes.info.ec/es/economia/comercio-ecuador-peru-incremento-200-siete-anos.html..
5
Inter alia: homogenized composite food preparations; soups and broths; ketchup and other sauces; mixed
condiments and seasonings; vinegar and substitutes; yeast and baking powders; stuffed pasta, whether or not
cooked; couscous; protein concentrates; and prepared fruit (frozen, cooked, prepared or preserved fruits, nuts and
peel, jam, paste, and marmalade). See FAOSTAT food group definitions for crops and livestock products at
http://faostat3.fao.org/search/food%20prep%20nes/E.
 Volume 16, Chapter 7, page 12

period ranges between 1 and 12. Initially, we use this index as the dependent variable in a Tobit

regression analysis (accounting for censoring in this variable). Formally, we describe this

relationship as:

∑𝑗 𝑦 = 𝐻𝐷𝐷𝐼ℎ = ∑𝒌 𝛽𝑘 𝑿𝑘ℎ + 𝑒ℎ (7.1)

𝑗ℎ

where 𝑋𝑘ℎ are covariates k for household h which explain dietary diversity, 𝛽𝑘 captures the

average effect of an explanatory variable on the HDDI across households, and eh is a normally

distributed error term with mean of zero and standard deviation of σ. It is important to highlight

that in this type of analysis as an explanatory variable Xk increases only the final, aggregated

effect on HDDI will be estimated, yet the pathways to this increase or decrease in dietary

diversity will be invisible. This may be problematic since it can hide important changes in

dietary patterns that occur in populations and also yield insignificant the effect of covariates that

may have a significant but contrary effect on different components of the dietary index. As an

example, households that were only consuming grains and tubers may start consuming animal

protein and vegetables as their income increases by 10 units, but they may reduce their

consumption of tubers if these are considered inferior goods. Similarly, households whose

income decreases by 10 units may stop consuming expensive vegetables and substitute them for

sugar and fat based foods. In both cases, the HDDI score will increase by one point, yet the 𝛽𝑘

parameter for income will not be significant, as no monotonous relationship with HDDI exists.

To visualize these relationships and dietary patterns, we estimate a model of joint demand

for isolated components of the HDDI, where the effect of each covariate Xk on the demand for

each component j can be different (the effect of covariate k on the demand for group j is denoted

by 𝛽𝑘𝑗 ). To examine the determinants of the demand for dietary components we use a

multivariate probit model (MVP). This discrete choice model assumes normality of the
 Volume 16, Chapter 7, page 13

distribution of the error terms and simultaneously examines the relationships between each

choice option and a set of explanatory variables by allowing errors to be freely correlated (Lin,

Jensen, & Yen, 2005). This model is superior to univariate models and multinomial models

because it explicitly recognizes and provides estimates of the correlation among the demands for

different demand components.

Data and Regression Variables

To analyze the dietary patterns of small rice farmers in Peru we used a survey of 497 rice

producers administered by the Peruvian Institute for Agrarian Innovation (INIA) and the

International Center for Tropical Agriculture (CIAT). This research included smallholders in the

main rice producing departments in the country (Amazonas, Cajamarca, La Libertad,

Lambayeque, Piura, and San Martin). The field work was carried out between October and

December 2012 in both the coastal and forest areas, and targeted rice farmers with less than 10

hectares. The distribution of the sample was relatively balanced across agro-ecological zones as

shown in Table 7.1 (48% forest and 52% coast), as well as across Northern and Southern

departments (48% North and 52% South).

The questionnaire consisted of various questions divided into twelve modules, including

a module on food security for calculation of the Household Diversity Index (HDDI). It asked

respondents about the household’s consumption of twelve food groups on the previous day:

cereals); roots, tubers, and plantains; vegetables; fruits; meat and poultry; eggs; fish and seafood;

legumes and dried fruit; milk and dairy products; foods based on oil, fat, and butter; sugar and

honey; and others such as soda, coffee, tea, and condiments.

 Volume 16, Chapter 7, page 14

A perusal of consumption by food groups (Figure 7.1) shows that about every household

in the sample consumes cereals, which is explained by the fact that they are all rice growers.

