Research on Rural Households’ Poverty Vulnerability in Poor Mountainous Areas: An Empirical Analysis in the Upper Reaches of the Min River, China

Abstract

Quantitatively measuring the poverty vulnerability level of rural households in mountainous areas and the factors influencing the poverty vulnerability level is significant for China to consolidate the existing poverty alleviation achievements and prevent farmers in mountainous areas from falling back into poverty in the future. This study selects the upper reaches of the Min River, a typical poor mountainous area, as the research area and decomposes poverty vulnerability into two dimensions: consumption level and consumption fluctuation. Based on microsurvey data, this paper uses an econometric model to measure the poverty vulnerability of farmers who lived in the upper reaches of the Min River in 2023 and analyses its influencing factors. The results show that Edu, Laborer, and Land significantly impact rural households’ per capita consumption level in the future. Agri-shock, in terms of homogeneous risk, and Hea-shock and Hou-shock, in terms of heterogeneous risk, significantly impact rural households’ future consumption variance. Deposit weakens the poverty vulnerability caused by homogeneous risk (Agri-shock). Physicap and M can reduce the poverty vulnerability caused by homogeneous risks and heterogeneous risks. The research results have important practical significance for the establishment of risk and vulnerability early warning mechanisms in the new poverty alleviation strategy in the upper reaches of the Min River and other contiguous poverty areas and have important strategic guiding significance for the sustainable development of rural areas.

1. Introduction

The phenomenon of poverty has always been present throughout the process of human civilization, and its solution path has become a common goal of the international community [1,2]. In recent years, various poverty alleviation policies implemented by the Chinese government addressing rural poverty have significantly reduced the number of poor people. By the end of 2020, China had eradicated regional overall poverty [3,4,5]. However, poverty in China’s rural areas is still severe, and the poverty alleviation task remains arduous. This is reflected in the fact that the causes of poverty are diverse and some poor people return to poverty after experiencing poverty alleviation due to the impact of diseases, disasters, and other risks [6,7,8]. To this end, the Chinese government has formulated the “Outline of Poverty Alleviation and Development in China’s Rural Areas (2011–2020)” and formulated the policy of “targeted poverty alleviation”. The prerequisite of targeted poverty alleviation is the accurate identification, namely, the identification of poor villages and poor households. To prevent the return to poverty to the greatest extent, accurate identification entails not only recognizing who is and who is not currently poor but also knowing which farmers, although not currently poor, may become poor in the future because of some risk impact. The concept of poverty vulnerability supports the accurate identification of poor farmers [9,10,11].

Early scholars regarded poverty as a static concept and judged whether a family was poor based on its current livelihood conditions [12,13,14]. However, families who are not currently poor may become poor in the future because of various risk impacts (such as crop yield reduction, major disease, or natural disaster) [15,16,17]. Based on this, the World Bank put forward the concept of “poverty vulnerability” in 2000 to describe the association between a household’s ability to cope with risks and future poverty. The agency posits that poverty vulnerability is the possibility of future welfare decline caused by various risk shocks to families or individuals [18,19,20]. Pritchett et al. defined poverty vulnerability as the probability that a household will fall into poverty for at least one year within the next few years [21,22,23]. Alwang et al. stated that poverty vulnerability is the fluctuation of a household’s income or consumption caused by risk shocks in the future [24,25,26]. Kuhl defined it as the possibility that a family’s consumption level will fall below the poverty line due to risk impact [7,27,28]. The World Bank economist Chaudhuri and colleagues divided the causes of poverty vulnerability into structural causes (forming the “low mean” of consumption) and risk causes (forming the “high fluctuation” of consumption) and then decomposed it into two dimensions: consumption level and consumption fluctuation [29,30,31]. They predicted the poverty vulnerability of households by estimating the future consumption level and consumption fluctuation [32,33,34]. On this basis, scholars have gradually carried out the measurement of poverty vulnerability. For example, Christiaensen et al. established a two-period model of household consumption decision-making and theoretically studied the key factors that influence the mean and variance of household consumption in the future [35]. Glewwe and Dercon et al. used the change rate of household consumption to analyze the poverty vulnerability of farmers in Peru and Ethiopia [36,37]. Amin et al. used panel data to analyze the association between the consumption and income of farmers in Bangladesh and then measured their poverty vulnerability [38]. Tai et al. analyzed the influence of farmers’ migrant work on poverty vulnerability in the western mountainous areas of China [39]. Wan et al. studied the poverty vulnerability in Shanxi and Shandong Provinces of China using panel data [40].

