Digital Inclusive Finance and Government Spending Efficiency: Evidence from County-Level Data in China’s Yangtze River Delta

Abstract

Amid the global drive to enhance public sector performance in the digital economy era, improving government spending efficiency has become a critical governance objective. This study investigates the impact of digital inclusive finance on government spending efficiency from a digital finance systems perspective using county-level panel data in China’s Yangtze River Delta for the period 2014–2022 and constructing the fixed-effects model and instrumental variable method to estimate the effect of digital inclusive finance and explore its underlying mechanisms. Heterogeneity across regions with varying economic development levels is analyzed, and fiscal pressure is examined as a potential mediating factor. The results indicate that (1) digital inclusive finance significantly enhances government spending efficiency, primarily through broad service coverage and deep usage of digital financial services such as mobile payments, digital credit, and insurance; (2) the positive effect is more pronounced in counties with lower government spending efficiency and economic development; and (3) fiscal pressure acts as a key transmission channel, with broader digital inclusive finance coverage helping to alleviate fiscal stress and improve government spending efficiency. These findings offer empirical insights into the role of digital finance in promoting effective and adaptive public financial governance.

1. Introduction

Government spending plays a fundamental role in ensuring the provision of social security and public services. However, in many regions, particularly in parts of Africa, rising levels of public expenditure have not yielded proportional improvements in human development indicators [1]. This discrepancy underscores a critical issue: the efficiency with which public funds are allocated and utilized may significantly influence the developmental effectiveness of government spending. In the context of the rapidly evolving digital economy, this study explores the potential of digital inclusive finance as a strategic instrument to enhance government spending efficiency. By doing so, it aims to provide valuable implications for policymakers seeking to improve fiscal governance and advance inclusive, high-quality development.

Digital inclusive finance refers to a technology-enabled financial system that delivers accessible, affordable, and user-centric services to underserved populations through mobile platforms, big data, cloud computing, and artificial intelligence. Conceptually, digital inclusive finance comprises three dimensions: (1) coverage breadth, denoting the geographical and demographic reach of digital financial services; (2) usage depth, reflecting engagement intensity across payment, credit, insurance, and investment domains; and (3) digitalization level, indicating the sophistication and automation of financial infrastructure. Unlike traditional inclusive finance, digital inclusive finance operates in real-time and is data-driven, offering enhanced scalability, responsiveness, and potential applications for dynamic public governance—such as precision subsidies and risk monitoring.

Parallel to this, government spending efficiency—defined as maximizing output at a given fiscal input or minimizing input for a fixed output level [2] has become an active research area. Foundational studies date back to the late 19th and early 20th centuries [3,4,5]. Recent empirical work has employed data envelopment analysis (DEA), stochastic frontier analysis (SFA), and other econometric approaches to measure efficiency and explore its determinants [6,7]. Key influencing factors include fiscal decentralization, administrative incentives, digital government initiatives, and economic scale [7,8,9].

In the era of digital transformation, researchers have begun exploring the implications of digital technologies for public sector performance. E-government platforms, for example, can improve tax administration, streamline procurement, and enhance service delivery [10]. While such studies mainly focus on intra-governmental efficiencies, digital inclusive finance presents a novel perspective—by bridging financial systems with citizen engagement and public service delivery, digital inclusive finance fosters co-governance models involving markets and governments alike.

A growing body of literature has examined digital inclusive finance’s macro-level impacts, including economic growth and social development [11,12], and carbon emission reduction [13]. Digital inclusive finance has been shown to lower local government debt risk [14,15], alleviate relative poverty [16], and expand spatial mobility, supporting inclusive development [17,18,19]. However, there is still a critical gap: the relationship between digital inclusive finance and government spending efficiency has not been rigorously quantified or systematically explored. Most existing studies are either conceptual or do not directly engage with fiscal efficiency outcomes at the local level.

To address this gap, this study seeks to answer the following research questions:

• RQ1: Does digital inclusive finance enhance government spending efficiency at the county level?
• RQ2: Through which mechanisms—particularly fiscal pressure—does digital inclusive finance influence government spending efficiency?
• RQ3: Does the impact of digital inclusive finance on government spending efficiency vary across regions?

This paper makes three main contributions. First, it extends the literature on digital finance and public governance by introducing a system-oriented framework that links digital inclusive finance with fiscal efficiency, moving beyond traditional analyses of economic output or poverty. Second, it provides a county-level empirical assessment—employing the SBM–GML method to capture county heterogeneity. Third, it enriches the policy discourse by revealing context-sensitive mechanisms, offering evidence-based insights for adaptive governance in digitally evolving economies.

2. Research Design

2.1. Theoretical Framework and Hypotheses

Within a systems-theoretic framework, digital inclusive finance platforms, information flows, and decision-making modules jointly form a ‘digital–fiscal–social’ coupled system that shapes the synergy and stability of the public expenditure system. Digital inclusive finance represents a critical concept within financial technology (fintech) and serves as a subset of digital finance. The primary focus of this concept is the utilization of digital technology to facilitate equitable access to financial services, particularly for marginalized populations that are often excluded from the traditional financial system. Consequently, digital inclusive finance is characterized by its emphasis on economic empowerment and social inclusiveness [20]. This enables digital inclusive finance to facilitate the establishment of a complete credit risk assessment system. The market is better equipped to accurately identify enterprise information, which improves the efficiency of fiscal spending to foster the growth of the real economy [21].

