Food Policy 85 (2019) 28–39

Contents lists available at ScienceDirect

Food Policy
journal homepage: www.elsevier.com/locate/foodpol

Impact of contract farming on land productivity and income of maize and T

potato growers in Pakistan
Muhammad Fawad Khana, , Yuko Nakanob, Takashi Kurosakic
⁎

a
Graduate School of Agriculture and Life Sciences, The University of Tokyo, 1-1-1, Yayoi, Bunkyo-ku, Tokyo 113-8657, Japan
b
Faculty of Humanities and Social Sciences, University of Tsukuba, 1-1-1 Tennodai, Tsukuba, Ibaraki 305-8577, Japan
c
Institute of Economic Research, Hitotsubashi University, 2-1 Naka, Kunitachi, Tokyo 186-8603, Japan

ARTICLE INFO ABSTRACT

Keywords: While the positive income effect of contract farming is well established in the literature, the heterogeneity of
Contract farming effects for different crops has not been widely investigated. This study compares the income and productivity
Grower-processor contracts effects of large-scale contract farming under contract schemes for potatoes and maize in Pakistan. In the area of
Staple food marketing the study, maize is a common staple crop, widely available in the market, while the potatoes produced by
Pakistan
contract farmers are a special variety not sold in the market. We find that potato contracting is associated with
significantly higher income for participating farmers, while maize contracting has no association with either
income or productivity. We found no negative association between contract participation and income from other
sources. Taken together, these results show a higher total household income for potato contract farmers. We also
show unique evidence that potato contract farmers tend to employ more skilled workers under a salary system
for the management of farms, which implies that contract farming may have positive spillover effects in the
skilled labor market. Neither scheme, however, is associated with larger employment of agricultural labor for
simple tasks.

1. Introduction Due to the importance of contract farming, there is ample literature

examining its impact on the productivity and welfare of small-scale
Food demand has changed considerably over the past six decades, farmers in developing countries (see Bellemare and Bloem, 2018;
from a focus on undifferentiated commodities towards high quality and Otsuka et al., 2016; Oya, 2012; Reardon et al., 2009; Ton et al., 2018;
diversified products mainly due to the rise in levels of income, trade Wang et al., 2014b for recent review). Many studies observe a positive
liberalization, and urbanization (Bellemare, 2012; Reardon and effect for contract farming on the income and productivity of partici-
Timmer, 2007). In response to this change in demand, supermarket pating farmers (Bellemare, 2012; Ching-Cheng et al., 2006; Dileep
chains and food processors now coordinate with food growers to im- et al., 2002; Key and McBride, 2003; Miyata et al., 2009; Rao and Qaim,
plement new standards and requirements, primarily through the in- 2011).1
stitution of contracts (Reardon and Timmer, 2007). A contract is an This study makes a three-fold contribution to the literature: First,
agreement between a farmer and a food processing/marketing firm we examine the heterogeneous impact of contract farming on the cul-
under which the farmer commits to produce an agricultural product for tivation of different crops. It has been pointed out that contract farming
the company on his/her land in a specific quantity and subject to other tends to be more effective in generating income effects for commodities
predetermined conditions on output prices, production process, etc. that have non-competitive markets (making side-selling less likely) and
Contract farming is considered an institution that aims to address var- whose production requires technical know-how (e.g., fresh fruits and
ious market imperfections such as high transaction costs, information vegetables, poultry and beef production for export markets, etc.) as
asymmetry, and missing markets (Abebe et al., 2013; Bellemare et al., compared to commodities with competitive markets and a low potential
2018; Grosh, 1994; Key and Runsten, 1999). for quality improvements (e.g., staple food crops) (Otsuka et al., 2016;

Corresponding author.
⁎

E-mail addresses: [email protected] (M.F. Khan), [email protected] (Y. Nakano), [email protected] (T. Kurosaki).
1
This may not be surprising given that farmers will participate only if they realize gain from their association with the company. Moreover, it is impossible for the
farmers to participate in the contract schemes that have previously failed (i.e., survivorship bias in the literature) (Bellemare, 2012; Bellmare & Boloem, 2018; Ton
et al., 2018).

https://doi.org/10.1016/j.foodpol.2019.04.004
Received 7 November 2018; Received in revised form 10 April 2019; Accepted 10 April 2019
Available online 19 April 2019
0306-9192/ © 2019 Elsevier Ltd. All rights reserved.
M.F. Khan, et al. Food Policy 85 (2019) 28–39

Simmons et al., 2005). As Ragasa et al. (2018) correctly notes, however, We also analyze the impact of contract farming on the employment
this heterogeneity is not well studied. This is partly because many of two types of workers: skilled workers employed under a salary
studies focus on high-value horticulture and industrial crops (for in- system3 for the management of farms (we call these “salaried workers”)
stance, Dolan and Humphrey, 2000; Gatto et al., 2017; Maertens and and unskilled workers temporarily hired for manual work (we call these
Swinnen, 2009; Miyata et al., 2009; Narayanan, 2014; Ochieng et al., “casual labor”). By doing so, we examine the spillover effect of large-
2017; Reardon and Berdegué, 2002; Wang et al., 2014a), although an scale contract farming towards the local labor market, which can po-
increasing number of studies deal with staple crops (Maertens and tentially increase the employment.
Velde, 2017; Ragasa et al., 2018; Soullier and Moustier, 2018).2 Most of Our results indicate that potato contract farming is associated with a
the studies focus on single contract farming schemes (for instance, large increase in income from the contracted crop, whereas maize
Bolwig et al., 2009; Sethboonsarng et al., 2008; Warning and Key, contract farming is not. We do not observe any significant increase in
2002) due to which heterogeneity could not be studied. Bellemare physical yield per acre for either of the schemes. Since there is no ne-
(2012), Bellemare and Novak (2017), Miyata et al. (2009), and Minten gative association between the potato contract and other sources of
et al. (2007) include multiple contract farming schemes in their data income, total household income is higher for the potato contract
but do not allow for heterogeneity of effects. The only exceptions are farmers than non-contract farmers. The main factors behind the in-
Narayanan (2014) and Simmons et al. (2005), who differentiate the creased income associated with the potato contract farming scheme are
impact of contract farming on different crops and observe significant the higher price and the increased potato crop share of the cultivated
differences across commodities. More evidence of the heterogeneous area. There is no significant difference, however, in total income be-
effects of contract farming, especially in the case of staple crops versus tween maize contract farmers and non-contract farmers. We also see
high value crops, would therefore add to the literature on contract evidence of an increase in demand for salaried workers under the po-
farming. tato contract farming scheme but observe no difference in the cost of
Secondly, only a small number of prior studies examine the impact casual labor for either scheme.
of contract farming on total household income and income from other The remainder of the paper is organized as follows: In Section 2, we
sources (Otsuka et al., 2016). One exception is Bellemare (2018), who discuss data and the contract farming schemes under study. Section 3
finds a significantly negative effect for contract farming on income from explains the econometric methodology used for estimation. Results and
non-agricultural sources. Since production under contract farming may associated discussion are presented in Section 4. Section 5 concludes
be more labor intensive, there is a possibility that the income from the paper and presents policy implications.
other sources are sacrificed. If this is the case, the impact of contract
farming on total household income may not be large compared to that
on income from contracted crops. 2. Data and study site
Thirdly, the impact of contract farming on the labor market is not
extensively examined in the literature. The existing literature focuses 2.1. Data
mainly on the direct poverty implications of contract farming by
looking at the participation of small-scale farmers in the contract We focus on two contract schemes, separately operated by Rafhan
scheme (e.g., Miyata et al., 2009; Glover, 1984; Bolwig et al., 2009; Maize Products Ltd. for maize production and PepsiCo. International Ltd.
Winters et al., 2005). Contract farming, however, could have spillover for potato production. We conducted a field survey in
effects in labor market. Production of contract crops that require the September–October 2016 in tehsil Depalpur in the Okara district of
adoption of labor-intensive technologies would increase the labor de- Pakistan, where both maize and potato contract farming schemes are in
mand of the contract households (Bellemare, 2018; Simmons et al., place.4 Potatoes and maize are the two most popular crops in the re-
2005). If this happens on a large scale, contract framing can increase gion. Recently, food processing and marketing firms have begun to
labor demand at the local level, as does other large-scale agricultural emerge in Pakistan’s food sector due to a rapidly increasing middle
investments (for instance, Herrmann and Grote, 2015; Nolte and class, urbanization, and market liberalization. For instance, Engro
Ostermeier, 2017). Foods (rice), PepsiCo. (potatoes), Rafhan (maize), Pakistan Tobacco
In order to fill these gaps in the literature, we compare large-scale Company (tobacco), K&N (meat and poultry), and Nestle (milk) are
commercial farmers (with an average landholding of roughly 68 acres) actively engaged in contract farming in the country (Chodavarapu
who participate in either of two contract farming schemes. We analyze et al., 2017). The Okara district is home to a number of such contract
the impact of the two contract farming schemes on price, yield, income farming schemes, mainly because of its highly productive lands with
from the contracted crop, and total household income as well as income abundant irrigation, a strong presence of large and progressive farmers,
from other sources. For this purpose, we estimate the bias-adjusted and proximity to the industrial centers of Lahore and Faisalabad.
treatment effect proposed by Oster (2017) by using the unique data set The study’s survey sample includes 104 contract farmers and 96
collected by the authors. In the area studied, maize is a highly com- non-contract farmers. Among the contract farmers, 83 are maize con-
mercial but not high-value crop that is available competitively in the tract farmers, 16 are potato contract farmers and five are both maize
spot market. In contrast, the potatoes grown by farmers under the and potato contract farmers. Contract farmers were randomly selected
contract are a special variety that is specifically suited to chip pro- from the lists of contract farmers provided by the contracting compa-
duction—a variety that is not available in the spot market and one that nies. Based on our discussions with company officials, we found that
requires some technical knowhow to produce. This contrast in contracts companies give preference to farmers with large land holdings. To en-
enables us to effectively compare the impact of contract farming on a sure that the control group would be comparable to the treatment
staple crop and a high-value crop. group, we first made a list of non-contract farmers with landholdings

