Impact of Natural Hazards on Agricultural Production Decision Making of Peasant Households: On the Basis of the Micro Survey Data of Hunan Province

Abstract

The agricultural production decision making of peasant households can have a direct impact on agricultural development trends and national food security, and the impact of natural hazards on the agricultural production decision making of peasant households merits significant consideration. On the basis of the micro survey data of Hunan Province in 2022, this paper empirically discusses the impact of natural hazards on peasant households’ agricultural production decision making by using a Probit model. The study found that: (1) natural hazards did not significantly affect peasant households’ willingness to engage in agricultural production; (2) peasant households will reduce the impact of natural hazards on agricultural production by planting “drought-resistant crop” or “flood-tolerant crop”; and (3) natural hazards will also reduce the probability of peasant households adopting green production decision making. Subsequently, the Logit model is used to test the robustness and the PSM model is used to correct the possible selectivity bias. The above results are supported. The results of a heterogeneity analysis indicate the following: (1) natural hazards can substantially affect the disaster adaptation decision making of young and middle-aged householders, and yet fail to significantly affect the disaster adaptation decision making of elderly householders; (2) natural hazards can severely influence the disaster adaptation as well as green production decision making of peasant households that have not received agricultural technology training but those who have received such training are not significantly impacted; and (3) natural hazards significantly affect the production decision making of peasant households who have not purchased agricultural insurance, and yet fail to significantly affect the production decision making of peasant households who have purchased agricultural insurance. Our findings can provide the government with an empirical basis for formulating pertinent policies.

1. Introduction

Against the backdrop of the normalization of extreme weather, frequent natural hazards have caused enormous losses to human society. In addition, the report “Human Losses Caused by Disasters in 2000–2019” issued by the United Nations Agency for International Disaster Reduction highlighted that there were 7348 major natural disasters worldwide from 2000 to 2019, with a total affected population of more than 4.2 billion and global economic losses of approximately $2.97 quadrillion US dollars; the economic impact of the disaster is staggering [1]. Due to the strong dependence of agricultural production on the natural environment, natural hazards have the greatest influence on agricultural production [2], with more serious consequences. In addition, the report of the World Food Programme (WFP) in 2018 indicates that the growth rate of crop output per hectare has slowed down significantly in comparison with the growth rate of the population. In accordance with the data issued by the Food and Agriculture Organization of the United Nations (FAO) in 2016, on the condition that the current situation of greenhouse gas emissions and climate change continues, the yield of major cereal crops will have declined to vary degrees by 2100 (20–45% for corn, 5–50% for wheat, and 20–30% for rice) [3].

China is a developing agricultural nation and one of the nations with the most devastating natural hazards. In accordance with the statistics of the United Nations Agency for International Disaster Reduction, China ranked first worldwide from 2000 to 2019 in the number of natural hazards that killed 10 or more people and affected 100 or more people. Moreover, confirming the data of the Ministry of Emergency Management of China, the affected area of crops in China reached 11,739,000 hectares in 2021 alone [4]. Currently, China’s agriculture is still on the basis of small-scale production in the unit of peasant households [5]. This mode of agricultural production has a low capacity to withstand the effects of natural hazards, so natural hazards have a significant impact on the agricultural production decision making of peasant households. As the micro subject of agricultural production decision making, the agricultural production decision making of peasant households can directly influence the trend of agricultural development as well as national food security.

In this context, this paper employs the micro-survey data of Hunan Province, adopts an econometric model to empirically analyze the impact of natural hazards on the agricultural production decision making of peasant households, as well as explores the heterogeneity of this impact. It is anticipated to provide a scientific basis for enhancing the capacity of rural households to prevent and withstand natural hazards, promoting the green and high-quality development of agriculture, and stabilizing food security. It can be said that the study of the impact of natural hazards on the agricultural production decision making of peasant households is not only on the basis of the realistic consideration of the national conditions of a “large nation with small-scaled peasant households”, but also a response to manageable agricultural risks required by the rural revitalization strategy.

2. Literature Review

The research on the economic behavior of peasant households provides a crucial lens for analyzing and comprehending the agricultural production decision making of peasant households. Moreover, the early academic circles mainly understood the economic behavior of peasant households from the perspectives of “rational peasant households” and “existing peasant households” [6]. The viewpoint of the “rational peasant” compares peasant households to capitalist enterprises, comprehends and analyzes the behavior of peasant households on the basis of the rational economic man, and presumes that peasant households pursue profit maximization [7,8,9]. In accordance with the viewpoint of the “existing peasant”, peasant households pursue existing needs rather than maximum interests. Small-scale peasant decision making follows the principle of risk avoidance and safety first, and the goal is to meet the family’s consumption needs [10,11]. In combination with the local situation in China, Huang Zongzhi put forward the analytical framework of the “commodity peasant” on the basis of the “rational peasant and the “existing peasant”. Huang Zongzhi assumes that the behavior of Chinese peasant households includes the dual logic of rationality and irrationality. Existing pressure compels Chinese peasant households to produce the bare minimum. During the same time, Chinese peasant households are continually pursuing plans that enable their families to earn greater profits [12]. The “existing peasant” provides the lowest logic for understanding the agricultural production decision making of peasant households. The irrational perspective includes peasant households’ habits, instincts, experiences, and social and cultural factors in the analysis framework. The rational perspective abstracts the concept of quantity and value from goods and behavior [13], and the logical and computable aspects of the agricultural production decision making of peasant households are studied in depth [14], except that, as agricultural production is an intertwined process of natural reproduction and economic reproduction, this paper not only analyzes the agricultural production willingness and disaster adaptation decision making of peasant households under the impact of natural hazards from the perspective of economic behavior but also focuses on the green production decision making of peasant households, striving to possess a deeper comprehension of how the agricultural production decision making of peasant households are made under the impact of natural hazards.