Moreover, digital inclusive finance represents a significant avenue for financial institutions to cultivate their operational efficiencies and improve their social governance capabilities, thereby possessing the potential to reinforce social governance [11]. The enhancement of overall social welfare directly results from improved government spending efficiency. First, digital inclusive finance offers comprehensive financial services via digital platforms, addressing the diverse financial needs of various demographic groups. The increased accessibility and reach of financial services respond to individuals’ ongoing aspirations for an improved quality of life, thereby fostering overall social welfare across multiple dimensions, including health, education, insurance coverage, and other relevant areas [22]. Second, digital inclusive finance has the potential to improve the efficiency of financial services and offer liquidity support for the advancement of the real economy. This, in turn, can foster entrepreneurial ventures and create employment opportunities, stimulating economic growth and contributing positively to enhancing social security systems [23]. Thirdly, digital inclusive finance diminishes the reliance on cash transactions. This shift complicates criminals’ ability to acquire substantial sums of illicit funds through theft, thereby reducing the overall crime rate [24,25]. Finally, digital inclusive finance, functioning as an information intermediary, has leveraged technologies like the Internet, big data, and artificial intelligence. These advancements facilitate the rapid capture, integration, and analysis of substantial volumes of financial data. Furthermore, they contribute to establishing a digital platform characterized by well-defined rights and responsibilities, thereby supplying the government with comprehensive information pertinent to social governance [11]. Information transparency facilitates the timely formulation and adjustment of government policies, improving the government’s governance capacity. It also reinforces the scientific basis of fiscal decision-making, supports the advancement of public initiatives, and contributes to increasing the efficiency of financial spending at the local government level.

Therefore, this paper proposes:

Hypothesis 1:

Digital inclusive finance can improve government spending efficiency.

Furthermore, digital inclusive finance has the potential to alleviate fiscal pressures on government entities, enhance government expenditure levels, and improve the efficiency of public spending. By lowering financing costs and access barriers, digital inclusive finance offers greater financial convenience and essential support to enterprises and individuals. It can also transform the avenues for economic activities, including entrepreneurship and innovation, through fiscal support from local governments, thereby reducing the fiscal burdens these local authorities face [16,26]. Meanwhile, digital inclusive finance can improve credit institutions’ risk management capabilities by offering various insurance and risk management instruments. This enhancement may diminish the general populaces and businesses’ reliance on local government debt relief measures [14,15]. The reduction in local government debt, coupled with a decrease in fiscal pressure, allows for a greater allocation of expenditures towards infrastructure development, the provision of public services, and other related areas. This, in turn, enhances the efficiency of fiscal spending by elevating the overall fiscal outlay [26].

Therefore, this paper proposes:

Hypothesis 2:

Fiscal pressure is a mediating factor in the relationship between digital inclusive finance and the efficiency of government spending. This relationship is characterized by the capacity of digital inclusive finance to alleviate fiscal pressure, thereby enhancing the efficiency of government spending.

The existing literature indicates that the socio-economic impacts of digital inclusive finance exhibit regional variability. According to economic opportunity theory, regions characterized by higher economic development are generally associated with increased employment and entrepreneurial prospects [24]. When job opportunities in the labor market align with the population’s demands, social security assurance is enhanced, and government spending efficiency is significantly improved. Under these circumstances, the marginal impact of the favorable advancements in digital inclusive finance may begin to wane. Conversely, in areas characterized by lower economic development and reduced efficiency in government spending, the initial level of digital inclusive finance is correspondingly low. An enhancement in the level of digital inclusive finance will likely exert a more pronounced influence on the government’s capacity for governance and fiscal pressures. The contribution of digital inclusive finance to improving government spending efficiency is expected to be significantly amplified.

Therefore, this paper proposes:

Hypothesis 3:

The impact of digital inclusive finance on government spending efficiency varies by region.

2.2. Model Construction

To study the impact effect of digital inclusive finance on government spending efficiency, the following measurement model is constructed:

$${GE}_{it}={\alpha }_0+{\alpha }_1{DIF}_{it}+{\alpha }_2\sum {Control}_{it}+{\gamma }_t+{\theta }_i+{\epsilon }_{it},$$

(1)

Among them, $i$ and $t$ represent the county and the year, respectively; the level of government spending efficiency is indicated and measured by the SBM–GML index. ${\alpha }_0$ is a constant term; ${DIF}_{it}$ represents the level of digital inclusive finance, the county-level digital inclusive finance index is adopted for measurement. ${\alpha }_1$ is the coefficient of the core explanatory variable focused on in this study. If it is greater than 0, it indicates that digital inclusive finance can improve government spending efficiency. If it is less than 0, it indicates that digital inclusive finance will inhibit government spending efficiency. If it is equal to 0, it indicates that digital inclusive finance has no relation to government spending efficiency. ${\alpha }_2$ is the coefficient of the control variable. ${Control}_{it}$ is the selected control variable. ${\gamma }_t$ is the fixed effect of the year. ${\theta }_i$ is the fixed effect at the county level. ${\epsilon }_{it}$ is a random disturbance term.

