The Impact of Digital Inclusive Finance on Employment in Micro- and Small Enterprises: An Empirical Analysis Based on the China Micro- and Small-Enterprise Survey (CMES) Database

Abstract

Digital inclusive finance has become an important policy instrument for promoting inclusive growth and decent employment. Using matched data from the China Micro- and Small-Enterprise Survey (CMES) and the Peking University Digital Financial Inclusion Index, this study applies a bivariate Probit model to 3263 micro- and small enterprises (MSEs) across 28 Chinese provinces to examine how digital inclusive finance relates to MSEs’ labor hiring decisions, distinguishing between formal and informal employment and exploring the underlying mechanisms. Digital inclusive finance is positively associated with formal employment and negatively associated with informal employment in MSEs. Mediation analysis shows that credit constraints are an important transmission channel, accounting for 9.17% and 10.58% of the associations with formal and informal employment, respectively. The breadth of coverage and depth of use of digital inclusive finance are positively associated with formal employment and negatively associated with informal employment, whereas the degree of digitalization exhibits the opposite pattern. Digital inclusive finance also alleviates supply-side credit constraints faced by MSEs but shows no significant relationship with demand-side constraints. These findings offer policy implications for promoting inclusive and decent employment in MSEs.

1. Introduction

The United Nations 2030 Agenda for Sustainable Development proposes 17 Sustainable Development Goals (SDGs), which encompass the three dimensions of sustainable development: economic, social, and environmental. SDG 8, “Decent Work and Economic Growth,” aims to “promote sustained, inclusive and sustainable economic growth, full and productive employment, and decent work for all.” This goal emphasizes improving equal access to employment opportunities, skills development, and financial services, particularly for vulnerable groups, including youth, women, persons with disabilities, and other disadvantaged populations. Furthermore, Targets 8.3 and 8.10 highlight the importance of supporting the development of micro, small, and medium-sized enterprises, promoting their formalization, and expanding access to banking, insurance, and financial services for all social groups. As an important part of the national economy and social development, Micro- and Small Enterprises (MSEs) play a significant role in improving people’s livelihoods, expanding employment, and releasing market potential [1,2,3]. According to data from the China Household Finance Survey of Southwestern University of Finance and Economics (CHFS2015), China’s MSEs have absorbed 237 million jobs and contributed 30% to GDP growth. The Central People’s Government of the People’s Republic of China website shows that at the end of 2017, there were about 28 million MSEs and 62 million individual industrial and commercial households in China. Micro, small, and medium-sized enterprises (including individually owned businesses) accounted for more than 90% of all market players and contributed more than 80% of the country’s employment. China is currently upgrading its industrial structure and pursuing high-quality economic growth. Against this backdrop, employment in MSEs matters greatly for the healthy, stable, and sustainable development of the national economy [4,5,6]. Examining employment at the MSE level also speaks directly to the claim, advanced by some overseas scholars, that China’s expansion has been a “jobless growth boom” [7,8]. Because China is the world’s second-largest economy, evidence on how digital inclusive finance relates to employment in its MSEs also carries weight beyond its borders. Such evidence gives other economies a reference point as they reconsider the role of digital finance in employment relationships, and it feeds into the wider debate on building a healthy and sustainable labor market.

Within MSEs, equal pay for equal work and equalization of employment opportunities are important ways to protect workers’ rights and interests and the basic path and strategic direction of current labor market-oriented reform. It is worth noting that, as an important component of the labor market, informal employment has been the subject of attention and discussion from many sides. Still, differences in the causes and perceptions of informal employment have triggered the discussion of scholars on the policy orientation toward informal employment. Two schools of thought have consequently emerged. The first treats informal employment as a way for workers to sustain their livelihoods in a segmented labor market. On this view, informal workers earn less than formal ones [9,10,11] and obtain significantly lower returns to education and experience [12,13,14]. Informal work also brings long hours [15,16], unstable jobs, and worse physical and mental health [17,18,19,20], together with a lack of labor protection and social security. Because many workers accept it only under employment pressure, their satisfaction is correspondingly low [21,22,23]. Scholars in this camp therefore urge governments to take active measures to reduce informal employment [24,25]. On the other hand, although a shortage of acceptable alternatives pushes many workers into informal jobs, others prefer non-standard arrangements, particularly part-time hours [26,27]. Informality can also be a way of sidestepping burdensome regulation and the cost, time, and effort that formal registration demands [28,29]. Informal employment has a less significant income gap than formal employment and is more autonomous, flexible, and competitive, so governments should guide and encourage its development [30,31,32,33,34]. Despite these inconsistent views, scholars broadly agree that, over the long run, informal employment is a product of lower levels of socio-economic development and may itself become a source of concern. In addition, informal employment may discourage the least qualified and exclude them from the primary sector of the labor market [35,36]. As the economy develops and related institutions improve, giving workers decent working conditions and social security has become both an internal driver and an unavoidable direction of labor market reform. Recent demographic change and the accelerating upgrading of China’s industrial structure have reinforced this shift toward a more formal employment structure, which in turn raises the bar for market-oriented labor reform. On the one hand, it is necessary to ensure the legitimate rights and interests and survival needs of informally employed workers. On the other hand, it is necessary to promote the formalization of informal employment in the context of the transformation and upgrading the labor structure [32,37,38,39].

Finance and employment are closely linked and inseparable [40,41]. At the micro level, a frictionless financial market lets firms draw funds from financial institutions continuously, which encourages investment and the expansion of production [42,43,44,45]. Financial friction, by contrast, creates credit constraints. Such constraints harm MSEs [46,47,48,49] and reduce the number of workers they employ [50,51]. Financial development reduces the cost of financing for MSEs [52,53], eases their credit constraints [54,55,56,57,58], and promotes employment and growth [28,59,60,61]. Greenwood et al. [62] found that financial development provides credit support to individuals who were previously financially excluded, energizing the production of social capital and creating jobs. Inclusive finance rests on the principle that financial services should reach every group that needs them at a price those groups can bear. Its priority clients are the ones the traditional system underserves: agriculture and rural households, small and medium-sized firms, and low-income urban residents [63,64,65]. Meanwhile, considering the relationship between inclusive finance and employment: on the one hand, inclusive finance can increase employment opportunities in enterprises by alleviating the long-standing financial inhibition of enterprises [66,67]. Evidence also indicates that inclusive finance helps optimize China’s employment structure, though unevenly: the effect is pronounced in the eastern provinces but negligible in the central and western regions [68,69,70]. On the other hand, as inclusive finance continues to expand and reshape the financial ecosystem, it supplies the funding MSEs need to grow steadily, and financial inclusion has in turn been linked to sustainable macroeconomic growth [71,72].

However, with the development of inclusive finance, shortcomings such as information asymmetry, higher transaction costs, insufficient breadth and depth of services, inadequate investment in hardware and software infrastructure, and low mobility have begun to emerge. As big data, cloud computing, and artificial intelligence have matured, traditional inclusive finance has increasingly merged with these technologies, and digital inclusive finance has emerged as a field in its own right [73]. Digital inclusive finance can be defined as “deploying cost-effective digital means to provide a range of formal financial services tailored to the needs of those currently excluded and underserved by finance, provided responsibly at a cost that customers can afford and providers can sustainably provide” [74]. By lowering barriers to saving, it gives low-income households a way to build a financial buffer, and the deposits that result give banks a steadier funding base in periods of stress [75,76,77]. It has likewise been associated with economic development [78,79,80], poverty reduction [81,82,83], and gains in individual well-being [21,82,84,85,86]. Empirical work further points to a role in environmental governance [87,88,89], while fintech has been shown to narrow income inequality indirectly through its effect on financial development [90,91,92]. As a core driver of inclusive finance [93,94,95], fintech has also been linked to job creation [96], economic growth [97], and broader development outcomes [14,37,92,98,99,100]. On employment specifically, the literature reports that digital inclusive finance raises employment [101,102,103], with the effect running mainly through the breadth of coverage and the depth of use. Hu and Lu [104], Shang and Liu [61] all find that digital inclusive finance markedly raises firms’ labor demand, particularly for high-skilled workers, mainly by easing financing constraints and encouraging digital transformation. The size of this effect nevertheless varies across firms and is notably weaker in non-high-tech firms and in firms paying lower wages.

Digital inclusive finance is likely to matter more for employment in MSEs than in medium-sized and large firms [105]. Larger firms draw on social capital such as political connections and implicit state guarantees, which gives them an edge in securing financial resources. MSEs hold little of this capital, so the spread of digital inclusive finance does more to improve their access to credit [106,107]. On the demand side, the “grassroots” orientation of digital inclusive finance is aimed precisely at relieving the credit constraints of MSEs. Medium-sized and large firms holding high-quality collateral can already obtain relationship-based loans, so they benefit far less. Moreover, the widely used “Peking University’s Digital Inclusive Finance Index” is built mainly on Alipay data, and listed companies fall outside its direct service scope. This suggests that digital inclusive finance is oriented above all toward “supporting small and micro” businesses. Therefore, with the development and improvement of digital inclusive finance, its inclusive features will have an important impact on labor employment in MSEs. Accordingly, we ask the following questions: Is the development of digital inclusive finance a major incentive for labor employment behavior in MSEs? Is there heterogeneity in its impact on the employment of formal and informal workers? What are the underlying mechanisms?

Compared with existing studies, the marginal contributions of this paper are summarized as follows. First, this paper approaches MSEs’ employment decisions through the heterogeneity of employment forms. That perspective is new to this literature and fills a gap left by studies that treat employment as homogeneous. Second, this paper analyzes theoretically how digital inclusive finance bears on MSEs’ labor employment decisions. It then matches the Peking University Digital Inclusive Finance Index with survey data on Chinese MSEs and tests that relationship empirically. Previous studies concentrated on macro- and meso-level outcomes such as economic growth and financing constraints. Few have examined labor employment in MSEs from a micro perspective. The findings of this paper provide valuable references for subsequent relevant empirical research. Third, this paper builds a mediation model along two dimensions, formal and informal employment. The model opens the “black box” of how digital inclusive finance reaches labor employment in MSEs and identifies the boundary conditions of that effect. It also supports a detailed theoretical and empirical analysis of the chain running from the development of digital inclusive finance, through an intermediate transmission factor, to formal or informal hiring.

2. Theoretical Model Analysis

Following Caggese et al. [108], we build an employment model for MSEs with heterogeneous employees. The model rests on three premises. First, wages are rigid and do not move with fluctuations in individual productivity. Second, informal employees can be converted into formal ones. Third, the cost of dismissal rises the longer an employee has been with the firm.

Assuming that the output of each employee is Bμ/m_t^1−α, where B denotes MSE-specific productivity, m_t denotes the number of employees, and μ signifies employee productivity. α ∈ (0, 1) is the elasticity of MSEs’ total output with respect to labor input. The newly hired informal employees of MSEs have an initial productivity μ^X, which follows a uniform distribution [μ^L, μ^H]. For each newly hired informal employee, there is a probability p of them becoming formal employees. After becoming formal employees, their productivity is μ^Y, which follows a uniform distribution [μ^L, δμ^H]. Where δ > 1, it indicates that formal employees have higher productivity compared to informal employees.

