Poverty alleviation and rural revitalization: Perspective of fiscal spending and data evidence from 81 Chinese counties

Abstract

This paper builds a theoretical model of government performance functions for poverty alleviation using county-level panel data from 81 counties in China from 2014 to 2019. It uses a Panel-Tobit model and mechanism tests to verify the effect of fiscal policies on poverty reduction, and consolidates the robustness of the results through a series of extended methods, such as endogeneity treatment, robustness tests, and heterogeneity analysis. The results show that (1) poverty-related allocations can significantly reduce poverty incidence, and the effect of poverty reduction is more pronounced in poor counties; (2) public spending can significantly reduce poverty incidence, and the effect of poverty reduction through public spending is more pronounced in the sample of poor counties and nonfunded pilot counties; (3) poverty reduction can affect poverty incidence through primary and secondary industry development, and the effect of poverty reduction through primary industry development is more significant, while public spending does not affect poverty incidence through primary and secondary industries; and (4) improving health services can reduce poverty to a large extent, while education development has no effect on poverty reduction due to the long return cycle. This study suggests increasing the size of poverty-specific allocations and public spending, strengthening industry support, and implementing differentiated policy initiatives according to local conditions to improve the impact of poverty reduction.

1. Introduction

Poverty is an issue to which every country attaches great importance. Solving the poverty problem can effectively increase national happiness, reduce the poverty gap and promote the consumption ability and consumption level of residents. Since 2013, the Chinese government has made poverty eradication an important policy goal and formulated and introduced a series of targeted plans and measures. As a result, China has achieved remarkable results in poverty reduction[1]. The number of people living in poverty has been reduced from 98.99 million at the end of 2012 to 5.51 million at the end of 2019, with the poverty rate falling from 10.2% to 0.6%, representing a decrease in poverty of more than 10 million per year for seven consecutive years [2]. By the end of 2020, China will officially declare that poverty has been completely eliminated. China's tremendous success in poverty reduction would not have been possible without substantial financial investment, especially guaranteed fiscal support. Since 2015, when China announced the start of its poverty alleviation program, fiscal spending on poor areas has totalled nearly 1.6 trillion yuan. Fiscal spending, as one of the most important instruments of China's fiscal policy, plays an important role in reducing poverty and promoting income growth of the population[3]. Therefore, this paper compiles research on the impact of fiscal spending on poverty alleviation at the county level, which is of great theoretical significance and practical value (see Fig. 1, Fig. 2, Fig. 3).

Fig. 1.

Relationship between poverty alleviation and rural revitalization.

Fig. 2.

Theoretical analysis and empirical processes.

Fig. 3.

Policy recommendations and practice recommendations from empirical analysis.

At present, the relationship between tax expenditures and the effect of poverty reduction is widely and enthusiastically discussed in academia, with the following three main views: (1) The effect of fiscal spending on poverty reduction is remarkable. Fiscal expenditures have the function of optimizing resource allocation, regulating income distribution, stabilizing economic development, improving the protection system, and promoting coordinated regional development. Scholars who hold these views believe that tax expenditures help increase the property income of the poor, increase the accumulation of human capital in poor areas, and improve production conditions in poor areas, thereby increasing productivity and promoting the development of agribusiness to reduce poverty[[4], [5], [6]]. Rapid economic growth and sound fiscal policies contribute to improving the welfare of the poor by using tax revenues in certain ways and forms to alleviate poverty and inequality[[7], [8], [9], [10]]. In addition, some scholars have validated the poverty-reducing effect of a particular tax policy. [11] found that the construction of agricultural distribution infrastructure helped to increase the income of urban and rural residents and reduce the incidence of poverty in China. Nizar (2018) argued that public spending on education and health in Tunisia significantly reduced inequality and extreme poverty. [12] noted that progressive tax policies help increase tax revenues and raise social welfare, which is well suited for poverty reduction. [13] found that lowering the corporate income tax can reduce the poverty level of Malaysian households. [14] argued that the strict blockade enforced during this period makes it difficult for poorer people in society to survive, and fiscal transfers can reduce this negative impact.

• (2)The poverty reduction effect of fiscal spending is not remarkable. The poverty alleviation effect of fiscal policy is not always effective because the stages of economic and social development vary from country to country, and there is variability in the causes that lead to the occurrence of poverty. In addition, poor people are often groups with low literacy, weak labor skills, and poor access to information and resources and are disadvantaged in accessing public resources such as education, health, and employment, making it difficult to obtain sufficient social resources and leading to the ineffectiveness of fiscal policies in reducing poverty[[15], [16], [17], [18]]; argue that fiscal spending in health and housing policy areas is conducive to poverty reduction but that fiscal spending in social areas is not effective in reducing poverty. [19] suggest that cash transfers are not significant in reducing long-term poverty. [20] finds that tight fiscal policies may exacerbate inequalities in resource allocation and are not conducive to addressing residual rural poverty. [21] find that restrictive public spending and revenues will have a significant impact on increasing poverty and social inequality. [22] found that fiscal decentralization did not improve the efficiency of local public goods and services in metropolitan, municipal and district assemblies (MMDAs) in the Ghanaian region.
• (3)There are regional, structural, and group differences in poverty reduction through fiscal expenditures. Many scholars have argued that the effectiveness of tax expenditures depends on whether policy goals are reasonably set since tax policy is often linked to corresponding policy goals[23]. [24] argues that the impact of capital-related budget expenditures on poverty reduction is stronger than the impact of recurrent budget expenditures on poverty reduction. [25] find a significant spatial spillover effect of rural public spending on poverty reduction. [26] suggest that the introduction of fiscal austerity measures has a greater impact on the poorer classes. more. [27] argues that fiscal transfers are affected by benefit targets and recipient groups, with significant differences in their impact on poverty reduction. Enami (2019) argues that transfers are generally more effective than taxes in reducing inequality. [28] argue that fiscal policy can mitigate the poverty trap problem caused by BGPs (High-growth balanced growth paths). [29] argues that regional government revenues and intergovernmental transfers contribute significantly to poverty reduction, while regional government expenditures do not.

