The role of time preferences in contract breach: Evidence from Chinese poultry farmers participating in contract farming

HOU Jing ,ZHOU Li ,Jennifer IFFT ,YING Rui-yao

1 Business School,Jiangsu Normal University,Xuzhou 221116,P.R.China

2 College of Economics and Management,Nanjing Agricultural University,Nanjing 210095,P.R.China

3 Department of Agricultural Economics,Kansas State University,Manhattan,Kansas 66506,USA

Abstract Farmers’ contract breach behavior is cited as one of the major stumbling blocks in the sustainable expansion of contract farming in many developing countries.This paper examines farmers’ contract breach decisions from the perspective of time preferences.The empirical analysis is based on a household survey and economic field experiments of poultry households participating in contract farming conducted in Jiangsu Province,China.A discounted utility model and a maximum likelihood technique are applied to estimate farmers’ time preferences and the effect of time preferences on contract breach in the production and sales phases are explored with a bivariate probit model.The results show that,on average,the poultry farmers in the sample are generally present biased and impatient regarding future utility.The regression results show that farmers with a higher preference for the present and a higher discount rate are more likely to breach contracts,and time preferences play a greater role in the production phase than in the sales phase.When considering heterogeneity,specific investments and transaction costs promote contract stability only for farmers with a low degree of impatience.Moreover,compared with large-scale farmers,small-scale farmers’ contract breach decisions are more significantly affected by their time preferences.These results have implications for contract stability policies and other issues that are impacted by the linking of behavioral preferences to agricultural decisions.

Keywords: time preferences,contract breach,contract farming,economic field experiments,China

1.Introduction

Contract farming is the dominant form of vertical coordination that contributes to the transition to modern agriculture (Wanget al.2014;Maertens and Velde 2017).The Food and Agriculture Organization of the United Nations defines contract farming as “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 2020).Participating in contract farming by smallholders may be beneficial for accessing better production technology (Key and McBride 2003;Ragasaet al.2018),reducing price risks and transaction costs (Fukunaga and Huffman 2009;Wanget al.2014;Bellemareet al.2021),and receiving higher returns(Trifkovic 2016;Soullier and Moustier 2018;Dubbert 2019;Yanget al.2021).However,some studies have demonstrated that the contractual relationship between a principal (firm) and agents (farmers) is not stable,and there is a high rate of contract breach in contract farming in practice (Guo and Jolly 2008;Wanget al.2011;Barrettet al.2012).For example,when the spot market price exceeds the contracted price,farmers may sell their contracted production to obtain a higher price from the spot market than from the contracting firm.Farmers’contract breach behavior is cited as one of the major stumbling blocks in the sustainable expansion of contract farming in many developing countries,including China.

This paper investigates the role of Chinese poultry farmers’ time preferences in their contract breach decisions,with the goal of providing empirical evidence for promoting the stability of contractual relationships.Generally,the main factor influencing contract breach is whether the benefits outweigh the costs (Gowet al.2000).When farmers choose to breach a contract in favorable market conditions,they obtain a short-term speculative profit (i.e.,contract breach benefits),but their behavior may be discovered by the contracting firm,resulting in substantial losses (i.e.,contract breach costs).In China,a complete breach by farmers is uncommon,and the most common type of breach is to “side sell” a small percentage of production or purchase a small portion of inputs from the market rather than the contracting firm.Farmers’benefits from the contract breach are relatively small in such situations.However,once the farmers’ contract breach behavior has been exposed,it may lead to longterm costs of the contract breach,which include the value of losses resulting from termination of the relationship and damage of the farmer’s reputation in local social networks(Guo and Jolly 2008).From this perspective,the longterm costs of contract breach may be greater than the short-term benefits.This presents a puzzle as to why farmers would implement such opportunistic behavior by breaching contracts,considering the substantial costs of contract breach.

Time preferences -how individuals weigh current versus future rewards -are likely to be crucial for understanding contract breach,given the long time intervals between the benefits and costs resulting from contract breach.Generally,farmers will compare the short-term benefits they can achieve by breaching contracts with the discounted expected future utility stream that they would lose (Guo and Jolly 2008).Farmers with higher levels of time preferences pay more attention to immediate interests and undervalue future utility;thus,although the potential costs of contract breach are substantial,farmers have incentives to breach contracts when the benefits of contract breach are higher than the present value of the expected costs.Consequently,the expected costs of contract breach are sensitive to the producers’ discount factors,which may influence a farmer’s decision of whether to breach contracts.Moreover,poor households in developing countries are generally biased to the present (Tanakaet al.2010),and poverty is not only found to be associated with high levels of impatience (Andersonet al.2004;Nguyen 2011),but also with stress and negative affective states,leading to adverse economic behavior such as short-sighted decision making (Haushofer and Fehr 2014).This shortsighted decision making is reflected in the production and sales process of farmers.Consider the example of the poultry industry,with its short production cycle and high price risk.Once price fluctuation is sufficiently large during the three-month production period,some contract farmers will not be patient enough to wait for the firms’ settlement date,and tend to sell some products to the market secretly.Moreover,when chickens become sick (e.g.,pullorum disease),some contract farmers -particularly those far from contracting firms -have no patience to wait for guidance from the technicians and tend to purchase veterinary drugs from nearby stores or retailers.Therefore,the farmers’ contract breach behavior is affected by their impatience.

Many scholars have studied the determinants of farmers’ decisions to breach a contract,which include specific investments (Williams and Karen 1985;Kumaret al.2013),transaction costs (Haji 2010),reputation mechanisms (Kreps and Wilson 1982;MacLeod 2007),and the design of the contract terms (Guoet al.2007;Guo and Jolly 2008;Kunteet al.2017).However,few studies provide a clear account of the role of time preferences in contract breach.Previous research has shown that time preferences play an important role in the adoption of new technologies and programs,machinery investments,input choices and farm rotations,all of which involve short-and long-run benefits and costs (Dufloet al.2011;Duquetteet al.2012).Nevertheless,empirical studies are rare due to the inherent complications in measuring individuals’preferences (Shavitet al.2014).Therefore,in this study,we attempt to obtain farmers’ time preferences from economic field experiments and analyze the role of their time preferences in contract breach by an econometric model.

