Can China’s Digital Inclusive Finance Alleviate Rural Poverty? An Empirical Analysis from the Perspective of Regional Economic Development and an Income Gap

Abstract

Digital inclusive finance (DIF) plays an active role in preventing poverty-stricken groups from returning to poverty and reducing poverty. This paper empirically tests the impact of DIF on rural poverty alleviation using panel data from 30 Chinese provinces from 2011 to 2020 as a sample. It employs multiple linear regression, mediation effect models, and threshold effect models. The results show that: (1) DIF and its three sub-indicators (coverage breadth, depth of use, and digitalization degree) have significant poverty reduction effects, and the findings hold even when endogeneity is taken into account; (2) a study of regional heterogeneity found that DIF and its sub-indices, coverage and depth of use in the eastern region, have the greatest effect on the poverty alleviation of rural residents, and the effects in the central and western regions have the least effect; (3) the mediation effect test found that DIF could indirectly promote poverty alleviation in rural areas by promoting regional economic growth and narrowing the urban-rural income gap. The Sobel test shows that the mediating effect of regional economic growth is greater than the mediating effect of the urban-rural income gap; (4) it is found through the threshold effect test that regional economic growth has a double threshold effect on rural poverty alleviation, and as the threshold value continues to increase, the poverty reduction effect increases in turn. Therefore, this paper puts forward policy suggestions for the aspects of accelerating the development of DIF in rural areas, implementing regionally differentiated poverty reduction strategies according to local conditions, promoting regional economic growth, and narrowing the urban-rural income gap.

1. Introduction

Eradicating poverty and achieving common prosperity are the fundamental goals of social development. At the beginning of reform and opening up, the poverty rate in rural China was up to 97.5%. After decades of development, remarkable results were achieved by China on the path of characteristic poverty alleviation: All the rural poor, meeting today’s standards, were lifted out of poverty, and absolute and regional overall poverty was eliminated. However, the return of absolute poverty to zero does not mean the end of the anti-poverty cause. Relative poverty in regions and households may remain for a long time, and a long-term mechanism for poverty reduction needs to be explored. The epidemic of COVID-19 has exacerbated the risk of a return to poverty and the extent of poverty for some groups. For this reason, China’s 14th five-year scheme states to “Improve the monitoring and assist mechanism to prevent the return of poverty, and realize the effective link between strengthening and broadening the accomplishments in poverty alleviation and revitalizing rural areas” (Accessed on 10 June 2021. http://www.gov.cn/xinwen/2021-03/13/content_5592681.htm).

In poverty reduction, resident income is an important measure of “abundance” and an important indicator of absolute poverty [1]. Financial services play a major part in poverty reduction. Inclusive finance, as the most important way to address the “last mile” of financial services, plays a key role in increasing residents’ income and reducing poverty [2]. In 2016, seven Chinese departments collectively published the “Execution Comments on Financial Boosting Poverty Alleviation”. The document calls for, on the basis of the growth of inclusive finance, we would do all we can to encourage financial services in poverty-stricken areas for villages, households, and people and strive to enable every eligible poor population to obtain loans according to their needs easily. Any poor person in need of financial services should be able to enjoy modern financial services easily. To this end, China has formulated a large number of documents and measures to encourage the growth of financial inclusion, including promoting the development of financial inclusion (2016–2020), the G20 advanced principles of DIF, and other documents in 2016; in 2022, the “Implementation Opinions on Promoting the High-quality Growth of Inclusive Finance” was published. In addition, an important discussion on “Strengthening Financial Services for Rural Revitalization” was specifically mentioned in document no. 1 of the central committee of the communist party of China. The above documents and policies underscore the government’s strong desire to address the shortage of rural financial services, consolidate the achievements of poverty alleviation, and support the revitalization of rural areas through inclusive finance.

As an outgrowth of the close connection between digital techniques and inclusive finance in the new era, DIF is not only an extension of the growth of inclusive finance but also a further innovation of the progression mode of inclusive finance, which could generally refer to all actions that use digital financial services to promote financial inclusion [3]. According to the report “Peking University DIF Index (2011–2020)”, China’s DIF business has leapt forward in the past decade, with the median of the index increasing from 33.6 in 2011 to 334.8 in 2020, an increase of nearly nine times [4]. The characteristics of the low threshold, low cost, and wide reach of digital inclusive financial services have attracted great attention in academic circles in terms of narrowing income disparity [5,6], encouraging residents’ consumption [7,8], and reducing poverty [9,10]. To further enrich the research on poverty reduction’s effectiveness and the mechanism of DIF, this study samples 30 Chinese provinces from 2011 to 2020 to comprehensively investigate the impact of DIF and its three sub-dimensions on poverty alleviation among rural residents. It also examines the mediating effect of regional economic growth and income disparity in this procedure.

The novelties of the paper are: (1) this study utilizes the DIF index and its three sub-indices to test the effect of DIF on poverty alleviation of rural residents. Furthermore, this paper discusses the effect of DIF on the poverty alleviation of rural residents from the whole and part of China; (2) from the view of regional economic development and income disparity, this paper analyzes the function of DIF in encouraging poverty alleviation in rural residents. The findings of the study have important implications for China to encourage the growth of inclusive finance further and help poverty alleviation; and (3) founded on traditional linear regression, the non-linear association between DIF and the poverty alleviation of rural residents is tested through the threshold effect.

The remainder of the article contains four sections. Section 2 contains a literature review and research hypotheses. Section 3 contains the selection of indicators, model setting, and data sources. Section 4 presents the empirical results. Section 5 focuses on the outcomes and suggestions for countermeasures.

2. Literature Review and Research Hypothesis

Financial development and poverty reduction have always been hot topics of research around the world. The research regarding these issues for the existing literature has gone through roughly three stages: traditional finance and poverty reduction, inclusive finance and poverty reduction, and DIF and poverty reduction [11,12]. First, the conclusions on the development and poverty reduction effects of traditional finance have not been unified and mainly include three viewpoints. The mainstream view holds the point that financial development could reduce the poverty level of a country or region by increasing residents’ income [13], stimulating consumption [14], promoting economic growth [15,16], and promoting fair income distribution [17]. The second view holds that financial development is not conducive to poverty reduction from the perspectives of financial allocation distortion, unequal opportunity, and financial exclusion [18,19]. The third view holds that financial development and poverty alleviation exhibit an inverted “U”-shaped non-linear relationship [20]. Secondly, most scholars’ research conclusions on the poverty reduction impact of inclusive finance are similar. Namely, inclusive finance alleviates destitution by enhancing the efficiency of financial services and risk management capabilities [21], promoting economic growth [22], narrowing income gaps [23], promoting residents’ entrepreneurship [24], and increasing residents’ income and employment rates [25].