In addition, to evaluate the validity of the research Hypothesis 2, with fiscal pressure serving as a mediating variable, the subsequent model has been developed:

$${MED}_{it}={\beta }_0+{\beta }_1{DIF}_{it}+{\beta }_2\sum {Control}_{it}+{\gamma }_t+{\theta }_i+{\epsilon }_{it},$$

(2)

$${GE}_{it}={\alpha }_0+{\alpha }_1{DIF}_{it}+{\alpha }_2{MED}_{it}+{\alpha }_3\sum {Control}_{it}+{\gamma }_t+{\theta }_i+{\epsilon }_{it},$$

(3)

Among them, ${MED}_{it}$ represents the mediating variable, namely fiscal pressure. The coefficient ${\beta }_1$ represents the direct impact of digital inclusive finance on the mediating variable. Both the mediating variable and digital inclusive finance are incorporated into the equation of Formula (3). If both ${\beta }_1$ and ${\alpha }_2$ are significant simultaneously, it indicates that the mediating effect holds. If it is not significant, Equation (3) will no longer be executed.

2.3. Variable Selection and Measurement

The explained variable is government spending efficiency (GE). This study draws upon the existing body of literature [10,27] to assess the efficiency of government spending at the input–output level. The analysis employs the SBM directional distance function alongside the GML index to evaluate the dynamic variations in government spending efficiency across two distinct periods, with 2014 as the baseline year. A GML index value exceeding 1 signifies an enhancement in the efficiency of government expenditure relative to the 2014 benchmark. In contrast, a GML index value below 1 indicates a decline in efficiency compared to the same reference year. The input indicators for calculating government spending efficiency are each county’s annual per capita fiscal spending. The output indicators cover areas such as economic development, educational resources, and medical and healthcare, specifically including the per capita regional GDP at the county level, the average value of the night light index, the proportion of students in ordinary middle schools, the proportion of students in ordinary primary and secondary schools, and the per capita number of medical beds. The construction and methodology for measuring government spending efficiency, as summarized in

Table 1. Indicator system and descriptions for measuring government spending efficiency.

	Variable	Obs	Mean	Std. Dev.
Input indicator	Per capita fiscal expenditure	1520	9032.482	5763.457
Output indicators of economic development	Per capita regional GDP	1520	64,026.680	51,038.430
	The average value of the night light index	1520	16.516	13.735
Indicators of educational resource output	The proportion of students currently enrolled in regular middle schools	1520	414.607	97.378
	The proportion of students enrolled in regular primary and secondary schools	1520	613.056	216.033
Medical and health output indicators	The number of medical beds per capita	1520	45.636	16.400

, are as follows.

According to the existing literature [28], the measurement steps are as follows:

(1)

Assuming each county is a decision-making unit, and each decision-making unit contains $N$ kinds of input factors $x_{in}=\left(x_{i1}{, x}_{i2},\dots ,{ x}_{iN}\right)\in R_N^{+}$, we can obtain $M$ kinds of expected outputs $y_{im}=\left(y_{i1}{, y}_{i2},\dots ,{ y}_{im}\right)\in R_M^{+}$, and $K$ kinds of unexpected outputs $b_{ik}=\left(b_{i1}{, b}_{i2},\dots ,{ b}_{ik}\right)\in R_K^{+}$; hence, the global production possibility set is constructed as Equation (4):

$$P^G(x)=\left\{\begin{array}{c}\left(y^t,b^t\right):\sum _{t=1}^T\sum _{i=1}^I{\beta }_i^t{y}_{im}^t\ge y_{im}^t,\forall m;\sum _{t=1}^T\sum _{i=1}^I{\beta }_i^t{b}_{ik}^t,\forall k\\ \sum _{t=1}^T\sum _{i=1}^I{\beta }_i^t{x}_{in}^t\le x_{in}^t,\forall n;\sum _{t=1}^T\sum _{i=1}^I{\beta }_i^t=1,{\beta }_i^t\ge 0;\forall i\end{array}\right.$$

(4)

(2)

$\left(x^{t,i},y^{t,i},b^{t,i}\right)$, $\left(g^x,g^y,g^b\right)$, and $\left(s_n^x,s_m^y,s_k^b\right)$ represent the input–output, direction, and relaxation vector, respectively, and $z_i^t$ is the weight of each cross-section. Global SBM directional distance functions are as follows in Equation (5):

$$\begin{array}{c}{S}_v^G\left(x^{t,i},y^{t,i},b^{t,i},g^x,g^y,g^b\right)\hfill \\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}=max{\displaystyle \frac{\frac{1}{N}{\sum }_{n=1}^N\frac{s_n^x}{g_n^x}+\frac{1}{M+K}\left({\sum }_{m=1}^M\frac{s_m^y}{g_m^y}+{\sum }_{k=1}^K\frac{s_k^b}{g_{ik}^b}\right)}{2}}\end{array}$$

(5)

$$s.t.{\sum }_{t=1}^T{\sum }_{i=1}^I{z}_i^t{x}_{in}^t+s_n^x=x_{in}^t, \forall n;$$

$${\sum }_{t=1}^T{\sum }_{i=1}^I{z}_i^t{y}_{im}^t-s_{nm}^y=y_{in}^t, \forall m;$$

$${\sum }_{t=1}^T{\sum }_{i=1}^I{z}_i^t{b}_{ik}^t+s_i^b=b_{ik}^t, \forall i;$$

$${\sum }_{i=1}^I{z}_i^t=1, z_k^t\ge 0,\forall i;$$

$$s_m^y\ge 0,\forall m;s_i^b\ge 0,\forall i$$

(3)

The Malmquist–Luenberger (ML) index has some unsolvable linear programming problems, but the GML index, based on the SBM directional distance function (DDF), can avoid misunderstanding linear problems. The method can measure government spending efficiency with dynamic characteristics by introducing the change in efficiency at time t and t + 1, and the results are more accurate. The expression in Equation is as follows:

$${GE}_t^{t+1}={GML}_t^{t+1}={\displaystyle \frac{1+S_v^G\left(x^t,y^t,b^t,g^x,g^y,g^b\right)}{1+S_v^G\left(x^{t+1},y^{t+1},b^{t+1},g^x,g^y,g^b\right)}}$$

(6)

where $S_v^G\left(·\right)$ is the global DDF, and the GML index measures the dynamic change in government spending efficiency between two periods.

