Food Policy 40 (2013) 14–24

Contents lists available at SciVerse ScienceDirect

Food Policy

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

Contract farming configuration: Smallholders’ preferences for contractdesign attributes

Gumataw K. Abebe a,b,⇑, Jos Bijman a, Ron Kemp a, Onno Omta a, Admasu Tsegaye c

a Wageningen University, Management Studies, Hollandseweg 1, 6706 KN Wageningen, The Netherlandsb Hawassa University, P.O. Box 05, Hawassa, Ethiopiac Addis Ababa University, P.O. Box 1176, Addis Ababa, Ethiopia

a r t i c l e i n f o

Article history:Received 29 July 2012Received in revised form 30 December 2012Accepted 14 January 2013Available online 1 March 2013

Keywords:Contract farmingContract designConditional logitPotatoEthiopia

0306-9192/$ - see front matter � 2013 Elsevier Ltd. Ahttp://dx.doi.org/10.1016/j.foodpol.2013.01.002

⇑ Corresponding author at: Wageningen UniveHollandseweg 1, 6706 KN Wageningen, The Netherla

E-mail addresses: gumataw@yahoo.com, gumataw

a b s t r a c t

While Contract Farming (CF) can enhance smallholders’ income in developing countries, empiricalresearch on the motivation of smallholders to participate in CF is scarce. This paper explores farmer pref-erences for particular contract design attributes. We combined analytical hierarchy process and discretechoice experiments to investigate the importance of contract design attributes. On the basis of data col-lected among potato farmers in Ethiopia, we found that input market uncertainty is more important thanoutput market uncertainty in smallholders’ decision to participate. Farmers tend to minimize their risk byopting for the buyer firm above the state and NGOs as providers of seed, inputs, and technical assistance.The results imply that the success of a CF scheme depends on the willingness of the firm to incorporatethe preferred contract design attributes. Institutional intervention in the input market could induce agri-business firms to offer attractive contracts for smallholders.

� 2013 Elsevier Ltd. All rights reserved.

Introduction

Participation in global markets calls for greater integration inagrifood value chains to respond to the quality and safety require-ments of international customers. Contract Farming (CF) has beenclaimed to have a positive impact on local economies by improvingthe welfare of rural households (Barrett et al., 2012; Bellemare,2010; Bijman, 2008; Grosh, 1994; Reardon et al., 2009; Singh,2002). However, CF also remains a much debated institutionalarrangement (Key and Runsten, 1999; Little and Watts, 1994;Oya, 2012; Singh, 2002). Discussions on CF mainly revolve aroundrecurrent issues, such as the role of private-led CF schemes inaddressing market failures (Grosh, 1994) and in reducing the riskof agribusiness firms with regard to production, land expropria-tion, and labor (Herath and Weersink, 2009), and emerging issues,such as agri-food globalization, private standards, and land grab-bing (Oya, 2012). Analyses of CF often use a political economy per-spective, an institutional economics perspective, or a combinationof both.

In the political economy view, CF is seen from the lens of un-equal power relations, conflict, and labor related issues (Little,1994; Wilson, 1986). The main concern is that CF can lead farmersinto problems such as loss of autonomy, increased production risk,

ll rights reserved.

rsity, Management Studies,nds.@gmail.com (G.K. Abebe).

and indebtedness (Little and Watts, 1994; Porter and PhillipsHo-ward, 1997; Rehber, 1998; Singh, 2002).

Conversely, the institutional economics view emphasizes therole of CF in addressing market failures (e.g., Barrett, 2008; Grosh,1994; Key and Runsten, 1999; Kirsten and Sartorius, 2002; Mintenet al., 2009; Sartorius and Kirsten, 2007). More specifically, this lit-erature focuses on the micro-functioning of CF schemes, dealingwith transaction costs resulting from uncertainty, risk, marketimperfections, and coordination failures.

Empirical studies in developing countries provide varied analy-ses about participation and welfare effect of CF. Several authorsfound that participation improves farmers’ income (e.g., Barrettet al., 2012; Bellemare, 2012; Warning and Key, 2002), althoughthe extent to which participation contributes to the welfare ofsmallholders continues to be a methodological question (Barrettet al., 2012). Evidence is mixed, however, concerning inclusion.While Warning and Key (2002), in Senegal Miyata et al. (2009)and Wang et al. (2011), in China, found no evidence of exclusionof smallholders from participation, others, such as Singh (2002),in India, Guo et al. (2005), in China, and Key and Runsten (1999),in Latin America, reported the opposite. The literature also docu-ments several problems affecting CF performance: high defaultrate, biased terms, delayed payments, cheating, and lack of com-pensation for crop failure (Guo et al., 2005; Singh, 2002). Further-more, Barrett et al. (2012) reported cases of high participationturnover due to lack of commitment to honor agreements by eitherparty.

G.K. Abebe et al. / Food Policy 40 (2013) 14–24 15

A general conclusion from the literature is that CF improves in-come. Even those who are critical of CF schemes generally agreethat participation improves household income (Little, 1994; Singh,2002). Indeed, farmers will only participate in CF if there is an ex-pected gain in doing so (Bellemare, 2012). Likewise, firms willchoose CF when the expected benefits from contracting exceedthose of the alternatives, such as buying on a spot market or pro-ducing on proprietary farms.

One question the existing literature does not address is aboutfarmers’ preferences for particular contract terms and provisions.While the main motivation of smallholders to enter into CF is theresolution of market failure, a closer look at participation decisionsmay disclose how different contract provisions are evaluated.Eventually, smallholders’ contract acceptance can be improvedby better aligning contract terms and provisions with farmers’preferences (Minten et al., 2009).

We argue that contract terms and conditions, hereafter calledcontract design attributes, can affect farmers’ decisions to partici-pate in CF by varyingly affecting their expected level of utility fromparticipation. In theory, contracting parties choose a contract de-sign that provides little incentive to opportunism. However, inpractice, contracts are biased toward agribusiness firms and oftenexpose smallholders to ex post risk (Singh, 2002), because firmschoose contract design attributes that will offer them the highestpayoffs without considering farmers’ expected utility level (Barrettet al., 2012). Masakure and Henson (2005) noted that contractsinvolving smallholders are rarely governed by explicit performanceand risk-sharing incentives. Hence, the likelihood that a contractdesign is attractive to smallholders remains uncertain. For the firm,this could lead to high transaction and coordination costs due topossible side-selling, default, and underinvestment (Delpierre,2009; Miyata et al., 2009).

In reviewing the CF literature, we noted several gaps. First,although many authors discussed the importance of contract de-sign attributes, surprisingly little attention has been paid to mea-sure the relative importance of these attributes directly fromfarmers’ perspective. Our study builds on Masakure and Henson(2005), who explicitly focused on ex ante aspects of smallholder’motivation toward CF. While these authors asked farmers abouttheir motivation to enter into CF, our study goes a step furtherby using an experimental approach to elicit their preferences oncontract design attributes. For example, while the authors reportedoral contracts as the preferred contract form by the buyer firm,they did not investigate whether this option was also preferredby the farmers. Second, there is a general assumption in the liter-ature that farmers are risk averse, and that their motivation to par-ticipate in CF is primarily to manage output price risks (e.g., Chavasand Holt, 1996; Michelson et al., 2011). Subsequently, agribusinessfirms tend to design contracts with pre-fixed price, quantity, andquality specifications. However, contract design is a complex pro-cess involving many trade-offs (Bogetoft and Olesen, 2002), andfarmers may have different risk preferences for the different mar-kets in which they operate. Third, previous studies on CF heavilyfocused on the income and broader welfare effects, as well on indi-vidual-specific characteristics, as key determinants for participa-tion. Yet, the effect of different contract design attributes onsmallholders’ contract choice has not been investigated.

The main objective of the present study is to explore the relativeimportance of different contract design attributes that could differ-entially affect the motivation of smallholders to participate in CF.Better information on farmers’ preferences can be used by agri-business firms to design better contracts as well as by policy mak-ers in developing an enabling institutional environment.

Our study fits the framework developed by Barrett et al. (2012),where participation decision is conceptualized as a sequence offour stages: firm choice of procurement location; firm contract

offer; smallholder contract acceptance; and firm and smallholders’decisions to honor the contract. In this framework, the fourth stage(contract compliance) is the outcome of the preceding stages,which reflect the attractiveness of the contract offer and the likeli-hood of the offer being accepted by farmers. Hence, our study is, ineffect, an attempt to understand the preferences of farmers towarda contract offer ex ante, and can be considered as a first order con-dition for causality studies such as Bellemare (2012) and Barrettet al. (2012).

