1. Introduction
Food safety governance is a common issue all over the world [
1]. In recent years, the problem of food quality and safety in China is particularly severe [
2]. China’s food safety incidents, such as Melamine milk powder, lean meat powder, Sudan red salted duck egg, and sick and dead pigs entering the market (the first three events involved the illegal addition of the harmful raw materials, the substances that can promote the growth of lean meat, and the carcinogenic chemicals. The event of dead pigs denotes that the dead pigs which should have been treated innocuously were illegally transported to the market for sale), have received particular public and media attention because they could cause severe health problems (more than 50 thousand people could get sick or die) [
3]. Therefore, the issue of how to effectively reduce food safety risk is very important in front of the Chinese government [
4]. As an international problem, the causes of food safety risk are complex and diverse, including natural and human factors [
5]. The former is mainly subject to the changes of natural ecological environment and human scientific and technological forces, while the latter can be alleviated by improving the food safety governance mechanisms.
As one of the important governance mechanisms reducing the risk of food safety, contract farming has rapidly become the dominant form of vertical coordination in developed countries [
6]. In recent decades, contract farming has become increasingly popular in China’s agri-food supply chain [
7,
8]. Contract farming is conducive to regulating farmers’ production behavior, promoting food safety, and ensuring public health [
9,
10]. For example, farmers participating in contract farming are required to use pesticides in accordance with the national regulations of pesticide use safety (such as the requirements for drug varieties, dosage, and withdrawal time), and agribusiness firms send technicians to guide and supervise farmers’ production behavior on a regular basis. Farmers who participate in contract farming can also access to new agricultural technologies [
11]. It is thus clear that contract farming not only helps farmers to improve their production experience and awareness of safety production, but also restricts farmers’ irregular production behavior and promotes the quality and safety of agricultural products. However, the enforcement of the contract is not optimistic in China’s agri-food supply chain [
12,
13]. For example, Guo [
13] found that the farmers’ contract performance rate was only 38 percent in a survey made in China. The issue of contract enforcement not only affects the sustainable expansion of contract farming in China, but also affects the role of contract farming in reducing the risk of food safety. Therefore, how to ensure the enforcement of the contract in China’s agri-food supply chain is of great significance to improve the quality and safety of agricultural products.
In China, contract arrangements between farmers and processing or distribution firms are the main types of vertical coordination [
14]. Farmers prefer and perform differently based on different contract farming models (i.e., they have autonomous choice in enforcement) [
15]. The existing literature has studied impact factors of contract enforcement, which include reputation mechanism [
16], specific investments [
17,
18], contract terms [
14,
19], and transaction costs [
20,
21]. However, the literature mainly focuses on the impact of external factors on contract performance and generally presupposes that the behavioral agents are rational, ignoring the agents’ personality traits such as risk attitude [
22,
23]. In reality, decision-making is a process of complex cognitive operations influenced by personal as well as environmental variables [
24]. Researchers suggest that risk attitude plays an important role in farmers’ investment decisions and agricultural production decisions [
25,
26]. Beyond decisions, risk perceptions may determine farmers’ health and coping behaviors, both toward the business and the environment [
27,
28,
29]. Thus, it is more appropriate to incorporate farmers’ risk attitude and contract arrangements into the study of contract enforcement. In determining the relationship between risk attitude and agricultural decisions, most empirical researches in the literature typically use two methods when estimating individual risk attitude [
30]. One is to depend on the assumption of objective function and advanced econometric technique to impute the coefficient of risk aversion that will fit the model; however, the assumption of a utility function form and arbitrary heuristics could cause bias. The other method is using wealth as a proxy for risk aversion; but this method could be problematic as it could potentially undermine the role of risk attitude in farmers’ decisions. One contribution to the existing literature is that all risk attitude parameters used in the analysis of contract enforcement are being elicited from an economic field experiment.
This paper attempts to examine the effect of farmers’ risk attitude and contract arrangements on their enforcement in the scenario of China’s agri-food supply chain, with the goal of exploring the governance path for effectively preventing food safety risk. Our data is from a household survey and economic field experiment conducted in Anhui and Jiangsu provinces of China. In this study, we first measure farmers’ risk attitude from an economic field experiment, and then extend their experiment results to the contract performance decisions. The enforcement of the contract in agri-food supply chain is a relatively less researched dimension of contract farming and food safety, and we expect the paper to be a good addition to the existing literature.
