Study on the Relationship between Agricultural Credit, Fiscal Support, and Farmers’ Income—Empirical Analysis Based on the PVAR Model

Abstract

The growth of farmers’ income is one of the most critical issues in China’s “Three Rural Issues,” and optimizing fiscal policy support and improving credit supply are crucial to improving farmers’ income. Based on the panel data of 30 Chinese provinces from 2003 to 2020, this paper develops a PVAR model in order to explore the relationship between agricultural credit, fiscal support for agriculture, and farmers’ income from a dynamic perspective, considering regional heterogeneity. The empirical results show the following factors for farmers’ income growth: (1) From the GMM estimation, the positive correlation between fiscal support for agriculture is stronger than that of agricultural credit. (2) From the impulse-response function, in the eastern region, the positive shock of agricultural credit is positively correlated in the short run, but it will be negatively correlated as that of fiscal support for agriculture in the long run; in the central region, the positive shocks of agricultural credit and fiscal support for agriculture are persistently positively correlated; in the western region, the positive shocks of agricultural credit are persistently negatively correlated, while fiscal support for agriculture will be positively correlated in contrast. (3) From the variance decomposition, agricultural credit contributes more to famer’s income growth in the short run, while fiscal support for agriculture contributes more in the long run. The policy implications for promoting farmers’ income growth include implementing regionally differentiated agricultural credit development strategies, reasonably enhancing fiscal support for agriculture, and optimizing the structure of fiscal support for agriculture.

1. Introduction

Increasing the income of rural residents is not only an essential part of the strategy to promote rural revitalization, but it is also a necessary way to build a new development pattern that promotes both the domestic and international production cycles. Since 2004, China has issued the No. 1 Central Document with the theme of the “three rural issues” (which are issues relating to agriculture, rural areas, and farmers) for 19 consecutive years. In 2019, at the 13th National People’s Congress, Premier Li Keqiang pointed out, for the first time in a government work report, that “we will continue to prioritize the development of agriculture and rural areas and strengthen both poverty eradication and rural revitalization.” In 2022, General Secretary Xi Jinping further made the statement to “comprehensively promote rural revitalization and adhere to the priority development of agriculture and rural areas,” in the report of the 20th National Congress of the Communist Party of China. In order to solve the problem of the “three rural issues” and realize rural revitalization, the first task is to solve the problem of capital sources and improve the efficiency of capital use. Agricultural credit and fiscal support for agriculture, as the primary sources of rural development funds, play a pivotal role in developing the rural economy and improving farmers’ income. At the national level, agriculture-related loans increased from CNY 777.8 billion in 2003 to CNY 33,870.9 billion in 2020, and the fiscal expenditure on agriculture, forestry, and water affairs increased from CNY 57.4 billion in 2003 to CNY 2214.6 billion in 2020. With the continuous promotion of rural revitalization strategies, agricultural credit and fiscal support for agriculture will be further improved, and improving the efficiency of these two funding mechanisms and guaranteeing the stable growth of farmers’ income will become important issues. Exploring the correlation and the regional heterogeneity between agricultural credit, fiscal support for agriculture, and farmers’ income increase will help us to optimize the arrangement of the relevant fiscal and credit funds and will help us to improve the efficiency of their use. This has vital theoretical and practical significance for establishing a high-quality mechanism to improve farmers’ income and realizing the rural revitalization strategy.

There are two main views on the relationship between agricultural credit and rural people’s income. The first view is that there is a positive relationship between agricultural credit and farmers’ income growth. Li (1998) and Hassan et al. (2011) argued that rural financial development is an important way to promote farmers’ income growth [1,2]. Poliquit (2006) and Ololade et al. (2013) stated that agricultural credit is essential for agricultural development, poverty reduction, and farmers’ income increase [3,4]. Wang Qian et al. (2021) concluded that the coordinated development of agricultural insurance, agricultural credit, and farmers’ income is beneficial to the growth of farmers’ income, based on the panel data of 25 Chinese provinces from 2009 to 2018 [5]. Through empirical tests, Chen Xinzhong (2021) and Liu Saihong et al. (2021) reached a similar conclusion that agricultural credit can alleviate farmers’ financial shortage constraints, increase farmers’ income levels, and contribute to regional poverty alleviation [6,7]. The second view is that there is a negative relationship between agricultural credit and farmers’ income increase. Machethe (2004) pointed out that the credit that is allocated through institutions in most developing countries has no significant positive correlation with the growth and income enhancement of rural economies [8]. Domestic scholars have reached a similar view. In the short term, the increase in agricultural credit inputs in the periphery is detrimental to local farmers’ income growth, and in the long term, the inhibitory effect of credit inputs on the growth of rural residents’ wage income becomes increasingly evident [9,10,11].

