Relationship between Disaster Shock Experience and Farmers’ Entrepreneurial Inclination: Crisis or Opportunity?

Abstract

In the context of climate change, it is important to examine the correlation between farmers’ disaster shock experience and their entrepreneurial inclination, as well as its adaptive mechanisms for rural development. We define farmers’ entrepreneurship as farmers engaging in self-employment or business operation, then analyze the positive and negative correlations between disaster shock experience and farmers’ entrepreneurship and propose that there is a U-shaped relationship between them. We develop a mechanism based on the mediating role of farmers’ risk appetite and the moderating role of government support. Based on a survey of 39,113 households from the 2019 China Household Finance Survey, we test our hypotheses using a probit model and structural equation modeling. The findings indicate that disaster shock experience strongly correlates with farmers’ entrepreneurship. Over time, a U-shaped trend emerges between them, in which disaster shock experience can deter farmers’ entrepreneurship through risk appetite, while government support can alleviate this inhibitory effect. Moreover, different types of disasters show the same U-shaped trend, but the relationship differs based on farmers’ gender and income levels. These findings suggest that rural areas should enhance their development capacity and disaster resilience, and the government should establish long-term support mechanisms for entrepreneurship. Additionally, farmers need to accurately understand and cope with disaster shock experience, so they can uncover the potential value and opportunities it presents.

1. Introduction

Farmers play a crucial role in revitalizing rural areas and their entrepreneurial and innovative behaviors are key drivers of rural prosperity. Currently, China’s entrepreneurial environment is generally positive, with the exception of some western regions that lag behind in terms of economic development. According to the Peking University Entrepreneurship Index evaluation and the Global Entrepreneurship Monitor China Report (2019–2020) [1,2], China’s provincial and regional entrepreneurship environment indexes are highly rated, ranking second worldwide. The development of the “digital countryside”, smart agriculture, and rural tourism presents opportunities for farmers to engage in entrepreneurship, leading to a significant number of farmers pursuing innovative practices. In 2022, more than 11 million people returned to their hometowns to undertake entrepreneurial ventures [3], demonstrating that rural entrepreneurship is thriving.

We define farmers’ entrepreneurship as farmers engaging in self-employment or business operation. Research on farmers’ entrepreneurship has mainly focused on the effects of factors such as entrepreneurs’ knowledge, social capital, and personality traits [4,5,6], as well as external factors such as infrastructure, financing environment, and government support [7,8,9,10,11]. There has been limited research, however, on the influence of changes in the natural environment, especially climate change and disaster shocks, on farmers’ entrepreneurship. The relationship between agriculture and changes in the natural environment is profound, and increasing disaster risks related to climate extremes significantly disrupt rural life and production, potentially limiting long-term rural social development. Although some studies have noted the influence of related experiences such as early childhood starvation on farmers’ entrepreneurship [12], few have explored the unique correlation between disaster shock experience and farmers’ entrepreneurship, which is dependent on subjects’ risk appetite and social behavior characteristics [13].

The occurrence of a disaster as a sudden shock can disrupt the developmental path of a group that faces hardship, and post-disaster recovery is contingent on the group’s ability to withstand such shocks [14]. Following a disaster shock, it is difficult for the affected group to participate in entrepreneurship given its high-risk nature, combined with the inadequate infrastructure and typically underdeveloped socioeconomic situations of rural areas. Nevertheless, the correlation between disaster shock experience and farmers’ entrepreneurship remains largely unexplored, and there is no consensus on the related mechanisms. Some researchers agree with the premise that disaster shock experience will reduce entrepreneurial intention and impede entrepreneurial decision making. For instance, disaster shock experience can make individuals more risk-averse, thus negatively affecting entrepreneurship (which is inherently high risk). Research on the effects of natural disasters and early famine experience supports this perspective [15,16,17]. By contrast, others suggest that although disaster shocks have immediate adverse effects on individuals, an indirect influence such as the conversion of psychological well-being and survival circumstances might increase the likelihood of individuals selecting self-employment. Natural disasters, for example, have been found to increase farmers’ propensity to start their own businesses, and nonfarm employment rises with the intensity of natural disasters [18,19]. Furthermore, severe disasters can entirely disrupt farmers’ livelihoods, forcing them to pursue alternatives, thus increasing their chances of pursuing entrepreneurial activities [20]. Yet, there remains a lack of empirical research on the two abovementioned perspectives on the correlation between disaster shock experience and farmers’ entrepreneurship. Such work could improve our understanding of the mechanisms of farmers’ entrepreneurial behavior considering the increasing probability of climate change–related extreme weather events and help to suggest targeted countermeasures. The ultimate goal is to identify ways that the government can accelerate rural innovation and entrepreneurship, specify the direction of rural development, promote the upgrading and adjustment of the industrial structure, and realize the connection between poverty alleviation and rural revitalization.

This research uses data from 39,113 farming households included in the 2019 China Household Finance Survey (CHFS) to study the correlation between disaster shock experience and farmers’ entrepreneurship. We identify a U-shaped relationship between disaster shock experience and farmers’ entrepreneurship. The experience of disaster shocks will initially inhibit farmers’ entrepreneurship but will then later promote it as time passes after exceeding the critical point, with the mediating role of risk appetite and the moderating role of government support promoting the leftward shift of the critical point. In addition, we found that different disasters have the same U-shaped relationship, women are more dampened by disaster shocks than men, and low-income households are more affected by disaster shocks than high-income households. This study’s marginal contributions are threefold. First, based on prior studies, we propose a U-shaped relationship between disaster shock experience and farmers’ entrepreneurship. Second, we enrich the process variables of the correlation between disaster shock experience and farmers’ entrepreneurship—namely, farmers’ risk appetite and government support. Third, we analyze the heterogeneous roles of different disasters, individual traits, and family factors between disaster shock experience and farmers’ entrepreneurship.

This paper has six main sections. Following the introduction, we discuss the theoretical relationship between disaster shock experience and farmers’ entrepreneurship and present our hypotheses. Section 3 describes the data and the econometric model used for analysis. Next, Section 4 presents the baseline regression results and addresses endogeneity issues. Section 5 explores the correlation between disaster shock experience and farmers’ entrepreneurship and tests for heterogeneity. Finally, Section 6 concludes with a discussion of the findings and the policy implications.

2. Theoretical Foundations and Hypotheses

2.1. Theoretical Foundations of Disaster Shock Experience and Farmers’ Entrepreneurship

French economist Richard Cantillon defined an “entrepreneur” as a risk-taker in business activities [19,21]. In the early stages of entrepreneurship research, individual traits were the primary focus. Entrepreneurial trait theory emphasized the differences between entrepreneurs and non-entrepreneurs, arguing that individuals with specific entrepreneurial psychological traits possess unique abilities that are relatively stable and not affected by context [20,22]. Later, research shifted toward entrepreneurial resources, including human and social capital [23]. More recently, contextual factors have gained attention, and there is increasing interest in exploring the institutional environments that guide entrepreneurship, as well as the combination of entrepreneurial opportunities and resources.

Farmers’ entrepreneurship has evolved into a form of entrepreneurship that is unique to the Chinese context, given the country’s distinct urban–rural structure [24]. Innovative entrepreneurial practices continue to emerge, such as Taobao village, which is driven by the development of e-commerce specific to the Chinese context. The development of entrepreneurial theories, such as the bottom-of-the pyramid theory and inclusive growth theory [25,26], fuels novel practices. With the addition of imprinting theory, researchers have focused on the distinct perspective of “experience”. Some have found that the business experience, CEO experience, and non-farm employment experience of village officials can be organically blended with entrepreneurship [27,28,29]. Technical accumulation, management experience, and market information from past labor experiences can significantly moderate entrepreneurial performance [30]. Furthermore, the experience of entrepreneurial failure has been shown to foster antecedents for future entrepreneurship and continuous growth in job performance [31,32]. With the widespread application of imprinting theory, studies of experience have become increasingly comprehensive. In addition to studies related to entrepreneurial labor, studies targeting childhood poverty experiences, early famine experiences, schooling experiences, second-generation growth experiences, and overseas experiences have gradually emerged in the research landscape [16,30,33]. The role of such experiences is often not a single positive influence, but a complex and diverse one.