The data for this study are derived from a household survey and economic field experiments on poultry farmers conducted in Jiangsu Province,China,in 2016.We apply a discounted utility model and a maximum likelihood technique to estimate farmers’ time preferences and explore the effect of time preferences on contract breach in the production and sales phases with a bivariate probit model.The results indicate that the poultry farmers in our sample are generally present biased and impatient regarding future utility.The empirical results show that farmers with a higher preference for the present and a higher discount rate are more likely to breach contracts,and time preferences play a greater role in the production phase than in the sales phase.When considering heterogeneity,specific investments and transaction costs are found to promote contract stability only for farmers with a low degree of impatience.Moreover,compared with large-scale farmers,small-scale farmers’ contract breach behavior is more significantly affected by time preferences.The results of this study provide several contributions.First,this study applies economic field experiment approaches to a contract farming field that has received little attention so far from this set of tools.In particular,we extend our experimental results to the contract breach decisions;thus,the findings from this study bridge the gap between field experiments and real-world behavior regarding farmers’ contract breach decisions.Second,traditional contract theory usually focuses only on breach of contract in the sales phase,while this study expands on that research by considering farmers’ contract breach decisions in the production phase.Third,this study expands the literature on contract breach by including farmers’ time preferences,specific investments,transaction costs,and farm scale as the factors that influence contract breach under the same empirical framework,so it also provides reference for improving the design of contract terms.In summary,this research contributes to an understanding of the degree to which time preferences predict individual contract breach behavior and the results have implications for contract stability policies and other issues that are impacted by the linking of behavioral preferences to agricultural decisions.

The remainder of this paper includes four sections.Section 2 presents the conceptual framework.Section 3 describes the data and methods,including sample selection,design of the field experiments,estimation of the preference parameters,and the empirical strategy.Section 4 presents and discusses the empirical results,while Section 5 summarizes the results and concludes with specific policy implications and directions for future research.

2.Conceptual framework

This section introduces a simple conceptual framework through which time preferences and the role of time preference parameters in the contract breach decision process can be understood.

2.1.Measurement of time preferences

Time preferences are described as a psychological attitude where individuals prefer the present over the future,usually characterized by a discount factor.Time preferences differ among people depending on various factors,such as an individual’s level of patience (Shavitet al.2014).Some studies have used the exponential discounting model to describe discount rates,in which it is assumed that subjects have constant discount rates for different time horizons.However,this model is often contradicted by experimental data (Fredericket al.2002).Some studies have focused on providing proof that discount rates are not constant but decrease over time (e.g.,O’Donoghue and Rabin 1999;Loewenstein and O’Donoghue 2004;Benhabibet al.2010;Wanget al.2016);that is,discount rates exhibit a “present bias” or preference for immediate utility.These characteristics can be expressed by a quasi-hyperbolic discounting model defined by Laibson (1997) and O’Donoghue and Rabin(1999).Quasi-hyperbolic discounting has been effectively applied to study contract design,retirement planning,social security systems,tax policy,work environment and other systems (e.g.,Diamond and Koszegi 2003;DellaVigna and Malmendier 2004;Schwarz and Sheshinski 2007;Tanakaet al.2010;Guo and Krause 2015).In particular,scholars have used quasi-hyperbolic specification to characterize and estimate farmers’ time preferences based on experimental data (e.g.,Tanakaet al.2010;Nguyen 2011;Liebenehm and Waibel 2014).

Therefore,in this study,the quasi-hyperbolic discounting model was used to characterize farmers’ time preferences (Laibson 1997;O’Donoghue and Rabin 1999;Liebenehm and Waibel 2014).The discount factor for immediate and delayed utilities is defined as:

whereβis a parameter reflecting present bias,ris the parameter for discount rate,trepresents time,βδdenotes the discount factor between now and the next period,andδrepresents the discount factor between any two future periods.The quasi-hyperbolic discounting is reduced to exponential discounting whenβ=1.We interpretβas an individual’s preference for the present,i.e.,the individual prefers the immediate reward over all future rewards.The smaller the value ofβ,the stronger the preference for the present.

The conventional method for estimating the time preference parameters requires participants to make a series of choices between two options (e.g.,receivingxCNY today oryCNY fortperiod in the future).Previous studies of time discounting have generally and implicitly assumed that individuals are risk-neutral.However,Andersenet al.(2008) assert that the estimated discount rates are significantly biased upward when assuming risk neutrality.Therefore,a more proper equation for estimating the time preference parameters should include a utility function (Nguyen 2011).

The existing literature suggests that farmers’ risk preferences are better captured by prospect theory,rather than expected utility theory,in the context of agricultural decision-making (Tanakaet al.2010;Nguyen 2011;Liu 2013;Liu and Huang 2013;Liebenehm and Waibel 2014;Maoet al.2019).As Liu (2013) points out,Chinese farmers generally have an expected income level that they intend to achieve,and they are more sensitive to losses than to gains at such an expected income level;thus,parameters other than risk aversion,such as loss aversion and probability weighting,should also be considered.Therefore,a farmer’s utility under prospect theory is defined as (Tversky and Kahneman 1992):

In eq.(2),xandyare the payoffs,andpand 1-pare the corresponding probabilities associated with these payoffs,v(x) is the value function,andπ(p) is the probability weighting function.Variableσreflects the curvature of the value function,and is interpreted as a proxy for risk aversion (Liebenehm and Waibel 2014).The higher the risk aversion parameter (σ),the lower the levels of risk aversion.The individual is risk-averse ifσ＜1,risk-loving ifσ＞1,and risk-neutral ifσ=1.Variableλrepresents the degree of loss aversion (Tversky and Kahneman 1992).Variableαreflects the accuracy of assessing probability events,and is associated with overweighting small probability events and underweighting large probability events (Prelec 1998).Ifσ＜1 andλ＞1,the value function is S-shaped and steeper for losses than for gains (Tversky and Kahneman 1992).Ifα＜1,the probability weighting function is inverted S-shaped,implying that individuals overweight small probabilities and underweight large probabilities.However,ifα=1 andλ=1,the utility function under prospect theory is reduced to the expected utility specification.

