Determinants of Farmers’ Confidence in Agricultural Production Recovery during the Early Phases of the COVID-19 Pandemic in China

Abstract

The COVID-19 pandemic has adversely impacted the agricultural supply chain, export of agricultural products, and overall food security. However, minimal exploration has been attempted of farmers’ confidence in agricultural production recovery after the COVID-19 pandemic. Therefore, this study intends to explore the determinants of farmers’ confidence in agricultural production recovery in China during the early stages of the COVID-19 pandemic. More specifically, we analyzed the relationship between risk expectation and social support on the farmers’ confidence in agricultural production recovery by using the ordered probit model. Cross-sectional survey data were collected from February to March 2020 from 458 farm households in the 7 provinces of China to produce the findings. We found that the risk expectation of farmers had a significant negative impact on farmers’ confidence in agricultural production recovery. Social support seemingly had a significant positive impact on the farmers’ confidence in agricultural production recovery, and could play a supportive role in moderating the relationship between risk expectation and farmers’ confidence in recovery. However, social support alleviates the adverse effect of risk expectation on farmers’ confidence in agricultural production recovery to a certain extent. In addition, there were intergenerational differences in the effects of risk expectation and social support on farmers’ confidence in agricultural production recovery. These results imply that policies establishing the risk early warning mechanisms for agricultural production and strengthening the social support from governments and financial institutions are likely to significantly impact agricultural development in the post-COVID-19 era. The formal and informal risk minimization mechanisms should extend their support to vulnerable sectors such as agribusiness.

1. Introduction

The COVID-19 pandemic has had a substantial negative impact on the development of the world economy and the Sustainable Development Goals set by the United Nations [1,2]. According to the World Bank, the actual global gross domestic product declined by 5.2% in 2020. Moreover, the COVID-19 pandemic has also dramatically impacted the global agricultural system [3], negatively impacting the supply, cultivation, transportation, and sales of agricultural products [4]. First, the pandemic has limited the movement of people, leading to labor shortages in agriculture and some disruptions to production [5]. Second, the pandemic has had a negative impact on the agricultural supply chain, and it has become extremely difficult to purchase agricultural materials, transport products, and sell agricultural products because of traffic controls, village closures, and road closures [6] Third, higher production costs, pandemic prevention measures, social security, taxation, and equipment maintenance during the suspension period have increased agricultural production costs [7]. Fourth, the direct consequence of rising costs and liquidity difficulties is cash flow tension. The sudden outbreak of the pandemic broke the subjects’ capital planning, resulting in the interruption of cash flows [8]. A large number of studies have analyzed the impact of the COVID-19 pandemic on the agricultural supply chain [9,10], rural poverty [11,12], export of agricultural products [13,14], food security [15,16], and food prices [17,18]. However, as a sudden public crisis, the COVID-19 pandemic has had a specific impact on the social economy and the physical health of the people; it will also have short-term and long-term negative consequences for people’s mental health [19]. Farmers are a relatively vulnerable group in terms of economic resources and social status. The COVID-19 pandemic has put farmers in a turbulent and transformative agricultural environment, which will increase their uncertainty and bring a particular impact on their confidence in agricultural production [20]. Although the pandemic is generally getting better at present, the fear and worry of psychological trauma caused by the pandemic will persist for some time. The recovery of farmers’ confidence affects all aspects of agricultural production; thus, how to repair the psychological trauma of farmers and quickly return to agricultural production has become an urgent and vital issue to be solved in order to ensure the stable development of agriculture [21,22]. However, most research is based on the macro perspective to consider the countermeasures of agricultural industry recovery. Very few studies have explored farmers’ confidence in recovery from the perspective of their mental health, or their confidence in production recovery in the context of the COVID-19 pandemic. Therefore, in order to accurately analyze farmers’ cognitive laws and behavioral tendencies under the influence of sudden public crisis events, this study used the survey data of the early COVID-19 pandemic in China to explore the farmers’ confidence in agricultural production recovery. The results are helpful to restore farmers’ confidence in government policies, credit, and themselves, and will be beneficial for farmers to resist unexpected shocks and improve the recovery speed, with significant value for understanding the economic recovery in the post-pandemic era.

Many studies have identified that disasters and uncertainties have specific negative impacts [23], and most studies have demonstrated that individual subjective expectations of disaster risks are an essential factor that affects post-disaster recovery [24]. However, existing studies have not formed consistent conclusions about the impact of risk expectations on disaster recovery. Interestingly, some studies have identified that the higher the loss and risk farmers expect from disasters, the lower their enthusiasm for production, and the less favorable it is to return to a standard production trajectory [25]. Several studies have also shown that the “ripple effect” of disaster events increases the expected risk of farmers in the short term (see Naylor et al. [26] for more details). The higher the farmers’ perception of the risk caused by the disaster, the more they tend to invest more actions in enabling them to resume production more quickly [27,28]. At the same time, disasters also cause impairment, which often creates income loss for some farmers [29]. When farmers are unable to bear losses, their production recovery will be affected. However, appropriate external support—such as village committees, township governments, banks, or other institutions—plays a positive role in alleviating the disaster’s impact and restoring agricultural production and life [30,31,32]. For example, Arouri et al. [33] found that credit is one of the main post-management strategies for farmers after disasters. When farmers face any disaster, credit support can help improve their families’ ability to withstand risks [34].

Existing studies have the following three shortcomings: First, most studies focus on the impact of COVID-19 on the agricultural industry from a macro perspective, and few studies have focused on the farmers’ confidence in agricultural production recovery since the outbreak of COVID-19. Second, most studies regard risk expectation and social support as the two independent factors affecting peasant households’ post-disaster recovery (Such as Sharma et al. [35], and Luckstead et al. [36]); in contrast, few studies bring risk expectation and social support into the same logical framework for analysis and discuss the moderating effect of social support. Third, there are apparent differences in farmers’ perceptions of the COVID-19 pandemic and their confidence in recovery between different generations. Discussing this issue from the perspective of intergenerational differences will help improve the pertinence of the relevant conclusions. Therefore, this study uses micro survey data of 458 peasant households and an ordered probit model to analyze the impact of risk expectation and social support on farmers’ confidence in agricultural production recovery, and critically explores whether social support plays a moderating role in the relationship between risk expectation and farmers’ confidence in agricultural production recovery. From this analysis, intergenerational differences are further explored in-depth. Our results can provide theoretical reference for restoring farmers’ confidence in agricultural production and order.