The concept of DIF was proposed for a short period. In 2016, the “G20 High-level Principles of DIF” announced at the G20 summit in Hangzhou proffered the concept of DIF for the first time. Driven by new technology, such as big data and cloud computing, new vitality into digital inclusive financial poverty alleviation was injected [3]. This also led to a research upsurge of scholars on the poverty reduction effect of DIF. Currently, the theoretical research on poverty reduction through DIF is relatively mature and has reached a consistent conclusion, which is that DIF mainly achieves poverty reduction through direct and indirect effects [26]. As far as the direct effect is concerned, most of the current literature is based on financial channels to achieve poverty reduction. Most scholars believe that DIF reduces financial transaction costs and credit costs to a greater extent through digital techniques. For instance, the internet, cloud computing, and big data reduce financial exclusion, increase the availability of financial services to poverty groups, and thus promote the entrepreneurship and employment of poor residents to achieve income growth for poverty alleviation purposes [27,28,29]. The second is poverty reduction through economic channels. E-commerce developed based on DIF, especially rural e-commerce, is playing an important role in poverty reduction. DIF may not only help the poor to start businesses online by alleviating financial constraints but also sell special products from poverty-stricken areas to the whole country with the help of e-commerce platforms developed based on DIF, increasing the income of poverty-stricken areas [30,31]. Finally, DIF could also guide residents’ online consumption by easing liquidity constraints and facilitating residents’ payments, thereby reducing the living cost of the poor and improving their quality of life [32]. Other scholars pointed out that DIF not only has a positive effect on poverty alleviation in the area but also has a positive space overflow effect on related regions, and pointed out that the poverty reduction effect and spatial spillover effect are significant in a short time, but not significant in the long term [33].

In terms of indirect channels, the poverty reduction effect of DIF primarily contains the following areas. The first is the inclusive growth effect of DIF. The rapid development of DIF has gradually transferred financial resources to groups and enterprises with financing difficulties and optimized the allocation of social, financial resources. On this basis, an inclusive economic increase could be achieved. The level of economic growth has improved, and the social trickle-down effect is obvious, which in turn, promotes the improvement of the overall level of social welfare of the poor and achieves a certain effect of poverty reduction [34]. The second is the income distribution effect of DIF. Some scholars believe that DIF could improve income distribution by lowering financial barriers and reducing financial exclusion [35]. In the early stage of financial development, owing to the profit-driven nature of financial capital, traditional financial institutions mainly served richer groups, and low-income groups were unable to obtain financial services, leading to a more prominent problem of income inequality between the rich and the poor [36]. DIF has the advantages of a low threshold, low cost, and no physical space constraints, and its service targets cover the “long-tail market” groups, such as low-income groups and SMEs, which are severely excluded by traditional finance, and improve the social income imbalance [37]. At the same time, the development of DIF could also increase the earnings of the poor and narrow income disparities between the rich and the poor by improving the employment level of the poor and adjusting the employment structure, thereby achieving poverty alleviation [38]. In addition, some scholars believe that DIF could also indirectly alleviate poverty through industrial structure upgrading [39] and improving the level of regional innovation [40]. Founded on the above analysis, this paper, therefore, proposes the succeeding assumptions:

H1. 

The progression of DIF could significantly promote poverty alleviation among rural residents.

H2. 

DIF is conducive to regional economic development, thereby promoting the poverty alleviation of rural residents.

H3. 

DIF could help to narrow income disparity, thereby facilitating poverty alleviation among rural residents.

3. Data and Empirical Model

3.1. Data Sources

This paper utilizes the panel data of 30 provinces in China from 2011 to 2020 as a sample for empirical analysis. The data stems from the “Statistical Yearbook” of each province, the “China Rural Statistical Yearbook”, the “China Statistical Yearbook”, and the “China Education Statistical Yearbook”. The missing data for individual years and areas are completed by interpolation. The DIF index is released by the DIF Research Center of Peking University. The index is constructed from three dimensions of DIF coverage, depth of use, and degree of digitization and contains 33 indicators, which are recognized and widely researched by Chinese scholars. Simultaneously, to eradicate the possible impact of heteroscedasticity, logarithmic processing is adopted for the dependent variable, DIF, three sub-indicators, and economic growth.

3.2. Variable Measurement

(1)

Dependent variable: Poverty alleviation level for rural residents (PCDI). In poverty alleviation work, residents’ income is an important measure of “abundance” and an important indicator of absolute poverty [1]. Considering the accessibility of poverty data, this study uses the net income per capita for rural residents (PCDI) to measure the poverty alleviation level of the rural inhabitants. The net income per capita for rural residents reflects the average level of income of all rural inhabitants in an area. The higher the net income per capita for rural inhabitants, the lower the poverty level of the rural inhabitants in this area, and the higher the poverty alleviation level of the rural residents. On the contrary, the lower the net income per capita for the rural inhabitants, the lower the poverty alleviation level of the rural residents. In 2013, the national bureau of statistics unified the indicator of rural per capita net income with urban per capita disposable income and changed the caliber to rural per capita disposable income. According to the relevant statistical yearbooks, there is not much difference between the two, so the per capita disposable income of the rural inhabitants in the corresponding year is used to replace the net income per capita for rural inhabitants.

(2)

Independent variable: digital inclusive finance (DIF). The DIF data is released by the Digital Finance Research Center of Peking University. The DIF index contains three sub-indicators: the breadth of coverage (lnDIF_B), depth of use (lnDIF_D), and degree of digitization (lnDIG). To a certain extent, it is scientific, reasonable, and authoritative and has been widely used by domestic scholars in research related to DIF [41]. Now, we must examine its impact on the poverty alleviation level of rural residents. Considering the dimensional difference between the data of each variable, this paper uses the logarithm of the original DIF index and its sub-dimension data.

(3)

Mediating variables: (1) The level of regional economic growth (RGDP). This study uses the per capita gross product (RGDP) of each province (autonomous region and municipality) to evaluate the regional economic development level [42]. Under normal circumstances, the increase in regional GDP per capita means that the economic growth of the area is improving, the income level of residents would increase, and poverty would be alleviated. Founded on the existing study, this paper utilizes the gross product per capita (RGDP) to evaluate the regional economic growth level. (2) The rural-urban income gap (Theil). This study picks the Theil index to evaluate income disparity. The indicators for evaluating income disparity mainly incorporate the Gini coefficient and the Theil index [43,44]. Among them, the Gini coefficient is a common indicator to evaluate income disparity. However, it is only susceptible to changes in the income of the bourgeoisie, and it is hard to differentiate whether the expansion of the Gini coefficient is due to the general increase in the income of all classes or the further widening of the wealth gap, which leads to the lower earnings of the low-earnings class and the higher income of the high-earnings class. Founded on the comparative study in this article, in comparison to the Gini coefficient, the Theil index not only fully considers the impact of people but also is more responsive to alterations in the earnings of different classes and two segments and reflects the alterations in the earnings of different classes and changes in the urban-rural population ratio well. The larger the Theil index is, the larger the income disparity between residents. Narrowing the income disparity between residents is not only helpful for poverty alleviation but also for social equity and stability, and economic growth.