To achieve our objective, we combined a literature review to de-fine contract design attributes, an analytical hierarchy process(AHP) method to identify the most important contract design attri-butes, and a discrete choice experiment (DCE) to elicit individualpreferences. Choice-based approaches are relatively new to theCF literature.

The remainder of the paper is organized as follows. ‘‘Literaturereview and conceptual framework’’ provides a literature reviewand the conceptual framework. ‘‘Method’’ presents the method,followed by ‘‘Econometric results and discussion’’, where we pres-ent the empirical results and discussion. ‘‘Conclusion’’ provides theconclusion.

Literature review and conceptual framework

The objectives of this literature review are to explore the factorsleading to CF, understand agricultural contract functions and con-cepts, and identify contract design attributes that could motivatesmallholders to participate in CF. We do not aim to provide a fullliterature review of the determinants and the effects of CF; readersare advised to read the overview by Little and Watts (1994), Kir-sten and Sartorius (2002), and Bijman (2008) or, more recently,Barrett et al. (2012).

Market imperfections and transaction costs – antecedents forparticipation in CF

Contracting between farmers and their buying firms can be con-ceptualized as a specific form of governance structure. Accordingto Transaction Cost Economics (TCE), governance structures areinstitutional arrangements that have evolved (or have been cho-sen) in order to prevent or reduce transaction costs (Williamson,1979). Although the TCE literature usually emphasizes asset spec-ificity as the main source of transaction cost, in agricultural trans-actions uncertainty is the most common determinant ofgovernance structure (Masten, 2000). Agricultural transactions in-volve high uncertainty because products are perishable and har-vested seasonally. When farm products are delivered to theprocessing industry, transactions involve high coordination costsbecause of aligning production, harvesting, collection, and process-ing. In developing countries, which are often characterized by highmarket failures, smallholders are exposed to additional risk anduncertainty (Delgado, 1999; Key and Runsten, 1999; Poole et al.,1998; Poulton et al., 2010). Production risks are not only resultingfrom uncontrollable factors such as weather conditions, the quan-tity and quality of output is also affected by the environmentaluncertainty related to failing input markets (e.g., unavailability offertilizers at crucial moments in the growth cycle of the plant).In addition, farmers face price uncertainty due to high fluctuationsin demand, and technological uncertainty due to insufficient assis-tance for using new crop varieties or inputs (Smale et al., 1994). Byentering in a CF scheme, smallholders have the opportunity to engagein the production of a remunerative crop, a production that otherwisewould entail high uncertainties that present prohibitive risks.

From the perspective of the agribusiness firm, CF can be anattractive governance structure as it allows to reduce the transac-

16 G.K. Abebe et al. / Food Policy 40 (2013) 14–24

tion costs related to procurement risks. Particularly for companiesprocessing agricultural products, uniform quality and consistencyof supply is of crucial importance (Sartorius and Kirsten, 2007; Ver-meulen et al., 2008). For preferred suppliers of fresh produce tosupermarkets, CF schemes can reduce their risk of sourcing prod-ucts that have the proper certificates, indicating the products havebeen produced under the strict quality requirements of the (for-eign) retailer (Jaffe et al., 2011).

1 The number of pairwise comparison is given by n(n � 1)/2; where n is a matrix ofcontract design attributes in each of the three sources of uncertainty, which is 3[4(4 � 1)/2].

Transactions involving seed potatoes

The transaction we have been studying is the production of seedpotatoes by smallholder farmers and the sale of these potatoes to atrading company. This trading company, in turn, sells the seedpotatoes to domestic and (mostly) foreign ware potato growers.The production of seed potatoes is more risky than most othercrops, for several reasons. First, seed potato production requiresintensive cultivation practices like selecting the appropriate plant-ing date and harvesting date, frequency of tillage, fertilizer applica-tion at the right moment in plant growth cycle, frequency offungicide application, and storage (Hirpa et al., 2012). There areseveral trade-offs in cultivation practices. For instance, focusingtoo much on yield growth may not lead to proper tuber size (as de-manded by the buyer). This is even more challenging as tuber sizeis not easy for farmers to observe on a daily basis. Second, seed po-tato production is costly as it requires large amount of plantingmaterial per unit of land (Batt, 2003). Third, potato is a seasonalproduct and rather perishable. This requires a timely coordinationto minimize the loss of value during the process of harvesting, stor-age, distribution and marketing.

The complex nature of seed potato production may therefore ex-pose smallholders to direct transaction costs, such as searching andselecting of the right quality seed and other inputs, and indirecttransaction costs resulting from missing input markets or the failureto identify appropriate trading partners. In addition, seed potatoproduction requires specific investment in human capital, due tothe specificity of cultivation practices. Thus, the high cost of seeds,the need for specific inputs, the special skills needed, and the limitedmarket opportunities of the harvested product all call for an institu-tional arrangement that sufficiently reduces direct and indirecttransaction costs, while maintaining the incentive structure for indi-vidual farmers. Such hybrid governance structure can be the CFscheme.

While TCE literature seeks to forecast or explain the incidence ofparticular governance structure (as the outcome of an economizingprocess), we are more interested in the details of the contractualarrangement and how the different contract attributes can accom-modate particular risks. The main functions of agricultural contractsinclude minimizing of coordination and transaction costs, providingincentives (including penalties), and sharing of risks (Bogetoft andOlesen, 2002; Grosh, 1994; Key and Runsten, 1999). To realize thesefunctions, a contract design may incorporate several instruments,such as risk-sharing mechanisms, incentive schemes, contract me-nus, repeated contracting and renegotiation options, and simplifiedand transparent contract terms (Bogetoft and Olesen, 2002).

In a CF scheme, the main contract design problem of the firmrelate to the quality and price of the product; the sufficiency ofsupply; the necessary inputs; and the coordination of production,harvesting and delivery (Key and Runsten, 1999). Often the firm’ssolution to this problem is to define profit maximizing contractterms assuming that the farmer will accept and honor them. How-ever, contract design is a multi-criteria decision problem involvingtrade-offs (Bogetoft and Olesen, 2002) and, hence, should alsoinclude the incentive considerations and risk-bearing capacities ofsmallholders (Lajili et al., 1997).

Table 1 shows the relationship between contract design con-cepts, functions, and attributes. While contract design attributescan be considered as factors that affect smallholders’ motivationto participate in CF, they can also be conceptualized as instrumentsthat are used by a firm to achieve coordination, motivation, andtransaction cost minimization objectives.

Conceptual framework

Our conceptual framework explains how farmers choose CF toreduce uncertainty, and how different contract design attributescould varyingly affect their motivation to participate in CF. In orderto understand smallholders’ motivation toward CF, we first pro-posed 12 contract design attributes that were adapted fromMasakure and Henson (2005). Using a Principal Component Analy-sis (PCA), they reported 11 attributes that varyingly influencedsmallholders’ choice to participate in CF in the context of a high-value fresh produce exports. We used a choice-based research de-sign, which requires respondents to choose among alternativesrather than to rank or rate them (Chang et al., 2012). A choice taskthat involves more than six attributes is not recommended (Greenand Srinivasan, 1990), as this tends to confuse respondents (Saw-tooth Software, 2008). Thus, we had to limit the contract designattributes to six. To do so, we carried out a pilot study among 20seed potato farmers, 60% of them had experience in CF. The resultswere analyzed using the Analytical Hierarchy Process (AHP) meth-od. AHP has been used in several applications of multi-criteriadecision making (Ghodsypour and O’Brien, 1998), and evaluates aset of alternatives in a hierarchical structure.

The farmers evaluated every pair of contract design attributesusing a pairwise comparison matrix (Sinuany-Stern et al., 2000).In the pairwise comparison, two contract design attributes fromeach of the three sources of uncertainty were shown on either sideof a 9-point scale, where 1 shows both contract design attributesare equally important and 9 represents that one of the contract de-sign attributes is extremely preferred over the other. Subsequently,farmers evaluated a total of 18 judgments1 (pairwise comparisons).The AHP uses the eigenvector method to yield priorities for criteriaand for elements by criteria, and then synthesizes the priorities ofthe elements by criteria into composite measures to arrive at a setof ratings for the elements (Sinuany-Stern et al., 2000); the bestalternative is the one with the highest rate. Table 2 details the rela-tive importance of different contract design attributes.

Concerning output market uncertainty, farmers consideredprice option and form of contract more important than contractduration and quantity. As to quality uncertainty, seed and productquality specifications were more important than quality controlmechanism and place of quality inspection. Regarding input mar-ket uncertainty, farmers considered input supply arrangementand technical assistance more important than transportation andcredit arrangements. The six highest weighted contract designattributes (Table 2, mean scores in bold) are used in our conceptualframework; they will be discussed individually below.