There are two main views on the relationship between fiscal support for agriculture and rural residents’ income. The first view is that there is a positive relationship between fiscal support for agriculture and farmers’ income growth. Scholars such as Borro (1981) and Maitra (2017) have demonstrated through empirical tests that fiscal support for agriculture is beneficial to the agricultural economy and increases farmers’ income [12,13]. Based on the national data of China from 1985 to 2015, Zhang Xiaohan et al. (2018) concluded that fiscal agricultural expenditure has a long-term positive correlation with the growth of farmers’ wage income, household business income, and property income [14]. Other domestic scholars have also concluded that fiscal support for agriculture and rural residents’ income show a positive relationship, based on data from provinces and cities in China [15,16,17]. The second view is that there is a negative relationship between fiscal support for agriculture and farmers’ income. Feng Mengli et al. (2020), based on a comparative study of the role relationship between different forms of fiscal support and farmers’ income, concluded that farmers’ subsidy funds have a negative relationship with farmers’ income level [18]. By comparing non-poor counties with similar levels of initial economic development, Zhou Minhui et al. (2016) found that there was no significant positive correlation between financial inputs and the net per capita income of farmers in poor counties [19]. Chen et al. (2019) used Chinese family panel studies (CFPS) data to obtain a similar conclusion that fiscal transfers are not effective in reducing poverty [20]. In summary, the literature has laid a foundation for research on the relationship between agricultural credit, fiscal support, and farmers’ income; however, there are still three shortcomings. First, it is still controversial as to whether the correlation between agricultural credit, fiscal expenditure, and farmers’ income is positive or negative over different periods, and more empirical evidence needs to be found in order to examine such questions. Second, most of the studies on farmers’ income are based on national or provincial perspectives. However, there are few studies on the correlation between agricultural credit, fiscal support, and farmers’ income in different economic regions. In addition, most studies choose only one of the two variables, agricultural credit or fiscal support to farmers, in order to analyze the relationship with farmers’ income, and they lack a dynamic perspective of the relationship of the interaction between the three. Therefore, this paper attempts to fill the gap in the existing studies, and its possible marginal contribution lies in the following two points: First, based on the interprovincial panel data of China from 2003 to 2020, this paper reveals the regional heterogeneity of the correlation between agricultural credit, fiscal support, and farmers’ income from the perspective of economic regions. Second, this paper treats agricultural credit, fiscal support, and farmers’ income as endogenous variables and uses impulse response and variance decomposition methods based on the establishment of the PVAR model to explore the relationship among them from a dynamic perspective, which provides theoretical references and practical suggestions for improving the effective utilization rate of financial and fiscal funds and promoting farmers’ income growth.

The remainder of this paper is organized as follows: Section 2 provides a literature review and the research hypotheses. Section 3 outlines the research design, and Section 4 discusses the empirical analysis. Section 5 provides the conclusion and recommendations.

2. Theoretical Analysis and Research Hypotheses

2.1. A study on the Correlation between Agricultural Credit and Farmers’ Income Growth

Academics usually argue that financial capital plays a central role in economic growth in developing countries and facilitates the equitable distribution of income [21,22,23], and this conclusion also applies to the rural economy. First, agricultural credit enhances household productivity and profitability by alleviating the subsistence formal credit constraints of farm households and is particularly effective in raising the income of rural low-income groups [24]. Second, the main source of farmers’ income is agricultural output. When farmers’ own savings cannot meet the financial needs of agricultural production, timely access to credit can help them to purchase the needed production and business materials that are more conducive to agricultural production activities [25,26]. Therefore, there is a positive correlation between the increase in the scale of agricultural credit and the agricultural output [27], which in turn enhances farmers’ income. With the increase in farmers’ income, path dependence and behavioral inertia increase farmers’ enthusiasm for production and can lead them to expand the scale of production and increase the demand for capital loans, thus forming a positive cycle for increasing farmers’ income. In addition, financial capital is an important prerequisite for entrepreneurship, and agricultural credit improves the availability of financial capital for rural residents and promotes the equalization of entrepreneurial opportunities, which in turn improves urban–rural income inequality [28,29]. Based on the above analysis, this paper proposes the following hypothesis:

Hypothesis 1. 

Agricultural credit inputs are positively associated with farmers’ income.

2.2. The Correlation between Fiscal Support for Agriculture and Farmers’ Income Growth

Fiscal participation is not only an important method to solve the current poverty alleviation funding, but it is also an important guarantee of the timely completion of the tasks of poverty alleviation [30]. The level of fiscal support for agriculture directly and indirectly increases the income level of rural residents. For the direct effect, the expenditure on agriculture, forestry, and water affairs has a function that is similar to a blood transfusion for the rural economy, and it can reduce the production cost of farmers, promote the increase in farmers’ wage income, and transfer income through the direct allocation of economic resources, such as agricultural production subsidies [31,32]. The indirect effect is an extension of the agricultural industry chain by improving the agricultural production conditions and supporting agricultural industrialization projects. Fiscal support can promote the cross-development of rural industries and the transformation and upgrade of traditional agriculture. In addition, the investment of government fiscal support in agriculture can lead to more financial and social capital participation [33], the improvement of the financial service environment for rural residents, the effective support of financial inclusion, and the broadening of channels of farmers’ capital sources, which in turn increases farmers’ income. Based on the above analysis, this paper proposes the following hypothesis:

Hypothesis 2. 

Fiscal inputs to support agriculture are positively associated with farmers’ income.

