Digital Inclusive Finance and Socially Sustainable Development: Empirical Evidence on Intergenerational Educational Mobility in China’s Ethnic Minority Regions

Abstract

Intergenerational educational mobility constitutes a cornerstone of sustainable social development—particularly in China’s ethnic minority regions, where socioeconomic development lags behind the national average. Digital inclusive finance—with its broad geographic reach and low entry barriers—offers a promising lever for breaking the cycle of low intergenerational mobility in these regions. This study systematically examines how the development of digital inclusive finance in China affects intergenerational educational mobility in ethnic minority regions and identifies the underlying mechanisms, thereby addressing SDG 4 (Quality Education) and SDG 10 (Reduced Inequalities). Using six waves of longitudinal data (2012–2022) from the China Family Panel Studies (CFPS), we examine the impact of digital inclusive finance on intergenerational educational mobility. To mitigate endogeneity, we construct an instrumental variable based on historical telecom infrastructure and employ two-stage least squares (2SLS) estimation. The findings indicate that digital inclusive finance significantly enhances intergenerational educational mobility in ethnic minority regions. Further mechanism analysis indicates that the effect operates primarily through increased household net income and, more tentatively, through improved access to education-related information—proxied here by traditional media usage. Accordingly, we recommend upgrading digital infrastructure in ethnic minority regions, improving residents’ digital literacy, and aligning digital inclusive finance with education subsidies and employment guidance. This helps households convert financial access into sustained investments in offspring’s education—boosting intergenerational educational mobility and advancing social sustainability.

1. Introduction

Intergenerational educational mobility is not merely a matter of equal opportunity—it serves as a key indicator of whether a society can achieve sustainable development [1]. Parental socioeconomic background increasingly influences offspring’s educational attainment: the intergenerational educational elasticity rose from 0.258 in 1989 to 0.31 in 2019 [2,3]. Since higher elasticity reflects lower mobility, this trend signals weakening intergenerational educational mobility [4]. This trend should be taken especially seriously in ethnic minority regions. Although average schooling in ethnic minority regions has grown faster than the national average, historical legacies [5], geographical isolation, and economic and cultural disadvantages [6] continue to limit upward educational mobility for offspring of low-education households. Meanwhile, high-education households use accumulated socioeconomic resources to sustain intergenerational advantage—creating a situation where immobility and blocked upward mobility coexist. Against this backdrop, examining whether digital inclusive finance can disrupt entrenched patterns and empower disadvantaged groups is theoretically important and policy-urgent for advancing social sustainability in ethnic minority regions.

Digital inclusive finance—characterized by broad geographic coverage, low marginal delivery costs, and high service penetration—presents a promising pathway to address the intergenerational immobility outlined above. Empirical evidence indicates that it contributes to alleviating credit constraints [7] and narrowing educational opportunity gaps [8]. Yet, existing research predominantly examines general financial accessibility, paying insufficient attention to intergenerational educational mobility in ethnic minority regions—and even less to its underlying mechanisms. Moreover, how digital inclusive finance affects educational mobility in ethnic minority regions remains underexamined.

In response to the aforementioned research gaps, this study systematically examines how China’s digital inclusive finance development affects intergenerational educational mobility in ethnic minority regions—and critically identifies the underlying causal mechanisms.

This study makes three major contributions. First, this is the first study to link digital inclusive finance with intergenerational educational mobility in China’s ethnic minority regions—and to show how household income and access to education-related information shape this relationship. Second, it identifies two mediating pathways—income growth and information access—through which digital inclusive finance fosters upward intergenerational educational mobility. Third, it reveals heterogeneous effects across four dimensions—labor migration, geographic location, urban–rural residence, and subjective self-perception—offering a policy-relevant basis for targeted interventions.

Using six waves of CFPS data (2012–2022) and a two-way fixed effects model, we show that digital inclusive finance significantly improves intergenerational educational mobility in ethnic minority regions—especially among historically disadvantaged subgroups. These results provide actionable evidence for digital finance policies aimed at educational equity, inclusive and sustainable social development.

2. Literature Review

2.1. Conceptualization and Measurement of Intergenerational Educational Mobility

Intergenerational educational mobility measures how much offspring’s education depends on individual effort and ability—rather than parental background [9,10]. Becker and Tomes [11] model this as intergenerational human capital transmission under credit constraints: imperfect capital markets link parental education to investment in children’s schooling, perpetuating inequality across generations.

Two dominant methodological paradigms exist for quantifying mobility [12]. First, regression-based approaches—using coefficients on parental education (e.g., in Mincer-type earnings or schooling equations) or elasticity estimates—capture the average marginal effect of parental background on offspring schooling [13]. While parsimonious and widely interpretable, these methods implicitly assume homogeneity in mobility barriers across educational thresholds (e.g., primary-to-secondary vs. secondary-to-tertiary transitions), obscuring critical heterogeneity in opportunity structures. Second, transition matrix-based methods—including educational transition matrices and sibling fixed-effects models—explicitly model mobility as discrete, stage-specific processes [14], while controlling for unobserved family-level confounders through within-family comparisons [15]. These approaches yield richer insights into the structural bottlenecks constraining upward mobility and improve identification of transmission mechanisms. Yet methodological pluralism introduces interpretive challenges: divergent estimates across frameworks often reflect distinct underlying assumptions—not merely statistical noise—leading to ambiguous policy guidance. Crucially, most existing measures were developed and validated for majority-population contexts; their psychometric properties—including measurement invariance across linguistic, cultural, and institutional settings—remain untested in China’s ethnically diverse regions. Direct application risks systematic bias due to differential item functioning. Consequently, measurement non-equivalence poses a first-order threat to causal inference in studies of ethnic minority mobility—not merely a technical concern, but a substantive validity issue. In the subsequent analyses, we seek to reduce measurement bias in part by disaggregating the sample and performing robustness tests, with the aim of strengthening inferential validity.

