The Transition to Modern Agriculture: Contract Farming in Developing Economies

Abstract

Recent years have seen considerable interest in the impact of contract farming on farmers

in developing countries, motivated out of belief that contract farming spurs the transition

to modern agriculture. In this paper, we provide a thorough review of the empirical

literature on contract farming in both developed and developing countries, paying careful

attention to broad implications of this research for economic development. We first find

empirical studies consistently support the positive contribution of contract farming to

production and supply chain efficiency. We also find that most empirical studies identify

a positive and significant effect of contract farming on farmer welfare, yet are often

unable to reach consistent conclusions as to significant correlates of contract participation.

We support our review with a meta-analysis of the empirical literature to identify study

characteristics that are conditionally correlated with particular empirical outcomes. Our

meta-analysis indicates that studies using larger, more recent datasets are more likely to

report a priori expected empirical results, but that empirical findings are not statistically

different across developmental status or agricultural commodities.

1
Contract farming (CF) has long been established in developed countries. In recent

decades, it has become more popular in developing countries, yet the fundamentals of CF

are different across developed and developing countries primarily because the latter is

inescapably linked to economic development. These differences raise important issues for

governments, especially those in developing countries, as to whether they should adopt a

policy to promote, regulate, or prevent the development of CF, or to leave the status quo.

In this article, we review and synthesize the existing research on CF, comparing studies

related to developed and developing countries, in order to identify general conclusions

developed by the literature. We focus on CF as a vehicle of transition to modern

agriculture in developing economies, using China as an example.

Driven by competition in the food market, the food supply chain has quickly

become vertically coordinated in developed countries, with CF as the dominant form of

coordination for most traditional commodities. It is generally agreed that two major

reasons explain why farmers opt to contract with a downstream processor and marketer:

risk reduction (Allen and Lueck 1995; Hennessey and Lawrence 1999) and transaction

cost reduction (Hobbs and Young 1999).

For developing countries, there are other potential benefits associated with CF.

Since farm scale tends to be small, farmers are generally less educated, production and

management technologies are less efficient, and infrastructure such as transportation, cold

storage, and information channels are underdeveloped; contracting with a large

agribusiness firm may be the only way farmers in developing countries can access higher

end markets and receive higher returns (Barrett et al. 2012). Transaction cost reduction is

2
also an important motive given relative scarcity of resources (Bijman 2008). These two

motives may be more important than the risk reducing motive (Wang et al. 2011).

The consequences and impacts from food supply chain coordination may also be

different in developed versus developing economies. In both economies, CF provides a

stable and consistent supply to the downstream food processors and consumers, gives the

processors more control, makes food traceable, and reduces transaction costs. In

developed countries, farmers are concerned about losing their independence and

important business decision skills (Schulze et al. 2006); yet, in developing countries,

farmers may acquire better production technology, achieve scale economies, and receive

a higher return to improve their welfare (Tripathi et al. 2005; Miyata et al. 2009). Indeed,

CF may help modernize the entire agricultural sector (Morrissy 1974), which is one

important aspect of economic development. However, there are also concerns associated

with CF, particularly salient for developing countries, including increased instability for

nonparticipants in the community; disruption of power relations in traditional household

culture; overreliance on cash crops that may leave households more vulnerable to food

shortages; and exploitation from large firms (Key and Runsten 1999).

To anchor our discussion of CF in developing countries, we focus on China as a

particular case. While there may be considerable heterogeneity across developing

countries with respect to CF, studies focusing on China may yield important implications

on CF in developing countries for the following reasons. First, China has a large number

of small and low income farms. Second, different from other developing countries in

Latin America or Africa that were open to foreign firms in the 1970s, China is a late

3
comer without a long tradition of contracting, providing an opportunity to understand

how the introduction of CF impacts development. Third, China has a large domestic food

market and newly established domestic firms to serve this market (many processors are

domestic firms). Fourth, China has a strong government with regulatory as well as

promotional powers.

In this paper, we make two contributions to the empirical literature on CF. We

first provide a thorough review of the existing empirical literature, focusing on factors

that motivate farmer participation in CF and the impact of CF on agricultural productivity

and farmer welfare in developing countries (section 3). Second, we conduct a meta-

analysis of the existing empirical literature to look for attributes that may explain why the

empirical literature has failed to identify consistent effects of several variables on the

decision to contract and its impact (section 4).

A Brief Conceptual Overview

We first provide a basic conceptual definition and a cursory description of several

important theoretical developments in the literature. Readers are referred to the

references provided in this section for further theoretical details.

Different to traditional marketing in which farmers produce under their own

decisions on variety, quantity, quality, and timing, and then sell to the open market at the

market price at harvest1, CF refers to “agricultural production carried out according to an

agreement between a buyer and farmers, which establishes conditions for the production

and marketing of a farm product or products” (FAO 2013). Typically, the farmer agrees

4
to provide certain quantities of a specific commodity at the specified quality standards

and time, while the buyer commits to pay at a specified price or pricing scheme. The

buyer may also supply some inputs or technical support to the farmer.

Reasons supporting vertical coordination in the food supply chain through CF are

similar to vertical integration, except the buyer does not have to assume the full

responsibility of financial investment and risks, rather sharing these with farmers

(Schrader 1986). Economic contracting theories have long been established in general

(Coase 1937), for vertical integration (Williamson 1971), and for marketing contracts

around risk and information (Laffont and Tirole 1986; Allen and Lueck 1995). The

application of such theory in agriculture can be traced back to the 1990s (Hennessy 1996;

Leathers 1999; Goodhue 2000). For a discussion and review, see Wu (2006).

CF has quickly spilled over from developed countries to developing countries,

especially since large agribusiness firms began to acquire fresh produce from Latin

America to supply their home markets. CF helps the modernization of small growers

(Morrissy, 1974), and also has economic, political, and social impacts (Glover 1984).

The Empirics of Contract Farming

There is a rich empirical literature investigating two primary facets of CF:

participation incentives and the impact on farms. The econometrics for the former is often

a simple binary outcome modeled with probit or logit regression. The latter, itself, has

many facets, including the effects of CF on farm growth, productivity, income, or product

quality, often evaluated by including a binary explanatory variable of CF participation.

5
Contract Participation

Many estimate the probability that a farmer will choose to contract as the first step

in a two-step econometric procedure focusing primarily on the impact of CF on farmer

welfare (e.g., Katchova and Miranda 2004; Simmons et al. 2005; Miyata et al. 2009;

Wang et al. 2011; Bellemare 2012; Ito et al. 2012). Others focus exclusively on farmers’

decision to contract (e.g., Guo 2005; Masakure and Henson 2005; Zhu and Wang 2007).

Under the presumption that CF improves welfare – a consensus reached theoretically and

empirically – it is vital to fully understand which factors are associated with farmers’

willingness to contract, in order to leverage welfare gains and economic development.

Demographic Factors

Age, gender, and education are often included in empirical studies. Unfortunately, there

fails to be a consensus as to both the sign and significance of each of these demographic

variables on the probability of participation. Many find that the age of the head of the

household has a significantly negative effect, such as Simmons et al. (2005) for seed corn

in Indonesia and Bellemare (2012) for several commodities in Madagascar. However,

Katchova and Miranda (2004) find that age has a significantly positive effect for soybean

in the United States. Simmons et al. (2005) find an insignificant effect for seed rice and

broilers in Indonesia, so do Katchova and Miranda (2004) for corn and wheat in the U.S.