3
In order to manage the farm activities, the farmers employ workers on a full-
2
There are several studies on the seed production of staple foods, for which time basis throughout the year and pay them a fixed amount of monthly salary.
the quality requirements are stricter than those for foods for consumption (e.g., Responsibilities of these workers include security of the farm, arrangement of
Abebe et al., 2013; Mishra et al., 2016; Simmons et al., 2005; Winters et al., inputs from the market, processing and sale of product, water management, etc.
2005). Leung et al. (2008) and Mishra et al. (2018) also provide examples of Labor-intensive farm work such as seed sowing, fertilizer or pesticide applica-
contract farming for high-quality organic rice in Laos and India, respectively. In tion, or weeding are done by casual labor.
4
these cases, staple crops under contract can be considered as high-value crops A tehsil is a unit of administration in the Indian subcontinent that consists of
rather than normal staple foods. one or more towns and villages around the town(s).

29
M.F. Khan, et al. Food Policy 85 (2019) 28–39

similar to contract farmers through local informants and then selected 3. Methodology
our sample of non-contract farmers randomly from the list. The farmers
in our samples of contract and non-contract farmers did not participate Before we analyze the impact of contract farming, we estimate a
in any other contract farming schemes. probit model to investigate if there is any constraint for the farmers to
participate in the contract program. In this model, a dummy variable
that takes a value of 1 if the farmer has a contract and 0 otherwise
2.2. Contract farming in the study site
serves as the dependent variable. The independent variables include the
years of schooling of the household head, his or her age, the number of
2.2.1. The maize contract farming scheme
adult family members, land size, land value per acre (a proxy for land
As mentioned earlier, the maize contract farming scheme under
quality), value of non-farm assets, and the number of livestock owned.
study is operated by the largest maize processor in the country—Rafhan
We then analyze the association between participation in contract
Maize Products Ltd. Rafhan is the largest wet-miller in Pakistan and a
farming and various outcome variables such as crop income, total
dominant buyer of maize, consuming about 10 percent of the total
household income, income from other sources, and the cost and number
maize produced in the country. The maize is used for industrial pro-
of hired laborers for both salaried and casual labor. Yi represents the
cessing rather than domestic consumption or direct exports, and the
outcome variables for household i; Ci is a dummy variable that equals 1
quality of maize is roughly similar to the maize traded in the open
for contract farmers and 0 for independent growers. Since we have two
market. This particular contract farming program began in the early
crops under contract, we define Ci in three different ways, as we will
1970s (CIMMYT, 1989). Presently, the company procures about 40
discuss later. Xi is a vector of the household and farm level character-
percent of its maize requirement through contract farming; the rest
istics mentioned above. The simple OLS regression can be estimated as
comes from the open market. According to the company’s officials, the
follows;
rationale for the contract farming of maize, for which the markets are
considerably developed, is to have an assured supply of the raw ma- Yi = + Ci + Xi + i
terial and to reduce cost by eliminating intermediaries.
Before cultivation begins, a field agronomist employed by the Obviously, however, the estimated coefficient β would be biased as
company finds farmers who are interested in contracting their maize a measure of contract impact because farmers self-select into the con-
crop. A written contract between the farmer and the company is then tract. Accordingly, we considered important controls based on findings
signed. The contract specifies the number of acres to be cultivated for from our interview with company officials. These include land holdings,
the company and sets the price for the farmer’s output as the market experience in farming, and the trustworthiness of the farmer. Among
price at the time of delivery. The company provides extension services these, the trustworthiness of the farmer is an attribute that we could not
(if needed by the farmer), transportation support, and packaging ma- control. Moreover, it is likely that typically unobservable characteristics
terial, but it does not provide seed for the main contracted variety (such as entrepreneurial ability, risk aversion, social networks, etc.)
(Yellow Maize), fertilizers, or pesticides, as these are available from the may bias our estimates.
market. The contract company makes full and immediate payments To address the issue of bias, we adopt the methodology developed
through an electronic payment system whereas, in the spot market, by Oster (2017), which is the extension of the idea of Altonji et al.
farmers sometimes receive split or delayed payments. As we will discuss (2005). According to Oster (2017), if the assumption of proportional
later, there seems to be no difference in the yield, price or income be- selection holds (i.e., that the relationship between the outcome and the
tween maize contract farmers and non-contract farmers. Nonetheless, observed control variables informs the relationship between outcome
this efficient payment system appears to motivate farmers to join the and unobservables), then changes in the magnitudes of the coefficients
maize contract. and value of R2 can tell us about the size of omitted variable bias. There
are two ways to conduct this test. One is to calculate the value of δ,
which is the degree of selection on the unobservables relative to the
2.2.2. The potato contract farming scheme observables that would be needed to drive our estimated coefficients to
PepsiCo. International Ltd. has a state-of-the-art potato chip plant in zero. δ is mathematically defined as:
Lahore that caters to markets in Pakistan and Afghanistan. Due to the
special quality requirements for chip-making and the need for a stable ( )(R R)
supply of potatoes, the company relies completely on contract farming ( )(Rmax R)
rather than direct procurement from the market. Depalpur tehsil has the
largest number of the company’s contract farmers, accounting for ap- where is the coefficient of the treatment dummy and R is the R2
proximately 55 percent of its potato supply. The contract farmers are value in the simple regression of outcome on treatment; and R cor-
carefully selected based on their land holdings, knowledge of potato respond to those in the regression with all observable controls included.
cultivation, and market reputation. is the targeted value of the coefficient (e.g., zero). Rmax corresponds
The company processes a special variety of potato known as Lady to R2 in a hypothetical regression containing all observable and un-
Rosetta, which is specifically suited to chip production, and provides observable controls. Oster (2017) recommends using 1.3R as the value
appropriate seed to the contract farmers. The company also provides for Rmax; a δ value of 1 is considered an appropriate cutoff. The second
regular extension services by sending extension specialists at least once approach to conducting this robustness test is to estimate coefficient
a week but does not offer fertilizer or pesticide. The pre-cultivation bounds. One bound is , the value of β when δ = 0. The other bound is
contract specifies the number of acres to be cultivated by the farmer for , the value of β when δ = 1 and R2 = Rmax. If the estimated coefficient
the company and fixes the price of the crop, which is usually much bounds interval does not include zero, the estimates are robust to un-
higher than the market price of potatoes because of the difference in observables. We use both approaches to test for the robustness of our
variety. The company performs a strict quality check at the farm gate estimates. While this test might seem like an attempt to establish
before granting permission for transport to its storage facility. To ensure causality, we interpret our results as associative rather than causal, as
compliance, the company has adopted a punishment strategy such that the underlying assumption of proportional selection cannot be tested.
farmers are blacklisted permanently in case of contract breach. Thus, Since we are testing many outcome variables, standard errors must
we have to note that the differences between maize and potato con- be adjusted to account for the fact that the outcome is a part of a family
tracts are not only the crop itself but also the provision of extension of hypotheses (Duflo et al., 2007). Here, we follow the step-down re-
services and seed, as well as the price setting, which partly arise from sampling procedure of Westfall and Young (1993). That is, we classify
the nature of the crop. the variables into five families: (1) crop performance (price, yield and