Other food groups that are consumed by a large majority of households are roots, tubers and

plantains (88%); soda, coffee, tea, and condiments (86%); sugars and honey (78%); and meats

and poultry (72%). A medium range of consumption is evidenced for legumes and beans (70%),

fats and oils (68%), fruits (65%), and vegetables (61%). Less than half of the households

consume fish and seafood, milk, dairy products, and eggs (48%). Examining food diversity in

terms of the number of food groups consumed by households, we find that the majority of

households consumed six to ten different food groups (76%), with only 8% of the sample

consuming food from five or less food groups.

[Take in Figure 7.1 Here]

Regression Analysis

As explained earlier in the methodological section, we first regress the HDDI on a set of

covariates through a tobit analysis. We then decompose this index into different food group

components and use them as dependent variables in a multivariate probit framework. The HDDI

index is originally calculated based on the twelve different food groups presented earlier. To

keep enough degrees of freedom to identify the coefficients of all parameters in the multivariate

regression, we aggregate several of the groups.6 First, we use five aggregated groups, combining

roots and tubers with legumes and beans (as less perishable goods); vegetables and fruits (as

highly perishable and micronutrient rich); meats (including poultry), eggs, and fish (as proteins);

milk and dairy products are left as a separate category; and sugars and honey are combined with

6
The cereals group was excluded from the disaggregated food group analysis because nearly all respondents
reported consuming cereals.
 Volume 16, Chapter 7, page 15

foods based on oils and fats, as well as with the coffee, tea, and soda group (these are more

processed foods that may be seen as inferior from a nutritional perspective).

While the former categories may make sense from a researcher’s perspective,

households’ preferences may deem some groups as “very different.” For example, tubers are

sometimes seen as inferior goods in market economies, where wheat and rice have become more

popular staples. In many Latin American countries, beans are a complement of rice in dishes, but

not of cassava or potatoes. Similarly, coffee and tea may have nothing to do with oil- and fat-

based foods. Thus, we perform a more disaggregated analysis with eight food groups by

separating roots and tubers from legumes and beans, as well as sugars and honey from oils and

fats, as well as from coffee, tea and soda.7

To capture the different socioeconomic constraints that households face in their access to

different food groups, as well as the heterogeneity of households that influence preferences and

gendered use of time, we include the variables presented in Table 7.2. In particular, we include a

measure of wealth called the progress out of poverty index (PPI) which determines the likelihood

that a household is living below the poverty line (Schreiner, 2012). This index is positively

related to a household’s dietary diversity (Hatloy et al., 2000). Its calculation is based on ten

questions related to household structure, characteristics and assets. A score of 44.6 tells us that

an average household in the sample has about 39% likelihood of being below the national

poverty line.

7
To disaggregate food groups in this way, we decrease the dimensionality of the covariate vector. In particular, the
model has difficulty converging when the dummy for coastal area is included. As will be shown later, this variable
has significant effects on the consumption of three food groups (roots, tubers, and legumes; fruits and vegetables;
and proteins) which are picked up by climatic shocks (that differ between the coastal and Selva/forest areas) and
different levels of education in the coastal and Selva/forest areas; yet it does not influence other alternatives or
coefficients. Thus we feel confident that the insights we gain by disaggregating the sugars, oils, and coffee group
will not change if the coastal variable is included. Disaggregating roots and tubers from legumes has no other issue
than the change in the effect of the previously mentioned correlated variables (education and severe flood pick up
the effect on protein consumption, flood alone picks up the effect on fruits and vegetables, and severe drought picks
up the effect on the roots group).
 Volume 16, Chapter 7, page 16

[Take in Table 7.2 Here]

To capture the use of women’s time, we include two variables: a dummy that indicates

whether a woman is the head of the household and the number of children that are between zero

and sixteen years of age. Being the head of the household highly correlates with being the

manager of the rice plots. Thus, female-headed household status may negatively influence

dietary diversity, as rice in parcels is not generally intercropped with other self-consumption

plants and women dedicated to manage rice plots may not have the time to grow home gardens.