Disasters pose a major threat to the development of human society, and the imprint of the disaster shock experience is reflected in the psychological and behavioral patterns of the people affected [34]. Frequent natural disasters such as earthquakes, hurricanes, and floods, as well as the emergence of viruses such as COVID-19, challenge farmers’ ability to cope with major disaster shocks [17,19,35,36]. Research on disasters should focus on not only their natural attributes but also their social attributes. Exploring the various aspects before and after the occurrence of a disaster from an economic perspective is an important direction for studying the social attributes of disasters [37]. Early studies focused on disaster management, such as disaster loss measurement and disaster risk assessment [38,39]. Since then, research on disasters’ effects on economic growth has become a mainstream trend, including research on both short- and long-term economic effects [40,41]. Researchers have also started to examine the effects of disaster shocks on individuals and households from a micro-level perspective [42,43], linking keywords such as “vulnerability”, “resilience”, and “toughness” with disasters [44,45,46] and studying the psychological perspectives of affected people. The severity of a disaster can be reflected in farmers’ post-disaster behavior in terms of their psychological health [34]. It is necessary to correctly understand and deal with disaster shock experiences to gain new ideas for supporting the development of disaster-affected people.

2.2. Hypothesis Regarding Disaster Shock Experience and Farmers’ Entrepreneurship

The experience of a disaster, which is an external shock, provides a strong incentive for individuals to modify their behavior in response to changes in their environment [47]. Surviving entrepreneurs who have lived through such disasters are a unique group [48]. One study suggested that those who were exposed to fatal starvation disasters during childhood developed long-term frugal behaviors, and experienced a reshaping of beliefs and preferences that would ultimately affect their professional decision making [49]. Populations affected by disasters can adjust to the declines in their income and survival environment by changing their behaviors or livelihood patterns, including their consumption decisions and migration patterns. Adversity serves as a catalyst for individual behavioral changes, motivating farmers to engage in entrepreneurship [50]. Furthermore, disasters enhance the correlation between entrepreneurial attitudes and intentions, thus preparing individuals for the transformation of entrepreneurial intention into entrepreneurial practice [51].

Entrepreneurship presents itself as a challenging career path [52]. Rural entrepreneurs face several constraints, including limited access to administrative resources, geographic location, economic costs, social network resources, and security conditions [53]. Capital is a crucial factor for farmer entrepreneurs, providing them with resources to engage in entrepreneurial activities. Various rural households, with different asset allocations, respond to disaster shocks in different ways. High-income households tend to rely on asset liquidation to secure stable incomes, whereas low-income households tend to reduce their consumption [54]. Disaster shocks also encourage more cautious saving [55]. Generally, farmers prefer low-risk agricultural production to maintain stable livelihoods, raising income through agricultural mechanization services and they might have a lower likelihood of starting their own businesses compared to other groups [56,57]. However, agriculture is not adequately prepared for disaster shocks due to the susceptibility of sustainable supply chains of agricultural products to interruption and the vulnerability of agricultural food production to external disturbances [19,58]. Thus, disaster shocks that disrupt opportunities for agricultural or other forms of livelihood might instead increase the likelihood of entrepreneurship, such as agricultural e-commerce [59]. Thus, self-employment is likely to become a new livelihood pathway.

In summary, the correlation between disaster shock experience and farmers’ entrepreneurship is not linear. Farmers’ employment choices after a disaster shock are highly elastic, and their employment decisions often depend on changes in labor market rewards [60]. As the negative effects of disasters become weaker over time, cross-sectoral labor mobility becomes a new option, with labor moving from the severely affected production sector to another production sector [61]. This means that the likelihood of farmers moving from agricultural production toward business and industry increases, which means farmers’ entrepreneurship might increase as well. Thus, there could be a U-shaped relationship between disaster shock experience and farmers’ entrepreneurship.

Figure 1 illustrates the relationship between disaster shock experience and farmers’ entrepreneurship. The red curve represents the tendency of disaster shocks to inhibit farmers’ entrepreneurship: the destructive nature of disaster shocks leads to a reduction in the resources and endowments required for entrepreneurship, while the high-risk nature of entrepreneurship is different from the risk preferences of farmers, which shows an overall inhibitory tendency. The green curve shows the U-shaped relationship between disaster shock experience and farmers’ entrepreneurship over time: when farmers first experience disaster shocks, their productive activities are destroyed, and their first choice is to work with lower risk and faster pay rather than start a business. Entrepreneurial momentum falls sharply as farmers build entrepreneurial capital through post-disaster recuperation (post-disaster reconstruction, or zone I in Figure 1). After a period of recovery, the negative effect of the disaster continues to diminish to a critical point, and farmers accumulate wealth through production and management to reach the initial conditions to be able to start a business. This represents the optimal point for farmers to start their own businesses after the disaster shock (zone II: entrepreneurial preparation). After the critical point, the proportion of farmers who are eligible to start their own businesses starts to grow significantly and rapidly as farmers start their own businesses and intersectoral transfers occur. Further, post-disaster social assistance provides more theoretical and practical knowledge and ways to increase income for farmers who start their own businesses (zone III: entrepreneurial practice). Over time, however, more and more farmers become engaged in industrial and commercial operations, leading to increased entrepreneurial competition among farmers. The market reaches saturation, and the growth rate of entrepreneurship slows down (zone IV: free choice).

Hypothesis 1.

There is a U-shaped relationship between disaster shock experience and farmers’ entrepreneurship over time.

2.3. Mediating Role of Risk Appetite

The aftermath of a disaster leaves a psychological imprint on farmers that results in a more cautious approach to future risk-taking. This can be detrimental to rural entrepreneurship [62]. Farmers are intrinsically motivated to turn their entrepreneurial intentions into practices through factors such as risk attitudes, resilience, and entrepreneurial spirit. However, external entrepreneurial forces such as government support, economic benefits, and social benefits are crucial for evaluating farmers’ entrepreneurial performance. A person’s attitude toward risk can affect their career choices [63]. Thus, a farmer’s risk appetite serves as the internal foundation for their entrepreneurship, and disaster shocks can affect their ability to identify business opportunities and execute them [50]. Therefore, we analyze the mediating role of risk appetite in the relationship between disaster shock experience and farmers’ entrepreneurship.

The human inclination toward risk-taking is subject to alteration through experience and perception. A profound, firsthand experience of disaster can lower risk appetite among the affected populace [64]. Through the short-term effects on noncognitive skills, individuals exhibit significant risk aversion characteristics [65]. Ahsan et al. [66] likewise found that natural disasters tend to lower people’s risk propensity. The experience of starvation during early childhood can also reshape one’s preferences, influencing career choices in the direction of professions with lower risk. Using an instrumental variable strategy, Blasio et al. [67] found that individuals who experienced earthquakes were more inclined to be risk-averse, which in turn inhibited their likelihood of becoming entrepreneurs. It has also been hypothesized that individuals who have just experienced a calamity shock might become more risk-averse, which, in turn, could encourage entrepreneurship, as shown by Cramer et al. [63]. Based on social learning theory, Zhou et al. [68] found that disaster shock experience positively affected risk perception. One’s conduct is directly influenced by one’s sense of the risk posed by natural disasters [69]. Therefore, disaster shock experience can play a crucial role in influencing farmers’ risk appetites when making business decisions. Research has confirmed that individual risk appetites can explain why people make different risk decisions and individual risk attitudes can favorably affect entrepreneurial activities [70]. Furthermore, risk propensity exhibits significant regional variation, with a higher likelihood of entrepreneurship in rural areas and a lower likelihood of entrepreneurship in urban areas. Given that starting a business is a high-risk activity, an entrepreneur’s risk attitude can influence every decision. Disaster shocks have a significant effect on an individual’s risk preferences, which play a crucial role in the relationship between disaster shock experience and farmers’ entrepreneurship.

Hypothesis 2.

The correlation between disaster shock experience and farmers’ entrepreneurship is influenced by risk appetite.

2.4. Moderating Role of Government Support

According to the career choice model [71], households only consider entrepreneurship when the benefits of being an entrepreneur outweigh those of being a wage earner. Expanding on this theory, Bianchi et al. [72] established a theoretical relationship between household financial resources and entrepreneurship. For instance, entrepreneurs who have access to financial resources are better able to allocate resources, resulting in greater output. Thus, the supply of resources can increase the likelihood of farmers establishing businesses. Drawing on these theories and considering the current situation in China, we identify government support as a moderating variable in farmers’ entrepreneurship.

