You are on page 1of 12

Journal of Cleaner Production 276 (2020) 124241

Contents lists available at ScienceDirect

Journal of Cleaner Production


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

Risk attitude, risk perception, and farmers’ pesticide application


behavior in China: A moderation and mediation model
Dan Pan a, Mimi He a, Fanbin Kong b, c, *
a
School of Economics, Jiangxi University of Finance and Economics, Nanchang, 330013, China
b
College of Economics and Management, Zhejiang A&F University, Zhejiang Province, 311300, China
c
Research Academy for Rural Revitalization of Zhejiang Province, Zhejiang A&F University, Zhejiang Province 311300, China

a r t i c l e i n f o a b s t r a c t

Article history: Understanding farmers’ pesticide application behaviors is essential for environmental sustainability and
Received 22 April 2020 food safety in China. Based on a nationally representative survey of 603 rice farmers from seven major
Received in revised form rice-producing provinces in China, this paper constructs a moderation and mediation model to examine
5 August 2020
the causal relationship between risk attitudes, risk perceptions, and farmers’ pesticide application be-
Accepted 14 September 2020
Available online 22 September 2020
haviors. The results show that risk-averse farmers are more likely to use more pesticides. Farmers’
perceptions of the risks posed by pesticides to profit-maximizing factors, namely food quality and human
Handling Editor: Bin Chen health, can decrease their pesticide expenditure, while their perceptions of risks to environmental fac-
tors, namely soil degradation, water pollution, and air pollution, are not significantly associated with
Keywords: their pesticide expenditure. Moreover, their perceptions of risks to food quality and human health can
Risk attitude alleviate the positive effect of risk attitude on pesticide expenditure, and can also play a partial mediating
Risk perception role in the relationship between risk attitude and pesticide expenditure. Risk management tools such as
Pesticide application behavior crop insurance, and educational programs to improve farmers’ risk perception, would be beneficial
Moderating effect
policies to help alleviate farmers’ excessive pesticide use.
Mediating effect
© 2020 Elsevier Ltd. All rights reserved.

1. Introduction and it aims to achieve zero growth in pesticide use by 2020 (MOA,
2015). However, thus far, the effectiveness of these policies is not
China has been the world’s largest pesticide user for several very significant. Although the intensity of pesticide application has
decades, and has experienced a sharp increase in pesticide appli- been decreased slightly in recent years, it is still much higher than
cation (Sun et al., 2012). The intensity of pesticide application in the world average (see Fig. 1).
2017 was 9.95 kilos per hectare, which is 3.75 times larger than the The above evidence poses a puzzle as to why farmers deviate
world average and 1.95 times higher than the amounts applied in from profit-maximizing behavior by using excessive pesticides,
1991 in China (Fig. 1). However, not only has the high intensity of especially considering that decreasing pesticide application would
pesticide application in China been shown to be a main contrib- greatly reduce farmers’ production input costs, improve their in-
uting factor in agricultural non-point source pollution (Hu et al., comes, and benefit their health and environment (Liu and Huang,
2007), water quality degradation (Zhang et al., 2015), and effects 2013). Hence, a thorough understanding of farmers’ pesticide
on human health (Lai, 2017), but it is also widely known to be application behavior is essential to ensure the success of any pol-
related to deteriorating food quality and safety (Zhao et al., 2018). icies aiming to decrease pesticide overuse (Che ze et al., 2020).
Estimates show that 0.2 million people meet with pesticide Previous research has suggested that farmers’ pesticide appli-
poisoning accidents in China annually (Wang et al., 2017). The cation behavior is mostly affected by observable characteristics:
Chinese government has responded to these concerns by imple- individual characteristics, such as age (Wang et al., 2015), gender
menting a series of policies to reverse the growth in pesticide use, (Wang et al., 2017), and education (Wu and Hou, 2012); household
characteristics, including off-farm employment (Zhang et al., 2019),
farm size (Wu et al., 2018), and cooperative membership (Zhou
* Corresponding author. et al., 2018); policy characteristics, including access to public
E-mail addresses: pandan@jxufe.edu.cn (D. Pan), 13155835871@163.com technical training (Huang et al., 2012), government subsidies (Fan
(M. He), kongfanbin@aliyun.com (F. Kong).

https://doi.org/10.1016/j.jclepro.2020.124241
0959-6526/© 2020 Elsevier Ltd. All rights reserved.
D. Pan, M. He and F. Kong Journal of Cleaner Production 276 (2020) 124241

12
China World
10
(kilos per hectare)
Pesticide dosage
8

0
1994

1999

2004

2009

2013
1991
1992
1993

1995
1996
1997
1998

2000
2001
2002
2003

2005
2006
2007
2008

2010
2011
2012

2014
2015
2016
2017
Fig. 1. Change trend of pesticide use level in China and the world from 1991 to 2016
Data source: FAO STAT.

et al., 2015), and governmental regulations (Zhao et al., 2018). three questions:
However, less attention has been paid to the role of psychological
determinants, especially farmers’ risk attitudes and risk percep- i. What is the impact of risk attitude and risk perception on
tions, in farmers’ pesticide application behavior (Wang et al., 2018). farmers’ pesticide application behavior?
Farmers’ risk attitudes are understood as psychological feelings, ii. Can risk perception moderate the relationship between
and they can be divided into three categories: risk-averse (avoiding farmers’ risk attitudes and pesticide application behavior?
any risk under any conditions), risk-neutral (indifferent to risk- iii. Can risk attitude impact farmers’ pesticide application
averse and risk-loving), and risk-loving (having a preference for behavior via risk perception?
taking risks) (Pennings et al., 2002). Feder (1979) has pointed out in
theory that pesticide is a risk-reducing input in that it decreases the This study aims to make two contributions to the current liter-
production variance resulting from destructive pests. Hence, risk ature. First, we implemented a joint analysis framework to inves-
attitude is the main driver of farmers’ pesticide application tigate both risk attitude and risk perception on farmers’ pesticide
behavior. Chen et al. (2013), Liu and Huang (2013) and Gong et al. application behavior, rather than solely focusing on the impact of
(2016) have found robust evidences to support the theory pro- either risk attitude or risk perception. Second, we examined the
posed by Feder (1979), which shows that a risk-averse farmer is moderation and mediation effects of risk perception in the rela-
more likely to use excessive pesticides. tionship between risk attitude and pesticide application behavior
Farmers’ risk perceptions refer to their perceptions of the like- based on a moderation and mediation model. Our findings may
lihood of adverse effects to humans or the environment associated provide valuable suggestions for the Chinese government to reduce
with a specific phenomenon (Asravor, 2019). Certain studies have pesticide dosage, and may also benefit other countries in relation to
come to the conclusion that farmers with a higher level of knowl- sustainable agricultural development.
edge about the negative consequences of pesticide overuse are Our study is based on a representative survey of 603 rice farmers
more likely to use smaller quantities of pesticides (Bagheri et al., from seven major rice-producing provinces in China. Rice is one of
2019; Chen et al., 2013). the most important crops in China, occupying about 30 million
However, while much of the existing literature has investigated hectares of farmland in 2016. In addition, pesticide overuse is very
the impact of either risk attitude or risk perception on farmers’ common in rice production in China (Sun et al., 2012).
pesticide application behavior, less attention has been paid to the The remainder of this paper proceeds as follows. Section 2
influence of both risk attitude and risk perception on farmers’ presents the theoretical framework and related hypotheses. Sec-
pesticide application behavior. In fact, as pointed out by Lopes tion 3 outlines the data and models used, and the selection and
(1987) and Lusk and Coble (2005), an individual’s behavior de- description of variables. The empirical results are shown in Section
pends not just on their risk attitude, but also their risk perception. 4. Section 5 discusses the empirical results. Finally, conclusions are
The interaction between risk attitude and risk perception is a provided in Section 6.
dominant factor influencing human behavior. Furthermore, previ-
ous studies have considered the roles of risk attitude and risk 2. Theoretical framework and hypotheses
perception to be independent and direct, while very little research
has examined the moderation and mediation effects of risk Based on the theoretical method proposed by Lusk and Coble
perception in the relationship between risk attitude and pesticide (2005), we developed the following model to analyze the impact
application behavior. Indeed, as shown by Dessart et al. (2019), of risk attitude and risk perceptions on farmers’ pesticide applica-
there is a causal ordering from risk attitude to risk perception to tion behavior.
human behavior. Supposing farmers’ utility function is UðWÞ, where W repre-
To address the two problems described above, this paper con- sents farmers’ wealth. Further assuming farmers’ perceived
structs a moderation and mediation model to examine the causal outcome of using pesticide can be denoted by a random variable Z
relationship between risk attitude, risk perception, and farmers’ and its variance is s2 , where Eðs2 Þ ¼ 0. As we stated forehead,
pesticide application behavior. Specifically, the study addresses pesticide overuse is risky because it will cause potential damage to
2
D. Pan, M. He and F. Kong Journal of Cleaner Production 276 (2020) 124241

