Digital Financial Inclusion and Multidimensional Poverty Alleviation: Empowering Young Migrant Workers in China

2.1 Definition and Poverty Measurement

The concept of poverty has been transformed from a narrow economic perspective of resource scarcity into a comprehensive, multidimensional understanding of social deprivation. Adam Smith pioneered the argument in The Wealth of Nations that poverty extends beyond the absence of basic necessities to encompass the inability to afford “necessaries which the customs of the country render it indecent for creditable people, even of the lowest order, to be without.” The social shame experienced by a respectable worker without a linen shirt was used to demonstrate the social dimension of a “decent life.” Building on this foundation, the concept of “relative deprivation” was introduced by Townsend (1979), who posited that poverty arises when individuals are unable to participate in customary social life because of limited resources. Poverty was redefined from absolute scarcity to a form of social exclusion. Sen (1999) further developed this understanding through the Capability Approach, in which poverty was conceptualized as the deprivation of substantive freedoms to engage in basic functional activities. Within this framework, emphasis was placed on moving beyond income or consumption metrics to assess whether genuine capabilities exist for individuals to pursue their valued life choices, with attention directed toward agency, resource-capability conversion mechanisms, and the structural foundations of freedoms, choices, and institutions.

As these theoretical foundations evolved, scholars and international organizations developed various frameworks for measuring MP. Traditional poverty metrics, such as income- or consumption-based poverty lines (Chen and Ravallion 2007), have been widely applied because of their simplicity and practicality. However, poverty has increasingly been recognized as a complex, cross-sectoral phenomenon that cannot be fully captured by monetary indicators alone. Exclusive reliance on income or consumption measures is viewed as inadequate because critical dimensions of deprivation, such as health, education, housing, and social participation, may be overlooked (Alkire and Santos 2014; Sen 1985). In response, MPI was developed to encompass various domains, including income, education, health, and quality of life (Alkire and Foster 2011; Cheng 2023; Neckerman et al. 2016; Wang et al. 2024). Although these frameworks have advanced the understanding of poverty, ensuring that the lived experiences of marginalized groups across different regions are accurately reflected remains a persistent challenge. Within this context, the United Nations Development Programme’s Global MPI has been designed to measure health, education, and living standards in alignment with the sustainable development goals, now covering more than 570 million people worldwide (Alkire and Kanagaratnam 2020). In China, measurable deprivation indicators were created through the transformation of the targeted poverty alleviation policy “two assurances and three guarantees” thereby enabling more precise identification of MP (Cheng et al 2021). Nonetheless, poverty continues to be challenged by the applicability of these indicators across diverse socioeconomic contexts, together with regional variations in data availability (Kana Zeumo et al. 2014).

2.2 MP Issues of Migrant Workers

Rural-to-urban migrant workers in China face complex and interconnected MP challenges as a result of institutional barriers and structural inequalities. The household registration system hukou, a distinctive policy in China, categorize individuals as rural or urban residents according to birthplace. Substantial restrictions are imposed on rural-to-urban migrant workers through this system, preventing them from obtaining equal rights and services available to urban residents. In particular, access to employment opportunities, income, social security, and public services such as healthcare and education is constrained despite their physical presence in urban areas. A sharp division between urban and rural populations is established through the hukou system, thereby generating significant disparities in living conditions and opportunities for migrant workers (Chiang et al. 2021).

This institutional constraint is compounded by the urban-rural economic dualism, which intensified socioeconomic vulnerability (Huang and Zhang 2019). Young vibrant workers are particularly susceptibility to social marginalization, discriminatory treatment, and obstacles to urban integration, thereby impeding their full economic and social participation (Tang et al. 2020). Although opportunities for mobility have been expanded by urban reindustrialization and service sector growth, the predominance of low-skilled employment has resulted in intense competition and the dual challenges of inadequate wages and job insecurity. Consequently, higher living expenses and heightened employment uncertainty are generated, further amplifying the vulnerability of this demographic (Huang and Zhang 2019). Income deprivation has been shown to arise from the combined effect of low wages and employment instability, while social protection deprivation has been manifested in limited health insurance participation (approximately one-third) and minimal urban homeownership (merely 5.7 %) (Chiang et al. 2021). Furthermore, the susceptibility of the new generation of migrant workers to in-work poverty has been heightened by restricted skill development and institutional embeddedness in informal employment (Tang et al 2020). Through these structural limitations, young vibrant workers experience multidimensional deprivation, and their vulnerability is reinforced across multiple aspects of life, including access to stable housing, healthcare, and social services.