Price optionPrice volatility is one source of uncertainty that may affect

smallholders’ participation decision in CF. Different price optionsmay entail different risks and rewards (Hueth and Ligon, 1999).Price option refers to the payment conditions farmers accept in ex-change for delivering an agreed product quality and quantity. Thecommon price options are fixed, variable, or formula (Bogetoft and

Table 1Contract design concepts, functions, and attributes.

Contractdesignconcepts

Contract functions Contract design attributes

Coordination Coordination of production, harvesting, and processing/marketing; allocation of risks

Form of contract; product quality specification, seed quality specification, qualitycontrol mechanism, place of quality inspection; input supply arrangement, technicalassistance, transportation, and credit

Motivation Provide proper incentives for effort and investment; reduceopportunism and renegotiation, allocation of value; continuity

Price option, form of contract, quantity, and contract duration; product qualityspecification, and quality control mechanism; specification of who bears what risk

Transactioncosts

Reduce direct and indirect cost of contracting; increasetransparency

Form of contract; product quality specification; sanctions; conflict resolutionprocedure

Table 2Pilot study results (n = 20).

Source of uncertaintya Contract design attributes Mean scoreb

Output market uncertainty Price option 0.369Form of contract 0.359Contract duration 0.149Contract quantity 0.123

Quality uncertainty Seed quality specification 0.479Product quality specification 0.217Quality control mechanism 0.182Place of quality inspection 0.122

Input market uncertainty Input supply arrangement 0.361Technical assistance 0.269Transportation arrangement 0.250Credit arrangement 0.121

a Based on Masakure and Henson (2005).b Figures in bold correspond to the attributes used in the DCE.

G.K. Abebe et al. / Food Policy 40 (2013) 14–24 17

Olesen, 2002; Miyata et al., 2009). For simplicity, we focus on fixedand variable price options.

If a contract specifies a fixed payment ex ante, farmers only bearthe production risk while the firm takes all the market risk. Byaccepting a lower expected price, farmers, in effect, agree to paya risk premium. Indeed, a fixed price option increases the firm’srisk exposure. However, the firm can employ different risk man-agement tools that are not available for farmers. Conversely, iffarmers consider a fixed price option unattractive, the firm canuse a variable price option. This strategy may reduce moral hazardproblems, by making both parties residual claimants, but may in-crease farmers’ price risk exposure (Wolf et al., 2001). Since weplan to conduct the experiment with seed potato farmers, wedisaggregated the variable price option into size-based and yield-based.

Obviously, there are trade-offs in choosing one price option overanother. While choosing a fixed price option provides farmersinsurance against downside price risks, this option would disfavorthem when the ex post spot market price by far exceeds the priceagreed in the contract. Based on evidence from several empiricalstudies (e.g., Minten et al., 2009; Miyata et al., 2009; Tripathiet al., 2005), we expect farmers to prefer a fixed price option overa variable one, ceteris paribus. Furthermore, seed potato produc-tion that targets a specific tuber size increases the intensity of farmmanagement practices. Hence, we expect farmers to prefer theyield-based over the size-based price option, ceteris paribus.

Form of contractAllocation of risks and rewards could be affected by the form of

the contract, which can be written or oral (Barrett et al., 2012).A written contract specifies detailed roles and responsibilities,procedures for monitoring, and penalties for performance non-compliance (Poppo and Zenger, 2002). Although a writtencontract could provide better enforcement possibilities, it remains

incomplete (Williamson, 1979). In an oral contract, reputationand repeated interactions are the main enforcement mechanisms(Wolf et al., 2001). Levin (2003) argues that oral contracts cansubstitute written contracts by promoting trust in the relationship,providing the incentive to pay promised compensation, and givingthe parties the option to walk away.

Due to limited prior exposure to working with agribusinessfirms, farmers are expected to prefer the coordination and motiva-tion provisions of a contract to be specified in a written form overan oral form, ceteris paribus.

Seed quality specificationSeed quality uncertainty shows the systematic link between in-

put and output markets; i.e., it implies that specific inputs are nec-essary to get definite output quality (Little and Watts, 1994; Scott,1984). From the farmers’ perspective, seed sourced from anony-mous suppliers may lack quality and could expose them to produc-tion risk, such as yield, and price risk, due to poor output quality.CF can reduce seed quality uncertainty if the buyer firm is supply-ing seed as part of the contract. Seed quality specification becomesmore important when output quality is difficult to measure (Good-hue, 2011). By supplying seed of a known quality, a part of thequality risk will be reduced, both for farmers and the buyer firm.

From the above discussion, two seed quality specification op-tions can be considered: seed sourced from the buyer firm or fromanother supplier. When seed is sourced from the buyer firm, qual-ity (and supply) of the seed is assured, but farmers are bound tosell the output only to the buyer firm (Henson et al., 2005). Con-versely, when seed is sourced from another supplier, the qualityof the seed is not guaranteed because the other supplier may havean incentive to cheat. Moreover, the market for good quality seedmay not be accessible for farmers.

The trade-off for the farmers would be whether to use buyerfirm supplied seed that could be overpriced because the firmmay have monopoly power in supplying the seed but reduces therisk of low quality seed, or to buy seed from other suppliers, at alower price but take the risk of low quality seed. Because the impli-cation of seed quality for both production and price risk is high, weexpect farmers to prefer the buyer firm despite concerns for higherseed costs, ceteris paribus.

Product quality specificationQuality uncertainty is one source of risk in agricultural transac-

tions (Wolf et al., 2001). The desire for high (specific) quality attri-butes increases the firm’s willingness to engage in CF (Goodhue,2011; Henson et al., 2005). Likewise, searching for buyers and get-ting to know their quality requirements is difficult in an imperfectmarket environment. Hence, CF is expected to reduce farmers’quality uncertainty because the quality demand of the buyer firmwill be known ex ante. Accordingly, two contract options can beconsidered: minimum quality for all deliveries or provisions forvariable quality.

Table 3Summary of the conceptual framework.

Categories Contractattributes

Attribute levels

Output marketuncertainty

Price option 1 A fixed price, for all deliveries2 Variable price, depending onyield3 Variable price, depending ontuber size

Form of contract 1 No written contract2 Written contract

Qualityuncertainty

Product qualityspecification

1 Minimum quality requirementsfor all deliveries2 Variable quality is accepted, withvariable price

Seed qualityspecification

1 Seed supplied by buyer firm2 Seed purchased from anothersupplier

Input marketuncertainty

Input supplyarrangement

1 Provision of inputs by the buyerfirm2 Provision of inputs by thegovernment3 Provision of inputs by an NGO

Technicalassistance

1 Provision of technical assistanceby the buyer firm2 Provision of technical assistanceby the government3 Provision of technical assistanceby an NGO

18 G.K. Abebe et al. / Food Policy 40 (2013) 14–24

Minimum quality for all deliveries refers to cases where farmersreceive the same payment per unit by virtue of meeting a pre-spec-ified minimum quality level. Consequently, farmers assume therisk of product rejection without receiving a premium for an aboveaverage quality. This option entails a low price risk related to animperfect quality measurement by the buyer, as there will be a sin-gle standard to measure quality. A firm that targets a single chan-nel may find this option appropriate. In contrast, a firm havingdifferentiated markets may accept or even prefer different qualitylevels. Because payment depends on performance, this arrange-ment may stimulate farmers to deliver a high quality product.However, this option requires several quality measurement crite-ria. Consequently, the cost of measuring quality is expected to behigh and may expose farmers to additional price risk (Hueth andLigon, 1999). For instance, a farmer who has delivered a high qual-ity product, after investing in quality improvements, could still re-ceive a low price due to a measurement error by the buyer firm.

Generally, the choice for farmers is between a fixed quality op-tion, which offers little incentive for improving quality and holdsthe risk of complete rejection, and a variable quality option, whichmay expose them to downside price risk because of a quality mea-surement error. However, in a CF scheme, the risk of incorrectlymeasuring quality tends to be more frequent than the risk of com-plete rejection, and thus we expect farmers to prefer a fixed qualitycontract over a variable one, ceteris paribus.

Input supply arrangementIn an imperfect input market, farmers may have limited access

to specialized inputs. In order to access such inputs, farmers mayconsider to participate in CF. This may give the firm a monopolypower over the provision of specialized inputs and a monopsonypower in the product market (Key and Runsten, 1999). To avoidbecoming fully dependent on the buyer firm, farmers may opt tosource key inputs from a third party. However, provision by publicagencies is often less efficient and effective (Dorward et al., 2004),which could endanger the CF relation between the buyer firm andsmallholders.