2.2. Mechanisms Shaping Intergenerational Educational Mobility

Existing scholarship identifies two countervailing forces shaping intergenerational educational mobility: (1) endogenous, family-based mechanisms that reinforce intergenerational continuity—often termed “solidification”—and (2) exogenous, context-driven mechanisms rooted in institutional and structural conditions that foster mobility. Yet these strands of research have largely evolved in parallel, with limited theoretical integration or empirical investigation into their dynamic interplay. At the micro level, family capital—comprising economic, cultural, and social dimensions— constitutes a central explanatory mechanism [16]. Complementing this account, behavioral genetics research shows that inherited traits—such as cognitive aptitude and socioemotional competencies—contribute to intergenerational educational transmission [17]. Nevertheless, dominant frameworks often conceptualize parental influence as unidirectional and deterministic, undertheorizing offspring’s agentic responses and the contingent, moderating effects of contextual factors (e.g., school quality, peer composition, neighborhood institutions). At the macro level, institutional reforms—particularly those designed to be universal and inclusive—and financial system development operate as critical levers of mobility. For instance, equitable policy interventions (e.g., expanded public schooling, need-based financial aid) generate both “environmental effects” (by leveling learning opportunities across socioeconomic strata) and “wealth effects” (by altering household resource allocation priorities), thereby attenuating the predictive power of parental education [18]. Similarly, deeper financial markets alleviate credit constraints for disadvantaged households, enhancing their capacity to invest in offspring’s human capital—even when immediate income is low—thus strengthening mobility pathways [19,20]. This reveals a tension: micro-level analyses foreground mechanisms of advantage reproduction, while macro-level analyses emphasize structural openings for upward mobility. Crucially, however, the conditions under which these forces synergize, counteract, or mutually condition one another—and whether their net effect can ever fully neutralize entrenched inequalities—remain empirically unresolved and theoretically underdeveloped.

Focusing on ethnic minority regions, the interplay among family capital, institutional design, and cultural identity exerts a heightened and empirically distinctive influence on intergenerational educational mobility. Xu Shanshan et al. [5] demonstrate that offspring from impoverished households in these areas face severely constrained mobility prospects—not primarily due to absolute resource scarcity, but because low parental educational expectations and limited access to pedagogically relevant learning materials systematically restrict educational aspiration formation and academic engagement. Complementing this, Chen Yiting [21] finds that parental involvement in school-based communication (e.g., teacher conferences, curriculum consultations) yields stronger positive effects on student outcomes than direct monetary transfers alone, underscoring the irreplaceable mediating roles of cultural capital (e.g., navigating bureaucratic procedures) and social capital (e.g., building trust-based relationships with educators). Further, Min Ruihua [22] and Liu Chundi [6] identify a culturally specific pathway: erosion of indigenous language proficiency and weakening cultural belongingness correlate with diminished parental valuation of formal schooling, reduced household education expenditures, and elevated dropout rates—constituting a “cultural exclusion → devaluation of schooling → educational disengagement” mechanism distinct from mainstream contexts. Collectively, these studies reveal a critical policy insight: macro-level economic empowerment—such as poverty-alleviation transfers or financial inclusion initiatives—does not automatically catalyze micro-level educational mobility in ethnic minority settings; its efficacy is powerfully conditioned by culturally embedded mediating variables, including linguistic capital [6], community-level social networks [23], and parental educational agency shaped by ethnocultural positioning [24]. In ethnic minority regions marked by significant cultural distance, does external financial empowerment—despite its intent—fail to translate into sustained household investment in education due to diminished parental willingness? Put differently, it remains empirically unclear whether financial support in these contexts reliably fosters educational aspirations and investment behavior.

2.3. Digital Inclusive Finance and Intergenerational Educational Mobility in Ethnic Minority Regions

The externally driven financial empowerment discussed earlier has evolved—under the impetus of digital technologies—into a broad-based, scalable institutional form: digital inclusive finance [25]. Enabled by internet platforms and data analytics, digital inclusive finance reduces transaction costs and broadens financial access for underserved populations [26,27]. Its broader socioeconomic implications have attracted growing scholarly attention. Empirical evidence demonstrates that it contributes meaningfully to more efficient labor resource allocation [28], advances the equalization of basic public services—including education and healthcare [29], and stimulates household educational investment [30]. Collectively, these findings lay a foundational empirical basis for assessing the non-pecuniary, structural contributions of digital inclusive finance to social development.

Yet, extant research on the nexus between digital inclusive finance and intergenerational mobility remains narrowly anchored in income outcomes [31]. Notably, studies by Zhou et al. [32], Zhang et al. [33], and Wei et al. [34] converge in showing that digital inclusive finance significantly enhances intergenerational income mobility—particularly among low-income households and rural residents. While valuable, such findings primarily pertain to outcome fairness [35]—post-distribution improvements, not equitable access to opportunity. Intergenerational educational mobility operates upstream, shaping individuals’ capacity to acquire human capital and enter skilled occupations. When educational opportunities remain tied to parental socioeconomic status, income-mobility efforts risk reinforcing structural inequities. Thus, evaluating digital inclusive finance solely through the lens of income transmission overlooks its potential role in reshaping opportunity structures at their origin. This study accordingly shifts analytical focus to intergenerational educational mobility—a more fundamental indicator of fair social openness—and thereby extends and deepens the prevailing assessment of digital inclusive finance’s societal impact.