2. Materials and Methods

2.1. Background Analysis and Hypothesis Development

2.1.1. The Impact of Risk Expectation on Farmers’ Confidence in Agricultural Production Recovery

Farmers’ confidence in agricultural production recovery refers to the psychological state in which farmers are optimistic about agricultural production prospects and believe that agricultural income can be restored to average levels before disasters or shocks [37,38,39]. Risk expectation refers to farmers’ subjective assessment and cognition of risk characteristics and disaster impacts in agricultural production [40,41]. According to the risk perception theory, farmers’ subjective perception of risk has a specific impact on their behavioral decisions [42]. A series of risks expected by farmers will have adverse effects on their psychology, such as depression, anxiety, or worry [43]. Supposing that farmers expect that the COVID-19 pandemic may bring more significant risks (such as impacts on the scale, cost, or product market of agricultural production), the negative psychological impact from the pandemic will inevitably be more significant, which is not conducive to farmers’ confidence in agricultural production recovery. In a study of 46 agricultural cooperatives in Shanghai, Gu and Wang [44] found that the risk caused by the COVID-19 pandemic posed significant cognitive impacts on farmers, and raised a certain amount of shock, which ultimately hinders the production recovery. In a review, Xu et al. [45] explored the risk perceptions of COVID-19 and the locust crisis and found that risk perception has limited the capacity of farmers for recovery. According to Perkins and Repper [46], internalities and externalities trigger the perception of risk and slow down the recovery process in terms of production recovery. Hypothesis one (H1) is presented based upon the above assumptions as:

Hypothesis 1 (H1).

Risk expectation has a negative impact on farmers’ confidence in agricultural production recovery.

2.1.2. The Impact of Social Support on Farmers’ Confidence in Agricultural Production Recovery

Social support refers to individuals’ material or emotional help through social networks such as communities or families [47]. According to the social support theory proposed by Cohen and Wills [48], social support can buffer the adverse effects of stress and frustration by providing information, emotional support, and material help; they also stated that the more social support individuals receive, the better they can cope with various challenges from their environment. In a study of disaster-affected households in the Vietnamese Mekong Delta, Nguyen-Trung et al. [32] found that, while connecting social capital during the rehabilitation process, linking social capital was critical for long-term recovery. In a study of Zhangye County, Gansu Province, China, Tan et al. [49] found that social supports can play a mediating role in minimizing external shock and drought risk. Therefore, our study proposes Hypothesis 2 (H2) as:

Hypothesis 2 (H2).

Social support has a positive impact on farmers’ confidence in agricultural production recovery.

Drawing on the practice of Zimet et al. [50] and Gottlieb and Bergen [51], our study divides social support into government support, support from relatives and friends, and financial support, according to the data sources. In this study, government support refers to the government’s various policy incentives for farmers to achieve specific agricultural development goals [52]. The impact of government support on the confidence in agricultural production recovery is shown in the following aspects: (1) policy support, (2) support from relatives and friends, and (3) financial support. The government issued a series of policies to fully support agricultural production recovery after the COVID-19 outbreak. The transportation procedures of agricultural products have been simplified to ensure the standard transfer of farmers’ production materials during the pandemic as service support. In addition, the government has provided agricultural production guidance and training [53]. Support from relatives and friends provides farmers with tangible resources, such as technical support, fundraising, market opportunities, and social connections; it can also provide intangible resources that are hard to replicate, such as emotional motivation, stress relief, and mental support [54]. Financial support can help farmers obtain financing and promote them to purchase seeds, fertilizers, pesticides, and other means of production for resuming agricultural production [55]. Thus, this study offers the subsequent hypotheses as follows:

Hypothesis 3 (H3).

Government support has a positive impact on farmers’ confidence in agricultural production recovery.

Hypothesis 4 (H4).

Support from relatives and friends positively impacts farmers’ confidence in agricultural production recovery.

Hypothesis 5 (H5).

Financial support has a positive impact on farmers’ confidence in agricultural production recovery.

2.1.3. Moderating Effects of Social Support

The COVID-19 pandemic is a global crisis event that has brought certain economic losses to farmers and caused multiple social and psychological shocks [18]. Farmers are subjected to high risks, influencing the success rate of their agricultural endeavors [56]. Social support can also alleviate the adverse impact of risk expectation on farmers’ confidence in agricultural production recovery by improving their risk-response ability and risk tolerance [57]. Specifically, economic support, credit support, and information support that farmers receive from the government, households, and financial institutions can help them to obtain timely resource supplements, realize risk sharing, and improve their risk-response ability [58]. On the other hand, farmers perceive psychological encouragement such as love, trust, and resonance from family, friends, communities, and others, allowing them to view risks more optimistically, with a positive attitude, as challenges that farmers should face directly rather than insurmountable obstacles [59]. This can increase farmers’ confidence in overcoming difficulties, and improve their risk tolerance [60,61]. Based on the above discussion, our study puts forward Hypothesis 6 as:

Hypothesis 6 (H6).

Social support positively affects the relationship between risk expectation and farmers’ confidence in agricultural production recovery.

Government support, support from relatives and friends, and financial support can also alleviate the negative impact of risk expectation on confidence in recovery [49]. Government support can effectively reduce farmers’ uncertain risks and losses through policy and service support [59]. In contrast, the support of relatives and friends can promote the psychological health of farmers through emotional support and encouragement, and increase the risk tolerance of farmers [62]. Financial support mechanisms are associated with providing specific fund-lending support to farmers. The more fund-lending support farmers get, the more they can resist the risks brought by external shocks [63,64]. Based on this premise, this article proposes the following hypotheses:

Hypothesis 7 (H7).

Government support positively affects the relationship between risk expectation and farmers’ confidence in agricultural production recovery.

Hypothesis 8 (H8).

Support from family and friends positively affects the relationship between risk expectation and farmers’ confidence in agricultural production recovery.

Hypothesis 9 (H9).

Financial support positively affects the relationship between risk expectation and farmers’ confidence in agricultural production recovery.