The available research shows that scholars still disagree on whether poverty reduction through fiscal expenditures is significant. Due to the great differences in the level of economic and social development among different regions, the level of public services and the causes of poverty have become complicated, and the nature of poverty has diversified. Therefore, there is still much room for research on poverty alleviation methods and the impact of fiscal expenditure. To objectively and accurately determine the role of fiscal expenditure in poverty alleviation in counties in China and compare and analyze the differences in the poverty alleviation effects of fiscal expenditure between different regions, this paper constructs a theoretical model of the government's poverty alleviation performance function from three perspectives: production function, poverty alleviation utility, and poverty alleviation performance. Based on county-level panel data of 81 counties in China from 2014 to 2019, we test the poverty alleviation effects of fiscal policy using Panel-Tobit model and mediating effects model, and consolidate the results through a series of extensions such as endogeneity treatment, robustness test, and heterogeneity analysis. The potential innovations of this paper are as follows: (1) The sample data are more unique. Compared with the provincial-level panel data, the county-level panel data used in this paper are more accurate and valid. In addition, the poverty incidence data comes from the business management module of the China Poverty Alleviation and Development Information System, which is more accurate compared to the published statistics. (2) The research perspective is distinctive. Most of the existing studies examine the role of fiscal expenditures in poverty alleviation from the provincial level, while there is little literature examining the effects of fiscal expenditures on poverty alleviation at the county level in China as the direct implementer of fiscal expenditures. This study examines the impact of fiscal expenditures on poverty alleviation in 81 counties in China and the results of the study can provide a more scientific reference for the formulation of county-level fiscal spending policies. (3) The research method is more accurate. Compared with the regression of poverty incidence using a static panel model, this paper uses poverty incidence as the explained variable, and poverty incidence is truncated between 0 and 1, so it is more appropriate to use a panel Tobit model.

The paper is divided into five parts: The first part is the introduction, the second part establishes the theoretical model, the third part presents the econometric model and the selection of variables, the fourth part analyzes the empirical results, and the fifth part concludes the paper.

2. Theoretical model

According to China's financial management system, the poverty alleviation effect of budget expenditure policy is mainly based on two approaches: poverty-specific allocations and public expenditure (excluding poverty-specific allocations). Poverty-specific allocations refer to special fiscal allocations for the poor that include direct transfers, reduced consumption expenditures, and improved vocational training opportunities. Public expenditures refer to the effect of reducing poverty by affecting the overall size or structure of the economy, which in turn encourages the poor to increase their incomes, improve prices of agricultural products, reduce transaction costs in agriculture, and ultimately achieve poverty reduction. Based on the traditional economic growth model, this paper introduces fiscal policy measures and local government performance to improve the model and then examines the impact of fiscal policy measures on poverty reduction. The model is constructed as follows.

2.1. Production function

Since the poverty-reducing effect of fiscal policy primarily examines the effectiveness of the use of fiscal resources, the subject of the effect is the government. Following [30,31]; the government production function for poverty reduction is defined as follows:

$$Y_{\mathrm{i}\mathrm{t}}=D_{it}{K}_{it}^{{\alpha }_k}{\left(G_{it}+E_{it}\right)}^{{\alpha }_{g+e}}{L}_{it}^{1-{\alpha }_k-{\alpha }_{g+e}}$$ (1)

where $Y_{\mathrm{i}\mathrm{t}}$ represents county-level per capita income, $D_{it}$ represents county-level economic development level, $K_{\mathrm{i}\mathrm{t}}$ represents county-level capital input, $L_{\mathrm{i}\mathrm{t}}$ represents county-level human capital, ${\alpha }_k$ and ${\alpha }_{g+e}$ represent the output elasticity of capital and the output elasticity of fiscal expenditure, respectively, and take values in the range of (0, 1). To facilitate the analysis, fiscal spending is further subdivided into poverty-specific allocations and public expenditures, where $E_{it}$ represents poverty-specific allocations and $G_{it}$ represents public expenditures of local governments, which include education expenditures, social security and employment expenditures, health expenditures, and agriculture, forestry, and water expenditures.

2.2. Effectiveness function

Based on the research results of [30]; Li and Yin (2012) and Fan and Gao (2019), it is assumed that the utility function of each county government can have two components, namely, real output and poverty alleviation performance, and the utility function of each county government is determined as follows:

$$U_{\mathrm{i}\mathrm{t}}=\left(1-{\beta }_{it}\right)\mathit{ln}{Y}_{it}+{\beta }_{it}\mathit{ln}{P}_{it}$$ (2)

where $U_{\mathrm{i}\mathrm{t}}$ represents the overall utility function of county governments, $Y_{it}$ represents the output per capita of county areas, $P_{it}$ represents the poverty alleviation performance of county areas, and ${\beta }_{it}$ represents the target bias coefficient, which satisfies $0<{\beta }_{it}<1$.

2.3. Performance functions

According to the China Rural Poverty Monitoring Report (2014) and Ren (2014), the county government poverty reduction performance function is assumed to obey the Cobb–Douglas form:

$$P_{\mathrm{i}\mathrm{t}}=\frac{1}{E_{it}^{\gamma }}\frac{1}{{\left(1+{\theta }_{\mathrm{i}\mathrm{t}}\right)}^{\phi }}$$ (3)

where $P_{\mathrm{i}\mathrm{t}}$ stands for poverty reduction performance at the county level, $E_{it}$ stands for poverty-specific allocation, ${\theta }_{\mathrm{i}\mathrm{t}}$ stands for poverty incidence, $\gamma$ stands for the performance elasticity of pro-poor expenditures, and $\phi$ stands for the performance elasticity of poverty incidence. $\gamma +\phi >1$, $\gamma >0\text{,}$ and $\phi >0$.