A nascent strand of literature begins to address this dimension. Qi Yudong et al. [19] employ intergenerational elasticity estimation to demonstrate that digital inclusive finance mitigates household credit constraints, enabling disadvantaged families to raise educational expenditures and attenuate the transmission of parental economic disadvantage to offspring’s schooling attainment. Complementing this, Ma Yechi [36] and Yan Siyu et al. [30] refine the causal pathway by emphasizing enhanced educational affordability (via improved liquidity) and reduced precautionary motives (via greater income stability), respectively. Together, these works articulate a coherent theoretical mechanism: digital inclusive finance → alleviation of financing constraints/augmentation of disposable resources/mitigation of uncertainty → increased educational investment → higher intergenerational educational mobility. Nevertheless, this framework remains largely agnostic to contextual heterogeneity—especially the distinctive socioeconomic, institutional, and cultural configurations of ethnic minority regions. Critically, it does not engage with the proposition advanced earlier—that financial empowerment may be culturally moderated—a gap this study explicitly addresses.

We focus on ethnic minority regions for three interrelated reasons: pronounced financial exclusion under conventional banking systems; persistent information asymmetries stemming from geographic isolation; and in certain communities, diminished perceived returns to formal education [37].

Two critical lacunae persist in current scholarship. First, analyses of intergenerational educational mobility in ethnic minority regions continue to privilege conventional determinants—such as parental education, local school quality, or fiscal transfers—while overlooking digital inclusive finance as a novel, system-level enabler of opportunity expansion. Second, existing empirical work tends toward national- or provincial-level aggregation, or treats ethnic regions as homogeneous extensions of general models. Yet digital inclusive finance operates distinctively in these contexts: it transcends geographical isolation, diversifies information sources beyond kinship or local networks, and bolsters household resilience against idiosyncratic shocks—factors that collectively amplify its potential to disrupt entrenched educational immobility. To date, however, no study has rigorously examined how—and under what institutional, cultural, and infrastructural conditions—digital inclusive finance reconfigures educational opportunity structures in ethnic minority regions. This study fills that void by conducting context-sensitive empirical analysis to determine whether, and through which precise channels, digital inclusive finance serves as a catalyst for breaking intergenerational educational rigidity in these settings—thereby contributing novel theoretical insights and policy-relevant evidence on digitally mediated equity advancement.

3. Theoretical Framework and Hypothesis Development

Intergenerational educational mobility advances two key SDGs: equitable education (SDG 4) and reduced inequality (SDG 10). When offspring’s education depends more on parental background than on their own abilities, schooling reinforces—rather than reduces—structural inequality.

Intergenerational educational elasticity—defined as the percentage change in offspring’s educational attainment (e.g., years of schooling or highest degree attained) associated with a 1% change in parental educational attainment—is a canonical metric for quantifying the strength of intergenerational educational persistence. A higher elasticity coefficient reflects greater transmission of educational advantage (or disadvantage) across generations, indicating lower intergenerational educational mobility; conversely, a lower elasticity signifies weaker ascription-based influence and stronger meritocratic openness.

When educational outcomes strongly reflect parental socioeconomic status, education reinforces stratification—not mobility—undermining SDG 4 and SDG 10 [38]. This study therefore frames its core inquiry as follows: Can digital inclusive finance disrupt the intergenerational transmission of educational disadvantage in ethnic minority regions—and thereby foster sustainable, self-reinforcing intergenerational educational mobility?

We posit that digital inclusive finance promotes intergenerational educational mobility in ethnic minority regions through three mutually reinforcing channels: (i) alleviating household financial constraints, (ii) expanding information accessibility, and (iii) diversifying household income sources. Collectively, these pathways enhance parental agency—in terms of willingness (motivation), capacity (resources), and informed judgment (awareness)—to invest in offspring’s education, thereby weakening the statistical and causal link between parental and offspring educational outcomes.

First, financial constraint alleviation. In ethnic minority regions, many families face real barriers to accessing traditional credit—often due to distance from financial institutions and limited assets to use as collateral [39,40]. Digital inclusive finance provides a targeted, evidence-informed response: mobile-based lending and algorithmic credit scoring expand financial access [28,41,42,43], enabling families to invest in children’s education—including tuition, tutoring, and university enrollment—and thereby advance intergenerational mobility [19,44,45].

This channel shows systematic variation across regions. In provinces with high labor outflow, prolonged physical separation weakens intergenerational educational support—financially, cognitively, and emotionally. Although remittances raise household income (consistent with remote transmission of family resources theory), they cannot replace parental presence in academic guidance, cultural capital transfer, or daily learning supervision [46]. Digital inclusive finance helps bridge this gap: mobile payments enable fast, low-cost, education-targeted transfers; digital insurance reduces dropout risks from health or seasonal shocks; and integrated financial-education platforms deliver remote parenting support. Thus, its mobility-enhancing effect is stronger in high labor-outflow provinces. Similarly, rural households face multiple barriers—few bank branches, rigid credit criteria, and low financial literacy—making them especially prone to financial exclusion [47,48]. By using widespread mobile infrastructure and behavioral data analytics, digital inclusive finance overcomes institutional and geographic barriers [49,50], offering tailored credit, savings, and insurance products that directly fund education—aligned with credit rationing theory. Consequently, its impact on educational mobility is greater in rural than in urban areas.

Second, enhanced information accessibility. Digital inclusive finance platforms deliver timely, localized educational information—scholarship opportunities, admission criteria, and peer attainment data—to households hindered by distance and language barriers [22,51]. By lowering the cost of accessing this information, they reduce asymmetries that distort parental expectations and limit educational decisions [52,53], enabling more informed school choices and sustained engagement with formal education.

Heterogeneity in this channel stems from regional and psychological factors. Southern ethnic minority regions—which constitute the vast majority of China’s ethnic minority population and host diverse local economies (e.g., tourism, e-commerce, agro-processing)—appear especially conducive to the educational empowerment potential of digital financial tools [54]. Given the limited statistical power arising from a small northern subsample (N = 144), we cannot reliably assess whether similar patterns hold there. Accordingly, our heterogeneity analysis focuses on the southern subsample; findings for northern regions are reported descriptively only. Furthermore, households with low self-perception—characterized by diminished belief in their capacity to influence educational outcomes—are systematically disadvantaged in traditional information environments. Digital inclusive finance counteracts this via “frictionless” design: algorithmic nudges, simplified interfaces, and positive reinforcement loops lower activation thresholds and automate opportunity identification. This “compensatory empowerment” effect implies disproportionately large mobility gains for low self-perception households—a proposition grounded in Bandura’s self-perception theory [20].