Scholars have also analyzed the factors influencing poverty vulnerability from the perspectives of external risk shocks faced by rural households (homogeneous risk and heterogeneous risk) and the internal farmers’ ability to deal with these risks. For example, Damien used survey data of Haitian farmers to analyze the influence of external risk shocks on farmers’ poverty vulnerability and found that, compared with homogeneous risks, heterogeneous risks, especially health-related heterogeneous risks, had a greater influence on the farmers’ poverty vulnerability [41]. In terms of rural households’ internal processing capacity, researchers have mainly analyzed the impact of rural households’ livelihood capital stock and income diversification on their poverty vulnerability. For example, Jalan et al. found that the nonproductive capital of the rich can cause the rich to decide to invest in risky activities and obtain more profits, while the poor set too much nonproductive capital aside to guard against risks and fall into poverty due to the lack of productive capital, resulting in lower income [42]. Udry found that poor households could maintain normal production and ensure smooth consumption through food savings (such as livestock and grain) and cash savings [43]. Tai et al. analyzed the impact of an increase in household income on poverty vulnerability by using microscopic data in the western mountainous areas of China [44]. Zhao et al. analyzed the cultivated land transfer’s influence on rural households’ poverty vulnerability and its regional differences [45]. However, there are few studies on the poverty vulnerability of farmers in concentrated contiguous poverty areas in China.

The upper reaches of the Min River are part of the Tibetan region of four provinces in China’s concentrated contiguous poverty areas. In recent years, various poverty alleviation projects have been vigorously promoted, and the lives of rural households have been greatly improved in the upper reaches of the Min River, China. By the end of 2020, with the realization of China’s goal of eliminating absolute poverty, all poor farmers in the upper reaches of the Min River had been lifted out of poverty [46]. However, risk shocks, such as frequent natural disasters (especially the collapse, landslide, debris flow, and other geological disasters caused by the 5.12 Wenchuan earthquake) and diseases, in this region have caused some people to return to poverty or caused poor people to become more poor [47]. In this context, what is the degree of farmers’ poverty vulnerability in the upper reaches of the Min River? What are the factors that affect their consumption level and consumption fluctuation? Based on this, this study selects the upper reaches of the Min River in the poor mountainous areas as the research area, adopts the perspective of vulnerability, takes rural households as the smallest research unit, and uses an econometric model to measure the farmers’ poverty vulnerability and analyze its key influencing factors and influencing mechanism based on the microsurvey data of farmers. The research results have important practical significance for the establishment of risk and vulnerability early warning mechanisms in the new poverty alleviation strategy in the upper reaches of the Min River and other contiguous poverty areas and have important strategic guiding significance for the sustainable development of rural areas.

2. Study Area

The upper reaches of the Min River refer to the area above Dujiangyan in the Min River Basin. It is undulating, with the lowest altitude of 762 m in the southeast and the highest altitude of 5870 m in the northwest. Its scope is basically consistent with the administrative areas of Heishui County, Songpan County, Li County, Wenchuan County, and MaoCounty in Aba Tibetan and Qiang Autonomous Prefecture of Sichuan Province in China. The total area is approximately 22,000 km² (Figure 1). Socially and geographically, people of Tibetan, Qiang, Han, and Hui ethnicities coexist in the region, and it is part of the Tibetan region of four provinces in China’s concentrated contiguous poverty areas. The upper reaches of the Min River are high altitude topographically and under-developed economically. The per capita consumption level was only RMB 12,199/year, 0.23 percentage points lower than the per capita consumption level of Chinese rural residents (RMB 15,916/year) in the same year.

3. Research Data and Research Methods

3.1. The Sources of Research Data

First, we selected Wenchuan and Heishui Counties as the sample counties according to the per capita disposable income in 2023. Then, sample towns and sample villages that have different economic levels (high, medium, and low economic levels) were selected from the sample counties, and 30 farmers were randomly selected as sample farmers in every sample village. An on-site questionnaire survey was distributed to some of the sample farmers as a presurvey. The main contents of the presurvey included the livelihood capital, income sources, main expenses, and possible risk impact of the farmers. The questionnaire was modified and improved through the presurvey, and the formal survey was finally executed in July 2024. A total of 406 questionnaires were distributed in this survey, and 395 valid questionnaires were collected, with an effective proportion of 97.29%. The basic sample region information is shown in

Table 1. Basic sample region information.