(4)

Control variables: This study draws on the existing research literature to select the control variables as follows [45,46]: (1) Transportation infrastructure (road). In order to better describe the traffic accessibility of each province, this study picks the proportion of the total mileage of graded highways and other highways in the area of the administrative area, that is, the density of provincial highways; (2) The level of the external openness (open) is measured with the percentage of total imports and exports to the region’s GDP; (3)The level of financial support for agriculture (GOV), which is expressed by the percentage of fiscal support for agriculture in general public budget expenditures, of which financial support for agriculture is calculated as the expenditure on agriculture, forestry, and water affairs; (4) Rural education development level (EDU), utilizing the average years of education to evaluate the level of education. EDU = (number of people with primary school education/total number of people) × 6 + (number of people with junior high school education/total number of people) × 9 + (number of people with high school education/total number of people) × 12 + (college degree and above/total number of people) × 16; (5) The employment level of the rural population (JOB) is evaluated with the proportion of rural private enterprises and rural individuals employed in the total population in rural areas.

The descriptive statistical results of the selected variables were shown in

Table 1. Descriptive statistics.

Variables		Sample Size	Mean	Standard Deviation	Min	Max
Dependent variable	LnPCDI	300	9.358	0.413	8.271	10.46
Explanatory variable	LnDIF	300	5.219	0.668	2.909	6.068
	LnDIF_B	300	5.075	0.820	0.673	5.984
	LnDIF_D	300	5.201	0.648	1.911	6.192
	LnDIG	300	5.510	0.698	2.026	6.136
Mediating variables	LnRGDP	300	9.834	0.862	7.421	11.68
	Theil	300	0.096	0.044	0.0195	0.227
Control variables	Open	300	0.411	0.462	0.012	2.397
	JOB	300	0.233	0.375	0.180	2.665
	EDU	300	9.232	0.887	7.514	12.80
	GOV	300	0.115	0.0328	0.041	0.204
	Road	300	0.952	0.5088	0.018	2.665

below.

3.3. Model Specification

According to the purpose of this paper, the benchmark regressions (1)–(4) are constructed to examine the effect of DIF development on the poverty alleviation level of rural residents. The regression model is set as follows.

$${\mathrm{PCDI}}_{\mathrm{it}}={\mathsf{\lambda }}_0+{\mathsf{\lambda }}_1{\mathrm{DIF}}_{\mathrm{it}}+{{\displaystyle \sum }}^{\hspace{1em}}{\mathsf{\lambda }}_{\mathrm{i}}{\mathrm{X}}_{\mathrm{it}}+{\mathrm{u}}_{\mathrm{i}}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(1)

$${\mathrm{PCDI}}_{\mathrm{it}}={\mathsf{\lambda }}_0+{\mathsf{\lambda }}_1\mathrm{DIF}\_{\mathrm{B}}_{\mathrm{it}}+{{\displaystyle \sum }}^{\hspace{1em}}{\mathsf{\lambda }}_{\mathrm{i}}{\mathrm{X}}_{\mathrm{it}}+{\mathrm{u}}_{\mathrm{i}}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(2)

$${\mathrm{PCDI}}_{\mathrm{it}}={\mathsf{\lambda }}_0+{\mathsf{\lambda }}_1\mathrm{DIF}\_{\mathrm{D}}_{\mathrm{it}}+{{\displaystyle \sum }}^{\hspace{1em}}{\mathsf{\lambda }}_{\mathrm{i}}{\mathrm{X}}_{\mathrm{it}}+{\mathrm{u}}_{\mathrm{i}}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(3)

$${\mathrm{PCDI}}_{\mathrm{it}}={\mathsf{\lambda }}_0+{\mathsf{\lambda }}_1{\mathrm{DIG}}_{\mathrm{it}}+{{\displaystyle \sum }}^{\hspace{1em}}{\mathsf{\lambda }}_{\mathrm{i}}{\mathrm{X}}_{\mathrm{it}}+{\mathrm{u}}_{\mathrm{i}}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(4)

Among them, the explanatory variables PCDI indicates the poverty alleviation level of rural residents; DIF indicates the DIF development index; DIF_B indicates the DIF coverage breadth; DIF_D indicates the digital inclusive finance usage depth, and DIG indicates the digitization degree. ${\mathrm{X}}_{\mathrm{it}}$ indicates the control variables in the paper which may affect poverty alleviation.

In this paper, we examine the mediating effect between DIF and rural residents’ poverty alleviation level from two perspectives: economic growth and income distribution. The model refers to the existing research results [47]. ${\mathsf{\lambda }}_1$ denotes the total effect of DIF on rural residents’ poverty alleviation level, ${\mathsf{\theta }}_1$ denotes the direct effect of DIF on rural residents’ poverty alleviation level effect, and ${\mathsf{\alpha }}_1\times {\mathsf{\theta }}_2$ denotes the mediating effect. In the case that ${\mathsf{\lambda }}_1$, ${\mathsf{\alpha }}_1$, and ${\mathsf{\theta }}_2$ are significant, if ${\mathsf{\theta }}_1$ is significant, it means there is a partial mediation effect; if ${\mathsf{\theta }}_1$ is not significant, it means there is a full mediation effect.

$${\mathrm{PCDI}}_{\mathrm{it}}={\mathsf{\lambda }}_0+{\mathsf{\lambda }}_1{\mathrm{DIF}}_{\mathrm{it}}+{{\displaystyle \sum }}^{\hspace{1em}}{\mathsf{\lambda }}_{\mathrm{i}}{\mathrm{X}}_{\mathrm{it}}+{\mathrm{u}}_{\mathrm{i}}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(5)

$${\mathrm{MED}}_{\mathrm{it}}={\mathsf{\alpha }}_0+{\mathsf{\alpha }}_1{\mathrm{DIF}}_{\mathrm{it}}+{{\displaystyle \sum }}^{\hspace{1em}}{\mathsf{\alpha }}_{\mathrm{i}}{\mathrm{X}}_{\mathrm{it}}+{\mathrm{u}}_{\mathrm{i}}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(6)

$${\mathrm{PCDI}}_{\mathrm{it}}={\mathsf{\theta }}_0+{\mathsf{\theta }}_1{\mathrm{DIF}}_{\mathrm{it}}+{\mathsf{\theta }}_2{\mathrm{MED}}_{\mathrm{it}}+{{\displaystyle \sum }}^{\hspace{1em}}{\mathsf{\theta }}_{\mathrm{i}}{\mathrm{X}}_{\mathrm{it}}+{\mathrm{u}}_{\mathrm{i}}+{\mathsf{\epsilon }}_{\mathrm{it}}$$

(7)

MED represents the mediating variables, specifically, the regional economic growth level (RGDP) and the income gap (theil). The rest of the variables are the same as above.