With the continuous progress of China’s industrialization and urbanization, the rural labor force’s development space in cities has expanded continuously [15]. The opportunity cost of the agricultural labor force has been rising, and the comparative income of agriculture has gradually been decreasing; consequently, the enthusiasm of the rural labor force for agricultural production decreased, as well as the fact that a large number of the rural labor force has moved from agricultural sectors to non-agricultural sectors and from rural areas to urban areas [16,17,18]. After decades of the large-scale reallocation of the factors of production in China’s rural areas, a labor mobility pattern has been formed in which labor with higher human capital is transferred to non-agricultural industries and urban areas [19,20], and low-human-capital labor is returned to agriculture and rural areas [21,22,23]. Consequently, the majority of people who are still engaged in agricultural production in rural areas today are those who have lost access to other employment opportunities or a small portion of the population that can generate other income locally [24] Although the starting points and purposes of engaging in agricultural production are distinct, both of them have a high degree of dependence on agricultural production. On the basis of the above analysis, the following assumptions are proposed:

Hypothesis 1 (H1).

Natural hazards have no significant impact on peasant households’ willingness to produce grain crops.

Globally, the occurrence, fatality, number of victims, and total estimated costs of natural hazards on economies are rising. Critical concerns have been raised as the rate of warming exacerbates the catastrophic weather events’ frequency, severity, and impacts [25]. As adaptation is frequently conceptualized as a location-specific phenomenon, some scholars have advocated for more meso- and even micro-level analyses in order to better comprehend the fundamental process of adaptation and encourage the formulation of more targeted and operational adaptation policies [26]. As the predominant disaster bearers of natural hazards and the main participants in coping with natural hazards, the selection of disaster adaptation strategies by peasant households is crucial for the risk reduction of natural hazards throughout the region [27,28], which is an important dimension that the government needs to consider. The planting strategy is one of the most significant disaster adaptation behavior strategies for peasant households, which is influenced by internal or external factors during agricultural production, planting distinct crops to achieve the goal of reducing loss and stabilizing income [29]. The selection of planting disaster-resilient varieties of crops is a disaster adaptation mode with a relatively low technical threshold and cost in planting strategies, which has comparatively universal explanatory power. For instance, in Ethiopia and South Africa, switching to drought-resistant crops is a frequently used drought-resistance strategy for local peasant households [30]. Numerous factors, including family economic conditions, the local natural environment, disaster risk awareness, as well as the acquired technology and information, influence the disaster adaptation decision making of peasant households [31,32,33], but when their agricultural production and family life are affected by natural hazards, peasant households are compelled to actively face and adopt corresponding strategies to reduce the negative effects of natural hazards [34]. On the basis of the above analysis, the following assumptions are proposed:

Hypothesis 2 (H2).

Natural hazards have a significant positive impact on peasant households’ disaster adaptation decision making.

China’s agricultural resource endowment is categorized as “a large population with relatively limited land”. “Relatively limited land” includes three aspects: reasonably limited cultivated land per capita, comparatively limited high-quality cultivated land, and relatively limited exploitable reserve resources. Chemical fertilizer input has played a crucial role in enhancing crop yield and ensuring national food security, which is a significant characteristic of modern agriculture [35]. Under the constraints of land resources, peasant households continue the agricultural production model of high output and high income by increasing chemical fertilizer input for a long time, which has caused the agricultural nonpoint source pollution problem to become more prevalent and to have a negative impact on agricultural production and food safety [36,37,38]. From this, it is clear that accelerating the green transformation of agriculture and promoting the high-quality development of agriculture has become the proper meaning of rural environment management and ensuring the quality and safety of agricultural products. In the special situation in which the small-scale peasant economy occupies an important position in China, peasant households contribute significantly to the transformation of the agricultural mode of production and the construction of an ecological environment [39]. Their active practice of the green production mode is the primary driving force for advancing the development of ecological agriculture and achieving agricultural sustainability. Nonetheless, under the impact of natural hazards, peasant households may engage in irrational decision making due to the loss of agricultural output as well as household income [40,41,42], which will impact the growth of sustainable agricultural production. On the basis of the above analysis, the following assumptions are proposed:

Hypothesis 3 (H3).

Natural hazards have a significant negative impact on the green production decision making of peasant households.

Investigating the heterogeneity of the impact of natural hazards on the agricultural production decision making of peasant households is beneficial in deeply analyzing the distinctions in the impact of natural hazards on the divergent agricultural production decision making of peasant households, to put forward more targeted countermeasures. In the first place, householders perform an outstanding role in the decision making process of agricultural production. Moreover, the differences between young and middle-aged householders in terms of production experience, risk perception, technological learning capability, and other factors will result in young and middle-aged homeowners [43,44,45] and elderly householders illustrating different preferences in agricultural production decision making. Particularly after natural hazards, these distinctions may become more pronounced. Secondly, agricultural technology training is not only the key force to supporting the development of modern agriculture, but also an essential policy tool for the government to support agriculture [46,47,48,49]. Agricultural technology training to enhance the quality of peasant households through on-site guidance and centralized training can not only promote the adoption of advanced agricultural technology in peasant households, encourage the improvement of the agricultural production mode, and enhance agricultural production efficiency and the rural ecological environment, but also help to strengthen peasant households’ awareness of hazard prevention and mitigation, enhance their skills in hazard prevention and mitigation, and minimize the impact of natural hazards on agricultural production [50,51]. Eventually, as a professional financial tool for agricultural production risk management, agricultural insurance can effectively spread the risks associated with agricultural production and operation [52]. It can enhance the capability of agricultural production to resist the risk of natural hazards through disaster loss compensation mechanisms and effectively reduce the negative impact of natural hazards. Agricultural insurance can also expand technological progress, promote the sustainable growth of the agricultural economy by increasing the agricultural total factor productivity, and thereby effectively increase peasant households’ enthusiasm for agricultural production [53,54,55]. On the basis of the above analysis, the following assumptions are proposed:

Hypothesis 4a (H4a).