2.3 Effect of DFI on the Comprehensive Poverty of Migrant Workers

DFI is a vital mechanism for promoting inclusive growth through the expansion of financial resource access among underserved populations. DFI delivers affordable and accessible financial services, enabling households to manage risks, stabilize consumption, and engage in productive activities. Corrado and Corrado (2017) demonstrated that household resilience is enhanced through DFI by strengthening long-term financial planning and economic participation, thereby supporting broader sustainable development objectives. Similarly, Suri and Jack (2016) reported that Kenya’s M-PESA substantially increased per capita consumption and elevated 194,000 households above poverty. Comparable outcomes have also been observed in other contexts, such as mobile financial services such as bKash in Bangladesh and the Pix payment system in Brazil, which have been used to extend financial access to millions of low-income households through cost-effective, technology-driven platforms (Duarte et al. 2022; Lee et al. 2020).

Collectively, these cases demonstrate how technology-driven financial services mitigate poverty by expanding financial access and reducing transaction barriers, particularly in regions underserved by traditional banking. this has been emphasized by Yang et al. (2025) that poverty is reduced through DFI by lowering service costs, improving access to financial information, and fostering transparency and social trust that facilitate economic participation of vulnerable groups. Similarly, Lee et al (2023) demonstrated that poverty alleviation is promoted through DFI via income growth and spatial spillover effects, while the relationship between DFI and poverty reduction has been shown to follow a “U-shaped” trajectory, reflecting both early-stage risks and long-term benefits of financial deepening.

Yang et al (2022) found that urban integration is facilitated through DFI with regard to migrant workers. Tsao and Tsaih (2023) developed an AI-based microloan platform to evaluate migrant workers’ credit scores and default risks, thereby improving the accessibility of financial services. However, Ozili (2020) cautioned that digital finance may intensify internal inequalities in the financial system and increase both the types and levels of risk faced by poor populations. Despite these divergent academic perspectives, empirical evidence regarding the impact of DFI on the MP of young migrant workers is limited. As this group is both digitally active and structurally disadvantaged, this study presents significant research value. From the Capability Approach perspective, the potential enhancement of substantive freedoms in areas such as education, health, and employment may be achieved through DFI, rendering it a crucial tool in alleviating MP. Accordingly, the following hypotheses are proposed:

Existing research revealed significant regional variations in the effects of DFI due to differences in institutional environments and development foundations. These variations manifest in the ways DFI reduces MP and promotes social integration. Wang et al. (2024) found that stronger MP-reducing effects of DFI occur in central and western regions, as well as in rural areas, through the enhancement of credit access, building of social capital, and the promotion of non-agricultural employment, particularly in regions underserved by traditional financial services. Yang et al (2022) found that inclusive digital finance services more effectively support migrant urban integration in Chinese cities outside the top 100 than in first-tier cities, indicating that city-level disparities influence the effectiveness of DFI in reducing poverty. Furthermore, it has been highlighted that digital infrastructure is more developed in southern China, where the impact of DFI on migrant urban integration is correspondingly greater. Collectively, these findings underscore the critical role played by regional disparities in infrastructure development, financial access, and urban structure in shaping DFI outcomes. Accordingly, the following hypotheses are proposed:

Regarding the susceptibility of DFI to “elite capture,” divided perspective has been expressed in the scholarly literature. Balakumar and Maitra (2023) argued that, in weak institutional environments, credit allocation is often influenced by politically connected families, thereby undermining financial inclusion. Gabor and Brooks (2016) argue that data commodification and closed algorithmic systems in digital finance may result in the concentration of financial power and perpetuation of inequality. Ozili (2018) warned that insufficient regulation could widen the “digital divide,” thereby leaving vulnerable populations exposed to fraud and gender-based discrimination. Conversely, digital finance has been demonstrated to substantially reduce geographic barriers and transaction costs, thereby improving financial access for disadvantaged groups. Mobile services, such as mobile payments, address traditional banking constraints, especially in remote regions, while digital risk control reduces information asymmetry, enabling low-income individuals, rural residents, and women to access financial services at lower costs (Munyegera and Matsumoto 2016; Khera et al. 2022; Wang et al. 2024; Yang et al 2022). Although young migrant workers demonstrate strong digital literacy, structural barriers are encountered, including household registration limitations, inadequate social security, and restricted formal employment opportunities, all of which constrain their access to essential life domains such as income, education, health, and information. If DFI platforms successfully overcome obstacles such as geographic isolation, administrative complexity, and identity verification requirements, opportunities can be expanded, service accessibility can be enhanced, information access can be improved, and job security can be strengthened, thereby fostering development potential. Accordingly, the following hypothesis is proposed:

Several studies have explored the nonlinear and contextual mechanisms through which DFI alleviates poverty. Lee et al (2023) found that DFI promotes poverty alleviation while exhibiting a U-shaped nonlinear relationship, implying that financial inclusion may initially increase poverty risks before long-term benefits are generated as digital infrastructure and governance improve. In contrast, Wang et al. (2024) demonstrated that the poverty-reduction effects of DFI depend on the broader financial environment; once the financial system has crossed a threshold of institutional maturity, the impact on MP becomes significantly stronger, particularly through credit access, social capital, and non-agricultural employment. Collectively, these findings imply that the effectiveness of DFI is both nonlinear and context-dependent, strengthening as institutional and environmental capacities improve. Accordingly, the following hypothesis is proposed:

3.1 Data Description

This study used data from the 2019 wave of the CHFS, administered by the Survey and Research Center for China Household Finance at Southwestern University of Finance and Economics. The CHFS implemented a comprehensive multistage sampling strategy based on probability proportional to size, covering 29 mainland Chinese provinces, with the exception of Xinjiang and Tibet. High standards of data quality and national representativeness are maintained, making the dataset a valuable resource for poverty, household finance, and social protection research.

The 2019 CHFS dataset comprises information from 34,643 households and 107,008 individuals. After data cleaning and selection criteria application, 11,066 valid observations representing young migrant workers are identified. In accordance with the classification of Chen et al (2025), young migrant workers are defined as individuals aged 18–34 who hold agricultural hukou registration and are employed in non-agricultural sectors.

For the measurement of DFI, the Peking University DFI Index is employed. This index is specifically designed for the Chinese context and provides a detailed assessment of digital financial services across regional and demographic variations. While the World Bank Group’s Global Findex offers broader international data on financial inclusion, the Peking University DFI Index incorporates three essential dimensions for analyzing the impact of DFI on marginalized groups: coverage breadth, usage depth, and digitalization level. These components are particularly relevant for evaluating financial inclusion among young migrant workers in China. Consequently, the Peking University DFI Index is selected for this study owing to its contextual specificity and detailed alignment with our research objectives.

3.2 Selection of Variables

3.3 Descriptive Statistical Analysis

Table 2 presents the descriptive statistics, showing that the MPI for young migrant workers is 32.8 %. The incidence of MP gradually declines from 93.8 % to 0 % as the deprivation threshold increases. At the individual level, the average age of the respondents was 26 years, with 52.3 % being male and 46.4 % being married. While 35.4 % of respondents possess financial knowledge, only 4.9 % demonstrate economic awareness. At the household level, the average family size is 4.6 members, with a debt ratio of 55.4 %, a child dependency ratio of 11.9 %, and an elderly dependency ratio of 4.8 %, indicating substantial financial and caregiving burdens among young migrant households.

The threshold value represents the number of dimensions in which an individual must experience deprivation to be considered poor; as the threshold increases, individuals must be deprived in more dimensions to qualify as poor. Figure 1 illustrates that MP incidence declines with higher threshold, indicating that fewer individuals experience deprivation across multiple dimensions. Following Yi and Lu (2022), the primary threshold is set at 0.4, while a more stringent threshold of 0.5 identifies individuals deprived in at least half of the dimensions, reflecting more severe poverty. Based on these criteria, the incidence of MP among young migrant workers is 32.02 % at the 0.4 threshold and 20.75 % at the 0.5 threshold, providing a foundation for subsequent analysis.

Figure 1:

Relationship between multidimensional poverty (MP) incidence and threshold value.

This study mapped the incidence of MP at the 0.4 and 0.5 thresholds across provincial-level administrative regions using sampling weights from the CHFS database, illustrating the prevalence of MP among young migrant workers in 2019. The visualization employs a standard map format (Figure 2), where darker shadings represent a higher incidence of MP. Provinces such as Guizhou, Yunnan, Gansu, Qinghai, and Hainan appear with darker color blocks, indicating that young migrant workers have elevated MP levels. This geographical representation provides an important reference for analyzing regional disparities in MP across China.

