Internet Use and the Poverty Vulnerability of Rural Households: From the Perspective of Risk Response

Abstract

Alleviating the poverty vulnerability of rural households helps to resist risk shocks and maintain livelihood security. From a risk-response-ability and -strategy perspective, this paper constructs a theoretical analysis framework for the impact of internet use on the poverty vulnerability of rural households using data from the 2018 China Family Panel Studies (CFPS) to conduct an empirical analysis. The results show that internet use has a significant impact on alleviating the poverty vulnerability of rural households. After a robustness test using the IV-probit two-step method, the results remained robust and reliable. When viewed from a regional perspective, the impact of internet use on the poverty vulnerability of rural households is reduced in the regions with a lower risk of poverty and higher use of the internet. An analysis of the influence mechanism shows that risk-response ability represented by human capital, social capital, and selfefficacy and risk-response strategy represented by nonagricultural employment and commercial insurance participation have a significant mediating effect between internet use and the poverty vulnerability of rural households. Therefore, based on improving network infrastructure, the government should guide rural households to improve their risk-response ability and implement risk-response strategies to prevent them from falling into poverty in the future.

1. Introduction

Poverty is a significant issue that nations all over the world must confront and address to achieve social development. It greatly impacts a country’s economy, people’s livelihoods, and long-term sustainable economic growth. Since 2012, China has achieved tremendous progress as it continues to pursue the targeted poverty alleviation program. According to China’s current criteria, all 98.99 million rural poor people had been lifted out of poverty as of February 2021, together with all 832 poor counties and 128,000 poor villages [1]. However, there is a need to realize that poverty is a changing condition due to the impact of risk shocks. The impoverished population may experience swings as a result of external shocks such as natural catastrophes, illnesses, and unemployment, which raises the possibility of falling back into poverty. According to data from the State Council Poverty Alleviation Office, approximately 2 million people who have been lifted out of poverty are at risk of falling back into poverty. This risk is higher in rural areas, where farmers are more likely to fall back into poverty as a result of disasters and illnesses due to a lack of livelihood assets. If this issue cannot be resolved, consolidating the results of poverty alleviation will be impossible. Scholars are accustomed to using poverty vulnerability to measure the possibility of families falling into poverty in the face of risk shocks. This dynamic measurement method can better reflect people’s welfare status, because empirical evidence shows that the most effective way to solve poverty is to prevent poverty before it occurs rather than reverse it afterwards [2]. Therefore, studying rural families’ poverty vulnerability represents a major change in understanding poverty and should be a key component of China’s poverty alleviation strategy.

With the rapid advancement of information technology, the internet has increasingly become a driving force in promoting reform and innovation, profoundly influencing rural residents’ production, life, and ideological concepts, thus playing an important role in rural revitalization and poverty alleviation strategies [3]. Specifically, “Internet +” has created a resource integration platform that spans time and place for targeted poverty alleviation. It has realized critical functions, such as data monitoring, information exchange, and resource integration, for poverty alleviation activities. Simultaneously, “Internet + education”, “Internet + entrepreneurship”, “Internet + health”, “Internet + tourism”, “Internet + rural e-commerce”, and other modes have increased the number of farming-related livelihood opportunities and employment channels and have become a key factor in raising farmer income and lowering poverty. Farmers can interact with, obtain, and consume information in real-time at a low cost using the internet. The internet increases human, social, financial, and other forms of livelihood capital and also emancipates farmers’ minds and renews their ideas. Using the internet or We Media creates more nonagricultural employment or selfemployment opportunities [4,5]. This will help rural households alleviate their current poverty and reduce the possibility of falling into poverty in the future. As a result, this study investigates the influence and function routes of internet use on the poverty vulnerability of rural households in the context of the “Internet +” era. It aims to provide a theoretical basis for the government to use the internet to alleviate and prevent rural households from going into dynamic poverty.

Based on the above analysis, this paper empirically examines the effect and mechanism of internet use on the vulnerability to poverty of rural households using the CFPS2018 data. The potential contributions are as follows:

First, taking the vulnerability of rural households to poverty as the research object, this study discussed the problem of alleviating the future poverty of rural households from the perspective of internet use. This is to provide policy references for the government to manage dynamic poverty in the internet era;

Second, based on the perspective of risk-response ability and strategy, we have constructed a theoretical analysis framework on the impact of internet use on the vulnerability of rural households to poverty. We analyzed the intermediary effect of risk-response ability and strategy in internet use to alleviate the vulnerability of rural households to poverty;

Third, because the explained variable of the poverty vulnerability is binary, an improved mediation effect model was used for the analysis. Simultaneously, in view of the possible endogenous problems, the IV-Probit two-step method was used for the robustness test, providing a reference research method for similar research.

The next section presents the literature review and the research framework of the relationship between internet use and the poverty vulnerability of rural households. Section 3 introduces the source of the research data, the descriptive analysis of variables, and the methodology selection. The empirical results of the model are presented in Section 4, and Section 5 provides a discussion. Section 6 summarizes our main findings.

2. Literature Review and Hypotheses Development

2.1. Overview of the Poverty Vulnerability Theory

Poverty is a social state of families that can be described and measured. While it is an expost measurement method [6], the poverty status of rural households is a stochastic phenomenon. Poor households today may be able to escape poverty with the assistance of the government or society, and some households who are not poor may also slip into poverty as a result of the effects of diseases or natural disasters [7]. As a result, poverty research should not only include postmeasurements based on poverty status but also dynamic predictions of whether rural households would slip into poverty in the future. Poverty vulnerability is seen as a dynamic and forward-looking indicator to measure family welfare [8]. The World Bank defines poverty vulnerability as the possibility that individuals or households will drop their living standards and quality to a socially recognized level due to the risk shocks that may occur in the future [9]. This definition reflects the predictive judgment of whether the family will likely fall into poverty in the future. An in-depth study of this definition is conducive to exploring how rural households can prevent welfare reductions under risk shocks.