Due to the vast territory of China, regional development is uneven. The impact of DIF on the poverty alleviation of rural residents might be related to the level of regional economic growth. During economic growth at different levels, the effect of DIF in reducing poverty and increasing income may be different. Therefore, this paper constructs the following panel threshold model to verify further whether the effect of DIF on the poverty alleviation of rural residents has a threshold effect based on the level of regional economic development.

$$\begin{gathered} {\mathrm{PCDI}}_{\mathrm{it}}=\partial +{\partial }_1{\mathrm{DIF}}_{\mathrm{it}}\ast \mathrm{I}{\left(\mathrm{Q}<{\mathsf{\gamma }}_1\right)}_{\mathrm{i},\mathrm{t}}+{\partial }_2{\mathrm{DIF}}_{\mathrm{it}}\ast \mathrm{I}{\left({\mathsf{\gamma }}_1\le \mathrm{Q}<{\mathsf{\gamma }}_2\right)}_{\mathrm{i},\mathrm{t}}+\cdots \\ +{\partial }_{\mathrm{n}}{\mathrm{DIF}}_{\mathrm{it}}\ast \mathrm{I}{\left(\mathrm{Q}\ge {\mathsf{\gamma }}_{\mathrm{n}}\right)}_{\mathrm{i},\mathrm{t}}+{\partial }_{\mathrm{n}+1}{{\displaystyle \sum }}^{\hspace{1em}}{\mathrm{contry}}_{\mathrm{i},\mathrm{t}}+{\mathsf{\epsilon }}_{\mathrm{i},\mathrm{t}} \end{gathered}$$

(8)

Among them, Q is the threshold variable, I( ) is an indicator function; if the formula in the parentheses holds, I( ) takes 1, otherwise takes 0; ${\mathsf{\gamma }}_1$,…, ${\mathsf{\gamma }}_{\mathrm{n}}$ is the threshold value of Q to be estimated; The rest of the variables are the same as above.

4. Empirical Results and Analysis

4.1. Baseline Analysis

To ensure the scientificity and rationality of the selection of the control variables and to avoid the existence of multicollinearity in the variables, a variable correlation test should be carried out before the empirical analysis.

Table 2. The results of the correlation test of the variables.

	LnPCDI	LnDIF	Open	JOB	EDU	GOV	Road
LnPCDI	1
LnDIF	0.754 ***	1
Open	0.499 ***	0.0740	1
JOB	0.604 ***	0.275 ***	0.662 ***	1
EDU	0.680 ***	0.324 ***	0.644 ***	0.642 ***	1
GOV	−0.463 ***	−0.0420	−0.709 ***	−0.471 ***	−0.546 ***	1
Road	0.524 ***	0.239 ***	0.499 ***	0.500 ***	0.406 ***	−0.721 ***	1

Note: *** indicates significance at the 1% level.

lists the results of the correlation test of each variable. The results show that there is a significant correlation between the control variables and the explained variables, indicating that the selection of control variables is scientific. The correlation coefficients among the other significantly correlated variables are all less than 0.75. That is, the multicollinearity may be small, indicating that these variables could be used for empirical testing.

Commonly used panel fitting models include the mixed least squares (POLS), random effects (RE), and fixed effects models (FE). The model most suitable for the sample data in this paper still needs to be tested. For this, the F test, tLM test, and Hausman test were used to select the model. The test results are shown in

Table 3. Optimal model test results.

Testing Method	p-Value	Conclusion
LM test	0.000	RE is better than POLS
F test	0.000	FE is better than POLS
Hausman test	0.000	FE is better than RE
Conclusion	FE is the optimal model

. According to the results in Table 3, the fixed effect model is the optimal model, so this paper selects the fixed effect model as the benchmark regression model for the fitting analysis.

The basic regression results of DIF and its three sub-dimensions (DIF_B, DIF_D, and DIG) on the poverty alleviation level of rural residents are shown in

Table 4. Regression Results of DIF on the Poverty Alleviation Levels of Rural Residents.

Variables	(1)	(2)	(3)	(4)	(5)
	LnPCDI	LnPCDI	LnPCDI	LnPCDI	LnPCDI
LnDIF	0.395 ***	0.272 ***
	(39.45)	(21.75)
LnDIF_B			0.202 ***
			(19.64)
LnDIF_D				0.266 ***
				(17.47)
LnDIG					0.170 ***
					(11.77)
Open		−0.160 ***	−0.073	−0.042	−0.061
		(−4.38)	(−1.26)	(−0.68)	(−0.82)
EDU		0.139 ***	0.189 ***	0.135 **	0.274 ***
		(3.72)	(4.81)	(3.12)	(5.50)
GOV		0.692	0.611	0.758	2.861 ***
		(1.29)	(1.06)	(1.23)	(4.04)
JOB		0.172 **	0.199 **	0.163 *	0.137
		(2.83)	(3.07)	(2.36)	(1.64)
Road		0.634 ***	0.740 ***	0.857 ***	1.032 ***
		(6.48)	(7.19)	(7.90)	(7.96)
_cons	7.296 ***	5.934 ***	5.787 ***	5.788 ***	4.513 ***
	(138.50)	(19.18)	(17.57)	(16.27)	(11.14)
N	300	300	300	300	300
adj. R2	0.836	0.893	0.878	0.860	0.798
F	1556.009	381.233	329.329	281.625	182.419

Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively, whereas the brackets represent the t values, the same as below.

. It can be seen from Model (1) that without adding the control variables, the estimated coefficient of DIF and poverty alleviation for rural residents is 0.395, and it has passed the 1% level significance test. It shows that DIF could significantly promote poverty alleviation for rural residents. Model (2) shows that the results are still significant after adding the control variables, and every 1% increase in the overall DIF index increases the poverty alleviation level of rural residents by 0.271%. A reasonable explanation is that with the continuous advancement of internet technology, inclusive finance has been endowed with digital attributes, which could provide financial services to rural groups in a more convenient and efficient way and meet the financial service needs of rural residents due to production and life. On the one hand, DIF enables traditional financial institutions to get rid of the restrictions of physical outlets, reducing the cost of providing services for financial institutions such as banks. At the same time, rural residents could handle basic financial services, such as account openings, transfers, deposits, and loans, etc. without leaving their homes, using mobile terminals, such as mobile phones and computers, as the medium, which reduce the use cost of rural residents. On the other hand, the “digitality” and “inclusiveness” of DIF could effectively reduce the financial exclusion faced by vulnerable groups such as rural residents. Moreover, financial institutions could establish electronic information platforms to collect the transaction and credit information of farmers, improve the personal credit portrait of rural residents, reduce the information asymmetry between credit demanders and credit suppliers, and thus improve the willingness of financial institutions to provide services and the market participation opportunities of rural credit groups. In general, the development of DIF could meet the financial service needs of all social strata so that disadvantaged groups, such as rural residents, are no longer excluded from the financial system, thereby increasing their income levels to achieve the goal of poverty alleviation.