In comparison with the elderly householders, the impact natural hazards have on the agricultural production decision making of young and middle-aged householders is greater.

Hypothesis 4b (H4b).

In comparison with households who have received agricultural technology training, the impact natural hazards have on households who have not is greater.

Hypothesis 4c (H4c).

In comparison with households who have purchased agricultural insurance, the impact natural hazards have on households who have not is greater.

3. Research Design

3.1. Data source

This paper’s research data come from a questionnaire survey conducted in Hunan Province (See Figure 1) by the research team. In addition, the selection of rural areas in Hunan Province for investigation is mainly based on the following three considerations: First, Hunan is a province with the frequent occurrence of natural disasters due to the special geographical environment and changeable climatic conditions, and it is also one of the provinces with the most frequent occurrence of natural disasters in China [56], which is an echo of our old saying “disaster-free year is nowhere”. Second, Hunan has been a major agricultural province since ancient times, and as one of the major grain-producing areas in China, it is recognized as “land of fish and rice”. At present, Hunan ranks first and second, respectively, in terms of rice area and yield in China, first and third in rape area and yield, and first steadily in both the area and output of Camellia oleifera. Vegetables, tea, freshwater products, citrus, traditional Chinese medicine, and other products have risen to the top of the country’s production rankings, assuming the crucial responsibility of ensuring national food security and an adequate supply of major agricultural products. Last, Hunan Province is located in the middle reaches of the Yangtze River in China, which is composed of basins, plains, hills, and mountains. In addition, the geomorphological sequence is relatively complete, and the research conclusions are widely representative. Moreover, the research group carried out a pre-survey in December 2021 and a household questionnaire survey province-wide from January to March 2022. In addition, the sampling was divided into three stages: In the first stage, 10 counties in Hunan Province were chosen as the sample areas on the basis of a comprehensive consideration of the agricultural production scale, agricultural production type, as well as topographic and geomorphic characteristics. In the second stage, random sampling was used to select two sample villages from each county, for a total of 20 sample villages. In the third stage, 20~25 peasant households were randomly selected from each sample village, with a total of 480 peasant households. Aiming to maximize the quality of the data, the research group invited experts to carry out training for interviewers before conducting the pre-survey and formal survey. At the same time, to overcome the dialect barrier, the students who were born and brought up in the Hunan province were predominantly recruited as interviewers. In addition, a total of 480 questionnaires were sent out in the form of one-to-one interviews. After eliminating the questionnaires with incorrect information logic and missing variable data, 453 valid questionnaires were acquired, with an effective rate of 94.38%.

3.2. Selection and Treatment of Variables

The variable being explained is the “agricultural production decision making of peasant households”. The production decision making of peasant households includes many aspects. This paper mainly focuses on the agricultural production willingness of peasant households, disaster adaptation decision making, and green production decision making. We utilize the item “whether your family chooses to reduce planting activities” to measure the agricultural production willingness. We measure the variable of “disaster-resistant variety selection” with the item “whether your family chooses to plant ‘drought-resistant crop’ or ‘flood-tolerant crop’” to characterize the disaster adaptation decision making of peasant households. To characterize the green production decision making of peasant households, the variable “fertilizer input” is measured by the item “whether your family increases chemical fertilizer input”. As shown in

Table 1. Variable definition and descriptive statistics.

Variable Name	Variable Definition and Assignment	Mean Value	Standard Deviation
Natural disaster	Whether peasant households have suffered natural hazards such as floods or droughts in recent years (No = 0; Yes = 1)	0.247	0.432
Agricultural production willingness	Whether peasant households reduce their planting activities (No = 0; Yes = 1)	0.391	0.488
Selection of disaster-resistant varieties	Whether peasant households choose to plant disaster-resilient varieties of crops (No = 0; Yes = 1)	0.45	0.498
Fertilizer input	Whether peasant households increase the input of chemical fertilizer (No = 0; Yes = 1)	0.329	0.47
Age	The actual age of householders	52.667	15.24
Gender	Gender of householders (female = 0; male = 1)	0.521	0.5
Marital status	Marital status of householders (unmarried = 0; married = 1)	0.841	0.366
Educational level	Education level of householders (illiteracy = 1; primary school = 2; junior high school = 3; senior high school or technical secondary school = 4; university and above = 5)	3.086	0.938
Family size	The number of peasant household members interviewed	4.587	1.755
Number of working population	The number of working population among the household members interviewed (the number of family members who have reached the age of 16 to the retirement age, and the number of family members with the capability to work)	2.795	1.463
Annual household income	Annual income level of households (below 10,000 yuan = 1; 10,001–30,000 yuan = 2; 30,001–50,000 yuan = 3; above 10,000 yuan = 5)	2.667	1.271
Village Consolidation	Whether the villages where the householders are located have consolidated in the last ten years (No = 0; Yes = 1)	0.737	0.441

.

The core explanatory variable is “natural disaster”. Since droughts and floods have been the most devastating natural hazards in China over the past four decades, accounting for more than half of the total grain loss [57], they are measured using the item “whether your family has suffered natural hazards including floods or droughts in recent years”.