Domestic and international experts have mainly discussed three aspects regarding the influencing factors of the poverty vulnerability of rural households: the characteristics of risk shocks, socioeconomic factors, and the coping ability and strategies of families. Risk shock is a significant factor in the vulnerability of rural households to poverty. Poor households exposed to risk shocks will forego investment in productive assets, increasing their chances of being impoverished [10]. Mekasha believes that serious disease will swiftly push farmers back into poverty [11].

Using a theoretical model, Carter suggested that low-income farmers can only employ asset smoothing to cut consumption expenditures in the face of risk shocks in order to keep household assets from slipping into the asset poverty trap [10]. The poverty vulnerability of rural households is also influenced by socioeconomic factors. Government transfer payments, social security expenditures, inclusive finance, the subsistence allowance system, the transfer of agricultural land, and health poverty alleviation project have all been shown to significantly impact farmers’ poverty vulnerability [12,13,14,15,16,17]. Because risk shocks and socioeconomic factors are uncontrollable for rural households, many researchers have focused on studying how rural households might reduce poverty vulnerability from the standpoint of risk response. Existing research focuses on two aspects: risk-response ability and risk diversification mechanism. The family’s human, material, financial, social, and other livelihoods, as well as psychological capital [5,14] can reflect the family’s ability to resist risk shocks and act as a buffer and protector when risk shocks occur. According to research, rural households should also adopt reasonable risk-sharing strategies in addition to improving their risk-response ability. It has been demonstrated that taking part in health insurance, buying commercial insurance, and selecting nonagricultural employment or diversified livelihood activities all lower the likelihood of farmers becoming vulnerable to poverty [5,7,18,19].

2.2. Research on Internet Use and Poverty Vulnerability

The promotion and use of the internet as an information and communication technology (ICT) tool has had a significant impact on the output and way of life of the entire society. Scholars have begun to pay attention to the association between internet use and rural household poverty in this context but have yet to reach a uniform conclusion. Some academics believe that the development and usage of the internet can help to close the digital divide, enhance the livelihoods of impoverished rural households, and effectively reduce poverty [20]. Specifically, internet finance, e-commerce, internet access, inclusive finance, etc., have all been demonstrated to be beneficial to the alleviation of rural poverty [21,22,23,24]. They are also crucial instruments in China’s use of internet thinking to improve the level of targeted poverty alleviation. Furthermore, expanding individual information channels, increasing employment and entrepreneurship opportunities, lowering transaction costs, private lending, nonagricultural employment, social capital, and human capital have all been identified as important paths through which internet use affects the poverty of rural households [4,5]. However, relatively little research has been carried out on internet use and vulnerability to poverty. According to Chapman, information technology can provide the necessary information to enable the rural poor to create more livelihood capital and capacity, which are conducive to developing diversifying livelihoods and immigrant migration strategies to help reduce the poverty vulnerability of the poor [25]. Zhang conducts empirical research based on Chinese data and finds that the use of the internet reduces the level of poverty vulnerability by increasing income levels, enhancing the ability of farmers to obtain information, and promoting nonagricultural employment [26]. However, some academics think that ICT cannot help the poor. This is largely due to so-called “digital poverty”, i.e., the lack of access to information and communication technologies, resulting in an insufficiency of useful information [27,28]. Furthermore, Galperin discovered that the influence of internet use is unpredictable in developing countries and does not significantly alleviate poverty [29].

Despite the abundance of study findings, the nature of the association between internet use and rural family poverty remains unknown. The great majority of research solely covers the effect of internet use on farmers’ already-existing poverty and its mechanism. There are very few studies on whether rural households are likely to fall into poverty in the future as a result of internet use. In light of this, this article examines the effect and mechanism of internet use on the poverty vulnerability of rural households and offers a theoretical foundation for the government to develop internet poverty alleviation strategies.

2.3. Hypotheses Development

This paper builds a theoretical analysis framework for the effect of internet use on the poverty vulnerability of rural households. It does this by scouring previous research literature (as shown in Figure 1). The internet has been widely used as a tool to alleviate poverty, and farmers now have access to education (such as distance education), healthcare (such as telemedicine), and market access (such as e-commerce) via the internet [30]. These options enhance the livelihood capacity of rural households and reduce their vulnerability to risk shocks. Furthermore, the use of the internet and the popularity of smart devices have increased the number of channels through which rural households may acquire information, enhanced the transparency of their market information, and lowered the transaction costs of market participation. This not only encourages rural households to adopt new technologies to improve productivity, but it also increases agricultural income, realizes accumulation strategies, and encourages rural households to diversify risks and realize coping strategies through income diversification and participation in agricultural credit or insurance. Based on the above analysis, the following hypothesis is proposed:

Hypothesis 1.

Rural households’ use of the internet has a significant impact on alleviating poverty vulnerability.