It should be noted that unexpected outputs were not included in our study, so k in the above equations is 0.

The core explanatory variable is the level of digital inclusive finance. This paper employs the county-level digital inclusive finance index, which the Digital Finance Research Center at Peking University and Ant Financial Services Group have collaboratively developed, to quantify this variable. The existing literature on the theme of digital inclusive finance in China mainly uses this data source [29]. This index encompasses three dimensions: the breadth of coverage, the depth of utilization, and the level of digitalization.

The mediating variable under consideration is the fiscal pressure experienced by local governments (FP). In accordance with established measurement techniques found in the existing literature [30,31,32], the fiscal gap method has been employed for this analysis. Specifically, this method quantifies the fiscal pressure by calculating the ratio of the disparity between general public budget expenditures and general public budget revenues to GDP. An increase in this indicator signifies a heightened financial burden on local governments.

The control variables in this study encompass population density (POP), the extent of financial development (FIN), the degree of fiscal dependence (FIS), and the status of industrial development (IND). Population density is quantified as the ratio of the total population within a county to the land area of the corresponding administrative region. Financial development is measured by the ratio of the year-end balance of various loans from financial institutions to GDP. Fiscal dependence is the ratio of local general public budget revenue to GDP. Lastly, industrial development is represented by the proportion of the added value generated by the secondary industry relative to that of the tertiary industry.

2.4. Regional and Temporal Scope of the Study

The Yangtze River Delta, situated in the lower reaches of the Yangtze River, is one of China’s most economically vibrant, open, and innovative regions. It plays a pivotal role in the country’s modernization and regional integration strategy and has been officially designated as one of six world-class urban agglomerations [33]. In 2019, the Outline of the Development Plan for the Integrated Development of the Yangtze River Delta Region formalized the scope of the Yangtze River Delta to include Shanghai, Jiangsu, Zhejiang, and Anhui provinces. This region was selected as the empirical setting for several reasons. First, the Yangtze River Delta exhibits advanced development in digital inclusive finance alongside notable intra-regional disparities in economic capacity, digital infrastructure, and fiscal governance. Such variation allows for the exploration of how digital inclusive finance may exert differential impacts across counties with varying baseline conditions. Second, governments in this region face dual pressures: while they possess abundant fiscal resources, they are also under growing demands to enhance the effectiveness and equity of public service delivery. These characteristics make the Yangtze River Delta an ideal case for examining the relationship between digital financial systems and government spending efficiency.

The study covers the period from 2014 to 2022, selected based on both data availability and policy relevance. The year 2014 marks the initial release of standardized county-level data from the Digital Inclusive Finance Index developed by the Ant Financial and Peking University. Furthermore, this period corresponds to China’s broader national strategy of digital infrastructure expansion, smart government reforms, and fintech integration. Analyzing this time frame enables the study to capture the formative stages of digital inclusive finance development and its evolving role in enhancing local governance performance.

2.5. Data Description

This study focuses on the reliability and integrity of the data by examining 191 counties within the Yangtze River Delta region from 2014 to 2022, resulting in a dataset comprising 1714 panel observations. The Digital Inclusive Finance Index utilized in this research is derived from a county-level index collaboratively developed by Ant Financial and the Digital Finance Research Center at Peking University. The night light data come from the improved DMSP-OLS-like data (1992–2022) set calculated by [34] by combining the calibrated DMSP-OLS data (1992–2013) and the DMSP-OLS-like (2013–2022) data converted from the SNPP-VIIRS data. Additional data were obtained from the “County Statistical Yearbook” and various provincial and municipal statistical yearbooks. Instances of missing data were addressed through linear interpolation. It is important to note that the GML index is based on data from 2014, resulting in the unavailability of specific variable values for that year. Consequently, the statistical analysis concerning the efficiency of government spending is based on 1520 observations. Descriptive statistics for each variable are presented in

Table 2. Descriptive statistical results of core variables.

	Variable	Obs	Mean	Std. Dev.
Explained variable	Government spending efficiency (GE)	1520	1.065	0.776
Core explanatory variable	Digital Inclusive Finance Index (DIF)	1651	108.647	20.509
	Coverage breadth (COV)	1651	95.585	14.878
	Usage depth (USE)	1651	139.078	33.117
	Digitalization level (DIG)	1651	96.489	33.629
Mediating variable	Financial pressure on local governments (FP)	1714	0.094	0.093
Control variable	Population density (POP)	1714	551.158	292.150
	Level of financial development (FIN)	1622	1.075	0.606
	Fiscal dependence (FIS)	1714	0.563	0.268
	Industrial development situation (IND)	1714	1.111	0.469

Notes: The data for some variables is missing in certain years.

. Overall, the average government spending efficiency in the Yangtze River Delta region is 1.065. The average value of the digital inclusive finance index is 108.647. Among them, the average values for coverage breadth, usage depth, and digitalization level are 95.585, 139.078, and 96.489, respectively. From the perspective of their standard deviations, the variation range of the digital inclusive finance level data is not particularly large.

3. Empirical Tests

3.1. Baseline Regression Results

Table 3. Baseline regression results.