The number of children in the household may also affect women’s ability to care for nutrition, as

more children mean greater time constraints. Only 8% of the households are headed by women

and, on average, they have one child between zero and sixteen years old.

Food preferences may be highly influenced by the education level of the household head

and his or her age. The literature suggests that higher education should be related to higher

dietary diversity. To the contrary, older household heads may face constraints in their diets due

to health issues that limit their intake of some food groups, ultimately reducing their HDDI. The

average age of a household head is 57 and the average education is between 6 and 7 years.

Two location variables capture whether a household is in the coastal area, as opposed to

the forest region (Selva), and whether a household is in one of the Northern departments that

border Ecuador. On the one hand, coastal location, as discussed earlier, may be linked to higher

availability of fruits and livestock in the market due to important agroexport industries and

commercial livestock farms. Northern regions, on the other hand, have better access to imports

from Ecuador and Colombia, with lower prices of multiple food products. These products,
 Volume 16, Chapter 7, page 17

though, are part of a more extended value chain than are the domestic products which are related

to higher processing with increased contents of sweeteners, salt and fats (Gomez et al., 2013).

Finally, we include measures of risk and instability in food consumption, such as whether

the household has faced a severe drought or a severe flood in the past five years, and the number

of months for which it lacked adequate food quantities during the past year. These variables may

negatively affect dietary diversity. In terms of weather shocks, 12 % of households have faced a

severe flood in the past five years and 22% have faced a severe drought, causing some food

scarcity.

Results

Results from the Tobit regression are presented in the first column of Table 7.3. As expected,

very few variables are statistically significant since aggregation likely hides some of the

explanatory power. However, it is worth mentioning that education of the household head is

positively correlated with household dietary diversity; for every extra year of schooling, the

household dietary diversity score increases by approximately 0.13 points. Also, those households

in the North have about 1.7 more points in the index (i.e., they eat, on average, 1.7 more food

groups) than those in the South. Furthermore, those households experiencing a severe drought in

the last five years have a 0.5 lower index than those who have not experienced a severe drought.

Next, we disaggregate the index to better understand the factors influencing the

consumption of different food groups and the complementarities among them. The results of the

first multivariate regression, using five food groups, are included in Table 7.3. Columns 3

through 7 show the simultaneously estimated coefficients of the explanatory variables on the

consumption of roots, tubers, legumes, and beans (column 3); fruits and vegetables (column 4);
 Volume 16, Chapter 7, page 18

meats, eggs, and fish (column 5); dairy products (column 6); and oils, fats, sugars, honey, coffee,

tea, and soda (column 7).

We find that wealthier households are more likely to consume fruits and vegetables, as

well as dairy products. Households with more educated heads are more likely to consume dairy

products. Female-headed households are less likely than male-headed and couple-headed

households to consume fruits and vegetables. Households with more children are more likely to

consume oils, fats, sugars, honey, coffee, tea, and soda. The “children” effect could be driven by

two distinct possibilities. On the one hand, parents with many children may have less time to

focus on nutrition and lower budgets to purchase nutritious but expensive foods such as meats

and vegetables. On the other hand, this effect could be driven by children’s own preferences. If

this is the case, older children may have a greater ability to steer a household’s consumption

basket. To distinguish between these, we run the regression including children of different ages:

zero to twelve years old, zero to fourteen years old, and zero to sixteen years old.8 We find when

older children are included, the higher the magnitude and the significance of the effect.

Furthermore, if only children from zero to six years old are included, the coefficient on this

group of foods is insignificant, yet the positive coefficient on dairy products becomes significant,

as well as a negative coefficient explaining the consumption of foods based on oils and fats.

Examining our geographic determinants, we find that households in the North, close to

the trade border, consume more foods from the following groups: fruits and vegetables, and oils,

fats, sugars, honey, coffee, tea, and soda. These households probably benefit from Ecuadorian

exports of fruits and vegetables, yet their preferences are also influenced by the market in terms

of consuming more processed foods with higher fat, salt, and sugar additives. Households

located in the coastal areas are much more likely than those in the Selva/forest areas to consume
8
Results are available upon request.
 Volume 16, Chapter 7, page 19

protein from meats, eggs, and/or fish, as well as fruits and vegetables. However, they are less

likely to consume roots, tubers, legumes, and beans. As mentioned earlier, the coastal area has

better access to markets than the forest area; it also exports a number of fruits and vegetables and

has large commercial livestock farms. While in the forest areas, roots and tubers are traditional

crops (cassava in lower altitudes and potatoes in higher altitudes) that are often intercropped with

beans and legumes, the more advanced markets in the coastal areas have provided for a higher

consumption of crops such as corn and wheat, in addition to rice.