2.4. Factors influencing pro-poor performance

Substituting Equations (1), (3) into Equation (2), the partial derivative of ${\beta }_{it}$ is obtained as

$$\frac{\partial U_{\mathrm{i}\mathrm{t}}}{\partial {\beta }_{it}}=-\mathit{ln}\left[D_{it}{K}_{it}^{{\alpha }_k}{\left(G_{it}+E_{it}\right)}^{{\alpha }_{g+e}}{L}_{it}^{1-{\alpha }_k-{\alpha }_{g+e}}\right]+\mathit{ln}\left[\frac{1}{E_{it}^{\gamma }}\frac{1}{{\left(1+{\theta }_{\mathrm{i}\mathrm{t}}\right)}^{\phi }}\right]$$ (4)

from $\frac{\partial U_{\mathrm{i}\mathrm{t}}}{\partial {\beta }_{it}}=0$, the collation yields

$$\ln \left(1+{\theta }_{it}\right)=-\frac{\gamma }{\phi }\mathit{ln}{E}_{it}-\frac{1}{\phi }\mathit{ln}{D}_{it}-\frac{{\alpha }_k}{\phi }\mathit{ln}{K}_{it}-\frac{{\alpha }_{g+e}}{\phi }\mathit{ln}\left(G_{it}+E_{it}\right)-\frac{1-{\alpha }_k-{\alpha }_{g+e}}{\phi }\mathit{ln}{L}_{it}$$ (5)

Equation (5) shows that the incidence of poverty ${\theta }_{\mathrm{i}\mathrm{t}}$ has a negative relationship with poverty earmarking $E_{it}$, public expenditure $G_{it}$, regional economic development level $D_{it}$, capital level $K_{it}$, and human capital $L_{it}$. This indicates that the two major fiscal expenditure policies, poverty earmarking and public expenditure, can reduce the incidence of poverty, and the higher the level of economic development of a region, the richer the resources, such as labor and capital, and the more favorable they are to improving poverty. Based on this, the following hypotheses are proposed in this paper:

Hypothesis 1

Increasing poverty-specific allocations will reduce the incidence of poverty.

Hypothesis 2

Increasing public spending will reduce the incidence of poverty.

3. Model and variables

In this section, we present the model and variables used to investigate the relationship among poverty-related allocations, public expenditures, and poverty incidence. We begin by introducing the econometric model of government performance functions for poverty alleviation. We then describe the variables used to operationalize the model, including explained, explanatory, and control variables. Next, we provide details on the data sources and methods used to construct our panel dataset from 81 counties in China over the period 2014 to 2019. Finally, we present descriptive statistics of the variables used in our analysis. This section serves as the foundation for our empirical analysis.

3.1. Model

The incidence of poverty in counties is influenced not only by poverty-related allocations, public expenditures, the level of regional economic development, the level of capital and human capital alone but also by other factors. Therefore, based on the above model, relevant control variables are added to obtain the basic regression model as follows:

$$\mathit{ln}\left(1+{\theta }_{it}\right)=\alpha +\beta \mathit{ln}{E}_{it}+{\gamma }_i\mathit{ln}{Controls}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$ (6)

$$\mathit{ln}\left(1+{\theta }_{it}\right)=\alpha +\beta \mathit{ln}{G}_{it}+{\gamma }_i\mathit{ln}{Controls}_{it}+{\mu }_i+{\lambda }_t+{\epsilon }_{it}$$ (7)

where $\mathit{ln}\left(1+{\theta }_{it}\right)$ represents the logged poverty incidence in region $i$ in period $t$, $\mathit{ln}{E}_{it}$ represents the logged poverty-specific allocation in region $i$ in period $t$, $\mathit{ln}{G}_{it}$ represents the logged public expenditure in region $i$ in period $t$, $\mathit{ln}{Controls}_{it}$ represents the control variables in region $i$ in period $t$, and ${\epsilon }_{it}$ represents the disturbance term of the random effect.

Since the explanatory variable in this paper is poverty incidence, its indicator value is in the interval [0,1], and the Panel-Tobit model has the advantage of dealing with truncated data between [0,1], our method is applicable to the case where both continuous and discontinuous variables are included in the explanatory variables and can correctly reflect the true relationship between the variables. Moreover, using the least squares method to estimate the parameters of the model containing truncated data is biased, and the estimates are inconsistent. Therefore, this paper uses the Panel-Tobit model for empirical analysis. The expression of this estimation method is as follows:

$$\mathit{ln}\left(1+{\theta }_i\right)=\alpha +\beta \mathit{ln}{E}_i+{\gamma }_i\mathit{ln}{Controls}_i+{\tau }_i+{\epsilon }_i$$ (8)

$$\mathit{ln}\left(1+{\theta }_i\right)=\alpha +\beta \mathit{ln}{G}_i+{\gamma }_i\mathit{ln}{Controls}_i+{\tau }_i+{\epsilon }_i$$ (9)

where ${\theta }_i$ is the incidence of poverty or per capita disposable income of rural residents, $E_i$ stands for poverty-specific allocation, $G_i$ stands for public expenditure, and the control variables are selected in turn as the value added of primary industry, value added of the secondary industry, number of students in general primary and secondary schools, number of beds in medical and health institutions, and number of units in social service institutions.

To further investigate the poverty reduction pathways of poverty-specific allocations and public expenditures, a mediating effects model was established to explore the impact of primary sector development and secondary sector development on poverty reduction, and a mediating effects model was constructed as shown below:

$$m_1=\alpha +\beta \mathit{ln}{E}_i+{\gamma }_i\mathit{ln}{Controls}_i+{\tau }_i+{\epsilon }_i$$ (10)

$$\mathit{ln}\left(1+{\theta }_i\right)=\alpha +\beta \mathit{ln}{E}_i+\delta m_1+{\gamma }_i\mathit{ln}{Controls}_i+{\tau }_i+{\epsilon }_i$$ (11)

$$m_2=\alpha +\beta \mathit{ln}{G}_i+{\gamma }_i\mathit{ln}{Controls}_i+{\tau }_i+{\epsilon }_i$$ (12)

$$\mathit{ln}\left(1+{\theta }_i\right)=\alpha +\beta \mathit{ln}{G}_i+\delta m_2+{\gamma }_i\mathit{ln}{Controls}_i+{\tau }_i+{\epsilon }_i$$ (13)

where $m_1$ represents primary sector development and $m_2$ represents secondary sector development.

3.2. Variables

According to equation (5), poverty incidence is closely related to poverty-specific distribution, public expenditure, regional economic development level, capital level, and human capital. Therefore, poverty incidence is selected as an explanatory variable, special allocations for poverty and public expenditure as explanatory variables, and value added of primary industry, value added of secondary industry, number of students in general primary and secondary schools, number of beds in medical and health institutions, and number of units in social welfare institutions as control variables. The reasons for this selection are as follows.