2. Conceptual Framework
The contract enforcement decision for a farmer is a tradeoff between the revenues and the costs [
31]. The farmer’s default revenues (
R) are determined by the default quantity (
Q) times the wedge between the contracted price (
pc) and the actual market price (
pm); thus, we can express it as:
. We assume the contractor will not trade with a farmer over some period of time if farmers’ default behavior has been exposed. The farmer’s default costs (
C) depend on the value of losses (
l) that result from termination or non-renewal of the contract, the damage of the farmer’s reputation in local business, and the probability of exposure and punishment for defaulting contract (
P). Thus, we can express these default costs as:
, where
is the value of farmer’s losses in the period
i,
δ is the discount factor, and
P is the probability of exposure and punishment that is affected by the efficiency of legal system and supervision and farmer’s risk attitude. The probability that the farmer is forced to pay compensation for losses resulting from contract default is determined by the cost of contract enforcement for the agribusiness firm and the adequacy of the legal system. The costs of contract enforcement are time, effort, and money that must be spent to take legal action; these are affected by efficiency of the legal system, contract provisions and design, and characteristics of the firm or farmer [
14]. Following the above discussion, we can specify the farmer’s contract enforcement strategy.
When the market price exceeds the contracted price, the farmer will weigh the short-term benefits achieved by defaulting contract (i.e., the default revenues) and the future utility stream of losses (i.e., the default costs). A farmer has incentive to default the contract only when the default revenues exceed the default costs. Considering that the default quantity is relatively small in reality, the decision-making in contract default or performance mainly depends on the default costs; the higher the costs, the greater the probability of contract performance.
From the above discussion, we can infer that the value of
l,
P, and the wedge between
and
will affect farmers’ contract performance rate. Given the state of the legal system and market conditions, these values are determined by contract arrangements, such as contract form, contract type, or contracted price [
14]. Notably, farmers’ risk attitude may affect the subjective utility of costs resulting from contract default and thus influence the contract performance decisions. For example, if a farmer is risk loving or fails to accurately evaluate probability information, he or she tends to undervalue the probability that the contractor will discover the default behavior. Consequently, the potential default costs decrease, leading to a substantial increase in the likelihood of contract default.
To analyze the effect of farmers’ risk attitude and contract arrangements on the enforcement of the contract in China’s agri-food supply chain, we developed a straightforward reduced form model as follows:
where
CE refers to contract enforcement and is measured by farmers’ contract performance decisions;
risk and
arrangements denote key explanatory variables and
Z represents a vector of control variables that are factors influencing contract enforcement. The variables are defined below.
The dependent variable, CE, is indicated by a discrete value of 0 or 1 and is derived from contracted farmers’ response to the survey questions. The value is 1 if farmer i fulfills the contract, and 0 otherwise. When the market price exceeds the contracted price, the farmer may sell a portion of products to the market. Moreover, we also observe whether the farmer carries out standardized production according to the contractual stipulation.
The independent variable,
risk, is defined as farmers’ risk attitude. The common approach uses expected utility (EU) to measure individual risk attitude, in which risk aversion is the sole parameter that determine the shape of the utility function; however, the existing literature suggests that farmers’ risk attitude is better captured by prospect theory rather than expected utility theory in the context of agricultural decision-making [
25,
30]. As Liu [
25] points out, Chinese farmers generally have an expected income level they want to achieve, and they are more sensitive to losses than to gains at the expected income level; thus, parameters other than risk aversion, such as loss aversion, should also be considered in the research. In prospect theory, three variables (i.e., risk aversion, probability weighting, and loss aversion) jointly determine the shape of the utility function [
32]. A farmer’s utility under prospect theory is defined below:
In equation (2),
denotes the expected value over binary monetary outcome
x and
y with corresponding probabilities
p and 1 −
p, respectively. Furthermore, a two-part power function assigns a value for gains (
x > 0) and losses (
x < 0) separately:
where
σ reflects the curvature of the value function and can be interpreted as a proxy for risk aversion (The higher the value function curvature (
σ), the lower the levels of risk aversion. The individual is risk-averse if
σ < 1, risk-loving if
σ > 1, and risk-neutral if
σ = 1.);
λ reflects the degree of loss aversion. Additionally, we assume that farmers place a decision weight on probability information
p that reflects the desirability of uncertain events. Thus, the probability weighting function is defined below:
where
α represents a proxy for probability weighting [
33]. (
α reflects the accuracy of assessing probability events, and is associated with overweighting small probability events and underweighting large probability events.