Against the background of climate change, sound government policies are essential for mitigating disaster shocks and achieving sustainable development goals by reducing vulnerability and increasing resilience [73,74]. By creating a favorable entrepreneurial environment, the government can reduce the effect of disasters on entrepreneurs, mitigate disaster shocks, and improve the expected outcomes of entrepreneurial ventures, thus significantly increasing the likelihood of entrepreneurial success. Government support primarily fosters the potential for farmers’ entrepreneurship by enhancing their financial status after experiencing a disaster. In the case of migrant workers who returned to their hometowns, it has been shown that government financial support provides a foundation for them to launch businesses [75]. A government’s post-disaster reconstruction policy provides protection for households against disaster shocks, curbing the reduction in household income, and the income gap among migrant workers, thereby controlling labor outflow, which minimizes the negative effects of labor shortages [76]. Government relief can also raise the hopes of affected farmers and positively affect entrepreneurship, not only at the material level (e.g., financial resources) but also at the psychological level of capital [77]. As a result, government support significantly mediates the relationship between the risk of disaster shocks and farmers’ entrepreneurship [78]. Studies examining government aid for entrepreneurship have commonly found that financial assistance is better suited for the secondary and tertiary sectors, whereas policy support and training are more fitting for entrepreneurship in the primary sector, particularly for farmers. Through government support, farmers who experience disaster shocks can acquire more capital to overcome barriers to entrepreneurial activity, thus reducing the hindering effect of such shocks and facilitating entrepreneurial endeavors.

Hypothesis 3.

Government support plays a positive moderating role between disaster shock experience and farmers’ entrepreneurship.

Figure 2 shows that the correlation between disaster shock experience and farmers’ entrepreneurship follows both direct and indirect paths. The mediating and moderating variables are classified based on whether the factors influencing farmers’ entrepreneurship are directly affected by the “experience” of disaster. The mediating variable is risk appetite, which is influenced by the “experience” of disaster. Moreover, the moderating variable is government support, which is triggered by “disaster shock”. Farmers’ risk appetite changes in response to their experience of disaster shocks, influencing their entrepreneurial activities. External factors, such as government support, regulate farmers’ entrepreneurship in response to “disaster shock”.

3. Materials and Methods

3.1. Data Sources

This study uses information obtained from the 2019 CHFS database pertaining to the entrepreneurial activities of 39,113 farm households. The CHFS uses a three-stage stratified sampling method, proportional to population size, aiming to collect micro-level data on Chinese households using modern survey techniques and tools. The sample covers 29 Chinese provinces, autonomous regions, and municipalities (Tibet, Xinjiang, Hong Kong, Taiwan, and Macao are excluded). The survey includes demographic characteristics, assets and liabilities, insurance and protection, expenditures and income, financial knowledge, grassroots governance, and subjective evaluations, among other areas. The CHFS uses various measures to control errors in the sampling and non-sampling processes; its data can be considered representative and reliable. The baseline regression data are sourced from 2019 data.

Data collation primarily involves amalgamating the database horizontally by family ID and eliminating the data with discrepant family ID, and retaining rural households with a rural household registration and farm households originally registered as rural households in the unified household registration. We also retain rural households between the ages of “18 and 65”, based on the legal age of majority and of retirement in China. Furthermore, when the sample is large, to mitigate the effect of outliers on the study results, continuous variables are usually shrunken, which prevents model errors and improves the significance of the model. Thus, in this study, 1% shrinkage is applied to the total income data. Finally, there are technical access failures, invalid information, nonresponse and other problems in the survey data process that can lead to missing data, which often results in unreliable output and inaccurate results [79]. After eliminating missing values using the deletion method, we obtained a total of 39,113 valid data sets.

3.2. Variable Selection and Assignment

3.2.1. Dependent Variable

The dependent variable examined in this study is farmers’ entrepreneurship. There are various definitions of farmers’ entrepreneurship, such as rural households engaging in non-farm business activities [80]. Following Yin [52], we define farmers’ entrepreneurship as farmers engaging in self-employment or business operation that does not include agricultural production and operation (e.g., agriculture, forestry, animal husbandry, fishery). If the farmer is engaged in self-employment or business operation, the farmer is considered to be engaged in business operation; otherwise, the farmer is considered to not be engaged in business operation.

3.2.2. Independent Variables

The independent variable utilized in this study is disaster shock experience. This paper adopts Kellenberg’s [81] method of measuring sudden-onset disasters in which the disaster shock experience is divided into three parts: disaster shock, disaster level, and disaster shock experience. Disaster shock is defined as whether a sudden major disaster shock event has occurred. We determine whether a disaster shock occurred based on the answer in the questionnaire—Yes = 1; No = 0—where 1 indicates that a major sudden disaster such as an earthquake, tsunami, typhoon, landslide, fire, major disease outbreak, or household economic shock, has occurred. Disaster shock experience is defined as having experienced a sudden major disaster shock event, and the year of disaster shock occurrence is selected as a proxy variable for the time of disaster shock, representing farmers who had experienced a disaster shock. A value of 1–6 was assigned to the questionnaire data; 0 means the disaster time is not in this interval; the larger the value, the further the time of experiencing disaster shock is from the year under investigation. In addition, statistics about the types of disaster shocks are gathered to measure the level of disaster shocks. Owing to the limited range of values, the degree of disaster shock was assigned 0, 1, or 2; the larger the value, the greater the level of shock experienced.

3.2.3. Other Variables

The mediating variable examined in this study is risk appetite, which is defined as a “tendency” based on past accumulation. To measure risk appetite, we define risk appetite as one’s expectation of a positive or negative outcome and follow Qi [66,82], selecting the question, “If you have a sum of money to invest, which investment project would you be most willing to choose?” The responses measuring investment appetite serve as proxies for risk appetite: 1 = high risk, high return project; 2 = slightly higher risk, slightly higher return project; 3 = average risk, average return project; 4 = slightly lower risk, slightly higher return project; 5 = not willing to take any risk; and 6 = do not know. Meanwhile, risk perception has a moderating effect on risk appetite [83]. We select the questions, “Which do you think is more risky, main board stocks or growth enterprise market stocks (GEM)?” and “Which do you think is more risky, equity funds or debt funds?” The responses to these two questions were geometrically averaged to measure the level of farmers’ perception of risk in multiple dimensions and further determine farmers’ risk appetite.

Our moderating variable is government support, which encompasses both project and financial assistance. Since farmers who experience disaster shocks might receive multiple subsidies, accounting for only one type of subsidy might lead to omitted variables. Therefore, we incorporate multiple subsidies into the analysis, based on responses to the question, “Which of the following social relief and subsidies did your household receive from the government last year, converted into cash or kind?” The responses include 1 = minimum living security (low-income security); 2 = temporary assistance; 3 = five-guarantee household subsidy; 4 = medical assistance; 5 = poverty alleviation; 6 = natural disaster assistance; and 7 = pension. We take the logarithm of the subsidy amount received as a proxy variable to measure the degree of government support.

To minimize bias, we include control variables at the individual, household, and regional levels, as shown in

Table 1. Variable definitions and assignments.

Variable Category	Variables	Variable Code	Variable Assignment
Dependent variable	Farmers’ entrepreneurship	Ent_p	Yes = 1; no = 0
Independent variables	Disaster shock	Dis_s	Yes = 1; no = 0
	Disaster level	Dis_l	[0, 2]
	Disaster shock experience	Dis_t	[0, 6]
Intermediate variable	Investment appetite	Ris_a	[1, 6]
	Risk perception	Ris_p	Main Board/Equity-biased funds = 1; GEM/Debt-biased funds = 2; Have not heard of the main board/equity-biased funds = 3; Have not heard of the GEM/debt-biased funds = 4; Have not heard of both = 5; Same size = 6;
Moderating variable	Government support	Gov_s	[0, 12.30]
Individual control variables	Gender	Gen	Male = 1; female = 0
	Age	Age	18–65 years old
	Education	Edu	Never went to school = 1; elementary school = 2; junior high school = 3; high school = 4; secondary school/vocational high school = 5; college/high school = 6; undergraduate = 7; master’s degree = 8; doctoral degree = 9
	Marital status	Mar_s	Married = 1; partnered but not married = 2; divorced = 3; separated but not divorced = 4; widowed = 5; unmarried = 6
Family control variables	Targeted poverty alleviation	Tar_p	Yes = 1; no = 0
	Total income	Tot_i	[−1.89, 13.06]
Regional control variable	Area	Area	[1, 4]

. At the individual level, the variables include respondent characteristics, such as the gender, age, education level, and marital status of the household head. At the household level, the variables mainly pertain to household characteristics, such as whether the household receives targeted poverty alleviation and annual household income. Finally, at the regional level, provinces are classified into four regions (eastern, central, western, and northeastern) to control for any potential regional differences.