food quality, human health, soil degradation, water pollution, and Previous studies have also shown that farmers’ perceptions of
air pollution, so we define CE as the certainty equivalent, which environmental damage are not significantly associated with the
represents the amount of money that makes farmers indifferent adoption of sustainable agricultural practices (Knowler and
between using pesticides with outcome Z and being paid with a Bradshaw, 2007; Chen et al., 2018). Therefore, it is hypothesized
certain amount of money CE. In other words, CE is the monetary that:
value of using pesticides to farmers, which can be written as:
H2. Farmers’ perceptions of the risks posed by pesticides to food
E½UðW þ ZÞ ¼ UðW þ CEÞ (1) quality and human health are negatively related to pesticide
application.
Taking a Taylor series extension of W on both sides of Equation
(1), CE can be rewritten as: H3. Farmers’ perceptions of the risks posed by pesticides in terms
of soil degradation, water pollution, and air pollution are not
 00  significantly associated with pesticide application.
1 U ðWÞ
CE ¼  s2  0 (2)
2 U ðWÞ Risk perception also moderates and mediates the relationship
between risk attitude and pesticide application behavior. The
00
ðWÞ
Where rðwÞ ¼ UU 0 ðWÞ denotes the risk aversion degree and s2 moderating effect means that when farmers realize that pesticide
represents the risk attitude degree. Equation (2) indicates that use will entail higher risks, even if they are risk-averse, they may
farmers’ pesticide application behavior is affected by two variables: reduce their pesticide usage. That is to say, the strength of the
relationship between risk attitude and pesticide application
risk attitude rðwÞ and risk perception s2 . We illustrated the impact
behavior is weakened as risk perception increases. The mediating
of these two variables on pesticide use below.
effect implies that risk attitude is indirectly associated with pesti-
Pesticides play a fundamental role in maintaining crop output
cide application behavior via farmers’ risk perception. Therefore, it
by controlling pest infestation (Shankar et al., 2008). Estimates
is hypothesized that:
show that pesticide prevents the loss of one-third of total grain
output globally. In China, the reduction in grain output loss H4. Risk perception weakens the relationship between risk atti-
resulting from pesticide use was 93.6 million metric tons in 2015, tude and pesticide application behavior.
which was 15.1% of total grain output (Sun et al., 2019). Therefore,
H5. Risk perception mediates the relationship between risk atti-
pesticides have been widely recognized as a risk-reducing input
tude and pesticide application behavior.
that can lower the variance in production output (Feder, 1979). As a
risk-reducing tool, pesticide dosage is higher among risk-averse Fig. 2 shows the theoretical model and hypotheses of this paper.
farmers than risk-neutral farmers (Salazar and Rand, 2020). As
shown in Equation (2), we can see that an increase in rðwÞ is
associated with a rise in the value of pesticide use CE and thus
results in more pesticide dosage. Therefore, it is hypothesized that: 3. Data, model, and variables
H1. Risk-averse farmers use more pesticides than risk-neutral
3.1. Sample selection and survey design
ones.
Although pesticide usage can prevent potential output loss, it We used a multi-stage sampling strategy to select the samples.
can also cause damage to food quality and human health and can First, to gain a general picture of farmers’ pesticide application
lead to soil degradation, water pollution, and air pollution. For behavior in China, we chose seven major rice-producing provinces
instance, massive pesticide overuse is linked to lower food quality distributed in different parts of China (Fig. 3): Jiangsu and Shan-
and lower food sale prices, thus decreasing farmers’ income. It is dong provinces, located in eastern China, have an output volume of
reported that only 10e20% of the pesticide can be absorbed by rice 19.314 million tons and 0.881 million tons respectively; Jiangxi,
plants, and the remainder is left in the ground, which then causes Hubei and Henan provinces, located in central China, have an
damage to food quality (Zhou et al., 2018). Pesticide overuse also output volume of rice ranked 3, 5 and 13 respectively; Sichuan
jeopardizes human health, via both pesticide exposure and diet- province, located in western China, has an output volume of rice
related diseases resulting from pesticide residues in food. Esti- ranked 6; Heilongjiang province, located in northern China, is the
mates show that 0.2 million people meet with pesticide poisoning second-largest rice production province. Hence, the seven selected
accidents in China annually (Wang et al., 2017). Lai (2017) has also provinces can nationally represent China. Second, two counties in
verified that a 10% growth in pesticide application will lead to a each of the sampled provinces were selected according to their per
14.22 million yuan increase in medical costs in China. Excessive capita income: one county that has relatively higher per capita
pesticide use has also been shown to have detrimental effects in income and the other has lower per capita income. Third, in a
terms of soil compaction and degradation, water quality, and air similar pattern, two towns in each sampled counties and two vil-
pollution (Sun et al., 2012). Dessart et al. (2019) have pointed out lages in each sampled town were selected. Finally, 10e15 farmers
that, in theory, risk perception is a strong positive predictor of were randomly chosen in each sampled village, resulting in a total
sustainable farming practices. Applying their theory to the context number of 603 rice farmers.
of pesticide application behavior, we can assume that farmers’ Face-to-face interviews were conducted with each sampled rice
perceptions of the risks mentioned above posed by pesticides may farmers during July and September 2017. Prior to the survey, a pilot
also influence their pesticide application behavior. As shown in survey with 20 rice farmers was conducted in Jiangsu province to
Equation (2), an increase in s2 indicates that farmers perceive that pre-test the questionnaire to delete unclear questions and add
pesticide overuse will cause more damage, and it is therefore linked additional questions. A mixed team of eight enumerators, majoring
with a decrease in the value of pesticide use CE and hence less in agricultural economics or ecological economics, was chosen to
pesticide dosage. However, the risk of pesticide damage in terms of collect the data. To improve data accuracy, all the enumerators have
soil degradation, water pollution, and air pollution is mostly a attended a two-day training workshop. In addition, to maintain
matter of public goods. As profit-maximizing individuals, farmers data quality, each farmer received 30 yuan, which was nearly a
may not take these risks into account in their production decisions. quarter of their daily income, to complete the questionnaire.
3
D. Pan, M. He and F. Kong Journal of Cleaner Production 276 (2020) 124241

Risk perception

Risk Risk
perception Risk Risk Risk
perception
perception perception perception
of food of human
of soil of water of air
quality health
damage degradation pollution pollution
damage
H2
H3

Risk attitude
H4, H5

Risk Risk Risk


averse neutral loving H1 Farmers’ pesticide
application behavior

Fig. 2. Theoretical model and hypotheses.

Fig. 3. The location of the study area.