3. Research Design

3.1. Model Setting

The PVAR model uses panel data and is based on multiple endogenous variables, which can truly reflect the interaction between the variables. It has the advantages of the traditional VAR model; that is, all of the variables are considered to be endogenous variables. Moreover, it makes up for the shortcomings of the VAR model’s inability to handle cross-sectional data and satisfies the requirement for high-performance time series [34]. Therefore, this paper refers to the model settings of related studies [35,36,37]. The agricultural credit (LOAN), the fiscal support for agriculture (SUP), and the rural residents’ income (NI) are selected as the endogenous variables of the system, and a PVAR model is constructed in order to study the relationship between agricultural credit, fiscal support for agriculture, and rural residents’ income. The specific form of the model is as follows:

$$\begin{gathered} Y_{it} = {\alpha }_i + {\gamma }_t + {\displaystyle \sum _i^D}{\beta }_j{Y}_{i,t-j} + {\mu }_{it} \\ {Y}_{it} = [Ln{\mathrm{NI}}_{it},Ln{\mathrm{LOAN}}_{it},Ln{\mathrm{SUP}}_{it}] \end{gathered}$$

where j is the lag order, i is the individual province, t is the year, ${\beta }_j$ is the coefficient matrix, ${\alpha }_i$ is the individual fixed effect, ${\gamma }_t$ is the time fixed effect, ${\mu }_{it}$ is the random error term, and NI, LOAN, and SUP denote the income of rural residents, the level of agricultural credit, and the level of fiscal support for agriculture, respectively.

3.2. Variable Selection and Data Description

Due to the change in the statistical caliber of the fiscal support data for agriculture after 2002, and considering the comparability and completeness of the data, this paper takes the data of 30 Chinese provinces (excluding China, Hong Kong, Macao, Taiwan, and Tibet areas) from 2003 to 2020 as the sample, i.e., a total of 540 samples are collected. Among them, each province has three variables, and each variable has 18 years of time-series data, i.e., there are 54 data points in total for each province and a total of 1620 observations in the full text. On this basis, and considering the interregional differences, the samples are grouped and analyzed according to the three major regions of Eastern, Central, and Western China. The subgroups are based on the regional divisions of the National Bureau of Statistics of the People’s Republic of China (NBSPRC), as shown in

Table 1. Regional division.

Region	Included Provinces
Eastern	Beijing, Tianjin, Hebei, Liaoning, Shanghai, Jiangsu, Zhejiang, Fujian, Shandong, Guangdong, Hainan
Central	Shanxi, Jilin, Heilongjiang, Anhui, Jiangxi, Henan, Hubei, Hunan
Western	Inner Mongolia, Guangxi, Chongqing, Sichuan, Guizhou, Yunnan, Shaanxi, Gansu, Qinghai, Ningxia, Xinjiang

Note: Table 1 is based on the regional division standard of the National Bureau of Statistics of China, which groups the 30 provinces in China according to three major regions: East contains 11 provinces, Central contains 8 provinces, and West contains 11 provinces.

.

In this paper, the rural residents’ income (NI), the agricultural credit (LOAN), and the fiscal support for agriculture (SUP) are selected as variables, as shown in

Table 2. Variable settings and descriptions.

Variables	Variable Symbols	Description	Unit
Income of rural residents	NI	Per capita disposable income of rural residents	Yuan/person
Agricultural Credit	LOAN	Agricultural-related loans/primary industry output	%
Fiscal support for agriculture	SUP	Agriculture, forestry, and water budget expenditures/general public budget expenditures	%

Note: Table 2 shows the settings and descriptions of the variables. The first column is the variable name, the second column is the variable symbol, the third column is the specific description, and the fourth column is the variable unit.

. Among them, the rural residents’ disposable income (NI) is usually measured by the per capita disposable income of the farmers. The specific calculation formula refers to the measurement of the National Bureau of Statistics of China, namely, the per capita disposable income of the farmers = (total income of rural households−household operating expenses−tax expenses−depreciation of productive fixed assets−survey subsidies–expenditure for giving gifts to friends and relatives within rural areas−transfer expenditures−property expenditures)/permanent resident population of rural households. The loans that are related to agriculture reflect the support for rural economic growth in the process of financial development [38]; therefore, this paper uses the ratio of agriculture-related loans to the output value of the primary industry in order to measure the level of agricultural credit (LOAN), with a higher ratio indicating a higher level of agricultural credit. The fiscal support for agriculture (SUP) is usually measured by the budget expenditures for agriculture, forestry, and water. However, considering that there are some differences in public budget expenditures among the provinces, this paper uses the ratio of the budget expenditures for agriculture, forestry, and water to the general public budget expenditures in order to measure the level of fiscal support for agriculture and reflect the degree of local government support for the development of the “three rural” industries. The per capita disposable income of the rural residents is obtained from the China Rural Statistical Yearbook, the data on the agriculture-related loans are obtained from the China Rural Financial Services Report, and the data on the agriculture, forestry, and water budget expenditure, the primary industry output value, and the general public budget expenditures are obtained from the China Statistical Yearbook. Considering that the data in this paper are not cross-sectional, the income and the output data are deflated by using the consumer price index (CPI) and the GDP deflator in order to remove the influence of price changes for a cross-period comparison [39]. In addition, we consider the following: (1) We aim to prevent possible heteroskedasticity and multicollinearity problems in the data. (2) The difference between the individual samples is extremely large in multiples, and the absolute differences between the data need to be reduced in order to prevent the effect of extreme values (e.g., at the national level, the minimum value of LOAN is 0.081, and the maximum value is 173.671). (3) The economic meaning of the regression coefficients that is obtained from the logarithm of the variables is the elasticity coefficient, which can play the role of decompartmentalization. Therefore, in this paper, all of the variables are taken in logarithm in the subsequent modeling [40].