Input supply, therefore, could be arranged by the buyer firm, thegovernment, or an NGO. When inputs are supplied by the buyer firm,the firm has the advantage of controlling the quality of inputs andkey farm management practices (Wolf et al., 2001). Alternatively,when inputs are supplied by the government or by an NGO, the firmcan allocate its resources to other activities and can avoid the risk ofcredit default and collection costs. The pros and cons of this type ofpublic–private partnerships are increasingly discussed in the litera-ture (Boselie et al., 2003; Harou and Walker, 2010, cited in Barrettet al., 2012; Poulton and Macartney, 2012).

Given the trade-off in each arrangement, we expect farmers toprefer the buyer firm to supply the inputs despite concerns thatthe firm could overcharge for these inputs, ceteris paribus.

Technical assistanceSimilar to the input supply arrangement, the need to access

information (on technology, timing, and quality, see Key and Run-sten, 1999) may motivate farmers toward CF. Access to new pro-duction techniques not only helps farmers to improve productionand market performance of the contracted crop, it can also havea positive spillover effect on other crops (Masakure and Henson,2005; Minten et al., 2009). Technical assistance can be arrangedin different ways. The buyer firm could provide all the requiredtechnical assistance. While this type of arrangement may allowfarmers to get technical assistance and research-based informa-tion, the buyer firm could overcharge farmers for this service.Alternatively, government or NGO extension agents could providetechnical assistance. However, they may be less effective in provid-ing contract-specific technical assistance.

Given the above considerations, we expect farmers to prefer thebuyer firm to provide technical assistance despite concerns thatthe firm may overprice the services, ceteris paribus.

In sum, our conceptual framework highlights the trade-offsfarmers encounter in evaluating different sets of contract designattributes. Table 3 summarizes the conceptual framework dis-cussed above.

Method

Discrete choice experiments

Choice experiments are based on the Lancaster’s (1966) theoryof consumer choice, where individuals derive utility from the dif-ferent characteristics a good possesses, and McFadden’s (1974)random utility theory, providing the econometric rationale ofchoice experiments. Following Lancaster (1966), attributes havebeen defined as characteristics of a good. However, recent studieshave extended this concept to include aspects of policy design (Co-lombo et al., 2005), agro-environmental scheme design (Ruto andGarrod, 2009), community forestry design (Gelo and Koch, 2012),and land use management contract design (Tesfaye and Brouwer,2011).

Discrete choices experiments (DCEs) are used when a choiceproblem involves two or more discrete alternatives. Among thediscrete choice models, logit is the most widely used model (Train,2003). In this model, the random term is assumed to be indepen-dently and identically distributed as type 1 extreme value distribu-tion; i.e., the model takes the assumption that unobserved factorsare uncorrelated over alternatives and have the same variance forall alternatives. In this study, DCE is used to construct alternativesthat are defined in terms of contract design attributes and the lev-els these attributes could take.

Empirical model specification

The model developed in this study considers the contract designattributes (levels) shown in Table 3, which are assumed to deliver a

G.K. Abebe et al. / Food Policy 40 (2013) 14–24 19

certain level of utility to the potato farmers when participating in aCF scheme.

While the multinomial logit model is the widely used functionalform in DCE, it does not accommodate preference heterogeneitywithin choice data and does not allow each respondent to respondto multiple choice sets (Ben-Akiva and Lerman, 1985; McFadden,1974). A conditional logit model is appropriate when the choiceamong alternatives is modeled as a function of the characteristicsof the alternatives rather than the characteristics of the individualmaking the choice. This makes the conditional logit model appro-priate for estimating behavioral models.

Our model specification follows Train (2003). In a sample con-sisting of N respondents with choice of J unordered alternativeson T choice tasks, the indirect utility that an individual farmer ndrives from choosing alternative j on a choice task t from a finiteset of J alternatives is given by

Vnjt ¼ ajn þ cjzn þ bnxnjt þ enjt ð1Þ

where Vnjt stands for the value (utility) of alternative j to individualn on choice task t, ajn is the alternative-specific intercept, cj capturespreference heterogeneity related to individual-specific characteris-tics, zn is a vector of individual-specific variables for individual n,bn captures systematic preference heterogeneity related to contractdesign attributes, xnjt is a matrix of contract design attributes spe-cific to alternative j, and enjt is a matrix of a random term to accountfor the aspects of utility that the researcher does not observe.

Eq. (1) presents a general model that allows estimating alterna-tive-specific effects (Gelo and Koch, 2012). First, we employ a pureconditional logit model by restricting cj = c and ajn = a to modelindividual choices solely as a function of the characteristics ofthe alternatives. That is, assuming that bn = b and the error termis independently and identically distributed as type 1 extreme va-lue and independent across alternatives, the logit choice probabil-ities can be derived by the following conditional logit model

Pnit ¼expðbxnit þ aþ chnÞ

XJ

j¼1

expðbxnjt þ aþ chnÞð2Þ

where Pnit be the probability of individual n choosing alternative ion choice task t among J alternatives. In the conditional logit model(Eq. (2)), the explanatory variables x assume different values in eachalternative at each choice set. However, the impact of a unit of x isusually assumed to be constant across alternatives, giving only asingle coefficient estimate for each x variable. Hence, the impactof a variable on the choice probabilities derives from the differencein the value of the characteristics across alternatives.

Second, we relax our assumption and include socio-economicfactors; that is, the conditional logit model given in Eq. (3) sepa-rates explanatory variables into alternative-specific attributes ofthe choices, such as contract design attributes, and characteristicsof the individual, such as age, sex, and education. This implies thatthe effect of individual-specific variables would be different foreach alternative as indicated in Eq. (1). Thus, the alternative-spe-cific conditional logit model probability is given by

Pnit ¼expðbnxnit þ ain þ ciznÞ

XJ

j¼1

expðbnXnjt þ ajn þ cjznÞð3Þ

Experimental design

DCE are used in this study to estimate the effect of differentcontract design attributes on the attractiveness of a contract.How well a discrete choice experiment performs partly depends

on the options used in the choice experiment, and how the optionsare grouped into choice sets (Street et al., 2005). The AHP results(Table 2) allowed us to determine a narrow range of contract de-sign attributes and levels that we then used to create choice setsusing the Sawtooth Software. We tested the experimentally gener-ated questionnaire for the clarity of choice sets. The final question-naire consisted of three versions, which vary in the order and typeof choice sets. Each respondent was randomly assigned to one ofthe three versions to make sure that the order of the choice setsdoes not affect respondents’ preferences and to get more variationin the preferred choices.

Regarding the sample selection, the strategy was to include allthe 72 farmers that already had a contract and the same numberfrom the pool of non-contracted farmers. For the latter, we ran-domly selected 120 farmers from the land ownership registerand continued the experiment until we reached to 72 farmers.Thus, a total of 144 respondents were selected, and each respon-dent was given 15 experimentally generated choice tasks, eachcontained two cards. Each card was described by one of the levelsof the six contract design attributes. Hence, respondents wereasked to evaluate their preferred set of contract design attributesin 15 different choice tasks, making it a panel of 2160 choice tasksand 4320 observations. The alternatives were constructed in a waythat respondents had to make trade-offs to avoid any dominantchoice tasks, in which one alternative is strictly superior to theother. During the experiment, the principal investigator explainedrespondents about each choice task before moving to the next taskto ensure that they understood the trade-offs between profiles.

Study context and description of the data

Ethiopia is endowed with good agro-ecological zones (coolhighlands) for the production of relatively disease free and highquality seed potatoes. Although some improved varieties exist,which were released by the state research institutes, the uptakefor such varieties is very low (Hirpa et al., 2012), because adapta-tion of these varieties to local conditions is low and seed supplysystems are poorly developed.

Following the 2005 Agriculture Development Led Industrializa-tion policy, the country has attracted many foreign firms. As of2009, the government had either granted or promised to grantaround three million hectares of land to foreign investors (Weiss-leder, 2009). As this practice has been scrutinized by activistsand international NGOs (Li, 2011; Oya, 2012), CF has becomeincreasingly attractive for foreign firms. Although the governmentconsiders CF as a suitable policy instrument to integrate farmersinto agricultural value chains (Minot, 2011), participation contin-ues to be low, with many problems of non-compliance and side-selling (Getaneh and Bekabil, 2008).

In 2006, a foreign-owned agribusiness firm started a CF schemeto produce seed potatoes for the export market. West Shewa Zone,one of the areas selected for this purpose, has a population of over2 million, with an average family size of 4.8 and an average farmsize of 1.4 ha (Deininger, 2003). Farmers in the area grow severalcrops in addition to potato. However, participation to the CFscheme was limited to farmers who had agricultural land alongthe irrigation canal, which was built by the local government.The foreign-owned firm provided different seed varieties, whichwere all imported, and other inputs to the contracted farmers. Fur-thermore, the firm assigned personnel to provide technical assis-tance and to monitor farm management practices.