30
M.F. Khan, et al. Food Policy 85 (2019) 28–39

Table 1 Table 2
Household and farm characteristics. Sources of household income and cost structure.
Maize contract Potato contract Non-contract Maize contract Potato contract Non-contract
farmers farmers farmers farmers farmers farmers

=1 if female headed 0.00 0.00 0.00 Farm Income (USD) 30,777*** 67,937*** 11,663
household [0.00] [0.00] [0.00] From Maize (USD) 18,394*** 25,810*** 9,305
Number of adult household 3.30 2.58** 3.61 From Potato (USD) 8,447*** 40,205*** −1,552
members [1.75] [1.20] [2.13] From Other Crops (USD) 3,936 1,922 3,910
Number of total household 5.07 4.37 4.92 Livestock Income (USD) 7,824 8,793 7,575
members [2.13] [1.88] [2.63] Other Income (USD) 2,925** 927 590
Age of household head 45.64 45.07 44.72 Total Household Income 41,526*** 77,656*** 19,828
[9.93] [5.12] [11.20] (USD)
Years of schooling of 11.15 11.49 10.88 Per Capita Income (USD) 11,993** 21,562*** 7,140
household head [2.72] [1.99] [3.67] Per Acre Farm Income 209** 469*** 120
Land owned (Acres) 84.06** 76.90* 51.59 (USD/Acre)
[148.11] [71.38] [62.84] Percentage of operated land 73.0 82.3** 68.4
Land operated (Acres) 138.16*** 153.19** 93.96 allocated to Potato1
*
[118.41] [106.72] [106.11] Percentage of operated land 66.7 75.4 67.2
Land value (USD/Acre) 20,231*** 19,809** 16,854 allocated to Maize2
[10,191] [6,913] [6,012] Number of Salaried Workers 6.2 8.4 4.8
Assets - Value of non-farm 69,232** 51,202 47,115
Cost structure (per acre) in maize crop
assets (USD) [81,056] [60,828] [74,840]
Price (USD per 100 kg) 20.8 21.1 20.5
Assets - Number of livestock 17.88 8.95 15.14
Yield (tons) 2.7* 2.6** 2.8
owned [34.37] [14.88] [45.01]
Gross Output Value (USD) 573.0 522.0 556.2
Length of association with 14.58 5.10 –
Paid Cost of Machinery 45.2 38.8*** 47.3
company (years) [11.34] [3.94]
(USD)
Observations 83 21 96
Paid Cost of Inputs (USD) 252 218.7*** 244.8
Paid Cost of Casual Labor 87.0 85.4 86.0
In this table, we compare the average of a variable among the maize (or potato) (USD)
contract farmers and that among non-contract farmers. If the difference is Paid Cost of Salaried 19.1 24.7 20.9
statistically significant, we put asterisks on the average among the maize (or Workers (USD)
potato) contract farmers. For instance, 2.58** in the Potato Contract Farmers Income (USD) 173.4 161.9 163.2
column, variable “Number of adult household members,” indicates that the
Cost structure (per acre) in potato crop
average among the potato contract farmers at 2.58 is significantly smaller than Price (USD per 100 KG) 6.7*** 16.1*** 5.6
3.61 (the average among non-contract farmers) with the statistical significance Yield (tons) 10.7*** 10.1 10.1
at the 5% level. Standard errors are reported in brackets. Gross Output Value (USD) 677.8*** 1080.4*** 563.6
The five contract farmers who contract for both maize and potatoes are in- Paid Cost of Machinery 72.1 74.4 71.9
cluded only in the potato contract farmers’ category for descriptive analysis (USD)
because of its small sample size. Paid Cost of Inputs (USD) 451.5 551.2*** 473.6
* Significant at 10% level. Paid Cost of Casual Labor 81.4 82.9 83.8
(USD)
** Significant at 5% level.
Paid Cost of Salaried 14.5 17.4 15.0
*** Significant at 1% level. Workers (USD)
Income (USD) 49.8*** 347.1*** −87.3

output value per acre), (2) costs (machinery, crop inputs, casual and For the meaning of asterisks, see the note to Table 1.
salaried labor cost), (3) farm productivity (income per acre and percent In calculating “Farm Income,” all the paid costs listed in the cost structure were
of land allocated to the crop), (4) aggregate income, (farm income, total subtracted. “Paid Cost of Casual Labor” is wages paid to workers doing casual
income, income from non-contracted crops, livestock income and other labor. “Paid Cost of Salaried Workers” is salaries paid to salaried workers.
income), and (5) labor usage (number of salaried workers employed). “Total Household Income” is the sum of “Farm Income,” “Livestock Income,”
This procedure gives us both standard errors and p-values adjusted for and “Other Income.”
1,2: the percentage of area under potato or maize cultivation (of the total land
multiple hypotheses testing.
operated) is calculated for the cropping seasons of each crop. For most of the
Moreover, to confirm the robustness of our estimates to changes in
farmers in the study, the alternative to potato or maize cultivation in each
specification and functional form, we estimate the average treatment season is leaving the land fallow.
effect on treated (ATT) of the contract farming by using the propensity * Significant at 10% level;
score matching (PSM) method (Rosenbaum and Rubin, 1983). In this ** Significant at 5% level;
case, we estimate the propensity score of entering into the contract p *** Significant at 1% level.
(Xi) by using the above mentioned probit model and estimate

ATTPSM = E[Yi Ci = 1, p(Xi)] E[Yi C i = 0, p(Xi)] results below.