The descriptive statistics are shown in

Table 3. Descriptive statistics of the variables.

Region	Variables	Average Value	Standard Deviation	Minimum Value	Maximum Value
National	NI	6594.307	3853.063	1590.330	22,910.360
	LOAN	4.162	8.659	0.081	173.671
	SUP	0.094	0.043	0.012	0.204
Eastern	NI	9041.558	4569.396	2702.193	22,910.360
	LOAN	6.456	13.654	0.081	173.671
	SUP	0.073	0.034	0.012	0.171
Central	NI	5869.348	2518.539	2185.076	10,837.010
	LOAN	2.726	2.700	0.218	13.560
	SUP	0.100	0.038	0.029	0.190
Western	NI	4674.298	2242.106	1590.33	10,597.130
	LOAN	2.911	2.244	0.097	9.460
	SUP	0.110	0.045	0.027	0.204

Note: The national data are obtained by combining the provincial datasets of the three regions. The total number of observations is 1620, of which 594 are in the eastern region, 432 are in the central region, and 594 are in the western region. The first column of the table shows the different ranges of the economic regions, the second column shows the names of the variables, the third column shows the mean of each variable, the fourth column shows the standard deviation of each variable, the fifth column shows the minimum of each variable, and the sixth column shows the maximum of each variable.

. For the NI mean, the eastern region is higher than the national average, while the central and western regions are lower than the national average. The income level of the eastern region has formed a large gap between the central and western regions. For the LOAN average, the national average is 4.162, and the levels in the eastern and western regions are higher than 1, which means that the agriculture-related loans are much larger than the output value of the primary industry. In addition, similar to the NI average, the eastern region is higher than the national average, while the central and western regions are lower than the national average. For the SUP mean, the national average is below 0.1, which means that less than 10% of the government’s general public budget expenditure is spent on agriculture, forestry, and water. Unlike the two variables that have been mentioned above, the western region has the highest SUP mean level, followed by the central region, and the eastern region has the lowest level.

4. Findings

4.1. Unit Root Test

In this paper, we use inter-period data, which may result in “pseudoregression” if the data are nonstationary, therefore, unit root tests are required before estimating the model. Thus, unit root tests are performed on the variable data using the same unit root test that is based on LLC, in addition to different unit root tests that are based on the Fisher ADF test, in order to investigate the smoothness of the data [41]. The results are shown in

Table 4. Unit root test.

Sequence	Inspection Method	National	East	Central	West
NI	LLC	−7.9661 ***	−5.0634 ***	−3.3162 ***	−3.4639 ***
	ADF	15.0111 ***	12.4180 ***	6.9706 ***	6.4275 ***
LOAN	LLC	−5.6720 ***	−4.9085 ***	−4.4036 ***	−2.4399 ***
	ADF	8.2262 ***	5.0202 ***	4.1139 ***	5.0566 ***
SUP	LLC	−5.6229 ***	−5.3287 ***	−1.6512 **	−1.3696 *
	ADF	14.6421 ***	9.4628 ***	7.4263 ***	8.3846 ***

Note: Table 4 shows the results of the smoothness test. The first column of the table shows the variable names; the second column shows the test method selection; and the third, fourth, fifth, and sixth columns show the results of the smoothness tests under the different economic regions. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively.

. All of the variables pass the smoothness test, and therefore, the series can be considered smooth for subsequent model estimation.

4.2. Selection of the Optimal Lag Order

In this paper, the optimal lag order of the PVAR model is selected based on the information minimization criterion; the three information criteria values of AIC, BIC, and HQIC are calculated for the model with lags of 1–3 periods; and the optimal lag order of the model is determined by using the minimum values of the combined three information criteria [42]. As shown in

Table 5. Estimated AIC, BIC, and HQIC values of the model with different lag orders.

Region	Lag Period	AIC	BIC	HQIC
National	1	−3.1896 *	−2.3288 *	−2.8512 *
	2	0.5936	1.5797	0.9824
	3	−1.4199	−0.2944	−0.9751
Eastern	1	−2.9017 *	−2.1451 *	−2.5948 *
	2	−0.7933	0.1667	−0.4036
	3	3.6790	4.8623	4.1597
Central	1	−3.8666 *	−3.1313 *	−3.5679 *
	2	−1.1612	−0.1856	−0.7650
	3	−2.1248	−0.8869	−1.6226
Western	1	−3.5401	−2.7835	−3.2333
	2	0.3934	1.3534	0.7831
	3	−4.5645 *	−3.3812 *	−4.0838 *

Note: Table 5 shows the estimated results of the AIC, BIC, and HQIC values of the model with different lag orders. The first column of the table shows the different economic regions, the second column shows the number of lags (range is 1–3 periods), and the third, fourth, and fifth columns show the estimated results of the three information criteria values of the AIC, BIC, and HQIC, respectively. * indicates the minimum value of each information criterion value for each region.

, the optimal lag order takes the value of one for the national, eastern, and central regions, and three for the western region. However, considering that the difference between the information criterion values of lag one and lag three for the western region is very small and that the sample length is long, lag one is chosen as the optimal lag for the PVAR model for the national, eastern, and western regions.