α < 1 implies that individual does not correctly estimate probabilities, resulting in an inverted s-shaped probability weighting function.) The smaller
α is, the larger the departure from linear probability weighting and hence, the stronger the tendency to overweight small probabilities and underweight large probabilities. Specially, if
and
, the above specification reduces to the standard expected utility specification.
Defaulting contracts is risky behavior for farmers because farmers may suffer a large loss resulting from default; thus, we assume that a more risk-averse farmer has a higher tendency to fulfill the contract. The loss aversion parameter is assumed to be positively related to contract enforcement. Farmers with higher levels of loss aversion tend to attach more value to the penalties and losses resulting from contract default, and thus have a higher tendency to fulfill the contract. Additionally, we expect that probability weighting parameter is positively related to contract enforcement. Generally, farmers who are unable to correctly appraise probability information are prone to underestimate the probability of exposure and punishment by contractor and thus have a low likelihood of contract performance.
The independent variable,
arrangements, is defined as contract arrangements or terms, which is an important factor influencing farmers’ decision-making [
14,
18]. In this paper, we include four indicators that describe the contract arrangements: (a) the form of contract (
form), (b) contract pricing mechanism (
pricing), (c) contract duration (
duration), and (d) bonus clause (
bonus). The variable
form measures the form of contract established between the contractor and the farmer; it takes on the value of 1 if the contract form is oral, and it is 0 if the contract form is written contracts. If a farmer signs an oral contract with an agribusiness firm, the court is difficult to determine the responsibility of default, which may lead to a decrease in farmers’ default costs [
14]; thus, we assume that the probability of contract performance is higher with written contracts than with oral contracts. The variable
pricing refers to the contract pricing mechanism specified in the contract; it has a value of 1 if the mechanism is floor pricing (contract with a floor pricing mechanism offers farmers a minimum price at the beginning of planting; at the time of the delivery, the contractors will use the minimum price if the market price is lower than the minimum price, otherwise they will use the market price), and it is 0 if the mechanism is fixed pricing. We expect that the floor pricing mechanism is associated with a high rate of contract performance because it provides farmers with income guarantees [
14]. The variable
duration denotes the length of the contract between the contractor and farmers. Farmers who sign long-term contracts with the contractor tend to consider the adverse effects of contract default from their own long-term interests; the longer the duration of the contract, the greater the losses from defaulting contract. Thus, we assume that contract duration is positively related to the probability of contract performance. The variable
bonus is binary and has a value of 1 if the contractor offered a bonus to farmers who complied with the contract and 0 otherwise. We expect that contract terms with a bonus are associated with a high rate of contract performance, because such bonus clause will increase the farmers’ income and improve the enthusiasm of farmers cooperating with contractor [
18].
In addition to risk attitude and contract arrangements, contract enforcement may be influenced by other factors such as demographic and socio-economic characteristics. The rest of the control variables consist of the age of household head (age), the education of household head (education), household labor (labor), and family asset (asset), farm scale (scale), planting experience (experience), the distance from farm to contractor (distance), the fluctuation of market price (fluctuation), and regional differences (region).
3. Materials and Methods
3.1. Data Collection
The data was derived from a household survey and economic field experiment of Chinese fruit farmers conducted in 2017. We chose apple and pear, two popular and common fruit in China as representative commodities. China produces more than 20 million tons of apples per year, making it the largest producer in the world. China’s pear production reached 18.7 million tons in 2016, accounting for 75% of world pear production. Therefore, the study on Chinese fruit industry has an important meaning.
Our study was in Anhui province and Jiangsu province. The study county in each province was selected based on a random pre-selection, which was followed by the purposive selection of counties with high rate of contract participation. Specifically, in our survey we chose two counties, Dangshan and Fengxian, which are located in Anhui Province and Jiangsu Province, respectively. In each county that was selected, we randomly picked two towns. In every town, we randomly picked two villages, and in each village we sampled 30 fruit farmers participating in contract farming. Notably, a pilot survey was conducted before formal interview. Prior to the interview process, we conducted several telephone calls with an agribusiness firm to explain the design and the purpose of the farmer survey, its implications, the interview approach, and the sample selection procedures; then we randomly interviewed ten target population—fruit households who have contracted with the agribusiness firm. After the pre-survey was done, the effectiveness of the queries and statements in the questionnaire is tested and improved.