Examining the effects of shocks on food consumption shows that having faced a severe

drought in the past five years has the most negative effects on current diet diversity. It

significantly reduces the likelihood of consuming protein, micronutrients, and starches (roots,

tubers, legumes, and beans; meats, eggs, and fish; and milk and dairy products). Having faced a

severe flood in the past five years does not show a significant effect on food consumption. It may

be though that the effect is picked up by the regional coastal dummy, since most floods occur in

the forest area. Finally, having faced additional months of food scarcity in the past year

significantly reduces the likelihood of consuming proteins in the form of meats, eggs, and fish.

Disaggregating the food options further9 (separating roots and tubers from beans and

legumes, as well as oils and fats from sugars and from the soda, coffee, and tea group), we find

that in addition to the effects found before, wealth significantly increases the consumption of

legumes and beans (but not of tubers and roots) and decreases the consumption of sugars and

honey. Households headed by women consume less legumes and beans, in addition to the

previously found lower consumption of fruits and vegetables. Households with more children eat

significantly less food based on oils and fats, yet they consume more foods in the soda, coffee,

and tea group. The age of the household head is found to significantly decrease the consumption
9
Results available upon request.
 Volume 16, Chapter 7, page 20

of oils and fats; sugars; and soda, coffee, and tea, but slightly increase the consumption of dairy

products. More educated households are shown to not only consume more protein, but also

consume more fats and oils. Shocks like drought and food scarcity decrease the consumption of

nutritious foods and increase the consumption of sugars.

Last, we examine whether significant complementarity or substitutability exists among

food groups. Table 7.4 presents the estimated correlation parameters (off-diagonal elements of

the variance covariance matrix). We find the highest and most significant complementarity is

between sugars and fats, as well as between sugars and the soda, coffee, and tea group. The

consumption of the latter food group is also highly correlated with the consumption of fat/oil

based foods. The consumption of sugars is to a lesser degree positively correlated with

consumption of fruits and vegetables and has a weak substitutionary relationship with

legumes/beans. Interestingly, milk and dairy products are strong substitutes of soda, coffee, and

tea. The consumption of dairy is also a significant complement of the consumption of legumes,

roots, tubers, other vegetables and fruits. Complementary relationships of lesser degrees exist

between legumes and roots and tubers, as well as with meats relative to fruits and vegetables.

[Take in Table 7.4 Here]

Conclusions

Micronutrient deficiencies and obesity are growing phenomena that have been linked to the

expansion of food value chains in developed and developing countries. These can cause chronic

diseases that are costly to society and individuals. We find that the strongest complementarities

among food groups exist among obesogenic foods, such as sugars, fat/oil based foods, and soda.

Those foods are highly consumed by households with children and adolescents, as well as by
 Volume 16, Chapter 7, page 21

households located in the north, near the border with Ecuador, where imported processed foods

are more accessible. This result is likely related to the global phenomenon of food adaptation,

where new foods available in the market and their appeal through all types of advertisement

influence eating behaviors and cause increases in consumption of processed and standardized

store-bought foods, which are typically rich in sugar and fats.

These results suggest a negative influence of trade on household nutrition and food

security. However, they also indicate that rice farmers living in the northern departments that

border Ecuador are more likely to eat fruits and vegetables. Thus, policy and program

recommendations should focus on nutritional information (such as labeling requirements) and

education so that consumers can make informed decisions. Further research could explore the

implications of taxing obesogenic foods or subsidies for healthy foods.

We also find that having faced a severe drought in the past five years or food scarcity in

the last twelve months are factors associated with decreased consumption of nutritious foods,

such as legumes, meats, and dairy. Simultaneous increases in sugar consumption are evidenced.