4.3. Model Stability Test

Since the validity of the impulse-response function of the PVAR model presupposes that the estimation of the model must be stable [43], i.e., the mode of the characteristic roots must be within the unit circle, we need to perform a stability test on the overall model. Since the model has three endogenous variables, and each variable is lagged by one period, there is one characteristic root for each variable. Figure 1 shows the test results from left to right for the national, eastern, central, and western regions. The characteristic roots of each variable in each region are distributed within the unit circle, therefore, we can assume that the model data satisfy the stability condition, and thus, it is reasonable to use the model for dynamic impulse response analysis.

4.4. GMM Estimation of PVAR Model Parameters

Since the PVAR is a dynamic model, there is an endogeneity problem due to the dynamic data structure. In addition, due to cross-sectional heterogeneity, the traditional maximum likelihood estimation method cannot be used to estimate vector autoregressive models based on panel data [44], therefore, this paper uses the generalized method of moments (GMM) estimation method in order to correct biased and inconsistent parameter estimation. The specific estimation results are shown in

Table 6. GMM estimation results.

Region	NI
	NI (t − 1)	LOAN (t − 1)	SUP (t − 1)
National	0.9022 ***	0.0206 ***	0.0279 ***
Eastern	0.9054 ***	0.0181 ***	0.0241 ***
Central	0.8773 ***	0.0275 ***	0.0346 ***
Western	0.9119 ***	0.0210 ***	0.0278 ***

Note: Table 6 shows the results of GMM estimation. The first column of the table shows the different economic regions, and the second, third, and fourth columns show the one-period lagged GMM estimates of each variable for each region. *** indicate significance at the 1% level.

.

The results of the GMM estimation show that, generally, there is a positive correlation between agricultural credit and the increase in farmers’ income, as well as a correlation between fiscal support for agriculture and the increase in farmers’ income. In addition, the correlation between fiscal support for agriculture and the increase in farmers’ income is stronger than that of agricultural credit. In terms of the different regions, the positive correlation between agricultural credit and the farmers’ income growth is the strongest in the central region and the weakest in the eastern region, as is the correlation between fiscal support for agriculture and the increase in the farmers’ income. The reason for this may be that the central region has at least two advantages over the western region. First, agriculture is more developed in the central region than in the western region. Four of the top ten provinces in China in terms of total agricultural output are in the central region. Referring to the measurement method of Li (2021) [45], the level of rural industrial integration is measured as 1.24 in the central region and 0.93 in the western region. Second, the level of human capital is higher in the central region than in the western region. For example, the number of years of education per capita is 9.20 in the central region and 8.69 in the western region. Rural industrial integration not only promotes the extension of the agricultural industry chain and the multifunctionality of agriculture, but also promotes the integration of agricultural services [46,47]. Therefore, the integration of rural industries can effectively play a regulatory role and can enhance the positive correlation between agricultural credit, fiscal support, and farmers’ income. While the integration of rural industries gives rise to new industries and new business modes, it also puts forward higher requirements for farmers’ human capital. The higher the stock of human capital, the higher the acceptance of industrial integration by the farmers and the more they can participate in rural industrial integration [48], which further enhances the positive correlation. However, it should be noted that the eastern region is stronger than the central and western regions in both of the advantages that have been mentioned above, however, the resulting correlation is the weakest, which is probably due to the lack of incentives and mechanisms to further promote farmers’ income in the eastern region. First, in the eastern region, as the main driving force of China’s economic growth, urban and rural areas present obviously different economic benefits from the same economic resource input. Therefore, the pursuit of rapid economic growth may lead to the neglect of the further income growth of the rural residents in the region. Second, the income of the farmers in the eastern region is already at a high level compared with that of the farmers in the central and western regions, and the average annual growth rate of the years of education per capita and the level of rural industrial integration are the lowest among the three regions, thus there is limited room for further improvement. Therefore, the moderating effect of rural industrial integration and human capital level on farmers’ income will gradually diminish, and the eastern region needs to find other potential moderating mechanisms for farmers’ income that have not yet been exploited. However, the limitation of GMM estimation is that it can reflect the relationship between variables from a macro perspective but cannot show the causal logic relationship between the variables, the dynamic transmission mechanism and influence path, or the contribution of shock variables. Therefore, this paper will further study the dynamic relationship between the above economic variables through the Granger causality test, impulse-response function, and variance decomposition.

4.5. Granger’s Causality Test

This paper further analyzes the short-term dynamic correlation between agricultural credit, fiscal support, and farmers’ income by imposing constraints on the estimated coefficients of the variables based on the GMM estimation results and by examining the Granger causality between the economic variables using the Wald statistic. The results are shown in

Table 7. Results of Granger’s causality test.