The shortcoming of the PSM method is that we can control only for
4. Results
observed heterogeneity. Recognizing this, we also conducted a sensi-
tivity analysis of the robustness of the estimates to unobserved het-
4.1. Descriptive analysis
erogeneity by estimating a critical odds ratio as proposed by
Rosenbaum (2002). In appendix Tables 1–3, we show the ATT results of
Table 1 presents a summary of basic household and farm char-
our PSM estimation.5 Although we do not discuss this in detail due to
acteristics for the three groups of farmers (i.e., maize, potato, and non-
the limitation of space, the results are largely consistent with the main
contract farmers). The results of t-tests of the mean differences between
each contract farming group and non-contract farmers are also re-
5
Results of the balancing test, the average treatment effect (ATE), and the ported. All heads of household in our sample are male. Years of
average treatment effect on the untreated (ATU) for the estimation of PSM are schooling and age of household head are nearly equal for all three ca-
available upon request. tegories of farmers. Moreover, for all three groups, the average

31
M.F. Khan, et al. Food Policy 85 (2019) 28–39

Table 3 The potato contract farmers earn the highest farm and total annual
Probit model estimation results for participation in contract. income, followed by the maize contract farmers. The potato contract
(1) (2) (3) farmers earned more than twice as much farm income per acre (469 US
dollars) as the maize contract farmers (209 US dollars) and about four
Independent variables Either or both Maize contract Potato contract times as much as the non-contract farmers (120 US dollars). The potato
contracts and maize contract farmers earn total income of roughly 78 thousand
HH head education −0.003 −0.008 0.006 and 42 thousand US dollars per annum, respectively, which is sig-
(0.032) (0.033) (0.052) nificantly higher than the earnings of the non-contract farmers (20
HH head age 0.020* 0.019* 0.034* thousand US dollars). We also observe that potato contract farmers
(0.011) (0.011) (0.019) allocate a significantly higher proportion of their land to potato culti-
No. of adult members −0.136** −0.112* −0.291**
vation, which seems to be lucrative for them.
(0.057) (0.058) (0.119)
ln (land owned) 0.113 0.121 0.188 The higher income in both contract groups can be further clarified
(0.093) (0.096) (0.176) by analyzing the cost structure of the maize and potato crops, as se-
ln (non-farm asset 0.196*** 0.195** 0.086 parately shown in Table 2. The maize contract farmers receive a similar
value)
price to non-contract farmers for their maize, while the potato contract
(0.075) (0.076) (0.113)
Number of livestock −0.002 −0.001 −0.008
farmers receive a significantly higher price than non-contract farmers
(0.002) (0.002) (0.009) for their potatoes. In terms of bivariate comparison, we observe dif-
ln (per acre land 0.662** 0.603** 1.185** ferences in maize and potato yield per acre between contract and non-
value) contract farmers, some of which are statistically significant. The di-
(0.280) (0.279) (0.600)
rection of the difference, however, is opposite to the expected pattern (a
Constant −12.30*** −11.57*** −20.23**
(4.113) (4.095) (8.851) natural expectation might be that the contract farmers of that crop
obtain higher yield) and the magnitude of the difference is small. The
Observations 200 184 117
R-squared 0.080 0.075 0.153 potato contract farmers earn significantly higher income per acre from
their potato crops than non-contract farmers. For the maize contract
Standard errors are reported in parentheses. farmers, although the per-acre income from their maize crop is higher
* Significant at 10% level. than that of non-contract farmers, the difference is not statistically
** Significant at 5% level. significant. The potato contract farmers appear to employ more salaried
*** Significant at 1% level. workers than the other groups, though the difference is not statistically
significant. There are no significant differences in paid costs of casual
schooling is approximately 11 years, which is much higher than the six- labor between potato contract farmers and non-contract farmers. We
year average for farmers in the country overall (Anríquez and Valdes, did not observe any significant differences in the paid costs of salaried
2006). workers or casual labor between the maize contract farmers and non-
The average land holding of 51.6–84.1 acres for farmers in all three contract farmers.
groups is relatively high given that approximately 96 percent of agri-
cultural farms in the country are less than 25 acres (Government of 4.2. Impact of contract farming
Pakistan, 2010). We should note that the land holdings in the non-
contract farmers group are large mainly because we deliberately sam- To give insight into the factors influencing participation in contract
pled large farms that would be comparable to the contract farms in the farming, Table 3 shows the estimation results of the probit model for
treatment group. Nevertheless, the amount of land owned, as well as participation in the contract. We found that the age of the household
the amount of land operated, is significantly higher (at the 10 percent head is significantly correlated with participation in both potato and
level of significance) for both categories of contract farmers versus the maize contract farming, which may reflect the desire of companies to
non-contract farmers in the sample.6 Moreover, it can be seen that the contract with experienced and trustworthy farmers. We also found that
quality of land (as measured through the present sale value of one acre the number of adult members of the household is negatively correlated
of land) is also significantly higher in both categories of contract with contracting. In addition, the quality of the farmer’s land, which we
farmers as compared to non-contract farmers. Since this would create a measure through per-acre sale value, is also strongly linked to partici-
bias for our estimation results, we control for the quality of land in our pation in the contract program. Furthermore, non-farm assets are po-
subsequent analysis. Farmers in our sample, on average, have 15 years sitively associated with participation in a contract program, which re-
of association with Rafhan, the maize buyer; some have as many as flects the better endowment of contract farmers. We see no association
35–40 years of contract farming association with the company. Pep- between the amount of land owned or education of household head and
siCo., the potato contract company, is a comparatively new entrant. Its participation. One possible explanation for this might be related to the
average length of association with the farmers in our sample is ap- fact that, as mentioned earlier, our control sample includes only farmers
proximately five years, with a maximum of 16 years. who are comparable to the contract farmers in terms of land size, which
Table 2 presents a summary of the farm and total income of contract may obscure any association between land-size and contract partici-
and non-contract farmers. Income is defined as gross output value pation.
minus paid-out costs for machinery, chemical inputs, and hired labor, Next, we analyze the association between contract farming and
including both salaried workers and casual labor.7 t-tests are conducted various outcome variables of interest. Since we included two different
for each contract farming group and the control group. Note that potato contract farming schemes in our study, we performed this analysis in
contract farmers, maize contract farmers, and non-contract farmers all three different ways. First, we pool all the contract farmers into one
grow both potatoes and maize. treatment group. Then, in order to investigate the heterogeneity of the
effects, we separately estimate the impact of the potato contract by
6 including only the potato contract farmers in the treatment group and
The amount of land operated equals total land owned plus amount of land
rented-in, minus amount of land rented-out. only non-contract farmers in the control group. Similarly, for the maize
7
Since salaried workers are hired throughout the year and responsible for all contract-related effects, we take the maize contract farmers as the
the crops, we gave weight to the total expenditure for hiring salaried workers treatment group and the non-contract farmers as the control group.
during the cultivating season, depending on the area cultivated for each crop, in The combined effect of two contracts is reported in Table 4. As
calculating the costs of salaried workers per acre. shown in Panel A of the table, the contract farmers receive a

32
M.F. Khan, et al. Food Policy 85 (2019) 28–39

Table 4
Estimation results for two contracts pooled.
Outcome Variables Unit Coefficient S.E δ for β = 0 Coefficient bounds N

A: Crop performance (Potato)

Potato Price USD Per 100Kg 2.57*** 0.55 2.51 [2.12, 2.57] 184
Potato Yield Tons per acre 0.42* 0.23 5.19 [0.36, 0.42] 184
Potato Income per Acre USD/Acre 161.09*** 46.52 3.12 [135.82, 161.09] 184

B: Crop performance (Maize)

Maize Price USD Per 100Kg 0.13 0.41 1.75 [0.06, 0.13] 177
Maize Yield Tons per acre −0.12 0.08 7.17 [−0.12, −0.12] 177
Maize Income per Acre USD/Acre 8.24 23.15 30.79 [8.24, 8.37] 177

C: Farm productivity
Total Farm Income per Acre USD/Acre 117.42* 43.42 3.32 [101.45, 117.42] 200
Percent of Land Allocated to Potato Percent 3.83 3.90 3.52 [2.97, 3.83] 184
Percent of Land Allocated to Maize Percent −2.44 3.83 −1.72 [−2.44, −3.81] 177

D: Aggregate income
Total Farm Income USD 20,024 6,438 3.13 [16,211, 20,024] 200
Total Household Income USD 22,426 6577 4.18 [19,279, 22,426] 200
Non-Potato-Maize Crop Income USD −1201 1282 −4.76 [−1458, −1201] 200
Livestock Income USD 164.69 1190 1.76 [73.41, 164.69] 200
Other Income USD 2236 1297 −21.02 [2236, 2398] 200