2.1.4. Analysis of Confidence in Agricultural Production Recovery among Different Generations of Farmers

Generations refer to groups with similar birth years and experiences of major social events at a critical growth stage [65]. The intergenerational differences between individuals result from the cohort effect, age effect, and period effect [66]. The period effect mainly reflects the “common changes” of all generations in society, so this article only considers the cohort effect and age effect. Farmers in different generations have differences in their formative experiences, socioeconomic mentality, sources of information, levels of education, and social capital [67,68]. This would seem to provide a theoretical explanation for the intergenerational differences in the impact of farmers’ risk expectations and social support on their confidence in agricultural production recovery.

The cohort effect refers to the differences in socio-environmental changes caused by major historical events in the social development of groups in the same period [69]. The cohort effect causes differences in behavioral cognition between generations [70]. Members of the same generation have experienced the same historical events and external influences, and they often have similar life experiences and core values. However, due to the differences in the historical background and living environment of the first generation and the new generation, they have different perceptions [71]. In addition, compared with earlier generations, the level of education of younger generations is generally higher, and they are more willing to try to accept new things than their elders [72]; they also have more sources of information and social capital. Age has essential impacts on the individual’s experience, which means that the individual’s psychological characteristics are differentiated by age differences in growth [73]. Because of their older age, the older generation of farmers tends to have a more conservative value orientation, while the younger generation of farmers tends to be more adventurous, open to new things, and have a higher risk tolerance [74].

Therefore, there are significant differences in the effects of risk expectation and social support on farmers’ confidence in agricultural production recovery for different generations. Academic circles typically divide farmers’ younger and older generations by the 1980s [75,76,77]; however, after considering “generational effects” with the time lag required to form values, the younger and older generations of farmers are divided by the birth year of the head of their household, with 1975 as the dividing point [69]. Thus, the study puts forward Hypothesis 10.

Hypothesis 10 (H10).

There are significant intergenerational differences in the impact of risk expectation and social support on farmers’ confidence in agricultural production recovery.

2.2. Data Source

Figure 1 represents the theoretical model of the study. To evaluate the impact of COVID-19 on Chinese farmers’ confidence in agricultural production recovery, our research team conducted a household survey from February to March of 2020. The survey was conducted entirely online due to the lockdown measures imposed by the Chinese government. This survey selected Hubei Province as the primary survey area, and Shaanxi, Chongqing, Hunan, Jiangxi, Anhui, and Henan provinces (see Figure 2) were used as supplementary regions for data collection. The main reasons for choosing those areas were as follows: First, Hubei province was the epicenter of the pandemic in China, and its agricultural production has suffered the most. Second, Hubei Province and its neighboring provinces are major agricultural production areas in China; therefore, the region is representative for the study of the pandemic on farmers’ confidence in agricultural production recovery.

A stratified sampling method was adopted to collect the empirical data. First, 2–3 counties were randomly selected from each province. Second, 2 townships were randomly selected from each county, followed by 3 villages from each township, and 4–7 households were randomly selected from each village. Then, the survey was carried out in the form of an online interview. The questionnaire included household head characteristics, family characteristics, planting characteristics, farmer risk expectations, social support, and confidence in agricultural production recovery. A total of 1456 questionnaires were disseminated during this survey, and received 503 responses. Finally, a set of 458 valid and useable questionnaires were selected after eliminating the invalid questionnaires containing missing values and outliers.

Table 1. Descriptive statistics.

Variables and Dimensions	N	%	Variables and Dimensions	N	%
Gender			Education level
Male	382	83.41	Illiterate	52	11.35
Female	76	16.59	Primary	109	23.80
Age			Junior secondary	170	37.12
Under 45	75	16.38	High school/technical secondary school	102	22.27
46–55	202	44.10	Junior college and above	25	5.46
56–65	147	32.10	Household income level
66 and above	34	7.42	Under RMB 50,000	95	20.74
Farm size			RMB 50,000–10,000	186	40.61
Under 3 mu	24	5.24	RMB 100,000–200,000	142	31.00
			Above RMB 2,000,000	35	7.64
3–10 mu	334	72.93
Above 10 mu	100	21.83

Notes: Data source was accessed from February to March of 2020.

shows that the household heads were mainly men (83.41%) over 45 years old (83.62%), while 16.59% of household heads were female. Second, the educational level of household heads was primarily concentrated in junior middle school, with junior middle school and below accounting together for more than half of respondents (72.27%); this reveals that the educational level of household heads is generally low. Third, from the perspective of farm size, farmers with a planting scale of 3–10 mu (1 mu = 1/15 ha) accounted for 72.93% of the total sample. Fourth, from the perspective of household income level, farmers with household incomes of less than RMB 100,000 (during the study period, USD 1 = RMB 7.03) accounted for 61.35% of the total sample, and those with incomes of RMB 100,000 to 200,000 accounted for 31% of the total sample. Overall, the characteristics of sample farmers were essentially consistent with the actual situation in rural areas of China.

2.3. Variable Measurements

2.3.1. Farmers’ Confidence in Agricultural Production Recovery

The explained variable of this study is farmers’ confidence in agricultural production recovery. The study defines the farmers’ confidence in agricultural production recovery as farmers reestablishing their agricultural expectations and their belief that their agricultural income will completely recover to its average level before disasters or shocks. Farmers’ confidence in agricultural production recovery was measured by asking, “Are you confident that agricultural incomes will recover under the COVID-19 pandemic shock?”; the responses to the question ranged from 1 (only a little) to 5 (a great deal) to evaluate farmers’ confidence. Overall, the agricultural production confidence of interviewed farmers showed good recovery, with an average value of 3.214.

2.3.2. Risk Expectation

The key explanatory variable of this article is risk expectation. This study defines risk expectation as farmers’ subjective assessment and cognition of risks in agricultural production. Drawing on the work of Safi et al. [78] and Mueller [79], risk expectation was measured by the following three questions: (1) “How much do you think the COVID-19 pandemic will negatively impact your family’s future agricultural production scale?”, (2) “How much do you think the COVID-19 pandemic will negatively impact your family’s future agricultural production costs?”, and (3) “How much do you think the COVID-19 pandemic will negatively impact your future market judgment?”; the responses to each question ranged from 1 (suffer little influence) to 5 (suffer large impact). We then combined all three risk targets in one compound risk expectation measure with principal component analysis. STATA software was used for the analysis of the variable. However, the risk expectation Likert scale had reliable internal consistency (Cronbach’s α = 0.585).