Region	Dependent Variable-Independent Variable	Card Side	p Value	Test Results
National	NI-LOAN	30.465	0.000	LOAN is NI’s Granger cause
	NI-SUP	22.716	0.000	SUP is NI’s Granger cause
Eastern	NI-LOAN	8.175	0.004	LOAN is NI’s Granger cause
	NI-SUP	4.121	0.042	SUP is NI’s Granger cause
Central	NI-LOAN	15.257	0.000	LOAN is NI’s Granger cause
	NI-SUP	17.892	0.000	SUP is NI’s Granger cause
Western	NI-LOAN	32.703	0.000	LOAN is NI’s Granger cause
	NI-SUP	14.421	0.000	SUP is NI’s Granger cause

Note: Table 7 shows the results of the Granger causality tests. The first column of the table shows the different economic regions; the second column shows the name of the dependent-independent variable; the third column shows the chi-square; the fourth column shows the p value; and the fifth column shows the test results, i.e., whether the independent variable is the Granger cause of the dependent variable.

. From the national perspective, and from the regional perspective of the three regions, there are significant Granger causes between agricultural credit, fiscal support, and farmers’ income, and they pass the 5% significance level test, indicating that there is a certain correlation between agricultural credit, fiscal support, and farmers’ income in the short run, and the model of this paper is reasonable.

4.6. Impulse-Response Function Analysis

The impulse response diagram reflects the response of a single standard deviation change in a perturbation term in a PVAR system in the current period in terms of the shocks of the future period’s living variables when the values of the other variables are held constant in the current and previous periods; this intuitively further portrays the dynamic interaction between the variables [49]. By observing the results of the unit root test and the model stability test on the variable data, it can be seen that the PVAR model that has been established in this paper is relatively stable and meets the requirements of the impulse-response function analysis and the subsequent variance decomposition. The impulse-response function analysis of this paper spans 10 periods and uses 500 Monte Carlo simulations. In order to explore the dynamic correlation between agricultural credit, fiscal support, and farmers’ income, the overall characteristics of the relationship between the economic variables and the regional heterogeneity are reflected. In this paper, impulse response analysis is conducted for the national sample and for the three regions, and the results are shown in Figure 2, Figure 3, Figure 4 and Figure 5. Each graph, from left to right, shows the impulse response results of fiscal support for agriculture, agricultural credit, and farmers’ income on farmers’ income.

4.6.1. Impulse Response Analysis of Agricultural Credit on Farmers’ Income

At the national level, when agricultural credit is subjected to a positive shock, farmers’ income does not respond in the current period, then responds positively, and peaks in period 3. Subsequently, it gradually declines but still remains at a high positive response level, indicating that agricultural credit inputs can rapidly raise farmers’ income in the short term from a national perspective, and this positive correlation is sustainable. By region, after a positive shock to agricultural credit, farmers’ income in the eastern region responds positively in period 1 and quickly peaks in period 2, before weakening and then responding negatively in period 10. In the central region, farmers’ income responds positively in period 1, rises slowly to reach a peak in period 7, and then weakens but still remains at a high level of positive response. In the western region, farmers’ income shows a negative response, peaks in period 3, and then gradually decreases to zero. This paper suggests that this may be due to the double-threshold effect of the correlation between agricultural credit and farmers’ income [50]. The western region is relatively backward in terms of development, its economic level does not cross the first threshold, and the development of agricultural finance is not perfect, which leads to the ineffective use of agricultural credit inputs and even squeezes the space of other relatively mature agricultural resources, therefore, the increase in agricultural credit exacerbates the poverty of the farmers [51]. The economic level in the central region crosses the first threshold but has not yet crossed the second threshold, therefore, agricultural credit can be effectively utilized in the region, and there is a positive correlation between agricultural credit and farmers’ income for a certain period, though this has a certain time lag. The eastern region is relatively developed, its economic level has already crossed the second threshold, and its financial development is perfect. Agricultural credit funds can be fully and effectively utilized and can rapidly promote farmers’ income growth in the short term; however, in the long term, they may make farmers path-dependent and weaken the endogenous development momentum. This in turn inhibit farmers’ income growth, which is, to some extent, consistent with Greenwood et al. (1990) [52], in the sense that the relationship between financial optimization and income growth is not simply linear but may present an inverted “U-” shaped relationship.

4.6.2. Impulse Response Analysis of Fiscal Support for Agriculture on Farmers’ Income

At the national level, after a positive shock to fiscal support, farmers’ income has no response in the current period. A positive response appears, quickly peaks in period 3, and then weakens, but it is still maintained at a certain level in period 10, indicating that there is a positive correlation between fiscal support and farmers’ income growth at the national level, and the correlation can last for a certain period. By region, after a positive shock to fiscal support for agriculture, farmers’ income in the eastern region continues to increase after a negative response in period 1, slows down in period 8, and maintains a high level of negative response. In the central region, a positive response in farmers’ income appears in period 1, gradually strengthens, and then gradually weakens after reaching a peak in period 5, but a high level of positive response remains. In the western region, a positive response in farmers’ income appears in period 1 and continues to increase, then peaks in period 7, and slightly decreases but remains present until period 10. However, the positive response level is weaker than that in the central region. This paper suggests that this may be due to the single-threshold effect of the correlation between fiscal support and farmers’ income. The economic level of the central and western regions has not yet crossed the threshold, there is a positive correlation between fiscal support for agriculture and farmers’ income, and the positive correlation will increase with the increase in the economic level. In addition, the central region includes large agricultural provinces, such as Henan, Heilongjiang, Hunan, and Hubei, which have developed primary industries and relatively good agricultural infrastructure facilities, therefore, the allocated fiscal support for agriculture can be invested in agricultural production more quickly and efficiently. In contrast, the efficiency of fiscal support for agriculture in the west is relatively low and lacks a sustainable and stable long-term mechanism [53], therefore, the positive response in the center is greater than that in the west. Through the measurement of the industrial advanced index [54], it has been found that the industrial advancement index in the eastern region increased from 0.85 in 2003 to 1.52 in 2020, which shows that the industrial structure in the eastern region has changed significantly, and that the development is no longer focused on agriculture. Therefore, too much fiscal support for agriculture in this region can instead squeeze the fiscal resources of the key industries. In contrast, Li Yanqiu (2021), and other studies [50], show that the correlation between fiscal support for agriculture and farmers’ income has a significant threshold effect. When the income of the rural residents is greater than CYN 13,262, the positive effect of fiscal support for agriculture changes in the opposite direction. The data of China’s National Bureau of Statistics show that the per capita disposable income of the rural residents in the eastern region reached CYN 19,930 in 2020, which exceeded the threshold value, therefore, there is a negative correlation between fiscal support for agriculture and farmers’ income in the eastern region.