3.2. Model specification Where Pesticidei denotes pesticide application behavior of farmer i;
RiskAi represents farmers’ risk attitude; RiskPi indicates farmers’
3.2.1. Moderating effect of risk perception on the relationship risk perception towards pesticide use, including risk perceptions of
between risk attitude and pesticide application behavior food quality damage, human health damage, soil degradation,
We used an interactive term model to evaluate the moderating water pollution, and air pollution; Controli denotes the control
effect of risk perception. The model is as follows: variables that influence farmers’ pesticide application behavior.
a1 and a2 represents the impact of farmers’ risk attitude and risk
Pesticidei ¼ a0 þ a1 RiskAi þ a2 RiskPi þ a3 RiskAi *RiskPi perception on their pesticide application behavior, respectively;
þ a4 Controli þ ε1 (3) RiskAi *RiskPi is the interaction term of farmers’ risk attitude and
risk perception and a3 denotes the moderating effect of risk

4
D. Pan, M. He and F. Kong Journal of Cleaner Production 276 (2020) 124241

perception on the relationship between risk attitude and pesticide 3.3. Variables selection
application behavior. If a3 and a1 are opposite, then risk perception
will weaken the relationship between risk attitude and pesticide (1) Farmers’ pesticide application behavior. Most previous
application; otherwise, it will strengthen the relationship between studies have defined pesticide application behavior as the
risk attitude and pesticide application. quantity or frequency of pesticide use (Wang et al., 2017; Lai,
2017). We have defined it as pesticide expenditure (yuan per
hectare). This is because there are different types of pesticide
3.2.2. Mediating effect of risk perception on the relationship
in China, and the active content varies among pesticide
between risk attitude and pesticide application behavior
types, so indicators such as the quantity or frequency of
Following the mediation effect model developed by Baron and
pesticide application may not be accurate or reveal real
Kenny (1986), we used the following three models to test the
pesticide application behaviors (Ying et al., 2017).
mediating effect of risk perception on the relationship between risk
(2) Farmers’ risk attitudes. Farmers’ risk attitudes were elicited
attitude and pesticide application behavior:
by asking the following question:
Pesticidei ¼ d0 þ d1 RiskAi þ d2 Controli þ ε2 (4)
There are two operational plans. In the first plan, you will
definitely win 2000 yuan, while in the second plan, you will either
RiskPi ¼ b0 þ b1 RiskAi þ b2 Controli þ ε3 (5)
win 3000 yuan or lose 1000 yuan. Which one will you choose?
If the farmer chooses the first, then he or she is regarded as risk-
Pesticidei ¼ g0 þ g1 RiskAi þ g2 RiskPi þ g3 Controli þ ε4 (6) averse; if the farmer chooses the second, he or she is viewed as risk-
In addition, because the mediator in our paper ee risk loving; if the farmer thinks there is no difference between the two
perception, is a categorical data, ordinary least squares (OLS) plans, and then he or she is regarded as risk-neutral.
method cannot be applied to estimate Equation (5), Logistic
regression is more suitable. Assuming the number of category of (3) Farmers’ risk perceptions. In terms of pesticide risk per-
risk perception is j, therefore, Equation (5) can be rewritten as: ceptions, existing studies have emphasized farmers’ per-
ceptions of risks to human health and food quality from the

PðRiskPi > jjRiskAi Þ


RiskPi ¼ logitPðRiskPi > jjRiskAi Þ ¼ ln ¼ b0 þ b1 RiskAi þ b2 Controli þ ε3 (7)
1  PðRiskPi > jjRiskAi Þ

Where Equation (4) examines the link between risk attitude and application of pesticides (Brainerd and Menon, 2014; Zhou
pesticide application behavior; Equation (5) investigates the rela- et al., 2018). We focus on perceptions of risks regarding not
tionship between risk attitude and risk perception. If d1 and b1 are only human health and food quality but also environmental
significant, Equation (6) continues to be estimated to test whether degradation. For this purpose, we constructed five questions
risk attitude and risk perception both related to pesticide applica- to measure farmers’ risk perceptions:
tion behavior. If b1 and g2 are both significant, then there is a i. Food quality damage risk perception (RiskP-food): Do you
mediating effect of risk perception on the relationship between risk think the frequent and massive use of pesticides causes damage
attitude and pesticide application behavior. Besides, when g1 is to food quality?
significant, then there is a partial mediating effect. When g1 is not ii. Human health damage risk perception (RiskP-health): Do you
significant, then there is a full mediating effect. think the frequent and massive use of pesticides causes damage
In general, the mediating effect of risk perception can be to human health?
denoted as b1 g2 . However, the b1 in Equation (5) is estimated iii. Soil degradation risk perception (RiskP-soil): Do you think the
through Logistic regression, while g2 in Equation (6) is computed frequent and massive use of pesticides causes damage to the
through OLS regression, so b1 needs to be standardized so that it is soil?
on the same scale of g2 . Following MacKinnon (2012), the stan- iv. Water pollution risk perception (RiskP-water): Do you think
dardized b1 can be denoted as: the frequent and massive use of pesticides causes damage to
water quality?
v. Air pollution risk perception (RiskP-air): Do you think the
frequent and massive use of pesticides causes damage to air
SDðRiskAi Þ
bstd
1 ¼ b1 (8) quality?
SDðRiskPi Þ
The answers to all five questions were arranged on a four-point
Where SDðRiskAi Þ is the standard deviation of RiskAi , which can be scale: “no idea”, “no impact”, “slight impact”, and “severe impact”.
written as:
(4) Control variables. This paper examines the link between risk
attitude, risk perception, and farmers’ pesticide application
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
behavior. As can be seen from Equations (4)e(6), the control
p2
SDðRiskPi Þ ¼ b21 varðRiskAi Þ þ (9) variables should impact farmers’ risk perception and their
3 pesticide application behavior simultaneously. Following
2
previous studies by Liu and Huang (2013), Wang et al. (2018),
Where p3 is the residual variance of standard logistic distribution. Sun et al. (2019), and Che ze et al. (2020), we chose the
So the mediating effect of risk perception can be denoted as bstd
1 g2 .
following as control variables: the individual and household