2.3.3. Social Support

Based on the research hypothesis illustrated in the previous section, we divided social support into three dimensions: government support, support from relatives and friends, and financial support. Government support was measured by asking “How much help could you receive from the government under the COVID-19 pandemic shock?” Support from relatives and friends was measured by asking “How much help could you receive from relatives and friends under the COVID-19 pandemic shock?” Financial support was measured by asking: “How much help could you receive from financial institutions under the COVID-19 pandemic shock?” The responses to each question ranged from 1 (only a little) to 5 (a great deal) to evaluate each dimension of social support. Interestingly, measuring each dimension of social support could be measured as a composite social support indicator. Therefore, we combined all three dimensions into one compound social support measure using principal component analysis. The social support Likert scale also had reliable internal consistency (Cronbach’s α = 0.605).

2.3.4. Control Variables

To reduce the possibility of missing variables as much as possible, we referred to relevant research on the decision-making behavior of farmers [80,81]. It is generally believed that planting characteristics, family characteristics, and demographic characteristics (such as age and gender) are important factors affecting residents’ post-disaster life and psychology [82,83]. Thus, we selected planting characteristics, household head characteristics, and farmer family characteristics as control variables. The household head characteristics included gender, age, and household head education level. Planting characteristics included farm size, whether a household grows food crops, and whether a household grows cash crops. Meanwhile, household characteristics included agricultural labor force, household income level, the proportion of agricultural income, and household size. The model in the study also included provinces as dummy variables to reduce the impact of regional differences, as suggested by Wang et al. [84].

2.3.5. Instrumental Variables

Additionally, instrumental variables were introduced because risk expectations may pose problems for farmers’ confidence in agricultural production recovery. We focused on the impact of farmers’ risk expectations on confidence in agricultural production recovery within the context of the COVID-19 pandemic. The “proportion of medium–high-risk areas” in a city where farmers live was selected as the instrumental variable of farmers’ risk expectations. If the proportion of medium–high-risk areas in the city where the farmer lives is relatively higher, the risk that the farmer expects to face will be more significant [85]. At the same time, the proportion of “medium- and high-risk areas” in the city where the farmers live is not directly related to the farmers’ confidence in agricultural production recovery, so it is strictly exogenous. The data for the instrumental variables were derived from the pandemic risk level data released by the National Health Commission of the People’s Republic of China in February 2020. The definition and statistical analysis results of each variable are shown in

Table 2. Definition and summary statistics of the selected variables.

Variables	Definition (Unit)	Mean	Std. Dev
Farmers’ confidence in agricultural production recovery	Are you confident that agricultural incomes will recover from the COVID-19 pandemic shock? (1 = only a little, 5 = a great deal)	3.214	1.022
Risk expectation	Principal component analysis	0.506	0.198
	How much do you think the COVID-19 pandemic will negatively impact your family’s future agricultural production scale? (1 = suffer little influence, 5 = suffer large impact)	3.048	1.102
	How much do you think the COVID-19 pandemic will negatively impact your family’s future agricultural production costs? (1 = suffer little influence, 5 = suffer large impact)	3.242	1.031
	How much do you think the COVID-19 pandemic will negatively impact your future market judgment? (1 = suffer little influence, 5 = suffer large impact)	2.799	1.066
Social support	Principal component analysis	0.466	0.199
Government support	How much help could you receive from the government during the COVID-19 pandemic shock? (1 = only a little, 5 = a great deal)	2.963	1.04
Support from relatives and friends	How much help could you receive from relatives and friends during the COVID-19 pandemic shock? (1 = only a little, 5 = a great deal)	2.419	1.06
Financial support	How much help could you receive from financial institutions during the COVID-19 pandemic shock? (1 = only a little, 5 = a great deal)	2.784	1.001
Gender	Male = 1; Female = 0	0.834	0.372
Age	The household head’s age in years	53.048	8.604
Education	Education level of the household head: illiterate = 1; primary = 2; junior secondary = 3; high school/technical secondary school = 4; junior college and above = 5	2.867	1.057
Farm size	Farm size: under 3 mu = 1; 3–5 mu = 2; 6–8 mu = 3; 8–10 mu = 4; Above 10 mu = 5	3.489	1.135
Agricultural labor force	Household agricultural labor force (number)	2.234	0.829
Household income level	Total household income in 2019: under RMB 50,000 = 1; RMB 50,000–100,000 = 2; RMB 100,000–200,000 = 3; above RMB 2,000,000 = 4	2.255	0.872
Proportion of agricultural income	The proportion of agricultural income in total household income: 0–20 = 1; 21–40 = 2; 41–60 = 3; 61–80 = 4; 81–100 = 5	2.338	1.100
Household size	The population of household families (number)	6.544	1.802
Household grows food crops	Yes = 1; No = 0	0.677	0.468
Household grows cash crops	Yes = 1; No = 0	0.236	0.425
Province dummy variables	Hubei Province = 1; Elsewhere = 0	0.400	0.490
Instrumental variable	The proportion of medium-risk areas and high-risk areas in the city where the farmers live (%): 0–20 = 1; 21–40 = 2; 41–60 = 3; 61–80 = 4; 81–100 = 5	3.948	1.000

Notes: Data source was accessed from February to March of 2020.

.

2.4. Common Method Bias Test

This study collected data using the scale questionnaire method. Potential systematic measurement errors will inevitably occur due to specific measurement methods (such as measurement environment, project context, or characteristics unique to the project), and are generally referred to as common method bias. This systematic error may lead to incorrect causal inference and seriously affect the accuracy of the results. The following measures were adopted in this study to avoid the influence of such systematic errors: Pre-control and post-test resolve potential common method bias. Initially, objective and open-ended questions were both inserted into the questionnaire. Concurrently, the Harman single-factor test was used to test the data [86]. According to the previous practice, exploratory factor analysis was executed on all latent variables under the extraction method of no rotation, and featured roots greater than 1. The results indicate that the first principal component explained 30.05% of the total variance—less than half of the total extraction variance accumulation; this signifies that there is no serious common method bias with the instituted common method of data.