For MSEs, they can dismiss informal employees without any cost. However, MSEs need to pay a certain cost C to lay off formal employees, while they can hire new informal employees by paying a fee D. For simplicity, the productivity of a new hire becomes observable only once the wage has been fixed and can no longer be adjusted. Meanwhile, the market interest rate r and the specific productivity B of MSEs remain constant. Therefore, the optimal wage w remains unchanged, and the optimal number of employees does not change over time, that is, m_t = m. In each stage of the development of MSEs, informal employees have a probability q of deciding to resign. For the remaining informal employees, a small fraction have a probability λ of becoming formal employees, while MSEs will hire some new informal employees.

We define V^Y(μ_t^Y) as the value of a formal employee in an MSE, which is a function of productivity:

$$V^Y({\mu }_t^Y)=({\displaystyle \frac{B}{m_t^{1-\alpha }}}{\mu }_t^Y-w)+{\displaystyle \frac{(1-q)}{1+r+\eta }}{E}_t[V^Y({\mu }_{t+1}^Y)]$$

(1)

In the above formula, r represents the market interest rate, and η denotes the degree of financing constraints faced by MSEs. Since the productivity of MSEs remains unchanged over time, we have μ_t^Y = μ^Y, from which we can deduce the following:

$$V^Y({\mu }^Y)=({\displaystyle \frac{B}{m^{1-\alpha }}}{\mu }^Y-w){\displaystyle \frac{1+r+\eta }{q+r+\eta }}$$

(2)

An informal employee with a productivity of μ^X who is not dismissed in the current period creates the following value:

$$V^X({\mu }^X)=({\displaystyle \frac{B}{m^{1-\alpha }}}{\mu }^X-w)+{\displaystyle \frac{(1-q)}{1+r+\eta }}\lambda \{E[V^Y({\mu }^Y)]\}+{\displaystyle \frac{1-q}{1+r+\eta }}(1-\lambda )V^X({\mu }^X)$$

(3)

It can be simplified as:

$$V^X({\mu }^X)={\displaystyle \frac{1+r+\eta }{r+\eta +\lambda +q(1-\lambda )}}\{({\displaystyle \frac{B}{m^{1-\alpha }}}{\mu }^X-w)+{\displaystyle \frac{(1-q)}{1+r+\eta }}\lambda \{E[V^Y({\mu }^Y)]\}\}$$

(4)

where E[V^Y(μ^Y)] represents the expected value created by the conversion of informal employees to formal employees:

$$E[V^Y({\mu }^Y)]=[{\displaystyle \frac{B}{m^{1-\alpha }}}({\displaystyle \frac{1}{2}}\delta {\mu }^H+{\displaystyle \frac{1}{2}}{\mu }_{min}^Y)-w]{\displaystyle \frac{1+r+\eta }{q+r+\eta }}$$

(5)

where μ_min^Y represents the minimum productivity required for a formal employee to avoid dismissal.

The optimal employment size m of an MSE is determined by the following conditions:

$$E_t\left[V^X\right({\mu }^X\left)\right]-D=0$$

(6)

Since V^X(μ^X) has a linear relationship with μ^X, employee productivity is uniformly distributed around μ^X. Therefore, the expected value of value creation by newly hired informal employees in MSEs is as follows:

$$E_t[V^X({\mu }^X)]={\displaystyle \frac{1}{2}}{V}^X({\mu }^H)+{\displaystyle \frac{1}{2}}{V}^X({\mu }_{min}^X)$$

(7)

Here, μ_min^X represents the minimum productivity level required for informal employees to avoid dismissal.

Under the condition of free entry, informal employees with excessively low productivity and no contribution value to MSEs will be dismissed. Their minimum productivity μ_min^X satisfies

$$V^X\left({\mu }_{min}^X\right)=0$$

(8)

If the value of a formal employee to an MSE is negative and the absolute value is greater than the cost of dismissal, then the MSE will dismiss the formal employee. The minimum productivity μ_min^Y of the formal employee satisfies

$$V^Y\left({\mu }_{min}^Y\right)=-C$$

(9)

From Formulas (2) and (9), we can derive the following:

$$({\displaystyle \frac{B}{m_t^{1-\alpha }}}{\mu }_{min}^Y-w){\displaystyle \frac{1}{\Delta }}+C=0$$

(10)

Here, Δ ≡ (q + r + η)/(1 + r + η), we can derive the following:

$${\mu }_{min}^Y={\displaystyle \frac{m^{1-\alpha }}{B}}(w-\Delta C)$$

(11)

Combining Equation (5) and Equation (11), we obtain the following:

$$E[V^Y({\mu }^Y)]={\displaystyle \frac{1}{2}}({\displaystyle \frac{B}{m^{1-\alpha }}}\delta {\mu }^H-w){\displaystyle \frac{1}{\Delta }}-{\displaystyle \frac{1}{2}}C$$

(12)

Combining the dismissal condition V^X(μ_min^X) = 0 for informal employees mentioned in the previous text, we can deduce that:

$${\mu }_{min}^X={\displaystyle \frac{m^{1-\alpha }}{B}}\{w-{\displaystyle \frac{1-q}{1+r+\eta }}\lambda E[V^Y({\mu }^Y)]\}$$

(13)

Combining Equations (11)–(13), we obtain the following:

$${\mu }_{min}^Y-{\mu }_{min}^X={\displaystyle \frac{m^{1-\alpha }}{B}}\{{\displaystyle \frac{1-q}{1+r+\eta }}\lambda E[V^Y({\mu }^Y)]-\Delta C\}$$

(14)

An increase in η raises Δ, and a higher Δ lowers E[V^Y(μ^Y)]. Therefore, μ_min^Y − μ_min^X will also decrease accordingly.

Assuming that in a new era, due to the development of digital inclusive finance, firm-specific productivity (B) of MSEs rises, leading to changes in the value created by employees. At this point, as μ_min^Y and μ_min^X decrease, MSEs will increase the employment of formal and informal employees.

Combined with Equation (4), the higher the degree of financing constraints faced by MSEs, the more their informal employees’ value is driven by current profitability (Bμ^X/m^1−α − w).

Furthermore, by taking partial derivatives of Equations (11) and (13), we can obtain

$${\displaystyle \frac{\partial {\mu }_{min}^Y}{\partial B}}=-{\displaystyle \frac{m^{1-\alpha }}{B^2}}(w-\Delta C)$$

(15)

$${\displaystyle \frac{\partial {\mu }_{min}^X}{\partial B}}=-{\displaystyle \frac{m^{1-\alpha }}{B^2}}\{w-{\displaystyle \frac{1-q}{1+r+\eta }}\lambda E[V^Y({\mu }^Y)]\}$$

(16)

Combining the above two equations, we obtain:

$${\displaystyle \frac{\partial {\mu }_{min}^X}{\partial B}}-{\displaystyle \frac{\partial {\mu }_{min}^Y}{\partial B}}=-{\displaystyle \frac{m^{1-\alpha }}{B^2}}\{\Delta C-{\displaystyle \frac{1-q}{1+r+\eta }}\lambda E[V^Y({\mu }^Y)]\}$$

(17)

Therefore, ∂μ_min^X/∂B − ∂μ_min^Y/∂B decreases as the degree of financing constraints increases.

On the one hand, digital inclusive finance has widened the range and the reach of inclusive financial services and rebuilt them into a broadly accessible, multi-tiered system, which raises MSE employment in the sectors it touches. On the other hand, digital finance with inclusive characteristics has improved credit allocation in the traditional financial sector. Unlike traditional finance, which relies on hard information such as corporate financial statements for credit extension, digital inclusive finance builds credit evaluation models based on the soft information accumulated by MSEs on the Internet. Because this lowers transaction costs, it eases the credit constraints MSEs face. Consequently, it allows for an increase in capital investment, promoting the employment of formal employees in these enterprises. Due to the substitution relationship between formal and informal employees, the development of digital inclusive finance has alleviated credit constraints for MSEs while also inhibiting the employment of informal employees.

3. Hypotheses Development

3.1. Digital Inclusive Finance and Labor Employment in MSEs: Based on Differences in Employment Forms

For MSEs, formal and informal employment may be related to each other. In some cases, they may exhibit a substitution relationship. At the early stage of employment, the income difference between formal and informal employees may be relatively small. However, this difference may increase with employees’ age and skill level. Formal employees may also have better access to social insurance and other employment benefits than informal employees. Therefore, labor costs may be an important consideration in MSEs’ employment decisions. When facing financial constraints, MSEs may be more likely to rely on informal employees because of their relatively lower employment costs.

The formalization of the labor market is a gradual process. Informal employment may also involve several disadvantages. For example, informal employees may receive lower wages and fewer employment benefits than formal employees performing similar work. This may affect employee satisfaction and loyalty. It may also create challenges for human resource management and labor relations. In addition, the comparative advantage of informal employment may partly rely on lower labor costs and weaker employment protection. Such an employment model may be less sustainable in the long term. Therefore, promoting a more formal and inclusive labor market may help improve workers’ welfare and support the sustainable development of enterprises. Based on these considerations, employment costs and financial conditions may be associated with MSEs’ choices between formal and informal employees. When enterprises have sufficient financial resources, they may have greater capacity to hire formal employees. They may also have greater capacity to provide corresponding wages, benefits, and social insurance.

MSEs are an important part of the “long tail” of credit demand. Digital inclusive finance may improve financial accessibility for these enterprises. Through information technology, digital finance can make greater use of available information. It may also reduce information asymmetry between borrowers and lenders. Therefore, digital inclusive finance may help ease financing constraints and improve access to financial services for MSEs [109]. The development of digital inclusive finance may also be associated with changes in MSEs’ employment decisions. On the one hand, improved access to financial services may provide MSEs with greater flexibility in managing employment costs. This may increase their capacity to provide wages, benefits, and social insurance for employees. Such changes may be associated with a greater tendency toward formal employment. On the other hand, improved access to financial services may provide MSEs with more resources to adjust their employment structure according to their business needs.

Accordingly, this paper puts forward the following hypotheses:

Hypothesis 1. 

The development of digital inclusive finance is positively associated with formal employee hiring and negatively associated with informal employee hiring in MSEs.

3.2. The Mechanism of Digital Inclusive Finance Affecting Labor Employment in MSEs: Credit Constraints

Credit constraints are an important factor in the financing decisions of MSEs. In traditional credit markets, information asymmetry between borrowers and lenders may limit enterprises’ access to external finance. Collateral is often used to reduce information asymmetry. However, many MSEs have limited collateral and relatively incomplete financial information. These characteristics may make it more difficult for them to obtain formal credit.

Digital inclusive finance may help address some of these limitations. It combines digital technologies with traditional financial services. It also makes greater use of information generated through online activities and business transactions. Such information may provide additional evidence for credit assessment. It may also help reduce information asymmetry between financial institutions and MSEs. For example, digital financial services can use transaction records and other forms of soft information to assess the creditworthiness of enterprises. Previous studies suggest that such information can improve the matching between borrowers and financial resources [110]. Digital technologies may also help financial institutions obtain more information about the business activities of MSEs. These developments may improve the accessibility of credit for enterprises that have limited traditional collateral or financial information. Improved access to finance may be associated with lower credit constraints among MSEs. When financing conditions improve, enterprises may have greater flexibility in allocating resources to production and employment. They may also have greater capacity to bear employment-related costs, including wages, benefits, and social insurance. Therefore, credit constraints may represent a potential mechanism linking digital inclusive finance with MSEs’ employment decisions. This mechanism may be reflected in the distinction between formal and informal employment. When credit constraints are less severe, MSEs may have greater capacity to employ formal employees. At the same time, their reliance on informal employees may decline. However, these relationships should be examined empirically rather than interpreted as direct causal effects.