The poverty vulnerability of rural households is primarily caused by risk shock. Risk shocks significantly raise the risk of middle-skilled households slipping into the asset poverty trap, whereas low-ability households with larger beginning asset allocations will still fall into chronic asset poverty [31]. On the contrary, rural households with strong risk response abilities can take various measures to mitigate risks in the face of shocks and even accumulate capital before the risks come, reducing the possibility of risk exposure. Specifically, the risk-response ability of rural households is expressed as human capital, social capital, and selfefficacy. Human capital reflects the education level and health status of rural households. It not only indicates farmers’ ability to earn a living but also decides whether they can make reasonable decisions to overcome challenges in the face of risk shocks. Social capital represents opportunities for rural households to receive assistance when deciding on livelihood plans or dealing with shocks; it also reflects their ability to deal with risks. Selfefficacy is a measure of the confidence and control of rural households in completing a certain behavior. It assesses the ability of rural households to actively modify their mindset and adopt effective measures to decrease poverty and vulnerability when confronted with risks or difficulties. It is an endogenous motivator.

In this age of intelligent information, rural households may quickly access useful information via the internet and obtain information-based knowledge resources. Additionally, education and medical services cost less, which is conducive to enhancing human educational capital and healthy human capital [32]. Rural households with high human capital can not only choose high-return livelihood strategies and earn high incomes but also pay more attention to human capital investments, such as education and health, forming a virtuous circle and reducing the possibility and destructiveness of risk shocks [33]. Additionally, they can also make rational production, consumption, and employment decisions in the face of risks based on knowledge and experience. Second, the internet alters the way people communicate with one another by erasing the boundaries of time and geography. The promotion and application of selfmedia social platforms such as QQ, WeChat, and microblogs can increase social capital by helping rural households maintain their original social network relationships and expand new social networks [34]. These social capitals provide not only rural households with job resources and business chances to enhance their income but also with financial assistance, information, and emotional support when risk arises, reducing the possibility of poverty in the future. Finally, rural households’ interpersonal engagement through the internet can reduce psychological stress and relieve loneliness, tension, and anxiety, allowing farmers to work and live more effectively, as well as promote selfefficacy [35]. Selfefficacy determines whether people can actively use internal and external resources to solve problems when faced with difficulties, challenges, and failures, which has an important impact on the vulnerability of rural households to poverty. Based on the above analysis, the following hypothesis is proposed:

Hypothesis 2.

Rural households’ use of the internet can further reduce their poverty vulnerability by improving their capacity to manage risks, which includes building up their human capital, social capital, and sense of selfefficacy.

Rural households must not only enhance their family’s risk-response abilities but must also plan risk-response tactics ahead of time and consciously disperse risks in order to cope better with risk shocks. Meanwhile, rural households’ use of the internet will encourage them to adopt risk response strategies to help alleviate their vulnerability to poverty. The internet can effectively alleviate the problem of information asymmetry caused by geographical isolation in rural areas so that rural households can search for more nonagricultural employment information and social information at a low cost, which improves the understanding of personal environment and transformation ability and provides the possibility to realize diversified nonagricultural employment [36]. Different from European and American countries, China’s agricultural business model is dominated by the small-scale peasant economy, with less arable land per capita and a higher degree of land fragmentation, which is not conducive to the promotion of modern agricultural technology, characterized by mechanization, informatization, and scale [37]. Therefore, the production efficiency and income of farmers engaged in agriculture are low. Moreover, because they are vulnerable to natural disasters, factor prices, and crop prices, their economic income is unstable and they are faced with high livelihood risks [38]. On the contrary, rural households who choose nonagricultural employment instead of solely agricultural strategies can not only raise their household income levels but also diversify the risk of income fluctuations and reduce their vulnerability to poverty [39]. Therefore, nonagricultural employment is an important risk-response strategy for Chinese farmers to reduce poverty vulnerability.

Second, with the continuous integration of internet technology and the insurance business, rural households can use the internet to obtain various insurance information and claim settlement cases, as well as engage with others online via interactive platforms. All of these help rural households realize the critical role of business insurance in risk diversification and prevention, and it increases their willingness and demand for commercial insurance, while rural households who buy commercial insurance can obtain corresponding economic compensation or credit enhancement financing after being impacted by risks, minimizing the risk of losing livelihood capital and falling into poverty vulnerability [40]. Based on the above analysis, the following hypothesis is proposed:

Hypothesis 3.

Rural households’ use of the internet can further reduce their poverty vulnerability by optimizing the risk response strategy, i.e., to participate in nonagricultural employment and commercial insurance purchases.

3. Data, Models, and Variables

3.1. Data Sources and Processing

The data in this article come from the China Family Tracking Survey (CFPS) in 2018. The sample data comprises over 16,000 homes in 25 provinces (autonomous areas and municipalities directly under the central government), and the sampling frame includes more than 95% of the population, resulting in broad coverage and strong representation. Because the research object is the poverty vulnerability of rural households, only the data of rural households are maintained in the original data. After removing some samples with plainly aberrant data or missing primary variables, 4633 valid samples were obtained.

3.2. Variable Selection

This paper divides all variables into three categories: explained variable, core explanatory variables, and control variables.

3.2.1. Explained Variable

This paper takes poverty vulnerability as the explained variable. Poverty vulnerability measuring approaches primarily comprise vulnerability as expected poverty (VEP), vulnerability as low expected utility (VEU), and vulnerability as uninsured exposure to risk (VER) [41]. This research chooses the VEP approach to quantify poverty vulnerability since it is an ex ante calculation method by comparing future income with a given poverty threshold and does not require panel data [41]. The basic idea is to calculate the average and variance of a family’s predicted income to determine the probability that the family’s income level is lower than a certain value (typically the poverty line) in the period of t + 1, as follows:

$$VE{P}_{it}=\mathrm{Pr}(Y_{i,t+1}\le Z)$$

(1)