The control variables part includes three perspectives: human characteristics, family characteristics, and village characteristics. Personal characteristic variables include age, gender, marital status, and education level. Moreover, family characteristic variables encompass family size, the number of the working population, and annual household income. The characteristic variable of the village is the Village Consolidation.

3.3. Model Setting

We consider that the assignment of peasant households’ agricultural production decision making is the binary variable: (0, 1). The binary choice model is selected to estimate the impact of natural hazards on the agricultural production decision making of peasant households so as to avoid the heteroscedasticity caused by the linear probe capability model. It should be noted that there are no distinctions between the Probit model and the Logit model, and there are few distinctions between them, except that the residual term of the Probit model accords with the standard normal distribution, whereas the residual term of the Logit model accords with the logical distribution. For the purpose of ensuring the reliability of the estimation results, this paper applies the benchmark regression through the Probit model and the robustness test through the Logit model [58]. In the first place, the benchmark econometric regression model is set as follows:

$$y_i=x_i^{\prime }\beta +\epsilon \left(i=1,\cdots ,n\right)$$

(1)

Among them, “y” is the explained variable, which represents the production decision of peasant households. “x” is the set of explanatory variables, including whether peasant households have been affected by natural hazards in recent years, as well as the individual characteristics of peasant households, family characteristics, village characteristics, etc. Since the value of “y” estimated by this model may be greater than 1 or less than 0, to keep the estimated value within the range of [0, 1], we consider the two-point function of “y”:

$$\left\{\begin{array}{c}P\left(y=1|x\right)=F\left(X,\beta \right)\\ P=\left(y=0|x\right)1-F\left(x,\beta \right)\end{array}\right.$$

(2)

selecting the appropriate form of the F function (the cumulative distribution function of random variables) for the purpose of ensuring that 0 ≤ $\widehat{y}$ ≤ 1. Due to the fact that E (y|x) = 1 × P(y = 1|x) + 0 × P(y = 0|x) = P(y = 1|x), $\widehat{y}$ is the probability of “y = 1”. If F is a cumulative distribution function that obeys standard normal distribution, subsequently the Probit model can be expressed as:

$$P\left(y=1|x\right)=F\left(x,\beta \right)=\varphi \left(x^{\prime }\beta \right)\equiv {\int }_{-\mathrm{\infty }}^{x^{\prime }\beta }\phi \left(t\right)dt$$

(3)

If F is a cumulative distribution function subject to logical distribution, the Logit model can also be indicated as:

$$P\left(y=1|x\right)=F\left(x,\beta \right)=\wedge \left(x^{\prime }\beta \right)\equiv \frac{e^{x^{\prime }\beta }}{1+e^{x^{\prime }\beta }}$$

(4)

The propensity score matching (PSM) method corrects possible selective errors. Initially, in accordance with the observable variables including the individual characteristics, family characteristics, and community characteristics of the sample, the Logit model is used to estimate the propensity score of each sample. Secondly, four distinct types of matching methods, including k-nearest neighbor matching, caliper matching, k-nearest neighbor matching within caliper, and kernel matching, are employed to match the samples. On the condition that the estimation results obtained by the above matching methods are consistent, accordingly, it is determined that the estimation results are robust and dependable. Finally, on the basis of the matched samples, we calculate the average treatment effect on the treated (ATT):

$$ATT=E\left(Y_{1i}\right|D_i=1)-E(Y_{0i}|D_i=1)$$

(5)

Among them, D_i is the core explanatory variable “natural disaster”, and Y_1i and Y_0i refer to the agricultural production decision making of peasant households when the sample suffers from natural hazards and when the sample does not, respectively.

4. Empirical Results

4.1. Benchmark Regression Results

The results of the benchmark regression are demonstrated in

Table 2. Benchmark regression results.

	Agricultural Production Willingness		The Selection of Disaster-Resilient Varieties		Chemical Fertilizer Input
	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
Natural disaster		0.065 (0.049)		0.179 *** (0.051)		0.193 *** (0.046)
Age	−0.012 *** (0.002)	−0.012 *** (0.002)	−0.001 (0.002)	−0.001 (0.002)	−0.001 (0.002)	−0.001 (0.002)
Gender	0.024 (0.071)	0.033 (0.071)	−0.056 (0.073)	−0.029 (0.072)	0.048 (0.072)	0.077 (0.070)
Marital status	0.064 (0.084)	0.059 (0.084)	0.037 (0.085)	0.020 (0.085)	0.126 (0.089)	0.102 (0.086)
Educational level	−0.018 (0.030)	−0.019 (0.030)	0.024 (0.032)	0.024 (0.032)	−0.018 (0.029)	−0.021 (0.029)
Family size	−0.033 * (0.017)	−0.032 * (0.017)	0.003 (0.018)	0.005 (0.017)	0.025 (0.016)	0.027 * (0.016)
Working population	0.045 ** (0.019)	0.044 ** (0.019)	0.027 (0.021)	0.025 (0.020)	0.009 (0.019)	0.008 (0.018)
Annual household income	0.002 (0.019)	0.003 (0.019)	0.004 (0.020)	0.008 (0.019)	−0.050 * (0.018)	−0.047 *** (0.018)
Economic assistance	0.034 (0.047)	0.036 (0.047)	−0.041 (0.051)	−0.036 (0.050)	0.017 (0.047)	0.020 (0.046)
Agricultural technology training	−0.006 (0.050)	−0.005 (0.050)	0.184 *** (0.050)	0.188 *** (0.050)	0.163 ** (0.047)	0.168 *** (0.046)

Notes: Standard errors in parentheses; *, **, and *** indicate the significance of 0.1, 0.05, and 0.01, respectively. Regression coefficients are marginal effects.