	(1)	(2)	(3)	(4)	(5)
	GE	GE	GE	GE	GE
DIF	0.003 **	0.004 ***
	(2.459)	(4.584)
COV			0.002 ***
			(4.821)
USE				0.003 **
				(2.528)
DIG					−0.000
					(−0.369)
POP		0.000	0.000	0.000	0.000
		(0.774)	(0.880)	(0.512)	(0.815)
FIN		−0.001	−0.001	−0.002	−0.000
		(−0.201)	(−0.133)	(−0.247)	(−0.047)
FIS		0.180 ***	0.182 ***	0.180 ***	0.186 ***
		(3.318)	(3.384)	(3.321)	(3.443)
IND		−0.001	−0.002	−0.006	−0.013
		(−0.122)	(−0.161)	(−0.534)	(−1.083)
Fixed effect	YES	YES	YES	YES	YES
Constant	0.702 ***	0.472 ***	0.576 ***	0.543 ***	0.814 ***
	(6.616)	(4.571)	(6.797)	(4.082)	(9.905)
N	1488	1415	1415	1415	1415
R2	0.013	0.251	0.251	0.246	0.243

Note: t statistics in parentheses, ** p < 0.05, *** p < 0.01.

delineates the findings of the benchmark regression model. Column (1) displays the regression outcomes without the inclusion of control variables, while column (2) presents the results after the incorporation of such variables. Furthermore, drawing upon prior research, the regression analysis utilized the three dimensions of digital inclusive finance—namely coverage breadth, usage depth, and digitalization level—as core explanatory variables. The findings are presented in columns (3) to (5) of Table 3.

The regression analysis indicates that irrespective of the presence of control variables, the coefficient associated with digital inclusive finance regarding the efficiency of government spending is significantly positive. This suggests that countries characterized by a higher level of digital inclusive finance exhibit, on average, greater efficiency in government spending. For instance, the regression results presented in column (2) indicate that a one-unit increase in the level of digital inclusive finance correlates with a 0.004 unit increase in the expenditure efficiency of county governments. This effect may be attributed to several factors, including the development of digital platforms, enhanced transparency and accessibility of government information, and improvements in government creditworthiness. A pertinent example can be observed in Pujiang County, Zhejiang Province, China, where convenient service stations have been established to facilitate communication among financial institutions, government entities, enterprises, rural communities, and the general public. By capitalizing on the benefits of financial services and engaging with pertinent departments, community livelihood concerns can be effectively communicated, and financial information can be rendered transparent and accessible. In 2024, Pujiang Rural Commercial Bank facilitated more than 210,000 government services for local residents through its convenience service stations, significantly enhancing the efficiency of governmental governance1.

The results of columns (3) to (5) indicate that both coverage breadth and usage depth significantly influence the efficiency of government expenditure; however, the level of digitalization does not appear to have a substantial effect on enhancing government expenditure efficiency. This finding underscores the notion that, rather than merely improving technical efficiency, the expansion of digital inclusive finance coverage and the optimization of the practical application of digital financial products and services have a more pronounced impact on enhancing the efficiency of fiscal expenditure at the county level. Overall, the Pujiang case exemplifies how real-world digital finance systems can facilitate coordination between government and stakeholders, yielding tangible gains in spending efficiency. Our econometric findings suggest that such improvements are not isolated. Indeed, the digital inclusive finance index captures the broad rollout and active use of digital platforms (e.g., mobile payment platforms and rural banking applications) across the counties studied. The significant coefficient of this index in our analysis provides evidence that the performance of these real digital systems is enhancing governance outcomes at scale.

3.2. Robustness and Endogeneity Tests

We employed four distinct methodologies to perform robustness and endogeneity assessments on the benchmark regression outcomes previously discussed. First, we utilize a two-stage least squares estimation method incorporating instrumental variables to examine the bidirectional causal relationship between digital inclusive finance and efficiency as measured by governmental indicators. Drawing upon existing literature and considering data availability, we employ the Internet penetration rate, mobile phone penetration rate, and the lagged value of digital inclusive finance from the previous period within the county as instrumental variables [16,18,26]. The three variables in question directly influence the advancement of digital inclusive finance; however, they do not directly affect the efficiency of government spending. The findings presented in

Table 4. Results of the instrumental variable method.

	(1)	(2)	(3)
	GE	GE	GE
DIF	0.012 ***
	(4.676)
COV		0.004 ***
		(4.240)
USE			0.009 ***
			(2.635)
Control variables	YES	YES	YES
Fixed effects	YES	YES	YES
Anderson LM	130.254	265.359	119.004
	[0.0000]	[0.0000]	[0.0000]
Cragg-Donald Wald F-statistic	40.836	94.979	36.927
	{12.83}	{12.83}	{12.83}
Constant	−0.502 *	0.180	−0.323
	(−1.792)	(1.057)	(−0.829)
N	1219	1219	1219
R2	0.377	0.418	0.412

Note: The Anderson LM test report shows the p value of the statistics, and the rest are t statistics in parentheses; * p < 0.1, *** p < 0.01; Cragg-Donald Wald F-statistic is the 15% maximal IV size in parentheses.

demonstrate that, following the implementation of the instrumental variable approach, the positive effects of digital inclusive finance, along with its breadth of coverage and usage, continue to be statistically significant regarding the efficiency of government spending.

Second, we assess the robustness of the positive influence of digital inclusive finance on the efficiency of government spending by systematically altering the primary explanatory variables to include the logarithmic values of digital inclusive finance index (DIF_1), the logarithmic values of various dimensions (COV_1, USE_1), and the lagged values (L.DIF, L.COV, and L.USE). The findings presented in

Table 5. Results of changing the core explanatory variables.