This is linked to increases in micronutrient and macronutrient deficiencies, as well as chronic

diseases. Thus, these shocks may set households on a path of ill health (chronic disease), in

addition to their economic and productivity losses. To prevent households from embarking on

such pathways, development and disaster practitioners might focus on how to make healthy

foods available and appealing in times of crises.

In terms of micronutrient and fiber deficiencies, having women as heads of households

that are highly engaged in rice production seems to be a risk factor for having households with

more micronutrient deficiencies, since these households consume significantly less fruits,

vegetables, legumes, and beans. This is likely related to the fact that rice production is a
 Volume 16, Chapter 7, page 22

monoculture and women who spend time producing rice have less time to grow household

gardens and otherwise obtain and prepare such foods. Facing drought may also increase this risk

of deficiencies. Furthermore, food scarce households are less likely to consume protein-rich

foods such as meats and dairy products. A longer time-series would be beneficial to understand

the latter result in more detail. A dynamic relationship may exist where past food scarcity may

have resulted in the liquidation of livestock assets, which are now unavailable for household

consumption.

Finally, small rice producer households in the Peruvian coastal area are definitely doing

better nutritionally, as well as households with relatively higher wealth. Such analysis and

discussion shows the nuanced results and recommendations that can be obtained by

disaggregating the household dietary diversity index.

References

Barrett, C. (2010). Measuring food insecurity. Science, 327, 825-828.

Behrman, J., & Deolalikar, A. (1987). Will developing country nutrition improve with income?

A case study for rural South India. Journal of Political Economy, 95(3), 492-507

Behrman, J., & Wolfe, B. (1984). More evidence on nutrition demand: Income overrated and

women's schooling underemphasized. Journal of Development Economics, 14, 105-128.

Chung, K., Haddad, L., Ramakrishna, J., & Riely, F. (1997). Identifying the food insecure: The

application of mixed-method approaches in India. Washington, DC: International Food

Policy Research Institute (IFPRI).

Coates, J. (2013). Build it back better: Deconstructing food security for improved measurement

and action. Global Food Security, 2, 188-194.

 Volume 16, Chapter 7, page 23

Davids, D. (2004). Changes in USDA food composition data for 43 garden crops, 1950 to 1999.

Journal of the American College of Nutrition, 23(6), 669-682.

Escobal, J., & Agreda, V. (1994). La comercialización de arroz en la Costa Norte del Perú: El

caso de los productores de Ferreñafe. In Escobal, J. (Ed.), Comercializacion agricola en

el Peru. Lima, Peru: GRADE, AID.

Food and Agriculture Organization of the United Nations [FAO]. (2001). The State of Food

Insecurity in the World 2001. Rome, Italy: FAO.

Food and Agriculture Organization of the United Nations [FAO]. (2002, June). Measurement

and assessment of food deprivation and under-nutrition. Proceedings of the International

Scientific Symposium. Rome, Italy: FAO.

Foote, J., Murphy, S., Wilkens, L., Basiotis, P., & Carlson, A. (2004). Dietary variety increases

the probability of nutrient adequacy among adults. Journal of Nutrition, 134, 1779-1785.

Greer,J., & Thorbecke, E. (1984). Pattern of food consumption and poverty in Kenya and effects

of food prices. Mimeographed. Cornell University, Ithaca, NY.

Gomez, M., Barrett, C., Raney, T., Pinstrup-Anderson, P., Meerman, J., Croppenstedt, A.,

Carisma, B., & Thompson, B. (2013). Post-green revolution food systems and the triple

burden of malnutrition. Food Policy, 42, 129-138.

Hatloy, A., Halland, J., Diarra, M., & Oshaug, A. (2000). Food variety, socioeconomic status and

nutritional in Koutiala (Mali). Public Health Nutrition, 3, 57-65.

Hawkes, C. (2006). Uneven dietary development: linking the policies and processes of

globalization with diet-related chronic diseases. Global Health, 2, 4.