5
D. Pan, M. He and F. Kong Journal of Cleaner Production 276 (2020) 124241

characteristics of gender, health status, education, off-farm farmers’ perceptions of the risks of pesticide use in relation to food
employment, number of agricultural laborers, and share of quality and human health are much higher than in relation to soil
agricultural income; and the farming characteristics of farm degradation, water pollution, or air pollution. For instance, 40.63%
size, rice-planting years, soil fertility, and technical training. and 37.48% of farmers state that pesticide use has a severe impact
on food quality and human health, whereas the corresponding
number drops to 35.99%, 24.21%, and 18.57% for soil degradation,
3.4. Data description water pollution, and air pollution respectively. Regarding the dif-
ferences in pesticide expenditure among different risk perceptions,
3.4.1. Description of the sample it can be concluded that an increase in risk perceptions can lead to a
The summary statistics of the basic characteristics of the 603 decrease in pesticide expenditure. For instance, farmers who state
respondents are shown in Table 1. Most respondents (76.1%) are that pesticide overuse has a severe impact on food quality spend
male and in good health. The majority of farmers have completed 1066.95 yuan per hectare on pesticides, which is nearly 20% lower
less than six years of education, and more than half of them (57.2%) than farmers who state that it has no impact. The ANOVA c2 test
have off-farm employment, which is in line with the phenomenon shows that the differences in pesticide expenditure among
that most Chinese farmers are with low education and engage in different risk perceptions regarding food quality and human health
off-farm jobs (Chadwick et al., 2015). The average number of agri- are statistically significant at the 1% level, whereas they are not
cultural laborers is two, and agricultural income accounts for 33% of statistically significant in relation to risk perceptions of soil
total household income. The average sampled household is degradation, water pollution, or air pollution.
assigned 1.11 ha of cultivated farmland, reflecting the smallholder
characteristics of Chinese agriculture. Farmers have been engaged 4. Empirical estimation results
in rice-planting for more than ten years on average. Most soil is of
medium quality, and nearly half of respondents (42.5%) have 4.1. Impact of risk attitude and risk perception on pesticide
attended rice-planting technical training programs. Therefore, application behavior (H1, H2, and H3)
based on the above statistics, we conclude that our survey samples
can be considered nationally representative. We first examine the impact of risk attitude on farmers’ pesti-
cide application behavior (Model 1 in Table 4). The estimated re-
3.4.2. Farmers’ risk attitudes and pesticide expenditure sults show that risk attitude has a statistically significant and
The descriptive statistics regarding farmers’ risk attitudes and positive effect on pesticide expenditure. Thus H1 is supported. Risk-
pesticide expenditure are displayed in Table 2. It can be seen that averse farmers spend more money on pesticides than risk-neutral
the majority of respondents (69.98%) are risk-averse. The pesticide or risk-loving farmers. Specifically, the estimated coefficient of
expenditure of risk-averse farmers is the highest (1257.51 yuan per risk attitude is 0.096. As shown in Tables 1 and 2, more than half of
hectare); the corresponding number drops to 908.43 yuan per our sampled farmers are risk-averse, and the average value of risk
hectare and 827.11 yuan per hectare for risk-neutral and risk-loving attitude is 2.482. Based on this finding, it can be estimated that if
farmers, respectively. The ANOVA c2 test shows that the differences the average farmer of our sample becomes risk-neutraldthe
in pesticide expenditure among different risk attitudes are statis- average value of risk attitude decreased from current 2.482 to 2,
tically significant at the 1% level, indicating that risk-averse farmers farmers’ pesticide expenditure could be reduced by 4.63%
use more pesticides than their counterparts. [9.6*(2.482e2)].
We then investigate the effect of risk perceptions on farmers’
3.4.3. Farmers’ risk perceptions and pesticide expenditure pesticide application behavior (Model 2 in Table 4). We find that the
Table 3 presents descriptive statistics regarding farmers’ risk perception of risks to food quality and human health has a statis-
perceptions and pesticide expenditure. tically significant and negative effect on pesticide expenditure,
As shown in Table 3, farmers’ perceptions of the risks of pesti- while the perception of risks in terms of soil degradation, water
cide use are very low, with most respondents stating that pesticide pollution, and air pollution has no effect on pesticide expenditure.
overuse has no or only slight impact on food quality, human health, Thus H2 and H3 are also verified. Specifically, the estimated coef-
soil degradation, water pollution, and air pollution. However, ficient of risk perception of food quality damage is 0.064. As

Table 1
Descriptive statistics.

Variables Definition Mean Std. Dev.

Pesticide expenditure The logarithm of pesticide expenditure during one rice-growing season in 2016 6.773 0.775
Risk attitude 1 ¼ risk-loving, 2 ¼ risk-neutral, 3 ¼ risk-averse 2.482 0.828
RiskP-food Risk perception of food quality damage: 1 ¼ no idea, 2 ¼ no impact, 3 ¼ slight impact, 4 ¼ severe impact 3.086 0.933
RiskP-health Risk perception of human health damage: 1 ¼ no idea, 2 ¼ no impact, 3 ¼ slight impact, 4 ¼ severe impact 2.993 0.949
RiskP-soil Risk perception of soil degradation: 1 ¼ no idea, 2 ¼ no impact, 3 ¼ slight impact, 4 ¼ severe impact 2.826 1.077
RiskP-water Risk perception of water pollution: 1 ¼ no idea, 2 ¼ no impact, 3 ¼ slight impact, 4 ¼ severe impact 2.539 1.074
RiskP-air Risk perception of air pollution: 1 ¼ no idea, 2 ¼ no impact, 3 ¼ slight impact, 4 ¼ severe impact 2.483 0.984
Agricultural income Share of agricultural income (%) 0.330 0.313
Off-farm employment Whether the head of the household has an off-farm job in 2016? 1 ¼ yes, 0 ¼ no 0.572 0.495
Agricultural laborers Number of agricultural laborers 2.148 0.957
Farm size Total cultivated farmland area (hectare) 1.11 3.26
Soil fertility 1 ¼ poor, 2 ¼ medium, 3 ¼ good 2.071 0.468
Technical training Whether the head of the household has participated in any rice planting technical trainings in 2016? 1 ¼ yes, 0 ¼ no 0.425 0.495
Gender 1 ¼ male, 0 ¼ female 0.761 0.427
Health 1 ¼ poor, 2 ¼ fair, 3 ¼ good, 4 ¼ excellent 1.909 0.790
Education 1 ¼ less than 1 year, 2 ¼ 1e5 years, 3 ¼ 6e8 years, 4 ¼ 9e12 years, 5 ¼ more than 12 years 2.524 0.990
Rice planting years 1 ¼ less than 3 years, 2 ¼ 3e10 years, 3 ¼ 10e15 years,4 ¼ more than 15 years 3.512 0.824

6
D. Pan, M. He and F. Kong Journal of Cleaner Production 276 (2020) 124241

Table 2
Statistics of farmers’ risk attitudes and pesticide expenditure.

Percentage (%) Pesticide expenditure (yuan per hectare) ANOVA c2 test

Risk-averse 69.98% 1257.51 3.65***


Risk-neutral 8.29% 908.43
Risk-loving 21.72% 827.11

Note: *** denotes statistical significance at the 1% level.

Table 3
Statistics of farmers’ risk perception and pesticide expenditure.

Variables Indicators Do not know No impact Slight impact Severe impact ANOVA c2 test

RiskP-food Percentage (%) 7.46% 17.08% 34.83% 40.63%


Pesticide expenditure 1451.25 1286.40 1072.50 1066.95 3.24***
RiskP-health Percentage (%) 7.13% 23.88% 31.51% 37.48%
Pesticide expenditure 1810.50 1098.15 1220.10 958.50 2.87***
RiskP-soil Percentage (%) 13.76% 29.19% 21.06% 35.99%
Pesticide expenditure 1440.15 1076.70 1105.35 1083.00 1.57
RiskP-water Percentage (%) 20.90% 28.52% 26.37% 24.21%
Pesticide expenditure 1319.85 1108.95 1209.90 924.75 1.26
RiskP-air Percentage (%) 17.25% 35.82% 28.36% 18.57%
Pesticide expenditure 1418.85 1081.80 1131.30 979.80 1.03

Note: *** denotes statistical significance at the 1% level.

Table 4
Impact of risk attitudes and risk perceptions on farmers’ pesticide application behavior.