Consequently,by incorporating the quasi-hyperbolic specification and the prospect theory into a utility function,the discounted utility model (Loewenstein and Prelec 1992) is constructed as:

whereD(ti) andPT(xi) are the quasi-hyperbolic discounting function and the utility function under prospect theory,respectively.

2.2.The role of time preferences in contract breach

We hypothesize that three key time periods (Ti,i=0,1,2) exist for a farmer.In periodT0,the farmer signs a production contract with an agribusiness firm.Next,the farmer engages in production or sales in periodT1,in which the farmer must decide whether to breach the contract.The costs of contract breach are assumed to occur in periodT2.For the farmer,breaching the contract in periodT1results in immediate returns (R) and contract breach costs (C) in periodT2.The returns of contract breach include earning the margin between the market price and the contract price in the sales period,saving feeding costs,and shortening the production period.The costs of contract breach include the value of losses resulting from termination of the relationship and damage to the farmer’s reputation in local social networks.Only when the short-term revenue of contract breach is higher than the present value of the expected cost of contract breach can the farmer breach a contract.For convenience,we assume that the time interval from periodT1to periodT2is the same as the time interval from periodT0to periodT1.According to the quasi-hyperbolic discounting model,when the farmer signs a contract in periodT0,the discount factor between periodsT2andT1isδ,the discount factor between periodsT1andT0isβδ,and the discount factor between periodsT2andT0isβδ2.Thus,the necessary condition for the farmer to plan to breach a contract in periodT0can be written as:R·βδ-C·βδ2＞0.This equation can be simplified to:R＞C·δ.Generally,the value ofRis less thanC·δ;that is,a farmer does not intend to breach a contract at the point of signing the contract with the agribusiness firm,otherwise the farmer is not likely to participate in contract farming.However,when the farmer actually produces or sells in periodT1,the farmer’s discount factor between periodsT2andT1is represented byβδ.At this point,the discounted value is simultaneously determined byβandδ(or the constant discount rater).Thus,the necessary condition for the farmer to breach a contract in periodT1can be written as:R＞C·δβ.This equation becomesC·δβ=C·δifβ=1,implying that the conditions for a farmer to breach a contract are identical in the long term and in the short term;however,C·δβ＜C·δifβ＜1,implying that the longterm and short-term conditions of farmers’ contract breach are distinct.Obviously,because of the existence of the present bias parameter (β),discount factors are higher in earlier periods,assuming the same time interval.WhenC·δβ＜R＜C·δ,the latter equation (i.e.,R＞C·δβ) can be satisfied,but the former equation (i.e.,R＞C·δ) cannot be satisfied.This phenomenon reflects the time-inconsistent preferences between long-term planning and short-term actual behavior (Ye and Cai 2008;Liet al.2016).

Considering these issues,we can infer that the farmers’ decision-making in actual production or sales is shortsighted under quasi-hyperbolic discounting.In other words,farmers are subject to the contradiction between long-term rational planning (i.e.,not breaching the contract due to the high costs) and short-term temptation (i.e.,breaching the contract for speculative profit).When making long-term production or sales plans,farmers are prone to judge short-term interests(e.g.,price margin) and future costs (e.g.,penalty and damage of reputation) more clearly,and thus tend to be rational.However,the existence of present bias makes farmers undervalue the future costs of contract breach and have insufficient patience towards short-term decisions.Moreover,with the degree of present-biased preferences increasing,farmers will pay more attention to the immediate rewards and exhibit greater impatience towards future events.Consequently,the discounted future costs decrease,leading to a substantial increase in the probability of contract breach.We anticipate that higher levels of present-biased preferences will result in a greater probability of contract breach for farmers.In particular,since the production phase occurs before the sales phase,the future costs of contract breach may be discounted more heavily at the time of production than the time of sales.Therefore,farmers are expected to have higher levels of standard discounting in the production phase than in the sales phase,and thus be more likely to breach contracts during the production period.

Additionally,risk preferences may play an important role in farmers’ contract breach behavior.The lower the degrees of farmers’ risk aversion and loss aversion,the smaller their perception and evaluation of contract breach cost;thus,the conditions of contract breach in the two equations above (i.e.,R＞C·δandR＞C·δβ) are more easily satisfied.In this study,we carefully measure and model farmers’ risk preferences -including risk aversion,probability weighting,and loss aversion.Breaching contracts is risky behavior for farmers because they may suffer a large loss resulting from contract breach;thus,we assume that a more risk-averse farmer has a lower tendency to breach the contract.Generally,farmers who fail to evaluate the likelihood of future events are prone to underestimate the probability of contract breach exposure;thus,they are more likely to breach a contract.Furthermore,farmers with higher levels of loss aversion tend to attach greater value to the penalties and losses resulting from contract breach,and thus have a lower tendency to breach a contract.Therefore,farmers’ loss aversion is assumed to be negatively related to contract breach.

In summary,a straightforward reduced form model is:

The dependent variable,breachrefers to the contract breach behavior of farmers who sign a contract with an agribusiness firm.The two independent variables,timeandrisk,are defined as farmers’ time and risk preferences,respectively;Xdenotes other control variables.A detailed description of the variables is presented in Section 3.3.

3.Data and methods

3.1.Sample selection

The data used in this study are derived from a household survey and economic field experiments on Chinese poultry farmers conducted by our research team in 2016.We chose Chinese poultry farmers as a test group due to the high production and consumption levels of poultry meat in China.China has become the world’s second-ranked producer of poultry.China’s poultry meat production was about 22 million tonnes in 2019,accounting for 29% of the national meat production (NBSC 2021).Also in 2019,China ranked the second in global poultry consumption,as China’s broiler meat consumption was about 14 million tonnes,accounting for 14% of global broiler meat consumption (USDA 2020).