2.5. Econometric Model

2.5.1. Ordered Probit Model

Since the dependent variable is an ordered variable of five categories, we adopted an ordered probit model. The empirical model was set as follows:

$$Y_i^{}=F({{\displaystyle \alpha }}_i+{{\displaystyle \beta }}_{i1}RE+{{\displaystyle \beta }}_{i2}SS+{{\displaystyle \beta }}_{i3}RE\times SS+{{\displaystyle \gamma }}_{ij}{{\displaystyle X}}_{ij}+{{\displaystyle \epsilon }}_i)$$

(1)

where $i$ representing the i^th farmer; $Y_i$ is the dependent variable representing the probabilities of farmers’ confidence recovery in agricultural production; $RE$ is the key independent variable representing risk expectation; $SS$ is the social support; $RE\times SS$ is the interaction term for risk expectation and social support; ${{\displaystyle X}}_{ij}$ represents the control variables, which include household head characteristics, planting characteristics, and family characteristics; $\beta$ and $\gamma$ are the estimated parameters; ${{\displaystyle \epsilon }}_i$ is the error term that obeys a standard normal distribution; $F(·)$ is a function, and its specific form is as follows:

$$F(Y_i^{*})=\left\{\begin{array}{lll}1& & Y_i^{*}<u_1\\ 2& & u_1<Y_i^{*}<u_2\\ ⋮\\ J& & Y_i^{*}>u_{J-1}\end{array}\right.$$

(2)

where $u_1<u_2<\cdots <u_{J-1}$ are the estimated parameters; and $Y_i^{*}$ is the latent variables, which should satisfy the following conditions:

$${{{\displaystyle Y}}_i}^{*}={{\displaystyle \alpha }}_i+{{\displaystyle \beta }}_{i1}RE+{{\displaystyle \beta }}_{i2}SS+{{\displaystyle \beta }}_{i3}RE\times SS+{{\displaystyle \gamma }}_{ij}{{\displaystyle X}}_{ij}+{{\displaystyle \epsilon }}_i$$

(3)

2.5.2. Conditional Mixed Process Estimation of Instrumental Variables

For the following reasons, endogeneity problems may exist between risk expectation and farmers’ confidence in agricultural production recovery: On the one hand, it is impossible to obtain all household head characteristic variables and family characteristic variables of respondents. Some unobserved factors (such as social environment) may influence farmers’ confidence in agricultural production recovery. On the other hand, risk expectation and farmers’ confidence in agricultural production recovery are subjective variables. Moreover, similar psychological mechanisms may lead to the possibility of reverse causality between risk expectation and farmers’ confidence in agricultural production recovery. Therefore, in order to reduce the possibility of endogeneity, instrumental variables should be introduced. As the explained variable is a multivariate discrete variable, the traditional instrumental variable method may estimate errors. We employed a conditional mixed process (CMP) estimation method, which is proposed to reduce the possibility of endogeneity problems [87]. Constructing recursive equations estimates a two-stage regression model based on seemingly unrelated regressions and maximum likelihood estimation. The empirical models are as follows:

$${{\displaystyle RE}}_i={{\displaystyle \alpha }}_i+\xi {{\displaystyle Z}}_i+{\displaystyle \sum _{i=1}^n{{\displaystyle \lambda }}_i{{\displaystyle X}}_i}+{{\displaystyle \epsilon }}_i$$

(4)

$${{\displaystyle Y}}_i={{\displaystyle \alpha }}_i+\gamma {{\displaystyle RE}}_i+{\displaystyle \sum _{i=1}^n{{\displaystyle \lambda }}_i}{{\displaystyle X}}_i+{{\displaystyle \epsilon }}_i$$

(5)

where ${{\displaystyle \alpha }}_i$ is a constant term; $Z$ is the instrumental variable; $\xi$, $\gamma$, and ${{\displaystyle \lambda }}_i$ are the estimated parameters; and ${{\displaystyle \epsilon }}_i$ is the error term that obeys a standard normal distribution.

2.6. Basic Steps of the Analysis

To explore the relationships between risk expectation, social support, and farmers’ confidence recovery in agricultural production, as well as their interaction mechanism, the empirical analysis steps of this paper were as follows: First, the benchmark model was used to explore the impact of risk expectation on farmers’ confidence in agricultural production recovery. Second, social support and its different dimension variables were introduced to analyze the impact of social support on farmers’ confidence in agricultural production recovery, and the moderating effect of social support. Finally, the intergenerational differences between risk expectation, social support, and farmers’ confidence in agricultural production recovery were analyzed.

3. Results

3.1. Basic Regression

Table 3. Impact of risk expectation on farmers’ confidence in agricultural production recovery.

Variables	Model 1	Model 2	Model 3
Risk expectation		−1.706 *** (0.330)	−1.733 *** (0.326)
Control variables
Gender	0.152 (0.128)	0.188 (0.127)	0.197 (0.127)
Age	0.017 *** (0.006)	0.012 ** (0.006)	0.015 ** (0.006)
Education	0.160 *** (0.059)	0.154 *** (0.059)	0.152 ** (0.060)
Farm size	−0.103 * (0.054)	−0.044 (0.056)	−0.048 (0.055)
Agricultural labor force	0.067 (0.084)	0.084 (0.082)	0.077 (0.082)
Household income level	0.176 *** (0.067)	0.141 ** (0.069)	0.151 ** (0.070)
Proportion of agricultural income	−0.048 (0.060)	−0.098 * (0.055)	−0.103 * (0.056)
Household size	0.039 (0.040)	0.010 (0.041)	0.013 (0.041)
Household grows food crops	−0.145 (0.121)	−0.164 (0.118)	−0.175 (0.119)
Household grows cash crops	−0.373 ** (0.146)	−0.343 ** (0.145)	−0.363 ** (0.146)
Province dummy variables	-	-	Yes
Pseudo R2	0.054	0.081	0.084
Wald chi2	69.09 ***	130.37 ***	134.33 ***
Log pseudo likelihood	−604.965	−587.729	−586.210

Note: ***, **, and * represent significance levels of 1, 5, and 10%, respectively. Robust standard errors are in parentheses.

estimates the results of the impact of risk expectation on farmers’ confidence in agricultural production recovery; model 1 only adds control variables, model 2 adds risk expectation variables based on model 1, and model 3 adds province dummy variables based on model 2.

For the impact of risk expectation on farmers’ confidence in agricultural production recovery, the risk expectations of farmers negatively affect farmers’ confidence in agricultural production recovery at the 1% significance level, which indicates that the greater the uncertainty and risk that COVID-19 is expected to bring to agricultural production, the slower the recovery of farmers’ confidence in agricultural production. This result confirms Hypothesis 1, and indicates that risk expectation does have a negative impact on farmers’ confidence in agricultural production recovery; one possible explanation for this is the direct impact of the pandemic, coupled with a series of measures taken to prevent its spread; moreover, this will significantly impact the entire chain and various agricultural production and operation fields. The greater the risk to agricultural production that farmers expect from these impacts, the lower their confidence in recovery.