Third, income diversification. Digital inclusive finance supports non-agricultural entrepreneurship via microloans and digital marketplaces, helping households diversify income beyond subsistence farming [33]. This income diversification reduces precautionary savings and frees up resources for education. It also sharpens parents’ awareness of labor market skills and returns—leading to more strategic educational investments [55,56].

Based on the foregoing theoretical analysis, we formulate the following hypotheses:

H1. 

Digital inclusive finance significantly enhances sustainable intergenerational educational mobility in ethnic minority regions.

H2. 

Digital inclusive finance enhances sustainable intergenerational educational mobility in ethnic minority regions through two empirically grounded pathways: the expansion of household net income and the improvement of information accessibility regarding educational opportunities and returns.

H3. 

The positive effect of digital inclusive finance on intergenerational educational mobility in ethnic minority regions is significantly stronger among four contextually distinct subpopulations: provinces characterized by high labor outflow; rural households; southern ethnic minority regions; households exhibiting low self-perception.

4. Empirical Strategy and Identification Approach

4.1. Data Sources and Sample Construction

Intergenerational educational mobility data are drawn from six waves (2012–2022) of the China Family Panel Studies (CFPS), a nationally representative biennial survey by Peking University. CFPS uses multistage probability sampling stratified by region, urban–rural status, and household size, enabling precise measurement of parental and offspring education and within-household analysis of intergenerational transmission over time.

Data on digital inclusive finance development are drawn from the DIFI, a rigorously constructed, province-level panel dataset co-developed by the Institute of Digital Finance at Peking University and Ant Group [26,57,58]. This paper matches the two databases—CFPS and DIFI—at the provincial level [8].

Furthermore, auxiliary provincial-level covariates—including real GDP growth rate, GDP per capita (in constant 2020 CNY), the share of fiscal expenditure allocated to education, and the count of regular higher education institutions—are sourced from authoritative and harmonized macroeconomic and educational databases: the National Bureau of Statistics of China (NBS), the China City Statistical Yearbook (annual editions), and the China Stock Market & Accounting Research (CSMAR) database.

To improve the accuracy and robustness of the empirical estimates, this study implements a rigorous sample selection and cleaning procedure. Specifically: (1) Intergenerational dyads are constructed by matching parental and offspring respondents using the official parent–child relationship codes from the CFPS; (2) To mitigate life-cycle-related bias, we exclude observations in which the offspring is under 16 years of age, the parent is over 65 years of age, or the intergenerational age gap is less than 16 years; (3) Observations with missing or implausible values in key variables—including years of schooling, age, and annual household income—are dropped; (4)To ensure data representativeness and comparability, this study confines the definition of ethnic minority regions to the Inner Mongolia Autonomous Region, Xinjiang Uygur Autonomous Region, Ningxia Hui Autonomous Region, Guangxi Zhuang Autonomous Region, Tibet Autonomous Region, Yunnan Province, Guizhou Province, and Qinghai Province. Relevant regional effects are controlled for in the models.

Following the above sample selection and cleaning procedures, the final analytical sample comprises 2547 intergenerational parent—child dyads—specifically, father–son and father–daughter pairs. This yields a robust micro-level dataset well suited for empirically testing the study’s central hypothesis: that the development of digital inclusive finance facilitates intergenerational educational mobility.

4.2. Model Specification and Variable Definition

This study adopts years of schooling as the primary proxy for human capital stock and estimates the effect of digital inclusive finance on intergenerational educational mobility in ethnic minority regions using a two-way fixed effects model. The baseline empirical specification is presented below:

$$c_{\mathit{Edu}}^{\mathit{ijt}}={\alpha }_0+{\alpha }_1{f}_{\mathit{Edu}}^{\mathit{ijt}}+{\alpha }_2{\mathit{DIFI}}_{j,t-1}+{\alpha }_3{f}_{\mathit{Edu}}^{\mathit{ijt}}\times {\mathit{DIFI}}_{j,t-1}+\beta X_{\mathit{ijt}}+{\lambda }_t+{\mu }_j+c_i+{\epsilon }_{\mathit{ijt}}$$

(1)

Here, $c_{\mathit{Edu}}^{\mathit{ijt}}$ denotes the natural logarithm of the offspring i’s years of schooling in province j in year t, and $f_{\mathit{Edu}}^{\mathit{ijt}}$ denotes the analogous measure for the parent(s). The effect of parental years of schooling on offspring’s years of schooling captures the degree of intergenerational transmission of education. Consistent with prior literature emphasizing the father’s predominant role in household educational decision-making—including resource allocation and investment—this study adopts paternal years of schooling as the proxy for parental human capital. To investigate how digital inclusive finance moderates the intergenerational elasticity of education, we further incorporate two key variables: (i) the provincial-level DIFI, and (ii) the estimated intergenerational elasticity of education. Specifically, our core explanatory variable is the natural logarithm of the provincial digital inclusive finance total index co-developed by the Institute of Digital Finance at Peking University and Ant Group. This index comprises three secondary dimensions: coverage breadth, usage depth, and degree of digitalization. In both the benchmark regression and mechanism analysis, we employ the provincial-level total index as the primary measure of digital inclusive finance to capture its aggregate impact. The variable ${\mathit{DIFI}}_{j,t-1}$ denotes the DIFI for province j in period t − 1; a one-period lag is applied to mitigate potential reverse causality arising from simultaneity between educational outcomes and financial development. The intergenerational educational elasticity is defined as the percentage change in the offspring’s years of schooling associated with a 1% increase in the parents’ years of schooling. In the empirical specification, we estimate a double-logarithmic regression model. A larger estimated elasticity coefficient indicates a stronger transmission of educational attainment from parents to offspring—and thus lower intergenerational educational mobility. The key interaction term in our empirical specification is $f_{\mathit{Edu}}^{\mathit{ijt}}\times {\mathit{DIFI}}_{j,t-1}$, and its estimated coefficient captures the moderating effect of digital inclusive finance on the intergenerational elasticity of education. A statistically significant negative estimate of ${\alpha }_3$ implies that higher levels of digital inclusive finance development attenuate the intergenerational transmission of educational attainment—from parents to offspring—thereby reducing the intergenerational elasticity of education and enhancing educational mobility across generations. $X_{\mathit{ijt}}$ denotes a vector of individual-, household-, and province-level covariates associated with intergenerational educational mobility. ${\lambda }_t$, ${\mu }_j$, and $c_i$ denote year, provincial, and offspring birth cohort fixed effects, respectively; ${\epsilon }_{\mathit{ijt}}$ is the idiosyncratic error term. All standard errors are clustered at the provincial level to account for within-province correlation in the disturbances.