The location of this study was Jiangsu Province,China.As a major poultry production province in China,Jiangsu Province has high levels of contract farming.In addition to the well-known leading poultry firms,such as Wens Group and Jiangsu Lihua Animal Husbandry Co.,Ltd.,there are also many medium-and small-sized poultry firms.We selected 11 counties in Jiangsu Province: five counties in the north (i.e.,Dafeng,Pizhou,Donghai,Xuyi,and Suyu),three counties in the middle (i.e.,Haimen,Gaoyou,and Jiangyan),and three counties in the south (i.e.,Taicang,Jintan,and Zhenjiang).In each selected county,we interviewed a leading poultry firm that offers production contracts to farmers.According to the list of contract farmers provided by each selected firm,we randomly sampled and interviewed 30 households from our target population -i.e.,poultry farmers participating in contact farmingby - including households who produced at least 2 000 poultry in 20151The poultry farmers in our study specifically refer to the farmers who produce broilers..The household survey included demographic information,production and sales details,as well as contractual arrangement information.We also conducted economic field experiments with the heads of the households at the end of the survey.Farmers were paid a monetary reward according to the experimental results.The average reward was 50 CNY,which equals approximately half a day’s income for one household in the study area.

Of the farmers surveyed,approximately 12% failed to complete the experiments and chose either Option A or Option B all of the time in the experiments.One major concern is that these farmers could have been illiterate and thus did not understand how the experiments work.If this was the case,their data would add noise to the estimates (Liu 2013).Therefore,we excluded these individuals from the sample,and ultimately obtained 290 usable observations.In our sample,approximately 22%of farmers breached contracts in the production phase,and 19% did so in the sales phase.Overall,the rate of contract breach was 27%.

3.2.Economic field experiments

To better estimate the time preference parameters,we followed Nguyen (2011) by employing a discounted utility model that incorporates quasi-hyperbolic discounting and prospect theory into a single framework.Specifically,we first conducted two field experiments addressing time and risk preferences,respectively;next,we applied a maximum likelihood technique to jointly estimate the parameters of the discounted utility model.

Design of economic field experimentsThe experiments used well-established approaches that have been shown to effectively reveal time and risk preferences in lab and field settings (e.g.,Tanakaet al.2010;Nguyen 2011;Liu 2013;Liu and Huang 2013;Liebenehm and Waibel 2014).We used a “switching Multiple Price List”(sMPL) design,which enforces monotonic switching such that participants cannot switch backward and forward within one series (Andersenet al.2006;Liebenehm and Waibel 2014).

The time preferences experiment was constructed as 15 series of five choices between a smaller reward delivered today (Option A) and a larger reward delivered at a delayed period (Option B).The experimental design is illustrated in Table 1.In every set of three series in this design,the same delayed reward at a delayed period(Option B) is contrasted with the same range of five current rewards (Option A).The delayed reward varies between 1 and 60 CNY,and the delay period varies between 3 days and 3 months.Participants must decide which option is preferred in each scenario.

The risk preferences experiment was constructed as three series of two pair-wise choices.Each scenario consists of a safe reward (Option A) and a risky reward(Option B).The design of the risk preferences experiment is illustrated in Table 2.For example,in scenario 1: if farmers choose Option A,they have a 30% probability of obtaining 20 CNY and a 70% probability of obtaining 5 CNY;but if farmers choose Option B,there is a 10%probability of obtaining 34 CNY and a 90% probability of obtaining 2.5 CNY.The probabilities are explained using 10 cards numbered 1 through 10.In each scenario,participants must make a choice between two options.

Table 1 Structure of the time preferences experiment

Table 2 Structure of the risk preferences experiment

Across the two experiments,participants completed 110 decision tasks in aggregate.The complete tables of the time and risk preference experiments are presented in Appendix A.At the end of each experiment,we paid a real monetary reward to encourage participants to reveal their true preferences (Andersenet al.2006).

Supplementary instructions on the experimentsBefore the start of the experiments,farmers were told they would receive 20 CNY to start if they agree to participate the experiments.Therefore,they had the opportunity to win up to 60 and 850 CNY,respectively,in the time and risk preference experiments;and they might have had the opportunity to lose up to 10.5 CNY in the risk preferences experiment.To ensure that most participants understood how the time and risk preference experiments work,we used a sample card to illustrate the procedure of the experiments and repeated the illustration two times.

At the end of each experiment,one pair of lotteries was randomly chosen to be paid with a monetary reward.For the time preferences experiment,the participants were asked to randomly draw one card out of the pair of 75 numbered cards after the completion of 75 decision tasks.The number on that card determined which scenario would be paid for the monetary reward,and the participants would gain the reward at the respective time in light of their choices.Regarding the risk preferences experiment,we prepared two pairs of numbered cards.After making all 35 scenarios in the experiment,participants were asked to first draw one card out of the first pair that contained 35 numbered cards;then,they drew another card out of the second pair that contained 10 numbered cards.The first card determined which scenario would be paid,and the second card was associated with the probabilities mentioned in the design of the experiment and determined participants’ monetary reward in that particular scenario.

We invited the village’s leading and firm technicians to witness the experiment,so that the farmers could trust us.Particularly,regarding the payment of reward in the time preferences experiment,we offered several alternative types of payment (e.g.,Wechat payment,AliPay wallet,and mobile-phone charge) to deliver money to those who chose to receive their money in the future2WeChat payment and Alipay wallet belong to different thirdparty mobile payment platforms,and they are equivalent to online wallets that can be directly used for consumption or withdrawal.Therefore,the WeChat payment or Alipay wallet can be regarded as directly delivering money to farmers.Mobile-phone charge refers to a way to directly recharge the phone bills to farmers’ mobile phone number,so as to allow farmers to call,send short messages,pay for mobile phone internet traffic,etc.Farmers could freely choose any payment mode according to their own preferences.In addition,farmers were told that technicians would deliver the money to their houses if they selected cash payment,in an attempt to equalize the pure transaction cost of receiving payment immediately or on the delayed date..As suggested by Tanakaet al.(2010) and Liebenehm and Waibel (2014),we announced and assigned a trusted agent to deliver payment.In our case,we specifically selected firm technicians,who often provide poultry raising guidance services for farmers,as the trusted agents.They are well known to all poultry farmers and are trusted,and would deliver the money within the time window selected by the farmers in the time preferences experiment.This option was aimed at eliminating any doubt the participants might have had that they could not obtain future payments if they chose Option B in the experiment.

