Introduction

China’s Hukou policy in the 1960s applied diverse identification and management policies to residents from urban and rural areas. Urban areas derive benefits from the unbalanced social resources that attract more rural residents, further accelerating the process of urbanization and exacerbating urban-rural disparities (He et al., 2016). Meanwhile, traditional financial resources are reluctant to favor the relatively vulnerable rural economies, which impedes rural economic development. The persistent inequality between rural and urban areas and the stagnant economic growth in rural areas have jeopardized the sustainable development of society (Wu & Treiman, 2007). Still, the emergence of digital financial inclusion has brought a transformative breakthrough. Digital financial inclusion incorporates digitization and inclusive finance, which emphasizes extending the scope of financial services and penetrating individuals or regions that traditional finance can’t serve (Hasan et al., 2022). As a result, digital financial inclusion enhances the openness and inclusiveness of the financial system and becomes a potential opportunity for impoverished regions to overcome financial isolation.

For the intent of clearly defining the significance of digital financial inclusion for economic development, the following research questions should be discussed: How does digital financial inclusion affect economic development in urban and rural areas? What is the effect of digital financial inclusion on the urban-rural disparity? Can digital financial inclusion’s impact diffuse across space?

In accordance with the above questions, we explore the provincial data from 2011 to 2019 and apply non-spatial and spatial panel models to specify the effects of digital financial inclusion on economic development and urban-rural inequality. Notable contributions of this research include: first, our work breaks new ground in constructing provincial-level urban and rural GDP per capita data by processing daytime and nighttime light-adjusted satellite images. Second, this research investigates the strengths of digital financial inclusion over traditional finance from both non-spatial and spatial perspectives, which complements the missing spatial explorations.

This paper also contains the following sections. Section 2 illustrates the previous literature. Section 3 demonstrates the variable selection and data processing. Section 4 explains the methodology, and Section 5 presents the empirical results. Last, Section 6 provides the conclusion and relevant suggestions.

Literature review

Financial development theory has stressed that finance can enhance the efficiency and depth of social financing and further boost economic development (Mayer, 1990; Fry, 1989). But for financially excluded areas and individuals, finance fails to work (Leyshon & Thrift, 1995; Salampasis & Mention, 2018). The combination of inclusive finance with the Internet has dramatically expanded financial services coverage and advanced financial product diversity, solving the information asymmetry issues and directly promoting economic growth (Loubere, 2017; Chen, 2016). Multi-country studies proved that digital financial inclusion is crucial for sustainable economic growth in low-income countries suffering from financial exclusion (Tay et al., 2022; Ozturk & Ullah, 2022; Geng & He, 2021). Furthermore, previous examinations of Chinese provinces found that diversified digital financial products can positively impact regional economic growth in remote and underdeveloped western regions (Liu et al., 2021; Ahmad et al., 2021; Sun & Tang, 2022) and reduce regional economic quality heterogeneities (Lv et al., 2022). In particular, digital financial inclusion can facilitate farmers’ application rate of agricultural techniques, which benefits rural economic growth (Ma & Li, 2021; Zhou et al., 2022; Xu & Wang, 2023). However, some studies worried that digital financial inclusion would attract more financial activities from the informal sector and can’t actually help financially excluded areas or individuals (Yu & Wang, 2021; Lai et al., 2022).

In addition to contributing to economic growth, financial development can also affect income inequality. The financial Kuznets curve (Greenwood & Jovanovic, 1990) emphasizes that at the beginning of the economic development process, financial development exacerbates inequality by favoring higher income groups. As financial development deepens, the intermediation and inclusiveness of finance redistributes capital resources, thereby improving income disparities. The findings of multi-country research validated that the inclusiveness of finance is especially profitable to low-income regions (Park & Shin, 2017; Fouejieu et al., 2020). The provincial-level investigations in China further verified that financial inclusion is conducive to narrowing inter-regional income inequality within rural regions (Liu et al., 2017) and mitigating the urban-rural gap (Wang et al., 2023). Digital financial inclusion, on the other hand, elevates financial inclusion to a more sophisticated level. The provincial studies have discussed that digital financial inclusion could improve rural residents’ income by supplementing access to loans (Ji et al., 2021; Li et al., 2021; Song & Guo, 2017). Additionally, digital financial inclusion can also reduce the urban-rural income disparity through accelerating industrial structural upgrading (Liu et al., 2023; Li et al., 2022).