Accordingly, this paper proposes the following hypothesis:

Hypothesis 2. 

The relationship between digital inclusive finance and MSEs’ formal and informal employee hiring decisions may operate through credit constraints.

4. Data and Econometric Strategy

4.1. Econometric Model Specification

To examine the impact of digital inclusive finance on labor employment in MSEs, we classify employees into two groups: “formal employees” and “informal employees” based on the differences in the forms of employment. These two forms of employment need not be independent: a firm may use both at once, and its reliance on informal workers may shape how many formal workers it takes on. Although separate Probit models can be used to estimate the decisions of MSEs to hire formal and informal employees, these two employment decisions may not be independent. In practice, formal and informal employment may be related. They may exhibit either substitution or complementary relationships within MSEs. Therefore, estimating the two employment decisions separately may fail to account for the potential correlation between their unobserved determinants. To capture this potential interdependence, we employ a bivariate Probit model. The bivariate Probit model is mainly used when two binary dependent variables are jointly affected by unobservable factors. It allows the error terms of the two equations to be correlated. This helps account for the potential dependence between the two employment decisions that may not be captured by separate Probit models. Therefore, the bivariate Probit model provides a more appropriate framework for jointly estimating MSEs’ decisions to hire formal and informal employees.

The bivariate Probit model consists of two equations: the first describes an MSE’s decision to hire formal employees and the second its decision to hire informal employees. In our specification, the two equations share the same control variables; the two hiring decisions are correlated, as are the error terms, whose covariance is treated as a fixed constant. The model is set up as follows:

Emp_m indicates the employment of formal employees in MSEs. Emp_m = 1 indicates that MSEs employ formal employees, and Emp_m = 0 indicates that MSEs do not employ formal employees. Emp_n indicates the employment of informal employees in MSEs. Emp_n = 1 indicates that MSEs employ informal employees, and Emp_n = 0 indicates that MSEs do not employ informal employees. Emp_m* and Emp_n* are the latent variables of labor employment decisions of MSEs. The latent variables of labor employment decisions are satisfied:

$$\{\begin{array}{c}Em{p}_m^{*}={\alpha }_{1}DIF+{\alpha }_{2}X+{\xi }_1\\ Em{p}_n^{*}={\beta }_{1}DIF+{\beta }_{2}X+{\xi }_2\end{array}$$

(18)

In Equation (18), DIF indicates the development level of digital inclusive finance. X is a control variables that represent other possible factors affecting the labor employment decisions of MSEs, including firm size and operating years. α and β are the parameters to be estimated, and it is assumed that the stochastic perturbation terms ξ₁ and ξ₂ follow a joint normal distribution with zero mean, unit variance, and correlation coefficient ρ. E(ξ₁) = E(ξ₂) = 0, Var(ξ₁) = Var(ξ₂) = 1 and Cov(ξ₁,ξ₂) = ρ. If ρ = 0, it indicates that the model can be reduced to two independent Probit models, which only need to be estimated separately. If ρ ≠ 0, it indicates that the error terms of the two equations are correlated, and the hiring decisions of formal and informal employees in MSEs are jointly influenced by several unobservable factors. When ρ > 0, unobservables that raise the probability of formal hiring also raise the probability of informal hiring. When ρ < 0, those same unobservables lower the probability of informal hiring.

The bivariate Probit model is defined as follows:

$${\mathit{Emp}}_m=\{\begin{array}{ccc}1& if& {\mathit{Emp}}_m^{*}>0\\ 0& if& {\mathit{Emp}}_m^{*}\le 0\end{array}{\mathit{Emp}}_n=\{\begin{array}{ccc}1& if& {\mathit{Emp}}_n^{*}>0\\ 0& if& {\mathit{Emp}}_n^{*}\le 0\end{array}$$

(19)

To address the correlation between two binary dependent variables, we employ a bivariate Probit model. This model assumes that the disturbances in the two regression equations are jointly normally distributed and uses the correlation coefficient ρ to account for unobserved covariates that affect both decisions simultaneously. In doing so, it alleviates both sample selection bias and endogeneity bias caused by the simultaneity of the two decisions. In addition, we use the uncentered variance inflation factor (VIF) to detect multicollinearity. The mean VIF of the model is 5.25, well below the conventional threshold of 10, suggesting that multicollinearity is not a serious concern.

Under this specification, and provided that a firm employs at least one worker, MSEs fall into three cases: they hire both formal and informal employees, only formal employees, or only informal employees. To improve the precision and efficiency of the estimates, the two decisions are estimated jointly by maximum likelihood. When Emp_m = 1, Emp_n = 1,

$$Pr(Em{p}_m=1,Em{p}_n=1)=F({\alpha }_{1}DIF+{\alpha }_{2}X,{\beta }_{1}DIF+{\beta }_{2}X;\rho )$$

(20)

When Emp_m = 1, Emp_n = 0,

$$Pr(Em{p}_m=1,Em{p}_n=0)=\Phi ({\alpha }_{1}DIF+{\alpha }_{2}X)-F({\alpha }_{1}DIF+{\alpha }_{2}X,{\beta }_{1}DIF+{\beta }_{2}X;\rho )$$

(21)

When Emp_m = 0, Emp_n = 1,

$$Pr(Em{p}_m=0,Em{p}_n=1)=\Phi ({\beta }_{1}DIF+{\beta }_{2}X)-F({\alpha }_{1}DIF+{\alpha }_{2}X,{\beta }_{1}DIF+{\beta }_{2}X;\rho )$$

(22)

The log-likelihood function is as follows:

$$LnL={\sum }_i^n\left\{\begin{array}{c}Em{p}_{m}Em{p}_{n}LnF({\alpha }_{1}DIF+{\alpha }_{2}X,{\beta }_{1}DIF+{\beta }_{2}X;\rho )\\ +Em{p}_m(1-Em{p}_n)Ln\left[\Phi \right({\alpha }_{1}DIF+{\alpha }_{2}X)-F({\alpha }_{1}DIF+{\alpha }_{2}X,{\beta }_{1}DIF+{\beta }_{2}X;\rho \left)\right]\\ +(1-Em{p}_m)Em{p}_{n}Ln\left[\Phi \right({\beta }_{1}DIF+{\beta }_{2}X)-F({\alpha }_{1}DIF+{\alpha }_{2}X,{\beta }_{1}DIF+{\beta }_{2}X;\rho \left)\right]\end{array}\right\}$$

(23)

In Equations (20)–(23), Φ(·) is the univariate standard normal cumulative distribution function and F(·) is the bivariate normal cumulative distribution function.

4.2. Data

China has experienced rapid development of digital finance and has established a relatively comprehensive digital financial infrastructure. This provides an appropriate context for examining the relationship between digital inclusive finance and the employment decisions of MSEs. In addition, the availability of the Peking University Digital Inclusive Finance Index allows us to match provincial-level digital financial development information with the MSE survey data. This enables us to examine the relationship between the regional digital financial environment and MSEs’ labor employment decisions, as well as to explore the potential role of credit constraints in this relationship.

We draw on two 2015 sources: the China Micro- and Small-Enterprise Survey (CMES), conducted by Southwestern University of Finance and Economics, and the Digital Inclusive Finance Index compiled by the Institute of Digital Finance at Peking University. The 2015 CMES wave spans 28 provincial-level units, and its sample is representative enough to support analysis of MSEs’ operations, financing, and labor employment. According to “Standards for the Classification of Small and Medium Enterprises” (MIIT Document No. 300, 2011), enterprises are categorized as large-, medium-, small-, and micro-sized enterprises based on industry characteristics and indicators such as the number of employees, operating revenue, and total assets. In this study, Micro- and Small Enterprises refer to enterprises classified as small or micro enterprises according to these official criteria. For example, in the industrial sector, enterprises with fewer than 50 employees or annual operating revenue below 5 million RMB are classified as Micro- and Small Enterprises, while other sectors have their own specific threshold criteria. This official classification standard guarantees the authority and comparability of CMES data in academic research and policy analysis.

The data were processed as follows. First, the MSEs’ questionnaire and Peking University’s Digital Inclusive Finance Index were matched by province. Second, samples with missing data or recorded as “don’t know” for variables characterizing MSEs, such as size and operating years, were excluded. This screening left a final cross-section of 3263 firms.

4.3. Variable Settings

4.3.1. Dependent Variable

MSEs’ employment choices are measured from responses to the CMES questionnaire item “Currently, what are the types of regular employees in your enterprise?” A firm is coded as employing formal workers if its staff include employees who have signed a full-time labor contract of more than one year with the respondent enterprise and are past probation. It is coded as employing informal workers if its staff include temporary workers on contracts of one year or less, full-time workers without a contract, probationary workers, or part-time workers. The two behaviors are counted separately and enter the employment decision equations as the two dependent variables. Because this coding reflects heterogeneity in employment forms, it keeps the two hiring decisions apart and gives a more objective picture of how digital inclusive finance relates to MSEs’ employment choices.

The two employment indicators are coded independently rather than as mutually exclusive categories. Firms employing both formal and informal workers are therefore assigned a value of one for both indicators. This treatment preserves mixed employment arrangements, which account for 20.69% of the sample. Firms employing only formal workers are coded (1, 0), firms employing only informal workers are coded (0, 1), and firms employing both types are coded (1, 1).

4.3.2. Core Independent Variables

Using an analytic hierarchy approach with coefficient-of-variation weighting, the digital finance group at Peking University compiled the Digital Inclusive Finance Index for 2011–2020, covering 31 provincial-level units, 337 prefecture-level cities, and roughly 2800 counties across China. In this study, the regional information of MSEs is only identified at the provincial level. Therefore, to examine the relationship between digital inclusive finance and the labor employment of MSEs, we match the 2015 provincial-level Digital Inclusive Finance Index with firm-level survey data. The provincial-level Digital Inclusive Finance Index provides a useful measure of the overall digital financial development environment and infrastructure of a region. The financing behavior and employment decisions of MSEs may be associated with the supply of digital inclusive financial services and the availability of digital credit in their respective provinces. In addition, MSEs generally lack direct firm-level measures of digital inclusive finance. The use of a provincial-level Digital Inclusive Finance Index therefore allows us to capture differences in the external digital inclusive financial environment across regions and reduces potential measurement-related endogeneity associated with firm-level measures of digital inclusive finance. Accordingly, the provincial-level Digital Inclusive Finance Index is suitable as a key explanatory variable for examining the relationship between digital inclusive finance and firms’ labor employment decisions.

4.3.3. Mediating Variable

In this study, credit constraints refer to the formal credit constraints faced by MSEs from financial institutions during their production and business operations. We follow the Direct Elicitation Method (DEM) proposed by Boucher et al. [111] and apply it to the relevant questionnaire items. This allows us to identify the types of credit constraints faced by MSEs and to construct binary variables indicating whether an enterprise is credit-constrained.