Sample Counties	Sample Towns	Sample Villages	Number of Valid Samples
Wenchuan (H)	Shuimo (H)	Laoren (H), Xianfengyan (M), Chenjiashan (L)	80
	Miansi (M)	Sanguanmiao (H), Lianghe (M), Gaodian (L)	90
	Yanmen (L)	Koushan (H), Qingpo (M), Tongshan (L)	86
Heishui (L)	Seergu (H)	Seergu	38
	Shashiduo (M)	Yangrong (H), Jiazu (L)	52
	Zhawo (L)	Kebie (H), Ruoduo (L)	49

Note: H, M, and L represent sample counties, sample towns, or sample villages with high, medium, and low economic levels, respectively.

.

3.2. Theoretical Research Framework

After relevant scholars proposed that poverty vulnerability is a forward-looking concept, theoretical research frameworks related to poverty vulnerability emerged accordingly. The two most representative ones are the following: one is the theoretical research framework emphasizing “external risks”, and the other is the theoretical research framework emphasizing “internal processing capacity”. Meanwhile, scholars have also noticed that the “external risks” encountered by farmers in different countries or regions and their “internal handling capabilities” for risks are not the same. Therefore, these two theoretical research frameworks are not fixed but open. Based on this, this paper combines these two theoretical research frameworks, namely, the “external risks” encountered by farmers and the “internal handling capacity” of farmers for risks, in order to reduce the poverty vulnerability of farmers to the greatest extent.

3.3. Model and Variables

According to the two-period model of rural household consumption decision-making established by Christiaensen et al., the main factor of the farmers’ consumption level is persistent income, and the main factor of the farmers’ consumption fluctuation is income change [35]. Based on this, this paper first establishes an econometric model of farmers’ future consumption level and consumption fluctuation, as described in Formulas (1) and (2).

According to Rajadel, assuming that the future consumption of farmers obeys a normal distribution [48], there are

$$\ln (C_i)={\alpha }_0+{\alpha }_1{I}_i^p+{\epsilon }_i$$

(1)

where $C_i$ is the per capita consumption level of the $i$-th farmer in the future, $I_i^p$ is the persistent income of the $i$-th farmer, ${\alpha }_0,{\alpha }_1$ are the parameters to be calculated in the model, and ${\epsilon }_i$ is the residual of the model $(i=1,2,\dots ,395)$.

With the gradual implementation of the policy of returning farmland to forest in the study area and the rapid development of urbanization, many rural surplus laborers choose to go out for work, so migrant income has gradually become a part of the total income of farmers, which is the first key reason for the change in farmers’ income. In addition, farmers may face risk shocks, such as diseases and children’s education, and farmers’ crops may present reduced production due to geological disasters or extreme weather. However, farmers’ savings and material capital can smooth the income changes caused by these risk shocks to a certain extent, which is the second main reason for farmers’ income changes. Therefore, in this paper, the change in farmers’ income mainly considers three aspects: migrant working, risk impact, and risk smoothing ability.

According to Chaudhuri et al. [35], if the future consumption fluctuation of farmers can be replaced by the square of the regression residual of their consumption level, then

$$C{F}_i={\beta }_0+{\beta }_1{M}_i+{\beta }_2{R}_i+{\beta }_3{S}_i+e_i$$

(2)

where $C{F}_i$ is the future consumption variance of the $i$-th farmer, $M_i$ represents the proportion of the migrant income in the total income of the $i$-th rural household, $R_i$ is the risk shock encountered by the $i$-th rural household, $S_i$ is the risk smoothing means of the $i$-th farmer, ${\beta }_0,{\beta }_1,{\beta }_2,{\beta }_3$ are the parameters to be calculated in the model, and $e_i$ is the residual of the model.