The core dependent variable in this article is the logarithm of the years of education of the offspring, which is used to measure the outcome variable of intergenerational educational mobility. The main explanatory variables include the logarithm of the years of education of the parents, the logarithm of the provincial digital inclusive finance development index lagged by one period, and their interaction term, to explore the impact of the development of digital inclusive finance on the intergenerational transmission effect of education in ethnic minority areas.

We include a comprehensive set of control variables to account for multi-level determinants of educational outcomes—spanning individual, household, and provincial dimensions. At the individual level, we control for the offspring’s and parents’ ages and their quadratic terms to control for potential nonlinear life-cycle effects. We further include binary indicators for offspring gender (1 = male), hukou status (1 = urban), Communist Party membership (1 = member), marital status (1 = married), and employment sector (1 = non-agricultural), reflecting key sociodemographic and occupational attributes associated with educational attainment. At the household level, we control for total financial assets (in yuan), education- and training-related expenditures (in yuan), number of offspring, and total household size—measures capturing family resource endowment, educational investment capacity, and dependency burden. At the provincial level, we incorporate macroeconomic and education-specific controls: real GDP growth rate, per capita GDP (in yuan), share of fiscal expenditure allocated to education, and count of regular institutions of higher education—thereby accounting for regional heterogeneity in economic development and public education provision. All variable definitions and summary statistics are reported in

Table 1. Descriptive Statistics of Main Variables.

	Variables	Sample Size	Mean	Standard Deviation	Minimum	Maximum
Offspring-Level Variables	Offspring’s Years of Schooling (years)	2547	10.2858	3.4546	0	20
	Offspring’s Gender (1 = Male)	2547	0.6435	0.4791	0	1
	Offspring’s Age (years)	2547	24.4609	6.3512	16	44
	Offspring’s Hukou Status (1 = Urban)	2547	0.4095	0.4918	0	1
	Offspring’s Communist Party Membership (1 = Member)	2547	0.0236	0.1517	0	1
	Offspring’s Marital Status (1 = Married)	2547	0.3408	0.4741	0	1
	Offspring’s Employment Sector (1 = Non-agricultural)	2547	0.0805	0.2721	0	1
Parent-Level Variables	Parent’s Years of Schooling (years)	2547	6.6761	3.937	0	16
	Parent’s Age (years)	2547	51.3396	6.9635	34	65
	Parent’s Hukou Status (1 = Urban)	2547	0.3353	0.4722	0	1
	Parent’s Communist Party Membership (1 = Member)	2547	0.0204	0.1414	0	1
	Parent’s Marital Status (1 = Married)	2547	0.9482	0.2217	0	1
	Parent’s Employment Sector (1 = Non-agricultural)	2547	0.0597	0.2369	0	1
Household-Level Variables	Log of Family Financial Assets (CNY)	2547	7.6329	4.1225	0	14.5574
	Log of Family Education and Training Expenditure (CNY)	2547	4.4306	3.709	0	11.1273
	Household Size (persons)	2547	4.8347	1.9456	1	13
	Number of offspring in Household	2547	0.5473	0.7678	0	4
Provincial-Level Variables	Log of Provincial DIFI	2547	4.9492	0.9826	2.9161	5.8724
	Provincial Real GDP Growth Rate	2547	0.1063	0.0469	0.0026	0.2006
	Log of Provincial Per Capita GDP (CNY)	2547	10.4719	0.3459	9.8494	11.4869
	Provincial Share of Education Expenditure in Total Fiscal Budget	2547	0.1793	0.0162	0.1068	0.1979
	Number of Regular Institutions of Higher Education	2547	68.6144	12.4397	7	85

.

5. Empirical Results

5.1. Benchmark Regression Results

Table 2. Benchmark Regression Results.

Dependent Variable: Years of Education of Offspring	(1)	(2)	(3)	(4)	(5)
Years of Education of Parents	0.1951 ***	0.1742 ***	0.1834 ***	0.2066 ***	0.2068 ***
	(4.0611)	(4.6500)	(4.4809)	(4.7696)	(4.9367)
DIFI	0.1192 ***	0.1099 ***	0.0756 **	0.3118 **	0.3114 **
	(5.1913)	(9.6942)	(3.2348)	(2.4963)	(2.5610)
Years of Education of Parents × DIFI	−0.0155	−0.0152 **	−0.0185 **	−0.0234 **	−0.0235 **
	(−1.5167)	(−2.0256)	(−2.5873)	(−3.1811)	(−3.2996)
Offspring Control Variables	No	Yes	Yes	Yes	Yes
Parent Control Variables	No	Yes	Yes	Yes	Yes
Family Control Variables	No	No	Yes	Yes	Yes
Regional Control Variables	No	No	Yes	Yes	Yes
Year Fixed Effects	No	No	No	Yes	Yes
Regional Fixed Effects	No	No	No	Yes	Yes
Cohort Fixed Effects	No	No	No	No	Yes
Observations	2547	2547	2547	2547	2547
adj. R2	0.105	0.178	0.191	0.197	0.196

Note: Standard errors clustered at the provincial level (cluster-robust) are reported in parentheses. ***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively. The same applies below.

reports the estimated effect of digital inclusive finance development on intergenerational educational mobility in ethnic minority regions. Column (1) presents the benchmark regression with no controls—only the parental education term, the lagged provincial DIFI, and their interaction.