3.3. Empirical Models

We examine the correlation between disaster shock experience and farmers’ entrepreneurship. For non-linear relationships, researchers often add a squared term to the model, but it is not rigorous to judge U-shaped and inverted U-shaped relationships only by determining whether the squared term is significant or not; Lind and Mehlum [84] argue that the squared term is significant when the true relationship is monotonic and convex. To test the relationship between the variables, we employ a probit model, and U- shaped and inverted U-shaped connection tests are conducted using the general framework developed by Lind [84]. Owing to the large standard errors associated with traditional regression analysis when estimating mediating roles, Iacobucci et al. [85] demonstrated through a series of Monte Carlo simulations that using regression analysis (REG) has serious drawbacks compared to using structural equation modeling (SEM); that is, the standard errors of the coefficients of the mediated paths obtained from regression analysis are consistently larger than those obtained through the SEM approach, and the assumptions are stringent. The Sobel test requires the assumption of an approximately normal symmetric distribution, and thus low-test power, but SEM can estimate all model parameters, including observable and latent variables. Thus, SEM is the best framework for mediating effects analysis [86,87]. Therefore, we test the mediating effects based on SEM and, to test the regression coefficients, we construct regression models of independent variables on dependent variables, independent variables on mediating variables, and independent variables and mediating variables on dependent variables. We also construct interaction terms for the regression tests of the variables using moderating effect test models. The specific models are as follows:

$$Y_{i,1}={\tau }_0+{\tau }_1{Z}_i+{\tau }_2{C}_i+{\mu }_i+{\epsilon }_i$$

(1)

$$Y_{i,2}={\alpha }_0+{\alpha }_1{Z}_i+{\alpha }_2{Z}_i\times Z_i+{\alpha }_3{C}_i+{\mu }_i+{\epsilon }_i$$

(2)

$$M_i={\beta }_0+{\beta }_1{Z}_i+{\beta }_2{Z}_i\times Z_i+{\beta }_3{C}_i+{\mu }_i+{\epsilon }_i$$

(3)

$$Y_{i,3}={\gamma }_0+{\gamma }_1{Z}_i+{\gamma }_2{Z}_i\times Z_i+{\gamma }_3{M}_i+{\gamma }_4{C}_i+{\mu }_i+{\epsilon }_i$$

(4)

$$Y_{i,4}={\theta }_0+{\theta }_1{Z}_i+{\theta }_2{Z}_i\times Z_i+{\theta }_3{T}_i+{\theta }_4{Z}_i\times T_i+{\theta }_5{Z}_i\times Z_i\times T_i+{\theta }_6{C}_i+{\mu }_i+{\epsilon }_i,$$

(5)

$Y_i$ serves as the dependent variable, denoting whether farmer $i$ is engaged in entrepreneurship, where $1$ denotes the correlation of whether or not disaster is experienced on rural household $i$, $2$ denotes the nonlinear relationship between disaster shock experience and rural household $i$, $3$ denotes the addition of mediating variables to $i$, and $4$ denotes the addition of moderating variables to $2$. $Z_i$ is the independent variable representing the experience of disaster shocks. $M_i$ is the mediating variable, indicating the mediating mechanism of risk appetite. $T_i$ is the moderating variable, indicating the moderating mechanism of government support. $C_i$ is a collection of control variables indicating observable individual and household characteristics of rural household $i.$ ${\mu }_i$ is the regional control variable controlling for measurement errors caused by regional differences; ${\epsilon }_i$ is an error term measuring unobservable factors affecting rural entrepreneurship, and $\tau ,\alpha ,\beta ,\gamma ,\theta$ are parameters requiring estimation. Equations (1) and (2) are subjected to U-shaped testing, where the significance of ${\tau }_1$ denotes the correlation between disaster shock experience and farmers’ entrepreneurship, while the significance of ${\alpha }_2$ indicates that the nonlinear regression of the chosen variables is consistent with a U-shape or inverted U-shape. Equations (2)–(4) test for mediating roles. If ${\alpha }_2$ is significant and ${\beta }_2,{\gamma }_2$ also pass the significance test, then the nonlinear mediating role between the two is significant. If ${\alpha }_2$ is not significant, ${\alpha }_2$ is significant and ${\beta }_1,{\gamma }_2$ are significant, then the linear mediating role between the variables is significant. The main test for the moderating role is the interaction term test, which tests the significance of ${\theta }_4,{\theta }_5$ in Equation (5). If it is significant, then the variable has a moderating role.

4. Empirical Results and Analysis

4.1. Descriptive Statistics

Table 2. Basic descriptive statistics of the variables.

Variables	Obs	Mean	SD	Min	Max
Ent_p	39,113	0.155	0.362	0	1
Dis_s	39,113	0.181	0.385	0	1
Dis_l	39,113	0.200	0.445	0	2
Dis_t	39,113	0.666	1.558	0	6
Ris_a	39,113	4.633	1.155	1	6
Ris_p	39,113	0.727	1.700	0	6
Gov_s	39,113	1.113	2.735	0	12.30
Age	39,113	44.59	13.29	19	65
Gen	39,113	0.503	0.500	0	1
Edu	39,113	3.214	1.535	1	9
Mar_s	39,113	1.977	0.798	1	6
Tar_p	39,113	0.390	1.783	0	12.26
Tot_i	39,113	10.63	1.301	−1.89	13.06
Area	39,113	2.184	0.977	1	4

Note: Obs: observations of variables; Mean: average value of the variables; SD: standard deviation; Min and Max: minimum and maximum values of the variables, respectively.

presents a preliminary analysis of the data showing that a total of 39,113 farmers were interviewed, with an almost equal gender distribution of 49.69% male and 50.31% female and an average age of about 44 years. The majority of respondents, about 84.31%, report nine years of compulsory education, while only 15.95% of households are married. Nearly 80% of the farm households have partners but are not married. About 73.70% of family members have jobs, whereas only 18.26% live and work outside the household, with most remaining in the household. The mean value of household income is 71,668.48 yuan, and the proportion of those receiving targeted poverty alleviation is about 5%. The mean value of farming households’ entrepreneurship is 0.155, indicating the concentration trend of entrepreneurship. Based on the 0–1 variable of farming households’ entrepreneurship, this indicates that the majority of farming households remain engaged in traditional production and operation (e.g., agriculture, forestry, animal husbandry, fishery), and the proportion of farming households’ entrepreneurship is 15.5%. This means agriculture is still the main source of livelihood in rural areas, while the low proportion of farmers’ entrepreneurship may be due to limitations in rural development conditions and the lack of entrepreneurial awareness, knowledge, capital, information, effective policy guidance, and other entrepreneurial resources. The influence of traditional concepts in rural areas may also hinder their entrepreneurship. The proportion of those who have experienced disaster is about 18%, and the mean value of the degree of disaster is about 0.2. The standard deviation of the time of disaster is 1.558, indicating that although people with disaster experience do not constitute the majority, the time of disaster fluctuates widely and is not concentrated around the same time. The mean value of farmers’ risk preference is 4.633, which, based on the scale of 1–6, indicates that the majority of farmers choose not to take any risk compared with gaining income through investment, accounting for 50.17%. Only 15% of households receive government support. The proportion of people in the eastern and western regions is basically the same, both around 34%, while 23.88% are in the central region and 7.34% in the northeastern region.

4.2. Baseline Regression Analysis of Disaster Shock Experience on Farmers’ Entrepreneurship

We use ordinary least squares (OLS) and probit regressions to analyze the correlation between disaster shock experience and farmers’ entrepreneurship. Given the possibility of multicollinearity among the selected variables, correlation and covariance tests are conducted on the independent variables. The results show a significant negative correlation between disaster shock experience and farmers’ entrepreneurship, and the reliability coefficient between any two variables other than those calculated is less than 0.4. This indicates that the model does not suffer from multicollinearity. The results of further covariance matrix and significance tests of the variables for reliability confirm this conclusion. Appendix A 

Table A1. Reliability coefficient.