Paxson et al. posit that $I_i^p$ is mainly determined by human capital, productive capital, and financial capital [39,40,49,50]. According to the real situation of the sample of rural households in the sample regions, this paper selects six indicators ($Age,Edu,Laborer,Land,Tool,Deposit$) to characterize the persistent income of farmers. Dercon et al. suggest that the risk shocks faced by rural households can be divided into two types: homogeneous risk and heterogeneous risk [37,40,49,50]. Homogeneous risks refer to the risks that affect farmers throughout the village within a certain period, such as price changes in agricultural products and crop damage caused by climate. Heterogeneous risks refer to the risks that affect only individual farmers, such as disease of family members and death of livestock. According to the actual survey situation, this paper selects $Agri\text{-}shock$ to represent homogeneous risk and selects four indicators ($Hea\text{-}shock,Edu\text{-}shock,Hou\text{-}shock,Haza\text{-}shock$) to characterize heterogeneous risk. The risk smoothing means of the farmers mainly include their human capital, material capital, and financial capital. This paper selects five indicators ($Age,Edu,Laborer,Physicap,Deposit$) to characterize farmers’ means of risk smoothing; the specific meaning of each indicator is presented in

Table 2. Measurement indicators and the assignment of consumption level and consumption variance.

Variable	Assignment
C	The per capita consumption level in a household (RMB/person)
Age	Household head’s age in a household (years)
Edu	Education years of household head in a household (years)
Laborer	Number of laborers of rural households (persons)
Land	Cultivated land and garden plot area in a family (hm2)
Tool	Whether there are large-scale productive tools (water pump, tractors, and so on) in a household (0 = no; 1 = yes)
Deposit	Whether there are any deposits in a household (0 = no; 1 = yes)
Physicap	Present values of housing, production tools, vehicles, and durable goods in a family (RMB 10,000)
M	Migrant income’s ratio in total income of rural households (%)
Agri-shock	Whether the crop losses are more than RMB 1000 due to prices or extreme weather (0 = no; 1 = yes)
Hea-shock	Whether the medical expenses of family members exceed RMB 3000 (0 = no; 1 = yes)
Edu-shock	Whether the family’s expenditure on children’s education exceeds RMB 3000 (0 = no; 1 = yes)
Hou-shock	Whether the consumption of house construction or durable goods exceeds RMB 5000 (0 = no; 1 = yes)
Haza-shock	Whether the damage caused by disaster impact exceeds RMB 5000 (0 = no; 1 = yes)

. Therefore, Formulas (1) and (2) are adjusted to Formulas (3) and (4), respectively (Figure 2).

$$\ln (C_i)={\alpha }_0+{\alpha }_{1}Ag{e}_i+{\alpha }_{2}Ed{u}_i+{\alpha }_{3}Labore{r}_i+{\alpha }_{4}Lan{d}_i+{\alpha }_{5}Too{l}_i+{\alpha }_{6}Deposi{t}_i+{\epsilon }_i$$

(3)

$$\begin{array}{l}C{F}_i={\beta }_0+{\beta }_1{M}_i+{\beta }_{2}Agri\text{-}shoc{k}_i+{\beta }_{3}Hea\text{-}shoc{k}_i+{\beta }_{4}Edu\text{-}shoc{k}_i+\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}{\beta }_{5}Hou\text{-}shoc{k}_i+{\beta }_{6}Haza\text{-}shoc{k}_i+{\beta }_{7}Ag{e}_i+{\beta }_{8}Ed{u}_i+{\beta }_{9}Labore{r}_i\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+{\beta }_{10}Physica{p}_i+{\beta }_{11}Deposi{t}_i+e_i\end{array}$$

(4)

3.4. Research Methods

On the basis of determining the variables and models of the future consumption level and consumption variance of farmers, this research applies the ordinary least squares method (OLS) to regress the model of farmers’ future consumption level and then uses the feasible generalized least squares method (FGLS) to regress the model of farmers’ future consumption variance to obtain the parameters to be estimated in the two models. All regression processes were carried out using Stata11.0 software.

4. Results

4.1. Variables’ Descriptive Statistical Characteristics

First, a descriptive statistical analysis of each variable in the model was conducted using Stata11.0 software to obtain the minimum (Min), maximum (Max), mean (Mean), and standard deviation (SD) of each variable in the model of the future consumption level and consumption variance of farmers (

Table 3. Descriptive statistical characteristics of variables.