The results indicate that, across all specifications, the parental education × DIFI interaction is negative and significant, indicating that digital inclusive finance weakens the intergenerational transmission of educational advantage—consistent with Hypothesis H1. A 10% increase in DIFI lowers the intergenerational educational elasticity by 2.6 percentage points—implying that higher digital inclusive finance penetration weakens the transmission of educational advantage from parents to offspring. This finding underscores the equity-enhancing function of digital inclusive finance; it expands educational opportunity for disadvantaged youth in ethnic minority areas and fosters more inclusive human capital accumulation.

5.2. Robustness Checks

To rigorously assess the reliability of the benchmark estimates and the robustness of our core inference—that digital inclusive finance enhances intergenerational educational mobility in ethnic minority regions—this study implements three complementary robustness checks: (1) refinement of birth cohort classification, (2) subsample restriction based on educational completion status, and (3) winsorization of extreme values.

Table 3. Robustness Checks.

Dependent Variable: Years of Education of Offspring	(1)	(2)	(3)
	Re-Estimated Cohort FE	Offspring Aged ≥ 22	1% Winsorization
Years of Education of Parents	0.2064 ***	0.2393 ***	0.2068 ***
	(4.8452)	(6.5638)	(4.9367)
DIFI	0.3164 **	0.5449 *	0.3114 **
	(2.7707)	(2.0764)	(2.5610)
Years of Education of Parents × DIFI	−0.0232 **	−0.0277 **	−0.0235 **
	(−3.1722)	(−2.9594)	(−3.2996)
Offspring Control Variables	Yes	Yes	Yes
Parent Control Variables	Yes	Yes	Yes
Family Control Variables	Yes	Yes	Yes
Regional Control Variables	Yes	Yes	Yes
Year Fixed Effects	Yes	Yes	Yes
Regional Fixed Effects	Yes	Yes	Yes
Cohort Fixed Effects	Yes	Yes	Yes
Observations	2547	1537	2547
adj. R2	0.197	0.200	0.196

reports the robustness check results.

First, to address potential sensitivity to cohort aggregation, we refine the birth cohort fixed effects by shifting from a conventional 10-year grouping (e.g., 1970–1979, 1980–1989) to a finer 5-year grouping (e.g., 1970–1974, 1975–1979). This adjustment follows Qi Yudong et al. (2025) [19] and responds to the accelerated pace of educational reform and labor market transformation in China: a 10-year window may obscure meaningful heterogeneity in policy exposure and socioeconomic context across adjacent cohorts. A 5-year classification better captures cohort-specific variation in digital financial access, schooling opportunities, and institutional environments—thereby mitigating confounding arising from macro-level structural shifts. Across all specifications, the interaction term between the DIFI and parental education remains statistically significant and negative, reinforcing the conclusion that digital inclusive finance strengthens upward educational mobility.

Second, to mitigate measurement error stemming from incomplete educational attainment, we restrict the offspring sample to individuals aged 22 years or older—effectively excluding those still enrolled in tertiary education or vocational training. This ensures that the dependent variable (offspring’s completed years of schooling) reflects finalized educational outcomes rather than interim status. The re-estimated coefficients retain both sign and statistical significance at conventional levels, confirming that our findings are not driven by age-related truncation bias.

Third, to assess sensitivity to outliers, we apply 1% two-tailed winsorization to the dependent variable and all key covariates. Re-estimation under this procedure yields coefficient estimates and standard errors virtually identical to those in the baseline model—both in magnitude and significance (all interaction terms remain significant at p < 0.05). This confirms that the estimated effect is not attributable to influential observations but reflects a consistent, population-wide pattern. Collectively, these checks substantiate the internal validity and generalizability of our central claim: digital inclusive finance serves as a robust catalyst for intergenerational educational advancement in ethnic minority areas.

Prior to examining the underlying mechanisms, we implement an instrumental variable (IV) approach to mitigate potential endogeneity—arising from omitted variables, reverse causality, or measurement error—in the relationship between digital inclusive finance and intergenerational educational mobility.

5.3. Mitigating Endogeneity

While the benchmark estimates withstand multiple robustness checks, a potential endogeneity concern remains between digital inclusive finance development and intergenerational educational mobility—arising from reverse causality or omitted variables. This creates a feedback loop wherein observed correlations may reflect mutual reinforcement rather than a causal effect of digital inclusive finance on mobility.

To mitigate this identification challenge, we adopt a Bartik-style instrumental variable (IV) strategy following Qi Yudong et al. [19]. Our instrument is the interaction between provincial historical telecommunications infrastructure—measured by the number of fixed-line telephones per 100 persons in 1984 (normalized)—and the national (excluding the province) mean of the DIFI. This instrument satisfies the relevance condition: 1984 fixed-line penetration predicts current digital financial development (confirmed in first-stage diagnostics). For the exclusion restriction, we assume 1984 telephone density affects current educational mobility only through digital inclusive finance. While plausible—fixed-line infrastructure from four decades ago is unlikely to directly influence today’s educational outcomes—this assumption cannot be empirically verified. Historical telecommunications development may correlate with persistent institutional quality or long-term educational investment patterns that independently shape mobility. Results should therefore be interpreted as consistent with a causal interpretation, not as definitive evidence of causality. For the endogenous interaction term (Parental Education × DIFI), we construct its instrument by multiplying parental education years with the baseline instrumental variable, thereby preserving orthogonality to the error term while maintaining relevance.