	(1)	(2)	(3)	(4)	(5)	(6)
	GE	GE	GE	GE	GE	GE
DIF_1	0.301 ***
	(4.623)
COV_1		0.187 ***
		(4.264)
USE_1			0.364 ***
			(3.377)
L.DIF				0.004 ***
				(5.344)
L.COV					0.002 ***
					(5.334)
L.USE						0.002 **
						(2.398)
Control variables	YES	YES	YES	YES	YES	YES
Fixed effects	YES	YES	YES	YES	YES	YES
Constant	−0.552 *	−0.052	−0.864 *	0.515 ***	0.621 ***	0.614 ***
	(−1.778)	(−0.237)	(−1.697)	(5.924)	(8.159)	(6.133)
N	1415	1415	1415	1398	1398	1398
R2	0.252	0.251	0.248	0.256	0.253	0.246

Note: t statistics in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.

reveal that the coefficients associated with digital inclusive finance, as well as its breadth of coverage and depth of usage, continue to meet the criteria for statistical significance, thereby affirming the robustness of the initial regression results. Additionally, the analysis of the lagged values indicates a significantly positive coefficient, suggesting that the digital inclusive finance index from the preceding period contributes to an enhancement in government spending efficiency. This implies that the impact of digital inclusive finance on improving government spending efficiency exhibits a degree of continuity over time.

Third, given that the current body of literature evaluates the efficiency of fiscal spending from the standpoint of overall developmental progress, we propose a redefinition of the input and output indicators to assess the efficiency of government spending. Specifically, the input indicator is defined as the total annual fiscal expenditure at the county level. The output indicator is modified to reflect the regional GDP level, replacing the previously used per capita regional GDP. All other indicators remain consistent with the previously established definitions. The findings presented in

Table 6. Results for replacing the explained variables.

	(1)	(2)	(3)	(4)
	GE	GE	GE	GE
DIF	0.003 ***
	(4.765)
COV		0.002 ***
		(4.241)
USE			0.002 *
			(1.953)
DIG				0.000
				(0.096)
Control variables	YES	YES	YES	YES
Fixed effects	YES	YES	YES	YES
Constant	0.588 ***	0.671 ***	0.686 ***	0.833 ***
	(7.685)	(9.910)	(7.158)	(13.696)
N	1415	1415	1415	1415
R2	0.356	0.356	0.352	0.351

Note: t statistics in parentheses; * p < 0.1, *** p < 0.01.

demonstrate that the coefficients associated with digital inclusive finance and its various dimensions continue to exhibit a high level of statistical significance.

Finally, considering that the outbreak of COVID-19 in March 2020 was an important and independent event that triggered global macroeconomic turmoil and affected government fiscal expenditures and reports, we therefore, following [35], conducted a sub-sample analysis covering the period from 2020 to 2022 to analyze this period’s extreme socio-economic development conditions in isolation.

Table 7. Results for changing the sample period.

	(1)	(2)	(3)	(4)
	GE	GE	GE	GE
DIF	0.024 **	0.026 **
	(2.299)	(2.037)
DIF_1			3.311 **
			(2.163)
L.DIF				0.050 **
				(2.452)
Control variables	NO	YES	YES	YES
Fixed effects	NO	YES	YES	YES
Constant	−1.973	−2.500	−15.248 **	−5.433 **
	(−1.514)	(−1.397)	(−2.011)	(−2.004)
N	567	534	534	537
R2	0.137	0.161	0.163	0.187

Note: t statistics in parentheses; ** p < 0.05.

shows the results obtained through regression analysis using the fixed effects model. The core explanatory variables in the first and second columns are digital inclusive finance, the first column presents the regression results without including control variables, the second column presents the regression results with all control variables included, the third column presents the logarithmic term of the digital inclusive finance index as the core explanatory variable, and the fourth column presents the lagged one-period value of the digital inclusive finance index as the core explanatory variable. The results indicate that the overall results remain unchanged. When this extreme period is considered in isolation, an increase in the level of digital inclusive finance helps to improve the efficiency of government expenditures.

4. Further Analysis

4.1. Mechanism Test

Table 8. Mechanism test results.

	(1)	(2)	(3)	(4)	(5)
	FP	GE	FP	GE	FP
DIF	−0.001 **	0.003 ***
	(−2.055)	(4.373)
FP		−0.531 **		−0.511 **
		(−2.313)		(−2.167)
COV			−0.000 ***	0.002 ***
			(−3.248)	(4.424)
USE					−0.000
					(−0.095)
Control variables	YES	YES	YES	YES	YES
Fixed effects	YES	YES	YES	YES	YES
Constant	0.254 ***	0.618 ***	0.249 ***	0.716 ***	0.219 ***
	(10.975)	(5.575)	(11.680)	(7.607)	(6.411)
N	1578	1415	1578	1415	1578
R2	0.395	0.256	0.401	0.256	0.386

Note: t statistics in parentheses, ** p < 0.05, *** p < 0.01.

presents the regression results of fiscal pressure as a transmission mechanism.