• (1)Explained variables. In this paper, poverty incidence was chosen as the explanatory variable. Poverty incidence, also known as the poverty headcount index, is the number of rural residents below the poverty line relative to the agricultural population, that is, the ratio of the poor population to the agricultural population. Poverty is often understood as a lack of food, shelter, clothing, and other necessities of life—the failure to live a decent life. DiNitto (2007) defines people living below the minimum standard of living in the United States as poor. The U.S. federal government defines poverty by calculating the cash income required to meet the minimum standard of living each year, also known as the “poverty threshold.” According to the Chinese government's Opinions on Establishing a Poverty Exit Mechanism, China uses poverty incidence as the main indicator to measure whether poor areas and people have escaped poverty, and it is appropriate to use poverty incidence as the main indicator of the effect of poverty reduction. To ensure the robustness of the regression results, this paper also uses the disposable income of the rural population as a proxy indicator for poverty incidence. In general, the higher the disposable income of the rural population is, the higher the income level of local farmers and the less likely they are to experience poverty.
• (2)Explanatory variables. In this paper, poverty-related allocations and public expenditures are chosen as explanatory variables. The poverty-specific allocations mentioned in this paper refer to the special funds in the budgets of governments at all levels that are dedicated to helping poor regions eradicate poverty and are mainly used to help poor regions accelerate their economic and social development, improve the basic production and living conditions of poor people, and help poor people raise their income levels. Therefore, it is reasonable to choose poverty-specific allocations as the central explanatory variable. The public expenditures mentioned in this paper refer to the general public budget expenditures that are not poverty-related and can have a poverty-reducing effect by influencing the overall scale or structure of the economy.
• (3)Control variables. The literature review and the present results show that the level of regional industrial development, health care, social services, and educational development can influence local economic and social development. The more balanced local economic development and the more equitable social security, the lower regional poverty tends to be. Based on previous studies, taking into account the accessibility of data and the robustness of the conclusion, relevant indicators such as the value added of primary industry, the value added of secondary industry, the number of students in general primary and secondary schools, the number of beds in medical and health facilities, and the number of units in social service facilities are selected as control variables.
• (4)County type. For further analysis of the empirical results, counties in this paper are divided into two categories: poor counties and nonpoor counties. Poor counties, also referred to as national key counties for poverty alleviation or national-level poor counties, are identified based on a standard set in China for supporting poor areas, and the qualification of poor counties is recognized by the Office of the Leading Group for Poverty Alleviation and Development of the Chinese government. Counties are divided into pilot counties and nonpilot counties for the purpose of integrating funds. In 2016, to further increase the investment of financial funds in poverty alleviation and fully realize the leading role and aggregation effect of financial funds in reducing poverty, the Chinese government issued the document “Opinions on Supporting Poverty-stricken Counties to Carry Out Pilot Projects of Integrating and Using Financial Funds Involving Agriculture”, which suggests that county governments should integrate the use of various types of financial funds at all levels, especially for the development of agricultural production and rural infrastructure in poverty-stricken areas, to further improve the efficiency of the use of funds.

3.3. Data

In November 2013, Chinese President Xi Jinping first proposed “precise poverty reduction” during a visit to Hunan Province in China. Since then, the Chinese government and businesses have stepped up their support for rural poverty alleviation and achieved remarkable results, declaring the complete eradication of poverty by 2020. Therefore, taking into account the authenticity and availability of data, the data collection period was set to 2014 to 2019. The poverty incidence data of poor counties were obtained from the business management module of the China Poverty Alleviation and Development Information System, which mainly contains data on poverty incidence, rural population, the number of people who escaped poverty in that year, and the number of people who did not escape poverty. The data on poverty incidence of nonpoor counties are mainly collected manually. First, government work reports and statistical yearbooks of nonpoor counties are collected to determine the rural population and the number of people who escaped poverty in the current year to calculate poverty incidence. Second, a small part of the data comes from news reports on the status of poverty reduction in each county. The poverty-specific allocations come from the Poverty Alleviation Fund project data published on the official websites of provincial government finance departments and poverty alleviation departments. Public expenditure data are mainly from the China Financial Statistics Yearbook. The disposable income of the rural population, the value added of primary industry, the value added of secondary industry, the number of students in general primary and secondary schools, the number of beds in medical and health institutions, and the number of units in social welfare institutions are mainly from the statistical yearbooks of cities and towns, the annual library of counties in the China Economic Network, and the Easy Professional Superior database (EPSDATA) (see Table 1).

Table 1.

Summary of empirical variables.

Category	Variable name	Symbol	Processing method
Dependent variable	Poverty incidence	$\theta$	Poor population divided by agricultural population
	Disposable income per rural resident	$y$	/
Independent variable	Poverty-specific allocations	$x_1$	/
	Public finance expenditure	$x_2$	/
	Development level of primary industry	$m_1$	Primary industry value added
Control variables	Development level of secondary industry	$m_2$	Secondary industry value added
	Educational development level	$z_1$	Number of students in primary and secondary schools
	Health and medical level	$z_2$	Beds in medical and health institutions
	Social service capability	$z_3$	Number of social service units
County categories	Whether it is a poverty-stricken county	$Pkx$	$Pkx=1$ means the county is a poverty-stricken county
	Whether it is a fund integration pilot county	$Tczh$	$Tczh=1$ means the county is a fund integration pilot county

3.4. Descriptive statistics

Table 2 shows the results of the descriptive statistics. There are large regional differences in the poverty incidence of 81 counties in China from 2014 to 2019, with the maximum value of poverty incidence by logarithm being 0.136 and the minimum value of poverty incidence by logarithm being only 0.001. Therefore, considering the different levels of economic development and natural resource endowment in different regions, differentiated measures should be carried out in different regions to achieve better effectiveness with the appropriate means. The comparison of the mean values of poverty-specific allocations and public expenditure shows that the average amounts of poverty-specific allocations have increased from 2014 to 2019, indicating that the Chinese government has invested significant financial resources in poverty alleviation since China increased its attention to rural poverty alleviation and development in 2014. The variance and mean values show that the 81 counties are more balanced in terms of public spending, level of health care, and socialized service capacity, without much variation.

Table 2.

Descriptive statistics.