Therefore,we can express the conditional log likelihood of choosing Option B for each participant as:

Using a maximum likelihood approach,we simultaneously estimated each individual’s time and risk preference parameters in the discounted utility model.The model estimates of preference parameters for a sample are presented in Table 3.Thus,the average individual estimates of time and risk preference parameters can be obtained.The cumulative distribution function (CDF) of time and risk preference parameters are plotted in Fig.1.For the individual estimates of time preference parameters,we found that the mean of the present bias parameter (β) was 0.643 (SD=0.126),and the mean of the discount rate parameter (r) was 0.307 (SD=0.275).Whenβis smaller andris larger,impatience is greater.The average individual estimates of parametersβandrsuggested that farmers have a great preference for the present and a high discount rate;that is,the average poultry farmer in our sample was generally impatient (i.e.,attaching great importance toimmediate rewards).With respect to risk preferences,our average individual estimates of the risk aversion parameterσ(mean=0.795;SD=0.210),the probability weighting parameterα(mean=0.778;SD=0.266),and the loss aversion parameterλ(mean=2.318;SD=2.028)indicated that the average poultry farmer in our sample was risk-averse in the gain domain;unable to correctly assess probability information;and paid more attention to losses than gains.

Table 3 Model estimates of the preference parameters for a sample1)

Fig.1 Cumulative distribution function plots of time and risk preference parameters.ECDF,empirical cumulative distribution function.

In order to further explore the differences in preference parameter estimates between farmers who breach contracts and farmers who fulfill contracts,we generated the descriptive statistics (Table 4),cumulative distribution function plots (Fig.2),and kernel density estimation plots(Fig.3) of the two samples.Thet-test results in Table 4 show that there are significant differences in time and risk preferences between the two samples.On average,contract breachers exhibited a greater preference for the present (smallerβ),higher discount rates (largerr),less risk aversion (largerσ),less loss aversion (smallerr),anda lower tendency to assess probabilities (smallerr).

Fig.2 Cumulative distribution function plots of preference parameters of the different farmers.ECDF,empirical cumulative distribution function.

Fig.3 Kernel density estimates of preference parameters of the different farmers.

Table 4 Comparison of preference parameters between farmers who fulfill the contract and those who breach the contract

3.3.Empirical strategy

Econometric model specificationWe designed this study to investigate the implications of farmers’ time preferences for their contract breach decisions.The dependent variable in our model refers to the contract breach behavior of poultry farmers who sign production contracts with agribusiness firms involved in contract farming.The production contracts in this study specifically refer to a form of vertical coordination in the poultry industry where the contracting firm provides inputs (e.g.,chicks,feed,and veterinary drugs),technical guidance and training,and unified recovery of the final products in the delivery period.The farmers are required to prepay a deposit,provide a poultry house built to a certain standard,and conduct epidemic prevention and feeding according to the unified requirements of the contracting firm (Hou 2020).In China’s poultry industry,although the inputs are provided by the contracting firm,they are provided on credit rather than for free.The price of the inputs is settled in the delivery period.

Previous studies on the fulfilment or breach of contract have generally considered only the sales phase (e.g.,Guoet al.2007;Guo and Jolly 2008;Kumaret al.2013).Motivated by speculative profit,farmers may conceal a portion of their production to reduce the delivery quantities for the contract firm and sell the hidden portions to the market when the market price exceeds the contracted price.Under the production contracts prevalent in this study,the contracting firm stipulates not only that farmers cannot “side sell” the contracted products,but also that farmers cannot purchase inputs through noncorporate channels (e.g.,a veterinary drug stores or small retailers);otherwise,they are deemed to have breached their contracts.In China’s poultry industry,the prices of inputs and outputs are all determined by the contracting firm,rather than following the spot market price (i.e.,a “virtual high-priced” pricing mechanism).Taking Wens Group as an example,the price of inputs provided by the firm is higher than the spot market price,and the price of outputs is also higher than the average market price.That is,the firm simultaneously increases the prices of inputs and outputs.In this way,only when the actual market price greatly exceeds the contracted price will farmers have the incentive to breach a contract in the sales phase.The purpose of this arrangement is to lock the poultry inputs and products in the alliance,making farmers’ speculative behavior in the sales phase unprofitable to a certain extent(Wan and Ou 2010).Under such a pricing mechanism,the price of inputs provided by the firm is generally higher than the spot market price.Hence,farmers may secretly purchase a portion of their veterinary drugs from a small retailer or veterinary drug store to reduce feeding costs.Moreover,another motivation for contract breach in the production phase is increasing the weight of poultry and making the poultry grow faster by obtaining hormone drugs that are not provided by the contracting firm.The issue of farmers using hormones or other drugs to promote poultry growth and shorten the production period is relevant for our study and considered as a production breach.

There are other manifestations of contract breach,such as the delayed delivery of products,unacceptable delivery quality,and diverting a portion of inputs provided by firms to noncontracted crops (Guoet al.2007;Barrettet al.2012).Given that we rarely observed these cases in our survey farms,this study considered only two typical contract breach behaviors: (a) production breach (breachp)and (b) sales breach (breachs).A production breach entails farmers purchasing a portion of their veterinary drugs from sources other than the contracting firm in the production phase.In a sales breach,farmers sell a portion of contracted poultry to someone other than the contracting firm in the sales phase.

Farmers’ contract breach behavior is indicated by a discrete value of 1 or 0.The value 1 means the farmer breached a contract,and 0 otherwise.We could have adopted two independent models to estimate the factors influencing production breach and sales breach.However,farmers’ contract breach behavior in the phases of production and sales are not independent under production contracts but have strong endogenous characteristics.If we developed two independent probit models for each dependent variable,the correlations between the disturbances would be ignored,resulting in inefficiency in model estimation (Greene 2011).This potential problem can be addressed by using a bivariate probit model.

Thus,a bivariate probit model was used in this study to analyze the factors influencing farmers’ contract breach behavior.The generic form is as follows:

wherebreach＊ipandbreach＊isare latent dependent variables and denote farmeri’s contract breach behavior in the production and sales phases,respectively;timeirepresents the key explanatory variables associated with farmeri’s time preferences;Zidenotes a vector of control variables related to farmeri’s risk preferences,household characteristics,production characteristics,corresponding contract terms,and other external factors;εipandεisare error terms,andρmeasures the correlation betweenεipandεis;Ifρ≠0,thenεipandεisare correlated.