In column (1), the impact coefficient of digital inclusive finance on fiscal pressure is significant at the statistical level of 5% and is negative. This means that digital inclusive finance has reduced the fiscal pressure on local governments. Column (2) simultaneously incorporates fiscal pressure and digital inclusive finance into the regression equation. The results indicate that reducing fiscal pressure can enhance the efficiency of government expenditure. Therefore, these results indicate that by reducing the fiscal pressure on local governments, digital inclusive finance indirectly promotes the improvement of government expenditure efficiency, thereby confirming Hypothesis 2. Furthermore, we discussed the promoting effect of various dimensions of digital inclusive finance on fiscal pressure. The results of columns (3)–(7) indicate that the coverage breadth has a reducing effect on the government’s fiscal pressure, and the reduction in the government’s fiscal pressure helps to promote the improvement of the spending efficiency level. However, the coefficient of the depth used is insignificant, indicating that this dimension has not significantly reduced the government’s financial pressure. Therefore, expanding the coverage of digital inclusive finance is the main factor in exerting the mediating role of the government’s fiscal pressure.

The above conclusion is like that obtained by [36] through a study using data from 185 cities in China. They believe that developing digital finance helps reduce financial risks and the volatility of financial markets and provides more financial options, alleviating fiscal pressure. Furthermore, the impact of digital finance on fiscal pressure mainly comes from the breadth of coverage. More groups’ access to and use of digital financial services can increase economic activities and tax revenue, improve financial conditions, and alleviate fiscal pressure. However, the impact of the depth of use and the level of digitalization on alleviating financial pressure is relatively limited.

4.2. Heterogeneity Test

The economic development at the county level within the Yangtze River Delta region exhibits notable disparities, and the efficiency of government spending is inconsistent across the area. These variations may influence the extent to which digital inclusive finance affects the efficiency of government spending. Consequently, this study employs indicators of economic development levels and government spending efficiency to conduct a heterogeneity analysis. The regions are categorized based on their average values within the Yangtze River Delta, with those exceeding the mean classified as the high-level group. In contrast, the remaining regions are designated as the low-level group. The findings of this analysis are presented in

Table 9. Results of heterogeneity test.

	(1)	(2)	(3)	(4)
	High-Level Group of Economic Development	Low-Level Group of Economic Development	High-Level Group of Government Spending	Low-Level Group of Government Spending
DIF	−0.000	0.002 *	−0.009	0.002 **
	(−0.213)	(1.946)	(−1.036)	(2.330)
Control variables	YES	YES	YES	YES
Fixed effects	YES	YES	YES	YES
Constant	0.987 ***	0.604 ***	1.819 **	0.724 ***
	(7.166)	(3.057)	(2.312)	(8.212)
N	495	920	367	1048
R2	0.176	0.310	0.215	0.257

Note: t statistics in parentheses; * p < 0.1, ** p < 0.05, *** p < 0.01.

.

First, the analysis of per capita GDP output value reveals that in counties classified within the low-income group, digital inclusive finance exerts a substantial positive influence on the efficiency of government spending. Conversely, in counties categorized as high-income, the effect of digital inclusive finance on government expenditure efficiency is not statistically significant. This discrepancy may be attributed to the fact that in regions with lower economic development, traditional financial services are inadequately accessible, thereby exacerbating the challenges associated with financing. The advancement of digital inclusive finance can reduce barriers to service access, stimulate the dynamism of the private sector, and enhance government revenue streams, thereby improving the efficiency of fiscal resource utilization. In contrast, regions with a higher degree of economic development tend to possess more robust financial infrastructure, enabling traditional financial institutions to address market needs more effectively. Consequently, in these areas, digital finance has already reached a mature stage of development, which may lead to diminished marginal returns from implementing digital inclusive finance initiatives.

Second, based on the analysis of government spending efficiency, it was observed that digital inclusive finance did not demonstrate a statistically significant positive impact in counties exhibiting relatively high levels of government spending efficiency. Conversely, in counties with lower government spending efficiency, the coefficient associated with digital inclusive finance was found to be significantly positive at the 5% statistical level, indicating a notable enhancement effect. This suggests that the advancement of digital inclusive finance has the potential to improve the spending efficiency of local governments and mitigate disparities in expenditure efficiency across counties. This phenomenon may be attributed to counties with higher government spending efficiency typically possessing more developed digital infrastructure and governance capabilities, which may result in diminishing marginal returns from digital inclusive finance initiatives.

The summary conclusion of the above research is that the digital inclusive finance index has a more significant positive effect on government spending efficiency in disadvantaged regions. We have not found literature that studies the heterogeneity impact of the digital inclusive finance index on government spending efficiency from these two perspectives. However, in the research on digital inclusive finance in other fields, we have found some literature that can support our result from a different angle. For example, the research conclusion of [16] is similar to ours. Their research found that compared with the eastern coastal areas, the impact of digital inclusive finance on the relative asset poverty alleviation of families in the central and western inland areas was more significant, because these regions’ families had more opportunities to access digital financial services than before. The research also indicated that the optimization effect of digital inclusive finance on labor allocation mainly occurred in regions with a lower level of traditional financial development [37]. The existing literature supports the heterogeneity research results of this paper from a different perspective.

5. Conclusions and Discussion

Crucially, the SBM–GML index and fixed-effects model capture the performance of genuine digital finance systems in operation. The digital inclusive finance index reflects real-world platforms—ranging from mobile payment networks to rural financial service kiosks—that governments use to streamline service delivery and data sharing. Consequently, the efficiency gains identified in our analysis correspond to substantive enhancements in the reach and functionality of these digital systems across the region. This paper uses the county-level panel data of the Yangtze River Delta region in China from 2014 to 2022. It adopts a multiple regression model to empirically test the impact effect and mechanism of digital inclusive finance on the efficiency of government spending. First, digital inclusive finance has significantly improved government spending efficiency. The coverage breadth and application depth of digital finance significantly promote the efficiency of government spending. However, the level of digitalization has no significant impact on the efficiency of government spending. Second, the positive impact of digital inclusive finance on the efficiency of government spending shows regional heterogeneity. Thirdly, digital inclusive finance can enhance the efficiency of government spending by reducing fiscal pressure, and the breadth of its coverage has the potential to play such a role. The above conclusions provide theoretical and empirical support for evaluating the government governance effect of digital inclusive finance and improving the efficiency of government expenditure. This paper puts forward the following suggestions.