Variables	N	Mean	SD	Min	Median	Max
$\theta$	486	0.040	0.032	0.001	0.027	0.136
$x_1$	486	7.860	1.019	3.481	7.866	10.077
$x_2$	486	12.550	0.430	11.515	12.537	14.097
$z_1$	486	1.850	1.261	0.240	1.601	11.714
$z_2$	486	7.190	0.600	5.257	7.230	8.661
$z_3$	486	2.780	0.581	0.000	2.890	4.043
$m_1$	486	11.870	1.179	2.691	11.994	13.399
$m_2$	486	13.050	1.188	4.711	13.198	16.165

4. Empirical analysis and results

The previous sections of this paper have laid out the theoretical framework and econometric design for examining the relationship between fiscal spending and poverty alleviation in Chinese counties. In this section, we present the empirical analysis and results from our study. Specifically, we explore the impact of poverty-related allocations and public spending on poverty incidence, coupled with the role of primary and secondary industry development, health services, and education in poverty reduction. To ensure the validity and reliability of our findings, we employ baseline regression, mechanism tests, endogeneity analysis, robustness tests, and heterogeneity analysis. Through these rigorous empirical analyses, we seek to provide insights into the effectiveness of fiscal policies in promoting poverty alleviation and rural revitalization.

4.1. Baseline regression

The results of the panel unit root test show that all variables reject the original hypothesis of the presence of a unit root at the 10% significance level, that all variables are stable and that the regression analysis can be performed with the data from the original variables. Since the poverty incidence data are truncated data that lie in the interval [0,1], the estimation of model parameters containing truncated data using the least squares method is biased, and the estimates are inconsistent; therefore, this paper uses a Panel-Tobit model for the empirical analysis. First, the poverty-reducing effect of poverty-related allocations is analyzed. Second, the poverty-reducing effect of public expenditures is analyzed. To compare the empirical results and analyze the changes in the coefficients, control variables are added step by step. Table 3, Table 4 report the results of the stepwise regression for the poverty-reducing effect of poverty-related allocations and public spending, respectively.

Table 3.

Baseline regression results.

Dependent	$\theta$	$\theta$	$\theta$	$\theta$
Method	Panel Tobit	Panel Tobit	Panel Tobit	Panel Tobit
$x_1$	−0.023*** (−14.87)	−0.022*** (−13.51)	−0.014*** (−7.23)	−0.014*** (−7.25)
$z_1$		−0.022 (−1.53)	−0.022 (−1.60)	−0.020 (−1.49)
$z_2$			−0.037*** (−7.57)	−0.037*** (−7.45)
$z_3$				0.008** (2.18)
_cons	0.259*** (16.77)	0.283*** (12.95)	0.474*** (14.53)	0.447*** (12.79)
${\tau }_i$	Yes	Yes	Yes	Yes
N	486	486	486	486

Notes: The parentheses in the table are z-statistic values, and ***, **, and * indicate passing significance tests at the 1%, 5%, and 10% levels, respectively.

Table 4.

Baseline regression results.

Dependent	$\theta$	$\theta$	$\theta$	$\theta$
Method	Panel Tobit	Panel Tobit	Panel Tobit	Panel Tobit
$x_2$	−0.079*** (−23.24)	−0.082*** (−22.06)	−0.075*** (−14.84)	−0.075*** (−14.77)
$z_1$		0.023* (1.88)	0.020* (1.65)	0.021* (1.71)
$z_2$			−0.009* (−1.96)	−0.009* (−1.92)
$z_3$				0.006* (1.79)
_cons	1.036*** (24.34)	1.039*** (24.48)	1.026*** (23.97)	1.003*** (22.50)
${\tau }_i$	Yes	Yes	Yes	Yes
N	486	486	486	486

Notes: The parentheses in the table are t-statistics or z-statistics, and ***, **, and * indicate passing significance tests at the 1%, 5%, and 10% levels, respectively.

The results in Table 3, Table 4 show that the coefficients of the main explanatory variables are significant, which is consistent with the results of the theoretical model and further supports the reliability of the results. In addition, the trend of the coefficient changes is consistent with the inclusion of the control variables, which initially suggests that our model is more robust. The baseline regression results suggest that (1) an increase in poverty-specific allocations contributes to a reduction in poverty incidence (coefficient = −0.014, p < 0.01). The possible reason is that increasing poverty-specific allocations will help improve agricultural production conditions, improve rural infrastructure, and promote agricultural science and technological innovation, effectively increasing the efficiency of agricultural production, expanding aggregate social demand, improving production and living conditions in poor rural areas, and further raising the income level of the poor, effectively reducing the incidence of poverty. (2) An increase in public spending contributes to the reduction of poverty incidence (coefficient = −0.075, p < 0.01). The likely reason is that government public spending significantly increases investment in education, health, water supply, roads, and other rural infrastructure through purchase spending or fiscal transfers, which promotes economic growth, improves income distribution inequality, reduces the income gap in poor areas, and gradually lifts the rural poor out of poverty.

(3) Improving medical and health care contributes to reducing poverty incidence (coefficient = −0.037, p < 0.01). Good health generally means higher labor productivity and income, while deteriorating health not only reduces labor productivity but also increases household expenses and even leads to poverty because people are unable to meet their medical expenses and make ends meet. Good medical health status helps to improve the labor productivity of the poor, increase labor income, and reduce household expenditure on serious diseases, thereby reducing the incidence of poverty. (4) The direction of the impact of educational attainment on poverty incidence is uncertain. One possible reason is that although higher educational attainment makes it easier for children of poor families to obtain a better education and complete a good education, thereby strengthening the endogenous motivations of poor families, short-term education does not seem to have a significant poverty-reducing effect because of the long cycle of rewards acquired through education.

4.2. Mechanism test

As China is a large agricultural country with a high proportion of traditional agriculture, the structure of the agricultural industry has not been upgraded and optimized compared with secondary and tertiary industries, which lag behind in development. The lack of large enterprises and brands operating in the characteristic beneficial industries and the backward development of the circular economy for agricultural products have seriously prevented the agribusiness industry from becoming larger and stronger. Most poverty-related allocations and public expenditures are related to industrial infrastructure construction, industry support, poverty alleviation workshops, cooperatives, industrial enterprise support, etc. The government generally introduces capital, technology and market elements through various industrial policies to promote enterprises in poor areas and push forward the development of related industrial supply chains to encourage poor people to increase their own development momentum. On this basis, this paper argues that poverty-related allocations and public expenditures have differential effects on poverty incidence by affecting the development levels of primary and secondary industries. This section discusses the effects of primary and secondary industry development levels on poverty incidence separately, using these levels as mechanism variables.