Variable	Min	Max	Mean	SD
C	333.33	667,400.00	8910.65	13,637.78
M	0.00	1.00	0.35	0.40
Age	19.00	81.00	48.71	13.00
Edu	0.00	16.00	5.88	3.69
Laborer	0.00	8.00	3.05	1.69
Land	0.00	1.63	0.30	0.31
Tool	0.00	1.00	0.29	0.45
Deposit	0.00	1.00	0.24	0.43
Physicap	0.36	333.39	30.33	34.63
Agri-shock	0.00	1.00	0.02	0.13
Hea-shock	0.00	1.00	0.36	0.48
Edu-shock	0.00	1.00	0.20	0.40
Hou-shock	0.00	1.00	0.13	0.34
Haza-shock	0.00	1.00	0.00	0.05

).

Table 3 shows that the per capita consumption level of farmers in the study area in 2023 was RMB 8910.65, which was 0.27 percentage points lower than the per capita consumption expenditure of rural residents across China in the same year (RMB 12,124). From the perspective of income, the average ratio of migrant work income to the total income of farmers was 35%. On the one hand, crop yield and income decrease in accordance with the decrease in cultivated land area and the impact of extreme weather on crop growth; on the other hand, new urbanization provides employment opportunities and positions for rural surplus labor, and the income from migrant working increases accordingly. Therefore, the ratio of rural households’ migrant working income to total household income increases year by year.

For human capital, the rural household heads are mainly middle-aged, with an average age of 48.71. The heads of rural households have only 5.88 years of education, on average, which means that the rural household heads’ overall education level is low, with the majority having a primary school education. The average number of laborers per household is generally approximately three people. For natural capital, the average land area (cultivated land and garden land area) owned by farmers is only 0.3 hm², as the implementation of the “returning farmland to forest” policy has greatly reduced the cultivated land area of farmers. In terms of material capital, 29% of farmers own large-scale productive tools, and the average present value of farmers’ material capital is RMB 303,300. In terms of financial capital, only 24% of farmers have deposits.

In terms of homogeneous risk, only 2% of farmers face the impact of crop damage. In terms of heterogeneous risks, 36% of farmers face the impact of disease expenditure, 20% of farmers face the impact of children’s education expenditure, and 13% of farmers face the impact of building houses (or renovating dilapidated houses or purchasing houses) or purchasing large durable goods, while none of the farmers suffer from damage to houses and other fixed assets due to disasters.

4.2. Results Analysis of Farmers’ Poverty Vulnerability

On the basis of the statistical analysis of each variable, OLS and FGLS regression analyses were carried out on Formulas (3) and (4), respectively, by using Stata11.0 software to obtain variables that have a significant impact on the future consumption level and consumption variance of farmers. The regression analysis results are described in

Table 4. Regression analysis results of the econometrical model for the rural households’ poverty vulnerability in the upper reaches of the Min River, China.

Variable	$\mathbf{Model} 1 \mathbf{ln}({\mathit{C}}_{\mathit{i}})$	$\mathbf{Model} 2 \mathit{C}{\mathit{F}}_{\mathit{i}}$	$\mathbf{Model} 3 \mathit{C}{\mathit{F}}_{\mathit{i}}$
Age	−0.001	−0.005	−0.006
Edu	0.040 ***	−0.181	−0.048 ***
Laborer	0.048 *	−0.051	−0.031
Land	0.342 **
Tool	0.078
Deposit	−0.106	−0.187 ***	−0.498
Physicap		−0.168 **	−0.198 ***
M		−0.045 *	−0.176 ***
Agri-shock		0.189 ***
Hea-shock			0.245 *
Edu-shock			−0.138
Hou-shock			0.129 **
Haza-shock			−0.160
Constant	8.949 ***	0.801 **	0.687 *
F statistics	3.39 ***	6.94 ***	5.74 ***
N	395	395	395
R-squared	0.095	0.112	0.130

Note: *, **, and *** represent significance at the 0.1, 0.05, and 0.01 levels, respectively.

. Model 1 is the result of the OLS analysis of Formula (3). Model 2 and Model 3 are the regression results of the FGLS analysis of Formula (4). The difference between Model 2 and Model 3 is that the risk shock variable of the former is homogeneous risk, namely, Agri-shock, while the risk shock variables of the latter are the four heterogeneous risks (Hea-shock, Edu-shock, Hou-shock, and Haza-shock).