First, the government should strongly support the development of digital inclusive finance. The coverage and application depth of digital inclusive finance can be enhanced by accelerating the development speed of digital finance in remote areas and encouraging society to learn and use digital finance widely. Furthermore, although the level of digitalization has not positively impacted the efficiency of government fiscal spending, the government should also enhance the construction of digital financial infrastructure, big data, artificial intelligence, and other aspects to improve the level of digitalization. On the one hand, it supports the overall development of digital inclusive finance by increasing digitalization. On the other hand, it is expected to highlight the positive role of this dimension in the effectiveness of government spending in the long term.

Second, governments should avoid a one-size-fits-all approach and instead tailor policies to support the development of digital inclusive finance in economically underdeveloped regions with low levels of fiscal effectiveness. In such regions, the expansion of digital financial services can play a disproportionately positive role in narrowing disparities in government spending efficiency. Therefore, differentiated policy instruments should be adopted, such as collaborating with financial institutions to establish dedicated loan schemes and preferential financing channels, as well as strengthening investments in digital infrastructure to accelerate the development of digital finance in these areas. As a forward-looking recommendation, we propose that local governments consider establishing a dynamic monitoring and assessment platform that integrates multifactorial indicators, such as digital financial inclusion coverage, education levels, healthcare access, and fiscal pressure to evaluate public spending efficiency in real time. This tool could be co-designed by policymakers, academic researchers, and IT specialists to generate actionable insights for precision-targeted policy interventions and more adaptive fiscal governance. Future research could further explore the system architecture, data integration models, and algorithmic frameworks required to operationalize such a tool at the county level.

Finally, based on the functional channels of digital inclusive finance, it is necessary to explore practical ways to reduce credit risks in the financial market and expand diversified financing channels, thereby meeting market credit demands and playing a role in alleviating the pressure of government fiscal debt. On the one hand, financial regulatory authorities should enhance regulatory efforts and jointly build an information-sharing platform with other government departments, financial institutions, enterprises, and other entities to alleviate information asymmetry, improve the efficiency of financial market supervision, and reduce credit risks in the financial market. On the other hand, financial institutions should actively develop diversified financial products to meet the personalized needs of customers and provide multiple credit channels.

From a theoretical standpoint, this study contributes to the emerging discourse on cyber–physical governance systems, where digital infrastructures and financial platforms interact with institutional logics and social behaviors. It extends the analytical scope of digital inclusive finance beyond economic outcomes, situating it firmly within the systems theory framework of public governance performance. This integrated perspective enriches both digital finance and public administration literature by providing a conceptual and empirical bridge. Practically, the study offers several policy implications. Governments should not only invest in digital financial infrastructure but also foster widespread and equitable access to ensure system-wide effectiveness. This includes promoting digital literacy, enhancing trust in digital platforms, and aligning fiscal strategies with digital transformation. Special attention should be paid to structurally disadvantaged regions, where the payoff of digital financial interventions is higher due to existing institutional gaps. Furthermore, the study suggests that public sector reform should treat digital inclusive finance as a subsystem of national governance, integrated with other digital systems such as e-government portals, real-time fiscal monitoring, and social service delivery platforms. Policymakers should also recognize the value of system-wide spillovers and pursue regional coordination strategies that enhance data interoperability and service consistency.

6. Limitations and Future Research

While this study provides empirical evidence on the positive effects of digital inclusive finance on government spending efficiency, several limitations must be acknowledged.

First, the analysis is geographically constrained to counties within the Yangtze River Delta, an economically advanced region characterized by high-quality digital infrastructure, widespread financial inclusion, and robust administrative capacity. These favorable conditions may amplify the effects of digital inclusive finance in ways that are not replicable in less-developed regions. Therefore, the conclusions drawn in this study are most applicable to areas with similar structural and institutional contexts. Caution should be exercised when extrapolating these findings to regions with lower digital penetration, weaker fiscal autonomy, or underdeveloped institutional settings.

Second, the study relies on county-level panel data, which, while suitable for regional comparative analysis, poses limitations in terms of variable granularity. Certain important social indicators, such as adult educational attainment, multidimensional poverty, or differentiated access to healthcare services, could not be incorporated due to data constraints. These omissions may limit the comprehensiveness of the welfare dimension and measurement accuracy.

In terms of future research directions, three promising areas are highlighted. First, studies could examine spillover effects by analyzing whether the development of digital inclusive finance in one region exerts influence on neighboring counties’ fiscal outcomes. Second, further exploration is needed into the value of data and its role in shaping policy outcomes, especially about transparency, media influence, and administrative efficiency. Relevant work by [38] on financial data valuation provides a useful reference in this context. Finally, interdisciplinary collaborations could explore the development of intelligent digital governance platforms that enable real-time monitoring of fiscal performance, integrating multifactorial indicators such as digital inclusion, education, and healthcare access. These platforms may serve as valuable decision-support tools for policymakers in the digital economy era.