Ito et al. (2012) find a nonlinear age effect for watermelon in China.

For gender, Wang et al. (2011), Bellemare (2012) and Wainaina et al. (2012)

document that females are significantly less likely to adopt CF than males in China,

Madagascar and Kenya, respectively. One possible explanation is that, in developing

6
countries, institutional forces may provide females with disadvantageous contract

opportunities or conditions. Conversely, Leung et al. (2008), Arumugam et al. (2011), Hu

(2012), Freguin-Gresh et al. (2012), and Wang et al. (2013) all document an insignificant

gender effect across both developed and developing countries and for a variety of

commodities. There is no apparent explanation for the differing conclusions.

A large number of studies have found that the education level of the head of the

household is not significantly related to participation in a contract (Guo et al. 2005; Wang

et al. 2011; Bellemare 2012; Ito et al. 2012; Wang et al. 2013). Yet, there are other

studies that find it significantly positive (Zhu and Wang 2007; Arumugam et al. 2011; Hu

2012), negative (Ramaswami et al. 2006; Miyata et al. 2009; Wainaina et al. 2012), or

dependent on commodity (Simmons et al. 2005; Katchova and Miranda 2004). One

possible explanation for these differing conclusions is that the education effect is

nonlinear (Miyata et al. 2009).

While one might anticipate differences in the relationship between these basic

demographics and contract participation across developed and developing countries, the

widespread failure to obtain a general conclusion is perplexing. One possibility is that

these differences arise because of statistical or modeling differences – a dimension that

we explore in our subsequent meta-study of contract participation findings. Another

possibility is that institutional differences across countries and across commodities within

countries may lead to heterogeneity. It suggests policymakers who seek to promote CF as

a means of increasing farmer welfare should use caution when designing policies

targeting farmer participation, and be wary of the implications of heterogeneity.

7
Economic Factors

Farm size, farmer experience, specialization, risk preference, and credit

constraints are also investigated. Farm size is measured by either the number of laborers

or land acreage. The effect of land acreage is found significantly positive in a large

number of studies like Zhu and Wang (2007), Lu and Ma (2010), Arumugam et al. (2011),

Wang et al. (2011), Bellemare (2012), Freguin-Gresh (2012), Hu (2012) and Wang et al

(2013); negative in Leung et al. (2009); and insignificant in Ito et al. (2012), Birthal et al.

(2005), Miyata et al. (2009) and Wainaina et al. (2012). Simmons et al. (2005) find

different signs for different commodities. The dominant result that larger farms are more

likely to contract is consistent with the common belief that they are more likely to be

offered a contract, for the transaction cost saving benefit of the processor. However, it is

likely that small farmers may gain more from contracting, and encouraging small farmers

to contract is important from a developmental perspective. The number of laborers is

generally insignificant (Simmons et al. 2005; Zhu and Wang 2007; Miyata et al. 2009;

Wang et al. 2011; Bellemare 2012; Hu 2012; Ito et al. 2012). This is either because it

does not affect the participation decision, or its effect can be represented by land acreage.

The effect of farm experience, measured by the number of years farming, is again

found divided. Bellemare (2012) finds a positive and significant effect, indicating that

more experienced farmers are more likely to contract. Yet, Zhu and Wang (2007) find a

negative effect, and Arumugam et al. (2011) fail to uncover a significant link. Two

potential explanations for these conflicting findings include the evidence of a nonlinear

relationship between experience and contracting (Ramaswami et al. 2006) – more

8
experienced farmers are less likely to contract, but at a diminishing rate; and the

commodity specific effect (Birthal et al. 2005) – experience increases the likelihood of

contracting for dairy and vegetable growers, but decreases for broiler farmers.

The effect of farmer specialization, measured as the share of income from their

primary crop to that from other farm and non-farm activities, on CF is also diverse, with

general findings being divided across significantly positive and insignificant effects.

Warning and Key (2002), Guo et al. (2005) and Freguin-Gresh et al. (2012) all document

a positive and significant effect, Zhu and Wang (2007), Arumugam et al. (2011) and Hu

(2012) find an insignificant effect, while Ramaswami et al. (2006) and Wainaina et al.

(2012) find a negative and significant effect. These effects, as indicated by both Katchova

and Miranda (2004) and Birthal et al. (2006), are commodity specific.

Simmons et al. (2005) consider farmer access to credit as one potential motive for

contract participation. They find that credit constraints are not significant in the corn and

rice industry, but positive for broiler growers. This significant effect (for broilers) is

intuitive because farmers with poor access to credit may be particularly vulnerable to

market fluctuations, and may find increased safety in a contract.2

The impact of farmer risk aversion and the degree of market risk on contract

participation has been studied by several researchers. Guo et al. (2005), Zhu and Wang

(2007) and Ito et al. (2012) all find that farmer risk aversion is not a significant predictor

of contract participation. Wang et al. (2011) find that risk aversion is negatively related to

contract participation, and Wainaina et al. (2012) find it positive. The Wang et al. (2011)

result seems counter-intuitive, yet as they point out, CF is relatively new in China which

9
makes risk averse farmers wary of contracting. Furthermore, both Guo et al. (2005) and

Wang et al. (2011) include measures of price volatility in their studies and find no

statistically significant effect on contracting. One interpretation of this literature is that

market risk may not be an important driving factor leading farmers to contract in China,

or that price risk in an economy where food price keeps rising is favorable to farmers

without being locked in a fixed price contract. Taken jointly, these results suggest two

interesting insights. First, the majority of authors seem to find no link between contract

participation and farmer risk aversion, consistent with the view that CF, at least in

developing countries, is less related to risk reductions than market access and cost

reductions. Second, policymakers in developing countries should be aware that the

introduction of CI may appear risky to farmers who are unaccustomed to contracting. If

policymakers wish to use contracting as a policy lever for improving farmer welfare,

minimizing initial fears of new ventures into contracting may be crucial.

Many empirical studies have also considered the impact of a farm’s assets on the

likelihood of participation, measured as either the value of household assets or the value

of farm equipment. The majority find that this variable is statistically insignificantly

related to contract participation (Simmons et al. 2005; Leung et al. 2008; Wang et al.

2011; Bellemare 2012; Hu 2012; Wainaina et al. 2012). Only Warning and Key (2002)

identify a significantly positive effect of the farm equipment assets on contracting for a

sample of Senegalese peanut farmers, indicating that farmers with more equipment may

have higher productivity and are more capable to repay the initial loan in the contract.

One possible explanation for the general insignificance of farm assets in the participation

10
regression is that farm assets may be an alternative measure of farm size, in which case

any size effect may have already been captured by the size of the land.

Several other authors have explored a few interesting participation motives in

contracting. Zhu and Wang (2007) find previous experience with CF contributes

positively, which suggests that farmers’ previous CF experience was likely successful.