(obs = 39,113)
	Ent_p	Dis_s	Dis_l	Dis_t	Ris_a	Gov_s	age	gen	Edu	Mar_s
Ent_p	1
Dis_s	−0.0401	1
Dis_l	−0.0394	0.956	1
Dis_t	−0.0364	0.892	0.858	1
Ris_a	−0.0842	0.0131	0.0109	0.0107	1
Gov_s	−0.0667	0.152	0.154	0.138	0.0446	1
age	−0.0632	0.00870	0.00590	0.00580	0.0988	−0.00190	1
gen	0.00100	0.00230	0.00170	0.00210	0.00110	0.00990	−0.0326	1
Edu	0.0837	−0.0604	−0.0605	−0.0504	−0.147	−0.107	−0.513	0.0907	1
Mar_s	−0.0150	0.0434	0.0466	0.0381	0.0187	0.0220	0.365	−0.0707	−0.268	1
Tar_p	−0.0562	0.0696	0.0750	0.0752	0.0346	0.355	−0.0115	0.0109	−0.0836	0.0317
Tot_i	0.0975	−0.0779	−0.0759	−0.0687	−0.127	−0.0853	−0.147	0.0122	0.225	−0.0726
area	−0.0759	0.0732	0.0725	0.0547	0.0327	0.101	−0.0310	0.00520	−0.0869	0.00410
	Tar_p	Tot_i	area
Tar_p	1
Tot_i	−0.0716	1
area	0.0867	−0.134	1

Note: Obs.: observations of variables.

shows the specific results. The variance inflation factor (VIF) values of the variables are all less than 10, and 1/VIF is less than 1, indicating the absence of cointegration issues. Validity and reliability between the variables are thus further verified.

Moreover, an overall regression analysis is performed on the variables, as shown in

Table 3. Correlation between disaster shock experience and farmers’ entrepreneurship.

	(1)	(2)	(3)	(4)
Variables	Ent_p	Ent_p	Ent_p	Ent_p
Dis_s	−0.167 ***	−0.106 ***	−0.038 ***	−0.022 ***
	(−8.01)	(−4.98)	(−8.50)	(−5.09)
Dis_l	−0.144 ***	−0.090 ***	−0.032 ***	−0.019 ***
	(−7.81)	(−4.82)	(−8.43)	(−4.99)
Dis_t	−0.093 ***	−0.061 ***	−0.021 ***	−0.013 ***
	(−4.98)	(−3.17)	(−5.38)	(−3.35)
Dis_t $\times$ Dis_t	0.011 ***	0.007 *	0.003 ***	0.002 **
	(2.93)	(1.86)	(3.21)	(2.00)
Constants	−0.989 ***	−6.943 ***	0.162 ***	−1.160 ***
	(−118.15)	(−16.05)	(78.63)	(−12.37)
Other controls	No	Yes	No	Yes
R2	0.002	0.030	0.002	0.024
Observations	39,113	39,113	39,113	39,113

Note: * p < 0.05, ** p < 0.01, *** p < 0.001; t-values in parentheses.

, demonstrating the influence of disaster shock experience on farmers’ entrepreneurship. Columns (1) and (2) show the probit model’s overall regression, testing the correlation between disaster shock experience and farmers’ entrepreneurship. The results indicate that disaster shock significantly inhibits entrepreneurship at the 1% statistical level. Specifically, for each disaster experienced, entrepreneurship declines by 16.7%, consistent with prior research. Furthermore, the inhibitory relationship intensifies as the likelihood of entrepreneurship among farmers decreases by 14.4% with the deepening of disaster severity.

To verify the U-shaped relationship between disaster shock experience and entrepreneurship, the U-shape test was conducted, verifying the duration of disaster. The assumption of a U-shaped relationship requires that the relationship trend decreases at low values and increases at high values of the interval and has only one extreme point. To test the existence of a U-shaped relationship, we can examine the sign of the second-order derivative, construct a confidence interval for the minimum point, and then determine whether this confidence interval is within the range of values taken. The U-shape test’s result has an extreme value of 4.1257 and significant at the 1% statistical level. However, the slope in the result has a positive sign in the interval, and the second order derivative is greater than 0, indicating a U-shaped relationship. The conclusion remains significant after adding control variables in Column (2).

Columns (3) and (4) show the baseline regression results using the OLS model. Compared with the probit model, the correlation between disaster shock experience and farmers’ entrepreneurship shows the same trend, and the level of disaster is also consistent with a negative correlation. Additionally, the relationship between disaster shock experience and farmers’ entrepreneurship is also consistent with a U-shaped relationship. Comparing the adjusted R², we find that the probit model fits better after adding control variables; thus, we use the probit model for the baseline regression. Columns (1) and (3) indicate that no control variables are added and Columns (2) and (4) indicate that control variables are added.

The regression results of both models are consistent with our theoretical expectation, thus verifying Hypothesis 1. According to the time of the disaster, farmers’ entrepreneurship shows a trend of first decreasing and then increasing.

4.3. Endogeneity Analysis of Disaster Shock Experience on Farmers’ Entrepreneurship

Considering the dissimilarity in disaster risk exposure between entrepreneurial and agricultural production activities, it is likely that many households opt for entrepreneurship as a means of mitigating future disaster consequences. Furthermore, entrepreneurship reduces household vulnerability and increases risk-taking capacity. Hence, reverse causality between entrepreneurship and disaster shock experience cannot be overlooked. Additionally, household decisions about whether to pursue entrepreneurship are often influenced by factors such as entrepreneurial ability, entrepreneurial confidence, successor traits, personal preferences, the speed and extent of the acceptance of new ideas, and the existence of an entrepreneurial diffusion effect in the surrounding environment. These factors might not be accurately observed, and there could be omitted variables and measurement errors owing to information asymmetry during data collection. It is important, then, to consider the presence of endogeneity between variables.

In light of the above, we use the instrumental variable approach to address the endogeneity issue. Following Liu [88], we select “1959–1961 mortality” as the instrumental variable for constructing the provincial disaster impact degree. There are two reasons for choosing this instrumental variable. First, 1959–1961 is the period described as having “three years of natural disasters” in China, and the extent of disaster damage during these three years was not the same everywhere. The influence of disaster shocks on those who experience them may also vary, and the degree of memory of the experience may also be different. Second, the extent of the disaster during 1959–1961 did not affect current entrepreneurial decisions. Thus, the instrumental variable satisfies both the relevance and exogeneity conditions. Initially, we use the Hausman test to test endogeneity, and the resulting statistical value of −21.02 does not reject the original hypothesis of “all dependent variables are exogenous”. This indicates the existence of an endogeneity problem. Consequently, we use the instrumental variable method to solve this problem. The results of the instrumental variable test indicate that the LM statistic is 50.092 with a p-value of less than 0.05. The original hypothesis of “unidentifiable” is rejected, and the p-value is less than 0.05, indicating that there is no overfitting problem.

Table 4. Regression results of instrumental variables.

	(5)	(6)	(7)
Variables	Ent_p	Ent_p	Ent_p
Dis_s	−0.605 ***
	(−3.88)
Dis_l		−0.452 ***
		(−3.85)
Dis_t			−0.223 ***
			(−4.20)
KP−LM statistics	50.092	61.397	38.987
	(0.000)	(0.000)	(0.000)
CD−Wald F statistics	16.714	20.491	13.005
Overidentification test	33.804	37.874	21.051
	(0.000)	(0.000)	(0.000)
Other controls	Yes	Yes	Yes
Observations	39,113	39,113	39,113

Note: *** p < 0.001; t-values in parentheses.

presents the regression results for specific instrumental variables, where Columns (5)–(7) represent the correlation of disaster shock, disaster level, and disaster shock experience with farmers’ entrepreneurship after the inclusion of instrumental variables. The regression results indicate that disaster shock experience still significantly affects entrepreneurship after applying instrumental variable analysis.

4.4. Robustness Tests for Correlation between Disaster Shock Experience and Farmers’ Entrepreneurship

To test the robustness of our results, we conduct robustness tests by replacing variables and models.