The experiment was conducted between September andNovember 2011. Half of the respondents were purposely selectedbecause of their experience in growing seed potatoes under a CFscheme. But the other half were randomly selected from the landownership register. From those who did not participate before,

Table 4Household and individual characteristics of sample farmers.

Variables Participatedbefore (n = 72)

Notparticipatedbefore (n = 72)

Mean Std.dev.

Mean Std.dev.

Age (year) 39.7 14.3 39.2 11.9Education (school year) 5.9 3.8 5.9 3.2Experience in agriculture (year) 18.3 12.3 18.6 11.3Farm size (ha) 2.9 2.4 2.4 2.2Distance from farm to main road (km) 2.7 1.9 2.3 1.3Sex (% male) 81.9 94Intention to participate in the next

season (% yes)80.6 82

Access to irrigation (% yes) 100 29

20 G.K. Abebe et al. / Food Policy 40 (2013) 14–24

82% of them indicated their intention to participate in the next sea-son. This helped us to attain the maximum level of realism to con-duct the experiment.

The questionnaire contained three parts: open questions, socio-economic data, and the DCE. The open questions were meant tounderstand respondents’ perception about CF. Accordingly, 98%of farmers responded that participation could bring some benefits,such as improved income, access to key inputs and technical assis-tance. However, 68% of participated and 38% of non-participatedrespondents expressed their concerns about CF, which includedpossible disagreement on contract terms, lack of trust in the rela-tionship, and low contract price.

Regarding the socio-economic variables, we expected that someindividual and household level characteristics affect the probabil-ity of choice. Therefore, we collected data on age, education level,sex, farm size, experience in potato farming and contract farming,access to irrigation, and distance from the main road (see Table 4).

Scanning through Table 4, the descriptive statistics for partici-pated and non-participated farmers are similar with the exceptionof access to irrigation. This is not surprising as the firm contractedwith farmers that can produce seed potatoes during the off-rainyseason.

Econometric results and discussion

We present the estimated utility function parameters of theconditional logit model (clogit) and the alternative-specific condi-tional logit (asclogit) model in Table 5. The parameters were esti-mated by the software package Stata 12.0. A variance estimator,provided by Stata 12.0, was applied to allow intragroup correlation,relaxing the usual requirement that the observations be indepen-dent (Baum et al., 2011).

2 We re-estimated the asclogit model with only ‘irrigation’ and with only‘experience in agriculture’, together with the contract design attributes; nonetheless,the model yielded no significant difference from the results reported in Table 5. Thisimplies that access to irrigation and experience in agriculture did not have influenceon the choice of contract design attributes.

Estimations of the conditional logit model

Referring to Table 5, with the exception of price option andproduct quality specification, all the other variables showed ex-pected signs. With regard to contract form, the probability thatfarmers choose the base alternative written contract is 95%; thatis, if a firm attempts to switch from a written contract to an oralcontract, the odds of being chosen by farmers is only 5%, ceterisparibus. As to the input supply and technical assistance, the buyerfirm is more likely to be chosen than the base alternative NGO;other things being equal, if a contract specifies input supply andtechnical assistance from the buyer firm, the odds of being chosenby farmers is 3.2 and 2.3 times higher than the base alternativeNGO, in that order. In addition, the probability that farmers choosetechnical assistance by government agents is 55% compared to the

base alternative NGO. Similarly, seed sourced from the buyer firmis five times more preferred than seed sourced from elsewhere,ceteris paribus. On the other hand, a fixed quality specification isless likely to be chosen than a variable quality specification. Thus,if the firm opts for a fixed quality specification, the odds of beingchosen by the farmers will be 25% less than the odds of farmerschoosing the base alternative size-based price option. However,the probability that farmers choose a yield-based price option is57%, compared to the alternative size-based price option; that is,if a firm switches its pricing strategy from a size-based to ayield-based one, the odds of choosing the latter is 32% higher thanthe odds of choosing the former. The relatively higher preferencetoward the yield-based price option over the size-based was ex-pected due to the high rejection risk in the latter.

Comparing estimations of the clogit model and asclogit model

Table 5 also provides a comparison of parameter estimationfrom the clogit model and asclogit model. In general, the signand significant level of the choice-specific variables are the samein both models. However, in the asclogit model (1) the significancelevel of technical assistance by the government has increased; (2)the odds of choosing oral contract, fixed quality specification, andyield-based price option have all increased; and (3) the odds ofchoosing seed source, inputs, and technical assistance from thebuyer firm have all decreased. Overall, the log likelihood has im-proved. This is expected as the inclusion of individual specific-vari-ables is likely to improve the estimations of the clogit model (Train,2003). However, the coefficients of the individual-specific variableswere not statistically significant, implying that the individual-spe-cific variables included in the model did not systematically affectthe probability of choice.2

Discussion

Smallholders producing seed potatoes in Ethiopia prefer a vari-able price contract. This result seems to contradict the commonassumption that smallholders in developing countries are riskaverse (Fafchamps, 1992), as well as the empirical findings of Miyataet al. (2009), Minten et al. (2009), Tripathi et al. (2005), and Bielzaet al. (2007). Our findings suggest that a pricing strategy contingenton certain performance criteria is more preferred than a fixed price.Our results correspond to the findings of Wang et al. (2011) whoreported smallholders’ preference for a floating price.

Inconsistency of findings across cases may suggest that motiva-tional differences vary depending on institutional settings. In thestudy area, a formula price was used by the firm, which was basedon the price of ware potatoes at the nearby local market and theyield and size of seed potatoes. However, failure to meet the spec-ified tuber size entails a significant loss for farmers; they receiveonly 25% of the price for ware potatoes at the local market.

There are several reasons why risk averse farmers may still optfor a variable price. First, farmers are suspicious that fixed priceswill also be low prices. Thus, farmers appear to perceive that con-tracts not only hold low prices because of the ‘insurance premium’they pay to the buyer, but also lead to underpayment because oncefarmer are locked into the contract the buyer can take advantage ofthe asymmetric power relationship. Second, farmers seem to be-lieve that they can outperform fellow farmers. As principal investi-gator in the experiment, the first author observed that farmers

Table 5Estimations of the clogit and asclogit models.

Variablea Clogit Asclogit

Coef. (std. err.) Odds ratio Coef. (std. err.) Odds ratio

Oral contract (written contract) �3.01*** (0.16) 0.05 �2.15*** (0.14) 0.12Minimum quality for all deliveries (variable quality specification) �0.87*** (0.13) 0.42 �0.64*** (0.09) 0.53Seed from buyer (seed from other supplier) 1.62*** (0.12) 5.06 1.02*** (0.08) 2.77Fixed price (size-based price) �0.29*** (0.11) 0.75 �0.33*** (0.10) 0.72Variable price (yield-based) (size-based variable price) 0.28*** (0.15) 1.32 0.57*** (0.14) 1.77Inputs by buyer (inputs by an NGO) 1.16***(0.11) 3.18 0.89*** (0.12) 2.44Inputs by the government(inputs by an NGO) 0.04 (0.12) 1.04 0.06 (0.11) 1.06Technical assistance by the buyer (technical assistance by an NGO) 0.84*** (0.10) 2.32 0.61*** (0.10) 1.85Technical assistance by the government (technical assistance by an NGO) 0.19* (0.10) 1.20 0.20** (0.09) 1.22Sex �0.06 (0.20)Age (ln) – �0.28 (0.33)Farm size (ln) – �0.12 (0.09)Experience in agriculture (ln) – 0.02 (0.17)Education – �0.02 (0.06)Prior experience in CF – 0.06 (0.14)Access to irrigation 0.07(0.13)Distance from the main road in km – �0.02 (0.04)Constant – 0.87 (0.94)Respondents 144 144Observations 4320 4320Log likelihood �1718.89 �833.51

* P < 0.1.** P < 0.05.*** P < 0.01; (robust) standard errors are given in parentheses; VIF, uncentered, 1.75.

a Variables in brackets are the base alternatives.

G.K. Abebe et al. / Food Policy 40 (2013) 14–24 21

were optimistic about their chances of meeting the quality require-ments set by the buyer firm. This suggests that the fixed price op-tion may penalize farmers who have entrepreneurial skills (as hasbeen argued by Rehber, 1998). Third, earlier studies have docu-mented trends of escalating food prices in Ethiopia (Alem andSöderbom, 2011), and have shown that the agricultural commoditymarket is characterized by speculative behavior (Tadesse and Gut-tormsen, 2011). We also learned from farmers that the price ofpotatoes had been on the rise, which could be another factor intheir decision toward a variable price.