Table 5 shows the empirical results on how the mediating effects of poverty reduction on poverty incidence affect the level of development of the primary and secondary sectors. Since the level of primary and secondary sector development and poverty bias are continuous variables, the regression is run using a fixed effects model when testing the mechanism. The results of using the level of primary or secondary sector development as an explanatory variable show that there is a positive relationship between poverty-specific allocations and the level of primary or secondary sector development (coefficient = 0.125, p < 0.01; coefficient = 0.142, p < 0.01), suggesting that the special allocation for poverty can positively promote the level of primary and secondary sector development. The coefficient on poverty-specific allocation remains significant when the level of primary or secondary industry development and poverty-specific allocation are simultaneously included in the regression equation (coefficient = −0.010, p < 0.01). The results of the Sobel and bootstrap tests suggest that both primary and secondary industry development levels satisfy the condition of a mediating variable. In other words, poverty-specific allocations can affect poverty incidence through the level of primary or secondary industry development, and primary industry development is more conducive to reducing poverty incidence. There are two possible reasons why the level of primary and secondary industry development affects poverty incidence: on the one hand, primary industry development can create more employment opportunities for the rural poor and increase their income to reduce relative poverty in rural areas. On the other hand, the development of primary industry can help promote the development of industrial integration in rural areas, and primary, secondary and tertiary industries can integrate and complement each other, extending the industrial chain, increasing the added value of agricultural products, promoting local economic development, and thus reducing the incidence of poverty.

Table 5.

Estimation results of the mediating effects of ×1 and $\theta$.

Dependent	$m_1$	$\theta$	$m_2$	$\theta$
Method	FE	Panel Tobit	FE	Panel Tobit
$x_1$	0.125*** (10.22)	−0.010*** (−4.80)	0.142*** (11.81)	−0.008*** (−3.66)
$m_1$	/	−0.029*** (−3.86)	/	/
$m_2$	/	/	/	−0.042*** (−5.46)
Control variables	Yes	Yes	Yes	Yes
${\tau }_i$	Yes	Yes	Yes	Yes
N	486	486	486	486
R2	0.874		0.883

Notes: The parentheses in the table are t-statistics or z-statistics, and ***, **, and * indicate passing significance tests at the 1%, 5%, and 10% levels, respectively.

Table 6 shows the empirical results of the mediating effect of public expenditure on poverty incidence by influencing the levels of primary and secondary sector development. Similar to the results in Table 5, the results of using the levels of primary or secondary sector development as the explanatory variables show that there is a positive relationship between public expenditure and the levels of primary or secondary sector development (coefficient = 0.393, p < 0.01; coefficient = 0.534, p < 0.01), indicating that public expenditure can positively promote the levels of primary and secondary sector development. The coefficient of public expenditure is still significant (coefficient = −0.010, p < 0.01) after adding the levels of development of both primary and secondary industry and public expenditure to the regression equation, but the coefficient of the levels of development of primary or secondary industry are no longer significant. Meanwhile, the results of the Sobel test and bootstrap test indicate that the levels of development of primary and secondary industry show weakly mediating variable results. In other words, public expenditure does not necessarily directly affect the incidence of poverty through the levels of primary or secondary industry development. The possible reason is that China's public expenditures can be divided into capital construction expenditures; support for agriculture, science, education, and health expenditures; pension and social welfare relief expenditures; national defense expenditures; administrative expenses; and foreign aid expenditures, and not all of these public expenditures can reduce poverty. Some studies show that the poverty reduction effect of public expenditures on agriculture is the most significant, followed by social relief public expenditures and infrastructure public expenditures, while the poverty reduction effects of public expenditures on science and technology, education, culture, and health are not significant.

Table 6.

Estimation results of the mediating effects of $x_2$ and $\theta$.

Dependent	$m_1$	$\theta$	$m_2$	$\theta$
Method	FE	Panel Tobit	FE	Panel Tobit
$x_2$	0.393*** (10.59)	−0.073*** (−12.68)	0.534*** (16.05)	−0.074*** (−11.37)
$m_1$	/	−0.006 (−0.83)	/	/
$m_2$	/	/	/	−0.002 (−0.25)
Control variables	Yes	Yes	Yes	Yes
${\tau }_i$	Yes	Yes	Yes	Yes
N	486	486	486	486
R2	0.925		0.942

Notes: The parentheses in the table are t-statistics or z-statistics, and ***, **, and * indicate passing significance tests at the 1%, 5%, and 10% levels, respectively.

4.3. Endogeneity

According to the basic regression setting in this paper, there are three possible sources of endogeneity: first, omitted variables. Although the model includes poverty incidence, poverty-specific allocation, education development, health care, and social service levels in the research framework, important variables are still inevitably omitted, and the resulting endogeneity problem introduces some bias into the estimation results. Second, there is reverse causality. The faster the incidence of poverty falls, the more likely a region is to receive more poverty-related allocations or public spending for its good work, further reducing the incidence of poverty. Third, measurement error should be taken into account. Since the data come from county statistical yearbooks, government work reports, and Easy Professional Superior database (EPSDATA), their statistical standards and statistical caliber may have some variation, which in turn leads to data measurement errors and thus endogeneity problems. For this reason, it is necessary to test and edit for endogeneity to obtain robust estimation results.

Although this paper uses a fixed-effects model that controls for individual effects, the effects of endogeneity on the results cannot be completely ruled out. The instrumental variable Tobit method (IVT) and the two-stage method are two accepted methods for testing the endogeneity of Tobit models[[32], [33], [34]]. To reduce the undesirable effects of endogeneity and further increase the reliability of the results, two instrumental variables were selected for endogeneity testing in this work. The first was selected as the lagged variable of poverty-specific allocations and public expenditures, and the second was selected as the average poverty incidence in all counties except this region. In Table 7, lagged and instrumental variables are used to construct IV-Tobit models to test and eliminate endogeneity problems and reduce estimation bias due to omitted variables, reverse causality, and measurement error.

Table 7.

Estimation results of the endogeneity test.