. Model 1, Model 3, and Model 5 report the effects of the control variables on agricultural production willingness, disaster-resilient variety selection, as well as chemical fertilizer input, respectively. In Model 1, age and family size have a substantial negative effect on agricultural production motivation, while the working population impacts agricultural production willingness in a significant manner. In Model 3, agricultural technology training substantially influences the selection of disaster-resilient plant varieties. Model 2, Model 4, and Model 6 report the impact coefficients of natural hazards on the agricultural production willingness of peasant households, disaster-resilient variety selection, and chemical fertilizer input, respectively.

The results illustrate that natural hazards did not remarkably affect the willingness of peasant households to produce (p > 0.1). In addition, the impact coefficient of natural hazards on the selection of disaster-resilient varieties is 0.179 (p < 0.001), indicating that peasant households would adopt adaptation strategies to reduce the negative effects of natural hazards on agricultural production. Furthermore, the impact coefficient of natural hazards on chemical fertilizer input is 0.193 (p < 0.001), demonstrating that natural hazards will discourage peasant households’ decision making regarding green production.

4.2. Robustness Test

The impact of natural hazards on the various production decision making of peasant households has been empirically analyzed with the Probit model. To ensure the stability and reliability of the results, this paper uses the Logit model to test the robustness. As demonstrated in

Table 3. Robustness test results.

	Agricultural Production Willingness	The Selection of Disaster-Resilient Varieties	Chemical Fertilizer Input
	Logit	Logit	Logit
Natural disaster	0.316 (0.232)	0.772 *** (0.233)	0.943 *** (0.239)
Control variable	Yes	Yes	Yes
Constant term	2.398 ** (0.934)	−1.077 (0.909)	−1.538 (0.938)

Notes: Standard errors in parentheses; ** and *** indicate the significance of 0.05 and 0.01, respectively. Regression coefficients are marginal effects.

, the impact coefficient of natural hazards on agricultural production willingness is 0.316, which fails the test of significance. Moreover, the impact coefficients of natural hazards on the selection of disaster-resilient varieties and chemical fertilizer input are 0.772 and 0.943, correspondingly, which are significant at the level of 1%. Thus, it can be seen that the sign direction and significance of the estimated coefficients of the Probit model and Logit model are the same, which can determine whether the model estimation results of the model are robust.

4.3. Correcting Selective Errors

This paper adopts the PSM model to establish a counterfactual framework of the impact of natural hazards on the agricultural production decision making of peasant households to correct possible selective errors. To ensure the high credibility of the propensity-score-matching results, it is essential to conduct a balance test and common support hypothesis test on the model. The objective of the common support test is to ensure that the majority of the treatment group samples can be matched with the control group samples [59]. In addition, the purpose of the balance test is to test whether there is a substantial distinction in the sample characteristics between the treatment group and the control group after propensity score matching. It is observed the propensity score distribution of the matched treatment group and the control group is further close (as demonstrated in Figure 2 and Figure 3). In addition, on the condition that using four methods of k-nearest neighbor matching, caliper matching, k-nearest neighbor matching, and kernel matching, 11 of the 453 samples do not fall within the common value range, which can be judged to pass the common support test. The t-test indicates that the absolute values of the standardization deviations of each matching variable in the treatment group and the control group are all less than 20%, which meets the standard [60] proposed by Rosenbaum and Rubin and is deemed to have passed the balance test.

Table 4. Estimation results of average treatment effect.

Variables	Agricultural Production Willingness	The Selection of Disaster-Resilient Varieties	Chemical Fertilizer Input
k-nearest neighbor matching (k = 4)	0.076 (0.060)	0.168 *** (0.060)	0.207 *** (0.058)
Caliper matching (cal = 0.01)	0.069 (0.057)	0.197 *** (0.057)	0.215 *** (0.056)
k-nearest neighbor matching within caliper (cal = 0.01, k = 4)	0.087 (0.060)	0.181 *** (0.060)	0.207 *** (0.059)
Kernel matching (default kernel function and bandwidth)	0.058 (0.053)	0.191 *** (0.055)	0.208 *** (0.054)

Notes: Robust standard errors in parentheses; *** indicates the significance of 0.01.

reveals the ATT values of natural hazards on the agricultural production decision making of peasant households estimated by the four matching methods: k-nearest neighbor matching, caliper matching, k-nearest neighbor matching within the caliper, and kernel matching. The ATT values of natural hazards on agricultural production willingness estimated by the above four matching methods are 0.076, 0.069, 0.087, and 0.058, correspondingly, which do not pass the significance test. The ATT values of natural hazards on disaster-resilient variety selection are 0.168, 0.197, 0.181, and 0.191, respectively, which are substantial at the level of 1%. The ATT values of natural hazards on fertilizer input are 0.207, 0.215, 0.207, and 0.208, respectively, which are considered at the level of 1%. After correcting for possible selective errors and eliminating the observable systematic differences between the treatment group and the control group, it can be seen that the impact of natural hazards on agricultural production willingness remains insignificant, whereas the impact of disaster-resilient variety selection and fertilizer input remains significant, which further verifies the reliability of the benchmark regression results.