E: Labor usage
Number of Salaried Workers Persons 0.71 0.77 1.86 [0.34, 0.71] 200
Cost of Salaried Workers USD/Year 830.24 1146 1.72 [362.50, 830.24] 200

For the potato crop, we have 184 observations because 16 non-contract farmers did not cultivate potatoes. Similarly, we have 177 observations for the maize crop
because 16 non-contract farmers, 5 maize contract farmers, and 2 potato contract farmers did not grow maize in the spring season. The control variables in the
regressions include the years of schooling of the household head, his or her age, the number of adult family members, land size, land value per acre (a proxy for land
quality), value of non-farm assets, and the number of livestock owned. Panels are based on the classification of variables into families for p-value adjustment.
Standard errors are obtained by bootstrapping with 50 replications, adjusted for multiple hypotheses testing. **Significant at 5% level;
* Significant at 10% level;
*** Significant at 1% level, based on the adjusted standard errors.

significantly higher price for their potato crop than the non-contract potato contract (Panel A). The potato contract, however, is associated
farmers. Income per acre and physical yield of potato crop are also with significantly higher gross output value per acre as well as the cost
higher for the contract farmers. In contrast, there is no significant dif- of inputs per acre (Panel A and B). Since the increase in gross output
ference in price, income per acre and physical yield between the value is larger than that of costs, the contract farmers have higher in-
treatment and control groups for maize (Panel B). We do not observe come per acre from the potato crop (Panel C).
any significant differences between contract farmers and non-contract Potato contract farmers also allocate about nine percentage points
farmers in total farm income, total household income or income from more of their land to potatoes than non-contract farmers (Panel C). As
other sources, such as other farm products, livestock income, and non- there is no significant difference in income from other sources between
agricultural income (Panel D). the contract and non-contract farmers, the gain in income from the
Table 4 also shows estimates of robustness to unobserved hetero- potato contract leads to a significant positive association between total
geneity as per the procedure of Oster (2017). Overall, both the value of annual income and potato contract farming (Panel D). The gain in in-
δ and the coefficient bounds point to robustness in our estimates. For come seems to be driven by the higher price of potatoes as well as the
instance, the value of δ = 2.51 in Panel A for the effect of contract larger proportion of land that is allocated to high quality potatoes. On
farming on potato price indicates that unobservables have to be 2.51 average, the potato farmers in our sample earn roughly 53,000 USD
times as important as the control variables in order to drive this more in household income than non-contract farmers with similar land
treatment effect to zero. Since this value is greater than 1, the effect can and household characteristics, an increase that is approximately 2.5
be considered robust to unobserved heterogeneity. Similarly, the coef- times the 19,800 USD income of the average non-contract farmer in our
ficient bounds interval (2.12, 2.57) does not contain zero, which also sample (Table 2).
implies that the estimate is robust.8 As for employment effects, the total cost of salaried workers and
To compare the effects of the potato contract farming scheme to casual labor per acre is not significantly higher (Panel B). Potato con-
those of the maize contract farming scheme, we first compare only the tract farmers, however, employ 2.6 more salaried workers, on average,
potato contract farmers with non-contract farmers. As shown in Panel A than similar non-contract farmers, which provides interesting evidence
of Table 5, potato contract farming is associated with a statistically of positive spillover effects for the potato contract on the employment
significant price increase of approximately 10 USD per 100 kg. This of skilled labor in the form of farm managers (Panel E). These results
higher price is mainly due to the special variety of potato that the are also robust to unobserved heterogeneity since the estimated values
contract farmers sell to the company. As for the physical yield of the of δ are greater than 1 and the coefficient bounds intervals do not
potato crop, there is no statistically significant association with the contain zero when the coefficient estimates are statistically significant.
In Table 6, we show a comparison between maize contract farmers
and non-contract farmers. As indicated, both the contract and non-
8
For some coefficients, δ is less than 0, which implies that the coefficient contract farmers receive a similar price (i.e., approximately 21 USD per
increases in magnitude due to the controls. Although this does not indicate that 100 kg of maize) (Panel A). The maize contract also does not have
the coefficient is unstable, it means that the method is not informative re- significant association with physical yield, cost of inputs, income per
garding omitted variable bias. Since we observe δ < 0 for insignificant coef- acre from the maize crop or total household income. This lack of any
ficients, this is not a major concern for our results. benefit of maize contract farming may be due to the fact that maize is a

33
M.F. Khan, et al. Food Policy 85 (2019) 28–39

Table 5
Estimation results for potato contract.
Outcome variables Unit Coefficient S.E δ for β = 0 Coefficient bounds N

A: Crop Performance
Potato Price USD Per 100Kg 10.43*** 0.52 1.01 [1.71, 10.43] 101
Potato Yield Tons per acre −0.04 0.32 −1.08 [-0.04, −0.08] 101
Potato-Gross Output Value USD/Acre 507.14*** 72.82 2.02 [487.76, 507.14] 101

B: Costs
Potato-Cost of Machinery USD/Acre 3.77 5.07 −13.39 [3.77, 4.17] 101
Potato-Cost of Inputs USD/Acre 68.06* 25.21 2.95 [59.92, 68.06] 101
Potato-Cost of Casual Labor USD/Acre −2.49 4.30 −4.38 [−2.49, −3.11] 101
Potato-Cost of Salaried Labor USD/Acre 2.91 4.55 5.81 [2.58, 2.91] 101

C: Farm Productivity
Potato Income per Acre USD/Acre 437.03*** 77.08 2.44 [437.03, 441.52] 101
Total Farm Income per Acre USD/Acre 333.23*** 76.64 2.97 [315.52, 333.23] 117
Percent of Land Allocated to Potato Percent 9.13** 6.31 3.14 [7.14, 9.13] 101

D: Aggregate Income
Total Farm Income USD 49,255** 10,199 2.00 [38,632, 49,255] 117
Total Household Income USD 52,873** 10,399 3.22 [48,450, 52,873] 117
Non-potato Crop Income USD 6782 6035 1.98 [3639, 6782] 117
Livestock Income USD 2917 2311 −6.18 [2917, 3514] 117
Other Income USD 700.23 636.46 −6.36 [700.23, 827.39] 117

E: Labor Usage
Number of Salaried Workers Persons 2.64* 1.31 5.98 [2.31, 2.64] 117

The control variables in the regressions include the years of schooling of the household head, his or her age, the number of adult family members, land size, land
value per acre (a proxy for land quality), value of non-farm assets, and the number of livestock owned. Panels are based on the classification of variables into families
for p-value adjustment. Standard errors are obtained by bootstrapping with 50 replications, adjusted for multiple hypotheses testing.
* Significant at 10% level.
** Significant at 5% level.
*** Significant at 1% level, based on the adjusted standard errors.

Table 6
Estimation results for maize contract.
Outcome variables Unit Coefficient S.E δ for β = 0 Coefficient bounds N

A: Crop Performance
Maize Price USD Per 100Kg 0.11 0.45 1.36 [0.03, 0.11] 162
Maize Yield Tons per acre −0.14 0.08 7.41 [-0.14, −0.14] 162
Maize-Gross Output Value USD/Acre 11.88 23.82 −8.50 [11.88, 13.65] 162

B: Costs
Maize-Cost of Machinery USD/Acre −2.66 2.05 527.04 [-2.66, −2.79] 162
Maize-Cost of Inputs USD/Acre 7.92 6.79 −5.16 [7.92, 9.20] 162
Maize-Cost of Hired Labor USD/Acre 1.79 1.44 −6.90 [1.79, 2.06] 162
Maize-Cost of Salaried Labor USD/Acre −0.36 2.19 −1.92 [-0.36, −0.56] 162

C: Farm Productivity
Maize Income per Acre USD/Acre 4.78 24.52 −12.89 [4.78, 5.39] 162
Total Farm Income per Acre USD/Acre 69.45 40.53 4.17 [59.77, 69.45] 184
Percent of Land Allocated to Maize Percent −3.53 4.08 −3.03 [-3.53, −4.63] 162
D: Aggregate Income
Total Farm Income USD 15,168 5932 3.62 [12,340, 15,168] 184
Total Household Income USD 16,969 6023 4.36 [14,229, 16,969] 184
Non-maize Crop Income USD 8781 4502 4.11 [7279, 8781] 184
Livestock Income USD −669.46 1003 −8.97 [−669.46, −750.07] 184
Other Income USD 2471 1395 −78.90 [2471, 2626] 184

E: Labor Usage
Number of Salaried Workers Persons 0.66 0.80 1.65 [0.27, 0.66] 184

The control variables in the regressions include the years of schooling of the household head, his or her age, the number of adult family members, land size, land
value per acre (a proxy for land quality), value of non-farm assets, and the number of livestock owned. Panels are based on the classification of variables into families
for p-value adjustment. Standard errors are obtained by bootstrapping with 50 replications, adjusted for multiple hypotheses testing. *Significant at 10% level;
**Significant at 5% level; ***Significant at 1% level, based on the adjusted standard errors.