Variables Model 1 Model 2 Model 3

Risk attitude 0.149***(0.034) 0.096***(0.035)


RiskP-food 0.096***(0.030) 0.064**(0.032)
RiskP-health 0.088***(0.033) 0.059*(0.034)
RiskP-soil 0.009 (0.036) 0.026 (0.037)
RiskP-water 0.007 (0.030) 0.0041 (0.030)
RiskP-air 0.006 (0.040) 0.016 (0.039)
Risk attitude  RiskP-food 0.091***(0.028)
Risk attitude  RiskP-health 0.068*(0.038)
Risk attitude  RiskP-soil 0.051 (0.040)
Risk attitude  RiskP-water 0.013 (0.033)
Risk attitude  RiskP-air 0.028 (0.041)
Agricultural income 0.639***(0.104) 0.642***(0.107) 0.595***(0.107)
Off-farm employment 0.227***(0.066) 0.231***(0.067) 0.208***(0.067)
Agricultural laborers 0.035 (0.030) 0.037 (0.031) 0.031 (0.031)
Farm size 0.829*(0.447) 1.042**(0.466) 0.956**(0.458)
Soil fertility 0.043 (0.054) 0.056 (0.056) 0.037 (0.057)
Technical training 0.360***(0.063) 0.332***(0.064) 0.328***(0.064)
Gender 0.260***(0.059) 0.221***(0.062) 0.214***(0.060)
Health 0.130***(0.035) 0.108***(0.038) 0.107***(0.038)
Education 0.017 (0.033) 0.032 (0.034) 0.036 (0.033)
Rice planting years 0.099**(0.040) 0.106***(0.039) 0.095**(0.038)
Constant 5.957***(0.237) 6.644***(0.259) 6.247***(0.235)
R2 0.241 0.241 0.271

Notes: (1) Standard errors are in parentheses; (2) ***, **, * denote statistical significance at the 1%, 5%, and 10%, respectively.

shown in Table 1, more than half of the farmers thought pesticide Some meaningful results are also observed among the control
use has a severe impact on food quality. Based on this finding, it can variables. The higher the proportion of agricultural income, the
be estimated that if farmers’ risk perception of food quality had more pesticides farmers use. This is consistent with Wang et al.
been enhanced from the average level to the level as those who had (2017), who stated that households with higher agricultural in-
all believed that pesticide use has a severe impact on food qual- comes rely more on agricultural production and therefore use more
itydthe average value of risk perception of food quality damage pesticides to avoid loss of agriculture income. Regarding the off-
increases from the current 3.086 to 4, the pesticide expenditure farm employment, there are two explanations for the impact of
could be reduced by 5.85% [6.4*(4e3.086)]. Similarly, the estimated off-farm employment on pesticide use in existing literature. On one
coefficient of risk perception of human health damage is 0.059. hand, off-farm employment could reduce pesticide use in that off-
According to this finding, farmers’ pesticide expenditure could be farm employment will lead to less time spent on agriculture and a
reduced by 5.95% [5.9*(4e2.993)] should their risk perception of lower probability of adopting sustainable agricultural technologies
human health damage be improved from the average level to the (Pan et al., 2017). On the other hand, off-farm employment could
level as those who had all believed that pesticide use has a severe lead to less pesticide use in that households with a higher pro-
impact on human healthdthe average value of risk perception of portion of off-farm employment are often with a lower proportion
human health increases from the current 2.993 to 4. of agricultural income and thus will apply less pesticide. In our

7
D. Pan, M. He and F. Kong Journal of Cleaner Production 276 (2020) 124241

results, the positive effect of off-farm employment on pesticide use partial mediating effect. H5 is supported.
is larger than its negative effect. Therefore, off-farm employment Further calculations demonstrate that the mediating effect of
shows a statistically significant and positive effect on pesticide std
the perception of risk to food quality is 0.021 (b1 g2 , where b1 ¼
std
expenditure, indicating that farmers who are engaged in off-farm 0:247, g2 ¼ 0.084), accounting for 14.1% (0.021/0.149) of the total
jobs are more likely to overuse pesticides. Farm size has a statisti- effect. That is, 14.1% of the impact of risk attitude on pesticide
cally significant and negative effect on pesticide expenditure, expenditure is via the perception of risk to food quality. Following
indicating that large-scale farmers use fewer pesticides than small- the same calculation method, the mediating effect of the percep-
scale farmers. This result is in line with Wu et al. (2018), who has
tion of risk to human health is 0.015 (bstd std
1 g2 , where b1 ¼ 0:245,
suggested that the benefits of decreasing pesticide use are larger for
g2 ¼ 0.063), accounting for 10.1% (0.015/0.149) of the total effect.
large-scale farmers, who are therefore more likely to use less
That is, 10.1% of the impact of risk attitude on pesticide expenditure
pesticide. Providing technical training to farmers can significantly
is via the perception of risk to human health.
decrease their pesticide expenditure. This is in line with findings by
In addition, we conducted a Sobel test to further examine the
Wang et al. (2018) and Huang et al. (2008). Higher pesticide
significance of the mediating effect (Sobel, 1982). Table 6 shows
expenditure is related to farmers that are female, have bad health
that the mediating effects of perceptions of risks to food quality and
status, and have more rice-planting years. Compared with male
human health are significant at the 5% and 10% levels, respectively,
farmers, female farmers are usually less-educated and less familiar
indicating that there are indeed mediating effects of these two risk
with the pesticide application technologies, and thus will apply
perceptions.
more pesticides in agricultural production (Yang et al., 2019).
Meanwhile, the coefficients of perceptions of risks in relation to
Farmers with higher rice-planting years are more likely to rely on
soil degradation, water pollution, and air pollution are not signifi-
their past pesticide use experience and tend to overuse pesticides
cant in Model 7, indicating that there is no mediating effect of these
(Zhang et al., 2015). Farmers with bad health status are more
three risk perceptions, which is consistent with the results of the
dependent on agricultural production and thus apply more pesti-
moderating effects model.
cides (Wang et al., 2017).

4.2. Moderating effect of risk perception (H4) 4.4. Robustness checks and their results

The above results show that risk attitude has a positive effect on To check the robustness of our results, we performed the
pesticide expenditure, while risk perception has a negative effect. following two robustness checks. First, we are concerned that our
One subsequent question is whether risk perception can moderate results may be biased resulting from omitted variables, so we add
the relationship between risk attitude and pesticide expenditure. more control variables into our models, including farmers’ knowl-
We performed an interaction term model to answer this question. edge about banned pesticides, cooperative membership, and age.
The estimation results are shown in Model 3 in Table 4. Second, some studies argue that data outliners will influence esti-
The coefficients of the interaction term of risk attitude with mation results (Liu and Huang, 2013), so we winsorize the bottom
perceptions of risks to food quality and human health are negative 10% of our continuous variables. Robustness check results of
and statistically significant, indicating that these two kinds of risk moderating effects and mediating effects are presented in Table 7
perceptions can alleviate the effect of risk attitude on pesticide and Table 8, respectively. We find that the coefficients of key vari-
expenditure. H4 is thus corroborated. Therefore, when a risk-averse ables essentially remain unchanged, indicating that our results are
farmer perceives higher risk in relation to food quality and human robust.
health, he or she will lower his or her pesticide expenditure.
Similarly to the results on the impact of risk perception on pesticide 5. Discussion
expenditure, the coefficients of the interaction term of risk attitude
with perceptions of risks regarding soil degradation, water pollu- The above empirical estimation results consistently suggest that
tion, and air pollution are not significant, meaning that there is no risk attitude and risk perception matters in farmers’ pesticide
moderating effect of these three risk perceptions on the link be- application behavior. Herein, we discuss the reasons for the above
tween risk attitude and pesticide expenditure. empirical results.
First, we found that risk aversion leads farmers to increase their
4.3. Mediating effect of risk perception (H5) pesticide use. This result is consistent with Liu and Huang (2013)
and Gong et al. (2016). Knight et al. (2003) have pointed out that
Table 6 presents the results regarding the mediating effect of farmers may seek a balance between expected income and income
risk perception. According to the model specification in Section variance, especially in developing countries such as China. Applying
3.2.2, we first tested the link between risk attitude and pesticide their theory to our context, the main reason why farmers who are
application behavior, and the results are shown in Model 1 in risk-averse have higher pesticide expenditure is that they want to
Table 5. Then we investigated the relationship between risk atti- prevent potential output loss. This reason is also verified by our
tude and risk perception, and the results are displayed in Models statistical analysis of farmers’ attitudes toward pesticide reduction.
2e6 in Table 5; Model 7 in Table 5 shows the results regarding the Specifically, we asked each farmer: If your pesticide use in rice
mediating effect of risk perception. production were decreased by 10%, would your rice output be
The coefficients of risk attitude on perceptions of risks to food reduced? The results in Fig. 4 demonstrate that most risk-averse
quality (Model 2) and human health (Model 3) are significantly farmers (71%) stated that there would be a definite reduction in
positive, and the coefficients of perceptions of risks to food quality rice output, while only 35% of risk-neutral farmers and 19% of risk-
and human health in Model 7 are also significant, indicating that loving farmers gave the same answer.
perceptions of risks to food quality and human health play a partial However, a number of previous studies have provided evidence
mediating role in the relationship between risk attitude and that the actual amount of pesticide use in China is much higher
pesticide application behavior. In addition, the coefficient of risk than the economically optimal level and thus, an increase of
attitude in Model 7 is significantly positive, but it is smaller than the pesticide use will not bring growth in yield. For example, the actual
corresponding coefficient in Model 1, which means that there is a amount of pesticides in rice, cotton, and maize production is 2.3,
8
D. Pan, M. He and F. Kong Journal of Cleaner Production 276 (2020) 124241

Table 5
The results of the mediating effects of risk perception.