In this equation, VEP_it denotes the vulnerability index of the family at the time of t. Y_i,t+1 is the per capita income level of the family, i, at the time of t + 1, and Z is the socially recognized poverty line. The vulnerability level of the family increases with a greater vulnerability index, and vice versa. Assuming that personal income follows a log-normal distribution, a family’s i poverty vulnerability can be represented as

$$\hat{V}u{l}_i=\mathrm{Pr}(Ln{Y}_i<LnZ|X_i)=\phi \left(\frac{LnZ-\hat{E}(Ln{Y}_i|X_i)}{\sqrt{\hat{V}(Ln{Y}_i|X_i)}}\right)$$

(2)

Among them, X_i represents the observable factors that affect the income of rural households, such as per capita property, family size, and so on (According to the sustainable livelihood theory, farmers’ livelihood results are importantly influenced by their livelihood capital and livelihood strategy. This study chooses selfowned land from natural capital, family size, family burden, years of education, and health level from human capital, selfowned housing from physical capital, farmer per capita property and government subsidies from financial capital, nonagricultural employment and land lease from livelihood strategy, together with social capital; these jointly construct the control variables that affect the income of farmers), while $\hat{E}(Ln{Y}_i|X_i)$ and $\hat{V}(Ln{Y}_i|X_i)$ represent the expected value and variance of the FGLS estimation equation obtained by constructing weights based on estimated income and squared residual error, respectively. Ln$Z|X_i$ is the logarithm of the poverty line. This study utilizes the World Bank’s standard of USD 2 per person per day as the poverty line vulnerability standard and 50% as the poverty threshold standard and chooses 50% as the poverty threshold standard to evaluate the poverty vulnerability of rural households. The value is 1 if there is a greater than 50% chance that the family’s future income will fall below the line denoting poverty vulnerability, indicating that the family is vulnerable to poverty; otherwise, the value is 0.

3.2.2. Explanatory Variables

As the internet plays a more important role in information acquisition and dissemination, this article employs “Internet use” and “Internet information dependency” as explanatory variables. When rural households use computers or mobile phones to access the internet, internet use is assigned a value of 1; otherwise, it is assigned a value of 0. The mean value of the importance of rural households utilizing the internet to receive information is used to measure “Internet information dependence”.

3.2.3. Control Variables

The control variables were set from the individual and family levels based on the research findings of previous academics in order to avoid the influence of other factors on the explained variables. At the family level, the control variables include per capita property, family size, selfowned housing, government subsidies, land rental, and family burden. The control variables at the individual level include the household head’s age, gender, and marital status. For a complete description of the indicators, see

Table 1. Variable list and descriptive statistics.

Variable	Variable Name	Variable Name	Mean	SD	Min	Max
Vep	Poverty vulnerability	If the income level of rural households in the future is less than the poverty vulnerability line, it is assigned 1; otherwise, it is assigned 0	0.073	0.260	0	1
Int1	Internet using	If someone in the household uses computer internet or mobile internet, it is assigned 1; otherwise, it is assigned 0	0.660	0.474	0	1
Int2	Dependence on network information	The mean value of the importance of family members using the network to obtain information	2.614	1.243	1	5
Pro	Family property	Divide total household cash and deposits by household size and then take the logarithm	5.677	4.027	0	13.017
Fam	Family size	The population of families	3.926	1.963	1	21
Sub	Government subsidies	If the family receives government subsidies, it is assigned 1; otherwise, it is assigned 0	0.609	0.488	0	1
Hou	Number of houses	Number of houses owned by families	1.061	0.527	0	5
Lease	Land lease	If the land is leased to others, it is assigned 1; otherwise, it is assigned 0	0.138	0.345	0	1
Land	Selfowned land	If the family is allocated collective land, it is assigned 1; otherwise, it is assigned 0	0.891	0.311	0	1
Bur	Family burden	The proportion of family members who do not work	0.487	0.299	0	1
Age	Age	Age of head of household	50.8	13.537	16	82
Sex	Sex	A male household head is assigned 1, while a female household head is assigned 0	0.571	0.495	0	1
Mar	Marital status	If the head of household is married, it is assigned 1; otherwise, it is assigned 0	0.871	0.335	0	1
Non	non- agricultural employment	The number of people in the household engaged in non-agricultural employment divided by family population	0.776	0.904	0	1
Insur	Commercial insurance purchase	If the family purchases commercial insurance, it is assigned 1; otherwise, it is assigned 0	0.266	0.442	0	1
Human	Human capital	If the frequency of the family learning or exercise rose between 2016 and 2018, it is assigned 1; otherwise, it is assigned 0	0.368	0.482	0	1
Social	Social capital	If the sum of the two indicators’ monthly expenditures on “expenses on favors and gifts” and “post and telecommunication expenses” of the family rose between 2016 and 2018, it is assigned 1; otherwise, it is assigned 0	6.033	2.067	0	10.374
Self	Selfefficacy	The mean value of selfefficacy among family members. “How confident you are about the future” is used to measure selfefficacy	4.151	0.748	1	5

.

3.2.4. Mediating Variables

According to the theoretical analysis, this research believes that internet use reduces poverty vulnerability by influencing the risk-response ability and response strategies of rural households. Three variables are included in the risk-response strategies of rural households: human capital, social capital, and selfefficacy. The risk response strategies of rural households include nonagricultural employment and commercial insurance participation. Table 1 has a description of the indicators.