The survey included demographic and socioeconomic information such as age, education, risk attitude, household labor, family asset, farm scale, production and sales details, and contract arrangements. In order to ensure smooth interaction between interviewers and farmers, easy-to-answer questions were presented first in the survey, such as farmers’ individual and household characteristics. The second part of the survey covered the main topic for this study and involved questions and statements designed to assess contract enforcement information. The last part of the survey included a field experiment addressing risk attitude. We applied for the ethical approval of the research involving human participants. The questionnaire was based on a literature review, related theories, peer review and revision, in combination with results of pre-survey, which indicates that it has a good content validity. To test the reliability of the questionnaire, we used Cronbach’s alpha as an indicator to test the reliability of samples. We observed that the overall Cronbach’s alpha coefficient is 0.71, implying that the questionnaire has a good and stable reliability.
In total, 235 usage observations were obtained, of which 118 (50.2%) farmers growing apples and 117 (49.8%) farmers growing pears.
3.2. Experimental Measure of Risk Attitude
Following Liu [
25] and Tanaka et al. [
34], we elicited farmers’ risk attitude by an economic field experiment. The field experiment on risk attitude took the form of a “switching Multiple Price List” (sMPL) design [
35,
36,
37]. MPL is the standard format in which the participant observes a fixed array of paired options and chooses one for each row. sMPL varies the standard MPL by asking the participant to simply choose which row he or she wants to switch at, assuming monotonicity of the underlying preferences to fill out the remaining choices for the participant [
35].
The risk attitude experiment was designed to estimate three prospect theory parameters. The experiment is illustrated in
Table 1. There are three series of paired lotteries. Each paired lottery consists of a safe reward (Option A) and a risky reward (Option B). Take Row 1 for example: If farmers choose option A, they have a 30% probability of winning 20 Yuan and a 70% probability of winning 5 Yuan; if farmers choose option B, there is a 10% probability of winning 34 Yuan and a 90% probability of winning 2.5 Yuan. Farmers must made a choice between two options in each series.
Across the risk attitude experiment, farmers aggregately completed 35 decision tasks. At the end of the experiment, one pair of lotteries was randomly chosen to be paid with real monetary reward. The real monetary reward was expected to encourage participants to reveal their true preferences [
35]. (Notably, the experimental tasks took an average of 10 minutes and the average reward was 35 Yuan.) Specifically, we prepared two pairs of numbered cards. After making all 35 choices in the experiment, participants were asked to first draw one card out of the first pair that contains 35 numbered cards. The number on that card determines which row would be paid the real money. They then drew another card out of the second pair that contains 10 cards numbered 1 through 10. Depending on the lottery they had chosen for that particular row, their rewards were determined by the second numbered card.
We estimated risk attitude parameters based on the farmers’ choices made in the experiment. Specifically, experiment results from Series 1 and Series 2 were used to jointly estimate the curvature of the utility function in the positive domain (
σ) and the probability weighting parameter (
α); experiment results from Series 3 were used to estimate the loss aversion parameter (
λ). For any participant who switch at row N, we could determine whether he or she prefers Option B over Option A at Row N and prefers Option A over Option B at row N-1. Thus, a set of two inequalities from this switching point could be obtained. For instance, if a participant switched from Option A to Option B at row 7 for both Series 1 and Series 2, then the following inequalities should be satisfied:
We could calculate the intervals of
σ and
α that satisfy this pair of inequalities by using the combination of switching points from Series 1 and Series 2; then, we followed Liu [
25] and Tanaka et al. [
34] to take the midpoint of each interval as the point estimate. After completing the estimates of
σ and
α, the inequalities involving
λ could be written out by using the switching point from Series 3. For instance, if a participant switched from Option A to Option B at row 7 for Series 3, then the following inequalities should be satisfied:
Thus, we could obtain the range of λ by solving for the above inequalities. Similarly, we used the midpoint of the interval as the point estimate of λ.
3.3. Empirical Methods
We used a probit model with a maximum likelihood estimator to explore the effect of farmers’ risk attitude and contract arrangements on contract enforcement. The dependent variable, CE, refers to contract enforcement. The independent variables include the key explanatory variables associated with farmer i’s risk attitude and contract arrangements, and a vector of control variables related to farmer i’s demographic and socio-economic characteristics.
The estimating equation is expressed below:
where
F( ) follows a cumulative normal distribution function giving the probability of contract performance and
xi is the hypothesized explanatory variables. The individual likelihood for farmer
i is described as follows:
Thus, we were able to obtain the likelihood function for all observations by assuming independence across individuals:
The probit model was estimated by Stata 15.0 software (StataCorp LP, TX, USA).