Table 5. Robustness test results.

	(8)	(9)	(10)	(11)
Variables	Ent_p	Ent_p	Ent_p Income	Ent_p Income
Dis_s	−0.070 **	−0.148 ***	−0.454 ***	−0.453 ***
	(−2.00)	(−4.43)	(−8.94)	(−7.02)
Dis_l			−0.386 ***	−0.385 ***
			(−8.70)	(−7.06)
Dis_t	−0.086 **	−0.094 ***	−0.222 ***	−0.221 ***
	(−1.91)	(−3.15)	(−4.65)	(−3.69)
Dis_t $\times$ Dis_t	0.017 *	0.014 **	0.024 **	0.024 *
	(1.68)	(2.19)	(2.41)	(1.89)
Constants	−6.803 ***	−6.849 ***	−6.943 ***	−1.160 ***
	(−15.16)	(−15.81)	(−16.05)	(−12.37)
Other controls	Yes	Yes	Yes	Yes
R2	0.043	0.032	0.102	0.316
Observations	45,818	39,113	5237	5237

Note: * p < 0.05, ** p < 0.01, *** p < 0.001; t-values in parentheses.

presents the results. Column (8) shows the regression results of replacing the explanatory variables. Drawing on Liu et al. [88], we use data from the China Household Tracking Survey database for the “China 1959–1961 famine event”, and “famine experience” is further validated as a proxy for disaster shock experience by replacing the independent variables. The results show that disaster shocks do inhibit farmers’ entrepreneurship. The results are significantly negative at the 5% level, and the U-shaped relationship between disaster shock experience and farmers’ entrepreneurship is significant. Column (9) shows the regression results using only the experience of natural disasters as the core independent variable. To exclude the influence of other disaster shock experiences, we only use natural disaster experience (e.g., earthquake, tsunami, typhoon, landslide) as the core independent variable for the empirical analysis. Column (10) shows the results of replacing the dependent variables with the logarithm of income earned from entrepreneurship to examine the performance of entrepreneurship, and we regress it using the Tobit model. Column (11) shows the regression results using the OLS model. The results are consistent with the previous conclusions, indicating that our results are robust.

4.5. Intermediary Relationship

Table 6. Mediating roles of risk appetite.

	(12)	(13)	(14)	(15)	(16)
Variables	Ris_a	Ent_p	Ent_p	Ris_a < 4	Ris_a > 3
Dis_s	0.039 ***	−0.037 ***	−0.108 ***	−0.079 *	−0.114 ***
	(2.58)	(−7.79)	(−5.07)	(−1.67)	(−4.76)
Dis_t	0.009 **	−0.009 ***	−0.063 ***	−0.077 *	−0.058 ***
	(2.11)	(−7.45)	(−3.31)	(−1.81)	(−2.66)
Dis_t $\times$ Dis_t	0.002 **	−0.002 ***	0.008 **	0.015 *	0.005
	(1.96)	(−5.89)	(1.98)	(1.71)	(1.22)
Ris_a		−0.026 ***	−0.078 ***
		(−15.88)	(−11.89)
Ris_p		0.006 ***	0.009 **
		(5.64)	(2.01)
Constants	4.626 ***	0.283 ***	−6.260 ***	−5.468 ***	−7.104 ***
	(681.59)	(35.02)	(−14.36)	(−5.90)	(−14.40)
Other controls	No	No	Yes	Yes	Yes
R2			0.034	0.018	0.032
Observations	39,113	39,113	39,113	6950	32,163

Note: * p < 0.05, ** p < 0.01, *** p < 0.001; t-values in parentheses.

presents the results for the mediation analysis of risk appetite. Columns (12)–(14) present the analysis of the mediating role of disaster shock experience and farmers’ entrepreneurship using SEM. The regression results of the mediating role of risk perception and investment propensity indicate that disaster shock experience inhibits farmers’ entrepreneurship by altering their risk appetite. The likelihood of farmers’ entrepreneurship decreases by 3.7% owing to risk appetite, which is statistically significant at the 1% level. Further analysis of the linear and nonlinear mediating roles of disaster shock experience validation results shows that the ratio of the mediating role to the total effect is 0.027, indicating that risk appetite has a mediating role of about 3% for disaster shock experience. The ratio of mediating role to direct role is 0.028, which means the mediating role is 0.028 times larger than the direct one. Therefore, risk appetite plays a mediating role, both linear and nonlinear, between disaster shock experience and farmers’ entrepreneurship. Hypothesis 2 is therefore supported.

Columns (15) and (16) show the correlation between disaster shock experience and entrepreneurship for different risk appetites, where the classification criterion is whether the farmer is risk-averse, with column (15) indicating risk-averse and column (16) indicating not risk-averse. The results show that the presence or absence of disaster has a smaller disincentivizing correlation with risk-preferring farmers and a larger disincentivizing correlation with risk-averse farmers. There is a U-shaped relationship between risk-preferring farmers and entrepreneurship, while only a linear inhibitory correlation is observed for risk-averse farmers. Figure A1a in the Appendix A shows the mediating role of risk preferences and the heterogeneity of different risk appetites. It clearly shows that the threshold of farmers’ entrepreneurship shifts to the left; the risk-averse farmers are less inhibited; and the U-shaped curve is more upward.

5. Further Analysis

5.1. Moderating Roles

Table 7. Moderating roles of government support.

	(17)	(18)
Variables	Ent_p	Ent_p
Dis_s	−0.129 ***
	(−5.38)
Dis_t		−0.084 ***
		(−3.86)
Dis_t $\times$ Dis_t		0.012 ***
		(2.70)
Gov_s	−0.033 ***	−0.031 ***
	(−8.19)	(−7.91)
Dis_s $\times$ Gov_s	0.030 ***
	(4.26)
Dis_t $\times$ Gov_s		0.023 ***
		(3.85)
Dis_t $\times$ Dis_t $\times$ Gov_s		−0.004 ***
		(−3.27)
Constants	−6.798 ***	−6.807 ***
	(−15.68)	(−15.70)
Other controls	Yes	Yes
R2	0.032	0.032
Observations	39,113	39,113

Note: *** p < 0.001; t-values in parentheses.

presents the moderating role of government support. Column (17) shows the correlation between disaster shocks and farmers’ entrepreneurship after adding government support. The results show that government support can significantly and positively moderate the inhibitory correlation between disasters and farmers’ entrepreneurship; the coefficient of the interaction term is positive at the 1% level of significance. Column (18) shows that the interaction term has an inverted U-shaped trend, indicating that government support mitigates the downward trend in the downward phase and the upward trend in the upward phase of farmers’ entrepreneurship. However, continued government support might lead to a dependence mentality, which is not conducive to entrepreneurship in the latter two stages. To measure the degree of farmers’ dependence mentality, we calculated farmers’ livelihood dependence on government support and livelihood dependence on entrepreneurship separately follow Xiong [89]. We found that an increase in livelihood dependence on government support can weaken livelihood dependence on entrepreneurship. As the dependence of farmers’ livelihood on government support increases, the dependence of farmers’ livelihood on entrepreneurship decreases by 6.1%, which means that the more farmers depend on government support, the less favorable it is for entrepreneurship. In addition, in terms of the critical point, compared with the disaster shock experience in Column (2), the critical point of farmers’ entrepreneurship is significantly earlier by about 0.85 years and significant at the 1% statistical level. This indicates that government support significantly contributes to the U-shaped relationship between disaster shock experience and farmers’ entrepreneurship, thus verifying Hypothesis 3.

5.2. Heterogeneity Analysis

5.2.1. Analysis of Different Disaster Shock Experiences

Disasters can encompass not only natural disasters, but also various social and uncertainty risks brought by such events. To examine the correlation between different disaster shock experiences and farmers’ entrepreneurship, we categorize disaster shocks into natural disasters (e.g., earthquake, tsunami, typhoon, landslide), major diseases (e.g., cancer), and household economic shocks (e.g., bankruptcy).