3.3. Model Selection

3.3.1. Baseline Regression Model

The probit model is a typical binary discrete choice model. Its advantage is that it is based on the cumulative normal distribution, which will avoid the unrealistic situation that the probability estimate value is less than 0 or greater than 1 in the linear probability model and can more accurately describe the decision-making process of our research object. Poverty vulnerability is the dependent variable in this article. When the anticipated income level of a rural household is less than the poverty vulnerability line, it is assigned a value of “1”, otherwise of “0”. As a result, the probit model is chosen to investigate the impact of internet use on the poverty vulnerability of rural households. The probit model expression is as follows:

$$\mathrm{Pr}(VE{P}_i=1)=\phi (\alpha +\beta T_i+\gamma X_i+\epsilon )$$

(3)

In Equation (3), VEP_i denotes whether rural households have poverty vulnerability, T_i denotes internet usage, which is the core explanatory variable, X_i is the control variable, and ε is the residual term.

3.3.2. Mediating Effect Model

The mediating effect model analyzes the influence path of the independent variable on the dependent variable. The majority of studies have adopted the three-step mediation method proposed by Baron and Kenny [42]; however, some academics have noted that this approach is only applicable when the mediator and dependent variables are both continuous variables. If any of the variables are binary, the regression coefficients are not on the same scale and cannot be simply compared [30]. Because the dependent variable in this paper, poverty vulnerability, is a binary variable, the traditional three-step mediation method cannot be applied. For the research on the intermediary effects, where dependent variables or intermediary variables are binary variables, the intermediary model is constructed by reference to the method of Iacobucci [43], as follows:

$$Y_i^{\prime }=i_1+cN{U}_i+{\lambda }_{1}Control+{\epsilon }_{1i}$$

(4)

$$M_i=i_2+\alpha N{U}_i+{\lambda }_{2}Control+{\epsilon }_{2i}$$

(5)

$$Y_i^{″}=i_3+c^{\prime }N{U}_i+b{M}_i+{\lambda }_{3}Control+{\epsilon }_{3i}$$

(6)

Y_i, NU_i, and M_i denote poverty vulnerability, internet use, and mediating variables, respectively, in Equations (4)–(6). i₁, i₂, and i₃ are constant terms, c, ɑ, and c’ are the coefficients of internet use, λ₁, λ₂, and λ₃ are coefficients of the control variable, b is the coefficient of the intermediate variable, and ε_1i, ε_2i, and ε_3i are random error terms. This study uses the research of Iacobucci (2012) for reference to implement the intermediary effect test [43] and transforms the regression coefficient of Equation (5) into Z_a = a/Se(a), transforming the regression coefficient of Equation (6) into Z_b = b/Se(b), and then calculates the value of Z_a × Z_b. Finally, we test the significance of Z_a × Z_b with the Sobel method. The following is the test statistic:

$$Z=\frac{Z_{a\times b}}{SE(Z_{a\times b})}=\frac{Z_a\times Z_b}{SE(Z_{a\times b})}=\frac{Z_a\times Z_b}{\sqrt{Z_a^2+Z_b^2+1}}$$

(7)

The specific operation is to measure the confidence interval of Z_a × Z_b using the Rmediation package of the R software 4.2.1 [44]. The mediating effect is significant when the confidence interval does not contain 0.

3.4. Descriptive Statistics

Table 1 shows the variable definitions and descriptive statistics. For 7.27% of rural households, the probability of their future income falling below poverty vulnerability is greater than 50%, and they are in a state of poverty vulnerability. Moreover, the standard deviation of poverty vulnerability (0.260) is greater than the mean value (0.073), indicating a large difference in the poverty vulnerability of rural households. In terms of internet usage, 65.96% of rural households utilize the internet, and their dependence on network information is moderate. Among other variables, 60.9% of rural households in the overall sample received government subsidies, 89.1% were assigned collective land, 13.8% leased their land to others, and 26.6% purchased commercial insurance. The mean value of the proportion of people without work in the family is 48.7%. Rural households own 1.061 houses per capita; the average number of people in each household is 3.329, and the mean value of the proportion of nonagricultural employment in the family is 77.6%.

4. Result

4.1. Total Sample Regression Results

Based on the data from the China Family Panel Studies in 2018, the probit regression model was used to test the impact of internet use on the poverty vulnerability of rural households. The estimated results are shown in

Table 2. Internet use and poverty vulnerability of rural households: regression results.

Variable	Model 1	Model 2	Model 3	Model 4	Model 5
	Full Sample	Full Sample	East	Central	West
Int1	−0.067 *** (0.007)		−0.052 *** (0.013)	−0.055 *** (0.013)	−0.084 *** (0.013)
Int2		−0.017 *** (0.004)
Pro	−0.015 *** (0.001)	−0.016 *** (0.001)	−0.012 *** (0.001)	−0.014 *** (0.002)	−0.018 *** (0.002)
Fam	0.019 *** (0.002)	0.016 *** (0.002)	0.015 *** (0.003)	0.013 *** (0.003)	0.027 *** (0.003)
Sub	−0.002 (0.008)	−0.003 (0.008)	0.012 (0.011)	−0.020 (0.013)	−0.005 (0.014)
Hou	−0.054 *** (0.008)	−0.056 *** (0.008)	−0.049 *** (0.014)	−0.042 *** (0.012)	−0.065 *** (0.013)
Lease	−0.060 *** (0.012)	−0.062 *** (0.012)	−0.033 * (0.017)	−0.068 *** (0.019)	−0.099 *** (0.026)
Land	0.057 *** (0.015)	0.064 *** (0.016)	0.050 ** (0.020)	0.047 * (0.025)	0.087 ** (0.040)
Bur	0.097 *** (0.014)	0.115 *** (0.015)	0.098 *** (0.022)	0.093 *** (0.024)	0.089 *** (0.028)
Age	0.002 *** (0.001)	0.002 *** (0.001)	0.002 *** (0.001)	0.002 *** (0.001)	0.001 ** (0.001)
Sex	0.010 (0.007)	0.007 (0.007)	−0.019 * (0.011)	−0.014 (0.012)	−0.004 (0.013)
Mar	−0.021 * (0.011)	−0.020 * (0.011)	−0.021 (0.018)	−0.029 * (0.016)	−0.006 (0.021)
Provincial fixed effect	Yes	Yes	Yes	Yes	Yes
Observation	4633	4633	1744	1273	1616

All results are marginal effects; ***, **, and * represent significance levels of 1, 5, and 10% respectively, and those in brackets are (robust) standard errors.