 Volume 16, Chapter 7, page 24

Hoddinott, J. (2012). Agriculture, health, and nutrition: toward conceptualizing the linkages. In

Fan, S., & Pandya-Lorch, R. (Eds.), Reshaping agriculture for nutrition and health

(Chapter 2). Washington, DC: International Food Policy Research Institute.

Kadiyala, S., Harris, J., Headey, D., Yosef, S., & Gillespie S. (2014). Agriculture and nutrition in

India: mapping evidence to pathways. Annals of the New York Academy of Sciences,

1331, 43-56.

Kennedy G.L., Pedro M.R., Seghieri, C., Nantel, G., & Brouwer, I. (2007). Dietary diversity

score is a useful indicator of micronutrient intake in non-breast-feeding Filipino children.

The Journal of Nutrition, 137, 472–477.

Lin, C.J., Jensen, K.L., & Yen, S.T. (2005). Awareness of foodborne pathogens among US

consumers. Food Quality and Preference, 16(5), 401-412.

Maxwell, D. (1999). Alternative food-security indicators: revisiting the frequency and severity of

coping strategies. Food Policy, 24(4), 411-429.

Mendez, M., & Popkin, B. (2004). Globalization, urbanization and nutritional change in the

developing world. In Globalization of food systems in developing countries: Impact on

food security and nutrition (Food and Nutrition Paper 83). Rome: FAO.

Popkin, B.M., & Nielsen, S.J. (2003). The sweetening of the world's diet. Obesity Research, 11,

1325-1332.

Radimer, K.L., Olson, C.M., & Campbell, C.C. (1990). Development of indicators to assess

hunger. Journal of Nutrition, 120, 1544-1548.

Rathnayake, K.M., Madushani, P.A.E., & Silva, K.D.R.R. (2012). Use of dietary diversity score

as a proxy indicator of nutrient adequacy of rural elderly people in Sri Lanka. BMC

Research Notes, 5, 469.

 Volume 16, Chapter 7, page 25

Ruel, M. (2003). Is dietary diversity an indicator of food security or dietary quality? A review of

measurement issues and research needs. FCND Discussion Paper, International Food

Policy Research Institute, Washington, DC. Retrieved from

http://www.ncbi.nlm.nih.gov/pubmed/12891828.

Shah, C.H. 1983. Food preference, poverty, and the nutrition gap. Economic Development and

Cultural Change, 32: 121-148.

Schreiner, M. (2012). Progress out of Poverty Index (PPI): A simple poverty measurement

scorecard for Peru. A Grameen Foundation Product. Retrieved from

http://www.progressoutofpoverty.org/country/peru.

Steyn, N.P., Nel, J.H., Nantel, G., Kennedy, G. and Labadarios, D. (2006). Food variety and

dietary diversity scores in children: are they good indicators of dietary adequacy? Public

Health Nutrition, 9(5), 644-650.

La Superintendencia Nacional de Aduanas y de Administración Tributaria (SUNAT). (2011).

Estadisticas de Comercio Exterior. Retrieved from http://www.sunat.gob.pe/estad-

comExt/modelo_web/web_estadistica.htm. Accessed April 24, 2015.

Vera, R. (2006). Country Pasture/Forage Resource Profiles: Peru. Retrieved from

http://www.fao.org/ag/AGP/AGPC/doc/counprof/Peru/Peru.htm April 24, 2015.

Webb, P., Coates, J., Frongilio, E.A., Rogers, B.L., Swindale, A., & Bilinsky, P. (2006).

Measuring household food security: Why it’s so important yet so difficult to do. The

Journal of Nutrition, 136(5), 1404S-1408S.

World Bank. (2007). From agriculture to nutrition: Pathways, synergies, and outcomes. Report

No. 40196-GLB. The World Bank, Washington, DC.