More crucially, new economic geography theory has suggested the vital impact of spatial dependence on regional disparities (Schmutzler, 1999; Krugman, 1998) due to the inter-regional interaction and competition for financial resources (Panov et al., 2019). The country-level spatial investigations indicated that digital financial inclusion has diverse spatial spillover effects on economic development in multiple countries (Shen et al., 2021). Provincial exploration in China further discussed the more valuable spatial spillover impacts of digital financial inclusion on western areas (Ding et al., 2022; Song et al., 2021). Also, although rural areas are not geographically advantageous compared to urban areas, most research highlighted the effectiveness of digital financial inclusion’s spillover effects in enhancing rural residents’ income and reducing farmers’ income vulnerability (Wang et al., 2023; Wang & He, 2020; Li et al., 2022).

According to the above literature, regional studies related to urban and rural areas predominantly focus on the inhabitant’s income due to the unavailability of urban and rural GDP per capita data. As a result, we will measure the rural and urban region’s GDP per capita to fill in the missing research field by processing satellite images. In addition, most of the past research has mentioned that digital financial inclusion is more beneficial for economic development than traditional finance. It’s crucial to recognize that the comparison between digital financial inclusion and traditional finance from the spatial perspective is still insufficient. Thus, the combined non-spatial and spatial investigation will be conducted in this study.

Data

Considering the importance of the MAUP (Modifiable Area Unit Problem), we conducted empirical studies of 31 provinces from 2011 to 2019 to exclude interference from missing values when employing other administrative levels. The following section will describe the applied variables.

Dependent variables

GDP data originate from the revised global 1km \(\times \) 1km grid real GDP images from 2011 to 2019 (Chen et al., 2022). Their data integrate official real GDP and luminosity-adjusted GDP growth rates, which provide a sweeping measurement of economic activities. Figure 1 exhibits the spatial distribution of GDP based on nighttime illumination. The urban and rural classification is obtained from China’s 30m annual land cover images (Yang & Huang, 2021). In accordance with the previous studies, we designate the cropland band as the rural area and the impervious band as the urban area (Gong & Howarth, 1990; Shi et al., 2018; Weng, 2012). An outstanding feature of this classification is that it can dynamically capture the urbanization process in response to actual land cover transitions. Furthermore, based on the provincial vector data from the GADM and the demographic data from the LandScan population global dataset (Bhaduri et al., 2002), we process data using zonal statistics to generate the GDP per capita in urban and rural regions. Last, the urban-rural gap is estimated by taking the ratio of urban GDP per capita to rural GDP per capita.

Fig. 1
figure 1

Luminosity-adjusted GDP in 2011 and 2019

Full size image

Independent variables

The key independent variable is digital financial inclusion, sourced from Peking University’s digital financial inclusion index (Guo et al., 2020). This index is derived from aggregating three primary elements and 33 sub-indicators using the hierarchical analysis.Footnote 1 In the light of traditional finance, the share of total deposits to total loans is computed, and data is collected from regional central banks.

Control variables

Regarding the control variables, the fixed asset investment over GDP is employed to indicate the level of physical capital investment. Furthermore, we calculate the number of patent applications granted per square kilometer to represent technological progress and the number of university graduates per square kilometer to measure educational performance. Last, fiscal expenditure is assigned to evaluate government effectiveness and total foreign direct investment to measure regional trade development level. Data for control variables are obtained from the National Bureau of Statistics and provincial statistical yearbooks. In this study, all variables are logarithmic except for the percentage form variables. Table 1 presents descriptive statistics.

Table 1 Descriptive statistics
Full size table

Methodology

This section centers around the methodological approach to exploring digital financial inclusion’s non-spatial and spatial impact on regional economic development and urban-rural inequality. In terms of the non-spatial analysis, the two-way fixed effect panel model is applied to control the unobserved heterogeneity. Mobile phone penetration rate and the number of Internet ports are selected as the instrumental variables to solve the endogeneity issues (Cheng & Zheng, 2023). Additionally, spatial modeling is performed to consider the spillover effects of digital financial inclusion (Shen et al., 2021). Figure 2 summarizes the methodological framework and detailed explanations are provided in the subsequent section.