. The results of the full-sample regression are reported for Models 1 and 2. In Model 1, the poverty vulnerability of rural households, internet usage, and all control variables were added to the model. In Model 2, the poverty vulnerability of rural households, the importance of using the internet to obtain information, and all control variables were added to the model. Models 4–6 report the probit regression coefficient of internet usage on the poverty vulnerability of rural households in the eastern, central, and western regions, respectively.

As shown in Table 2, the coefficient of internet usage and the importance of using the internet to obtain information was significant at the level of 1% and was less than zero, which meant that internet usage by rural households could significantly reduce the possibility of falling into poverty in the future. Thus, Hypothesis 1 is supported by the evidence. In addition, it can be seen that, in Models 3–5, internet usage can significantly reduce the poverty vulnerability of rural households in the eastern, central, and western regions at the level of 1%. The degree of influence was the west, central, and east in order of magnitude.

In terms of control variables, per capita property, selfowned house, land lease, and marital status of the head of household have significantly negative effects on the poverty vulnerability of rural households. In contrast, family size, family burden, and the age of the head of the household have significantly positive effects. While government subsidies and the gender of the head of the household have no significant statistical significance.

4.2. Endogeneity Test

Two-stage least-square (2SLS) tests were used for the endogeneity test in this paper. The results are shown in

Table 3. IV probit two-stage regression results.

Variable	Vep
	Phase I	Phase II	Phase I	Phase II
Int1		−1.354 *** (0.448)
Int2				−0.369 ** (0.121)
Instrumental variable	0.652 *** (0.049)		0.785 *** (0.040)
cons	0.662 *** (0.049)	−1.894 *** (0.558)	2.333 *** (0.150)	−1.813 *** (0.602)
Control variable	yes	yes	yes	yes
F statistic	174.46		154.62
Wald test	2.06		2.84
Wald test p value	0.1507		0.0919
n	4633	4633	4633	4633

(1) All results are regression coefficients; (2) ***, and ** indicate significance levels of 1, and 5% respectively, and those in brackets are (robust) standard errors.

. First, the regression coefficient of county-level internet usage and the mean value of county-level network information dependence in the first stage was significantly positive. According to the weak instrumental variable test, the F value of the first-stage regression test is 174.46 and 154.62, which far exceeded the threshold of the 10% bias level (16.38), indicating that there is no weak instrumental variable problem. Second, the regression coefficient of the second stage was −1.354 and −0.369, significant at the level of 5%. However, demonstrating the reliability of the benchmark regression results, that is, internet usage, has a significant inhibitory effect on individuals’ willingness to have children. The Wald test results were 2.06 and 2.84, respectively, which did not pass the significance test of 5%, indicating that internet usage and network information dependence are exogenous variables, that is, the original model does not have endogeneity, and its regression results are robust and reliable.

4.3. Test of Intermediary Effect

This paper uses human capital, social capital, and selfefficacy as mediating variables and conducts mediation effect tests according to the steps analyzed in the mediation effect model above to investigate whether internet use has an impact on poverty vulnerability through the risk response ability of rural households. The results are reported in

Table 4. Intermediary effect results based on risk-response ability.

Variable	Model 6	Model 7	Model 8	Model 9	Model 10	Model 11
	Human	Vep	Social	Vep	Self	Vep
Int1	0.449 *** (0.046)	−0.404 *** (0.070)	0.242 *** (0.053)	−0.292 *** (0.086)	0.070 *** (0.027)	−0.424 *** (0.069)
Human		−0.170 *** (0.069)
Social				−0.438 *** (0.081)
Self						−0.107 *** (0.029)
Control variable	yes	yes	yes	yes	yes	yes
Regional fixed effect	yes	yes	yes	yes	yes	yes
cons	0.863 *** (0.125)	−1.490 *** (0.223)	5.730 *** (0.243)	−0.543 *** (0.275)	3.906 *** (0.062)	−1.238 *** (0.273)
n	4633	4633	4633	4633	4633	4633

(1) All results are regression coefficient; (2) *** represent significance level of 1%, and those in brackets are (robust) standard errors.

below. The results in Models 6, 8, and 10 show that, after controlling for other influencing factors, the regression coefficients of internet usage on human capital, social capital, and selfefficacy were 0.449, 0.537, and 0.070, respectively, with standard errors of 0.046, 0.086, and 0.027. Models 7, 9, and 11, respectively, reflect the regression results of human capital, social capital, and selfefficacy on the poverty vulnerability of rural households under the control of internet usage and it's covariates. It can be seen that the regression coefficients for human capital, social capital, and selfefficacy were −0.170, −0.273, and −0.107, respectively, with standard errors of 0.069, 0.013, and 0.029. The Remediation software package of R software was used for testing, and the 95% confidence intervals for human capital, social capital, and selfefficacy were [−0.130, −0.025], [−0.160, −0.059], and [−0.014, −0.002], respectively. None of them contain 0, indicating that human capital, social capital, and selfefficacy have significant negative mediating effects between internet usage and the poverty vulnerability of rural households. Thus, Hypothesis 2 is supported by the evidence.