Farmers’ preference for a written contract could have at leasttwo explanations. First, in a thin market environment, farmersneed a guaranteed market for their product before they actually in-vest in production (Kirsten and Sartorius, 2002). A written contractcould better serve this purpose than an oral contract as the formerdetails the coordination and motivation aspects of a contract exante which could be used to resolve conflicts ex post. Second, atrust-based relationship is established primarily through repeatedtransactions (Fafchamps and Minten, 1999). Since agribusinessfirms generally work in areas where they have not previously beenoperating, local farmers may not trust them (Guo et al., 2005;Singh, 2002). As a result, farmers may seek all contract provisionsto be specified in a written form. Consistent with our result, Harouand Walker (2010, cited in Barrett et al., 2012) reported Ghanaianfarmers’ regret for accepting oral contracts from agribusiness firms.In our study area, the agribusiness firm used a written contract.

The preference for seed supplied by the buyer firm is also con-sistent with the risk averse attitude of farmers in the input market;i.e., the buyer firm is considered more reliable than any othersource. This was expected as seed quality has a substantial effecton yield and price risk (Scott, 1984). Being the residual claimantof the contract, the firm has an incentive to supply good qualityseeds. Furthermore, seed sourced from the buyer firm can facilitaterisk sharing when producing a crop with specific attributes. Allcontracted farmers in the study area used firm-supplied seed.

The results on product quality specification show that farmersgenerally prefer a variable quality scheme rather than a fixedone. This implies farmers’ willingness to accept a price risk thatis due to a possible quality measurement problem. This result

could at least relate to two factors. First, although the pricing scaleadopted could still penalize them, a variable quality specificationcontract appears to be understood by farmers as a way to avoidthe major risk that not all produce will be purchased. Second, asdiscussed above, farmers seem to believe that they can deliver anabove average quality product. The implication for agribusinessfirms is that a differentiation marketing strategy may be necessaryto induce contracted farmers to deliver a high quality product. Thepractice in the study area was mixed; while the agribusiness firmused a fixed size-based output quality specification, the yield-based one was variable. This practice seems to confirm the firstpoint. During the experiment, we learned from farmers about thedifficulty of managing tuber size, which was also the main sourceof conflict between the farmers and the agribusiness firm. Whilethe firm put a severe penalty on growing oversized potatoes, farm-ers preferred to sell the oversized seed potato in the alternativemarket as consumption (ware) potatoes because the price theycould receive by selling in this market was much higher than whatthey would receive from the firm, which was only 25% of the localprice. In contrast, when the farmers produced seed potatoes withinthe acceptable size range, they had a strong incentive to sell to thefirm as the price they could receive from the firm was much higherthan what they could get by selling in the local market.

With regard to input supply arrangement, the result confirmsthe risk averse attitude of farmers in the input market. This findingis in line with most of the literature, which concludes that inputmarkets in developing countries are missing or imperfect (Faf-champs, 1992; Key and Runsten, 1999). Although public sectorintervention is justified when markets fail, it is often ineffectivedue to unreliable delivery schemes and political interferences(Dorward et al., 2004; Poulton and Macartney, 2012). Moreover,growing non-traditional crops, such as tomato, seed potato, andother vegetables, uses large amounts of pesticides and fertilizercompared to traditional and staple crops (Singh, 2002). Hence, itcould be convenient and efficient for smallholders to receive inputsfrom the buyer firm. In the study area, contracted farmers usedfirm-supplied inputs.

Similar to the input supply arrangement, farmers prefer thebuyer firm over the state or NGO in receiving technical assistance.

22 G.K. Abebe et al. / Food Policy 40 (2013) 14–24

This result was expected as private firms are generally consideredmore trustful and effective than public agencies in deliveringtimely extension services (Bellemare, 2010; Umali-Deininger,1997). However, when the buyer firm is unable to offer technicalassistance, farmers prefer government agencies over NGOs. This re-sult shows an apparent tension between farmers’ preferences andNGO practices. The focus of NGOs is often on resource-poor house-holds in addressing technology gaps left by the state (Farrington,1995; Farrington and Bebbington, 1993; Umali-Deininger, 1997).Furthermore, the work of NGOs often lacks consistency due tothe multiple roles they play: partly responding to government fail-ure, partly addressing market failure, and partly engaging in advo-cacy (Bebbington, 1997). As a result, smallholders may notconsider NGOs as a reliable source of technical support in CFschemes. In the study area, smallholders have had experienceswith several NGOs.

In general, farmers’ choice of contract design attributes appearsto be in line with the predictions of the transaction costs econom-ics theory. Thus, farmers’ choice of written contract over oral form,variable over fixed quality specification, and firm-supplied seed,inputs, and technical assistance over other sources of supply areall motivated by transactional risks related to uncertainty aboutbuyer behavior, price risks, and missing markets. Likewise, farm-ers’ choice of variable price over fixed price is related to the incen-tive regime.

Regarding the reliability of the estimations in Table 5, the unc-entered vif test result shows 1.75 on average, implying no concernof multicollinearity. Also, the likelihood ratio test for the asclogitmodel shows no significance difference between the restrictedmodel, with only contract design attributes, and the unconstrainedmodel, with all the variables.3

Conclusion

Contract Farming (CF) is becoming increasingly important indeveloping countries, partly because of domestic and foreignsupermarkets requiring sophisticated supply chains, partly be-cause of the growing use of quality certificates and corporate socialresponsibility guarantees by downstream customers (Henson et al.,2005; Swinnen, 2007). Recent discussions in the general CF litera-ture mainly focus on smallholders’ participation, contractual rela-tions, and on how to measure the welfare impact of CF. With theobjective of providing insights that can help to improve existingand future CF schemes, this study has explored smallholders’ moti-vation for participation. We hypothesized that farmers’ motivationto participate in CF largely depends on the nature of the contractdesign attributes. We tested this hypothesis by applying the DCEmethod, which followed a literature review and an AHP approach.

The study shows that smallholders are generally positive aboutthe prospect of CF to improve their livelihood, although the major-ity of them suspect contracts to favor agribusiness firms. In termsof motivation, this study shows that farmers’ willingness to partic-ipate in CF increases if a contract design has the following attri-butes: a written form; inputs, technical assistance, and seedsupplied by the buyer firm; and variable output quality and vari-able price options.

3 Hausman McFadden test for the independence of irrelevant of alternatives (IIA)assumption is not possible in our study as each respondent had to face two mutuallyexclusive alternatives (e.g., fixed versus variable price, written versus no writtencontract, etc.). To test the IIA assumption, we need at least three alternatives that theinclusion of a third alternative should not change the odds ratio of the other pairs ofchoices. One possibility was to include a no-choice option in the design. However, thisdid not seem realistic in our study as it provides no information on the impact ofattributes on a choice set (Enneking et al., 2007), and the tendency to choose the no-option is highly likely (Dhar, 1997).

Our study provides several contributions to the literature. First,we show that input market uncertainty is more important thanoutput market uncertainty in smallholders’ decision to participatein CF. In the input market, farmers consider CF as a mechanism ofrisk-sharing to reduce input supply and seed quality uncertainty.In the output market, farmers appear to be worried about the riskof underpayment by the firm once they are locked into a fixed-price contract. They tend to prevent this type of risk by optingfor a variable quality specification and a variable output price.The significance of this result for CF schemes includes the follow-ing. From the behavior of smallholders in the output market, wemay conclude that fear of underpayment by the firm as well as fac-tors outside of contractual relations such as institutional factors(e.g., rising food prices) and individual factors (e.g., entrepreneurialattitude of farmers) are most important. These factors, as a result,tend to discourage farmers from participation because CF limitssmallholders’ freedom to make autonomous decisions. Conversely,in the input market, smallholders seem to have been constrainedby several problems, such as the unavailability of (quality) inputs,the lack of information on where to get and how to use them, andthe lack of access to credit for buying these inputs. The conditionsin the input market, therefore, tend to encourage farmers towardparticipation in CF. Furthermore, our findings imply that theattractiveness of a CF scheme partly depends on the strength ofthe institutional environment in solving input market constraints,and partly on the willingness of agribusiness firms to choose a pric-ing strategy based on variable prices.

Second, our study provides a new dimension of analyzing farm-ers’ decision to participate in CF using a multi-category discretechoice model, where the choice of a contract is modeled as (1) aset of different contract design attributes and (2) a combinationof contract design attributes and individual-specific characteristics.Parameter estimates of the (alternative-specific) conditional logitmodel show that contract choice is largely determined by the char-acteristics of the choice sets rather than the characteristics of thefarmer making the choice. This implies that heterogeneity in a con-tract design may not be necessary in a CF scheme.