4.4. Heterogeneity Analysis

4.4.1. Heterogeneity Analysis of the Age of Householders

The householder plays a significant role in the decision making process of agricultural production of peasant households. In addition, the empirical results indicate that the impact of natural hazards on the agricultural production willingness of middle-aged and young householders as well as elderly householders is not substantial, and the coefficients are 0.076 (p > 0.1) and 0.073 (p > 0.1), correspondingly. Natural hazards can substantially affect the selection of disaster-resilient varieties of young householders, and yet have no considerable effect on the selection of disaster-resilient varieties of elderly householders, with coefficients of 0.102 (p < 0.001) and 0.223 (p > 0.1), respectively. Furthermore, the impact of natural hazards on the fertilizer input of middle-aged and young householders was greater than that of elderly householders, with coefficients of 0.247 (p < 0.001) and 0.149 (p < 0.001), correspondingly.

4.4.2. Heterogeneity Analysis of Whether Emergency Assistance Is Obtained

The impact of natural hazards on the agricultural production willingness of peasant households who have received emergency assistance and those who have is not substantial, and the coefficients are 0.065 (p > 0.1) and 0.075 (p > 0.1), respectively. The impact of natural hazards on the disaster adaptation decision making of peasant households who have not received emergency assistance is marginally greater than that of peasant households who have received emergency assistance, as measured by coefficients of 0.179 (p < 0.001) and 0.157 (0.05 < p < 0.1), respectively. Moreover, the impact of natural hazards on the fertilizer input of peasant households without emergency assistance is greater than that of peasant households with emergency assistance, with coefficients of 0.594 (p < 0.001) and 0.156 (0.05 < p < 0.1), correspondingly.

4.4.3. Heterogeneity Analysis of Whether Emergency Assistance Is Obtained

The coefficients of 0.065 (p > 0.1) and 0.075 (p > 0.1) for the effect of natural hazards on the agricultural production willingness of peasant households who have received emergency assistance and those who have are not statistically significant. In addition, the impact of natural hazards on the disaster adaptation decision making of peasant households who have not obtained emergency assistance is slightly greater than that of peasant households who have received emergency assistance, with coefficients of 0.179 (p < 0.001) and 0.157 (0.05 < p < 0.1), appropriately. Moreover, the impact of natural hazards on the fertilizer input of peasant households without emergency assistance is greater than that of peasant households with emergency assistance, with coefficients of 0.594 (p < 0.001) and 0.156 (0.05 < p < 0.1), appropriately.

4.4.4. Heterogeneity Analysis of Whether Agricultural Insurance Is Purchased

As shown in

Table 5. Results of heterogeneity analysis.

	Agricultural Production Willingness	The Selection of Disaster-Resilient Varieties	Chemical Fertilizer Input
	Probit	Probit	Probit
Sample of young and middle-aged householders	0.076 (0.084)	0.102 *** (0.083)	0.247 *** (0.064)
Sample of elderly householders	0.073 (0.060)	0.223 (0.065)	0.149 *** (0.065)
Samples without agrotechnical training received	0.087 (0.056)	0.218 *** (0.057)	0.186 *** (0.051)
Samples with agrotechnical training received	0.105 (0.095)	0.090 (0.110)	0.170 (0.103)
Samples without agricultural insurance purchased	0.104 * (0.058)	0.196 *** (0.063)	0.175 *** (0.051)
Samples with agricultural insurance purchased	0.032 (0.097)	0.006 (0.089)	−0.109 (0.084)

Notes: Robust standard errors in parentheses; * and *** indicate the significance of 0.1 and 0.01, respectively. Regression coefficients are marginal effects.

. The impact coefficients of natural hazards on the production willingness, disaster-resilient variety selection, and fertilizer input of peasant households without agricultural insurance are 0.104 (0.05 < p < 0.1), 0.196 (p < 0.001), and 0.175 (p < 0.001), whereas the impact coefficients of natural hazards on the production willingness, disaster-resistant variety selection, and fertilizer input of peasant households with agricultural insurance are 0.032 (p > 0.1), 0.006 (p > 0.1), and −0.109 (p > 0.1), appropriately. This indicates that natural hazards can substantially affect the agricultural production decision making of peasant households who have not purchased agricultural insurance and yet fail to dramatically impact the production decision making of peasant households who have purchased agricultural insurance.

5. Discussion

Against the backdrop of the normalization of extreme weather, frequent natural hazards have had numerous negative effects on agricultural production. This paper empirically analyzes the impact of natural hazards on the agricultural production decision making of peasant households as well as its heterogeneity. Moreover, the predominant effects include the impact of natural hazards on peasant households’ agricultural production willingness, disaster-resilient variety selection, and chemical fertilizer input, in which disaster-resilient variety selection reflects peasant households’ disaster adaptation decision-making, and chemical fertilizer input represents the green production decision making of peasant households. The research revealed the following: (1) Natural hazards did not notably affect the willingness of peasant households to produce. On the one hand, the majority of rural residents who can find subsistence jobs in the cities have moved to the cities, and those who still stay in the rural areas are resistant to finding more suitable livelihood strategies in the cities. For them, agricultural production is not only an expression of a natural affinity for land and nature but also a desperate measure. On the other hand, for most peasant households, the role of agricultural production has shifted from subsistence supply to income augmentation. In addition, the impact of natural hazards will reduce household income, at most, and yet with little possibility of falling into a state of unsustainable life; accordingly, there is no need to change the willingness of agricultural production. (2) Peasant households will adopt adaptation strategies with the hope of reducing the impact of natural hazards on agricultural production. Risk-averse peasant households will take disaster adaptation measures to reduce agricultural production losses in the face of natural hazards. In accordance with the theory of the rational peasant, the disaster adaptation decision making of peasant households is made on the basis of the comprehensive consideration of numerous factors, including cost–benefit, operability, production, living habits, etc. From this perspective, planting disaster-resilient crop varieties is a cost-effective strategy for disaster adaptation that does not necessitate additional investments in infrastructure construction, agricultural production equipment, or other aspects, nor does it necessitate difficult planting techniques. (3) Peasant households will remarkably increase the input of chemical fertilizer after suffering from natural hazards. This may be due to the desire of peasant households to earn more income to make up for the loss of agricultural production, and yet this behavior will inevitably cause environmental pollution. Several studies have demonstrated that China’s agricultural fertilizer input has increased from 8 million tons in 1978 to nearly 60 million tons in 2019, and the fertilizer consumption per unit of sown area is up to 434.25 kg N/hm², far exceeding the internationally recognized upper limit of 225 kg N/hm² of the fertilizer safety level [61]. The excessive input of chemical fertilizer will not only increase the concentration of heavy metals in soil but may also lead to water pollution and eutrophication in the vicinity of farmland, resulting in severe negative externalities for the ecological environment [62]. From this point of view, guiding peasant households to engage in acceptable green production decision making is even related to the green transformation and high-quality development of China’s agriculture.