The 2015 CMES questionnaire first asked MSEs about their demand for bank (credit union) loans when investigating their financing conditions. For MSEs that had not applied for loans, the questionnaire further asked about the reasons for not applying. Based on their responses, we identify the specific types of credit constraints faced by MSEs. MSEs that “needed a loan, applied for one but were rejected” or whose “loan application met only part of their financing needs” are classified as facing supply-side credit constraints. For enterprises that “needed a loan but had never applied,” the questionnaire asked why they had not done so. The options were as follows: (1) not knowing how to apply for a loan; (2) expecting that the loan application would not be approved; (3) finding the application process cumbersome; (4) considering the loan interest rate too high; (5) finding the repayment period or method unsuitable for their needs; (6) not knowing any staff members at the bank or credit union; (7) lacking collateral or a guarantor; and (8) being concerned about their ability to repay the loan. Among these responses, enterprises selecting reasons (1), (2), (3), (6), (7), or (8) are classified as facing demand-side credit constraints, as these reasons primarily reflect constraints related to transaction costs, perceived risks, and other non-price factors.

Accordingly, the credit constraint variables are coded as binary variables, with a value of 1 indicating that the MSEs face the corresponding type of credit constraint and 0 otherwise.

4.3.4. Control Variables

This paper selects a series of control variables from the enterprise level that may have an impact on labor employment. The control variables are defined as follows. (1) Size: the natural logarithm of the firm’s current asset scale (in yuan). Larger firms tend to have steadier production and labor demand and greater capacity to bear risk, which allows them to absorb the fixed costs and compliance expenditures that formal employment entails. They are therefore more likely to hire formal employees. By contrast, smaller firms rely more on informal employment to preserve labor flexibility. (2) Operating years: the survey year (2015) minus the firm’s founding year, plus one. Firms with longer operating histories generally have more mature organizational and employment arrangements, together with a more stable reputation and cash flow, all of which facilitate formal employment relationships. By contrast, newly established and younger firms face greater employment uncertainty and are more likely to use informal hiring. (3) Property right: a dummy variable equal to 1 for state-owned enterprises and 0 otherwise. State-owned enterprises face stronger institutional constraints and social responsibility requirements, which push them toward more standardized employment practices and a greater propensity to hire formal employees. By contrast, non-state-owned enterprises have more latitude to use informal employment when balancing labor costs against flexibility. (4) Science and technology: a dummy variable equal to 1 for high-tech MSEs and 0 otherwise. High-tech firms depend more heavily on specialized skills and continuous innovation and therefore require human capital that is stable and cumulative, which makes them more likely to attract and retain talent through formal employment. By contrast, non-high-tech firms tend to have more flexible labor structures and may employ a higher share of informal workers. (5) Association: a dummy variable equal to 1 if the firm has joined an industry association and 0 otherwise. Membership gives firms access to industry norms, policy information, and management experience, which strengthens compliance awareness and standardizes employment practices and thereby raises the likelihood of formal hiring. By contrast, firms outside such associations face weaker external normative constraints and are more likely to adopt informal arrangements. (6) Profitability: a dummy variable equal to 1 if the firm was profitable in the previous fiscal year and 0 otherwise. Profitable firms have stronger cash flow and payment capacity and are better able to bear the recurring costs of formal employment, such as wages and social insurance contributions. By contrast, loss-making firms are more likely to cut rigid labor costs by turning to informal employment. (7) Taxation: the firm’s subjective assessment of its tax and fee burden on a five-point scale, with larger values indicating a heavier burden. A heavier tax burden raises the overall cost of employing labor. Firms that perceive the burden as heavy are therefore more likely to reduce formal hiring and expand informal hiring in order to lower fixed labor costs. (8) Compensation: the natural logarithm of ordinary employees’ average pretax monthly income in the previous year (in yuan). Compensation reflects both labor costs and the firm’s willingness to pay for labor quality. Firms that pay more are more inclined to secure higher-quality human capital through formal employment and to maintain stable labor relations. By contrast, cost-sensitive firms may rely more on informal employment. (9) Employee innovation capability: the degree to which the firm values employees’ innovation capability, measured on a six-point scale, with larger values indicating greater emphasis. Innovation activities are typically characterized by high uncertainty and project-based tasks, so firms that attach greater weight to employees’ innovation capability may prefer flexible staffing arrangements that allow them to reallocate labor efficiently and contain fixed labor costs. Such firms are therefore expected to make relatively greater use of informal employment. By contrast, firms that attach less importance to innovation face more routine labor demands and are more likely to meet them through formal employment. (10) Enterprise management system: the strictness with which the firm implements its basic management systems, measured on a five-point scale, with larger values indicating stricter enforcement. A sound management system does not necessarily imply exclusive reliance on formal employment. It may instead enable the firm to allocate formal and informal workers more efficiently according to different operational needs, thereby maintaining organizational efficiency and labor flexibility at the same time. Firms with better-enforced systems are thus expected to combine a stable formal core with a relatively larger share of informal workers. By contrast, firms whose systems are loosely enforced or exist in name only have weaker capacity to manage a differentiated workforce and depend more on conventional formal arrangements.

The empirical model also controls for region fixed effects. Industry dummies follow the classification used in the Standard Provisions for the Classification of Small and Medium-Sized Enterprises, which distinguishes 19 industries, so 18 industry dummies are included to absorb industry fixed effects.

In addition, we use the uncentered variance inflation factor (VIF) to detect multicollinearity. The mean VIF of the model is 5.25, well below the conventional threshold of 10, suggesting that multicollinearity is not a serious concern.

The definitions and descriptive statistics of the variables are shown in

Table 1. Definitions of the dependent, independent, mediating and control variables.

Variable Type	Variable Name	Variable Explanation
Dependent Variables	Formal employment	Whether the enterprise employs formal staff (Yes = 1, No = 0)
	Informal employment	Whether the enterprise employs informal staff (Yes = 1, No = 0)
Independent Variable	Digital inclusive finance	Provincial Peking University Digital Inclusive Finance Index in 2015
Control Variables	Size	Natural logarithm of the enterprise’s current assets (Yuan)
	Operating years	2015 − year of business commencement + 1
	Property rights	State-owned enterprise = 1; otherwise = 0
	Science and technology	High-tech enterprise = 1; otherwise = 0
	Industry association	Member of an industry association = 1; otherwise = 0
	Profitability	Positive profit in the previous year = 1; otherwise = 0
	Taxation	Self-assessed tax burden: 1 = very light … 5 = very heavy
	Employee compensation	Natural logarithm of average pre-tax monthly employee income in the previous year (Yuan)
	Employee innovation capability	Importance attached to employees’ innovation ability: 1 = very important … 6 = no need
	Enterprise management system	Strictness of basic management systems: 1 = very strict … 5 = nominal
	Eastern region	Eastern region = 1; otherwise = 0
	Central region	Central region = 1; otherwise = 0
Mediating Variable	Credit constraint	Whether the MSE faces formal credit constraints from financial institutions (Yes = 1, No = 0)

.

Table 2. Descriptive statistics of the variables.

Variables	Mean	Std. Dev.
Formal employment	0.7680	0.4222
Informal employment	0.4389	0.4963
Digital inclusive finance	234.4276	25.8031
Credit constraint	0.2951	0.4562
Size	14.3845	2.4603
Operating years	9.6859	7.6509
Property rights	0.0138	0.1166
Science and technology	0.0641	0.2449
Industry association	0.3690	0.4826
Profitability	0.5020	0.5001
Taxation	3.1615	1.1336
Employee compensation	7.4585	2.1412
Employee innovation capability	2.1918	1.2256
Enterprise management system	2.2326	0.8499
Eastern region	0.5728	0.4947
Central region	0.1952	0.3964

shows that only 43.89% of the sampled MSEs employ informal workers. The dispersion of the Digital Inclusive Finance Index confirms that provinces differ considerably in how far digital inclusive finance has developed. Credit constraints average 0.2951, so close to 30% of these firms are credit constrained. Firm size averages 14.38 and operating years 9.69; both are basic firm characteristics that cannot be ignored when studying what drives MSE employment. Ownership averages 0.01, meaning that state-owned firms make up only a small share of the sample, while the science-and-technology dummy averages 0.06, corresponding to 206 technology-based firms. Since technological progress can substitute for part of the workforce, it may in turn alter labor demand. Industry-association membership proxies to some degree for competitive pressure within a firm’s industry and hence for its hiring behavior. Profitability averages 0.50, so roughly half the firms failed to turn a profit in the previous year; because profit is a precondition for hiring, this bears directly on employment decisions. Taxation averages 3.16, suggesting that these firms do not regard their tax burden as heavy. Turning to employee characteristics, average compensation is 7.46 in logarithms, while the weight placed on employees’ innovation capability and on the enforcement of basic management systems averages 2.19 and 2.23 respectively. Both lie toward the strict end of their scales, indicating that the sampled MSEs value employee innovation and enforce their management systems fairly rigorously.

Table 3. Labor employment decisions in MSEs.

Labor Employment in MSEs	Formal Staff Employed		No Formal Staff Employed		Add Up the Total
	Number	Percentage (%)	Number	Percentage (%)	Number	Percentage (%)
Informal staff employed	675	20.69	757	23.20	1432	43.89
No informal staff employed	1831	56.11	0	0	1831	56.11
Add up the total	2506	76.80	757	23.20	3263	100

breaks the sample down further: 20.69% of MSEs employ both formal and informal workers. A plausible explanation is that informal employees, by their cost advantage, can compensate for the cost constraints of the firm’s production and operation and act as a substitute for formal employees. A further 56.11% employ formal workers only, so more than half of these firms rely exclusively on formal employment, while 23.20% employ informal workers only. This may be because, as the marketization process deepens and the social security system for employees begins to be gradually improved, the number of enterprises that “purely employ informal employees” begins to decrease.

5. Econometric Results

5.1. Main Results

Table 4. Digital inclusive finance and labor employment decisions of MSEs.