Models (3)–(5) show that the DIF_B, DIF_D, and DIG are significantly and positively correlated with the poverty alleviation level of rural residents, and all of them pass the significance test at the 1% level. It can be seen from the estimated coefficients that in the three sub-dimensions, the poverty reduction effects of the depth of use, the breadth of coverage, and the degree of digitization are weakened in turn. Among them, the depth of use has the best effect, and the level of poverty alleviation increases by 0.266% for every 1% increase in the depth of use of the DIF index. The reason may be that the increase in DIF_D means that financial institutions could provide customers with more personalized and diverse financial services. The improvement of DIF_B allows rural groups who were originally “marginalized” due to geographical constraints to obtain more equitable and efficient financial services anytime, anywhere, at a lower cost to improve their living space, thereby increasing their income levels. The improvement of DIG renders the advantages of digital finance more prominent. In addition to providing more convenient service methods for capital demanders, DIF could effectively reduce credit thresholds by virtue of digital technology, optimize the risk management and control efficiency and credit evaluation system of financial institutions, and improve the service enthusiasm of capital suppliers.

As far as the control variables are concerned, it can be seen from Model (2) that the estimated coefficient of the degree of openness (open) to the poverty alleviation level of rural residents is −0.034, and it passed the 1% level significance test. It indicates that the improvement of the degree of opening to the outside world would have a restraining effect on the growth of rural residents’ income, thereby increasing the poverty level of rural residents. The reason may be that although a high level of opening to the outside world is conducive to the introduction of foreign advanced agricultural production technologies and production concepts, rural residents have limited education levels, and their thinking is relatively conservative, rendering it difficult to digest and absorb foreign technologies in a short period of time. At the same time, the improvement of the level of opening to the outside world means that international trade is more frequent, and the expansion of the import scale would impact China’s agricultural product market to a certain extent, causing the outflow of funds from the country’s rural areas, which is ultimately not conducive to increasing farmers’ income, thereby reducing the poverty level of rural residents. The educational level (EDU), employment level of rural residents (JOB), and transportation infrastructure (road) all have significant positive effects on the poverty alleviation level of rural residents. The scale of government expenditure (GOV) has a positive but insignificant impact on the poverty alleviation level of rural residents, which may be related to the unreasonable structure of China’s fiscal support for agriculture and the inaccuracy of support for agriculture.

4.2. Robustness and Endogeneity Test

To guarantee the persuasiveness and reliability of the empirical outcomes, this study uses the following four methods to test the robustness. (1) Exclude the sample of municipalities directly under the central government. Considering that the economic growth degree and infrastructure construction of Beijing, Shanghai, Chongqing, and Tianjin are quite different from other provinces in China, it may affect the empirical results. Therefore, the four municipalities directly under the central government in China were excluded from the regression sample, and the fixed effect model was utilized for testing. The regression outcomes are shown in

Table 5. Robustness and Endogeneity Test results.

Variable	(1)	(2)	(3)	(4)	(5)	(6)
					Phase I	Phase II
	LnPCDI	2011–2016	2017–2020	LnPCDI	LnDIF	LnPCDI
LnDIF	0.251 ***	0.219 ***	0.965 ***			0.498 ***
	(18.54)	(17.42)	(31.36)			(14.41)
L.LnDIF				0.236 ***
				(23.22)
Inter					0.402 ***
					(10.37)
Control Variable	Control	Control	Control	Control	Control	Control
_cons	5.926 ***	6.916 ***	3.741 ***	6.399 ***	−0.762	6.275 ***
	(18.61)	(19.44)	(18.90)	(23.37)	(−1.26)	(29.73)
N	260	180	120	270	300	300
adj. R2	0.901	0.907	0.973	0.912	0.438	0.794
F	359.363	235.4	325.4	353.7	34.19

Note: *** indicates significance at the 1% level, whereas the brackets represent the t values.

Model (1). After excluding the municipalities, the results remain robust. (2) Divide the time interval. At the G20 summit held in Hangzhou, China, in September 2016, the concept of DIF was proposed for the first time, and the document “G20 Advanced Principles of Digital Financial Inclusion” was disclosed for the first time. Since then, China’s DIF has flourished further. Taking 2016 as the node, the two periods before and after the node would inevitably have a large difference in the level of DIF development, and the latter period would inevitably have a higher level of development and a better poverty reduction effect. The time interval of the third DIF index released by the Digital Inclusion Research Center of Peking University is 2011–2020. Therefore, this study selects data from two time periods, 2011–2016 and 2017–2020, to re-verify the effect of DIF development on the poverty alleviation level of rural residents. It can be seen from Models (2) and (3) in Table 5 that after dividing the time interval, the direction and significance level of the estimated coefficient of the core explanatory variables are unchanged, and the estimated coefficient of the data after the time node is 0.965, which is much larger than the coefficient before the node of 0.219. The results show that the development level of DIF is higher after the node, and the poverty reduction effect is better, which further verifies the robustness of the empirical results. (3) The explanatory variable with a lag of one period. Although this paper analyzes the relationship between DIF and the poverty alleviation level of rural residents, in order to obtain unbiased estimates, more control variables are selected to alleviate the endogeneity problem caused by the omission of important variables. However, the model setting still needs to face the simultaneous endogeneity problem that the two are mutually causal. To eliminate the endogeneity problem caused by the two-way causality, this paper selects the explained variables with a lag of one period for regression. From the regression results of Model (4), it can be seen that the explained variables of one lag period also have a significant positive impact on the poverty alleviation level of rural residents, which further proves the robustness of the results. (4) The instrumental variable method. In order to further solve the endogeneity problem, this paper uses the number of internet access ports (inter) as an instrumental variable for DIF and uses the 2SLS method to estimate. The results from Model (5) show that there is a significant positive relationship between the first stage internet network end interface and the core explanatory variables in this paper, and the F-statistic for the first stage is 34.19, both much greater than 10, indicating that there is no weak instrumental variable problem, which shows that the selected instrumental variables are reasonable. The regression results from Model (6) show that the estimated coefficient of digital inclusion finance is significantly positive at the 1% level after the introduction of the instrumental variables, indicating that digital inclusion finance has a significant positive contribution to the poverty alleviation level of rural residents, further illustrating the robustness of the results.