Table 4. Two-Stage Least Squares (2SLS) Estimation Results.

	(1)	(2)	(3)
Dependent Variable	First Stage	First Stage	Second Stage
	DIFI	DIFI × Years of Education of Parents	Offspring’s Years of Schooling
Instrumental Variable (IV)	−0.0003 ***	−0.0451 ***
	(−3.79)	(−3.61)
Years of Education of Parents	−0.0014 **	4.9727 ***	0.1164 ***
	(−2.07)	(7.5716)	(5.9382)
DIFI			0.0354 ***
			(3.4367)
Years of Education of Parents × IV	0.0005	0.9771 ***
	(0.1371)	(7.5180)
DIFI × Years of Education of Parents			−0.0139 ***
			(−4.2147)
Offspring Control Variables	Yes	Yes	Yes
Parent Control Variables	Yes	Yes	Yes
Family Control Variables	Yes	Yes	Yes
Regional Control Variables	Yes	Yes	Yes
Year Fixed Effects	Yes	Yes	Yes
Regional Fixed Effects	Yes	Yes	Yes
Cohort Fixed Effects	Yes	Yes	Yes
Observations	2547	2547	2547
adj. R2	0.840	0.197	0.197
First-Stage F-Statistic	575.1	1950.7

reports the two-stage least squares (2SLS) estimation results.

The first-stage regression results confirm the strength and relevance of our instrumental variable. The coefficient on the IV is −0.0003 (p < 0.01), and the F-statistic equals 575.1—well above the Stock–Yogo critical value, decisively rejecting the weak-instrument hypothesis. For the endogenous interaction term (Parental Education × DIFI), the corresponding instrument—constructed as the product of parental education years and the base IV—yields an even stronger first-stage F-statistic of 1950.7, further validating instrument strength.

The negative sign of the IV coefficient warrants brief interpretation: while historical fixed-line infrastructure is generally associated with long-term economic development, its marginal predictive power for contemporary digital inclusive finance may be attenuated after controlling for current regional economic conditions. In contexts where mobile-internet-based financial services have largely displaced legacy telecommunication infrastructure, the 1984 telephone density may capture residual or diminishing returns—yielding a statistically robust but economically nuanced negative association. Crucially, the sign carries no implication for instrument validity; identification relies solely on statistical relevance and exogeneity, both of which are satisfied.

In the second stage, the 2SLS estimate of the interaction term is −0.0139, remaining precisely estimated and highly significant. Relative to the baseline OLS coefficient of −0.0235, this reduction in magnitude reflects attenuation of upward bias induced by reverse causality: regions with higher intergenerational educational mobility may attract greater digital financial investment, leading OLS to overstate the causal effect of digital inclusive finance on mobility. The directional consistency—i.e., the IV estimate being less negative than the OLS estimate—provides corroborating evidence that the benchmark regression suffers from endogeneity-driven overestimation. This pattern suggests the IV estimates partially correct upward bias, though the exclusion restriction remains unverifiable. Thus, the 2SLS results support—rather than establish—a causal relationship.

5.4. Heterogeneity Analysis

To test Hypothesis H3, this study examines the heterogeneous effects of digital inclusive finance on intergenerational educational mobility across four dimensions: labor outflow intensity, urban–rural residence, geographical region (north vs. south), and individual self-perception. The results are presented in

Table 5. Heterogeneity Analysis.

Dependent Variable: Offspring’s Years of Education	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)
	High Labor-Outflow Provinces	Low Labor-Outflow Provinces	Rural	Urban	North	South	Low Self-Perception	High Self-Perception
Parent’s Years of Education	0.1806 ***	0.2376 **	0.2168 ***	0.0816 *	6.6500	0.1975 ***	0.2472 ***	0.1397
	(8.3716)	(6.6533)	(4.8170)	(1.8974)	(1.9427)	(5.8558)	(10.0674)	(0.6868)
DIFI	−4.4563 **	−0.0378	0.6689 ***	−0.2196	−0.8762	0.4633 ***	1.1048	−0.3411
	(−5.5161)	(−0.2061)	(4.2399)	(−0.9783)	(−0.4398)	(12.3408)	(1.3591)	(−1.2168)
DIFI × Years of Education of Parents	−0.0280 **	−0.0220	−0.0243 **	0.0017	−1.1615	−0.0217 **	−0.0443 ***	−0.0098
	(−4.2141)	(−1.8005)	(−2.3867)	(0.1062)	(−1.9278)	(−4.5604)	(−11.2793)	(−0.2538)
Offspring Control Variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Parent Control Variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Family Control Variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Regional Control Variables	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Year Fixed Effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Regional Fixed Effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Cohort Fixed Effects	Yes	Yes	Yes	Yes	Yes	Yes	Yes	Yes
Observations	887	1646	1693	854	144	2403	799	450
adj. R2	0.225	0.206	0.175	0.247	0.484	0.183	0.346	0.220

Note: The northern ethnic minority subsample has only 144 observations. Results are reported descriptively for completeness; cross-regional comparisons with the southern subsample are not statistically reliable.

.

First, labor outflow. The interaction term is significant only in high-outflow provinces (coefficient = −0.0280, p < 0.05), not in low-outflow ones. This indicates digital inclusive finance is most effective where labor migration disrupts traditional educational support—by enabling remote remittances and education budgeting.

Second, urban–rural residence. The interaction term is significant in rural but not urban subsamples, confirming that digital inclusive finance delivers higher marginal returns where formal financial infrastructure is weakest.

Third, geographic region. The interaction term is significant in the southern subsample (N = 2403; −0.0217, p < 0.05) but not in the northern subsample (N = 144). Given the northern sample’s small size, comparisons across regions lack statistical power—so we interpret the southern result as evidence of the effect in that context, not as evidence of regional heterogeneity. The northern estimate is reported descriptively only.