Variables	(1)	(2)	(3)	(4)	(5)	(6)
	Formal	Informal	Formal	Informal	Formal	Informal
Digital inclusive finance	0.0054 ***	−0.0060 ***	0.0048 ***	−0.0048 ***	0.0053 ***	−0.0076 ***
	(0.0010)	(0.0009)	(0.0010)	(0.0009)	(0.0015)	(0.0013)
Credit constraints			−0.1956 ***	0.3291 ***	−0.1446 **	0.2384 ***
			(0.0550)	(0.0506)	(0.0571)	(0.0527)
Size			0.0530 ***	−0.0169 *	0.0660 ***	−0.0320 ***
			(0.0104)	(0.0094)	(0.0118)	(0.0102)
Operating years			−0.0014	0.0020	0.0010	0.0000
			(0.0034)	(0.0030)	(0.0035)	(0.0031)
Property right			0.5314 *	−0.1666	0.4618	−0.1071
			(0.2797)	(0.2103)	(0.2848)	(0.2127)
Science and technology			0.4996 ***	−0.2229 **	0.4057 ***	−0.1250
			(0.1293)	(0.0955)	(0.1421)	(0.1031)
Industry			0.2363 ***	−0.0189	0.2023 ***	−0.0073
Association			(0.0546)	(0.0478)	(0.0556)	(0.0487)
Profitability			0.0505	−0.0084	0.0503	−0.0137
			(0.0504)	(0.0454)	(0.0511)	(0.0459)
Taxation			0.1375 ***	−0.1036 ***	0.1272 ***	−0.0748 ***
			(0.0220)	(0.0206)	(0.0259)	(0.0238)
Employee compensation			0.0290 ***	0.0085	0.0248**	0.0129
			(0.0112)	(0.0107)	(0.0117)	(0.0114)
Employee innovation capacity			−0.1045 ***	0.0595 ***	−0.0980 ***	0.0639 ***
			(0.0206)	(0.0198)	(0.0212)	(0.0203)
Enterprise management			−0.0784 ***	0.1162 ***	−0.0837 ***	0.1178 ***
system			(0.0302)	(0.0284)	(0.0313)	(0.0290)
Industry	No	No	No	No	Yes	Yes
Region	No	No	No	No	Yes	Yes
Cons	−0.5370 **	1.2518 ***	−1.4106 ***	0.9959 ***	−1.7126 ***	1.8630 ***
	(0.2310)	(0.2045)	(0.2987)	(0.2677)	(0.4259)	(0.3889)
Obs	3263		3263		3263
Log-L	−3201.7840		−3035.0172		−2960.5168
Wald’s chi2	51.03		408.13		1834.07
ρ	−0.9934 (0.2004)		−0.9916 (0.0048)		−0.9895 (0.0029)
p-Value	Prob > χ2 = 0.0000		Prob > χ2 = 0.0000		Prob > χ2 = 0.0000

Note: The symbols *, **, and *** denote statistical significance at the 10, 5, and 1% levels, respectively.

reports the estimates. Columns (1)–(2) include the core variable alone, columns (3)–(4) add the control variables, and columns (5)–(6) further control for industry and region fixed effects. The estimated ρ is −0.9934, −0.9916, and −0.9895 across the three specifications and is significant in every likelihood-ratio test, pointing to a “substitution” relationship between the two hiring decisions. Digital inclusive finance is positively related to the probability of hiring formal employees and negatively related to the probability of hiring informal employees, whether or not controls and fixed effects are included. Both associations are significant at the 1% level. As digital inclusive finance develops, MSEs become more likely to hire formal employees and less likely to take on informal ones. Thereby, Hypothesis 1 is verified. A reasonable explanation is that digital inclusive finance improves labor employment in MSEs at an affordable cost. On the demand side, enterprises can then hire more formal employees to strengthen their competitiveness. On the supply side, becoming a formal employee means an increase in social security and labor compensation, achieving a win-win situation between “enterprises” and “employees”. Considering that there is a substitution relationship between formal and informal employees, we find that, at the level of labor employment, the development of digital inclusive finance promotes the “formalization of informal employment” in MSEs. For policymakers, this finding carries a clear implication. They should keep developing the digital inclusive financial system, widen the reach of MSE financing, and improve the supply of digital financial services. Using financial support to guide firms toward standardized employment can then advance three goals at once: better financing, stronger enterprises, and higher-quality jobs.

For formal and informal employment respectively (columns 5–6 of Table 4), the coefficients on credit constraints are −0.1446 and 0.2384, significant at the 5% and 1% levels. Firms under tighter credit constraints are therefore less likely to employ formal workers and more likely to rely on informal ones. Hypothesis 2 is therefore supported. One possible explanation is that, on one hand, credit constraints restrict the expansion of firm size and reduce the employment scale of formal workers. On the other hand, formal workers command higher wages, while informal workers are cheaper and easier to deploy. Firms therefore substitute informal workers for formal ones to keep basic production going. Although this adjustment can alleviate firms’ short-term capital pressure to a certain extent, it may also undermine employee stability, human capital accumulation, and firms’ long-term productive efficiency, which is detrimental to the sustainable development of enterprises. From a policy perspective, the financing support system for MSEs therefore needs further work: wider financing channels, lower financing costs, and looser credit constraints. Easing these pressures would allow firms to expand formal employment and would raise standards across the labor market, so that enterprise development and high-quality employment advance together.

For the remaining control variables (columns 5–6 of Table 4), the coefficients of firm size are 0.0660 for formal employment and −0.0320 for informal employment, respectively, both significant at the 1% level. These results suggest that larger MSEs are more likely to adopt formal employment arrangements and reduce their reliance on informal employment. An explanation is that, according to the theory of labor demand and the “capital–skill” complementary hypothesis, enterprise expansion requires a corresponding increase in skilled labor and a more stable capital–labor structure. Therefore, as firms grow in scale and strengthen their operational capabilities, they tend to establish more standardized employment practices to support production and development. The coefficient of science and technology on formal employment is 0.4057 and is significant at the 1% level, so technology-based enterprises are more likely to employ formal workers. Research and development in such firms demands specialized skills and accumulated knowledge, which plausibly explains the pattern. Accordingly, such enterprises rely more heavily on stable formal employment relationships to reduce innovation costs associated with employee turnover, thereby increasing their demand for human capital. The coefficient of industry association is 0.2023, which is significant at the 1% level, indicating that MSEs belonging to an industry association are more likely to employ formal workers. Membership gives a firm better access to market information, financing resources, and policy support, and so strengthens its capacity to expand formal employment. The coefficient of taxation is 0.1272 for formal employment and −0.0748 for informal employment. Taken together, the two signs imply that larger and more profitable firms bear heavier tax burdens. Such firms are nevertheless better able to employ formal workers and less dependent on informal ones. In contrast, the coefficients of enterprises’ emphasis on employee innovation capabilities and the soundness of enterprise management systems are −0.0980 and −0.0837 for formal employment, and 0.0639 and 0.1178 for informal employment, respectively, all significant at the 1% level. These results suggest that MSEs placing greater emphasis on employee innovation capabilities and possessing more sound management systems tend to rely less on formal employment and more on informal employment. One possible explanation is that innovation activities are often characterized by high uncertainty and project-based tasks, leading firms to adopt more flexible employment arrangements in order to improve labor allocation efficiency and reduce fixed labor costs. Moreover, sound management systems do not necessarily imply exclusive reliance on formal employment. Instead, they may enable firms to allocate formal and informal workers more efficiently according to different operational needs, thereby maintaining both organizational efficiency and labor flexibility. These findings suggest that standardized management is not intended to eliminate informal employment, but rather to optimize the allocation of different forms of employment. For policymakers, improving labor protection systems and regulatory frameworks for informal employment may help balance enterprise flexibility with worker protection.

To facilitate the interpretation of the economic magnitude, we further report the average marginal effects (AMEs) from the bivariate Probit model. As shown in

Table 5. Average marginal effects of digital inclusive finance on MSEs’ employment decisions.

	(1)	(2)
Variables	Formal	Informal
Digital inclusive finance	0.0015 ***	−0.0028 ***
	(0.0004)	(0.0005)
Credit constraints	−0.0399 **	0.0876 ***
	(0.0157)	(0.0192)
Size	0.0182 ***	−0.0118 ***
	(0.0032)	(0.0037)
Operating years	0.0003	5.75 × 106
	(0.0010)	(0.0011)
Property right	0.1273	−0.0394
	(0.0786)	(0.0782)
Science and	0.1118 ***	−0.0459
technology	(0.0392)	(0.0379)
Industry	0.0558 ***	−0.0027
association	(0.0153)	(0.0179)
Profitability	0.0139	−0.0051
	(0.0141)	(0.0169)
Taxation	0.0351 ***	−0.0275 ***
	(0.0071)	(0.0087)
Employee	0.0068 **	0.0047
compensation	(0.0032)	(0.0042)
Employee innovation	−0.0270 ***	0.0235 ***
capacity	(0.0058)	(0.0074)
Enterprise	−0.0231 ***	0.0433 ***
Management system	(0.0086)	(0.0106)

Note: The symbols ** and *** denote statistical significance at the 5 and 1% levels, respectively.

, the AME of digital inclusive finance is 0.0015 for formal employee hiring and −0.0028 for informal employee hiring. This indicates that a one-unit increase in the Digital Inclusive Finance Index is associated with an approximately 0.15 percentage-point higher probability of formal employee hiring and a 0.28 percentage-point lower probability of informal employee hiring, holding other variables constant.

The AMEs of credit constraints are −0.0399 for formal employee hiring and 0.0876 for informal employee hiring. These results suggest that greater credit constraints are associated with a lower probability of formal employee hiring and a higher probability of informal employee hiring. Overall, the marginal-effect results are consistent with the direction of the baseline bivariate Probit estimates and provide a more intuitive interpretation of their economic magnitude.

5.2. Validation of Mechanism of Action

The analysis above shows that digital inclusive finance is positively associated with formal hiring and negatively associated with informal hiring in MSEs. To identify the channel behind these associations, we estimate the following mediation model:

$$Em{p}_i=cDIF+\alpha X_i+{\epsilon }_{i1}$$

(24)

$$Creditrationin{g}_i=aDIF+\beta X_i+{\epsilon }_{i2}$$

(25)

$$Em{p}_i=c^{\prime }DIF+bCreditrationin{g}_i+\gamma X_i+{\epsilon }_{i3}$$

(26)

Emp_i denotes the labor employment decision of enterprise i, with whether to hire formal or informal employees as the division criterion. DIF is the development level of digital inclusive finance. X_i denotes a series of control variables affecting the firm’s labor employment decisions. Creditrationing_i denotes the level of credit constraints of firm i. Meanwhile, a, b, c, c′ are the parameters to be estimated, and ε_i1, ε_i2, ε_i3 are the random disturbance terms. Model (24) and model (26) are estimated using a bivariate Probit model, and model (25) is estimated using a Probit model. Mediation is tested in the following sequence. We first examine c; if it is insignificant, the procedure stops, and if it is significant, we move on to a and b. Where c is significant and a and b are both significant, we then inspect c′: a significant c′ indicates partial mediation and an insignificant one full mediation. Where c is significant but at least one of a and b is not, we fall back on the Sobel test and judge the mediation effect significant or otherwise according to its outcome. The mediation proportion is computed by stepwise regression. Three regressions are run in sequence: the total effect, the independent variable on the mediator, and the independent and mediator variables together. These yield the coefficients c, a, b, and c′. The mediation effect is calculated as a × b. When the product ab has the same sign as the total effect c, the mediation proportion is computed as ab/c.

Table 6. Mechanisms of digital inclusive finance affecting the role of labor employment decisions of MSEs.