Fig. 2
figure 2

The flowchart of methods

Full size image

Spatial weight matrix

Due to Hainan province being geographically isolated from the mainland, we refer to Anselin (2005) suggestion and construct a Thiessen polygon based on the distance from each region’s center point to guarantee a complete spatial structure. The Thiessen polygon is shown as a blue buffer on the original map in Fig. 3. In addition, we adopt the Queen contiguity weight matrix, which is widely used in previous studies to avoid the asymmetry issue of distance weight matrix like the K-nearest neighbor weight matrix.

Fig. 3
figure 3

Thiessen polygon of provinces

Full size image
Table 2 Global Moran’s I
Full size table

Global Moran’s I

Before conducting spatial panel examination, we apply Moran’s I to measure the spatial dependence that the equation is as follows:

$$\begin{aligned} I=\frac{\sum _i \sum _j W_{i j} z_i \cdot z_j / S_0}{\sum _i z_i^2 / n} \end{aligned}$$
(1)

In the equation, \(W_{i j}\) represents the spatial weight, \(S_0\) is the sum of the weight, and \(z_i\) represents the GDP per capita’s deviation from the mean in the region i. The spatial randomness is rejected when Moran’s I is positive and significant (Anselin, 2019), which means there is a similar performance of GDP per capita between regions and neighbors. Based on the Queen contiguity weight matrix, Table 2 indicates that from 2011 to 2019, all dependent variables except urban GDP per capita have significant spatial dependence. Accordingly, we conduct further spatial analysis on GDP per capita, rural GDP per capita, and urban-rural gap.

Spatial Durbin model

To effectively control the unobserved heterogeneity, we first estimate the relationship between variables by controlling the region and time uniqueness. The expression is as follows:

$$\begin{aligned} Y_{i t}=r_{i} +\nu _{t}+\beta X_{it}+\varepsilon _{it} \end{aligned}$$
(2)

In the equation, \(Y_{i t}\) represents the dependent variable at regions i and time t, \(X_{it}\) represents the explanatory variables, \(\nu _{t}\) and \(r_{i}\) represent time-fixed and region-fixed effects, respectively, and \(\varepsilon _{it}\) represents the error term. However, the above non-spatial model omits the spatial dependence and heterogeneity, which will bias the estimation. Therefore, the spatial Durbin model (Elhorst, 2010) is applied, which is shown in the following equation:

$$\begin{aligned} Y_{i t}=\rho W Y_{i t}+\alpha i_n+ \beta X_{it}+\gamma W X_{it} +\varepsilon _{it} \end{aligned}$$
(3)

Unlike the non-spatial model, the SDM includes W for the spatial weight matrix, \(\rho \) is the coefficient of spatial lag of the dependent variable, and \(\gamma \) is the coefficient of spatial lag of the explanatory variable. \(\alpha \) is the constant term multiplied by the \(n\times 1\) vector \(i_n\), and n is the number of regions.

In addition, the likelihood ratio (LR) test and the Wald test in Table 3 suggest that the SDM is the optimal model compared to the SEM and the SAR, whereas the Hausman test prefers the fixed effects model. Taking into account the above indication, the two-way fixed effect SDM is adopted to examine the spatial spillover effect of digital financial inclusion.

Table 3 LR test, Wald test, and Hausman test
Full size table

Effect composition

The spatial Durbin model also effectively captures the direct and spatial spillover effects. We rewrite the equation as follows:

$$\begin{aligned} Y_{i t}=(I-\rho W)^{-1}\left( \alpha i_n+\beta X_{it}+\gamma W X_{it} +\varepsilon _{it}\right) \end{aligned}$$
(4)