The dependent variablesbreachipandbreachisare observed throughbreach＊ipandbreach＊is,respectively.Then the coefficients of the bivariate probit can be estimated by the full information maximum likelihood.

Variable settingsThe main characteristic of interest in this study is the farmers’ time preferences.Following Nguyen (2011),we used experimental economics methods to elicit farmers’ time preferences and measure them by using the following variables.The variableβis defined as the degree of “present bias”,and a smallerβis associated with a larger preference for the present.Individuals with more present-biased preferences show higher levels of impatience in making short-term decisions and tend to pay more attention to the immediate utility(Loewenstein and O’Donoghue 2004;Wanget al.2016).Therefore,we expect that higher levels of present-biased preferences (i.e.,a smallerβ) will result in a greater probability of contract breach among the farmers.Variablerrefers to the standard discount rate corresponding to the long-term discount factor.A higher discount rate(i.e.,a largerr) is associated with undervaluing future events in making long-term decisions (Tanakaet al.2010;Liebenehm and Waibel 2014).Moreover,future utilities are discounted more heavily at the time of production than the time of sales.Thus,we expect that farmers with a higher discount rate are more likely to breach contracts,particularly in the production phase.

In addition to time preferences,other factors may influence contract breach.Previous work indicates that risk preferences play a role in farmers’ agricultural decision-making (e.g.,Wanget al.2011;Liu 2013;Liu and Huang 2013;Wanget al.2014;Brick and Visser 2015;Aliet al.2021).Therefore,we also carefully measured and modeled farmers’ risk preferences -including risk aversion,probability weighting,and loss aversion -to use as control variables in the regressions.The variablerisk aversion(σ) reflects the concavity of the utility function.Breaching contracts is risky behavior for farmers because they may suffer a large loss resulting from contract breach.We assume that a more risk-averse farmer has a lower tendency to breach the contract.The variableprobability weighting(α) represents the accuracy of assessing the probability of events.The smallerαis,the stronger the tendency to overweight unlikely but desirable events and to underweight likely but undesirable events(Liebenehm and Waibel 2014).Given that contracting firms regularly send technicians to inspect and sample products,the probability of being discovered by the firm is relatively high.We expect that farmers who fail to accurately evaluate probability information are prone to underestimating the probability of contract breach exposure;thus,they are more likely to breach a contract.The variableloss aversion(λ) reflects an individual’s sensitivity to loss compared to gain,and is assumed to be negatively related to contract breach.Generally,farmers with higher levels of loss aversion tend to attach greater value to the penalties and losses resulting from contract breach,and thus have a lower tendency to breach contracts.

The distance between a farmer’s farm and a buyer’s trading place,which reflects the transaction costs of a farmer cooperating with a buyer to some extent,has been proposed as an important factor that influences contract stability (Haji 2010).In this study,the variabledistancemeasures the distance from the farmers’ poultry farms to their contracting firms.In order to accurately calculate the distance between two locations,we used the global positioning system (GPS) to collect latitude and longitude data.Generally,the farther the distance to the contracting firm,the higher the transportation costs for farmers(Haji 2010).Moreover,with increasing distance,the supervision of the contracting firm may be weaker.Thus,the distance to the contracting firm is expected to have a positive effect on the farmers’ contract breach behavior.

Contract terms that require farmers to make specific investments have been demonstrated to contribute to contract stability (Williams and Karen 1985;Guo and Jolly 2008).In China,to suppress farmers’ opportunistic behavior,most contracting firms require farmers to deliver a certain amount of deposit as the prepayment for inputs,and they refuse to refund the prepaid deposit if farmers breach contracts (Wan and Ou 2010).In this study,the variabledepositrefers to the prepayment for inputs required by firms,and it can be regarded as a type of specific investment.We assume that the amount deposited by the farmer at the contracting firm has a negative effect on farmers’ contract breach behavior.

The remaining control variables areage,education,labor,experience,farm scale,contract duration,bonus,andprice risk.Descriptive statistics of these variables are shown in Table 5.The average household head is 49 years old and has 7 years of formal schooling,and the average number of years of feeding poultry for contract farmers is approximately 6 years.Regarding the contract terms,the average deposit is approximately 6 CNY per poultry;and the bonus that farmers received from the contracting firms is less than 1 CNY per poultry.Additionally,the average distance to the contracting firm is approximately 17 km.

4.Results and discussion

4.1.Main observations

A bivariate probit model was used to estimate the effect of time preferences on farmers’ contract breach decisions during the production and sales phases.Table 6 presents the regression results3We used bootstrapped methods when estimating the contract breach model of whole sample.Particularly,we pooled contract breach during production and sales phases in a general analysis by using a probit model before separating them.We found that the significances of the core explanatory variables in the two tables were consistent..There are three models in Table 6.In model (1),the regressors only include the core explanatory variables,namely,time preference parameters (βandr);in model (2),the regressors include the time and risk preference parameters (β,r,σ,α,λ);in model (3),all the control variables based on the model (1)are added.

Regarding the coefficient estimates of time preferences,according to the results from all three models in Table 6,one of the time preference indicators,present bias parameter (β),has a significant negative effect on farmers’contract breach behavior in the production and sales phases.In accordance with our expectations,farmers with higher levels of present-biased preferences are more likely to prefer the immediate utility and thus breach a contract.Moreover,in all specifications the other indicator of time preferences,the discount rate parameter (r) that reflects the long-term discount factor,shows a significant positive effect on the production and sales breaches.Also consistent with our expectations,a higher discount rateresults in a greater tendency to breach contracts.These results regarding time preference parameters suggest that farmers with higher levels of impatience are more likely to breach a contract.This viewpoint can be corroborated by the interviews with the farmers in this survey.For example,we found that some farmers purchase veterinary drugs from noncorporate channels due to insufficient patience and quicker access.Specifically,some farmers use hormone drugs such as olaquindox in the production phase to increase poultry growth rates and sell the poultry sooner4Olaquindox is an anti-bacterial and somatotropic hormone drug that increases the growth rate of poultry and prevents diseases such as avian pasteurellosis.Olaquindox is prohibited as a pharmaceutical feed additive in the poultry industry in China,but 4% of farmers in our sample were found to use this drug..From these perspectives,farmers with higher levels of impatience attach greater importance to the immediate benefits (e.g.,increasing the weight of poultry,shortening the production period,and saving feeding cost),and thus are more likely to breach contracts (e.g.,purchasing drugs from a veterinary drug store or small retailer),even if the potential costs of contract breach are substantial5We conducted some robustness checks,and the results show that the estimated coefficients of core explanatory variables are consistent with our main findings (see Appendix B for details)..