Variables	(1)	(2)	(3)	(4)	(5)
	Formal	Informal	Credit Constraints	Formal	Informal
Digital Inclusive Finance	0.0016 ***	−0.0031 ***	−0.0376 ***	0.0015 ***	−0.0028 ***
	(0.0004)	(0.0005)	(0.0004)	(0.0004)	(0.0005)
Credit constraints				−0.0399 **	0.0876 ***
				(0.0157)	(0.0192)
Industry	Yes	Yes	Yes	Yes	Yes
Region	Yes	Yes	Yes	Yes	Yes
CVs	Yes	Yes	Yes	Yes	Yes
Obs	3263	3263	3263	3263	3263
Log-L	−2970.9365		−1742.058	−2960.2895
Wald’s chi2	1848.49		402.68	1869.80
ρ	−0.9877 (0.0030)			−0.9876 (0.0031)
p-value	Prob > χ2 = 0.0000		Prob > χ2 = 0.0000	Prob > χ2 = 0.0000

Note: The symbols ** and *** denote statistical significance at the 5 and 1% levels, respectively.

reports the AMEs. Digital inclusive finance is associated with a higher probability of formal employment and a lower probability of informal employment (columns 1–2). It is also associated with a lower probability of facing credit constraints (column 3). After credit constraints are included (columns 4–5), the AMEs of digital inclusive finance become smaller in magnitude but retain the same signs, while credit constraints themselves are negatively associated with formal employment and positively associated with informal employment. These patterns are consistent with a partial credit-constraint channel. Using the AME-based product formula, the mediated shares are approximately 9.17% for formal employment and 10.58% for informal employment. Because the sample is cross-sectional, we interpret these estimates as suggestive mechanism evidence rather than as definitive causal mediation effects.

5.3. Endogeneity Discussion

It is worth noting that there may be an endogeneity problem in model (1). Observable factors enter the specification as the core and control variables, but several unobservable determinants of MSE employment are omitted, which may bias the estimates. If the unobservables that affect MSE employment leave digital inclusive finance untouched, that is, if digital inclusive finance is exogenous, the estimated effect on employment choices is consistent. Therefore, to solve the possible endogeneity problem in the model, we introduce an instrumental variable in the bivariate Probit model to deal with it. We use the telephone penetration rate as an alternative instrumental variable for digital inclusive finance. Telephone penetration rate refers to the average number of telephone sets owned per 100 people among the total population of an administrative region during the reporting period. Regarding relevance, the telephone penetration rate in 2010 captures the stock of communication infrastructure across provinces, including trunk transmission capacity, switching capacity, and last-mile access networks. These infrastructure assets were subsequently upgraded and incorporated into the broadband and mobile networks that underpin digital financial services. Therefore, provinces with more extensive telephone networks faced lower marginal costs in deploying digital payment, mobile banking, and online lending services, while their residents were also more accustomed to remote transactions, thereby reducing behavioral frictions in the adoption of digital financial services. Historical communication infrastructure is closely associated with the subsequent diffusion of Internet services and digital financial technologies; therefore, regions with higher telephone penetration are expected to exhibit faster development of digital inclusive finance [112,113]. Regarding the exclusion restriction, two features suggest that this instrument is unlikely to affect the employment structure of Micro- and Small Enterprises (MSEs) through channels other than digital inclusive finance. First, the instrument is predetermined: it is measured in 2010, predating the development of digital inclusive finance, the subsequent rise in mobile payments in China, and the employment outcomes observed in our data by five years. It therefore cannot capture firms’ contemporaneous hiring decisions. Second, the instrument reflects communication capacity during the era of voice-based telecommunications rather than the platform-based digital environment prevailing in 2015. It was not the medium underlying e-commerce, platform-based sales, online recruitment, or digital labor matching. Therefore, unlike contemporaneous Internet penetration, the direct employment channels associated with digital connectivity are unlikely to be captured by telephone penetration measured in 2010. Indeed, fixed-line telephone usage in China had already passed its national peak before 2015 and was gradually being replaced by mobile telecommunications. Consequently, the provincial distribution of telephone penetration in 2010 primarily reflects the historical capacity of communication networks rather than the digital business environment faced by firms in 2015.

We further control for firm-level covariates as well as industry and region fixed effects. Because the exclusion restriction cannot be tested directly with a single instrument, the IV estimates are interpreted as supporting evidence rather than as definitive proof of causality.

It is calculated as follows:

Telephone penetration rate = (Total number of telephone sets/Total population of the administrative region) × 100

Because the bivariate Probit model offers no direct correction for endogeneity, we follow Roodman [114] and adopt a two-step procedure. The first step regresses digital inclusive finance on the instrument and the remaining exogenous variables and retains the fitted values. The second step enters those fitted values as the core regressor and re-estimates the system with the conditional mixed process (CMP) estimator.

Based on the results of the one-stage regression (

Table 7. Results of one-stage regression of bivariate Probit model considering endogeneity of variables.

Variables	Digital Inclusive Finance
Telephone penetration in 2010	0.6565 ***
	(0.0035)
Other control variables	YES
Obs	3263
R2	0.9151
F-value	35,158.78
p-Value	Prob > F = 0.0000

Note: The symbols *** denote statistical significance at the 1% levels, respectively.

), the coefficient of the impact of 2010 telephone penetration on digital inclusive finance is 0.6565, which is significant at the 1% level. This result indicates that historical telephone infrastructure can significantly promote the development of digital inclusive finance. We also use an IV-Probit model to estimate the relationship between digital inclusive finance and the credit constraints of MSEs (column 3 of

Table 8. Results of the two-stage estimation and validation of the mechanism of action considering the endogeneity of variables.

	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Variables	Formal	Informal	Credit Constraints	Formal	Informal	Formal	Informal
Digital inclusive finance	0.0050 ***	−0.0086 ***	−0.0130 ***	0.0043 ***	−0.0077 ***	0.0052 ***	−0.0079 ***
	(0.0016)	(0.0015)	(0.0016)	(0.0016)	(0.0015)	(0.0018)	(0.0017)
Credit constraints				−0.1435 **	0.2386 ***	−0.0977 *	0.2792 ***
				(0.0571)	(0.0527)	(0.0566)	(0.0544)
Industry	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Region	Yes	Yes	Yes	Yes	Yes	Yes	Yes
CVs	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cons	−1.6071 ***	2.0926 ***	1.4088 ***	−1.4715 ***	1.9021 ***	−1.9192 ***	2.1707 ***
	(0.4592)	(0.4180)	(0.4608)	(0.4610)	(0.4217)	(0.5028)	(0.4827)
Obs	3263	3263	3263	3263	3263	3263	3263
Log-L	−14,180.645		−12,898.993	−14,170.438		−23,385.658
Wald’s chi2	71,252.02		403.05	71464.77		7019.19
p-Value	Prob > χ2 = 0.0000		Prob > χ2 = 0.0000	Prob > χ2 = 0.0000		Prob > χ2 = 0.0000

Note: The symbols *, **, and *** denote statistical significance at the 10, 5, and 1% levels, respectively.

). (This study uses the 2010 telephone penetration rate as an instrumental variable for the core explanatory variable, digital financial inclusion. First-stage results are obtained from the ‘iv probit’ command with the ‘first’ option. The coefficient on the 2010 telephone penetration rate is highly significant at the 1% level, satisfying the relevance condition and indicating no weak-instrument problem. The Wald test of exogeneity yields a χ² statistic of 0.67 with a p-value of 0.4134; thus, the null that the core regressor is exogenous cannot be rejected at the 5% level. Digital financial inclusion therefore exhibits no material endogeneity bias, and the IV-Probit estimates are robust and valid.) The second-stage regression results and mechanism of the bivariate Probit model using the Conditional Mixed Process Estimator (CMP) are shown in columns (1) to (5) of Table 8.

After the potential endogeneity in digital inclusive finance is accounted for, the results still show a positive association with formal employment and a negative one with informal employment among MSEs. Both coefficients are significant at the 1% level. Meanwhile, the results provide evidence consistent with the mechanism that the development of digital inclusive finance is associated with reduced credit constraints among MSEs, which is in turn associated with increased formal employment and reduced informal employment.

5.4. Robustness Tests

5.4.1. Alternative Measures of Formal and Informal Employment

To further examine the robustness of our findings, we use the number of formal and informal employees as alternative measures of labor employment. Specifically, the dependent variable in Equation (1) is the natural logarithm of the number of formal employees plus one, while the dependent variable in Equation (2) is the natural logarithm of the number of informal employees plus one.

The results are reported in

Table 9. Robustness test based on the number of formal and informal employees.

	(1)	(2)	(3)	(4)
Variables	ln(Formal + 1)	ln(Informal + 1)	ln(Formal + 1)	ln(Informal + 1)
Digital inclusive finance	0.0045 **	−0.0167 ***	0.0049 ***	−0.0179 ***
	(0.0018)	(0.0029)	(0.0018)	(0.0030)
Credit constraints	−0.1423 *	0.5615 ***	−0.1464 *	0.5749 ***
	(0.0733)	(0.1161)	(0.0752)	(0.1159)
Industry	Yes	Yes
Region	Yes	Yes
CVs	Yes	Yes	Yes	Yes
Cons	−2.0529 ***	1.9319 *	−2.4958 ***	2.6939 ***
	(0.7021)	(1.1468)	(0.5265)	(0.7889)
Obs	3263	3263	3263	3263
Log-L	−9735.2098		−5673.0907	−4499.2548
Wald’s chi2	874.271
ρ	−0.5203 (0.0140)
p-Value	Prob > χ2 = 0.0000

Note: The symbols *, **, and *** denote statistical significance at the 10, 5, and 1% levels, respectively.

. Columns (1) and (2) report the results from the bivariate Tobit model. Columns (3) and (4) report the results from separate Tobit models for formal and informal employees, respectively. The results show that the coefficient of digital inclusive finance is positive and statistically significant for the number of formal employees. In contrast, its coefficient is negative and statistically significant for the number of informal employees.

These results are consistent with our baseline findings. They suggest that the main findings remain broadly stable when employment is measured by the number of formal and informal employees rather than by the original binary measures.

5.4.2. Lagged Digital Inclusive Finance Index

When the lagged index is used (

Table 10. Robustness test considering the time lag in the effect of digital inclusive finance.

	(1)	(2)	(3)	(4)	(5)
Variables	Formal	Informal	Credit Constraints	Formal	Informal
Digital inclusive finance	0.0059 ***	−0.0084 ***	−0.0124 ***	0.0053 ***	−0.0076 ***
	(0.0015)	(0.0013)	(0.0014)	(0.0015)	(0.0013)
Credit constraints				−0.1446 **	0.2384 ***
				(0.0571)	(0.0527)
Industry	Yes	Yes	Yes	Yes	Yes
Region	Yes	Yes	Yes	Yes	Yes
CVs	Yes	Yes	Yes	Yes	Yes
Cons	−1.8448 ***	2.0522 ***	1.2619 ***	−1.7126 ***	1.8630 ***
	(0.4229)	(0.3854)	(0.4252)	(0.4259)	(0.3889)
Obs	3263	3263	3261	3263	3263
Log-L	−2970.6704		−1740.6665	−2960.5168
Wald’s chi2	1969.86		407.93	1834.07
ρ	−0.9898 (0.0029)			−0.9895 (0.0029)
p-value	Prob > χ2 = 0.0000		Prob > χ2 = 0.0000	Prob > χ2 = 0.0000

Note: The symbols ** and *** denote statistical significance at the 5 and 1% levels, respectively.

), digital inclusive finance remains positively associated with the hiring of formal employees and negatively associated with the hiring of informal employees in MSEs, with credit constraints again acting as a mediator. Re-matching the 2014 Digital Inclusive Finance Index to the 2015 CMES data therefore leaves the findings unchanged.