Variables First stage Second stage Third stage

Model 1: Model 2: Model 3: Model 4: Model 5: Model 6: Model 7:


Pesticides expenditure RiskP-food RiskP-health RiskP-soil RiskP-water RiskP-air Pesticides expenditure

Risk attitude 0.149*** (0.034) 0.560*** (0.111) 0.553*** (0.103) 0.197 (0.196) 0.149 (0.096) 0.066 (0.097)0.127*** (0.034)
RiskP-food 0.084*** (0.030)
RiskP-health 0.063* (0.034)
RiskP-soil 0.014 (0.036)
RiskP-water 0.000 (0.030)
RiskP-air 0.013 (0.039)
Agricultural income 0.639*** (0.104) 1.072*** (0.284)> 0.268 (0.287) 0.114 (0.278) 0.447 (0.275) 0.467* (0.277) 0.600*** (0.106)
Off-farm employment 0.227*** (0.066) 0.022 (0.176) 0.082 (0.176) 0.156 (0.170) 0.202 (0.168) 0.047 (0.168) 0.230*** (0.067)
Agricultural laborers 0.035 (0.030) 0.184** (0.082) 0.034 (0.088) 0.111 (0.084) 0.077 (0.078) 0.001 (0.081) 0.028 (0.030)
Farm size 0.829* (0.447) 4.736*** (1.280) 0.521 (1.277) 1.688 (1.238) – 0.284 (1.224) 2.088* (1.235) 1.035** (0.450)
Soil fertility 0.0426 (0.0544) 0.188 (0.169) 0.236 (0.170) 0.0610 (0.159) 0.207 (0.163) 0.148 (0.161) 0.042 (0.056)
Technical training 0.360*** (0.063) 0.525*** (0.169) 0.456*** (0.169) 0.415** (0.166) 0.134 (0.162) 0.509*** (0.166) 0.327*** (0.064)
Gender 0.260*** (0.059) 0.292 (0.189) 0.343* (0.196) 0.176 (0.188) 0.311* (0.186) 0.139 (0.185) 0.237*** (0.060)
Health 0.130*** (0.035) 0.127 (0.105) 0.634*** (0.110) 0.455*** (0.101) 0.333*** (0.101) 0.321*** (0.102) 0.107*** (0.037)
Education 0.017 (0.033) 0.126 (0.089) 0.136 (0.086) 0.064 (0.086) 0.136 (0.084) 0.004 (0.084) 0.028 (0.033)
Rice planting years 0.099** (0.040) 0.062 (0.098) 0.107 (0.100) 0.0710 (0.099)< 0.079 (0.096) 0.102 (0.095) 0.092** (0.038)
Prob > chi2 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Pseudo R2 0.241 0.059 0.062 0.025 0.021 0.022 0.257

Note: (1) Standard errors are in parentheses; (2) ***, **, * denote statistical significance at the 1%, 5%, and 10%, respectively.

Table 6 limited knowledge of scientific pesticide management further in-


The results of the Sobel test of mediating effects. duces risk-averse farmers to increase their pesticide use.
Mediators bstd SEðbstd g2 SEðg2 Þ Sobel test Second, we found that farmers’ perceptions of risk to profit-
1 1 Þ
maximizing factors can decrease their pesticide expenditure, while
RiskP-food  0:247 0.049 0.084 0.03 2.445**
their perceptions of environmental risks have no impact on pesticide
RiskP-health  0:245 0.046 0.063 0.034 1.750*
expenditure. This result is in line with previous studies suggesting
1 bstd g
2 std SDðRiskAi Þ that farmers’ sustainable practices are influenced by concerns about
Notes: (1) ZSobel ¼ rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi , SEb1 ¼ SEðb1 Þ ; (2) **,
SDðRiskPi Þ
2 profit-maximizing factorsdtheir perceptions of risks to food quality
1 ðSEðg2 ÞÞ
bstd2 g22 ðSEbstd
2
þ 1 Þ
* denote statistical significance at the 5%, and 10%, respectively.
and human healthdrather than by worries about environmental
factorsdtheir perceptions of risks in terms of soil degradation, water
pollution, and air pollution (Wang et al., 2017; Knowler and
Bradshaw, 2007; Chen et al., 2018). A possible explanation for this
2.8, and 1.2 times higher than the optimal level (Zhang et al., 2015). result is that risks to food quality and human health have a direct
Sun et al. (2020) found that the marginal effect of pesticide impact on farmers’ well-being, such as decreasing income and
expenditure on crop yield decreases as pesticide expenditure ex- damaging health, and thus farmers may take these risks into account
pands. Huang et al. (2015) found that a reduction of pesticide use by in their pesticide use. However, the risks to soil degradation, water
10e35 percent had no adverse effect on rice yields. The contra- pollution, and air pollution are more like public goods, which do not
diction between the findings of these previous studies and the directly affect farmers’ well-being, and thus farmers may not take
farmers’ response to pesticide reduction in Fig. 4 indicate that these risks into account in their production decisions.
farmers lack sufficient knowledge of pesticide use. Most farmers One might be concerned that, apart from the fact that envi-
think that using more pesticide always result in higher yields, and a ronmental risks are mostly public goods, another contributor to the
reduction of pesticide use will definitely lead to a yield loss. This insignificant effect of environmental risks on pesticide use is that

Table 7
The results of robustness checks of moderating effects.

Robustness check 1: Robustness check 2:


Variables Adding more control variables Winsorizing the bottom 10% of data

Model 1 Model 2 Model 3 Model 4

Risk attitude 0.116***(0.034) 0.085**(0.034) 0.098***(0.032) 0.075**(0.032)


RiskP-food 0.074**(0.030) 0.054*(0.032) 0.101***(0.029) 0.080***(0.030)
RiskP-health 0.064*(0.034) 0.058*(0.034) 0.055*(0.031) 0.067**(0.032)
RiskP-soil 0.020 (0.036) 0.033 (0.038) 0.003 (0.036) 0.012 (0.037)
RiskP-water 0.003 (0.030) 0.006 (0.030) 0.047 (0.029) 0.044 (0.030)
RiskP-air 0.006 (0.039) 0.010 (0.039) 0.020 (0.039) 0.020 (0.040)
Risk attitude  RiskP-food 0.089***(0.028) 0.081***(0.027)
Risk attitude  RiskP-health 0.072*(0.038) 0.059*(0.036)
Risk attitude  RiskP-soil 0.048 (0.040) 0.029 (0.038)
Risk attitude  RiskP-water 0.014 (0.033) 0.010 (0.031)
Risk attitude  RiskP-air 0.016 (0.041) 0.009 (0.041)
Constant 6.579***(0.329) 6.437***(0.300) 6.167***(0.270) 5.987***(0.239)
Control variables Yes Yes Yes Yes

Note: Standard errors are in parentheses. ***, **, * denote statistical significance at the 1%, 5%, and 10%, respectively.