Although it is obvious that previous positive or negative experience will strongly

influence the future decision, given that many studies find evidence that CF increases

farmer welfare, the Zhu and Wang’s (2007) result indicates that future contracts may be

adopted more readily. This result is also consistent with Wang et al. (2011) in that risk

averse farmers are less likely to contract given uncertainties of entering into a contract

without much precedent. Government promotion policy is another factor that contributes

to CF participation in China (Guo 2005; Zhu and Wang 2007), indicating that recent

government efforts to encourage CF have been successful. If China is representative of

developing countries, these two studies provide crucial evidence that government can

play an important role in promoting CF in developing countries.

The distance to market may be a particularly important factor for farmers in

developing countries, but is found negative in Kenya (Wainaina et al. 2012) and positive

in Lao (Leung et al. 2008). Both results are potentially intuitive. On the one hand,

farmers that do not have access to a main road are less attractive partners for the

processor, while on the other hand, farmers who are farther from the market may find

additional security in contracting given their relative remoteness, and may be more likely

to contract. Hence, CF effects may also be dependent on infrastructural development.

11
Discussion

Despite the variety of commodities and countries used as CF case studies, we are unable

to draw simple, meaningful conclusions as to the direction of impact that many variables

have on the probability of farmer participation. It seems that simple arguments such as

developed versus developing, plant versus animal products, or countries in one continent

versus another are insufficient for explaining the lack of general consensus within the

literature. It appears that fundamental institutional or cultural differences that are perhaps

country and/or commodity specific determine the relationship between the variety of

explanatory factors considered in this literature, and contract participation.

These results are somewhat perplexing from a developmental policy perspective.

As it is well supported that CF can improve farmers’ welfare, developmental policy

should focus on increasing CF participation. Yet, if the empirical literature is unable to

identify generally significant motivating factors of CF participation, policy is not well

informed. We return to this issue with a statistical analysis in order to further sift through

the empirical studies reviewed thus far in an effort to obtain insight into possible

explanations for the differences in results produced by these empirical studies.

Welfare Impacts on Farms

The fact that CF has rapidly emerged and developed implies welfare gains for firms and

farms in general. The government and the public often care more about the welfare

impact on farmers. We summarize the findings from the literature for developed and

developing countries separately, emphasizing the latter.

12
 There is a relatively small body of literature on the impact of CF on farmers’

income in developed countries. Hu (2012) found improved returns to corn and soybean

farms in the United States, but not wheat farms. This is not a surprise because the price

set up in the contract is based on market price, so there should be no expected price

advantages in CF. The welfare gain from involvement in CF comes instead from risk

reduction and transaction cost savings (Hennessy 1996; Martin 1997; Gray and Boehlje

2005; Key 2013). Further, farmers in developed countries have negative concerns that

CF may make them lose their independence (Hobbs and Young 1999; Schulze et al.

2006). No empirical estimation of such welfare impact is found, but it is reflected in the

farmers’ attitudes in Germany and the United States.

A large number of studies report a positive income effect from CF in developing

countries, where governments and international non-government organizations (NGOs)

pay more attention. These include studies on Kenya (Wainaina et al. 2012), India (Singh

2002; Tripathi 2005; Ramaswami et al. 2006; Kalamkar 2012), Senegal (Warning and

Key 2002), Lao (Leung et al. 2008), Madagascar (Bellemare 2012), Nicaragua

(Michelson 2013), and China (Zhu 2007; Miyata et al. 2009; Xu and Wang 2009).

The increase in farmer income from CF comes from several sources. The primary

source is farmer access to the market. In many developing countries, farms are small and

farmers lack education, technology, and financial resources. Their agricultural products

can only be used by themselves or sold locally at low prices. CF provides the opportunity

to produce and sell higher valued commodities, or the same commodities but at higher

quality (Masakure and Henson 2005; Simmons et al. 2005; Bijman 2008). The second

13
source is farmer access to better technology and inputs provided by the contracting firms,

which upgrades their productivity (Gulati et al. 2007; Leung et al. 2008; Miyata et al.

2009). The third is that farmers receive other supports – such as loans and insurance –

from financial institutes, government agencies, and the NGOs when contracted (Zhu and

Wang 2007; Bijman 2008; Michelson 2013).

However, in areas where the market is more developed, the income effect may not

be that significant. For example, Wang et al. (2011) found in China where there exists an

open market, the contract price is set at the market level and contracted farmers do not

necessarily have higher profits than non-contracted farmers. The observed income effect

can also be heterogeneous, not uniformly positive (Ito et al. 2012); they find it is effective

only for small farms in China. Kalamkar (2012) observed that net return for non-

contracted farmers is higher than those contracted in India. A study in South Africa by

Freguin-Gresh et al. (2012) also found that the often identified income effect may be

misleading if not taking into account endogeneity in participation (see our discussion

below). Interestingly, these examples are all for three of the five major emerging

economies. Hence, it is possible that CF is effective at improving farmers’ income in real

primitive rural areas.

Other than the income effect for farmers, market risk and transaction cost

reductions are also welfare benefits to farmers as well as for firms (Wang et al. 2011).

Furthermore, there are social benefits, such as empowering women in traditional culture

(Raynolds 2002).

Productivity and Efficiency Impact on the Agricultural Industry

14
In addition to financial and social benefits, the impact of CF on productivity is also

important, because it increases total welfare instead of just redistributing it among

different groups. Morrison Paul et al. (2004) and Key and McBride (2007) find a positive

impact of CF on productivity in the U.S. In developing countries, the agricultural sector

tends to be laidback and slow in receiving technology transfers from developed countries.

Advanced farming technology such as mechanization is often based on large scale

farming; these techniques are difficult to apply on small scale, fragmented operations in

developing countries. CF helps facilitate technology transfers and improves productivity.

Often, the processor has technical standards for the crop or livestock growers. To

help growers achieve this standard, the processor may provide inputs such as seeds and

chemicals for crop growers, and baby animals, feed, and veterinary assistance for

livestock growers. They also provide technical training and other consulting services as

part of the contract. Such production efficiency improvements are found among poultry

growers in India (Ramaswami et al. 2006), Indonesia (Simmons et al. 2005) and

Bangladesh (Begum et al. 2012), and dairy growers in India (Birthal et al. 2005). Studies

also find crop yields increased for potato contractors in India (Tripathi et al. 2005), new

technology was adopted in Lao (Leung et al. 2008), and costs decreased but output

increased in China (Zhu 2007). Among all thirteen studies we found addressing

production efficiency, the results consistently show a positive contribution of CF.

On top of technical efficiency, financial constraints have always prevented farms

from gaining higher economic efficiency. This is especially true for small and poor farms

without credit or collateral to obtain financing in developing countries. CF can help

15
farmers receive credit from financial institutions, and in-kind credit such as seeds,

fertilizers, and other inputs directly from the firms (Simmons et al. 2005; Ma et al. 2011).

Beyond the farm gate, CF also improves the efficiency for downstream links of

the supply chain. The agribusiness firms can now have a reliable supply of raw materials

of their required variety and quality with less price uncertainty (Allen and Lueck 1995;

Ma et al. 2011). The transaction cost is reduced through contracting (Gray and Boehlje

2005; Liu et al. 2009), and CF also protects the industry investment in research and

development (Hobbs and Young 1999).