Fourth, subjective self-perception. The interaction term is significant and larger in the low self-perception group than in the high self-perception group, suggesting that digital inclusive finance disproportionately benefits individuals with lower psychological capital—a pattern consistent with its compensatory function. However, because the measure relies on a single item, this interpretation is necessarily tentative.

In summary, the heterogeneity analysis reveals stronger effects among residents of high-labor-outflow provinces and rural households. Significant effects also appear in the southern ethnic minority subsample—but northern comparisons are infeasible due to the small sample (N = 144). We also find larger effects among those with lower subjective social standing, though this finding is based on a single-item measure. Collectively, these results align with the theoretical framework and underscore the policy relevance of digital inclusive finance for advancing educational equity and mitigating intergenerational inequality.

6. Mechanism Analysis

This section investigates the underlying mechanisms through which digital inclusive finance fosters intergenerational educational mobility in ethnic minority regions. Drawing on both theoretical reasoning and extant empirical evidence, we posit that digital inclusive finance operates via two primary channels: (i) alleviating household financial constraints—thereby expanding capacity and willingness to invest in offspring’s human capital—and (ii) enhancing information accessibility—enabling more timely and informed educational decision-making.

Table 6. Mediation Mechanism Analysis.

Dependent Variable	(1)	(2)	(3)	(4)
	ln (Total Household Net Income)	Information Accessibility
		Television	Newspapers & Periodicals	Radio
DIFI	0.8170 **	9.6751 ***	6.6499 **	5.2847 **
	(2.4417)	(5.1206)	(2.4230)	(2.7448)
Offspring Control Variables	Yes	Yes	Yes	Yes
Parent Control Variables	Yes	Yes	Yes	Yes
Family Control Variables	Yes	Yes	Yes	Yes
Regional Control Variables	Yes	Yes	Yes	Yes
Year Fixed Effects	Yes	Yes	Yes	Yes
Regional Fixed Effects	Yes	Yes	Yes	Yes
Cohort Fixed Effects	Yes	Yes	Yes	Yes
Observations	2481	1552	1552	1552
adj. R2	0.294	0.048	0.029	0.050

presents the results of formal mediation analyses examining these pathways, specifically the mediating roles of household net income (Column 1) and multidimensional information accessibility—measured separately by perceived importance of television, newspapers and periodicals, and radio as education-related information sources (Columns 2–4).

Household net income is derived from the “total household net income” variable in the CFPS survey. To mitigate estimation bias arising from right-skewed distribution, we apply the natural logarithmic transformation. This variable captures post-deduction disposable income and serves as a robust proxy for household economic capacity—the foundational resource underpinning educational investment.

Information accessibility is proxied by households’ self-reported use of TV, newspapers, and radio for education-related information. These traditional media indicators are used because longitudinal Internet usage data in the CFPS remain scarce for ethnic minority regions.

Column 1 shows that digital inclusive finance is associated with higher household net income (0.817, p < 0.05), supporting the income channel in Section III. Columns 2–4 show positive, significant associations with TV, newspaper, and radio use for education-related information (all p < 0.05), consistent with the information channel in Section III. However, these media proxies capture only conventional-channel exposure—not digital access or information quality—so findings for this mechanism should be interpreted with caution.

In summary, the evidence supports household net income as a mediating pathway and offers preliminary support for the information accessibility channel—measured indirectly via traditional media usage. Collectively, these findings underscore digital inclusive finance’s role in expanding economic participation among historically excluded populations and, more cautiously, in advancing informational equity.

Data limitations prevent analysis of other channels—such as cultural identity and social capital—because the CFPS lacks consistent longitudinal measures. We therefore focus on household income and information accessibility as foundational preconditions for educational investment. Future research with richer survey instruments should examine how digital inclusive finance interacts with sociocultural factors to shape educational trajectories.

7. Research Conclusions and Policy Implications

This study finds that digital inclusive finance is associated with upward intergenerational educational mobility in China’s ethnic minority regions. A 10% increase in the Digital Inclusive Finance Index (DIFI) reduces the intergenerational elasticity of schooling by 2.6 percentage points, operating primarily through higher household income and, secondarily, improved access to education-related information—measured here via self-reported use of traditional media. Effects are strongest among rural households and provinces with high labor outflow. Significant effects also appear in the southern ethnic minority subsample—but the northern subsample is too small for reliable regional comparisons. We also find larger effects among those with lower subjective social standing, though this finding is based on a single-item measure.

Based on the above findings, this paper proposes the following evidence-informed policy recommendations:

First, prioritize digital infrastructure investment—broadband connectivity and mobile payment systems—in rural ethnic minority regions, complemented by policy-based credit facilities to reduce the cost of education-related financial services.

Second, financial capability programs should be institutionalized in schools and community centers, with curricula co-developed by local stakeholders to ensure cultural relevance and contextual appropriateness.

Third, digital inclusion initiatives should be systematically integrated with complementary policies—including means-tested education subsidies, entrepreneurship support programs, and social assistance—to ensure that expanded financial access translates into sustained educational investment and improved labor market outcomes.

This study has several limitations. First, years of schooling—used as a proxy for educational attainment—fails to capture variation in school quality, especially in ethnic minority regions where institutional resources differ substantially. Second, two key measures are constrained by data availability: information access is measured via traditional media usage (not digital access), and subjective social standing is assessed with a single survey item—insufficient to represent multidimensional constructs like self-efficacy. Third, the provincial-level DIFI obscures sub-provincial disparities in financial access, and the CFPS lacks granular data on specific digital financial products. Fourth, although our instrumental variable approach mitigates endogeneity, the exclusion restriction is unverifiable—so results are suggestive of a causal relationship, not definitive proof. Future research should integrate platform-level digital finance data with household surveys and use quasi-experimental designs (e.g., difference-in-differences) to better identify long-term causal effects.