34
M.F. Khan, et al. Food Policy 85 (2019) 28–39

normal staple crop that has little potential for quality improvement and observe significant differences between potato contract farmers and
is also widely available in the spot market. There was also no statisti- non-contract farmers in the employment of casual labor. As for the
cally significant difference in income from other sources, such as in- maize contract, it does not associate with higher employment numbers
come from crops other than maize, livestock income and income from for either casual labor or salaried workers. This is consistent with the
non-agricultural sources, for the maize contract farming scheme (Panel findings of Herrmann and Grote (2015), who found a limited impact of
D). Moreover, neither the number of salaried workers nor the cost of large-scale agricultural investment through the employment of casual
hired labor is statistically different, which suggests no spillover effect in labor.
labor market. One of the limitations of our study is the small sample size, which
To further confirm the relative profitability of the maize and potato gives rise to the question of low statistical power. Upon examination,
contract farming schemes, we descriptively examine the income and we did find that the tests for yield, production costs, and income from
costs of the five households who participated in both contract farming other sources have low power due to their large standard errors. Thus,
schemes. The within-household comparison allows us to control un- in these cases, we cannot rule out the possibility of imprecise coefficient
observable characteristics at the household level. As shown in Table A4 estimates. We find, however, sufficient statistical power in the tests for
in the appendix, we observe that potato contract farming results in a income per acre as well as total household income, which suggests that
much higher income per acre (260 USD) than maize contract farming our main argument regarding the impact of contract farming on farm
(30 USD). This supports our argument that the potato contract farming productivity and the welfare of the farmers is supported despite the
scheme which involves production of high quality crop through pro- small sample size.
vision of quality seeds and regular technical advises to farmers is more We observe no significant difference between the income or pro-
lucrative than the contract farming scheme involving a normal staple ductivity of maize contract farmers and that of non-contract farmers.
crop production where no allied services such as provision of high The question, then, arises: Why do farmers engage in maize contract
quality seed or technical advice are provided to the participating farming if there is seemingly no clear advantage to doing so? As we
farmers. discussed earlier, interviews with farmers and company field re-
presentatives revealed that farmers preferred to have the maize con-
5. Conclusion and policy implications tracting company buy their maize because of the company’s efficient,
transparent, and electronic payment system. Farmers choose to join the
In Pakistan, contract farming offers an alternative to the current maize contract program mainly because of their confidence in the
system of marketing, which lacks efficiency and organization company as a reliable buyer. It is also notable that, contrary to the
(Chaudhry and Ahmad, 2000). Apart from assuring markets for farmers, theoretical prediction, the farmers in our study area had no incentive to
companies offer their contract farmers extension services and logistical side-sell their maize products since the agreed contract price is the same
support for transporting their crop. This study empirically analyzes the as the price in the open market. Given that this scheme has continued
impact of contract farming on the economic welfare of participating since the 1970s, our study shows that the contract farming of a staple
farmers. Based on results, we found that potato contract farming, where crop is viable so long as it develops an effective institution, even if there
farmers produce high quality potatoes by receiving support from the is no clear income gain for the contract farmers.
company in the form of seed and regular technical services, is asso- Potato contract farming in Pakistani Punjab has improved land
ciated with significantly higher income but not with physical yield per productivity by enabling farmers to shift from low value-added vari-
acre. eties to a high value-added variety. In the historical context of Pakistani
For maize contract farmers, we did not observe any increase in in- Punjab, shifts to high value-added crops have contributed to pro-
come or productivity. This may be because the maize is also widely ductivity gains (Kurosaki, 2003). Such historical improvement mostly
available in the open market and have little potential for quality im- occurred due to the development of an irrigation and transportation
provement. Such a result is consistent with the findings of Ragasa et al. infrastructure. This study shows the possibility of a contract-farming-
(2018) and Soullier and Moustier (2018), who also show that the induced shift towards high value-added agriculture, with improved
contract farming of undifferentiated staple crops has limited prospects marketing.
of profitability.
No negative association was seen between contract farming and Acknowledgements
income from other sources, such as income from non-contract crops,
livestock income and non-agricultural income. This suggests that labor We thank two anonymous reviewers of this journal for their useful
reallocation for the production of contract crops, which are potentially comments. We are grateful to the priority project under Hitotsubashi
more labor intensive and require more careful management, is not a Institute for Advanced Study (HIAS), Hitotsubashi University, Japan for
significant factor. The absence of changes in income from other sources funding the data collection for this research. We also thank Sachiko
differs from the findings of Bellemare (2018), which is likely because Miyata of World Bank for sharing questionnaire of their contract
the large-scale farmers in our study do not rely on family labor for farming study with us.
agriculture but rather outsource the work to skilled salaried workers
and casual laborers. Declarations of interest
We also see evidence of an increased number of salaried workers
employed by the farmer for the potato contract. In contrast, we do not None.

Appendix

See Tables A1–A4.

35
 M.F. Khan, et al.

Table A1
Average treatment effect on the treated (ATT) estimation results for two contracts pooled – PSM.
Treatment: Either or Both Contracts
Control: Non-Contract

Kernel matching NN matching

Unit Treatment Control ATT S.E. Rosenbaum ATT S.E. Rosenbaum N Off support in PSM
bounds bounds
critical level critical level
of odds ratio of odds ratio

Potato Price USD Per 100Kg 8.33 5.69 2.64*** 0.51 3.00 2.76*** 0.5 3.50 184 6
Potato Yield Tons per acre 10.46 10.25 0.21 0.22 n.a 0.34* 0.27 1.50 184 6
Potato Income per Acre USD/Acre 85.86 −73.77 159.63*** 55.02 2.50 174.32*** 47.706 3.00 184 6

36
Percent of Land Allocated to Potato Percent 75.24 70.93 4.32 4.34 n.a 4.92 4.10 n.a 184 6
Maize Price USD Per 100Kg 20.82 20.70 0.12 0.37 n.a 0.18 0.43 n.a 177 6
Maize Yield Tons per acre 2.69 2.81 −0.12 0.08 n.a −0.11* 0.10 1.50 177 6
Maize Income per Acre USD/Acre 175.15 158.23 16.92 25.13 n.a 14.90 27.57 n.a 177 6
Percent of Land Allocated to Maize Percent 68.47 69.74 −1.26 4.41 n.a −0.78 4.74 n.a 177 6
Total Farm Income per Acre USD/Acre 249.72 134.73 115.00*** 41.73 2.00 119.55*** 45.58 2.00 200 6
Total Farm Income USD 34,816 15,172 19,644*** 6756 1.50 16,886*** 7,982 1.75 200 6
Total Household Income USD 45,702 23,909 21,793*** 7506 1.50 18,267** 8,766 1.50 200 6
Non-Potato-Maize Crop Income USD 3518 5397 −1878 2243 n.a −3791 2756 n.a 200 6
Livestock Income USD 8375 8078 297 1550 n.a −409 1934 n.a 200 6
Other Income USD 2510 658 1852* 1070 1.00 1790 1144 n.a 200 6
Number of Salaried Workers 6.34 5.76 0.58 0.82 n.a 0.48 1.04 n.a 200 6
Cost of Salaried Workers USD/Year 7964 7320 645 1165 n.a. 628 1507 n.a. 200 6

Standard errors are obtained by bootstrapping with 50 replications.