In the heterogeneity analysis, the samples were divided in accordance with the age of the householder, and whether agricultural technology training and agricultural insurance are acquired. In the first place, the research results of the heterogeneity of the age of the householders on the impact of natural hazards on the agricultural production decision-making of peasant households indicate the following: (1) The impact of natural hazards on the agricultural production willingness of young and middle-aged peasant households as well as elderly householders is not substantial. For young and middle-aged householders who nevertheless stay in rural areas, agricultural production is not the predominant means of livelihood [63]; natural hazards do not alter their means of subsistence, and thus fail to notably affect the willingness to engage in agricultural production. Due to their age, physical strength, health, and skills, elderly homeowners have lost more opportunities for growth. It is challenging to replace the role of agricultural production in their livelihood strategies in other ways even if they are impacted by natural hazards. Additionally, elderly householders have a relatively deep degree of attachment to the land, and engaging in agricultural production has become an integral part of their lives. (2) Natural hazards can substantially affect the selection of disaster-resilient varieties for young and middle-aged householders, and yet not for elderly householders. In comparison with the elderly, young and middle-aged people have closer ties with the outside world, enhanced access to data and resources, and relatively higher efficiency. Hence, after suffering from natural hazards, young and middle-aged householders can quickly engage in disaster adaptation decision making by utilizing the knowledge and resources they possess, thereby reducing and improving their capability to withstand natural disaster risks. On the contrary, elderly householders have experienced many natural hazards since they carried out agricultural production activities in their early years. As a result of acclimating to the natural environment, the risk perception level is low. The impact of natural hazards on the chemical fertilizer input of young and middle-aged householders is higher than that of elderly householders. In addition, the agricultural production decision-making experience of elderly householders is richer than that of young and middle-aged householders [64]. At the same time, the economic capacity of elderly householders is weaker than that of young and middle-aged householders. In the aftermath of natural hazards, the chemical fertilizer input of elderly homeowners is lower than that of younger and middle-aged homeowners due to their decision-making experience and economic capacity. Secondly, the research results of the heterogeneity on whether agricultural technology training is received in terms of natural hazards’ effects on the agricultural production decision making of peasant households indicate that natural hazards have no significant effect on the willingness of agricultural production of peasant households who have received agricultural technology training and those who have not. The predominant reason is that the content of agricultural technology training primarily involves new varieties, new technologies, meteorological knowledge, and pest control and fertilization technology in agriculture. Consequently, it has little effect on agricultural production determination. In addition, the impact of natural hazards on the selection of disaster-resilient varieties as well as fertilizer input of peasant households who have not received training in agricultural technology is greater than that of peasant households who have received agricultural technology training. This is due to the fact that peasant households who have received agricultural technology training relatively master more agricultural production technologies, which can mitigate the negative effects of natural hazards on agricultural production to a greater degree, to avoid “stress” disaster adaptation decision making. In addition, agricultural technology training can raise awareness of the green production of peasant households and then affect their green production decision making. Some studies have pointed out that agricultural technology training can effectively guide peasant households to apply fertilizer reasonably and improve their yield per unit area, which promotes the environmentally friendly agricultural production process and improves the efficiency of rural ecological environment governance [65,66,67,68,69]. Thirdly, the research results of the heterogeneity on whether agricultural insurance is purchased regarding the effects of natural hazards on the agricultural production decision making of peasant households indicate that natural hazards can substantially affect the production decision making of peasant households who have not purchased agricultural insurance and yet fail to markedly affect the production decision making of peasant households who have purchased agricultural insurance. As a special economic compensation system, the agricultural insurance system plays a crucial role in reducing agricultural production risks and ensuring agricultural production and stabilizing the income of peasant households [70,71,72,73,74,75]. It can reduce the losses caused by natural hazards to peasant households through preventive and disaster-alleviative measures. Thus, natural hazards have failed to substantially affect the agricultural production decision making of peasant households who have purchased agricultural insurance. On the other hand, when natural hazards strike, peasant households who have not purchased agricultural insurance will suffer greater losses, which will lead to a decline in the willingness to engage in agricultural production. On the condition that the opportunity cost of their labor is somewhat zero and they lack good employment opportunities, and they have to continue to engage in agricultural production [12], by selecting a drought-resistant crop or flood-tolerant crop and increasing fertilizer input, they will avoid agricultural production risks and boost agricultural output.