In order to investigate whether internet usage has an impact on poverty vulnerability through the risk-response strategies of rural households, the mediation effect tests were conducted based on the steps analyzed in the mediation effect model above. The results are reported in

Table 5. Intermediary effect results based on risk response strategies.

Variable	Model 12	Model 13	Model 14	Model 15
	Non	Vep	Insur	Vep
Int1	1.030 *** (0.047)	−0.189 *** (0.074)	0.460 *** (0.052)	−0.401 *** (0.069)
Non		−0.591 *** (0.076)
Insur				−0.275 *** (0.093)
Regional fixed effect	yes	yes	yes	yes
Control variable	yes	yes	yes	yes
cons	0.589 *** (0.136)	−1.130 *** (0.237)	−0.743 *** (0.132)	−1.451 *** (0.224)
n	4633	4633	4633	4633

(1) All results are regression coefficient; (2) *** represent significance level of 10%, and those in brackets are (robust) standard errors.

below. The results in Models 12 and 14 show that, after controlling for the other influencing factors, the regression coefficients of internet usage on nonagricultural employment and commercial insurance participation were 1.030 and 0.460, respectively, with standard errors of 0.047 and 0.052. Models 13 and 15, respectively, reflect the regression results of nonagricultural employment and commercial insurance participation on the poverty vulnerability of rural households under the control of internet usage and it’s covariates. It can be seen that the regression coefficients for nonagricultural employment and commercial insurance participation were −0.591 and −0.275, respectively, with standard errors of 0.076 and 0.093. The Remediation software package of R software was used for testing, and the 95% confidence interval of the nonagricultural employment and commercial insurance participation was [−0.747, −0.474] and [−0.203, −0.055], respectively. None of them contain 0, indicating that nonagricultural employment and commercial insurance participation have significant negative mediating effects between internet usage and the poverty vulnerability of rural households. Thus, Hypothesis 3 is supported by the evidence.

In order to compare the size of the mediating effects of the different mediating variables, we calculated their Z-statistics. According to Formula (7), the Z-statistics of human capital, social capital, selfefficacy, nonagricultural employment, and commercial insurance participation are −2.377, −3.463, −2.071, −7.322, and −2.788, respectively, all of which are less than −1.96, further demonstrating the reliability of the calculation results obtained using the multiplication and integration method. From the absolute value of the Z-statistics, it can be seen that the extent of the mediating effect is in the order of nonagricultural employment, social capital, commercial insurance purchase, human capital, and selfefficacy.

5. Discussion

In the face of the lack of research on the dynamic poverty of rural households in the context of internet usage, this study built a theoretical framework to determine the impact of internet use on the vulnerability of rural households to poverty from the perspective of risk-response abilities and strategies. The study used the probit model and improved the intermediary effect model to empirically analyze the impact of internet use on the poverty vulnerability of rural households, providing theoretical and practical guidance for the governance of dynamic poverty in the context of the internet.

5.1. Theoretical Significance

First, the results show that internet use significantly and negatively affects the poverty vulnerability of rural households. This is after controlling for other variables. Consistent with the findings of Chapman and Zhang, the more frequently rural households use the internet or rely on network information, the less likely they are to fall into poverty vulnerability in the future [25,26]. Moreover, among all the explanatory variables, the impact coefficient is second only to household burden, which means that, in the information age, the use of the internet has indeed played an important role in alleviating the poverty of rural households in the future. This may be because the popularization and application of the internet have broadened the channels for rural households to obtain information, making it easier and faster for rural households to obtain information resources, such as education, medical care, consumption, and employment. Mastering this information can effectively enhance human capital, social capital, and other livelihood capabilities, allowing families to mitigate risks and reduce vulnerability when exposed to external shocks.

Second, this paper also further discusses the regional differences in the impact of internet use on rural households’ poverty vulnerability. The findings show that the impact of internet use on the poverty vulnerability of rural households is reduced in the regions with a lower risk of poverty and higher use of the internet. According to the full-sample data, the distribution of poverty and poverty vulnerability among rural households in the eastern, central, and western regions increases in turn, while the distribution of computer internet usage data decreases in turn, and the gap between the three indicators in the western region and the other two regions is large. It means that rural households in the eastern and central regions have little pressure to reduce poverty, with internet penetration being quite high, resulting in a minor influence of internet use on the poverty vulnerability of rural households. However, people in the western region are under great pressure to reduce poverty now and in the future, and their internet penetration is relatively low, so there is more potential for improvement, and the region could benefit more from the national “Internet +” poverty reduction strategy. As a result, internet use is more effective in alleviating poverty vulnerability.

Third, this study verifies the mediating role of the three risk-response abilities, including human capital, social capital, and selfefficacy, for internet usage and the poverty vulnerability of rural households. Risk shocks, such as natural disasters, economic volatility, and household changes, may push a household into poverty vulnerability regardless of whether it is currently in poverty. Therefore, there is a need to study how a rural household can prevent dangers in advance to reduce vulnerability. The existing research mainly discusses this topic from the perspective of risk-prevention ability and risk-sharing strategy. According to this study, risk-response ability is mostly represented in human capital, social capital, and selfefficacy. Improving these risk-prevention abilities can lower the likelihood of rural households becoming vulnerable when a risk arises. At the same time, existing research shows that people’s use of the internet can improve human capital, social capital, and selfefficacy [32,34,35]. As a result, this study constructed a theoretical framework of internet use, risk-response ability, and the poverty vulnerability of rural households and empirically verified that these three risk-response abilities significantly mediate the poverty vulnerability of rural households through internet usage. The enlightenment derived from this conclusion in terms of management is that rural households can consciously improve their human capital, social capital, and sense of selfefficacy through the use of the internet, thus improving their risk-response ability and lowering the likelihood of future poverty.