Dependent	$\theta$	$\theta$	$\theta$	$\theta$
Method	IV-Tobit	IV-Tobit	IV-Tobit	IV-Tobit
$x_1$	−0.047*** (−10.33)		−0.051*** (−5.29)
$x_2$		−0.060*** (−9.88)		−0.107*** (−16.89)
Control variables	Yes	Yes	Yes	Yes
${\tau }_i$	Yes	Yes	Yes	Yes
N	324	405	405	405
Amemiya-Lee-Newey minimum chi-square statistic	2.253	3.698	0.087	0.170
AR	35.60***	99.59***	160.60***	145.07***
Wald test	34.70***	97.61***	28.01***	65.76***

Notes: (1) This table reports the endogeneity test results. (2) *, **, and *** represent 10%, 5%, and 1% significance levels, respectively. (3) The Amemiya-Lee-Newey minimum chi-square statistic "/" indicates that the number of instrumental variables is comparable to the number of endogenous variables that are exactly identified.

The estimation results of the endogeneity tests are shown in Table 7. In the results of the validity tests of the instrumental variables, the F-statistics are all much larger than 10, which means that the instrumental variables are selected to meet the correlation requirement. The results of AR tests and Wald tests show that the instrumental variables are not weak, and from the Amemiya-Lee-Newey results, it is clear that all instrumental variables are exogenous. The results in Table 7 show that the coefficients and significance of poverty incidence did not change significantly, and the endogeneity problem did not fundamentally disturb the basic results of this paper.

4.4. Robustness tests

In empirical analysis, bias or error may occur in the selection of measures or measurement variables. Therefore, to ensure the robustness of the empirical evidence, we adopted alternative variables, substitution of empirical methods, and selection of subsamples for robustness testing.

First, we replaced the explanatory variable of poverty incidence. The disposable income of rural residents in the empirical analysis can reflect the income levels of local rural residents, and it is more reasonable to use the disposable income of rural residents to replace the incidence of poverty. Since rural residents' disposable income and poverty-specific allocations and public expenditures are continuous variables, the regression results are re-estimated using a fixed-effects model. The results are shown in Table 8. It can be seen that poverty-specific allocations and public expenditures make a positive contribution to rural residents' disposable income, and the incidence of poverty will gradually decrease as rural residents' disposable income increases, which is consistent with the original empirical findings and further verifies the robustness of the regression results.

Table 8.

Robustness analysis of alternative regression methods.

Dependent	${\theta }^{\prime }$	${\theta }^{\prime }$	${\theta }^{\prime }$	${\theta }^{\prime }$
Method	FE	FE	FE	FE
$x_1$	0.275*** (23.77)	0.181*** (13.71)
$x_2$			0.831*** (35.22)	0.695*** (20.26)
Control variables	No	Yes	No	Yes
${\tau }_i$	Yes	Yes	Yes	Yes
N	486	486	486	486
$R^2$	0.852	0.888	0.913	0.919

Notes: The parentheses in the table are t-statistics or z-statistics, and ***, **, and * indicate passing significance tests at the 1%, 5%, and 10% levels, respectively.

Second, we replaced the regression methods. In this paper, the panel probit model and Tobit model are used in turn for the regressions, as shown in Table 9. The regression results show that both poverty-specific allocations and public expenditures have a significant negative effect on poverty incidence, which is consistent with the baseline regression findings and further verifies the robustness of the regression results.

Table 9.

Robustness analysis of alternative regression methods.

Dependent	$\theta$	$\theta$	$\theta$	$\theta$
Method	Panel Probit	Panel Probit	Tobit	Tobit
$x_1$	−0.098*** (−5.26)		−0.014*** (−7.25)
$x_2$		−0.884*** (−10.82)		−0.075*** (−14.77)
Control variables	Yes	Yes	Yes	Yes
${\tau }_i$	/	/	Yes	Yes
N	486	486	486	486

Notes: The parentheses in the table are t-statistics or z-statistics, and ***, **, and * indicate passing significance tests at the 1%, 5%, and 10% levels, respectively.

Finally, we replaced the sampling intervals. In this work, we excluded the 2014, 2016, and 2019 data sequentially before regressing the subsample, as shown in Table 10. The coefficients of the regression results changed, but the significance and negative effects did not change, which is consistent with the baseline regression results and again confirms the robustness of the regression results.

Table 10.

Robustness analysis of subsample regression.

Dependent	$\theta$	$\theta$	$\theta$	$\theta$	$\theta$	$\theta$
Method	Panel Tobit	Panel Tobit	Panel Tobit	Panel Tobit	Panel Tobit	Panel Tobit
Sample	Year ≠ 2014	Year ≠ 2016	Year ≠ 2019	Year ≠ 2014	Year ≠ 2016	Year ≠ 2019
$x_1$	−0.009*** (−4.21)	−0.011*** (−4.91)	−0.011*** (−5.65)
$x_2$				−0.064*** (−12.60)	−0.073*** (−13.23)	−0.072*** (−14.59)
Control variables	Yes	Yes	Yes	Yes	Yes	Yes
${\tau }_i$	Yes	Yes	Yes	Yes	Yes	Yes
N	405	405	405	405	405	405

Notes: The parentheses in the table are t-statistics or z-statistics, and ***, **, and * indicate passing significance tests at the 1%, 5%, and 10% levels, respectively.

4.5. Heterogeneity analysis

The various regions of China differ significantly in terms of the level of economic development, natural resource endowment, and institutional environment. The counties are further divided into two types: poor counties vs. nonpoor counties and pilot counties for capital integration vs. nonpilot counties for capital integration. From the regression results in columns (1) through (4) in Table 11, we find that both poverty-specific allocations and public spending have a significant inhibitory effect on poverty incidence depending on whether they are poor counties. A comparison of the empirical results for poor and nonpoor counties shows that the negative effect of poverty-specific allocations on poverty incidence is stronger in the sample of poor counties than in the sample of nonpoor counties, while the negative effect of public spending on poverty incidence is also stronger in the sample of poor counties than in the sample of nonpoor counties. As in the baseline regression, there is a significant negative relationship between poverty-related allocations and public spending on poverty incidence. One possible reason for this is that, in general, the larger the amount of poverty-related funding that poor counties receive, the more projects and funding that can be allocated to the poor, and the larger their effect on reducing poverty incidence.

Table 11.

Estimation results of heterogeneity analysis.