The supply chain efficiency gain can further trickle down to the consumers. The

fragmented small farm operation with heterogeneous commodity quality is the main

reason behind food safety problems. CF can cope with this fragmentation problem. For

example, in China where farmland is not allowed to be sold and farm size is very small,

CF enables farmers to pool their land and animals together. The food traceability

resulting from CF can bring benefits to end consumers (Wang et al. 2013; Yu et al. 2013).

Farmer cooperatives can play an important role in CF, helping small farmers to

gain more bargaining power in negotiations with large firms. This further reduces

transaction costs for the firms to deal with individual farms, and improves the contract

compliance rate in countries where contract violations are hard to prosecute (Guo et al.

2005; Guo and Jiang 2007; Ma and Xu 2008).

Endogeneity and Self-Selection into Contracts

There is concern that using the CF participation indicator as a binary independent

variable may lead to an endogeneity problem in the impact regression, because farmers

16
who choose to participate may have intrinsic characteristics that lead to higher welfare.

To deal with this issue, the two stage regression with instrumental variables is adopted by

Key and McBride (2007) and Miyata et al. (2009); a multi-step Heckman-type selection

correction model is used by Freguin-Gresh et al. (2012); and propensity score matching

techniques are deployed by Ito et al. (2012). Bellemare (2012) uses a contingent

valuation survey on farmers’ willingness to pay for the contract instead of revealed

participation. All papers continue to find a positive effect of CF on farmer welfare.

The self-selection arguments presented by these studies are compelling, and

provide ample warning that future empirical research should not neglect the potential bias

arising from the inclusion of a binary indicator for contract participation into a farmer

welfare regression without careful thought, even though the empirical findings may not

revert the direction of the results.

Contract Farming in China

So far, we have explored recent empirical research on CF across both developed and

developing countries. We now focus explicitly on CF in China and the potential lessons

therein. Our focus on China as a special case arises because China is a developing

economy, and has a large agricultural market as well as strong government regulatory

presence. Furthermore, the empirical literature on CF focusing exclusively on China is

relatively large compared to that on other specific developing countries.

Because of its large population and land policy 3 , the size of Chinese farms is

extremely small, with an average of 0.3 acres (Wang 2013). Despite the recent rapid

increase of Chinese income, rural income falls far behind urban income. In 2012, average

17
per capita rural income is about $1,277, only a third of its urban counterpart at $3,542

(NBSC 2013). The Chinese government has encouraged CF as a way of raising farmers’

income by growing higher-valued agricultural products. Just like the Japanese One

Village One Product Movement (OVOP) (Fujita 2007), the Chinese government supports

village level governments to invest directly or make policies to attract private investors to

invest in processing enterprises. The firms are called “dragon heads”, who will process

and market the branded product, and the governments will organize villagers to contract

so that a significant acreage in a village, sometimes the whole village, is used to produce

the particular product. Successful cases have been published extensively. For example,

Yin and Jin (2007) report that 39 percent of the planted crop acreage in Hubei province is

under the OVOP system, as was 62 percent of the aquaculture acreage.

From our review, we have identified several important recurring themes with

regards to empirical findings in China. First, heterogeneity appears to play an important

role in various facets of CF in China. Second, the Chinese CF experience has provided

opportunity for empirical assessment on effects from both government support and

previous CF experience on farmers’ participation. Third, there exist a lot of problems in

contract compliance.

While many of the CF participation studies for China reach consistent conclusions

as in other countries, Miyata et al. (2009) and Ito et al. (2012) both find empirical

evidence of nonlinearities in the effect of farmer demographics on the participation

decision. Ito et al. (2012) also find that CF may only have a positive effect on farmer

income for a certain subset of farmers. Our view is that these findings of heterogeneity

18
should not be taken lightly, from either an econometric perspective or a policy

perspective. Specifically, evidence of heterogeneity implies econometric misspecification

by assuming homogeneity, which potentially leads to a bias in econometric estimates and

to erroneous policy recommendations. The policy perspective is that, especially in China,

policymakers should use care to ensure that CF is encouraged only in areas in which

there is specific empirical evidence that CF can increase farmer welfare, or encourage

contracting with the right demographic sets of farmers.

As we mentioned previously, but reiterate here given its importance for CF in

China, Guo (2005) and Zhu and Wang (2007) study both the importance of previous CF

experience and government support on the contract participation. The Zhu and Wang

(2007) finding that previous experience in CF leads to a higher probability of contracting

indicates that the initial experience was positive. This is implicit evidence that CF has

been successful at improving farmer welfare – at least in China – and is consistent with

explicit econometric results of, for example, Miyata et al. (2009). This finding is also

consistent with the Wang et al. (2011) finding that more risk averse farmers were less

likely to participate in a contract. Since CF in China is relatively new, previous

experience in contracting is valuable for future contracting participation.

Although most of the publications on government supported CF in the form of

OVOP are descriptive, the Guo (2005) and Zhu and Wang’s (2007) result that the

presence of government support for contracting increases the likelihood a farmer chooses

to contract is particularly encouraging. A direct implication of this research is that

governments would be well advised to increase CF participation. It is clear that, at least

19
from the Chinese experience, that government policies aimed at increasing CF

participation have been successful.

A low contract compliance rate has been troublesome in developing countries

(Carney 1988; Reardon and Barrett 2000; Narayanan 2010), especially in China (Lei

2004; Fang and Wang 2013). Because the market prices of agricultural products have

increased rapidly with Chinese income growth, farmers tend to sell their products outside

the contract at a higher price. With a large number of small farms, the contract

enforcement cost is high, allowing the contract breaching rate to reach 80 percent in some

cases (Liu 2002). Guo (2006), Guo and Jiang (2007), Guo and Jolly (2008) and Wang et

al. (2011) find the contract design with a flexible upside price and quality related price

premium scheme may help facilitate contract compliance. Hence, these authors suggest

that there should be a benefit sharing mechanism built into the contract, so that farmers

perceive that the contract is fair. Ma and Xu (2008) find that cooperatives between

farmers and the processing firms may also help with contract compliance.

Meta-Analysis

Our review thus far has focused broadly on different facets of CF across both developed

and developing countries, and across different commodities. In some cases, we find a

strong consensus in the empirical literature, yet in other cases, we have found mixed

empirical evidence. In the latter cases, given the scope of empirical research surveyed, it

is often difficult to identify common patterns across studies that report similar

conclusions. To assist in these efforts, we conduct a meta-analysis along several different

20
dimensions of CF in order to identify factors correlated with different econometric

outcomes. Meta-analysis has a long history in applied economic research as providing a

statistical means of synthesizing established bodies of research, and as such providing an

important complement to our existing review. See, for example, Stanley (2001) for a

basic overview of meta-analyses in applied economic research.

Our focus in this study captures elements of both the participation equations and

the farmer welfare equations. We focus our statistical analyses on variables in the

literature whose effects on the outcome are ambiguous across studies. For example, in the

papers surveyed on contract participation, farm size is found to make a positive, negative,

or insignificant contribution across different studies. One aspect of our meta-analysis is to

identify characteristics across studies that are associated with each empirical outcome.