Table 5 Descriptions and summary statistics of the variables

Table 6 Bivariate probit model estimation results for the whole sample1)

The data in Table 6,also show that one of the risk preference indicators,the risk aversion parameter (σ),has a significant positive effect on farmers’ contract breach behavior in the production and sales phases.As hypothesized,a less risk-averse farmer is more likely to breach the contract.However,the other two risk preference indicators,namely,the loss aversion parameter (λ) and probability weight parameter (α),are not statistically significant in both phases.These findings are not in line with our expectations.Generally,we expect that loss aversion triggers compliance,i.e.,a more loss-averse farmer tends to attach greater value to the penalties and losses resulting from contract breach,and thus has a lower probability of contract breach.In reality,however,when poultry show signs of disease in the production phase,a more loss-averse farmer may be more inclined to purchase drugs from nearby stores or rush to sell the poultry to avoid the loss from the disease or death of poultry.Thus,the combination of disparate negative and positive effects may lead to the insignificance of loss aversion parameters in the empirical results.Moreover,farmers who fail to evaluate the likelihood of future events tend to underestimate the probability of contract breach exposure and may be more likely to breach a contract;although farmers who can accurately evaluate the likelihood of future events may be better able to assess the probability information about rising prices in the spot market,and thus are more prone to take risky actions (e.g.,breaching a contract).Consequently,the combination of negative and positive effects may lead to the insignificance of the probability weighting parameter in the empirical results.In short,we should have expected these results because in reality different factors can tend to cancel each other out.

Importantly,in order to reflect the scale and distribution of the explanatory variables,we standardized the explanatory variables before model estimation.The results in Table 6 show that the present bias parameter(β) has the greatest influence on farmers’ contract breach behavior,which implies that present-biased preferences may be the root cause of contract breach.In particular,we found that the standardized effect of time preference parameters in the production phase is larger than that in the sales phase.One possible reason for this result is that some poultry farmers who have greater impatience may be more worried about their poultry being vulnerable and falling sick (or dying) during the production phase;thus,they would tend to buy drugs from the nearby market rather than waiting for the guidance of firms.Another explanation may be that the contracted price of inputs is higher than the spot market price and the use of hormone drugs will shorten the production period.This phenomenon would lead to a great motivation for contract breach in the production phase.Moreover,farmers’decision-making in the sales phase occurs in a more distant future than that in the production phase;thus,farmers are more likely to breach contracts during the production period because of discounting.

Another observation from Table 6 is that neither of thedepositnordistancevariables has a statistically significant influence on farmers’ contract breach behavior.This result is contrary to our expectations,and may reflect the lack of homogeneity of the sample.For example,Jiangsu Province has various scales of contracting firms,and the contractual arrangements and enforcement among these firms differ.Thus,future work should focus on exploring these mechanisms through a heterogeneity analysis.

4.2.Subsample regressions

As discussed above,although previous studies have suggested that specific investments and transaction costs should be related to contract stability (e.g.,Williams and Karen 1985;Haji 2010),our empirical findings do not reveal such a relationship.Thus,the mechanisms of specific investments and transaction costs are considered in the context of time preferences,that is,we expect different groups of farmers to use different criteria when deciding to breach contracts.

Following Pennings and Leuthold (2000),we adopted a cluster analysis to test for heterogeneity using squared Euclidean distances and k-medians.Farmers’ time preference parameters (i.e.,present biasanddiscount rate) were included in the cluster analysis,which differentiated two distinct subsamples.Subsample 1 comprises 213 farmers with a relatively low degree of present-biased preference and discount rate.Subsample 2 comprises 77 farmers with a relatively high degree of present-biased preference and discount rate.The former subsample is defined as less impatient farmers and the latter subsample is defined as more impatient farmers.To explore the differences between these two subsamples,we analyzed the characteristics of the farmers regarding their levels of impatience by using the Mann-Whitney U test.The results show that the two subsamples differ significantly in their contract breach behavior.In particular,these two segments significantly differ regardingdepositanddistance.These results may imply that different factors can influence farmers’ contract breach behavior.Therefore,we divided the sample into two segments and estimated the parameters for each segment with a bivariate probit model.The regression results are presented in Table 7.

The data in Table 7 indicate some differences in the two subsamples.Specifically,the variablesdepositanddistanceplay a role in subsample 1,and this result is consistent with the findings of other studies (e.g.,Guo and Jolly 2008;Haji 2010).Had we treated the sample as homogenous,we would have concluded that neitherdepositnordistanceinfluences the farmers’contract breach behavior.However,when considering heterogeneity,these factors are clearly important determinants of contract breach for less impatient farmers.However,with respect to farmers who have more impatience (i.e.,subsample 2),the mechanisms of specific investments and transaction costs do not play a significant role.One possible reason is that farmers with greater impatience attach a higher value to immediate interests and undervalue the costs resulting from contract breach.This phenomenon leads to the role of traditional mechanisms,such as reputation in contract stability for the farmers,having less of an influence on behavior.The results show that the two groups of farmers have different decision structures,and the factors influencing their contract breach decisions differ based on the levels of their time preferences.In summary,the role of specific investments and transaction costs in farmers’ contract breach decisions may be weakened for more impatient farmers,but they do play a role for less impatient farmers.