* Significant at 10% level.
** Significant at 5% level.
*** Significant at 1% level.
Food Policy 85 (2019) 28–39
 M.F. Khan, et al.

Table A2
. ATT Estimation Results for Potato Contract – PSM.
Treatment: Potato Contract
Control: Non-Contract

Kernel matching NN matching

Unit Treatment Control ATT S.E. Rosenbaum ATT S.E. Rosenbaum N Off Support in PSM
bounds bounds
critical level critical level
of odds ratio of odds ratio

Potato Price USD Per 100Kg 16.10 5.57 10.53*** 0.48 9.00 10.39*** 0.47 10.00 101 2
Potato Yield Tons per acre 10.14 10.26 −0.12 0.32 n.a 0.00 0.47 n.a 101 2
Potato-Gross Output Value USD/Acre 1100.9 568.97 531.97*** 95.51 6.50 531.98*** 104.91 5.25 101 2

37
Potato-Cost of Machinery USD/Acre 74.68 73.44 1.24 6.14 n.a −1.09 6.40 n.a 101 2
Potato-Cost of Inputs USD/Acre 553.63 488.73 64.90*** 23.19 4.00 73.45*** 27.67 3.00 101 2
Potato-Cost of Casual Labor USD/Acre 81.9 84.2 −2.30 5.08 n.a −2.04 5.61 n.a 101 2
Potato-Cost of Salaried Labor USD/Acre 23.81 23.16 0.65 4.93 n.a 1.43 7.32 n.a 101 2
Percent of Land Allocated to Potato Percent 80.67 72.02 8.65 4.91 n.a 3.28 7.96 n.a 101 2
Potato Income per Acre USD/Acre 366.91 −100.59 467.49*** 96.50 6.00 460.22*** 110.81 4.75 101 2
Total Farm Income per Acre USD/Acre 486.12 124.33 361.79*** 129.91 2.75 334.35** 123.79 2.25 117 2
Total Farm Income USD 67,000 13,851 53,149** 22,896 3.50 52,260** 21,080 3.50 117 2
Total Household Income USD 77,742 21,205 56,537** 22,401 4.00 57,015*** 20,723 5.00 117 2
Non-potato Crop Income USD 25,778 16,309 9,469 9,402 n.a 9,004 10,075 n.a 117 2
Livestock Income USD 9718 6951 2766 3323 n.a 3951 3051 n.a 117 2
Other Income USD 1024 402 621 729 n.a 804 734 n.a 117 2
Number of Salaried Workers 7.11 5.38 1.72 1.84 n.a 3.02** 1.69 2.25 117 2

Standard errors are obtained by bootstrapping with 50 replications. *Significant at 10% level.
** Significant at 5% level.
*** Significant at 1% level.
Food Policy 85 (2019) 28–39
M.F. Khan, et al. Food Policy 85 (2019) 28–39

Table A3
ATT estimation results for maize contract – PSM.
Treatment: Maize Contract
Control: Non-Contract

Kernel matching NN matching

Unit Treatment Control ATT S.E. Rosenbaum ATT S.E. Rosenbaum N Off Support in
bounds bounds PSM
critical level critical level
of odds ratio of odds ratio

Maize Price USD Per 20.77 20.69 0.08 0.35 n.a −0.02 0.39 n.a 162 5
100Kg
***
Maize Yield Tons per 2.67 2.80 −0.13 0.09 n.a −0.18 0.09 1.75 162 5
acre
Maize-Gross Output Value USD/Acre 572.03 554.06 17.97 21.12 n.a −0.9 23.66 n.a 162 5
Maize-Cost of Machinery USD/Acre 45.33 48.13 −2.80 2.52 n.a −3.52 3.23 n.a 162 5
Maize-Cost of Inputs USD/Acre 252.53 247.76 4.77 6.54 n.a 5.58 8.35 n.a 162 5
Maize-Cost of Hired Labor USD/Acre 87.38 85.36 2.01 1.64 n.a 2.19 1.84 n.a 162 5
Maize-Cost of Salaried Labor USD/Acre 15.30 16.15 −0.85 3.00 n.a −1.31 3.87 n.a 162 5
Percent of Land Allocated to Maize Percent 67.25 69.59 −2.35 4.58 n.a −2.80 5.53 n.a 162 5
Maize Income per Acre USD/Acre 170.91 156.65 14.25 21.44 n.a −4.42 26.08 n.a 162 5
Total Farm Income per Acre USD/Acre 195.84 134.27 61.57 45.67 n.a 59.51 55.09 n.a 184 5
Total Farm Income USD 28,850 15,329 13,520* 7642 1.25 12,870* 8855 1.25 184 5
Total Household Income USD 39,731 24,799 14,932 9,152 n.a 13,185 10,508 n.a 184 5
Non-maize Crop Income USD 11,463 4524 6938 4854 n.a 6601* 5932 1.00 184 5
Livestock Income USD 8106 8749 −644 3019 n.a −1320 3371 n.a 184 5
Other Income USD 2775 720 2055 1620 n.a 2020 1565 n.a 184 5
Number of Salaried Workers 6.46 6.16 0.30 1.64 n.a −0.52 1.90 n.a 184 5

**
Standard errors are obtained by bootstrapping with 50 replications. Significant at 5% level.
* Significant at 10% level.
*** Significant at 1% level.

Table A4
Income and cost structure of five farmers who contract with both companies.
Potato Maize

Price (USD per 100 kg) 16.54 20.48

Yield (tons per acre) 10.56 1.81
Gross Output Value (USD per acre) 1005.33 374.52
Paid Cost of Machinery (USD per acre) 66.67 39.28
Paid Cost of Inputs (USD per acre) 559.24 190.00
Paid Cost of Casual Labor (USD per acre) 78.78 86.29
Paid Cost of Salaried Workers (USD per acre) 40.55 29.10
Income (USD per acre) 260.10 29.85

Appendix A. Supplementary material

Supplementary data to this article can be found online at https://doi.org/10.1016/j.foodpol.2019.04.004.

References 99 (2), 357–378.

Bolwig, S., Gibbon, P., Jones, S., 2009. The economics of smallholder organic contract
farming in tropical Africa. World Dev. 37 (6), 1094–1104.
Abebe, G.K., Bijman, J., Kemp, R., Omta, O., Tsegaye, A., 2013. Contract farming con- Chaudhry, M.G., Ahmad, B., 2000. Pakistan. In: Ali, M. (Ed.), Dynamics of Vegetable
figuration: smallholders’ preferences for contract design attributes. Food Policy 40, Production, Distribution and Consumption in Asia. Asian Vegetable Research and
14–24. Development Center (AVRDC), pp. 271–302.
Altonji, J.G., Elder, T.E., Taber, C.R., 2005. Selection on observed and unobserved vari- Ching-Cheng, C., Chi-Chung, C., Min-Ching, C., Wei-Chun, T., 2006. Is contract farming
ables: assessing the effectiveness of Catholic schools. J. Political Econ. 113 (1), more profitable and efficient than non-contract farming: a survey study of rice farms
151–184. in Taiwan. Paper presented at the American Agricultural Economics Association
Anríquez, G., Valdes, A., 2006. Determinants of farm revenue in Pakistan. Pakistan Annual Meeting, Long Beach, California.
Develop. Rev. 45 (2), 281–301. Chodavarapu, S., Giertz, A., Jaeger, P., 2017. Agribusiness in South Asia. In: Lopez-
Bellemare, M.F., 2012. As you sow, so shall you reap: the welfare impacts of contract Acevedo, G., Medvedev, D., Palmade, V. (Eds.), South Asia’s turn: Policies to boost
farming. World Dev. 40 (7), 1418–1434. competitiveness and create the next export powerhouse. World Bank Publications,
Bellemare, M.F., 2018. Contract farming: opportunity costs and trade-offs. Agric. Econ. Washington, DC.
49, 279–288. CIMMYT., 1989. Maize research and development in Pakistan. Mexico, D.F., CIMMYT.
Bellemare, M.F., Lee, Y.N., Novak, L., 2018. Contract Farming as Partial Insurance. Dileep, B.K., Grover, R.K., Rai, K.N., 2002. Contract farming in tomato: an economic
Working Paper, University of Minnesota. analysis. Indian J. Agric. Econ. 57 (2), 197.
Bellemare, M.F., Bloem, J.R., 2018. Does contract farming improve welfare? A review. Dolan, C., Humphrey, J., 2000. Governance and trade in fresh vegetables: the impact of
World Dev. 112, 259–271. UK supermarkets on the African horticulture industry. J. Develop. Studies 37 (2),
Bellemare, M.F., Novak, L., 2017. Contract farming and food security. Am. J. Agric. Econ. 147–176.