5.4.3. Robustness Test Using Province-Level Clustered Standard Errors

In the baseline analysis, the standard errors are estimated under the conventional specification. However, observations from the same province may share common economic, institutional, and financial characteristics, which could result in within-province correlation in the error terms. To account for this potential correlation, we re-estimate the baseline models using province-level clustered standard errors.

The results are reported in

Table 11. Robustness checks with province-clustered standard errors.

Variables	(1)	(2)	(3)	(4)	(5)
	Formal	Informal	Credit Constraints	Formal	Informal
Digital Inclusive Finance	0.0059 ***	−0.0084 ***	−0.0124 ***	0.0053 ***	−0.0076 ***
	(0.0020)	(0.0011)	(0.0026)	(0.0020)	(0.0011)
Credit constraints				−0.1446 **	0.2384 ***
				(0.0666)	(0.0593)
Industry	Yes	Yes	Yes	Yes	Yes
Region	Yes	Yes	Yes	Yes	Yes
CVs	Yes	Yes	Yes	Yes	Yes
Cons	−1.8448 ***	2.0522 ***	1.2619 *	−1.7126 ***	1.8630 ***
	(0.5234)	(0.3330)	(0.7634)	(0.5344)	(0.3308)
Obs	3263	3263	3261	3263	3263
Log-L	−2970.6704		−1740.6665	−2960.5168
Wald’s chi2	154.664			166.363
ρ	−0.9897746 (0.0043117)			−0.9895417 (0.0042332)
p-value	Prob > χ2 = 0.0000			Prob > χ2 = 0.0000

Note: The symbols *, **, and *** denote statistical significance at the 10, 5, and 1% levels, respectively.

. Columns (1) and (2) report the bivariate Probit estimates for formal and informal employment, respectively, while column (3) reports the Probit estimates for credit constraints. Columns (4) and (5) further report the bivariate Probit estimates of formal and informal employment after controlling for credit constraints. Compared with the baseline results, the coefficient of digital inclusive finance remains positive and statistically significant for formal employment and negative and statistically significant for informal employment. In addition, digital inclusive finance remains negatively associated with credit constraints. After credit constraints are included, the coefficients retain their original signs and remain statistically significant.

These results are broadly consistent with the baseline findings, indicating that the main conclusions are robust to clustering the standard errors at the province level.

5.4.4. Alternative Measure Using Separate Probit and Logit Models

To avoid relying on the cross-equation correlation of the bivariate Probit model, we estimate separate Probit and Logit models for formal employment, informal employment, and credit constraints (

Table 12. Robustness test using separate Probit and Logit estimates.

	(1)	(2)	(3)	(4)	(5)	(6)
Variables	Formal	Informal	Formal	Informal	Credit Constraints	Credit Constraints
Estimator	Probit	Probit	Logit	Logit	Probit	Logit
Digital inclusive finance	0.0052 ***	−0.0073 ***	0.0089 ***	−0.0120 ***	−0.0124 ***	−0.0216 ***
	(0.0015)	(0.0013)	(0.0027)	(0.0021)	(0.0014)	(0.0025)
Industry	Yes	Yes	Yes	Yes	Yes	Yes
Region	Yes	Yes	Yes	Yes	Yes	Yes
CVs	Yes	Yes	Yes	Yes	Yes	Yes
Cons	−1.6865 ***	1.6534 ***	−2.8442 ***	2.7333 ***	1.2619 ***	2.1640 ***
	(0.4373)	(0.3870)	(0.7639)	(0.6329)	(0.4252)	(0.7355)
Obs	3261	3261	3261	3261	3261	3261
Pseudo R2	0.0974	0.0615	0.0964	0.0619	0.1204	0.1212

Note: The symbols *** denote statistical significance at the 1% levels, respectively.

). Across these single-equation specifications, the associations retain the same signs and remain statistically significant, indicating that the main conclusions do not hinge on the joint-estimation assumption.

5.4.5. Alternative Measure Using the Internet Development Index

Finally, we employ an alternative measure of digital development to examine whether the main findings are sensitive to the measurement of digital development. Specifically, we replace the Digital Inclusive Finance Index with the Internet Development Index. The results are reported in

Table 13. Robustness test using an alternative measure of digital development.

Variables	(1)	(2)	(3)	(4)	(5)
	Formal	Informal	Credit Constraints	Formal	Informal
Internet penetration	0.0134 ***	−0.0161 ***	−0.0258 ***	0.0122 ***	−0.0142 ***
	(0.0032)	(0.0030)	(0.0032)	(0.0032)	(0.0030)
Credit constraints				−0.1495 ***	0.2477 ***
				(0.0570)	(0.0527)
Industry	Yes	Yes	Yes	Yes	Yes
Region	Yes	Yes	Yes	Yes	Yes
CVs	Yes	Yes	Yes	Yes	Yes
Cons	−1.2403 ***	0.9919 ***	−0.1756	−1.1804 ***	0.8851 ***
	(0.2964)	(0.2772)	(0.3173)	(0.2972)	(0.2792)
Obs	3263	3263	3261	3263	3263
Log-L	−2976.5322		−1746.8698	−2965.5116
Wald’s chi2	2840.36		409.91	2776.51
ρ	−0.9897246 (0.0028294)			−0.9892947 (0.002822)
p-Value	Prob > χ2 = 0.0000			Prob > χ2 = 0.0000

Note: The symbols *** denote statistical significance at the 1% levels, respectively.

. The coefficient of the alternative measure remains positive for formal employment and negative for informal employment. It is also negatively associated with credit constraints. The signs and statistical significance of the key coefficients are consistent with the baseline results, providing additional evidence that the main findings are robust to an alternative measure of digital development.

5.5. Examination of Heterogeneity

5.5.1. Different Dimensions of Digital Inclusive Finance

We further examine the labor employment effects of digital inclusive finance for MSEs across different dimensions. Breadth of coverage counts regional electronic accounts, such as mobile payment accounts and the bank accounts linked to them, and so captures the supply-side reach of digital financial infrastructure. Depth of use records how intensively those services are actually drawn on, covering payments, monetary funds, credit, insurance, and related products, and therefore reflects demand-side service capacity. The degree of digitalization captures the cost and efficiency of regional digital inclusive finance through indicators of mobility, affordability, and convenience. To further explore the impact of digital inclusive finance on labor employment in MSEs, we empirically examine each of the three dimensions of digital inclusive finance, and the regression results are reported in

Table 14. Different dimensions of digital inclusive finance and labor employment decisions of MSEs.

Variables	(1)	(2)	(3)	(4)	(5)	(6)
	Formal	Informal	Formal	Informal	Formal	Informal
Breadth of coverage	0.0048 ***	−0.0061 ***
	(0.0012)	(0.0011)
Depth of use			0.0032 ***	−0.0043 ***
			(0.0009)	(0.0008)
Degree of digitalization					−0.0079 ***	0.0052 ***
					(0.0017)	(0.0016)
Industry	Yes	Yes	Yes	Yes	Yes	Yes
Region	Yes	Yes	Yes	Yes	Yes	Yes
CVs	Yes	Yes	Yes	Yes	Yes	Yes
Cons	−1.4777 ***	1.3538 ***	−1.0722 ***	0.8923 ***	2.6892 ***	−2.0716 ***
	(0.3426)	(0.3169)	(0.2919)	(0.2710)	(0.7063)	(0.6531)
Obs	3263	3263	3263	3263	3263	3263
Log-L	−2960.5392		−2963.7435		−2966.3304
Wald’s chi2	1932.06		1768.40		2828.08
ρ	−0.9893 (0.0028)		−0.9894 (0.0029)		−0.9898 (0.0028)
p-Value	0.0000		0.0000		0.0000

Note: The symbols *** denote statistical significance at the 1% levels, respectively.

. As columns (1)–(2) and (3)–(4) show, the breadth of coverage and the depth of use are both significant at the 1% level. Each is positively associated with the hiring of formal employees and negatively associated with the hiring of informal employees in MSEs. Accordingly, the breadth of coverage and the depth of use of digital inclusive finance appear to have broadened the content and boundaries of financial services and eased firms’ credit constraints. Demand for formal employees rises as a result, while demand for informal employees falls.

The degree of digitalization is negatively associated with the employment of formal employees and positively associated with the employment of informal employees in MSEs. One possible explanation for this result is that, on the one hand, the digitalization of inclusive finance may place higher demands on local digital financial infrastructure. When the infrastructure does not adequately match the degree of digitalization, a “digital divide” may emerge. One likely manifestation is a widening income gap, which may be linked to tighter credit constraints for MSEs and, in turn, to lower formal employment (Table 14, column 5). This may also be associated with a greater tendency for MSEs to rely on informal employees due to cost considerations. On the other hand, deeper digitalization delivers financial products and services more conveniently and at lower cost. Without adequate regulation, however, it may also push firms toward financialization, drawing digital financial resources away from the real economy and opening a gap between digital financial development and enterprise development. Such a divergence may be associated with MSEs’ employment decisions and a greater reliance on informal workers as firms seek to manage the costs associated with industrial investment constraints (Table 14, column 6). Therefore, although the deepening of digitalization is associated with improved financing conditions for MSEs, attention should also be paid to the potential “digital divide” and the possible underutilization of funds. These findings highlight the potential importance of introducing and improving relevant regulatory policies.

5.5.2. Regional Heterogeneity

Considering regional differences in the development of digital inclusive finance and in local economic conditions, we examine whether the association between digital inclusive finance and MSEs’ employment decisions differs across eastern, central, and western China. Differences in statistical significance across separately estimated regional subsamples do not necessarily imply statistically significant differences between groups. We therefore estimate pooled interaction-term models rather than relying on subsample significance alone.

Specifically, we estimate bivariate Probit models on the full sample and successively include a regional indicator for the eastern, central, or western region, together with its interaction with the digital inclusive finance index. In each specification, the main coefficient on digital inclusive finance captures the association for firms outside the indicated region, while the interaction term captures the differential association for firms in that region relative to the rest of the sample.

Table 15. Digital inclusive finance and labor employment decisions of MSEs in different regions.

Variables	(1)	(2)	(3)	(4)	(5)	(6)
	Formal	Informal	Formal	Informal	Formal	Informal
Digital inclusive finance	0.0081 **	−0.0114 ***	0.0043 ***	−0.0053 ***	0.0052 ***	−0.0050 ***
	(0.0041)	(0.0038)	(0.0012)	(0.0010)	(0.0012)	(0.0011)
region	0.6274	−0.7144	−2.5087 *	2.6036 **	0.5238	0.2082
	(0.9413)	(0.8688)	(1.4014)	(1.2860)	(1.1594)	(1.0782)
Digital inclusive finance * region	−0.0031	0.0042	0.0117 *	−0.0128 **	−0.0022	−0.0011
	(0.0043)	(0.0040)	(0.0066)	(0.0060)	(0.0054)	(0.0050)
Industry	Yes	Yes	Yes	Yes	Yes	Yes
CVs	Yes	Yes	Yes	Yes	Yes	Yes
Cons	−2.2583 **	2.4865 ***	−1.4441 ***	1.2452 ***	−1.6542 ***	1.1317 ***
	(0.9056)	(0.8410)	(0.3467)	(0.3075)	(0.3499)	(0.3173)
Obs	3263		3263		3263
Log-L	−2961.5926		−2960.9299		−2967.6302
Wald’s chi2	1807.38		3168.48		2743.83
ρ	−0.9881385 (0.003013)		−0.9902354 (0.0029417)		−0.9891235 (0.0029508)
p-Value	Prob > χ2 = 0.0000		Prob > χ2 = 0.0000		Prob > χ2 = 0.0000

Note: The symbols *, **, and *** denote statistical significance at the 10, 5, and 1% levels, respectively.

reports the results.