For the impact of the household head characteristics on farmers’ confidence in agricultural production recovery, the age and education level of the household head positively affect farmers’ confidence in agricultural production recovery at 5% levels of significance. This suggests that the older a household head and the greater their farming experience, the more substantial their ability to manage risk and resist setbacks, and their agricultural recovery is also faster. The higher a household head’s education level, the broader their horizons, the more progressive their thinking, and the greater their capability to resist risks and maintain production order. In this model, the coefficient of gender is positive but insignificant, which is inconsistent with the findings of Bonanno et al. [83]; a possible explanation for this is that the proportion of male household heads in the survey data is too large, resulting in sample homogeneity.

For the impact of the family characteristics on farmers’ confidence in agricultural production recovery, household income level positively affects farmers’ confidence in agricultural production recovery at the 5% significance level; this indicates that the higher the family’s income level, the more capable they are of withstanding losses, and the faster they are able to regain confidence in agricultural production. The proportion of agricultural income negatively affects farmers’ confidence in agricultural production recovery at the 10% significance level; this indicates that farmers with a higher proportion of agricultural income will suffer more significant economic losses after a pandemic; simultaneously, the cost of reconstruction will be higher, and the recovery of agricultural confidence will be slower. In this model, the coefficients for the agricultural labor force and household size are positive but insignificant; one possible explanation for this is that with the increase in population and the application of modern agricultural technology in recent years, the size of the agricultural labor force has a more limited effect on agricultural production. In addition, we found that the larger the family, the greater the pressure to support and care for its members. However, traditional agricultural production cannot support large families, so most families have prominent non-agricultural employment situations; this means that agricultural losses caused by a pandemic can be accommodated by accumulating funds from off-farm employment.

Farm size is not significant for the impact of the planting characteristics on farmers’ confidence in agricultural production recovery; this indicates that most of the respondents were farmers with small farms who are not sensitive to external shocks. Whether a household grows cash crops was negative at the 5% significance level, while the coefficient of whether a household grows food crops was insignificant; a possible explanation for this is that cash crops are more affected by a pandemic than food crops.

3.2. The Impact of Social Support on Farmers’ Confidence in Agricultural Production Recovery

This study adds social support and interaction terms between risk expectations and social support based on model 3 to measure the impact of social support on farmers’ confidence in agricultural production recovery, as well as its moderating effect. The interaction effect can represent the moderating effect [82,88], and the results of the model are shown in

Table 4. The impact of social support on farmers’ confidence in agricultural production recovery.

Variables	Model 4	Model 5	Model 6	Model 7
Risk expectation	−0.923 ***	−1.387 ***	−1.140 ***	−1.467 ***
	(0.343)	(0.341)	(0.333)	(0.338)
Social support	2.392 ***
	(0.319)
Government support		0.192 ***
		(0.058)
Relatives and friends support			0.417 ***
			(0.055)
Financial support				0.255 ***
				(0.059)
Risk expectation × social support	4.711 ***
	(1.435)
Risk expectation × government support		0.730 ***
		(0.283)
Risk expectation× relatives and friends support			0.790 ***
			(0.270)
Risk expectation × financial support				0.598 **
				(0.299)
Other controls	Yes	Yes	Yes	Yes
Province dummy variables	Yes	Yes	Yes	Yes
Pseudo R2	0.145	0.102	0.134	0.107
Wald chi2	205.95 ***	165.84 ***	213.99 ***	164.09 ***
Log pseudo likelihood	−546.711	−574.355	−554.174	−571.182

***, and ** represent significance levels of 1, 5, and 10%, respectively. Robust standard errors are in parentheses.

.

Model 4 shows that social support positively affects farmers’ confidence in agricultural production recovery at the 1% significance level; this indicates that the more social support farmers receive, the better their ability to restore their confidence in agricultural production; therefore, Hypothesis 2 has been verified. The impact of the COVID-19 pandemic on farmers’ psychology and income can be mitigated by appropriate social support [89]. The interaction term between risk expectation and social support is positive at the 1% significance level; this indicates that the negative impact of risk expectations on the restoration of farmers’ agricultural production confidence will continue to weaken as their level of received social support increases. In short, social support plays a significant positive moderating role in the impact of risk expectations on farmers’ confidence in agricultural production recovery; therefore, Hypothesis 6 has been verified. Farmers expect that the risks posed by the COVID-19 pandemic in agricultural production will make it difficult for their agricultural income to recover to its average level, which will therefore affect their agricultural production confidence. Supposing that emotional and material support can be assessed from social networking, the ability of farmers to control and cope with risks could be more manageable, thereby alleviating the adverse effects of risk expectations on the restoration of agricultural production confidence.

Our analysis additionally explored the effects of different dimensions of social support. First, we introduced government support and its interaction with risk expectations. As shown in model 5, government support positively affects farmers’ confidence in agricultural production recovery, and plays a positive moderating role in the impact of risk expectations on farmers’ confidence in agricultural production recovery; thus, Hypothesis 3 and Hypothesis 7 have been verified. Second, we introduced support from relatives and friends, and its interaction with risk expectations. As shown in model 6, support from relatives and friends positively affects farmers’ confidence in agricultural production recovery, and plays a positive moderating role in the impact of risk expectations on farmers’ confidence in agricultural production recovery; therefore, Hypothesis 4 and Hypothesis 8 have been verified. Third, we introduced financial support and its interaction with risk expectations. As shown in model 7, financial support positively affects farmers’ confidence in agricultural production recovery, and plays a positive moderating role in the impact of risk expectations on farmers’ confidence in agricultural production recovery; thus, Hypothesis 5 and Hypothesis 9 have been verified.

3.3. Analysis of Confidence in Agricultural Production Recovery among Different Generations of Farmers

Since the explained variable is an ordered discrete variable, an ordered probit model was used. However, many studies have pointed out that using an OLS model, ordered probit model, and ordered logit model will not significantly impact the sign and significance of the variable coefficients compared with just the latter two models. The regression results of an OLS model can present the marginal effect more intuitively [90]. For robustness, OLS and ordered probit models were compared to analyze the intergenerational difference in the study.