4.7. Variance Decomposition

This paper further evaluates the contribution of endogenous variables to the prediction variance based on exploring the dynamic correlation between agricultural credit, fiscal support, and farmers’ income. The endogenous variables are decomposed by the variance decomposition of the unit increments into their own and the other two variables’ contributions in a certain proportion in order to express the dynamic characteristics of the model [55], as shown in

Table 8. Variance decomposition.

Period	National			Eastern			Central			Western
	NI	LOAN	SUP	NI	LOAN	SUP	NI	LOAN	SUP	NI	LOAN	SUP
1	1	0	0	1	0	0	1	0	0	1	0	0
2	0.868	0.097	0.035	0.854	0.125	0.021	0.821	0.125	0.054	0.896	0.062	0.042
3	0.717	0.145	0.137	0.714	0.172	0.113	0.653	0.183	0.165	0.745	0.111	0.143
4	0.594	0.146	0.261	0.591	0.163	0.246	0.538	0.179	0.284	0.615	0.127	0.257
5	0.502	0.128	0.370	0.489	0.136	0.376	0.461	0.153	0.386	0.520	0.125	0.355
6	0.437	0.109	0.454	0.410	0.108	0.482	0.409	0.126	0.465	0.453	0.115	0.432
7	0.390	0.093	0.517	0.350	0.086	0.563	0.373	0.103	0.524	0.407	0.104	0.489
8	0.356	0.080	0.564	0.306	0.069	0.625	0.347	0.086	0.567	0.373	0.094	0.533
9	0.331	0.070	0.600	0.272	0.057	0.671	0.328	0.073	0.598	0.349	0.086	0.566
10	0.311	0.062	0.627	0.246	0.048	0.706	0.315	0.065	0.621	0.331	0.079	0.591

Note: Table 8 shows the results of variance decomposition. The first column of the table shows the period, and the second, third, fourth, and fifth columns show the results of variance decomposition for the national, eastern, central, and western variables NI, respectively.

.

At the national level, the contribution of agricultural credit was larger in the early period and declined after reaching a peak of 14.6% in period 4, with a contribution of 6.2% in period 10. The contribution of fiscal support for agriculture is always increasing, and its growth rate is much higher than that of agricultural credit and exceeds the contribution of agricultural credit in period 4, with the largest contribution of 62.7% in period 10. The results by region are similar to those at the national level, whereby agricultural credit has a greater degree of contribution to farmers’ income in the short term, while fiscal support for agriculture can make a greater contribution to farmers’ income growth in the long term, with the contribution of fiscal support for agriculture in the eastern region reaching 70.6% in period 10, which is the highest among the three regions. Compared with the contribution of agricultural credit and fiscal support to farmers’ income, the contribution of agricultural credit in the eastern and central regions is surpassed by fiscal support in period 4, while the contribution of agricultural credit to farmers’ income in the western regions is surpassed by fiscal support in period 2. This is perhaps because the western region is less economically developed, the rural fiscal support system is not perfect, and the use of agricultural loans may lack the corresponding supervision mechanisms [6], which may lead some of the funds of the agricultural loans to flow in other directions. Thus, the contribution of agricultural credit to farmers’ income is weaker in the western region than it is in the eastern and central regions.

5. Conclusions and Policy Insights

5.1. Conclusions

This paper explores and investigates the relationship between agricultural credit, fiscal support for agriculture, and farmers’ income in China from 2003 to 2020 from a dynamic perspective by constructing a panel vector autoregressive (PVAR) model and draws the following conclusions:

First, both agricultural credit and fiscal support inputs are significant Granger causes of farmers’ income, and the positive correlation between fiscal support for agriculture and farmers’ income growth is stronger than that of agricultural credit, which, to some extent, indicates that both can effectively contribute to farmers’ income growth and the rationale behind the local government’s adoption of economic strategies of agricultural credit and fiscal support inputs to regulate farmers’ income.