4.2.1. Results Analysis of Average Consumption Value

Model 1 shows that, during the study period, Edu, Laborer, and Land have a significant positive impact on the per capita consumption level of rural households in the future. Specifically, with a one-year increase in Edu, the future per capita consumption level of rural households will improve by 4%. The future per capita consumption level of rural households will improve by 4.8% for each additional laborer. For every 1 hm² increase in Land, the future per capita consumption level of rural households will increase by 34.2%. An increase in Edu means an improvement in the educational level of the rural household head, which offers the possibility of more jobs or positions with more technical content, so the income of the rural households will increase, and the corresponding consumption level will increase. With more laborers, the income of farmers and, in turn, the consumption level will increase. With more Land, the area and yield of crops planted by farmers will increase, so agricultural income will increase, and the consumption level will increase in response.

4.2.2. Results Analysis for Consumption Fluctuations

Model 3 shows that, during the study period, Edu has a significant negative effect on the future consumption fluctuations of farmers, indicating that improving the educational level of the rural household head can weaken the poverty vulnerability caused by heterogeneous risks. Specifically, when faced with heterogeneous risks, the future consumption variance of rural households with highly educated household heads is 4.8% lower than that of rural households whose heads are less educated. However, Model 2 shows that, during the study period, Edu has no significant effect on the future consumption variance of farmers, which means that the improvement in the household head’s education level cannot weaken the poverty vulnerability caused by homogeneous risks.

Model 2 shows that, during the study period, Deposit has a significant negative effect on the future consumption variance of rural households, indicating that the increase in farmers’ savings can weaken the poverty vulnerability caused by the homogeneous risk Agri-shock. Specifically, when faced with homogeneous risks, the future consumption variance of farmers with savings is 18.7% lower than that of farmers without savings. However, Model 3 shows that, during the study period, Deposit has no significant influence on the future consumption variance of farmers, which means that the increase in farmers’ savings cannot weaken the poverty vulnerability caused by heterogeneous risks.

Model 2 and Model 3 show that, during the study period, Physicap and M have a significant negative impact on the future consumption variance of rural households, which means that an increase in Physicap and M can reduce the poverty vulnerability caused by homogeneous risks and heterogeneous risks. Specifically, with an RMB 10,000 increase in Physicap, the future consumption variance of farmers facing homogeneous risk and heterogeneous risk shocks will decrease by 16.8% and 19.8%, respectively. When M increases by 1%, the future consumption variance of farmers facing homogeneous risk and heterogeneous risk shocks will decrease by 4.5% and 17.6%, respectively.

Model 2 shows that, during the study period, the homogeneous risk Agri-shock has a significant positive effect on the future consumption variance of rural households. Specifically, the future consumption variance of farmers facing Agri-shock increases by 18.9% compared with those without Agri-shock, which indicates that Agri-shock is the main factor that causes poverty vulnerability among farmers.

Model 3 shows that, during the study period, among the four heterogeneous risks, Hea-shock and Hou-shock have a significant positive impact on the future consumption variance of rural households. Specifically, the future consumption variance of farmers facing Hea-shock increases by 24.5% compared with those not facing Hea-shock, and the future consumption variance of farmers facing Hou-shock increases by 12.9% compared with those not facing Hou-shock. This indicates that Hea-shock and Hou-shock are also the main factors that cause poverty vulnerability among farmers.

4.3. Robustness Test

To verify the accuracy of the above regression results, this paper adopts the following two methods to conduct robustness tests on the regression results of farmers’ poverty vulnerability. The test results are shown in

Table 5. Robustness test.