Fourth, this study verifies the mediating role of two risk factors: choosing off-farm employment and purchasing commercial insurance, on internet use and the poverty vulnerability of rural households. Before a risk arrives, taking the appropriate risk-response strategy may also alleviate the poverty vulnerability of rural households in addition to improving their risk-response ability. This study constructs a theoretical framework of internet usage, risk-response strategy, and farmers’ poverty vulnerability and empirically tests that the two risk-response strategies of choosing nonagricultural employment and purchasing commercial insurance have significant intermediary effects on internet usage and farmers’ vulnerability to poverty. As a result, farmers can consciously learn about nonagricultural employment and the purchase of commercial insurance via the internet and adopt effective risk-response strategies, which can also prevent families from falling into poverty in the future.

Finally, in terms of the control variables, per capita property, selfowned housing, land lease, and the marital status of the head of the household have significantly negative effects on the poverty vulnerability of rural households. This means that increasing the per capita property of peasant households, increasing the number of houses, renting land to others, and the married status of the household heads may smooth the impact of risk shocks on poverty vulnerability. The possible reasons include two aspects: on the one hand, there is a “welfare dependence” effect of government subsidies. After receiving government subsidies, the recipients will have psychological dependence, become depressed in spirit, and will have an abnormal value orientation and negative attitude toward life [45], which will lead to the recipients’ lack of progress in thought and reduced labor supply behavior, making it difficult to improve the family’s poverty situation. On the other hand, people receiving government subsidies are generally poor, and receiving subsidies can only help them maintain their basic living standards. As a result, when the risk comes, these people are very likely to fall into poverty.

5.2. Practical Significance

This study uses CFPS2018 data to analyze the impact and mechanism of internet use on the vulnerability of rural households to poverty from the perspective of risk-response abilities and strategies. It provides some enlightenment on managing farmers’ poverty alleviation practice and policy design.

First, the government should speed up efforts to upgrade internet infrastructure, improve the access rate and service performance of the internet, and increase the coverage of the network to rural areas. In addition, the government should provide financial subsidies or preferential tariff packages to purchase network equipment for vulnerable rural households to increase the penetration rate of the internet. Finally, it is necessary to rely on the government network platform to release information on education, medical care, insurance, agricultural technology, employment, and other topics, as well as to improve farmers’ skills in internet usage through training so that farmers can make better use of the internet.

Second, government departments should use the internet to help farmers improve their risk-response abilities. For example, training in e-commerce and network applications should be provided to encourage rural households to actively participate in projects such as “Internet + education” and “Internet + medical care”. This will help them to obtain a high-quality education and medical resources at a low cost. The government should also promote online social platforms in rural areas to assist rural households in developing online social capital. The government should also encourage rural households to participate in poverty alleviation projects, such as “Internet + E-commerce”, to improve their income and selfefficacy.

Third, the government should help rural households use the internet to implement risk-response strategies. For example, the internet employment information service platform should be used to provide timely information on nonagricultural employment, entrepreneurship, and business transactions. This will encourage rural households to pursue nonagricultural employment or livelihood diversification. At the same time, in order to reverse farmers’ conflicting views on insurance and make them understand the role of insurance in future risk shocks, the internet should be used to publicize and popularize information about risk prevention and insurance purchases so that rural households can rationally choose to purchase based on their own needs.

5.3. Limitations

This study has the following limitations. First, the China Family Panel Studies in 2018 is the latest data covering internet usage, poverty vulnerability, and the mediating variables; therefore, this study uses the cross-sectional data of the CFPS in 2018, which cannot reflect the development process of longitudinal dynamics. Therefore, after the release of CFPS 2020, further research can be carried out with the panel data. Second, this paper focuses primarily on the impact and mechanism of internet use on the poverty vulnerability of rural households rather than conducting in-depth research on the impact paths of different groups. Because the different groups of rural households may use the internet differently, this may affect poverty vulnerability in the different paths taken. As a result, quantitative research in this field could be conducted in the future.

6. Conclusions

Based on the perspective of risk-response abilities and strategies, this study analyzes the impact of internet use on the poverty vulnerability of rural households. Specifically, the robust model verified that internet usage significantly reduces the poverty vulnerability of rural households. For every 1% increase in the use of the internet by rural households, there is a 0.067% decrease in the probability of falling into poverty vulnerability. The impact coefficient is second only to family burden. From the regional perspective, the impact of internet use on the poverty vulnerability of rural households is highest in the western region, followed by the central region and then the eastern region. An analysis of the influence mechanism shows that the risk-response abilities represented by human capital, social capital, and selfefficacy, and the risk-response strategies represented by nonagricultural employment and commercial insurance participation have a significant mediating effect between internet usage and the poverty vulnerability of rural households.

With the rapid development of information, internet technology continues to penetrate all fields of the economy and society, changing people’s production, life, and even their thinking and behavior. Most scholars are keen to discuss the problem of ICT in alleviating poverty among rural households. However, for rural households, the question of how to use the internet to prevent them from falling into poverty in the future is more worthy of attention. Our research has certain enlightening significance, i.e., in the internet era, improving farmers’ use of information and communication technology can help them reduce poverty vulnerability, and using the internet to improve risk-response abilities and optimize risk-response strategies is an important impact path. This conclusion aims to draw the attention of the Chinese government and rural households to alleviate poverty and vulnerability through internet usage and to take active and effective countermeasures to prevent risks in advance.