Third, unlike most studies on CF, which have a developmenteconomics or political economy perspective, our study takes amanagerial perspective. The study shows the applicability of DCEmodels to investigate the attractiveness of different contract de-sign attributes in CF relations. Knowledge on the preferences offarmers for participation could help agribusiness firms to designbetter contracts that would minimize the problem of side-selling,contract non-compliance, and low levels of participation. Our find-ings suggest that an optimal contract can be designed through bal-ancing the risk averseness of farmers in the input and outputmarket and entrepreneurial desire of farmers in the output market.

Our findings have two policy implications. First, smallholdershave shown strong motivation to receive technical assistance andkey agricultural inputs from agribusiness firms. Hence, to promotesmallholder participation in CF and global value chains, publicagencies should support agribusiness firms that engage in CF, forinstance, by establishing infrastructure, facilitating access to credit,and providing other investment incentives. This finding also con-tributes to the discussion on public–private partnership, whichemphasizes how institutional arrangements between public andprivate sector actors can best be aligned to appropriately allocateresources for development projects (e.g., Hodge and Greve, 2009;Poulton and Macartney, 2012). Second, the risk averse attitude ofsmallholders in the input market calls for institutional interven-tions to reduce the risks. One intervention area could be strength-ening of collective action, such as producer organizations, tosupply key agricultural inputs to smallholders. This type of inter-vention may also induce agribusiness firms to offer more compet-itive contracts to smallholders.

G.K. Abebe et al. / Food Policy 40 (2013) 14–24 23

Finally, the study raises several issues that deserve further re-search. First, we found that smallholders are generally risk averse.It would be interesting to know if farmers change their risk atti-tude in different institutional settings with different crops. Second,it may be relevant to study the perception of agribusiness firms to-ward contract design attributes that smallholders considered moreimportant. Third, it may be important to further explore the appli-cability of DCE models in CF schemes. Fourth, although our resultsare robust regarding smallholders’ choices toward contract designattributes, there could be some bias in the estimates due to a po-tential problem of endogeneity, such as from omitted or unob-served variables. For instance, we attempted to capturesmallholders’ entrepreneurial attitude using the variable ‘experi-ence in agriculture’. However, this may be a poor proxy, and fur-ther research may select other variables that could bettermeasure entrepreneurship traits. Measuring entrepreneurial atti-tude is a common problem in the CF literature (see Barrett et al.,2012; Bellemare, 2012). Future research may also focus on contractschemes used for potatoes delivered to the wet market, the pro-cessing industry, and fast food restaurants, because productionand transaction characteristics may differ from those of seedpotatoes.

Acknowledgements

We gratefully acknowledge the Wageningen University Interna-tional Research and Education Fund (INREF) and the NetherlandsFellowship Program of NUFFIC for funding this research. Theauthors are very grateful to the three anonymous reviewers andthe editor for their insightful comments. The views expressed inthis paper are the sole responsibility of the authors.

References

Alem, Y., Söderbom, M., 2011. Household-level consumption in Urban Ethiopia: theeffects of a large food price shock. World Development 40, 146–162.

Barrett, C.B., 2008. Smallholder market participation: concepts and evidence fromeastern and southern Africa. Food Policy 33, 299–317.

Barrett, C.B., Bachke, M.E., Bellemare, M.F., Michelson, H.C., Narayanan, S., Walker,T.F., 2012. Smallholder participation in contract farming: comparative evidencefrom five countries. World Development 40, 715–730.

Batt, P.J., 2003. Building trust in the Filipino seed potato industry. Journal ofInternational Food and Agribusiness Marketing 13, 23–41.

Baum, C., Nichols, A., Schaffer, M., 2011. Evaluating One-Way and Two-Way Cluster-Robust Covariance Matrix Estimates, German Stata Users Group Meeting, July2011. <http://fmwww.bc.edu/repec/dsug2011/desug11_schaffer.pdf>.

Bebbington, A., 1997. New states, new NGOs? Crises and transitions among ruraldevelopment NGOs in the Andean region. World Development 25, 1755–1765.

Bellemare, M.F., 2010. Agricultural extension and imperfect supervision in contractfarming: evidence from Madagascar. Agricultural Economics 41, 507–517.

Bellemare, M.F., 2012. As you sow, so shall you reap: the welfare impacts of contractfarming. World Development 40, 1418–1434.

Ben-Akiva, M., Lerman, S.R., 1985. Discrete Choice Analysis: Theory and Applicationto Travel Demand. MIT Press, Cambridge, MA.

Bielza, M., Garrido, A., Sumpsi, J.M., 2007. Feasibility of a cash forward contract: anapplication to the French and Spanish potato sectors. Agribusiness 23, 245–261.

Bijman, J., 2008. Contract Farming in Developing Countries: an Overview. WorkingPaper, Wageningen University.

Bogetoft, P., Olesen, H.B., 2002. Ten rules of thumb in contract design: lessons fromDanish agriculture. European Review of Agricultural Economics 29, 185–204.

Boselie, D., Henson, S., Weatherspoon, D., 2003. Supermarket procurement practicesin developing countries: redefining the roles of the public and private sectors.American Journal of Agricultural Economics 85, 1155–1161.

Chang, J.B., Moon, W., Balasubramanian, S.K., 2012. Consumer valuation of healthattributes for soy-based food: a choice modeling approach. Food Policy 37, 335–342.

Chavas, J.P., Holt, M.T., 1996. Economic behavior under uncertainty: a joint analysisof risk preferences and technology. The Review of Economics and Statistics,329–335.

Colombo, S., Hanley, N., Calatrava-Requena, J., 2005. Designing policy for reducingthe off-farm effects of soil erosion using choice experiments. Journal ofAgricultural Economics 56, 81–95.

Deininger, K., 2003. Land Policies for Growth and Poverty Reduction. The WorldBank, Washington, DC.

Delgado, C., 1999. Sources of growth in smallholder agriculture in Sub-SaharanAfrica: the role of vertical integration of smallholders with processors andmarketers of high value-added items. Agrekon 38, 165–189.

Delpierre, M., 2009. Farm input use in a context of liquidity constraints and contractunenforceability. Journal of African Economies 18, 496–528.

Dhar, R., 1997. Consumer preference for a no-choice option. Journal of ConsumerResearch 24, 215–231.

Dorward, A., Kydd, J., Morrison, J., Urey, I., 2004. A policy agenda for pro-pooragricultural growth. World Development 32, 73–89.

Enneking, U., Neumann, C., Henneberg, S., 2007. How important intrinsic andextrinsic product attributes affect purchase decision. Food Quality andPreference 18, 133–138.

Fafchamps, M., 1992. Cash crop production, food price volatility, and rural marketintegration in the third world. American Journal of Agricultural Economics 74,90–99.

Fafchamps, M., Minten, B., 1999. Relationships and traders in Madagascar. TheJournal of Development Studies 35, 1–35.

Farrington, J., 1995. The changing public role in agricultural extension. Food Policy20, 537–544.

Farrington, J., Bebbington, A., 1993. Reluctant partners? Non-governmentalorganizations, the state and sustainable agricultural development. Routledge,New York.

Gelo, D., Koch, S.F., 2012. Does one size fit all? Heterogeneity in the valuation ofcommunity forestry programs. Ecological Economics 74, 85–94.

Getaneh, W., Bekabil, F., 2008. Integrating Small Scale Farmers into Bread WheatMarketing Chain through Contract Farming in Ethiopia. Paper presented at theAfrican Association of Agricultural Economists conference. Accra, Ghana,August. <http://ideas.repec.org/p/ags/aaae07/52101.html>.

Ghodsypour, S.H., O’Brien, C., 1998. A decision support system for supplier selectionusing an integrated analytic hierarchy process and linear programming.International Journal of Production Economics 56, 199–212.

Goodhue, R.E., 2011. Food quality: the design of incentive contracts. Annual Reviewof Resource Economics 3, 119–149.

Green, P.E., Srinivasan, V., 1990. Conjoint analysis in marketing: new developmentswith implications for research and practice. The Journal of Marketing, 3–19.

Grosh, B., 1994. Contract farming in Africa: an application of the new institutionaleconomics. Journal of African Economies 3, 231–261.

Guo, H., Jolly, R.W., Zhu, J., 2005. Contract farming in China: Supply chain or ball andchain? In: Paper Presented at 15th Annual World Food and AgribusinessSymposium. IAMA, Chicago, June.

Harou, A., Walker, T., 2010. The Pineapple Market and the Role of Cooperatives.Working Paper, Cornell University.

Henson, S., Masakure, O., Boselie, D., 2005. Private food safety and quality standardsfor fresh produce exporters: the case of Hortico Agrisystems, Zimbabwe. FoodPolicy 30, 371–384.