Dependent	(1)	(2)	(3)	(4)
method	Panel Tobit	Panel Tobit	Panel Tobit	Panel Tobit
style	$Pkx=1$	$Pkx=1$	$Pkx=0$	$Pkx=0$
×1	−0.056*** (−12.11)		−0.005*** (−4.87)
×2		−0.142*** (−16.09)		−0.033*** (−10.72)
Control variables	Yes	Yes	Yes	Yes
${\tau }_i$	Yes	Yes	Yes	Yes
N	150	150	336	336

Notes: The parentheses in the table are t-statistics or z-statistics, and ***, **, and * indicate passing significance tests at the 1%, 5%, and 10% levels, respectively.

From the regression results in columns (1) to (4) in Table 12, it can be seen that the coefficient of poverty-specific allocation in the capital integration pilot counties (coefficient = −0.031, p < 0.01) is larger than the coefficient of poverty-specific allocation in the noncapital integration pilot counties (coefficient = −0.006, p < 0.01), indicating that the implementation of the capital integration pilot policy will facilitate the reduction of poverty incidence. The public expenditure coefficient (coefficient = −0.097, p < 0.01) is also greater in the capital integration pilot counties than in the noncapital integration pilot counties (coefficient = −0.030, p < 0.01). A possible explanation is that, compared with the nonfund integration pilot counties, the fund integration pilot counties are able to unify the use of various types of poverty-specific allocations and focus on agricultural production development and rural infrastructure construction, which helps the poverty-specific allocations to be effective.

Table 12.

Estimation results of heterogeneity analysis.

Dependent	(1)	(2)	(3)	(4)
method	Panel Tobit	Panel Tobit	Panel Tobit	Panel Tobit
style	$Tczh=1$	$Tczh=1$	$Tczh=0$	$Tczh=0$
×1	−0.031*** (−9.48)		−0.006*** (−5.39)
×2		−0.097*** (−14.23)		−0.030*** (−7.84)
Control variables	Yes	Yes	Yes	Yes
${\tau }_i$	Yes	Yes	Yes	Yes
N	300	300	186	186

Notes: The parentheses in the table are t-statistics or z-statistics, and ***, **, and * indicate passing significance tests at the 1%, 5%, and 10% levels, respectively.

5. Conclusions and policy recommendations

5.1. Conclusions

This paper consolidates the robustness of the results by constructing a theoretical model of the government's poverty reduction performance function using county-level panel data for 81 counties in China from 2014 to 2019, a Panel-Tobit model and mechanism tests, and a series of extensions such as endogeneity treatment, robustness tests, and heterogeneity analysis. This leads to the following results.

First, poverty reduction can reduce poverty incidence. The empirical results show that earmarking funds have a significant negative effect on poverty incidence. The results of the robustness test and heterogeneity analysis indicate that the negative effect of poverty-specific allocations on poverty incidence is stronger in poor counties and pilot counties with funding integration than in nonpoor counties and pilot counties without funding integration.

Next, public spending can reduce the incidence of poverty. The empirical results show that public spending has a significant negative effect on poverty incidence. The results of robustness tests and heterogeneity analysis show that the negative effect of public spending on poverty incidence is also stronger in poor and capital-integrated pilot counties than in nonpoor and noncapital-integrated pilot counties.

Second, primary and secondary industry development can reduce poverty incidence. Based on the results of the mechanism test, this paper concludes that poverty-related allocations can negatively affect poverty incidence through the level of primary and secondary industry development and that the poverty reduction effect of primary industry development is higher than that of secondary industry development. In addition, public spending does not necessarily have a direct effect on poverty incidence through primary and secondary industry development.

Finally, improving medical care can help reduce poverty incidence. Good medical health helps improve labor productivity and income among the poor, which in turn reduces poverty incidence. In contrast, the impact of education development levels on poverty incidence is unclear, mainly because the cycle of returns to education is long and short-term education does not effectively reduce poverty.

5.2. Policy recommendations

Based on the above findings, this paper makes several policy recommendations. First, the scale of poverty-related allocations and public spending should be increased. Investment in poverty-related allocations and public spending should be continuously increased, the structure of financial stock should be optimized, poverty reduction should be made a key area of financial investment protection, and the share of poverty-related allocations and public spending should be further increased to achieve sustainable and stable growth in poverty-related allocations and public spending. At the same time, social capital should be actively encouraged to participate in poverty alleviation, and financial resources should be used as levers to encourage social and financial capital to further participate in poverty alleviation to realize financial and fiscal linkage mechanisms, and build a capital investment guarantee mechanism that is compatible with the task of poverty eradication.

Second, industry support should be strengthened. The empirical results show that both primary and secondary industry development can effectively reduce poverty incidence. Relying on the advantages of local natural resources and ecological resources, develop suitable special agricultural industries, and gradually form a more complete industrial system on this basis. Appropriately expand the scale of industrial development support funds, increase support for the whole industrial chain of modern agriculture and the integration development of agricultural and rural industries, and promote the development of a diversified rural economy. The development of processing enterprises of special industries should be supported to create an innovative development model of regional special industries in terms of scale, intensification, greening, upgrading, branding and whole industrial chains.

Third, differentiated policy initiatives should be implemented according to local conditions. The empirical results show that the poverty-reducing effects of poverty-related allocations and public spending are stronger in poor counties and capital-integrated pilot counties than in nonpoor counties and noncapital-integrated pilot counties. It is recommended that while maintaining policy continuity and stability, local governments should relax the scope of appropriations according to the effects of policy implementation, optimize the allocation of funds to agriculture, and focus funds on important issues.

Funding information

This work was supported by the National Natural Science Foundation of China (Grant Nos. 72063018 and 72163015), the Programming Project of Jiangxi Province Social Science (Grant No. 22YJ06), the earmarked fund for the Jiangxi Agriculture Research System (Grant No. JXARS-08), and Jiangxi Postgraduate Innovation Special Fund Project (Grant No. YC2022-B033).

Author contribution statement

Changsong Wang: Conceived and designed the experiments; Performed the experiments; Analyzed and interpreted the data; Wrote the paper.

Xihui Chen: Contributed reagents, materials, analysis tools or data; Wrote the paper.

Jin Hu: Performed the experiments; Analyzed and interpreted the data; Wrote the paper.

Muhammad Shahid: Conceived and designed the experiments; Analyzed and interpreted the data; Wrote the paper.

Data availability statement

Data will be made available on request.

Declaration of competing interest

The authors declare that there are no conflict of interests, we do not have any possible conflicts of interest.