The interaction terms are statistically significant only for the central region: digital inclusive finance is more positively associated with formal employment and more negatively associated with informal employment in central China than elsewhere (columns 3–4). By contrast, the eastern- and western-region interactions are statistically insignificant (columns 1–2 and 5–6). These patterns suggest that the employment composition response is comparatively stronger in the central region. One interpretation consistent with the institutional setting is that central China combines relatively abundant labor and industrial relocation opportunities with a thinner traditional financial supply, so that digital inclusive finance may relax financing constraints more effectively and support formal hiring. In the eastern region, financial supply is already denser and flexible employment is more prevalent, which may offset positive and negative forces. In the western region, weaker industrial capacity and thinner digital human capital may limit how far financial resources translate into formal jobs. The western estimate is also less precise because the sample is smaller, so an insignificant coefficient should not be read as definitive evidence of no effect.

Overall, the formal interaction tests support regional heterogeneity concentrated in the central region, while evidence of statistically distinguishable differentials for the eastern and western regions remains limited.

5.5.3. Differences in the Type of Credit Constraints

The preceding results indicate that digital inclusive finance eases MSEs’ credit constraints. Which part of those constraints it eases, the supply side or the demand side, is a separate question, so we re-estimate the model after classifying the constraints by type.

Table 16. Digital inclusive finance and supply-oriented credit constraints of MSEs.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	Full Sample	Full Sample	Full Sample	Full Sample	The Eastern Region	The Central Region	The Western Region
Digital inclusive finance	−0.0142 ***				−0.0132 ***	−0.0156 **	−0.0340 ***
	(0.0016)				(0.0017)	(0.0069)	(0.0060)
Breadth of coverage		−0.0120 ***
		(0.0013)
Depth of use			−0.0075 ***
			(0.0010)
Degree of digitalization				0.0081 ***
				(0.0017)
Industry	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Region	Yes	Yes	Yes	Yes	Yes	Yes	Yes
CVs	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cons	0.8160 *	−0.0307	−1.0930 ***	−5.9404 ***	0.7969	0.5334	4.3626 ***
	(0.4729)	(0.3995)	(0.3486)	(0.7498)	(0.5493)	(1.7402)	(1.3950)
Obs	3263	3263	3263	3263	1869	637	757

Note: The symbols *, **, and *** denote statistical significance at the 10, 5, and 1% levels, respectively.

and

Table 17. Digital inclusive finance and demand-oriented credit constraints of MSEs.

Variables	(1)	(2)	(3)	(4)	(5)	(6)	(7)
	Full Sample	Full Sample	Full Sample	Full Sample	The Eastern Region	The Central Region	The Western Region
Digital inclusive finance	−0.0020				−0.0014	−0.0053	0.0048
	(0.0020)				(0.0022)	(0.0089)	(0.0068)
Breadth of coverage		−0.0018
		(0.0016)
Depth of use			−0.0008
			(0.0013)
Degree of digitalization				0.0010
				(0.0023)
Industry	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Region	Yes	Yes	Yes	Yes	Yes	Yes	Yes
CVs	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cons	−0.4797	−0.5606	−0.8024 **	−1.4405	−0.7602	−0.5081	−1.3452
	(0.5426)	(0.4294)	(0.3661)	(0.9226)	(0.6550)	(1.9396)	(1.5740)
Obs	3263	3263	3263	3263	1869	637	757

Note: The symbols ** denote statistical significance at the 5% levels, respectively.

show that the association is concentrated on supply-side constraints rather than demand-side ones. Therefore, the credit constraints alleviation effect of digital inclusive finance from the demand side has become an endogenous demand and a driving force for the next step in the development of digital inclusive finance. Specifically, digital inclusive finance reduces supply-side credit constraints for MSEs, and this holds in the eastern, central, and western regions alike. The breadth of coverage and depth of use of digital inclusive finance can effectively alleviate the credit constraints of MSEs. A deepening degree of digitalization works in the opposite direction and tightens supply-side constraints, plausibly because digitalization has to be matched by adequate digital infrastructure. Where it is not, a digital divide opens up, and digitalization aggravates rather than relieves MSEs’ credit constraints. This is consistent with the findings of the previous study, which puts higher demands on the development of digital inclusive finance at the regulatory level.

6. Conclusions

This paper matches the China Micro- and Small-Enterprise Survey (CMES) with the Peking University Digital Inclusive Finance Index. Taking the heterogeneity of hiring forms as its starting point, it examines how digital inclusive finance relates to MSEs’ labor employment decisions and probes the channel behind that relationship. The main findings of this paper are as follows:

First, the development of digital inclusive finance is positively associated with formal employee hiring and negatively associated with informal employee hiring in MSEs. The development of digital inclusive finance is also associated with MSEs’ labor employment decisions through its relationship with credit constraints. The mediating effects of credit constraints account for 9.17% and 10.58% of the respective relationships between digital inclusive finance development and formal and informal employee hiring.

Second, coverage breadth and usage depth are positively associated with formal employment and negatively associated with informal employment, whereas the degree of digitalization shows the opposite pattern. At the regional level, formal interaction tests indicate that the association is comparatively stronger in central China, whereas eastern- and western-region differentials are statistically limited; western estimates should also be interpreted with caution given the smaller sample.

Third, the development of digital inclusive finance is negatively associated with supply-oriented credit constraints among MSEs, while its relationship with demand-oriented credit constraints is not statistically significant. This pattern is also observed across the eastern, central, and western regions. The breadth of coverage and depth of use of digital inclusive finance are negatively associated with supply-oriented credit constraints among MSEs, whereas the degree of digitalization is positively associated with supply-oriented credit constraints.

7. Discussion

7.1. Theoretical Implications

This study contributes to the existing literature in several ways. First, it extends the literature on digital inclusive finance by examining its relationship with the labor employment decisions of MSEs. Previous studies have mainly focused on the role of digital inclusive finance in areas such as financing constraints, entrepreneurship, and firm performance [115,116,117]. Relatively limited attention has been paid to how the development of digital inclusive finance is associated with firms’ employment decisions. By focusing on MSEs, this study provides further evidence on the potential relationship between digital financial development and enterprise employment behavior.

Second, this study provides a more nuanced perspective on enterprise employment by distinguishing between formal and informal employees. Rather than treating employment as a single outcome, we examine the two types of employment separately. This distinction helps reveal that the relationship between digital inclusive finance and employment may differ across formal and informal employment [118,119,120]. Therefore, the findings provide a more detailed perspective for understanding the labor allocation decisions of MSEs.

Third, this study incorporates credit constraints into the analysis and further distinguishes between supply-side and demand-side credit constraints. This approach helps clarify the potential mechanism through which the digital financial environment may be associated with MSEs’ employment decisions. By considering different types of credit constraints, this study extends the existing literature beyond the general relationship between financing conditions and employment and provides a more detailed understanding of the role of credit constraints in MSEs’ labor employment decisions.

Overall, this study enriches the literature by connecting digital inclusive finance, credit constraints, and MSEs’ formal and informal employment decisions within a unified analytical framework.

7.2. Practical Implications

Realizing the full potential of digital inclusive finance for MSEs’ employment structure calls for joint effort. Government authorities, financial institutions, and MSEs themselves need to improve the environment for digital inclusive finance, raise the efficiency of financial services, and promote standardized employment practices. Specific recommendations for government agencies, financial institutions, and MSE managers are set out below: (1) Government agencies: Government agencies should improve the development system of digital inclusive finance and promote coordinated regional development. Existing research shows that digital inclusive finance can promote the recruitment of formal employees and reduce reliance on informal employees by easing the supply-side credit constraints of MSEs, but this effect varies across regions. Therefore, government authorities should continue to improve the construction of digital financial infrastructure, expand the coverage of digital inclusive finance to less developed regions, and narrow regional development gaps. Meanwhile, attention should be paid to the rising financing threshold that may arise from digital transformation. By improving the corporate credit system, strengthening data sharing, and optimizing policy support, authorities can reduce financing barriers for MSEs and guide enterprises to achieve standardized employment. (2) Financial institutions: Financial institutions should improve the capacity of digital financial services to alleviate financing constraints of Micro- and Small Enterprises. Digital inclusive finance primarily alleviates the supply-side credit constraints of MSEs, but it has limited impact on demand-side credit constraints. Therefore, financial institutions should further leverage technologies such as big data and artificial intelligence to improve credit evaluation systems and enhance financing availability for MSEs. Meanwhile, they should develop innovative financial products and service models based on the actual operational needs of enterprises, reduce financing costs, and better exploit the role of digital finance in promoting formal employment of enterprises. (3) Micro- and Small Enterprises: MSEs should improve the utilization level of digital inclusive finance and optimize the employment structure of enterprises. MSEs should proactively use digital credit, online financial services, and other tools to improve capital turnover capacity and alleviate financing pressure so as to provide support for formal employment. Meanwhile, enterprises should avoid relying solely on flexible employment to reduce costs in the process of digital transformation. Instead, they should integrate digitalization with operational management improvement, gradually improve labor relations management, and realize the coordinated unification of enterprise development and standardized employment.

7.3. Limitations and Future Research

This study has several limitations that also provide directions for future research. First, the empirical analysis is based on cross-sectional data from 2015. The cross-sectional nature of the data limits the causal interpretation of the findings. Although the empirical analysis examines the relationships between digital inclusive finance, credit constraints, and MSEs’ labor employment decisions, it cannot fully account for unobserved firm-level heterogeneity or capture changes in these relationships over time. Future research could use panel data to examine the dynamic relationships among these variables and apply more rigorous identification strategies to further explore causal relationships.

Second, the Digital Inclusive Finance Index used in this study is measured at the provincial level. Although the provincial index can reflect the overall digital financial development environment and infrastructure in a region, it may not fully capture differences in digital financial access and usage across individual MSEs. Future studies could incorporate firm-level measures of digital financial use, such as digital payment, online lending, and digital financial service usage, to provide a more precise assessment of the relationship between digital inclusive finance and MSEs’ employment decisions.

Third, this study focuses primarily on credit constraints as a potential mechanism linking digital inclusive finance to MSEs’ labor employment decisions. Other potential mechanisms may also exist, such as improvements in information accessibility, reductions in transaction costs, and changes in enterprise risk management. Future research could examine these mechanisms in greater depth to provide a more comprehensive understanding of how digital financial development is associated with employment decisions.

Finally, this study has not developed a formal theoretical model to fully represent the relationships examined. The current study mainly relies on theoretical analysis and empirical testing based on the available survey data. Future research could develop a more comprehensive theoretical framework and formal model. This would help clarify the underlying mechanisms and provide a stronger theoretical foundation for the empirical analysis. Such efforts could further improve our understanding of the relationships examined in this study.