Herath, D., Weersink, A., 2009. From plantations to smallholder production: the roleof policy in the reorganization of the Sri Lankan tea sector. World Development37, 1759–1772.

Hirpa, A., Meuwissen, M.P.M., Van der Lans, I.A., Lommen, W.J.M., Oude Lansink,A.G.J.M., Tsegaye, A., Struik, P.C., 2012. Farmers’ opinion on seed potatomanagement attributes in Ethiopia: A conjoint analysis. Agronomy Journal104, 1413–1424.

Hodge, G.A., Greve, C., 2009. PPPs: the passage of time permits a sober reflection.Economic Affairs 29, 33–39.

Hueth, B., Ligon, E., 1999. Producer price risk and quality measurement. AmericanJournal of Agricultural Economics 81, 512–524.

Jaffe, S., Henson, S., Diaz Rios, L., 2011. Making the Grade-Smallholder Farmers,Emerging Standards, and Development Assistance Programs in Africa – AResearch Program Synthesis. � World Bank. <https://openknowledge.worldbank.org/handle/10986/2823>.

Key, N., Runsten, D., 1999. Contract farming, smallholders, and rural development inLatin America: the organization of agroprocessing firms and the scale ofoutgrower production. World Development 27, 381–401.

Kirsten, J., Sartorius, K., 2002. Linking agribusiness and small-scale farmers indeveloping countries: is there a new role for contract farming? DevelopmentSouthern Africa 19, 503–529.

Lajili, K., Barry, P.J., Sonka, S.T., Mahoney, J.T., 1997. Farmers’ preferences for cropcontracts. Journal of Agricultural and Resource Economics, 264–280.

Lancaster, K.J., 1966. A new approach to consumer theory. The Journal of PoliticalEconomy 74, 132–157.

Levin, J., 2003. Relational incentive contracts. The American Economic Review 93,835–857.

Li, T.M., 2011. Centering labor in the land grab debate. The Journal of PeasantStudies 38, 281–298.

Little, P.D., 1994. Contract farming and the development question. In: Little, P.D.,Watts, M.J. (Eds.), Living Under Contract. University of Wisconsin Press,Madison.

Little, P.D., Watts, M.J., 1994. Living Under Contract. University of Wisconsin Press,Madison.

Masakure, O., Henson, S., 2005. Why do small-scale producers choose to produceunder contract? Lessons from nontraditional vegetable exports from Zimbabwe.World Development 33, 1721–1733.

Masten, S.E., 2000. Transaction-cost economics and the organization of agriculturaltransactions. In: Baye, Michael, Maxwell, John (Eds.), Industrial OrganizationAdvances in Applied Microeconomics. Emerald Group Publishing Limited, pp.173–195.

24 G.K. Abebe et al. / Food Policy 40 (2013) 14–24

McFadden, D., 1974. Conditional logit analysis of qualitative choice behavior. In:Zarembka, P. (Ed.), Frontiers in Econometrics. Academic Press, New York, pp.105–142.

Michelson, H., Reardon, T., Perez, F., 2011. Small farmers and big retail: trade-offs ofsupplying supermarkets in Nicaragua. World Development 40, 342–354.

Minot, N., 2011. Contract Farming in sub-Saharan Africa: Opportunities andChallenges. Prepared for the policy seminar organized by ACTESA-MSU-IFPRIAfrican Agricultural Markets Programme. Kigali, Rwanda, April. <http://www.aec.msu.edu/fs2/aamp/Kigali%20Conference/Minot_Contract_farming_(AAMP%20Kigali).pdf>.

Minten, B., Randrianarison, L., Swinnen, J.F.M., 2009. Global retail chains and poorfarmers: evidence from Madagascar. World Development 37, 1728–1741.

Miyata, S., Minot, N., Hu, D., 2009. Impact of contract farming on income: linkingsmall farmers, packers, and supermarkets in China. World Development 37,1781–1790.

Oya, C., 2012. Contract farming in Sub-Saharan Africa: a survey of approaches,debates and issues. Journal of Agrarian Change 12, 1–33.

Poole, N.D., Del Campo Gomis, F.J., Juliá Igual, J.F., Vidal Giménez, F., 1998. Formalcontracts in fresh produce markets. Food Policy 23, 131–142.

Poppo, L., Zenger, T., 2002. Do formal contracts and relational governancefunction as substitutes or complements? Strategic Management Journal 23,707–725.

Porter, G., PhillipsHoward, K., 1997. Comparing contracts: an evaluation of contractfarming schemes in Africa. World Development 25, 227–238.

Poulton, C., Macartney, J., 2012. Can public–private partnerships leverage privateinvestment in agricultural value chains in Africa? A preliminary review. WorldDevelopment 40, 96–109.

Poulton, C., Dorward, A., Kydd, J., 2010. The future of small farms: new directions forservices, institutions, and intermediation. World Development 38, 1413–1428.

Reardon, T., Barrett, C.B., Berdegué, J.A., Swinnen, J.F.M., 2009. Agrifood industrytransformation and small farmers in developing countries. World Development37, 1717–1727.

Rehber, E., 1998. Vertical integration in agriculture and contract farming. RegionalResearch Project NE-165: Private Strategies, Public Policies, and Food SystemPerformance, University of Massachusetts, Department of Resource Economics,Amherst, MA 01003.

Ruto, E., Garrod, G., 2009. Investigating farmers’ preferences for the design of agri-environment schemes: a choice experiment approach. Journal of EnvironmentalPlanning and Management 52, 631–647.

Sartorius, K., Kirsten, J., 2007. A framework to facilitate institutional arrangementsfor smallholder supply in developing countries: an agribusiness perspective.Food Policy 32, 640–655.

Sawtooth Software, 2008. The CBC System for Choice-Based Conjoint AnalysisTechnical Paper. Sawtooth Software, Sequim, WA.

Scott, C., 1984. Transnational corporations and asymmetries in the Latin Americanfood system. Bulletin of Latin American Research 3, 63–80.

Singh, S., 2002. Contracting out solutions: political economy of contract farming inthe Indian Punjab. World Development 30, 1621–1638.

Sinuany-Stern, Z., Mehrez, A., Hadad, Y., 2000. An AHP/DEA methodology forranking decision making units. International Transactions in OperationalResearch 7, 109–124.

Smale, M., Just, R.E., Leathers, H.D., 1994. Land allocation in HYV adoption models:an investigation of alternative explanations. American Journal of AgriculturalEconomics 76, 535–546.

Street, D.J., Burgess, L., Louviere, J.J., 2005. Quick and easy choice sets: constructingoptimal and nearly optimal stated choice experiments. International Journal ofResearch in Marketing 22, 459–470.

Swinnen, J.F.M., 2007. Global supply chains standards and the poor: how theglobalization of food systems and standards affects rural development andpoverty. CABI Publishing, Wallingford.

Tadesse, G., Guttormsen, A.G., 2011. The behavior of commodity prices in Ethiopia.Agricultural Economics 42, 87–97.

Tesfaye, A., Brouwer, R., 2011. Testing participation constraints in contract designfor sustainable soil conservation in Ethiopia. Ecological Economics 73, 168–178.

Train, K.E., 2003. Discrete Choice Methods with Simulation. Cambridge UniversityPress.

Tripathi, R., Singh, R., Singh, S., 2005. Contract farming in Potato production: Analternative for managing risk and uncertainty. Agricultural Economics ResearchReview 18.

Umali-Deininger, D., 1997. Public and private agricultural extension: partners orrivals? The World Bank Research Observer 12, 203–224.

Vermeulen, H., Kirsten, J., Sartorius, K., 2008. Contracting arrangements inagribusiness procurement practices in South Africa. Agrekon 47, 198–221.

Wang, H.H., Zhang, Y., Wu, L., 2011. Is contract farming a risk managementinstrument for Chinese farmers? Evidence from a survey of vegetable farmers inShandong. China Agricultural Economic Review 3, 489–505.

Warning, M., Key, N., 2002. The social performance and distributional consequencesof contract farming: An equilibrium analysis of the Arachide de Bouche programin Senegal. World Development 30, 255–263.

Weissleder, L. 2009. Foreign direct investment in the agricultural sector in Ethiopia.Ecofair Trade Dialogue Discussion Papers No. 12. <http://ecofair.boellnet.de/pics/en/FDIs_Ethiopia_15_10_09_c.pdf>.

Williamson, O.E., 1979. Transaction-cost economics: the governance of contractualrelations. Journal of Law and Economics 22, 233–261.

Wilson, J., 1986. The political economy of contract farming. Review of RadicalPolitical Economics 18, 47–70.

Wolf, S., Hueth, B., Ligon, E., 2001. Policing Mechanisms in Agricultural Contracts.Rural Sociology 66, 359–381.