Data

One key aspect of conducting a meta-analysis is the selection of the relevant literature.

To maximize our sample of observations as well as ensure that our sample is not biased

by the exclusion of any relevant work, we have included studies spanning developed and

developing economies, different commodities, published and unpublished, and using both

Chinese and non-Chinese data, given our interest in CF in China.

The dependent variables we consider are measures of the impact of farmer age,

farmer education, farm size, and farm specialization on the probability of contract

participation, as well as the effect of CF on farmer income. More specifically, we

construct binary indicators for whether or not each study reported the a priori expected

result for each regressor. For example, economists expect farmer age to be negatively

21
correlated with contracts participation because younger farmers are more likely to

contract all else being equal. Hence, we define an indicator that takes a value of unity for

each study in our analysis that reports a negative age coefficient in the participation

regression. All other reported results, negative or insignificant, are classified as zero. We

expect contract participation to be positively correlated with farmer education, farm size,

and farmer specialization, and define these binary dependent variables accordingly. We

also expect that CF increases farmer income, so we define our income dependent variable

as a binary indicator that takes a value of unity for any study that reports a positive and

significant coefficient on the contract variable in its farmer income regression.

The independent variables we consider include the age of the data used by each

study, an indicator for whether or not the study is published, the sample size used in the

regression, an indicator for whether or not the study focuses on a developing country, an

indicator for whether or not the study is focused on China (i.e., uses Chinese data), and an

indicator for whether the commodity of interest is an animal protein. 4 Our goal is to

explore qualitative differences in the results across studies that may be attributed to

differences in data used by each study, as well as systematic differences across

developing or developed countries, or China.

Table 1 contains the definition and descriptive statistics for both the independent

and dependent variables used in this meta-analysis. Roughly one third of the papers in

our meta-analysis identify a significantly negative effect of farmer age, and a

significantly positive effect of farmer education and farm specialization on contract

participation. Sixty percent of the studies find that farm size significantly increases the

22
probability of contracting, and nearly sixty percent identify a significantly positive effect

of contracting on farmer income. In terms of the independent variables, we find the

average age of the data used in the literature is 8.73 years old (years between the most

recent data used in the study and the year 2013), with the most recent study being 1 year

old and the oldest 19 years. Fifty-four percent of these studies are published, twelve

percent of which use Chinese data. The sample size ranges from 50 to 4707 observations

with an average of 986. Sixty-nine percent of these studies focus on developing countries,

and a quarter focus on animal commodities, such as poultry, livestock, dairy, and fish.5

The sample size for each meta-analysis regression varies, with largest sample size

being 23 studies and the smallest 17. Since our dependent variables are binary, the ideal

model is a logit regression. Given the small sample size and the data demanding

maximum likelihood estimation of a logit regression, we also include ordinary least

squares results from a linear probability model (LPM) to benchmark the behavior of our

logit estimates.6 We find that our estimates are consistent across both estimators.

Results

Table 2 contains the results of our five meta-regressions using both least squares (LPM

columns) and maximum likelihood (Logit columns). For the LPM models, we report the

estimated coefficients and their standard errors; for the logit regressions, we report the

average marginal effect and its standard error obtained by using the delta method.

In our first model, we regress the farmer age indicator on six independent

variables for a meta-sample of 22 empirical studies. We find that the age of the data is

significantly negative in the logit model, and across both LPM and logit regressions, the

23
coefficient on the size of the sample is significantly positive. These coefficients indicate

that older datasets are less likely to identify the negative effect of farmer age on contract

participation by about 5 percent, while studies using larger samples are more likely to

find such an effect by about 2 percent. The effect of farmer age on the probability of

contract participation is not significantly linked to whether or not the paper is published,

whether or not the paper uses Chinese data, or the type of commodity. The LPM

identifies an insignificant effect of the developing country indicator, while is excluded

from the logit due to non-convergence.

Model 2 is for the education equation for a sample of 23 meta-studies. Both the

LPM and logit regressions have similar coefficients of determination, with only the

sample size variable being statistically significant. These results indicate that studies that

use larger samples are more likely to find a significantly positive effect of farmer

education on contract adoption, by about 2.5 to 3 percent. The insignificance of our other

independent variables indicates that the effect of farmer education on the probability of

contracting is not systematically related to any of these variables.

The farm size indicator meta-regression has a sample of 20 studies. The LPM

model finds that sample size is the only factor that is significantly related to the empirical

finding that larger farms tend to contract, while it is insignificant in the logit regression.

In the logit regression, we find that the age of the data is significantly negative. No other

independent variables are found related to the variation across empirical studies regarding

the impact of farm size on contract participation. Again, the coefficients of determination

are similar across models.

24
 The final meta-study regarding contract participation is a regression of the

specialization indicator on a set of six independent variables with 17 observations. We

omit the animal protein indicator since including this indicator renders the likelihood

function ill-posed in this specification. We find in the LPM regression that studies

focused on animal protein commodities are significantly less likely to find a significant

specialization effect, of about 60 percent. This is the only regression for which this

variable is statistically significant; we do not find that sample size is a significant factor

explaining these empirical results. We do not find any statistical significance in the logit

regression, indicating that the studies that have reported a significantly positive effect of

farmer specialization on the contract participation decision do not vary systematically

along the dimensions we consider here.

The last meta-regression examines study characteristics that affect the discovered

income effect of CF. The sample contains 17 meta-studies. We drop the developing

country indicator from both regressions since it renders our likelihood function ill-posed

in the logit case, and both regression estimators return marginal effects larger than unity

(indicating probability changes greater than unity). We find that Chinese studies are

significantly less likely to identify a statistically significant income effect of CF in the

logit model, by nearly 40 percent. We do not find that any of the other factors considered

are significantly related to the income effect in empirical CF studies.

Discussion

Looking across each of the five meta-regressions we conduct, it is clear that larger

samples are often significantly related to the recovery of an empirical result that is both

25
consistent with prior expectations in sign, as well as statistical significance. We also find

some evidence that the age of the data may be an important driver of some of the

conflicting empirical results identified in our detailed review. For the age and size

impacts on contract participation, the data age show a significant and negative effect,

which means analysis using more recent data support our prior expectation. However, the

data age factor is insignificant for the income equation, which indicates that studies using

more recent data are not more likely to find a significant income effect.

It is interesting to notice the insignificance of the other regressors in most other

models. In particular, for the models that include the developing country indicator, we do

not find any statistical significance of this indicator. At least, this finding clearly implies

that conflicting results in empirical studies do not conflict simply because of differences

across developed and developing countries. This is consistent with our review of the

empirical literature that indicated that there may be important differences within

developing countries, and that a simple developed versus developing country argument

does not suffice in explaining these differences. Yet, everything else the same (for the

independent variables we use as controls), we might interpret the insignificance of the

developing country indicator as evidence of intrinsic similarities in both types of

countries with respect to CF. It is also clear that animal based commodities are not

significantly different from other (plant based) commodities.

It is interesting to note that differences in empirical studies are not driven solely

because of differences between Chinese and non-Chinese data, or by whether or not the

research is published. The latter is reassuring because it indicates that there is not a

26
significant publication bias that sometimes arises in empirical research. The former result

is interesting because it indicates that China may not be systematically different from

other countries, at least in terms of CF. This further buttresses our consideration of China

as a representative developing country in terms of CF.