4.3. Heterogeneity Analysis

China has an enormous territory, and the level of development varies greatly from province to province. To verify whether there is regional heterogeneity in the impact of DIF on the poverty alleviation level of rural residents, this study refers to the division of provinces and regions by scholars and divides the whole sample into three sub-samples of eastern, central, and western areas for regression, respectively. Owing to the readability and space limitations of the experimental results, we omit the control variables. The regression outcomes are presented in

Table 6. Regression outcomes of regional heterogeneity of DIF and the poverty alleviation level of rural residents.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
	Eastern Part	Middle Part	Western Part	Eastern Part	Middle Part	Western Part	Eastern Part	Middle Part	Western Part
LnDIF	0.299 ***	0.229 ***	0.204 ***
	(12.69)	(11.32)	(9.01)
LnDIF_B				0.298 ***	0.200 ***	0.114 ***
				(14.52)	(13.12)	(6.67)
LnDIF_D							0.292 ***	0.239 ***	0.176 ***
							(9.84)	(8.38)	(7.53)
Control variable	Control	Control	Control	Control	Control	Control	Control	Control	Control
_cons	6.836 ***	5.658 ***	5.193 ***	6.997 ***	5.944 ***	4.918 ***	6.967 ***	5.283 ***	5.210 ***
	(13.30)	(11.23)	(10.50)	(14.85)	(12.90)	(8.58)	(11.45)	(8.85)	(9.44)
N	120	90	90	120	90	90	120	90	90
adj. R2	0.907	0.914	0.920	0.923	0.930	0.891	0.876	0.878	0.902
F	177.662	144.793	137.472	215.954	179.935	98.691	128.392	98.280	111.507

Note: *** indicates significance at the 1% level, whereas the brackets represent the t values.

.

The outcomes in Table 6 show that the estimated coefficients of the overall DIF index and its two sub-dimensions (DIF_B, DIF_D) on the poverty alleviation level of rural inhabitants in the eastern, central, and western regions are positively correlated and decrease in turn, all of which have passed the significance test. The outcomes present that the contribution of DIF to the poverty alleviation level of rural residents in the eastern area is remarkably higher than that in the central and western areas. The reason may be that the eastern region mostly consists of economically developed provinces, their rural digital infrastructure construction is relatively complete, and the internet coverage is relatively high. The superior economic conditions help to deepen digital financial reform and lower the threshold for rural residents to access financial services. Moreover, most of the well-known large-scale financial technology companies in China are located in the eastern provinces, which also improves the productivity of allocating resources in the eastern region to a certain extent. In addition, affected by the degree of economic growth, rural residents in the eastern area have relatively high personal endowments, so it is easier to break through self-exclusion and use digital financial products to improve their living standards. The central and western regions are mostly agricultural provinces with relatively backward economies. Before the advent of DIF, owing to the restrictions of economic growth and geographical environment, rural inhabitants in the middle and western areas suffered from severe financial exclusion. With the transformation of the information industry, the emergence of DIF has provided rural inhabitants in the middle and western areas with lower thresholds and more convenient financial services and improved the rural financial development environment in the middle and western areas to a certain extent. However, the various network infrastructures in the rural areas of the central and western regions are not perfect, and there is still much room for progress in the progression level of DIF. At the same time, due to the limitations of economic and educational conditions, the human capital of rural residents in the middle and western areas is relatively weak, and the lack of financial literacy leads them to face a more serious “technical threshold”. To some extent, this issue weakens the impact of DIF on the poverty alleviation of rural residents in the middle and western areas. To provide better play to the pro-poor characteristics of DIF and promote the process of common prosperity at this stage, we should focus on the efficiency of financial poverty alleviation in the middle and western areas, keep improving the infrastructure construction in underdeveloped regions, and provide a good growth soil for the development of inclusive finance. Simultaneously, it is essential to improve the inclusiveness of education in rural regions, alleviate the self-exclusion of rural residents in underdeveloped areas, and encourage them to use digital financial services; for example, internet insurance and internet wealth management to avoid risks and increase income.

4.4. Mediation Effect Analysis

4.4.1. DIF, Regional Economic Development, and the Poverty Alleviation Level of Rural Residents

Table 7. DIF and poverty alleviation level of rural residents: a test of the mediating effect of the regional economic development level.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	RGDP	lnPCDI	RGDP	lnPCDI	RGDP	lnPCDI	RGDP	lnPCDI
LnDIF	0.165 ***	0.182 ***
	(12.37)	(13.88)
LnDIF_B			0.130 ***	0.127 ***
			(12.87)	(11.44)
LnDIF_D					0.161 ***	0.162 ***
					(10.76)	(11.24)
LnDIG							0.097 ***	0.093 ***
							(7.51)	(8.19)
RGDP		0.546 ***		0.577 ***		0.648 ***		0.796 ***
		(10.91)		(10.49)		(12.50)		(15.42)
Control variable	Control	Control	Control	Control	Control	Control	Control	Control
_cons	8.853 ***	1.102 *	8.860 ***	0.671	8.757 ***	0.117	7.945 ***	−1.814 ***
	(26.87)	(2.16)	(27.42)	(1.20)	(25.15)	(0.22)	(21.98)	(−3.63)
N	300	300	300	300	300	300	300	300
adj. R2	0.786	0.929	0.793	0.917	0.764	0.916	0.715	0.899
F	170.705	508.492	177.634	429.656	150.599	423.971	118.378	348.526
Sobel test	0.1202 ***		0.1034 ***		0.1225 ***		0.1096 ***
	(z = 8.647)		(z = 8.653)		(z = 8.424)		(z = 7.215)
Goodman-1 test	0.1202 ***		0.1034 ***		0.1225 ***		0.1096 ***
	(z = 8.633)		(z = 8.639)		(z = 8.412)		(z = 7.204)
Goodman-2 test	0.1202 ***		0.1034 ***		0.1225 ***		0.1096 ***
	(z = 8.661)		(z = 8.667)		(z = 8.437)		(z = 7.226)
Intermediary effects	33.845%		36.853%		32.74%		41.622%

Note: * and *** indicate significance at the 10%, and 1% levels, respectively, whereas the brackets represent the t values.

shows the test results of the regional economic growth as an intermediary variable. Models (1), (3), (5), and (7) examine the effects of DIF and three sub-dimensions (DIF_B, DIF_D, and DIG) on the mediating variables. Models (2), (4), (6), and (8) empirically test the mediating effect of RGDP as a mediating variable. The regression outcomes indicate that in the process of DIF promoting the poverty alleviation of rural residents, the level of regional economic growth is a mediating variable, and its mediating effect is significant. Similarly, in the process of the three sub-indicators of DIF promoting the poverty alleviation of rural residents, the level of regional economic growth is a mediating variable, and the mediating effect is also significant. The Soble test shows that the mediating effects of DIF and its three sub-dimensions, DIF_B, DIF_D, and DIG, account for 33.845%, 36.853%, 32.74%, and 41.622% of the total effect, respectively.