9
D. Pan, M. He and F. Kong Journal of Cleaner Production 276 (2020) 124241

Table 8
The results of robustness checks of mediating effects.

Variables Model 1: Model 2: Model 3: Model 4: Model 5: Model 6: Model 7:


Pesticides RiskP-food RiskP-health RiskP-soil RiskP-water RiskP-air Pesticides expenditure
expenditure

Robustness check 1: Adding more control variables


Risk attitude 0.134*** (0.033) 0.496*** (0.112) 0.557*** (0.105) 0.198** (0.098) 0.120 (0.098) 0.073 (0.098) 0.116*** (0.034)
RiskP-food 0.074** (0.030)
RiskP-health 0.064* (0.034)
RiskP-soil 0.021 (0.036)
RiskP-water 0.003 (0.030)
RiskP-air 0.006 (0.039)
Control variables Yes Yes Yes Yes Yes Yes Yes
Prob > chi2 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Pseudo R2 0.253 0.068 0.0690 0.032 0.030 0.032 0.267
Robustness check 2: Winsorizing the bottom 10% of data
Risk attitude 0.122*** (0.031) 0.618*** (0.116) 0.607*** (0.108) 0.182* (0.100) 0.175* (0.100) 0.135 (0.100) 0.098*** (0.032)
RiskP-food 0.101*** (0.029)
RiskP-health 0.055* (0.031)
RiskP-soil 0.003 (0.036)
RiskP-water 0.047 (0.029)
RiskP-air 0.020 (0.039)
Control variables Yes Yes Yes Yes Yes Yes Yes
Prob > chi2 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Pseudo R2 0.257 0.068 0.075 0.029 0.023 0.028 0.277

Note: Standard errors are in parentheses. ***, **, * denote statistical significance at the 1%, 5%, and 10%, respectively.

80% 73% 71%


70% Risk-averse
60% Risk-neutral 57%
Risk-loving
50%
40% 35%
30%
20% 19%
20%
9% 8% 8%
10%
0%
No idea Not reduced Definitely reduced

Fig. 4. Farmers’ perception of output loss with 10% decrease of pesticides.


Note: the survey question is “If your pesticide use in rice production were decreased by 10%, would your rice output be reduced”.

farmers are not clear enough about the specific environmental (Zhou et al., 2018). Based on a nationally representative survey of
risks. We provide the following two explanations to address this 603 rice farmers from seven major rice-producing provinces in
concern. First, in our face-to-face interviews, the environmental China, this paper takes risk perceptions as the moderator and
risks of soil degradation, water pollution, and air pollution were mediator to investigate the impact of risk attitudes and risk per-
carefully elaborated to make sure farmers understand the specific ceptions on farmers’ pesticide application behavior. The paper
meaning of these environmental risks. Second, as shown in Tables 1 draws the following conclusions:
and 2, more than half of our sampled farmers stated that pesticide
use has a slight or severe impact on soil degradation, water pollu- i. Farmers’ risk aversion increases their pesticide expenditure,
tion, and air pollution. However, even though farmers realize the that is, risk-averse farmers are more likely to use more
environmental risks of pesticide use, there is still an insignificant pesticides.
relationship between environmental risks and pesticide expendi- ii. Farmers’ perceptions of risk to profit-maximizing factors,
ture. This finding further confirms that the characteristic of the namely the perceptions of risks to food quality and human
public goods of environmental risks, rather than farmers’ low health associated with pesticides, can decrease their pesti-
cognition of environmental risks, is the major reason for the cide expenditure, while the perceptions of environmental
insignificant effect of environmental risks on pesticide use. risks, namely the perceptions of risks in terms of soil
degradation, water pollution, and air pollution, are not
significantly associated with pesticide expenditure.
6. Conclusions ii. The perceptions of risks to food quality and human health
have a moderating effect on the relationship between risk
Regulating farmers’ pesticide application behavior is a funda- attitude and pesticide application behavior. The negative “risk
mental part of environmental sustainability and food safety in perceptionepesticide expenditure” effect can offset the
China, as pesticide use has not only led to severe environmental
degradation, but has also threatened food safety and human health
10
D. Pan, M. He and F. Kong Journal of Cleaner Production 276 (2020) 124241