The lesson learned from our meta-analysis of the existing empirical literature is

clear: factors contributing to farmer contract participation are not very different across

developed versus developing countries or commodities. More recent studies and those

using larger samples tend to give us results consistent with conventional knowledge.

There is still ample room for future empirical research to focus more heavily on particular

countries or commodities to further our collective understanding of these facets.

Conclusion

Recent years have seen movement of contract farming to developing countries, followed

by a large body of empirical research aimed at quantifying many effects of CF on farmer

welfare. In many studies, explicit attention is given to the understanding of why some

farmers choose to contract, while others do not. This body of research has focused widely

across both developed and developing countries, as well as commodities.

In some instances, the empirical literature has drawn clear conclusions. For

example, most contract participation studies find that neither the number of laborers

employed on the farm nor farm assets contribute effectively to contracting, but that

government policies that encourage contracting are effective drivers of contract adoption.

Our review has also shown that farmer selection into contracts is likely non-random.

27
These clear conclusions are important because they indicate stability across countries and

commodities of CF along these important dimensions.

Another consensus among countries and commodities is that CF has a significant

effect on improving farm efficiency as well as the efficiency of the supply chain. While

segmented small operations are pooled together when contracting with a large firm, they

can use larger equipment, adopt state of the art technology in production, reduce

transaction cost in the supply chain, and make food traceable. This indicates that CF can

serve as a vehicle to modernize small farm based agriculture in developing countries.

In other cases, researchers have been unable to draw clear conclusions across

either countries or commodities. As we have described, the reason for this lack of

consensus on certain facets of CF may stem from heterogeneity across countries and/or

commodities. For instance, some cultural or institutional factors may be unique across

countries; similarly, these same factors may have heterogeneous impacts on different

commodities within a particular country. In China, for example, the literature is clear that

risk averse farmers are less likely to contract because of general unfamiliarity of Chinese

farmers with contracting. This finding, however, need not be true outside of China.

Further evidence of heterogeneity comes from studies that have reported statistically

significant nonlinearities with respect to the effects of education and farmer experience

on contract adoption, as well as the effect of contracting on farmer income.

To address some of these issues, we develop a brief meta-analysis of the existing

research as a means of identifying systematic differences across studies that may drive

differences in empirical findings. We find that the age of the regression sample and the

28
size of the dataset are significantly correlated with the identification of statistically

significant empirical results that are consistent with a priori expectations. It is also clear

from our meta-regressions that the empirical differences culled from the existing

literature are not easily explained by simple comparisons of developed and developing

countries, or plant and animal commodities. The insignificance of these indicators

confirms our intuition culled from the available empirical research, and further

strengthens this intuition given our control of other factors.

Through this review and statistical analysis of the existing literature, several

important insights have been made clear. First, CF is largely successful in improving

farmer welfare. From a developmental perspective, this bears important implications

because policymakers can be reasonably assured that investment of resources into

developing CF is a fruitful policy venture. Second, our review has described which

aspects of farmer contract participation are generally significant, and which are somewhat

contentious. Third, our meta-study has made clear that with the exception of larger, more

recent samples, there does not appear to be systematic differences across

developing/developed countries or commodities.

29
References
Allen, D. W. and D. Lueck, (1995). “Risk preferences and the economics of contracts,”

American Economic Review, 85(2): 447-451.

Arumugam, N., F. M. Arshad, E. Chiew and Z. Mohamed, (2011). “Determinants of

fresh fruits and vegetables (FFV) farmers’ participation in contract farming in

peninsular Malaysia,” International Journal of Agricultural Management and

Development, 1(2): 65-71.

Barrett, C. B., M. E. Bachke, M. F. Bellemare, H. C. Michelson, S. Narayanan and T. F.

Walker, (2012). “Smallholder participation in contract farming: comparative

evidence from five countries,” World Development, 40(4): 715-730.

Begum, I. A., M. J. Alam, J. Buysse, A. Frija and G. Van Huylenbroeck, (2012).

“Contract farmer and poultry farm efficiency in Bangladesh: a data envelopment

analysis,” Applied Economics, 44(28): 3737-3747.

Bellemare, M. F., (2012). “As you sow, so shall you reap: the welfare impacts of contract

farming,” World Development, 40(7): 1418-1434.

Bijman, J., (2008). “Contract farming in developing countries: an overview,” working

paper, Department of Business Administration, Wageningen University,

Wageningen, Holland.

Birthal, P. S., P. K. Joshi and A. Gulati, (2005). “Vertical coordination in high value

commodities: implications for the smallholders,” MTID Discussion Paper 85,

Washington D.C., IFPRI.

30
Carney, J. A., (1988). “Struggles over crop rights and labour within contract farming

households in a Gambian irrigated rice project,” Journal of Peasant Studies, 15(3):

334-349.

Coase, R., (1937). “The nature of the firm,” Economica, 4(16): 386-405.

Fang, X. and H. H. Wang, (2013). “The three-party game of contract farming: low

contract compliance rates in grain transactions in China,” China: An International

Journal, 11(3):1-13.

FAO (2013). “FAQ: What is contract farming?” Contract Farming Resource Centre,

Food and Agriculture Organization of the United Nations.

http://www.fao.org/ag/ags/contract-farming/faq/en/, viewed on September 20,

2013.

Freguin-Gresh, S., M. d’Haese and W. Anseeuw, (2012). “Demythifying contract

farming: evidence from rural South Africa,” Agrekon, 51(3): 24-51.

Fujita, M., (2007). “Spurring economic development by capitalizing on brand agriculture:

turning development strategy on its head,” In Bourguignon, F. and B. Pleskovic

eds. Rethinking Infrastructure for Development, Washington, World Bank, pp.

205–229.

Glover, D. J., (1984). “Contract farming and smallholder outgrower schemes in less-

developed countries,” World Development, 12(11/12): 1143-1157.

Goodhue, R. E., (2000). “Broiler production contracts as a multi-agent problem: common

risk, incentives and heterogeneity,” American Journal of Agricultural Economics,

82(3): 606-622.

31
Gray, A. W. and M. D. Boehlje, (2005). “Risk sharing and transactions costs in producer

processor supply chains,” Choices: The magazine of food, farm, and resource

issues, 4th Quarter, 20(4): 281-286.

Gulati, A., N. Minot, C. Delgado, and S. Bora, (2007). “Growth in high-value agriculture

in Asia and the emergence of vertical links with farmers,” Global Supply Chains,

Standards and the Poor: How the Globalization of Food Systems and Standards

Affects Rural Development and Poverty. Eds by J. F. M. Swinnen, CABI,

Wallingford, UK (2007): 91-108, Chapter 8.

Guo, H., (2005). “An analysis of the influencing factors of Chinese farmers’ participation

in contract farming,” Chinese Rural Economy, 2005(3): 24-32. (In Chinese).

Guo, H., (2006). “Dragon-head companies and enforcement of contract-farming between

the agribusiness firms and farmers: theory and practice in Zhejiang province,”

Issues in Agricultural Economy, 2006(2): 36-42. (In Chinese).