Table 5. Analysis of intergenerational differences.

Variables	The New Generation		The First Generation
	Oprobit	OLS	Oprobit	OLS
Risk expectation	−2.349 *** (0.893)	−1.443 ** (0.559)	−0.795 ** (0.390)	−0.678 ** (0.301)
Social support	2.061 ** (0.958)	1.201 * (0.711)	2.379 *** (0.346)	1.875 *** (0.251)
Risk expectation× social support	8.474 ** (3.477)	4.380 ** (2.023)	4.396 *** (1.604)	3.139 *** (1.157)
Other controls	Yes	Yes	Yes	Yes
Province dummy variables	Yes	Yes	Yes	Yes
Number of obs	383	383	383	383
Pseudo R2	0.251	-	0.140	-
R2	-	0.490	-	0.340
Wald chi2	79.58 ***	-	161.24 ***	-
Log pseudo-likelihood	−77.113	-	−456.817	-

Note: ***, **, and * represent 1, 5, and 10% significance levels, respectively. Robust standard errors are in parentheses.

displays the results of confidence in agricultural production recovery among different generations of farmers.

First, risk expectation has a negative impact on the confidence in agricultural production recovery of the older and younger generations at 1% and 5% significance levels, respectively. The impact of risk expectations on the younger generation of farmers is more significant than on the older generation, demonstrating that the older generation of farmers has a longer planting time and a richer planting experience; therefore, risk expectations have a lesser impact on their confidence in agricultural production recovery. Second, social support significantly impacts the recovery confidence of farmers of both generations; however, it can be observed from the coefficients in the model that the impact of social support on the older generation is more significant than on the younger generation. One possible explanation for this is that, compared with the older generation of farmers, the younger generation have fewer pre-existing sources of support. However, the younger generation of farmers may have the more substantial social capital capacity for support. The younger generation of farmers has the characteristics of high social expectations and low life satisfaction; therefore, the social support of the younger generation of farmers has a more negligible effect on confidence recovery. Finally, social support has a positive moderating effect on the relationship between the risk expectations and the confidence in agricultural production recovery of both the younger and older generations of farmers. The impact on the younger generation is more significant than on the older generation; one possible explanation for this is that the younger generation of farmers has a more vital ability to integrate resources, and can more reasonably use any social support they receive to alleviate the adverse impact of risk expectation on farmers, enhancing the positive effects of social support on the relationship between risk expectations and farmers’ confidence in agricultural production recovery. Therefore, Hypothesis 10 has been verified.

3.4. Endogeneity Test

We used conditional mixed process (CMP) estimation to tackle the endogenous problems.

Table 6. Conditional mixed process estimation of the impact of risk expectation on farmers’ confidence recovery in agricultural production.

Variables	First Stage	Second Stage
Risk expectation	-	−4.736 *** (0.966)
Proportion of medium–high-risk areas	0.036 *** (0.010)	-
Other controls	Yes	Yes
lnsig_2	-	−1.758 *** (0.032)
atanhrho_12	-	0.672 ** (0.324)
Wald chi2	231.07 ***
Log pseudo-likelihood	−429.546

Note: *** and ** represent significance levels of 1, 5, and 10%, respectively. Robust standard errors are in parentheses.

shows the results of the CMP estimation of instrumental variables. The first-stage model takes risk expectation as an explained variable, while the second-stage model takes farmers’ confidence in agricultural production recovery as an explained variable. From the estimation results of the first-stage model, it can be seen that the instrumental variable has a positive impact on risk expectations at a 5% significance level, so the correlation is satisfied. The CMP estimates refer to the endogeneity test parameter atanhrho_12 to distinguish the homogeneity of the variable; if it is significantly different from 0, the parameter indicates that the benchmark model has an endogeneity problem. The atanhrho_12 parameter is significant at the 1% level, which rejects the original hypothesis that the “risk expectation” variable is an exogenous variable, indicates that the instrumental variables selected in this paper are exogenous, and signifies that there are no problems of weak instrumental variables. This shows that the ordered probit estimation results are somewhat biased due to the reverse causality between risk expectations and farmers’ confidence in agricultural production recovery. However, it can be observed from the results of the second-stage model that even if the estimated values of key variables are errors, they still have a certain level of economic significance. Seemingly, the risk expectations will indeed negatively affect farmers’ confidence in agricultural production recovery.

3.5. Robustness Check

Table 7. Robustness check.

Variables	Probit	IV-Probit	Probit	IV-Probit	Probit	IV-Probit	Probit	IV-Probit
Risk expectation	−2.263 *** (0.429)	−5.578 *** (0.746)	−2.448 *** (0.397)	−5.677 *** (0.703)	−2.185 *** (0.402)	−5.580 ** (0.666)	−2.594 ** (0.412)	−5.902 *** (0.574)
Social support	3.174 *** (0.464)	2.292 *** (0.659)
Government support			0.224 *** (0.068)	0.162 ** (0.064)
Support from relatives and friends					0.510 *** (0.075)	0.360 *** (0.107)
Financial support							0.272 *** (0.073)	0.195 ** (0.070)
Risk expectation × social support	9.196 *** (2.654)	6.235 ** (2.590)
Risk expectation × government support			0.777 ** (0.360)	0.493 ** (0.293)
Risk expectation × relatives and friends support					0.762 * (0.407)	0.482 (0.314)
Risk expectation × financial support							1.103 *** (0.421)	0.769 * (0.364)
Other controls	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Province dummy variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Wald chi2	128.28 ***	370.21 ***	125.85 ***	387.52 ***	146.23 ***	466.65 ***	124.62 ***	405.44 ***
Log pseudo-likelihood	−222.131	−64.368	−244.758	−86.870	−225.483	−67.436	−241.923	−83.422

Note: ***, **, and * represent significance levels of 1, 5, and 10%, respectively. Robust standard errors are in parentheses.

displays the results of the robustness test. We used other explained variable measurement methods to conduct empirical tests and test the estimation results’ robustness. We divided the answers to the explained variables into two categories: Farmers with confidence classified as “only a little”, “relatively small”, or “average” comprised the “unconfident” sample group. Farmers with confidence classified as “relatively large” or “very large” comprised the “confident” sample group. As a result, the explained variable changed from “ordered” to “two classifications” variables. The results of the probit model and IV-probit model used for robustness estimation showed that risk expectations have a negative impact on farmers’ confidence in agricultural production recovery. Social support and its different dimensions play a positive moderating role in the relationship between risk expectations and farmers’ confidence in agricultural production recovery. These results are consistent with previous research.