To clarify the effects from region 1 to n at the time t, Eq. 5 shows the partial derivative matrix of the model:

$$\begin{aligned} \begin{aligned} \left( \begin{array}{ccc} \frac{\partial Y_1}{\partial X_{1 t}} &{} \ldots &{} \frac{\partial Y_1}{\partial X_{n t}} \\ \ldots &{} \ldots &{} \ldots \\ \frac{\partial Y_n}{\partial X_{1 t}} &{} \ldots &{} \frac{\partial Y_n}{\partial X_{n t}} \end{array}\right) _t&=[I-\rho W]^{-1}\left[ \beta I+W \gamma \right] \end{aligned} \end{aligned}$$
(5)

The direct effect is the diagonal elements of the above matrix, and the indirect impact is the non-diagonal elements of the matrix (LeSage, 2008; Elhorst, 2010). Moreover, direct effect refers to the impact of explanatory variables on local economic development. In contrast, the indirect effect refers to the local explanatory variable’s impact on the surrounding region’s economic development, also called the spillover effect.

Empirical results

Non-spatial panel results

The Hausman test suggests the preference of the fixed-effect model,Footnote 2 so to mitigate the bias of unobserved heterogeneity, we control for both time-fixed effect and region-fixed effect. The empirical results are shown in Table 4. First, digital financial inclusion positively impacts overall economic development, which is even greater than traditional finance. Second, digital financial inclusion and traditional finance can significantly promote economic development in rural areas. However, there’s insufficient evidence to support a significant effect of digital financial inclusion in urban regions, and traditional finance has a significantly negative impact. Last, digital financial inclusion and traditional finance serve to benefit the reduction of urban-rural disparity.

Table 4 Non-spatial panel results
Full size table

The potential explanations of the above results are: first, digital financial inclusion has expanded the reach of financial services to a greater extent at a lower cost, effectively covering rural areas that are hard for traditional financial services to penetrate. Lower-threshold online lending and money funds have substantially enhanced living conditions in rural areas with high concentrations of poverty (Lee et al., 2023), further benefiting the economic development performance of the rural regions. Moreover, online borrowing relieves the financial pressure on small and medium-sized enterprises(SMEs), attracting more investment opportunities for rural areas (Lu et al., 2022). Meanwhile, these investment opportunities also bring more employment opportunities, promote the return of human capital, contribute to upgrading industrial structures, and ultimately profit rural economic development (Yang & Zhang, 2020). During this process, the distribution of resources has also been reallocated, further alleviating the urban-rural gap to a certain extent. Regarding urban areas, less diversified traditional financial services fail to meet the growing demand of investors. In the meantime, higher transaction thresholds and strict risk control are dragging down the economic growth of urban areas (Chen, 2006; Allen et al., 2007).

In the context of other control variables, physical capital investment significantly improves overall economic development. Still, it harms rural economic development and deteriorates the economic development gap between urban and rural regions. The underlying reason is that the expansion of the real estate market bubble is gradually worsening the equal distribution of investment resources (Cheng, 2017), and abundant resources are polarized towards urban areas, which is detrimental to rural economic development. Fiscal expenditure also harms rural regions’ economic development due to the proliferation of local debt in recent years (Pan et al., 2017; Wu et al., 2018). Despite the small magnitude of the technology coefficient, upgrading technological development can contribute to higher labor output and economic efficiency.

Table 5 Robustness test for non-spatial panel model
Full size table

Robustness test of non-spatial panel results

In order to validate the robustness of the above results, we only keep the key independent variables. The results in Table 5 show the consistency of our findings, except the impact of digital financial inclusion on urban-rural inequality is not robust. Consequently, we apply the 2SLS approach and use mobile phone penetration rate and the number of Internet ports as the instrumental variables. The Wald-F statistic indicates the selected instrumental variables’ validity and verifies the reliability of the results. Although digital financial inclusion still reveals an insignificant effect on the urban-rural disparity, we will conduct the spatial examination in the following section.

Spatial panel results and effect composition

Table 6 presents the estimation of SDM and Table 7 offers the effect composition. The results underline the following insights: first, digital financial inclusion has a beneficial spillover impact on promoting rural regions’ economic development, which is stronger than traditional finance. Also, the indirect effect of digital financial inclusion is more substantial than the direct effect. Particularly, digital financial inclusion has a spatial spillover effect on mitigating urban-rural inequality that traditional finance lacks.