Variable	5% Winsorization			Replacing the Core Explanatory Variable
	Model 1	Model 2	Model 3	Model 1	Model 2	Model 3
Age	−0.003	−0.010	−0.009	−0.005	−0.002	−0.003
Edu	0.060 ***	−0.181	−0.053 ***	0.120 ***	−0.330	−0.150 ***
Laborer	0.062 *	−0.080	−0.060	0.037 *	−0.025	−0.160
Land	0.461 **			0.241 **
Tool	0.126			0.054
Deposit	−0.305	−0.364 ***	−0.420	−0.004	−0.528 ***	−0.760
Physicap		−0.152 **	−0.208 ***		−0.034 **	−0.307 ***
M		−0.006 *	−0.019 ***		−0.013 *	−0.386 ***
Agri-shock		0.327 ***			0.090 ***
Hea-shock			0.446 *			0.315 *
Edu-shock			−0.008			−0.102
Hou-shock			0.320 **			0.028 **
Haza-shock			−0.053			−0.280
Constant	−0.003	−0.010	−0.009	−0.005	−0.002	−0.003
F statistics	0.060 ***	−0.181	−0.053 ***	0.120 ***	−0.330	−0.150 ***
N	0.062 *	−0.080	−0.060	0.037 *	−0.025	−0.160
R-squared	0.461 **			0.241 **

Note: *, **, and *** represent significance at the 0.1, 0.05, and 0.01 levels, respectively.

. Specifically, one approach is to apply a 5% winsorization for key numerical variables, and the other approach is to replace the core explanatory variable. Compared with Table 4, the significance and direction of the explanatory variables have not changed significantly, indicating that the results of the benchmark regression model are robust and reliable.

5. Discussion

The research results of this study show that, during the study period, Edu, Laborer, and Land have significant positive impacts on the future per capita consumption level of farmers, which is consistent with the research results of Tai et al. and Xu et al. [44,50]. This is mainly because the study area of this paper and that of Tai et al. and Xu et al. both belong to poor mountainous areas of China, where farmers have a certain dependence on land resources or agriculture, so Land has significant effects on the future per capita consumption level of rural households [44,50]. Edu can determine the employment choice of the family labor force to a large extent, thus determining the total income of the family, so Edu also has a significant impact on farmers’ future per capita consumption level. The more laborers there are, the higher the total income of the rural households will be, so the consumption level will increase accordingly. Therefore, Laborer has a significant impact on the future per capita consumption level of farmers. According to the current situation in the study area, rural education should be revitalized, and the quality of rural education should be improved in the nine-year compulsory education stage by improving the professional quality of rural teachers, increasing the policy support for rural education, and guiding funds and talents into rural education to improve the overall education level of rural household members. In the secondary and higher education periods, various scholarships and grants are provided to alleviate the financial pressure that rural students face to attend school, thereby enabling them to receive higher education. In addition, the agricultural and nonagricultural skills of the rural labor force can be improved through various trainings, and the characteristic agricultural products can gradually realize industrialization to increase the income of farmers.

M was also found to weaken the poverty vulnerability caused by homogeneous and heterogeneous risks during the study period, which is consistent with the research results of Xu et al. [50]. This may be due to the fact that the total income of farmers in the study area is complementary to the agricultural income and nonagricultural income. Therefore, when farmers face Agri-shock, they may increase their total income through migrant work, thereby weakening the poverty vulnerability caused by Agri-shock. To prevent family poverty in the face of various heterogeneous risks, family members may spend more time and energy engaging in migrant work and thereby increase the total income and reduce poverty vulnerability caused by heterogeneous risks to a certain extent.

As a homogeneous risk, Agri-shock had a significantly positive effect on the future consumption variance of rural households during the study period, which is consistent with the study results of Kochar and Tai et al. but different from the research results of Xu et al. [44,50,51,52]. This is mainly because the agricultural income of farmers in Kochar’s and Tai et al.’s study areas accounted for a large proportion of their total income, so Agri-shock was bound to cause income changes, which in turn caused changes in consumption variance [44,51,52]. However, the proportion of agricultural income within the total household income of farmers in Xu et al.’s study area was small, while the proportion of migrant income was large [50]. Therefore, Agri-shock had little effect on the total income of rural households, so its impact on the future consumption variance of farmers was not significant. To improve the agricultural anti-risk ability of rural households in the research region, the government should strengthen the publicity of agricultural insurance, gradually enhance awareness of agricultural insurance among rural households, and improve the financial subsidy level for agricultural insurance in the study area. Governments at all levels should also formulate clear agricultural insurance development plans and ensure their smooth implementation.