The second finding concerns the relationship between agricultural credit and farmers’ income. At the national level, a positive shock to agricultural credit generates a positive response in farmers’ income, and the response can last for a certain period. By region, the farmers’ income in the eastern region can quickly yield a strong positive response in the short term, but in the long term, it yields a negative response. In contrast, farmers’ income in the central region can continue to produce a positive response over a certain time period, while in the western region, it continuously produces a negative response. This suggests that the correlation between agricultural credit and farmers’ income may have a double-threshold effect. When the economic level is below the first threshold, and the regional financial development base is weak, agricultural credit will have a negative correlation with farmers’ income growth. When the economic level crosses the first threshold and is below the second threshold, agricultural credit will have a positive correlation with farmers’ income growth. Furthermore, when the economic level crosses the second threshold, the short-term positive correlation between agricultural credit and farmers’ income is further accentuated; however, the possible path dependence creates the negative correlation between agricultural credit and farmers’ income in the long run.

The third finding concerns the relationship between fiscal support for agriculture and farmers’ income. At the national level, farmers’ income responds positively to a positive shock of fiscal support for agriculture, and the response can last for a certain period. By region, the farmers’ income in the eastern region presents a persistent negative response. In contrast, the farmers’ income in the central and western regions presents a continuously positive response, and the positive response is stronger in the central region than it is in the western region. This indicates that the correlation between fiscal support for agriculture and farmers’ income may have a single-threshold effect on farmers’ income. When the economic level of the region is below the threshold, the improvement of the economic level will bring about the improvement in the agricultural infrastructure facilities and the fiscal fund allocation system in the central region compared with the western region, which will enhance the correlation between fiscal support for agriculture and farmers’ income. In contrast, when the economic level exceeds the threshold value, the industrial structure of the eastern region will develop in an advanced direction, and investment in agricultural support will squeeze the fiscal resources of the key development industries, therefore there is a negative correlation between the investment in agricultural support and the growth of farmers’ income.

Fourth, farmers’ income growth, both in the national sample and in the regional samples, show similar performance, i.e., in the short run, agricultural credit makes a larger contribution, while in the long run, fiscal support for agriculture makes a larger contribution.

5.2. Policy Insights

The findings of this study have the following policy implications:

First, a regionally differentiated agricultural credit development strategy should be implemented. For the eastern region, the further development of the agricultural credit services needs to be carefully considered. When it is necessary to boost farmers’ income in the short term, we can take measures to increase the agricultural credit input, however, at the same time, we should also consider in advance the negative impact of this measure in the long term and take mitigation measures in order to reduce this impact. For the central region, we can enhance the supply of agricultural credit and financial services, improve the breadth of agricultural credit services, and promote the growth of local farmers’ income. For the western region, the agricultural credit services should not be blindly carried out, but the construction and the improvement of the agricultural credit infrastructure facilities and departments should be prioritized in order to lay the foundation for agricultural credit development.

Second, we should reasonably enhance and optimize the structure of the fiscal support for agriculture. Because there is a certain threshold effect in the correlation between fiscal support for agriculture and farmers’ income, it is necessary to encourage tilt fiscal support for agriculture from the east to the central and western regions in order to prevent the mismatch of fiscal resources. It is also necessary to improve the supporting construction of the agricultural sector in the western regions, improve the allocation efficiency of the fiscal support for agriculture, give full play to the positive correlation between fiscal support for agriculture and farmers’ income, and further promote rural revitalization.

6. Study Limitations and Future Research Suggestions

6.1. Research Insufficiencies

This study focuses on the relationship between agricultural credit, fiscal support, and farmers’ income from a dynamic perspective, and there are two limitations in the analysis. First, in the GMM estimation analysis, the correlation between agricultural credit, fiscal support, and farmers’ income varies in different regions, and the theoretical logic of this paper is that the level of rural industrial integration and human capital can moderate the relationship between agricultural credit, fiscal support, and farmers’ income. In other words, if a region has a higher level of agricultural industry integration and human capital, the positive correlation between agricultural credit, fiscal support for agriculture, and farmers’ income will be stronger in that region, and vice versa. However, when the income of farmers in a certain region has reached a high level, and the level of rural industrial integration and human capital has limited room to rise, the moderating effect will gradually diminish. For the above logical chain, there is no corresponding empirical analysis in this paper. Second, in the impulse-response function analysis, this paper cites the research on the threshold effect in order to explain the regional heterogeneity of the correlation between agricultural credit, fiscal support, and farmers’ income. This entails the implicit assumption that the strength of the correlation is fixed within a certain interval of economic income, i.e., it presents a homogeneous effect within the same sample interval and a heterogeneous effect between sample intervals. However, the strength of the correlation between agricultural credit, fiscal support, and farmers’ income increase may also vary with the economic level and is not a fixed coefficient within a certain sample interval.

6.2. Future Research Suggestions

In order to address the abovementioned research limitations, this paper proposes the following three suggestions for future research: First, the moderating effects of the relationship between agricultural credit, fiscal support for agriculture, and farmers’ income can be tested empirically by considering rural industrial integration and human capital level as moderating variables. Moreover, this analytical framework can be used to explore other potential moderating mechanisms. Second, semiparametric analysis can be conducted in order to explore whether the strength and the significance of the relationship between agricultural credit, fiscal support, and farmers’ income vary with the level of the economy. In addition, there may be regional spillover effects of agricultural credit and fiscal support on farmers’ income, i.e., the level of agricultural credit and fiscal support in one region may affect the income level of the farmers in the neighboring regions, which can be investigated through spatial econometric analysis.