6. Conclusions and Policy Implications

6.1. Conclusions

China is one of the countries most affected by natural disasters worldwide due to its vast territory, complex geographical environment, large fluctuations in climate, poor ecological stability, and high frequency and intensity of disasters [57,76,77]. To this end, China has issued a series of hazard prevention and mitigation policies. For example, in 2016, the CPC Central Committee and the State Council issued “opinions on promoting the Reform of the System and Mechanism of Hazard Prevention, Mitigation, and Relief”. In 2022, the National Disaster Reduction Committee issued the “14th Five-Year Plan for Comprehensive Hazard Prevention and Reduction”. However, there are still shortcomings and deficiencies in China’s hazard prevention, mitigation, and relief system [78], and the situation of “rural unprotected” has not yet been fundamentally changed. The relationship between agricultural production and the natural environment is quite close, which makes agricultural production extremely vulnerable to the impact of natural hazards [79], and subsequently affects the production decision making of peasant households. In the context of China’s food security strategy of basic grain self-sufficiency and absolute ration safety, the production side of the grain supply plays a crucial role in ensuring China’s food security [80]. Accordingly, it is of considerable practical relevance to explore the impact of natural hazards on the agricultural production decision making of peasant households. On the basis of the micro survey data of Hunan Province, this paper empirically discusses the effects of natural hazards on the agricultural production decision making of peasant households and its heterogeneity by employing Probit, Logit, and PSM methods. The results indicate the following: (1) Natural hazards do not substantially affect the willingness of peasant households to engage in agricultural production. To reduce the effects of natural hazards on agricultural production, peasant households will adopt disaster adaptation decision making. Natural hazards are not conducive to peasant households adopting green production decision making. (2) The impact of natural hazards on the agricultural production decision making of peasant households is heterogeneous. Moreover, natural hazards can substantially affect the disaster adaptation decision making of young and middle-aged householders, but fail to substantially affect the disaster adaptation decision making of elderly householders. Natural hazards can significantly impact the disaster adaptation as well as green production decision making of peasant households that have not received agricultural technology training but those who have received such training are not significantly impacted. Natural hazards significantly affect the production decision making of peasant households who have not purchased agricultural insurance and yet fail to significantly affect the production decision making of peasant households who have purchased agricultural insurance. In addition, the research results can provide the government with an empirical basis for formulating pertinent policies.

6.2. Policy Implications

There is no doubt that the government’s proactive agricultural policies can support agricultural production affected by hazards and further protect food security [81,82]. Consequently, the government can reduce even further the negative impact of natural hazards on the agricultural production decision making of peasant households in the following aspects.

First, we should give full play to the guiding role of agricultural subsidies and optimize peasant households’ agricultural production decision making. We should give further play to the guiding function of agricultural subsidies, reasonably set the number of subsidies for resource-saving and environment-friendly agricultural production modes, strengthen the guidance of agricultural green production behavior, and consolidate the achievements of the “three-in-one” reform of agricultural subsidies, to realize the green transformation and high-quality development of agriculture. For example, we tend to increase subsidies for the techniques promotion of improved varieties and methods, and appropriately reduce subsidies for agricultural consumption materials such as chemical fertilizers, pesticides, agricultural films, and so on. At the same time, we should also pay attention to the reasonable compensation for the economic losses caused by the green production of peasant households and improve the enthusiasm of peasant households to implement agricultural green production.

Second, we should strengthen the popularization of agricultural technology and strengthen the education and guidance for peasant households. On the one hand, we should build a co-operative research and development mechanism for agricultural disaster-resilient technologies with enterprises as the main body, the market as the guidance, and the in-depth integration of production, education, and research; strengthen the breeding and promotion of disaster-resilient varieties of crops; optimize the farming systems and planting methods; speed up the development and application process of new technologies and materials for agricultural hazard prevention and mitigation [83], strive to solve the technical problems in all aspects of agricultural production; and constantly improve the ability of agricultural hazard prevention and reduction. On the other hand, expert lectures, agricultural technology training, “village to village” radio, television media, and other means of publicity enhance peasant households’ awareness of green production and thus improve the possibility of peasant households adopting the green mode of production. At the same time, we should also pay attention to the dissemination of knowledge about climate change and natural hazards to improve the ability of peasant households to reduce hazards and increase income.

Third, we should optimize the compensation mechanism for natural hazards and improve the guarantee capacity of agricultural insurance [84,85]; build a tightly and firmly woven multi-level compensation system for natural hazards; identify the positioning of various mechanisms such as financial assistance [86], agricultural insurance [87], social donation [88], and family mutual assistance; and give full play to the compensation synergy of various mechanisms. Among them, special attention should be paid to giving full play to the role of agricultural insurance. Agricultural insurance can ensure that peasant households will not engage in irritable production decision making under the impact of natural hazards and plays an important role in stabilizing agricultural production [75]. Therefore, on the one hand, we should continue to improve the coverage of agricultural insurance and the participation rate of small and scattered peasant households and strengthen the ability of more peasant households to resist natural disaster risks while constantly enriching the agricultural insurance fund pool. On the other hand, it is necessary to broaden the service field and business scope of agricultural insurance, to realize the goal of raising agricultural insurance with general insurance and ensure the sustainable development of agricultural insurance.

As with other studies, there are also some limitations to this study’s research. Limited by the research data, only the cross-sectional data were employed to analyze the impact of natural hazards on the agricultural production decision making of peasant households at a certain time, rather than investigating the long-term impact of natural hazards on the agricultural production decision making of peasant households. In addition, the utilization of panel data will enable us to find the time effect of natural hazards on the agricultural production decision making of peasant households. For the purpose of accomplishing this, the research team will conduct a future follow-up survey on the samples. Meanwhile, propensity score matching (PSM) cannot solve the endogenous problems caused by unobserved variables, such as peasant households’ agricultural ability.