Figure 3 shows the correlations between different disaster shock experiences and farmers’ entrepreneurship. Natural disasters and major disease shocks both have the same inhibitory correlation with entrepreneurship (dashed lines). The U-shaped trend of disaster shock experience remains unchanged, with the tipping point for each being earlier than the tipping point for all disaster shock experiences; the tipping point reaches 3.5 years after experiencing a natural disaster and 3.0 years after experiencing a major disease shock. This suggests that the more disasters experienced, the longer the time needed to recover, and the later the threshold for entrepreneurship. The inhibitory correlation of household economic shocks is not significant. This is probably because the business and industrial operations conducted via entrepreneurship have become part of their livelihoods and do not change because of changes in household economic conditions. Farmers whose livelihoods are based on business and industrial operations are even less likely to change their existing production patterns to start a business. However, the threshold for experiencing household economic shocks is 2.25 years, which is in the range of 1.4–2.6 years for the left half of the U-shaped branch. This indicates that the correlation of household economic shocks is limited and that farmers can adjust quickly and reach the lowest point after about two years.

Appendix A 

Table A2. Correlation of different disasters on entrepreneurship.

Variables	(25)	(26)
	Ent_p	Ent_p
Natural disaster	−0.148 ***	−0.094 ***
	(−4.43)	(−3.15)
Dis_t $\times$ Dis_t		0.014 **
		(2.19)
Major disease	−0.091 ***	−0.081 ***
	(−3.18)	(−2.99)
Dis_t $\times$ Dis_t		0.013 **
		(2.32)
Household economic shocks	0.032(0.53)	0.221 ***(3.84)
Dis_t $\times$ Dis_t		−0.054 ***
		(2.58)
Constants	−6.304 ***	−6.309 ***
	(−14.47)	(−14.48)
Other controls	Yes	Yes
R2	0.034	0.034
Observations	39,113	39,113

Note: ** p < 0.01, *** p < 0.001; t-values in parentheses.

shows the specific regression results. Column (25) shows the correlations between different disaster shock experiences and farmers’ entrepreneurship using the probit model. The regression results indicate that disaster shocks, except for household economic shocks, can significantly inhibit entrepreneurship. Experiencing a natural disaster decreases the likelihood of starting a business by 14.8%, and experiencing a major illness decreases the likelihood of starting a business by 9.1%, which is significant at the 10% confidence level. Column (26) shows the coefficients of the variables obtained from the U-shaped test of the probit model. The regression results show that the correlations of natural disaster experience and major illness experience with farmers’ entrepreneurship are consistent with the U-shaped feature. As the time of disaster changes, the correlations between disaster shock experiences and farmers’ entrepreneurship are also weakened and then promoted.

5.2.2. Analysis of Individual- and Family-Level Differences

The inclusive entrepreneurship hypothesis [25,26] posits that innovative entrepreneurship among underprivileged groups is essential for sustainable economic growth; it is a significant strategy for eliminating poverty and promoting inclusive development. Here, we explore how various disadvantaged groups respond to disasters. The analysis of household heterogeneity is primarily based on household income and the level of poverty alleviation being targeted, while the examination of individual heterogeneity is primarily based on gender difference.

To divide the sample, we use the mean value of the logarithm of annual per capita household income as the dividing point to distinguish between high- and low-income groups [90]. Additionally, we use the mean value of the degree of targeted poverty alleviation as the dividing point to distinguish between low- and high-level targeted poverty alleviation households.

Table 8. Correlation between different genders, income levels, and degree of targeted poverty alleviation and farmers’ entrepreneurship.

	(19)	(20)	(21)	(22)	(23)	(24)
Variables	Man	Woman	Below Average	Above Average	Below Average	Above Average
Dis_s	−0.102 ***	−0.110 ***	−0.170 ***	−0.059 **	−0.122 ***	0.256 **
	(−3.40)	(−3.62)	(−4.85)	(−2.17)	(−5.58)	(2.56)
Dis_t	−0.013 **	−0.012 **	−0.021 ***	−0.004	−0.017 ***	0.019 *
	(−2.44)	(−2.28)	(−4.71)	(−0.67)	(−4.15)	(1.67)
Dis_t $\times$ Dis_t	0.002	0.001	0.003 ***	−0.000	0.002 **	−0.003
	(1.60)	(1.22)	(3.48)	(−0.07)	(2.53)	(−1.28)
Constants	−6.489 ***	−7.409 ***	−9.952 ***	−3.877 ***	−6.790 ***	−11.928 ***
	(−10.77)	(−11.88)	(−12.90)	(−7.14)	(−15.48)	(−4.44)
Other controls	Yes	Yes	Yes	Yes	Yes	Yes
R2	0.029	0.031	0.052	0.008	0.021	0.056
Observations	19,677	19,436	16,183	22,930	37,220	1893

Note: * p < 0.05, ** p < 0.01, *** p < 0.001; t-values in parentheses.

presents the heterogeneity results for gender, income level, and targeted poverty alleviation degree. Columns (19) and (20) show the regression results according to gender, revealing that the correlation of disaster shocks is more inhibitory for female entrepreneurship than for male entrepreneurship. This could be attributable to “son preference” mentality in traditional Chinese families, which might result in women being treated harshly owing to resource shortages following a disaster. Furthermore, the incidences of disability and illiteracy are higher among women [91], making it more difficult for them to start and maintain entrepreneurial activities. The nonlinear regression of disaster shock experience on different genders was not significant, and only a linear relationship existed. Appendix A Figure A1b presents a more intuitive result.

Columns (21) and (22) indicate the correlation between different income levels and entrepreneurship. Comparative analysis reveals that low-income households are more vulnerable to the negative correlations of disaster shocks and exhibit a U-shaped trend, with the threshold almost coinciding with the threshold of entrepreneurship for all households. High-income households are relatively less negatively affected by disaster shocks, and the nonlinear trend is not significant. A possible reason is that disasters change households’ asset allocation strategies and keep their balance sheets relatively balanced [47]. For low-income families, a sudden disaster that severely disrupts their already low means of production can force them to allocate other parts of their original start-up capital, thus hindering entrepreneurial activities until they can get over the hump and meet their start-up conditions. In addition, high-income households often smooth future consumption through savings and investment decisions to protect against uncertainty risks. This behavior tends to mitigate the negative correlations of sudden disasters, and so they are less inhibited. Appendix A Figure A1c shows a more intuitive result.

Columns (23) and (24) present the analysis of entrepreneurship heterogeneity across different levels of targeted poverty alleviation. Comparing the two columns, the inhibitory correlation of disaster shocks is higher for households with low levels of poverty alleviation, and the U-shaped trend is more pronounced after the disaster. This is probably because families with poverty alleviation experience hardship and families with low levels of poverty alleviation receive less financial support from poverty alleviation programs. Families with low levels of poverty alleviation are not entirely dependent on targeted poverty alleviation and therefore rely on their own entrepreneurial capital accumulation. The high level of poverty alleviation for households receiving more targeted poverty alleviation funding is more strongly facilitated by disaster shocks. Moreover, the nonlinear regression of disaster shock experience is not significant; thus, there is only a linear relationship. Appendix A Figure A1d shows a more intuitive result. A possible reason is that targeted poverty alleviation makes households more capable of entrepreneurship, not only to achieve the goal of poverty alleviation but also to prevent households from returning to poverty. Thus, targeted poverty alleviation policies can greatly influence farmers’ entrepreneurship.

6. Conclusions, Policy Implications, and Discussion

6.1. Conclusions

As the probability of disaster events increases under global climate change, studying the correlation between disaster shock experience and farmers’ entrepreneurship can provide a basis for formulating countermeasures. Based on CHFS 2019 data, our analysis identifies a U-shaped relationship between disaster shock experience and farmers’ entrepreneurship. In other words, the experience of disaster limits entrepreneurship in the short term but promotes it in the long term as the correlation of the disaster shock decreases. In addition, the relationship between disaster shocks and disaster level and entrepreneurship is linearly suppressed, which is consistent with existing research [15,16,17]. The same conclusions are obtained when we replace the database with the independent variable of disaster shock experience, replace the dependent variable with the logarithm of farmers’ entrepreneurship income, and replace the regression model based on robustness tests that address possible sources of endogeneity, such as bidirectional causality. We further find that risk appetite plays a mediating role between disaster shock experience and farmers’ entrepreneurship by changing the judgment of entrepreneurial risk. In addition, government support plays a positive moderating role by changing the financial situation for entrepreneurship and weakening the negative correlation of disaster shock experience. Here, the critical point of the U-shaped curve shifts to the left with government support. The heterogeneity analysis reveals that different disaster shock experiences also have a U-shaped relationship with farmers’ entrepreneurship. The inhibition correlation of entrepreneurship is 0.8% higher for females than males, and the inhibition correlation of disaster shock experience is significantly higher for low-income households than for high-income ones. Disaster shocks are also more conducive to entrepreneurial activities for families receiving high levels of targeted poverty alleviation.