38
M.F. Khan, et al. Food Policy 85 (2019) 28–39

Duflo, E., Glennerster, R., Kremer, M., 2007. Using randomization in development eco- Oster, E., 2017. Unobservable selection and coefficient stability: theory and evidence. J.
nomics research: a toolkit. Handbook Develop. Econ. 4, 3895–3962. Business Econ. Statistics 1–18.
Gatto, M., Wollni, M., Asnawi, R., Qaim, M., 2017. Oil palm boom, contract farming, and Otsuka, K., Nakano, Y., Takahashi, K., 2016. Contract farming in developed and devel-
rural economic development: village-level evidence from Indonesia. World Dev. 95, oping countries. Ann. Rev. Resource Econ. 8, 353–376.
127–140. Oya, C., 2012. Contract farming in sub-Saharan Africa: a survey of approaches, debates
Glover, D.J., 1984. Contract farming and smallholder outgrower schemes in less-devel- and issues. J. Agrarian Change 12 (1), 1–33.
oped countries. World Dev. 12 (11–12), 1143–1157. Ragasa, C., Lambrecht, I., Kufoalor, D.S., 2018. Limitations of contract farming as a pro-
Government of Pakistan., 2010. Census of agriculture 2010. Islamabad: Pakistan Bureau poor strategy: the case of maize outgrower schemes in upper West Ghana. World Dev.
of Statistics. 102, 30–56.
Grosh, B., 1994. Control farming in Africa: an application of the new institutional eco- Rao, E.J., Qaim, M., 2011. Supermarkets, farm household income, and poverty: insights
nomics. J. African Econ. 3 (2), 231–261. from Kenya. World Dev. 39 (5), 784–796.
Herrmann, R., Grote, U., 2015. Large-scale agro-industrial investments and rural poverty: Reardon, T., Barrett, C.B., Berdegué, J.A., Swinnen, J.F.M., 2009. Agrifood industry
evidence from sugarcane in Malawi. J. African Econ. 24 (5), 645–676. transformation and small farmers in developing countries. World Dev. 37 (11),
Key, N., McBride, W., 2003. Production contracts and productivity in the US hog sector. 1717–1727.
Am. J. Agric. Econ. 85 (1), 121–133. Reardon, T., Berdegué, J.A., 2002. The rapid rise of supermarkets in Latin America:
Key, N., Runsten, D., 1999. Contract farming, smallholders, and rural development in challenges and opportunities for development. Develop. Policy Rev. 20 (4), 371–388.
Latin America: the organization of agro processing firms and the scale of out grower Reardon, T., Timmer, C.P., 2007. Transformation of markets for agricultural output in
production. World Dev. 27 (2), 381–401. developing countries since 1950: How has thinking changed? Handbook of Agric.
Kurosaki, T., 2003. Specialization and diversification in agricultural transformation: the Econ. 3, 2807–2855.
case of West Punjab, 1903–1992. Am. J. Agric. Econ. 85 (2), 372–386. Rosenbaum, P.R., 2002. Observational Studies. Springer, New York.
Leung, P., Setboonsarng, S., Stefan, A., 2008. Rice contract farming in Lao PDR: Moving Rosenbaum, P.R., Rubin, D.B., 1983. The central role of the propensity score in ob-
from subsistence to commercial agriculture (Working paper 90). Asian Development servational studies for causal effects. Biometrika 70 (1), 41–55.
Bank Institute, Tokyo. Sethboonsarng, S., Leung, P., Stefan, A., 2008. Rice Contract Farming in Lao PDR: Moving
Maertens, M., Swinnen, J.F.M., 2009. Trade, standards, and poverty: evidence from from Subsistence to Commercial Agriculture (Discussion paper 90). ADB Institute,
Senegal. World Dev. 37 (1), 161–178. Tokyo.
Maertens, M., Velde, D.V., 2017. Contract-farming in staple food chains: the case of rice in Simmons, P., Winters, P., Patrick, I., 2005. An analysis of contract farming in East Java,
Benin. World Dev. 95, 73–87. Bali, and Lombok Indonesia. Agric. Econ. 33 (s3), 513–525.
Minten, B., Randrianarison, L., Swinnen, J., 2007. Spillovers from high-value agriculture Soullier, G., Moustier, P., 2018. Impacts of contract farming in domestic grain chains on
for exports on land use in developing countries: evidence from Madagascar. Agric. farmer income and food insecurity Contrasted evidence from Senegal. Food Policy 79,
Econ. 37 (2–3), 265–275. 179–198.
Mishra, A.K., Kumar, A., Joshi, P.K., D’Souza, A., Tripathi, G., 2018. How can organic rice Ton, G., Vellema, W., Desiere, S., Weituschat, S., D’Haese, M., 2018. Contract farming for
be a boon to smallholders? Evidence from contract farming in India. Food Policy 75, improving smallholder incomes: what can we learn from effectiveness studies? World
147–157. Dev. 104, 46–64.
Mishra, A.K., Kumar, A., Joshi, P.K., D’Souza, A., 2016. Impact of contracts in high Wang, H., Moustier, P., Loc, N.T.T., 2014a. Economic impact of direct marketing and
yielding varieties seed production on profits and yield: the case of Nepal. Food Policy contracts: the case of safe vegetable chains in northern Vietnam. Food Policy 47,
62, 110–121. 13–23.
Miyata, S., Minot, N., Hu, D., 2009. Impact of contract farming on income: linking small Wang, H., Wang, Y., Delgado, M.S., 2014b. The transition to modern agriculture: contract
farmers, packers, and supermarkets in China. World Dev. 37 (11), 1781–1790. farming in developing economies. Am. J. Agric. Econ. 96 (5), 1257–1271.
Narayanan, S., 2014. Profits from participation in high value agriculture: evidence of Warning, M., Key, N., 2002. The social performance and distributional consequences of
heterogeneous benefits in contract farming schemes in Southern India. Food Policy contract farming: an equilibrium analysis of the Arachide de Bouche program in
44, 142–157. Senegal. World Dev. 30 (2), 255–263.
Nolte, K., Ostermeier, M., 2017. Labour market effects of large-scale agricultural in- Westfall, P.H., Young, S.S., 1993. Resampling-Based Multiple Testing: Examples and
vestment: conceptual considerations and estimated employment effects. World Dev. Methods for p-value Adjustment. Wiley Series, New York.
98, 430–446. Winters, P., Simmons, P., Patrick, I., 2005. Evaluation of a hybrid seed contract between
Ochieng, D.O., Veettil, P.C., Qaim, M., 2017. Farmers’ preferences for supermarket con- smallholders and a multinational company in East Java Indonesia. J. Develop. Studies
tracts in Kenya. Food Policy 68, 100–111. 41 (1), 62–89.

39