4. Discussion

Several empirical studies have shown that the COVID-19 pandemic has brought risks to farmers, and that these risks have had a specific impact on their mental health and the recovery of production and life standards [91,92]. However, the farmers’ confidence in agricultural production recovery is still unclear. Based on the background of the COVID-19 pandemic and the survey data of 458 households in 2020, this study investigated farmers’ confidence in agricultural production recovery, and analyzed the impact of risk expectation and social support, along with the moderating effect of social support, on this basis, while generational differences were also taken into account. Endogeneity analysis and robustness tests were carried out on the results, making up for the shortcomings of previous research.

First, risk expectation has a significant negative impact on the farmers’ confidence in agricultural production recovery, and this finding is consistent with previous research on risk expectations and disaster recovery [93]. Al-Nammari and Alzaghal [94] stressed the importance of public perception and attitudes to disaster risks in developing and implementing disaster management plans, building resilience, and adaptation measures. Seemingly, Murakami et al. [95] found that the fear caused by the risk perception of residents in the hardest hit areas had a negative and significant impact on the public’s mental health. Farmers’ risk expectations indeed have an adverse effect on their confidence in recovery. After the outbreak of the COVID-19 pandemic, farmers expected that the pandemic would bring certain risks and losses to the scale of agricultural production, agricultural production costs, and the agricultural product market. The more significant the farmers expect these risks and losses to be, the less conducive to their confidence in recovery.

Second, social support has a significant positive impact on farmers’ confidence in agricultural production recovery, which is consistent with the findings of Hsueh [96]. Shimda [97] also illustrated that social support plays an essential role in the process of recovery. However, in contrast to these studies, we also analyzed the role of government support, support from relatives and friends, and financial support as dimensions of social support. We found that government support, support from relatives and friends, and financial support significantly positively impacted farmers’ confidence in agricultural production recovery. This finding indicates that the government’s relevant policy support and service support to guide farmers to resume agricultural production comprehensively made up for farmers’ losses to a certain extent, and can effectively promote confidence recovery [98]. Emotional support and encouragement from relatives and friends, as well as information resources, also helped farmers to increase their confidence in overcoming difficulties [99]. The financing support obtained from financial institutions can alleviate the financial difficulties faced by farmers during emergencies [100].

Third, social support has a significant moderating effect on the relationship between risk expectation and farmers’ confidence in agricultural production recovery. Government support, support from relatives and friends, and financial support all play a moderating role in the relationship between risk expectation and confidence in recovery; this is similar to the findings of previous studies; for example, Zuo et al. [101] showed that social support could work as a protective factor in alleviating homelessness. Kaniasty et al. [31] stressed that salutary direct effects of supportive behaviors on post-disaster psychological distress were also highly evident. This indicates that support from governments, relatives, friends, and financial institutions can help farmers emotionally, and economically alleviate the pandemic’s risks and its negative impact on farmers’ confidence in recovery.

Finally, through the analysis of intergenerational differences, we found significant differences in the impacts of risk expectation and social support on farmers’ confidence in recovery between the older and younger generations [75]; a possible explanation for this is that the educational experience, social background, and values of the younger and older generations of farmers are different, leading to different risk expectations. At the same time, the younger and older generations of farmers have different social relations and information sources, allowing them to obtain additional social support. Therefore, there were significant intergenerational differences in the impacts of risk expectation and social support on farmers’ confidence in recovery, which is consistent with the findings of Sarkar et al. [102].

Based on the above results, our analysis suggests the following policy recommendations: First, strengthen the agricultural production risk mechanism, dynamic monitoring, and real-time early warning mechanisms; establish awareness of existing risks in agricultural production, identify problems, and resolve conflicts and unstable factors; simultaneously, extend the support of public services to farmers, strengthen risk-taking networks, and plan holistically to reduce risk expectations. Second, the government should play a leading role in the social support system, and improve its policy support and service support systems. It is necessary to fully mobilize the family’s functions so that family members, relatives, and friends can help one another and give full play to the essential protection function of relatives and friends in the recovery process. At the same time, it is necessary to strengthen the loan and financing support of financial institutions—such as the Agricultural Bank and insurance companies—to provide farmers with comprehensive and multilevel services and assistance. Third, the government should implement diversified agricultural policies based on differences between new and first generations of farmers concerning the intergenerational differences. For the younger generations of farmers, it is vital to strengthen their tolerance to disaster risk, while the older generations of farmers need to strengthen their external support resource integration capabilities.

5. Conclusions

Overall, we empirically analyzed the relationship between risk expectations, social support, and farmers’ confidence in agricultural production recovery. Our results show that the following: First, the risk expectation of farmers has a significant negative impact on their confidence in agricultural production recovery; the higher the risk that farmers expect to agricultural production from COVID-19, the harder it will be to restore their confidence in agricultural production. Second, social support has a significant positive impact on farmers’ confidence in agricultural production recovery. Social support also has a significant moderating effect on the relationship between risk expectation and farmers’ confidence in agricultural production recovery. In short, social support can alleviate the adverse impact of risk expectation on agricultural production confidence restoration. Different dimensions of social support—including government support, support from relatives and friends, and financial support—have a positive regulating effect on the relationship between risk expectation and farmers’ confidence in agricultural production recovery. Third, the analysis of intergenerational differences illustrates that risk expectations and social support have significant differences between the older and younger generations in terms of the impact on farmers’ confidence in agricultural production recovery. In addition, household head age, household head education level, household income level, and the proportion of agricultural income also significantly impact farmers’ confidence in agricultural production recovery.

The present study has several limitations. First, although “whether to plant food crops” and “whether to plant cash crops” were included in the model as the control variables affecting the farmers’ confidence in agricultural production recovery, the effects of crop types were not explicitly discussed; the COVID-19 pandemic occurred in the spring of 2020, farmers grow different types of crops, and their confidence in recovery is also different; as such, future research should take complete account of specific crop types. Second, the farmers’ confidence in recovery should also be specifically related to the COVID-19 pandemic, which should be included in future studies.