Guo, H., and W. Jiang, (2007). “Practice and implications of the contract farming model

of association + firm + cooperative + farms,” Chinese Rural Economy, 2007(4):

48-52. (In Chinese).

Guo, H., and R. W. Jolly, (2008). “Contractual arrangements and enforcement in

transition agriculture: theory and evidence from China,” Food Policy, 33(6): 570-

575.

Guo, H., R. W. Jolly and J. Zhu, (2005). “Contract farming in China: supply chain or ball

and chain,” Proceeding of the Minnesota International Economic Development

32
 Conference (MIDC05), University of Minnesota, Minneapolis, MN. April 29-30,

2005.

Hennessy, D. A., (1996). “Information asymmetry as a reason for food industry vertical

integration,” American Journal of Agricultural Economics, 78(4): 1034-1043.

Hennessy, D. A. and J. D. Lawrence, (1999). “Contractual relations, control, and quality

in the hog sector,” Review of Agricultural Economics, 21(1): 52-67.

Hobbs, J. E. and L. M. Young, (1999). “Increasing vertical linkages in agrifood supply

chains: a conceptual model and some preliminary evidence,” Research discussion

paper 35.

Hu, W., (2012). “Effect of contract farming on the farmers’ average return: the case of

the grain industry in the USA,” Paper presented at the Agricultural and Applied

Economics Association 2012 Annual Meeting, August 12-14, 2012, Seattle,

Washington. No. 124659.

Ito, J., Z. Bao and Q. Sun, (2012). “Distributional effects of agricultural cooperatives in

China: exclusion of smallholders and potential gains on participation,” Food

Policy, 37(6): 700-709.

Kalamkar, S. S., (2012). “Inputs and services delivery system under contract farming: a

case of broiler farming,” Agricultural Economics Research Review, 25, 515-521.

Katchova, A. I. and M. J. Miranda, (2004). “Two-step econometric estimation of farm

characteristics affecting marketing contract decisions,” American Journal of

Agricultural Economics, 86(1): 88-102.

33
Key, N., (2013). “Production contracts and farm business growth and survival,” Journal

of Agricultural and Applied Economics, 45(2): 277–293.

Key, J. and D. Runsten, (1999). “Contract farming, smallholders, and rural development

in Latin America: the organization of agroprocessing firms and the scale of

outgrower production,” World Development, 27(2): 381-401.

Key, N. and W. D. McBride, (2007). “Production contracts and farm productivity:

examining the link using instrumental variables,” working paper, prepared for

presentation at the annual 2007 AAEA meeting.

Laffont, J. J. and J. Tirole, (1986). “Using cost observation to regulate firms,” Journal of

Political Economy, 94(3): 614-641.

Leathers, H. D., (1999). “What is farming? Information, contracts, and the organization

of agricultural production: discussion,” American Journal of Agricultural

Economics, 81(3): 621-623.

Lei, H., (2004). “An analysis of default in contract farming,” Financial and Economic

Sciences, 2004(3): 76-77.

Leung, P., S. Sethboonsarng and A. Stefan, (2008). “Rice contract farming in Lao PDR:

moving from subsistence to commercial agriculture,” No. 90, ADB Institute

Discussion Paper.

Liu, F., (2002). “Incomplete contract and obstacles to contract compliance: the case of

contract farming,” Economic Research Journal, 91(4): 22-30.

Liu, W., B Liu, and Z. Ren, (2009). “Cost control methods on the green agricultural

product supply chain under order2farm agriculture model” Journal of

34
 Southwestern Jiaotong University: the Social Science Edition, 10(4): 114-119. (In

Chinese).

Lu, K. and J. Ma, (2010). “Empirical studies on the choice and influencing factors of

farmers’ participation in contract farming,” Journal of Agrotechnical Economics,

Issue 9. (In Chinese).

Ma, J., and X. Xu, (2008). “An analysis of market structure and the contract compliance

in contract farming,” Issues in Agricultural Economy, 2008(3): 35-41. (In

Chinese).

Ma, J., Y. Zhang, and C. Yu, (2011). “Innovative strategy and case study for the contract

farming based value chain,” Rural Financial Studies, 2011(7): 11-17.

Martin, L. L., (1997). “Production contracts, risk shifting, and relative performance

payments in the pork industry,” Journal of Agricultural and Applied Economics,

29(2): 267-278.

Masakure, O. and S. Henson, (2005). “Why do small scale producers choose to produce

under contract? Lessons from nontraditional vegetable exports from Zimbabwe,”

World Development, 33(10): 1721-1733.

Michelson, H. C., (2013). “Small farmers, NGOs, and a Walmart world: welfare effects

of supermarkets operating in Nicaragua,” American Journal of Agricultural

Economics, 95(3): 628-649.

Miyata, S., N. Minot and D. Hu, (2009). “Impact of contract farming on income: linking

small farmers, packers, and supermarkets in China,” World Development, 37(11):

1781-1790.

35
Morrison Paul, C. J., R. Nehring and D. Banker, (2004). “Productivity, economies, and

efficiency in U.S. agriculture: a look at contracts,” American Journal of

Agricultural Economics, 86(5): 1308-1314.

Morrissy, J. D., (1974). Agricultural Modernization Through Production Contracting.

Praeger, New York.

Narayanan, S., (2010). “Relationship farming: the problem of enforcement in contract

farming arrangements in India,” Working paper, Cornell University, 2010.

NBSC (2013). “China's economy achieved a stabilized and accelerated development in

the year of 2012,” National Bureau of Statistics of China, January 2013.

http://www.stats.gov.cn/english/pressrelease/t20130118_402867147.htm

Ramaswami, B., P. S. Birthal and P. K. Joshi, (2006). “Efficiency and distribution in

contract farming: the case study of Indian poultry growers,” MTID Discussion

Paper 91.

Raynolds, L. T., (2002). “Wages for wives: renegotiating gender and production relations

in contract farming in the Dominican Republic,” World Development, 30(5): 783–

798.

Reardon, T. and C. B. Barrett, (2000). “Agroindustrialization, globalization, and

international development: an overview of issues, patterns, and determinants,”

Agricultural Economics, 23(3): 195-205.

Schrader, L., (1986). “Responses to forces shaping agricultural marketing: contracting,”

American Journal of Agricultural Economics, 68(5): 1161-1166.

36
Schultz, B., A. Spiller and L. Theuvsen, (2006). “Is more vertical integration the future of

food supply chains? Empirical evidence and theoretical considerations from

German pork production,” International Agri-Food Chains and Networks:

Management and Organization, Wageningen, 49-63.

Simmons, P., P. Winters and I. Patrick, (2005). “An analysis of contract farming in East

Java, Bali, and Lombok, Indonesia,” Agricultural Economics, 33, 513-525.

Singh, S., (2002). “Contracting out solutions: political economy of contract farming in

the Indian Punjab,” World Development, 30(9): 1621-1638.

Stanley, T. D., (2001). “Wheat from chaff: meta-analysis as quantitative literature review,”

Journal of Economic Perspectives, 15(3): 131-150.

Tripathi, R. S., R. Singh and S. Singh, (2005). “Contract farming in potato production: an

alternative for managing risk and uncertainty,” Agricultural Economics Research

Review 18, 47-60.