positive “risk attitudeepesticide expenditure” effect to some Foundation in Jiangxi Province (No. 2018ACB21004), and the Sci-
extent. ence and Technology Research Foundation of Jiangxi Province (No.
iv. The perceptions of risks to food quality and human health GJJ170349).
also have a partial mediating effect on the relationship be-
tween risk attitude and pesticide application behavior. Sta-
References
tistically, perceptions of risks to food quality and human
health make contributions of 14.1% and 10.1% respectively to Asravor, R.K., 2019. Farmers’ risk preference and the adoption of risk management
the positive impact of risk attitude on pesticide expenditure. strategies in Northern Ghana. J. Environ. Plann. Manag. 62 (5), 881e900.
Bagheri, A., Emami, N., Damalas, C.A., Allahyari, M.S., 2019. Farmers’ knowledge,
attitudes, and perceptions of pesticide use in apple farms of northern Iran:
Our findings have some important policy implications regarding impact on safety behavior. Environ. Sci. Pollut. Res. 26 (9), 9343e9351.
environmental and agricultural sustainability in China: Baron, R.M., Kenny, D.A., 1986. The moderator-mediator variable distinction in so-
cial psychological research: conceptual, strategic, and statistical considerations.
J. Pers. Soc. Psychol. 51 (6), 1173.
i. Given that risk-averse farmers are more likely to use excessive Brainerd, E., Menon, N., 2014. Seasonal effects of water quality: the hidden costs of
pesticides, there seems to be a need to offer farmers some risk the Green Revolution to infant and child health in India. J. Dev. Econ. 107,
management tools, such as crop insurance, to mitigate their 49e64.
Chen, R., Huang, J., Qiao, F., 2013. Farmers’ knowledge on pest management and
potential output loss from reducing pesticides. This is especially pesticide use in Bt cotton production in China. China Econ. Rev. 27, 15e24.
important in China, where access to the formal insurance mar- Chen, X., Zeng, D., Xu, Y., Fan, X., 2018. Perceptions, risk attitude and organic fer-
ket is limited and most farmers are risk-averse. tilizer investment: evidence from rice and banana farmers in Guangxi, China.
Sustainability 10 (10), 3715.
ii. Given that most farmers lack sufficient knowledge of pesticide Cheze, B., David, M., Martinet, V., 2020. Understanding farmers’ reluctance to
use that further induces risk-averse farmers to use more pesti- reduce pesticide use: a choice experiment. Ecol. Econ. 167, 106349.
cides, effective knowledge training programs should be devel- Chadwick, D., Jia, W., Tong, Y., Yu, G., Shen, Q., Chen, Q., 2015. Improving manure
nutrient management towards sustainable agricultural intensification in China.
oped and implemented to improve farmers’ knowledge of Agric. Ecosyst. Environ. 209, 34e36.
scientific pesticide management. Dessart, F.J., Barreiro-Hurle , J., van Bavel, R., 2019. Behavioural factors affecting the
ii. The confirmed role of perceptions of risks to food quality and adoption of sustainable farming practices: a policy-oriented review. Eur. Rev.
Agric. Econ. 46 (3), 417e471.
human health in relation to pesticide expenditure, and the
Fan, L., Niu, H., Yang, X., Qin, W., Bento, C.P., Ritsema, C.J., Geissen, V., 2015. Factors
moderating and mediating effects of these two risk perceptions affecting farmers’ behaviour in pesticide use: insights from a field study in
on the relationship between risk attitude and pesticide expen- northern China. Sci. Total Environ. 537, 360e368.
Feder, G., 1979. Pesticides, information, and pest management under uncertainty.
diture, indicate that educating farmers about the risks to food
Am. J. Agric. Econ. 61 (1), 97e103.
quality and human health associated with pesticide overuse Gong, Y., Baylis, K., Kozak, R., Bull, G., 2016. Farmers’ risk preferences and pesticide
could be an effective tool to ensure agricultural sustainability. use decisions: evidence from field experiments in China. Agric. Econ. 47 (4),
iv The insignificant effect of risk perceptions regarding soil 411e421.
Hu, R., Cao, J., Huang, J., Peng, S., Huang, J., Zhong, X., Zou, Y., Yang, J., Buresh, R.J.,
degradation, water pollution, and air pollution suggests that 2007. Farmer participatory testing of standard and modified site-specific ni-
educational efforts should focus on the economic losses asso- trogen management for irrigated rice in China. Agric. Syst. 94 (2), 331e340.
ciated with pesticide overuse, such as in relation to food quality Huang, J., Hu, R., Cao, J., Rozelle, S., 2008. Training programs and in-the-field
guidance to reduce China’s overuse of fertilizer without hurting profitability.
and human health, rather than on the environmental benefits of J. Soil Water Conserv. 63 (5), 165e167.
reducing pesticides. Huang, J., Xiang, C., Jia, X., Hu, R., 2012. Impacts of training on farmers’ nitrogen use
in maize production in Shandong, China. J. Soil Water Conserv. 67 (4), 321e327.
Huang, J., Huang, Z., Jia, X., Hu, R., Xiang, C., 2015. Long-term reduction of nitrogen
Our study adds to the literature by examining the impact of fertilizer use through knowledge training in rice production in China. Agric.
psychological determinants, especially farmers’ risk attitudes and Syst. 135, 105e111.
risk perceptions, on farmers’ pesticide application behavior. Also, Knight, J., Weir, S., Woldehanna, T., 2003. The role of education in facilitating risk-
taking and innovation in agriculture. J. Dev. Stud. 39 (6), 1e22.
we have identified the moderating and mediating effects of risk
Knowler, D., Bradshaw, B., 2007. Farmers’ adoption of conservation agriculture: a
perceptions. A limitation of this study is that we only focus on rice review and synthesis of recent research. Food Pol. 32 (1), 25e48.
farmers in seven provinces in China. Further studies incorporating Lai, W., 2017. Pesticide use and health outcomes: evidence from agricultural water
pollution in China. J. Environ. Econ. Manag. 86, 93e120.
other crops and larger sample sizes are necessary to test the
Liu, E.M., Huang, J., 2013. Risk preferences and pesticide use by cotton farmers in
external validity of our findings. China. J. Dev. Econ. 103, 202e215.
Lopes, L.L., 1987. Between hope and fear: the psychology of risk. Adv. Exp. Soc.
CRediT authorship contribution statement Psychol. 20, 255e295.
Lusk, J.L., Coble, K.H., 2005. Risk perceptions, risk preference, and acceptance of
risky food. Am. J. Agric. Econ. 87 (2), 393e405.
Dan Pan: Conceptualization, Methodology, Formal analysis, Ministry of Agriculture, 2015. Action Plan for Halting Pesticide Use Increase by
Writing - original draft, Writing - review & editing. Mimi He: 2020. http://jiuban.moa.gov.cn/zwllm/tzgg/tz/201503/t20150318_4444765.
htm. Accessed date: 12 July 2019.
Software, Data curation, Writing - original draft. Fanbin Kong: MacKinnon, D., 2012. Introduction to Statistical Mediation Analysis. Routledge.
Conceptualization, Project administration, Funding acquisition, Pan, D., Kong, F., Zhang, N., Ying, R., 2017. Knowledge training and the change of
Writing - review & editing. fertilizer use intensity: evidence from wheat farmers in China. J. Environ.
Manag. 197, 130e139.
Pennings, J.M.E., Wansink, B., Meulenberg, M.T.G., 2002. A note on modeling con-
Declaration of competing interest sumer reactions to a crisis: the case of the mad cow disease. Int. J. Res. Market.
19 (1), 91e100.
Salazar, C., Rand, J., 2020. Pesticide use, production risk and shocks. The case of rice
The authors declare that they have no known competing in-
producers in Vietnam. J. Environ. Manag. 253, 109705.
terests or personal relationships that could have appeared to in- Shankar, B., Bennett, R., Morse, S., 2008. Production risk, pesticide use and GM crop
fluence the work reported in this paper. technology in South Africa. Appl. Econ. 40 (19), 2489e2500.
Sobel, M.E., 1982. Asymptotic confidence intervals for indirect effects in structural
equation models. Socio. Methodol. 13, 290e312.
Acknowledgments Sun, B., Zhang, L., Yang, L., Zhang, F., Norse, D., Zhu, Z., 2012. Agricultural non-point
source pollution in China: causes and mitigation measures. Ambio 41 (4),
This study was supported by the National Youth Talent Support 370e379.
Sun, S., Hu, R., Zhang, C., Shi, G., 2019. Do farmers misuse pesticides in crop pro-
Program, the National Natural Science Foundation of China (No. duction in China? Evidence from a farm household survey. Pest Manag. Sci. 75
71863016; No. 71673123), the Outstanding Youth of Natural Science (8), 2133e2141.

11
D. Pan, M. He and F. Kong Journal of Cleaner Production 276 (2020) 124241

Sun, S., Zhang, C., Hu, R., 2020. Determinants and overuse of pesticides in grain Natl. Acad. Sci. Unit. States Am. 115 (27), 7010e7015.
production. China Agr. Econ. Rev. 12 (2), 367e379. Ying, R., Zhou, L., Hu, W., Pan, D., 2017. Agricultural technical education and agro-
Wang, J., Chu, M., Deng, Y., Lam, H., Tang, J., 2018. Determinants of pesticide chemical use by rice farmers in China. Agribusiness 33 (4), 522e536.
application: an empirical analysis with theory of planned behaviour. China Agr. Yang, M., Zhao, X., Meng, T., 2019. What are the driving factors of pesticide overuse
Econ. Rev. 10 (4), 608e625. in vegetable production? Evidence from Chinese farmers. China Agr. Econ. Rev.
Wang, J., Tao, J., Yang, C., Chu, M., Lam, H., 2017. A general framework incorporating 11 (4), 672e687.
knowledge, risk perception and practices to eliminate pesticide residues in Zhang, C., Guanming, S., Shen, J., Hu, R., 2015. Productivity effect and overuse of
food: a Structural Equation Modelling analysis based on survey data of 986 pesticide in crop production in China. J. Integr. Agr. 14 (9), 1903e1910.
Chinese farmers. Food Contr. 80, 143e150. Zhang, J., Wang, J., Zhou, X., 2019. Farm machine use and pesticide expenditure in
Wang, Y., Wang, Y., Huo, X., Zhu, Y., 2015. Why some restricted pesticides are still maize production: health and environment implications. Int. J. Environ. Res.
chosen by some farmers in China? Empirical evidence from a survey of vege- Publ. Health 16 (10), 1808.
table and apple growers. Food Contr. 51, 417e424. Zhao, L., Wang, C., Gu, H., Yue, C., 2018. Market incentive, government regulation
Wu, L., Hou, B., 2012. China’s farmer perception of pesticide residues and the impact and the behavior of pesticide application of vegetable farmers in China. Food
factors: the case of Jiangsu Province. China Agr. Econ. Rev. 4 (1), 84e104. Contr. 85, 308e317.
Wu, Y., Xi, X., Tang, X., Luo, D., Gu, B., Lam, S.K., Vitousek, P.M., Chen, D., 2018. Policy Zhou, J., Liu, Q., Liang, Q., 2018. Cooperative membership, social capital, and
distortions, farm size, and the overuse of agricultural chemicals in China. Proc. chemical input use: evidence from China. Land Use Pol. 70, 394e401.

12

You might also like