 Volume 16, Chapter 7, page 26

Table 7.1. Departments and agro-ecological zones

Department Percentage Households Agro-ecological Zone Location

Amazonas 10% Forest North

Cajamarca 18% Forest North

La Libertad 15% Coast South

Lambayaque 18% Coast South

Piura 19% Coast North

San Martin 20% Forest South

 Volume 16, Chapter 7, page 27

Table 7.2. Descriptive statistics

Variable Description Mean Std. Dev. Min Max

HDDI Household dietary diversity Index 8.22 2.11 2 12

PPI Progress out of poverty index 44.60 9.49 15 78

WomanHead Female-headed household 0.08 0.28 0 1

Children 0–16 Number of children, 0 to 16 years old 0.93 1.20 0 6

AgeHead Age of household head 57 13 15 91

EduHead Years of schooling of household head 6.51 3.64 0 18

North Located in Northa 0.47 0.50 0 1

Coast Located on Coast (vs Selva) 0.52 0.50 0 1

SevereFlood 1 if household experienced a severe 0.12 0.32 0 1

flood in the last five years.

SevereDrought 1 if household experienced a severe 0.22 0.41 0 1

drought in last five years.

Food scarcity Number of months not having enough 0.71 1.70 0 12

(No. of months) food to feed all household members in

past year.

Notes: a) See Table 1 for departments located in North and South.

 Volume 16, Chapter 7, page 28

Table 7.3. Tobit regression of the HDDI and multivariate probit regression with 5 food groups

Tobit MVP

HDDI (1) (2) (3) (4) (5)

PPI 0.012 –0.005 0.022* –0.011 0.017* –0.005

(0.015) (0.014) (0.012) (0.019) (0.0098) (0.015)

FemaleHead –0.193 –0.297 –0.45** –0.227 –0.264 0.313

(0.336) (0.324) (0.227) (0.337) (0.221) (0.401)

kid016 –0.013 –0.029 0.031 –0.006 0.054 0.256*

(0.123) (0.128) (0.090) (0.137) (0.079) (0.149)

AgeHead 0.006 0.002 0.0007 0.004 0.007 –0.005

(0.009) (0.009) (0.007) (0.012) (0.006) (0.009)

EduHead 0.13*** 0.002 0.031 0.069 0.07*** -0.028

(0.029) (0.029) (0.024) (0.042) (0.019) (0.030)

North 1.68*** –0.150 0.602*** 0.497* –0.0171 1.44***

(0.196) (0.196) (0.158) (0.281) (0.126) (0.306)

Coast –0.112 –0.69*** 0.59*** 2.10*** –0.0822 –0.399

(0.216) (0.233) (0.171) (0.604) (0.139) (0.258)

SevereFlood 0.017 3.720 -0.034 0.117 0.144 0.257

(0.318) (145.5) (0.220) (0.293) (0.200) (0.396)

SevereDrought –0.532** –0.364* –0.188 –0.48* –0.38*** 0.357

(0.224) (0.208) (0.168) (0.266) (0.147) (0.256)

Food scarcity –0.046 –0.024 –0.046 –0.21*** –0.045 0.041

(0.056) (0.052) (0.040) (0.061) (0.037) (0.050)

Log Likelihood –979.67012 –768.05541 PseudoR2 0.0559

Standard errors in parentheses; (1)=Roots, tubers, legumes, (2)=Fruits and vegetables, (3)=Meats, eggs and fish,

(4)=Milk and dairy, (5)= Oils, fats, sugar, honey, coffee, tea, and soda

*** p<0.01, ** p<0.05, * p<0.1

 Volume 16, Chapter 7, page 29

Table 7.4. Estimated Correlation Coefficients of Food Options

Roots & Fruits & Meats, Legumes, Oils & Sugar & Soda,

tubers vegetables egg, fish beans Dairy fats honey coffee, tea

Roots/tubers

Fruits/vegetables. 0.14

Meats/egg/fish –0.08 0.28**

Legumes/beans 0.38*** 0.1 –0.02

Dairy 0.30*** 0.22** 0.02 0.26***

Oils/fats –0.09 0.19** 0.07 –0.05 –0.05

Sugar/honey –0.05 0.50*** 0.03 –0.17* –0.04 1.23***

Soda/coffee/tea 0.15 0.22** –0.07 0.14 –0.34*** 0.76*** 0.97***

*** p<0.01, ** p<0.05, * p<0.1

 Volume 16, Chapter 7, page 30

Figure 7.1. Distribution of households’ demand for food by food group

100%

90%

80%

70%

60%

50%

40%

30%

20%

10%

0%

View publication stats