The heterogeneous risk Hea-shock had a significantly positive effect on the future consumption variance of rural households during the study period, which is consistent with the research results of Kochar and Tai et al. [44,51,52]. This is mainly because, on the one hand, Hea-shock reduces the household labor force and the total household income; on the other hand, a large amount of money is spent on disease treatment, which increases household consumption and therefore has a significant effect on the future consumption variance of rural households. To reduce the impact of the heterogeneous risk Hea-shock on farmers’ consumption, the relevant departments in the study area need to further improve the rural basic medical insurance system, expand the types of chronic diseases and the scope of drugs that can be reimbursed, increase the reimbursement ratio of hospitalization and serious diseases, gradually establish an outpatient reimbursement system, and simplify the reimbursement process to reduce the economic burden of residents in the study area.

The heterogeneity risk Hou-shock had a significant positive effect on farmers’ future consumption variance during the study period, while Edu-shock had no significant effect on farmers’ future consumption variance, which is consistent with the research results of Xu et al. [50]. For farmers in poor mountainous areas, building houses (or renovating dilapidated houses, buying houses) or buying large durable goods is a large expenditure. Therefore, Hou-shock entails consumption fluctuations; that is, it has a significant impact on the future consumption variance of rural households. However, the average education level of farmers in both study areas is low and basically at the level of compulsory education; therefore, Edu-shock has little effect on farmers’ consumption fluctuations. To reduce the poverty vulnerability caused by Hou-shock, the relevant departments should gradually optimize the subsidy standards for building houses (or renovating dilapidated houses or buying houses) in rural areas, especially in poor mountainous areas, to ensure the housing security of rural households and reduce their consumption fluctuations caused by building houses (or renovating dilapidated houses or buying houses).

This investigation indicated that there have been a few large-scale geological disasters in the study region in recent years, so the heterogeneous risk Haza-shock has no significant effect on the future consumption fluctuations of farmers during the study period. However, the study area is part of an area where geological disasters, such as landslides and debris flows, occur frequently. Therefore, relevant departments should make corresponding disaster prevention and mitigation plans to prevent or reduce the loss of life and property caused by various disasters.

Physicap was found to reduce poverty vulnerability linked to homogeneous and heterogeneous risks during the study period, which is different from the research results of Tai et al. and Xu et al. [44,50]. A possible reason is that when rural households in the research area are affected by different kinds of risk shocks, they choose to sell stored grain, farmed poultry, and other large-scale production equipment, transportation, and other physical capital to smooth the impact of risk shocks on rural household consumption. However, Tai et al.’s research results showed that to accumulate sufficient physical capital, farmers chose not to sell physical capital to smooth the impact of risks on their families even when they were affected by risks [44]. Xu et al.’s research results showed that because there was no local second-hand commodity market or leasing market, rural households could not smooth the impact of risk shocks by selling or mortgaging physical capital [50].

Deposit was found to weaken the poverty vulnerability caused by homogeneous risk during the study period, which is consistent with the research results of Tai et al., while the research results of Xu et al. showed that Deposit could weaken the poverty vulnerability caused by both homogeneous risk and heterogeneous risk [44,50]. This difference is mainly because although farmers in the study areas considered in this paper and by Tai et al. had deposits, these deposits mainly came from agricultural income (the average ratios of agricultural income in total income were 65% and 82%, respectively) and thus were not large, such that the deposits could alleviate only the poverty vulnerability caused by Agri-shock to a certain extent [44]. However, these meager savings had no effect on the poverty vulnerability caused by heterogeneous risks. In the research region of Xu et al., the average proportion of rural households’ out-migration income was as high as 64% of their total income, so the amount of savings was relatively high and therefore could reduce the poverty vulnerability caused by both homogeneous risks and heterogeneous risks [50].

6. Conclusions

Based on the microsurvey data of rural households in the upper reaches of the Min River in 2023, this study analyzes the poverty vulnerability of farmers in the region and its influencing factors. The research results have important practical significance for the establishment of risk and vulnerability early warning mechanisms in the new poverty alleviation strategy in the region. Moreover, the measurement method of poverty vulnerability in this paper can provide some reference for research on the poverty vulnerability of farmers in other poor mountainous areas in China. However, the types of capital and risk shocks that affect the consumption level and consumption fluctuation of farmers undergo a dynamic process of change. Therefore, follow-up studies should undertake dynamic, longitudinal research on farmers in different years, analyze the change process and trends of various kinds of capital and risk shocks to reduce poverty vulnerability, realize the sustainable development of farmers, and propose targeted policies and recommendations.