The above conclusions could be interpreted as promoting the development of DIF could promote regional employment and venture, promote demand, stimulate spending, and at the same time, form a financial agglomeration effect and alleviate financial mismatches, thereby improving the level of regional economic development. With the improvement of the regional economic development level, the construction of digital infrastructure in rural areas is further improved. The high coverage of the mobile communication network provides a guarantee for DIF to exert its digital advantages, shortens the distance between rural residents and digital financial services, and is conducive to improving the poverty alleviation efficiency of digital finance. In addition, when economic development is at a high level, the degree of inclusive education also increases, which creates favorable conditions for rural residents to accumulate human capital and exert their own “blood production” ability. On the whole, regional economic development has an intermediary effect in the process of DIF and three sub-dimensions, alleviating the poverty of rural residents.

4.4.2. DIF, the Urban-Rural Income Gap, and the Poverty Alleviation Level of Rural Residents

Table 8. DIF and poverty alleviation level of rural residents: a test of the mediating effects of income disparity.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	Theil	lnPCDI	Theil	lnPCDI	Theil	lnPCDI	Theil	lnPCDI
LnDIF	−0.017 ***	0.223 ***
	(−8.24)	(17.81)
LnDIF_B			−0.014 ***	0.162 ***
			(−8.94)	(15.07)
LnDIF_D					−0.014 ***	0.213 ***
					(−6.13)	(15.77)
LnDIG							−0.010 ***	0.125 ***
							(−5.56)	(9.80)
Theil		−2.983 ***		−2.974 ***		−3.883 ***		−4.564 ***
		(−8.29)		(−7.41)		(−10.64)		(−10.47)
Control variable	Control	Control	Control	Control	Control	Control	Control	Control
_cons	0.308 ***	6.852 ***	0.301 ***	6.681 ***	0.339 ***	7.106 ***	0.396 ***	6.322 ***
	(6.22)	(23.30)	(6.22)	(20.84)	(6.48)	(22.37)	(7.90)	(16.78)
N	300	300	300	300	300	300	300	300
adj. R2	0.463	0.917	0.483	0.901	0.403	0.905	0.388	0.862
F	44.495	431.052	47.789	355.125	36.114	373.271	34.262	244.547
Sobel test	0.0778 ***		0.0668 ***		0.0815 ***		0.0613 ***
	(z = 6.147)		(z = 6.347)		(z = 6.046)		(z = 4.34)
Goodman-1 test	0.0778 ***		0.0668 ***		0.0815 ***		0.0613 ***
	(z = 6.137)		(z = 6.335)		(z = 6.036)		(z = 4.333)
Goodman-2 test	0.0778 ***		0.0668 ***		0.0815 ***		0.0613 ***
	(z = 6.158)		(z = 6.36)		(z = 6.056)		(z = 4.348)
Intermediary effects	21.917%		23.807%		21.776%		23.284%

Note: *** indicates significance at the 1% level.

presents the outcomes of the mediation effect test with theil as the mediating variable. Models (1), (3), (5), and (7) examine the regression results of DIF and the three sub-dimensions of theil. Models (2), (4), (6), and (8) examine the mediating role of theil as a mediating variable. The regression outcomes present that in the process of DIF promoting poverty alleviation in rural residents, with theil as the mediating variable, there is a significant mediating effect. Similarly, when the three sub-dimensions of DIF promote poverty alleviation in rural residents, with theil as the mediating variable, the mediating effect is also significant. The Soble test shows that the mediating effect of DIF and its three sub-dimensions, DIF_B, DIF_D, and DIG, account for 21.917%, 23.807%, 21.776%, and 23.284% of the total effect, respectively. These regression outcomes show that the narrowing of income disparity plays an important role in the process of DIF helping rural residents to alleviate poverty. On the whole, the outcomes of the tests based on the mediating effects clearly reveal the path of poverty alleviation in China, i.e., the development of DIF, the reduction in income disparity, and the alleviation of poverty among rural residents.

4.5. Expanded Research

According to the results of the regional heterogeneity research, in the eastern area with a higher degree of economic growth, DIF plays a stronger role in increasing income and reducing poverty. That is to say, the impact of DIF on the poverty alleviation of rural residents might be related to the level of economic growth. During economic growth at different levels, the effect of DIF on increasing income and reducing poverty may be different. Therefore, this paper constructs the following panel threshold model to further verify whether the effect of DIF on poverty alleviation of rural residents has a threshold effect based on the level of economic development. It can be seen from

Table 9. Outcomes of the test for the threshold effects of the regional economic development level.

Threshold Variables	Models	F-Value	p-Value	Number of BS	1% Threshold	5% Threshold	10% Threshold
Total digital inclusive finance index	Single threshold	43.00 ***	0.010	500	42.8889	36.5331	31.8878
	double threshold	34.13 ***	0.006	500	28.2859	22.0959	20.1348
	three threshold	30.08	0.992	500	138.126	119.1617	107.0717
Breadth of coverage	Single threshold	55.89 ***	0.000	500	37.4749	28.1436	24.6416
	double threshold	44.11 ***	0.000	500	28.8145	21.3555	18.5889
	three threshold	35.39	0.924	500	105.324	88.0746	81.6542
Depth of use	Single threshold	41.78 ***	0.002	500	38.9808	30.2118	27.5251
	double threshold	30.98 **	0.018	500	32.9611	25.7827	22.1627
	three threshold	25.72	0.560	500	97.9315	86.6842	77.2437
Degree of digitization	Single threshold	47.75 ***	0.010	500	46.272	36.5125	33.6711
	double threshold	32.24 **	0.042	500	39.1731	31.0936	27.0367
	three threshold	36.64	0.714	500	125.2511	104.4925	95.1005

Note: The number of BS is the number of threshold self-sampling. **, and *** indicate significance at the 5%, and 1% levels, respectively.

that DIF and its three sub-dimensions passed the double-threshold effect test, however, they did not pass the triple-threshold test; DIF and each dimension index have double-threshold effects.

Table 10. Threshold estimation results.

Threshold Variables	Threshold Estimates	95% Confidence Interval
Total digital inclusive finance index	First threshold estimate: 10.6127	10.5918	10.6275
	Second threshold estimate: 11.2599	11.2288	11.3129
Breadth of coverage	First threshold estimate: 10.6127	10.5918	10.6275
	Second threshold estimate: 11.2599	11.2288	11.3129
Depth of use	First threshold estimate: 10.4332	10.4152	10.4403
	Second threshold estimate: 11.2599	11.2288	11.3129
Degree of digitization	First threshold estimate: 10.4332	10.4152	10.4403
	Second threshold estimate: 10.6717	10.6275	10.6783

reports the threshold estimation results. Table 10 and

Table 11. Regression results of the threshold effects.