6.2. Policy Implications

Our findings have policy implications that can guide decision making in promoting farmers’ entrepreneurship and building resilience in the face of disasters.

First, given the destructive characteristics of disaster shocks, rural areas need to improve their development capacity and disaster shock–bearing capacity. According to the characteristics of different disasters, disaster prediction, damage assessment, and post-disaster management should be undertaken in different categories to minimize disaster losses. We should strengthen the disaster awareness and disaster management skills of those who guide agricultural and other production in rural areas and teach farmers approaches to pre-disaster prevention and post-disaster recovery in their daily work to help minimize losses. At the same time, rural areas need to rely on their resource endowment advantages to identify breakthrough points for post-disaster recovery, address post-disaster livelihoods in terms of cross-industry mobility, and improve disaster response capacity by strengthening regional endowment advantages to increase economic resilience and improve disaster carrying capacity. This is also an important basis for improving the risk preferences of farm households.

Second, the U-shaped relationship suggests that, with proper understanding and response, disaster experiences can be converted into opportunities. The threshold of the negative correlation of disaster shocks is overcome, and the cognitive advantage of disaster experience is maximized to promote entrepreneurship and a correct understanding of the related risks and opportunities. We clarify the characteristics of high investment, easy failure, and long-term entrepreneurship. The obstacles to farmers’ entrepreneurship include factors such as poor infrastructure, low economic development, and ineffective policy systems. In addition, factors like family size, labor force, income, and arable land area directly influence farmers’ entrepreneurship, and personal factors such as gender, age, education level, and professional expertise of farmers’ entrepreneurship should be taken seriously. Farmers’ ability to cope with these characteristics can be improved through government entrepreneurship education, innovation training, guidance in the creation process, and failure coverage. This can help advance the critical point of the U-shaped curve and fully achieve entrepreneurial practice.

Third, the positive regulation of government support suggests that the government should serve as a “visible hand” and establish a long-term support mechanism for farmers’ entrepreneurship, which could accelerate the leftward shift of the threshold of farmers’ entrepreneurship and promote stabilization. Policy support can increase external input for farmers’ entrepreneurship, establish a government–farmer internal/external linkage model, reduce credit constraints for entrepreneurship, enrich financing channels for farmers, provide more favorable market supervision, simplify the procedures for farmers’ entrepreneurship, establish entrepreneurship Q and A services, provide mechanisms to guide more farmers to start their own businesses, and help improve innovation in rural areas.

Finally, farmers’ individual characteristics affect heterogeneity in their roles. It is important, therefore, to focus on both the stigma and value of disaster experiences and maximize the benefits. Individuals with disaster experience will form a specific cognitive imprint, and its persistence will help them discover the upper limit of their abilities and accelerate the improvement of their skills, so that the old imprint can be integrated with the new one through new entrepreneurial situations to offset the negative correlation of disaster experience and optimize entrepreneurial skills and performance. In addition, family heterogeneity characteristics show that the pre-accumulation of entrepreneurship is often a key factor determining the outcome of entrepreneurship. Farmers should undertake good entrepreneurial planning in the preparatory stage; determine the required amount of capital, technology, and ability; and consider the possible risks and challenges.

6.3. Discussion

We find that disaster shock experience significantly affects farmers’ entrepreneurship and that risk appetite and government support change the likelihood of entrepreneurship. This study provides a reference for research related to farmers’ entrepreneurship. We explore its pathways from the perspective of disaster shock experience, expand the factors influencing farmers’ entrepreneurship, and provide suggestions for practitioners who work with farmers in their local communities who have been affected by disasters. It is beneficial to translate professional guidance into practice. In truth, this study does not unravel all the mechanisms that influence farmers’ entrepreneurship, and it would be worth further exploring the factors that influence a shift in the critical point of the curve in the U-shaped relationship. Social capital, household asset allocation, basic public services, and adoption behavior also play roles in coping with disaster shocks and entrepreneurship. In addition, it would be worth investigating whether different entrepreneurial approaches, such as self-fulfillment and survival, as well as the assessment of entrepreneurial performance will show different U-shaped trends. It would also be interesting to investigate how differences in the U-shaped curves of different industries and regions are reflected in practice and what theoretical mechanisms can explain the changes. We discuss this study’s limitations in more detail below.

Regarding sample selection, we only take Chinese farmers’ entrepreneurship as the research object and do not consider other country contexts. We also do not explore entrepreneurship among nonfarmers. Considering that China’s national situation is not exactly the same as that of other countries, and entrepreneurship policies can vary among different countries, the findings are not fully generalizable to other countries. Thus, more comprehensive data covering different country characteristics are needed to obtain more generalizable conclusions.

We use the CHFS database, which contains data on farmers’ entrepreneurship and disaster shock experiences. However, the measurement of disaster shocks and farmers’ entrepreneurship is relatively general. While a more specialized database would help to better understand the relationship between disaster shocks and farmers’ entrepreneurship, it is difficult to expand the database given the costs involved. Future research may be aimed at using more direct data to further validate the conclusions of this study.

Although we take an empirical approach to disaster shock experience, the regression model suffers from small variance. Even though the model’s explanatory power is improved by methods such as adding control variables and interaction terms to the model setting, there still might be unobserved omitted variables. More descriptive statistics and reliability tests could improve the study’s credibility. In addition, the analysis for causal regulation in this study uses only probit and SEM, which cannot fully explain the validity of the causality claim. Future studies can further analyze the relationship between disaster shock experience and farmers’ entrepreneurship through qualitative research, like case studies, which can better response to controversial issues in the literature. Furthermore, using more effective causal identification methods can provide a deeper understanding of the role of disaster shock experience [92,93,94]. In-depth longitudinal analyses of respondents’ disaster shock experiences could be performed after more targeted research to trace the formation, development, and evolution trends of disaster stigma to further reveal the effects on entrepreneurship.

We define entrepreneurship in this study as farmers engaging in self-employment or business operations. However, there is no in-depth discussion of how the entrepreneurial spirit is embodied in entrepreneurial farmers. An entrepreneur creates a new organization; this is the essential difference between an entrepreneur and a manager [90]. Entrepreneurship is an activity, while the entrepreneurial spirit is a form of expression, and the entrepreneurial spirit can be expressed in both entrepreneurial and nonentrepreneurial activities, such as household management. Although not all farmers who start businesses have an entrepreneurial spirit, they can develop it through long-term learning in entrepreneurial processes. Further, it is more likely that entrepreneurship will be displayed in entrepreneurial business than in the form of entrepreneurship in family management. Therefore, future research can further explore farmers’ entrepreneurial spirit in entrepreneurial processes. In addition, we have only discussed farmers engaged in non-farm types of entrepreneurship, while farmers engaging in agricultural entrepreneurship is also an important form of entrepreneurship. The studies on this type of entrepreneurship are more relevant to the needs of agricultural transformation.

We propose that disaster shock experience is a broad concept. Rising temperatures caused by climate change could lead to fires, hurricanes and tsunamis could disrupt production activities, and pandemics such as COVID-19 could pose significant challenges to people’s livelihoods. Moreover, it has been verified that crime rates increase in the context of disasters. Any of the above events could significantly affect farmers’ future development. It is important, then, to study the mechanisms of their effects on entrepreneurship.

We examine the U-shaped trend of disaster shock experience from the perspective of individual entrepreneurs without considering the mechanisms and possible outcomes of disaster shock experiences with regard to entrepreneurship continuity. This could be combined with branding theory to more deeply analyze the correlation of “experience” with entrepreneurship. Therefore, the long-term effect of “experience” on entrepreneurship could be analyzed in conjunction with imprinting theory, and the invisible role of experience could be explored in depth. In addition, our hypotheses are based on the perspective of entrepreneurial success without considering entrepreneurial failure. Future research can combine disaster shock experience and entrepreneurial failure, analyze their theoretical patterns and interaction effects, explore the role of individual learning in entrepreneurial failure, and discover the dynamic changes in the imprint of previous experiences in subsequent experiences.