4.3.Further discussion

In this section,we further explore the characteristics of farmers with different farm scales with respect to their contract breach decisions.In this study,farmers with a farm scale that is higher than the average farm scale are defined as relatively large-scale farmers,otherwise they are defined as relatively small-scale farmers.These two subsamples consist of 104 farmers and 186 farmers,respectively,and the regression results of the subsamples are presented in Table 8.The results show that the present bias parameter (β) has a significant negative effect on the contract breach behavior of both groups of farmers in the production phase;however,in the sales phase,it has no significant effect on the contract breach behavior of large-scale farmers,but it has a significant effect on small-scale farmers.Generally,different farm scales are associated with different agricultural production objectives of the farmers;for example,large-scale farmers may be more focused on long-term benefits,while smallscale farmers may pay more attention to short-termgoals and effectiveness.Compared with large-scale farmers,small-scale farmers have lower incomes,less savings and capital,and higher marginal utility of their immediate income (i.e.,they are more sensitive to the immediate income).Therefore,the higher the degree of present-biased preferences of small-scale farmers is not surprising,as they attach more importance to the shortterm utility rather than the long-term benefits of contract fulfillment.As a result,compared with large-scale farmers,small-scale farmers’ contract breach decisions are more affected by their present-biased preferences.

Table 7 Bivariate probit model estimation results by considering time preference heterogeneity1)

Table 8 Bivariate probit model estimation results for subsamples1)

Moreover,the results show that the risk aversion parameter (σ) has a significant positive effect on the contract breach behavior of both groups of farmers in the production phase.However,in the sales phase,the risk aversion parameter (σ) has no significant effect on the contract breach behavior of large-scale farmers,but it has a significant effect on small-scale farmers.This may be due to the differences in the transaction costs incurred by farmers with different farm-scales when conducting poultry market transactions.Small-scale farmers usually incur a higher information cost shared by unit agricultural products,they have less ability to participate in market competition and negotiation status,and it is difficult for them to take on the costs and risks of transportation in the wholesale market.Therefore,in order to reduce market risk and transaction costs,smallholders with a higher degree of risk aversion are more willing to cooperate with contracting firms for a long time,and thus are less likely to breach contracts.With the increase of farm scale,farmers’ capacity for taking on risk and transaction negotiation will be improved,and their information cost shared by unit agricultural products will also be reduced;thus,the role of the farmers’ risk aversion in their contract breach decision may be weakened.

We also noticed another interesting result in Table 8.The variablecontract durationhas a significant negative effect on the sales breach of small-scale farmers,but a significant positive effect on production and sales breaches of large-scale farmers.This result implies that poultry firms should provide diversified contracts for the different groups of farmers.For example,they could provide long-term contracts to small-scale farmers to help them avoid sales risks and thus reduce the rates of contract breach,although providing long-term contracts to large-scale farmers may be detrimental to their performance.

This study may be limited by using only the reported contract breach,although several steps were taken to mitigate potential bias.Moreover,farmers’ contract breach decisions can be influenced by many other factors in addition to the independent variables used in this study,such as market structure and the level of competition.These factors should be considered in future empirical research.Additionally,future theoretical studies regarding time preferences may attempt to set a conceptual model that clearly outlines the timing and expected amounts of the contract breach decision-making process for different types of farmers (e.g.,sophisticated and naive),as well as the corresponding expectations about the time preference parameters.In such an improved conceptual model,the expected results of different types of farmers under hyperbolic discounting and quasi-hyperbolic discounting could be discussed.

5.Conclusion and implications

The objective of this study was to validate the role of farmers’ time preferences in their decisions to breach contracts.A better understanding of farmers’ time preferences is conducive to improving the design of contract terms and relevant agricultural policies to influence farmers’ decision-making.We conducted a survey and economic field experiments with Chinese poultry farmers and employed a discounted utility model comprised of a quasi-hyperbolic discounting function and a prospect theory-based utility function to estimate the present values of future utility.By using economic field experiments and a maximum likelihood approach,we could elicit and estimate the time preference parameters that reflect the behavioral characteristics of Chinese poultry farmers participating in contract farming.We observed,on average,that the farmers in our sample have a strong preference for the present and a high discount rate.We used a bivariate probit model to capture farmers’ decision-making process as it related to contract breach in the production and sales phases.The main findings of these analyses are that farmers with higher levels of impatience are more likely to breach contracts,and time preferences play a greater role in the production phase than in the sales phase.These results are consistent after the addition of control variables to the model.When considering heterogeneity,specific investments and transaction costs promote contract stability only for farmers with a low degree of impatience.Moreover,the results of this study indicate that the contract breach behavior of small-scale farmers is more significantly affected by time preferences than that of large-scale farmers.

Therefore,future research on contract stability should pay more attention to the influence of factors related to farmers’ time preferences.Experimental studies from developing countries have demonstrated a negative relationship between wealth indicators and impatience;hence,government policies that increase farmers’ income(e.g.,providing farmers more income security and policy insurance regarding poultry raising) may mitigate their present-biased preferences and thereby increase contract stability.Furthermore,the design of contract provisions should consider smallholders’ time horizons in particular.Contract provisions designed to increase the farmers’immediate fulfillment rewards (e.g.,changing the payment time of subsidies and bonuses by installments) may incentivize farmers with high present-biased preferences to abide by their production contracts.In particular,the use of commitment devices may decrease the probability of contract breach.Given that present bias is severe in many cases,farmers may be better off “tying their hands”early on when contract breach appears less tempting.Additionally,we found that farmers with a higher degree of risk aversion are less likely to breach contracts,especially among small-scale farmers.Poultry firms should design contract terms that are conducive to helping the farmers share risks,so as to motivate risk-averse farmers to continuously participate in contract farming and improve their fulfilment rate.For example,firms can provide longterm contracts with fixed price clauses and fixed subsidies to small-scale farmers.Signing such contracts may help smallholders avoid sales risks and improve their perception of the long-term benefits of contract farming participation,and thus reduce the rate of contract breach.

Acknowledgements

This work was supported by the National Natural Science Foundation of China (72003082 and 71573130),the Project of Philosophy and Social Science Research in Colleges and Universities in Jiangsu Province of China (2020SJA1015),the Priority Academic Program Development of Jiangsu Higher Education Institutions,China (PAPD),and the China Center for Food Security Studies,Nanjing Agricultural University,China.

Declaration of competing interest

The authors declare that they have no conflict of interest.

Appendicesassociated with this paper are available on http://www.ChinaAgriSci.com/V2/En/appendix.htm