Several interpretations of the above results encompass: first, digital financial services such as microcredit and agricultural development funds offer supplementary financing for farmers to invest in agricultural technology innovation (Zhou et al., 2022). With the application and diffusion of agricultural technology, neighboring areas progressively embrace innovative technology, which indirectly profits farmers’ income and contributes to the economic development of adjacent regions. Moreover, with the assistance of digital financial inclusion, the expansion of SMEs has also brought employment opportunities to the surrounding regions, aiding in the furtherance of human capital resource return to the rural areas (Zhou et al., 2015). Commencing with the redistribution of financial resources, digital financial inclusion indirectly reduces the imbalance of technical and human resources between rural and urban areas, eventually relieving the economic development gap (Liu et al., 2023).

Table 6 Spatial panel results
Full size table
Table 7 Effect composition
Full size table

Regarding the control variables, physical capital investment has a negative spatial spillover impact on rural economic development. The process of fixed asset construction attracts the accumulation of investment and human capital, leading to resource insufficiency in the surrounding areas and harming their economic development (Glaeser et al., 2017). Technology innovation has an advantageous spatial spillover impact on urban-rural gap reduction. It is attributable to the fact that technological developments and innovations have effectively enhanced the outcome of economic activities. Although urban areas have the advantage of human capital and infrastructure, over-saturated markets and high costs restrict technology diffusion. However, rural areas have underlying strengths of lower costs, and technology-led investments can accelerate the upgrading of industrial structures in rural areas, mitigating the urban-rural gap (Jiang et al., 2022; Wang et al., 2023).

Robustness test of spatial analysis

With the aim of guaranteeing the above results are reliable, we switch the rook contiguity weight matrix and K-nearest (K=5) neighbor weight matrix. Both the spatial panel results and the effect composition remain robust.Footnote 3

Conclusion

By examining provincial data from 2011 to 2019, this paper clarifies digital financial inclusion’s non-spatial and spatial impact on regional economic development and urban-rural disparity. The key conclusions are as follows: first, digital financial inclusion significantly contributes to rural economic development. The low-cost and low-threshold financial services provided by digital financial inclusion effectively address loan limitations caused by traditional finance to farmers and SMEs, further attracting investment for agricultural technology advancement and enterprise expansion. In addition, digital financial inclusion has more substantial spatial spillover effects on rural economic development than traditional finance. Digital financial inclusion accelerates technology spillovers and enhances economic quality in local and neighboring regions, which creates mutually beneficial outcomes.

Last, digital financial inclusion has advantageous spatial spillover effects in mitigating urban-rural economic development imbalance. Digital financial inclusion adequately harmonizes the skewed distribution of financial resources between urban and rural areas but also substantially promotes industrial upgrading through the reconfiguration of technological and human resources, ultimately lessening the deterioration of the urban-rural gap. Therefore, promoting digital financial inclusion in rural areas merits the government’s consideration.

Taking into account the above conclusion, this study suggests the following policy recommendations. In view of the remarkable effects of digital financial inclusion in rural areas and inequality reduction, the government should delve into providing the necessary foundation for developing the demand side of digital financial inclusion. More specifically, the government should proceed to construct ICT infrastructure in rural regions while elevating the popularization and education of financial services to the public and raising rural residents’ adoption rate of digital financial services.

Second, regional cooperation is conducive to the diffusion of the spillover effect of digital financial inclusion. As a result, the government can artificially stimulate the supply side by creating financing funds to attract investment from the digital financial market for regional cooperation centered on fintech upgradation. The fintech-led collaboration will further accelerate the cross-regional spillover effects of technological advancement, indirectly enhancing the outcome of digital financial inclusion.

Last, in consideration of the mismatch between traditional finance and economic growth in urban regions, it is crucial for policymakers to loosen restrictions appropriately. The government can minimize artificial intervention in the financial market within the scope of risk control, promoting financial product innovation and lowering the financial entry threshold to satisfy the growing financial demands in urban areas.

Despite the valuable insight provided in this study, we also admit the following constraints. The scope of this study is limited to China, so comparative research with other countries will be the potential future work. Besides the urban-rural division, the geographical heterogeneity of digital financial inclusion is worth investigating in future research, which will aid in thoroughly characterizing the role of digital financial inclusion in regional development.