Variables	(1)	(2)	(3)	(4)
	LnPCDI	LnPCDI	LnPCDI	LnPCDI
LnDIF_1(RGDP ≤ 10.6127)	0.235 ***
	(18.04)
LnDIF_2(10.6127 < RGDP ≤ 11.2599)	0.256 ***
	(19.05)
LnDIF_3(11.2599 < RGDP)	0.285 ***
	(22.39)
LnDIF_B1(RGDP ≤ 10.6127)		0.174 ***
		(8.63)
LnDIF_B2(10.6127 < RGDP ≤ 11.2599)		0.198 ***
		(10.39)
LnDIF_B3(11.2599 < RGDP)		0.233 ***
		(12.92)
LnDIF_D1(RGDP ≤ 10.4332)			0.193 ***
			(9.81)
LnDIF_D2(10.4332 < RGDP ≤ 11.2599)			0.224 ***
			(11.79)
LnDIF_D3(11.2599 < RGDP)			0.256 ***
			(13.72)
LnDIG1(RGDP ≤ 10.4332)				0.106 ***
				(4.88)
LnDIG2(10.4332 < RGDP ≤ 10.6717)				0.138 ***
				(7.24)
LnDIG3(10.6717 < RGDP)				0.159 ***
				(8.73)
Control variables	Control	Control	Control	Control
Constant	6.350 ***	6.332 ***	6.033 ***	5.842 ***
	(17.61)	(15.86)	(15.45)	(12.47)
Observations	300	300	300	300
R-squared	0.930	0.925	0.906	0.867
F	308.1	280.1	202.0	129.3

Note: *** indicates significance at the 1% level, whereas the brackets represent the t values.

show that when DIF and its three sub-dimensions are used as the independent variables, the regional economic development level has a double threshold effect. Therefore, this study utilizes the threshold regression model for empirical analysis, and the outcomes are shown in Table 11. Model (1) represents the regression outcomes when the DIF index is the core explanatory variable; when the RGDP is ≤ 10.6127, the regression coefficient of DIF is 0.235, which is significantly positive at the 1% level. When 10.6127 < RGDP ≤ 11.2599, the regression coefficient is 0.256, which is significantly positive. When RGDP is > 11.2599, the regression coefficient is 0.285, which is significantly positive. This means that the income-increasing and poverty-reducing effect of DIF increases with the level of economic development. Models (2)–(4) report the regression results when DIF_B, DIF_D, and DIG are the core explanatory variables, respectively, which are consistent with the above conclusions from the data in the table. The higher the level of economic growth, the stronger the promotion impact of DIF_B, DIF_D, and DIG on the poverty alleviation of rural residents. The threshold regression model once again proves that the level of regional economic growth is one of the key factors affecting the impact of DIF on decreasing poverty and increasing income.

5. Research Conclusions and Recommendations

DIF has been developing rapidly over recent years. The vigorous progression of DIF has improved the coverage and availability of financial services and provided convenient financial services to all segments of society, especially the rural poor who have been excluded from traditional finance. Using China’s provincial panel data from 2011 to 2020, this study empirically tested the linear effect of DIF development on the poverty alleviation level of rural residents and employed the mediating effect model and the threshold model to validate the role of DIF on the poverty alleviation level of rural residents. The conclusions are as follows: (1) DIF and its three sub-dimensions have significant effects on increasing income and reducing poverty. Among the three sub-dimensions, DIF_D plays the predominant role, followed by DIF_B, and then DIG; (2) The regional heterogeneity study found that DIF and its sub-dimension DIF_B and DIF_D in the eastern area have the greatest effect on poverty alleviation for rural residents, and the effects on the central and western areas are weak in turn; (3) The mediating effect test found that DIF could indirectly promote rural poverty alleviation by facilitating regional economic development and narrowing income disparity. The Sobel test showed that the mediating effect of regional economic growth is better than the mediating effect of income disparity; (4) It was found through the threshold effect test that regional economic growth has a double threshold effect on rural poverty alleviation, and as the threshold value continues to increase, the poverty reduction effect increases in turn.

Founded on the research outcomes, the following recommendations are proposed: First, we must improve the construction of rural digital infrastructure, accelerate the progression of DIF in rural regions, and help rural residents alleviate poverty. The government should perform policies rationally to guide the information industry and financial institutions to participate in digital construction in rural areas. Simultaneously, increasing financial assistance for the construction of informatization in rural areas, strengthening the construction of rural network infrastructure, and promoting the development of DIF in rural areas is essential. Second, implementing regionally differentiated poverty reduction strategies according to local conditions is important. For the more developed eastern regions, the advantages of their economic development and digital environment should be fully considered, and take the innovation of financial tools and improvement of financial service efficiency as the development focus of DIF. Meanwhile, it would provide full play to the regional role of the center and help other regions to promote poverty reduction. For the undeveloped central region, the government and financial institutions could increase digital inclusive financial services, establish industrial poverty alleviation projects, and use policy advantages to guide enterprises and financial institutions to participate in poverty alleviation industries so as to meet the technical and financial needs of industrial development effectively. For the western region with serious poverty accumulation and a lack of financial resources, poverty alleviation should be based on development. Formulating support policies and promoting the deep integration of characteristic regional industries and financial institutions creates a sustainable long-term poverty alleviation mechanism. Using digital technology and internet financial platforms promotes the effective connection of supply and demand information and enhances the added value of products and the market competitiveness of the industry to accomplish hematopoietic poverty alleviation in the western region. Finally, promoting regional economic development and narrowing income disparity play the indirect role of DIF in poverty reduction. Combining the comparative advantages of the region, developing characteristic industries, and promoting industrial transformation to improve the level of economic growth is essential. Simultaneously, coordinating the development of urbanization and rural revitalization to narrow income disparity further, provide full play to the poverty reduction effect and spillover effect of urbanization, and advance high-quality rural development, thus improving the earnings level of poor groups and realizing poverty alleviation is recommended.

This paper also has some limitations. On the one hand, although the per capita net income of rural residents reflects the poverty level of rural residents to a certain extent, it cannot fully reflect the poverty of rural residents, and there are certain limitations. A more comprehensive poverty indicator system could be constructed in future research. On the other hand, the research sample data in this paper is limited to China. In the future, data samples from more countries around the world could be added to increase the influence of the research results and help more countries alleviate poverty. In addition, in recent years, the rapid development of DIF has aroused an upsurge of research by scholars from all over the world. The study found that DIF is not only related to poverty alleviation but also plays an important role in residents’ consumption [8,32], residents’ employment [38], economic development [40], sustainable environmental management [48], enterprise innovation [49], and environmental protection [50]. In the future, global scholars could further expand the scope of research and thus provide full play to the contribution of DIF to the sustainable development of humankind.