Wainaina, P. W., J. J. Okello and J. Nzuma, (2012). “Impact of contract farming on

smallholder poultry farmers’ income in Kenya,” Contributed paper presented at

the 2012 Triennial Conference of International Association of Agricultural

Economists, Foz do Iguacu, Brazil, August 18-24, 2012. No. 126196.

Wang, H. H., (2013). “Agricultural risks and risk management in the current context of

Chinese economy,” Agricultural Finance Review, 73(2): 245-254.

Wang, H. H., Y. Zhang and L. Wu, (2011). “Is contract farming a risk management

instrument for Chinese farmers?” China Agricultural Economic Review, 3(4):

489-504.

37
Wang, H. H., H. Yu and B. Li, (2013). “Is dairy complex a solution to food safety for raw

milk? A comparison of farmers’ perceived and realized safety effects,” CAER-

IFPRI Annual International Conference on Transforming China’s Agricultural

and Rural Sector: Challenges and Solutions, Wuhan, China, October 17-18, 2013.

Warning, M. and N. Key, (2002). “The social performance and distributional

consequences of contract farming: an equilibrium analysis of the Arachide de

Bouche program in Senegal,” World Development, 30(2): 255-263.

Williamson, O. E., (1971). “The vertical integration of production: market failure

considerations,” American Economic Review, 61(2): 112-123.

Wu, S. Y., (2006). “Contract theory and agricultural policy analysis: a discussion and

survey of recent developments,” Australian Journal of Agricultural and Resource

Economics, 50: 490-509.

Xu, J. and X. Wang, (2009). “An empirical analysis of the impact of contract farming and

its organization models on farmers’ income,” Chinese Rural Economy, 2009(4):

39-47. (In Chinese).

Yin, L. and L. Jin, (2007). “A current situation analysis of the ‘one village one product’

model for agricultural development in Hubei province,” Rural Economy, Science,

and Technology, 2007(7): 72-73.

Yu, H., H. H. Wang and B. Li. (2013) “What can we do to ensure food safety of raw milk

in China?” Organized Session Paper, Agricultural and Applied Economics

Association Annual Meetings, Washington, DC, August 2013.

38
Zhu, H. (2007). “An assessment of the effects of adopting contract farming structure in

the tomato industry in Xinjiang,” Journal of Agrotechnical Economics, 3, 89-95.

(In Chinese).

Zhu, H. and X. Wang, (2007). “An analysis on the influencing factors of tomato growers’

participation in contract farming in Xinjiang Autonomous Region,” Chinese Rural

Economy, 7, 67-75. (In Chinese).

39
 Table 1: Descriptive Statistics

Description Mean Std Dev Min Max

Binary Dependent Variables

Age Indicator, =1 if age negative/significant 0.36 0.49 0 1

Education Indicator, =1 if education positive/significant 0.30 0.47 0 1

Farm Size Indicator, =1 if farm size is positive/significant 0.60 0.50 0 1

Specialization Indicator, =1 if specialization

positive/significant 0.29 0.47 0 1

Income Indicator, =1 if contracting increased farmer

income 0.59 0.51 0 1

Independent Variables

Data Age Age in years of most recent data used in study 8.73 4.18 1 19

Published Indicator, =1 if the article is published in a

journal 0.54 0.51 0 1

Sample Size Sample size used in regression 986 1387 50 4707

Developing Indicator, =1 for developing country study

Country 0.69 0.47 0 1

China Data Indicator, =1 for data collected from China 0.12 0.33 0 1

Animal Protein Indicator,=1 if the commodity is fish, poultry,

livestock or dairy 0.23 0.43 0 1

40
 Table 2: Impact of Study Characteristics on Contract Farming Empirics: Results from Meta-Analysis Regressions

Age Education Farm Size Specialization Income

Variable LPM Logit LPM Logit LPM Logit LPM Logit LPM Logit

Data Age -0.017 -0.054* -0.011 -0.004 -0.062 -0.061+ 0.043 0.032 0.027 0.033

(0.034) (0.029) (0.038) (0.031) (0.047) (0.035) (0.035) (0.029) (0.034) (0.030)

Published -0.239 0.109 0.197 0.247 0.309 0.302 -0.169 0.090 0.112 0.161

(0.311) (0.294) (0.268) (0.220) (0.335) (0.257) (0.332) (0.229) (0.344) (0.264)

Sample Size 0.027** 0.020** 0.024** 0.032* 0.017+ 0.026 -0.005 0.003 0.002 0.001

(0.010) (0.008) (0.011) (0.018) (0.011) (0.024) (0.012) (0.011) (0.011) (0.009)

Animal Protein -0.032 0.098 0.156 0.242 -0.334 -0.262 -0.630+ -0.127 -0.137

(0.300) (0.337) (0.317) (0.272) (0.286) (0.197) (0.396) (0.375) (0.267)

Developing 0.514 0.125 0.469 0.073 0.181 0.317 0.227

(0.355) (0.346) (0.505) (0.416) (0.589) (0.385) (0.363)

China Data 0.226 0.155 -0.103 -0.025 -0.144 -0.180 -0.215 0.003 -0.384 -0.388+

(0.276) (0.201) (0.270) (0.223) (0.292) (0.225) (0.385) (0.305) (0.348) (0.259)

Constant -0.005 -0.056 0.837 0.040 0.428

 Table 2: Impact of Study Characteristics on Contract Farming Empirics: Results from Meta-Analysis Regressions

Age Education Farm Size Specialization Income

Variable LPM Logit LPM Logit LPM Logit LPM Logit LPM Logit

(0.464) (0.540) (0.589) (0.505) (0.472)

n 22 22 23 23 20 20 17 17 17 17

R2 0.436 0.349 0.344 0.325 0.436 0.421 0.314 0.118 0.272 0.237

Table reports marginal effects and standard errors for least squares and logit regression models. Logit marginal effects are average

marginal effects, and standard errors are the standard errors of the marginal effect obtained via the Delta method. Statistical

significance at the 1%, 5%, 10%, and 15% levels are denoted by ***, **, *, + respectively. Further regression output details are

available upon request.

1
Although cash forward contracts for a few major agricultural commodities have existed for a long time in developed countries, they

are not the same as contract farming because the latter tend to be for specialty crops or commodities with special quality features that

do not have a liquid open market.

2
Some contracts also provide farmers with a means of obtaining loans. This provides further motive for credit constrained farmers to

participate in a contract.
3
Farm land is not owned by individual farmers, and thus the ownership cannot be transferred. Farmers have the use right.
4
We explored more detailed indicators for commodity type, separately identifying field crops, vegetables, and animal products from a

base group of other products. We found that this level of division rendered our likelihood function ill-posed in our logistic regressions,

so we report only those results from division of commodity based on animal protein.

5 We also tried to further separate the plant based commodities into field crops and horticultural crops. The results do not show any

significance and cause some convergence failure in maximum likelihood estimation.

6 We also explored defining our dependent variables across negative, positive, and insignificant separately and estimating the model as

a multinomial logit. These results are generally qualitatively consistent with the binary logit results reported here, and are hence

omitted.