Figure 2:

Incidence of MP among young migrant workers across provincial regions in 2019. Note: This map was derived from a standard map acquired from the Map Technical Review Center of China’s Ministry of Natural Resources, bearing the review number GS(2020)4,619. The foundational map boundaries remain unaltered.

3.4 Model Setting

4.1 Baseline Regression Analysis

This analysis first estimates the baseline relationship between DFI and MP using a multiple regression model and then examines DFI’s impact on MP incidence using Probit models at 0.4 and 0.5 poverty thresholds, with the results presented in Table 3.

The analysis revealed a consistent and statistically significant negative relationship between DFI and MP. The ordinary least squares estimation shows that a one-unit increase in DFI is associated with a 0.136-unit reduction in the MPI. Probit model results further confirm this relationship, with significant negative coefficients of −0.924 and −1.412 at the 0.4 and 0.5 poverty thresholds, respectively. Collectively, these findings provide robust evidence supporting Hypothesis 1, indicating that DFI effectively alleviates MP among young migrant workers.

4.2 Robustness Test

To confirm the baseline results are robust and not dependent on model choice or sampling methods, three distinct robustness checks were conducted.

4.3 Heterogeneity Analysis

Given China’s substantial regional disparities in digital infrastructure, fiscal support, and other socioeconomic factors, the impact of DFI on MP is likely to be heterogeneous across regions. To examine this, the sample is divided into western and eastern-central subgroups, and separate regression analyses are conducted for each. Table 8 presents the results of these subgroup analyses.

Table 8 shows that at the 0.5 MP threshold, the overall DFI coefficient is −1.412 and statistically significant at the 1 % level. In the eastern-central region, the coefficient is −1.116 (1 % significance), whereas in the western region, it is −2.373 (5 % significance), indicating a stronger poverty-reducing effect of DFI in the western region. However, the DFI effect in the western region is not statistically significant in other scenarios. These findings provide empirical support for Hypothesis 2.

5.1 Sub-indices of the DFI

Panels A, B, and C of Table 9 report the effects of DFI’s sub-indices, coverage breadth, usage depth, and digitalization, respectively. Columns (1)–(3) present regressions without IV using MPI and the two MP thresholds as dependent variables, whereas Columns (4)–(6) show the corresponding IV regression results. The findings indicate that all three sub-indices significantly reduce MP among young migrant workers, with digitalization exhibiting the strongest and most consistent effect, followed by coverage breadth and usage depth across all model specifications.

5.2 Elite Capture

To investigate potential elite capture effects, this study uses “institutional employment” as a proxy, defined as households with at least one member working in a government or public institution. Individuals with institutional employment may enjoy privileged access to financial services through stable incomes, professional networks, and governmental connections, potentially leading to preferential financial terms or enhanced access to formal financial products. An interaction term between DFI and institutional employment is included in the analysis to determine whether elite capture moderates the relationship between DFI and MP.

Table 10 shows that DFI retains a significant negative relationship with MP across both models, with coefficients of −0.917 and −1.422, each significant at the 1 % level. The analysis further indicates that neither institutional employment nor its interaction with DFI is statistically significant. These findings support Hypothesis 3, indicating that DFI’s poverty-reducing effects occur independently of elite employment status, demonstrating that the benefits of digital financial inclusion are broadly inclusive.

The use of institutional employment as a proxy for elite capture has certain limitations. Although it captures some aspects of privileged access to financial resources, it may not reflect all dimensions of elite capture, especially those arising from political connections or extensive social networks. Moreover, employment in the government or public sector does not always confer substantial financial advantages. Owing to data constraints, institutional employment serves only as an approximate measure and may not fully capture the broader manifestations of elite capture.

5.3 Nonlinear Effects

This analysis investigates nonlinear effects by including a DFI squared term. Table 11 shows that the linear DFI coefficient is significantly positive in models (1), (2), and (3), while the quadratic term is significantly negative, indicating a concave (inverted-U) relationship between DFI and poverty. This suggests that at lower levels of DFI, MPI or MP may initially increase, but beyond a specific threshold, the marginal effect turns negative and DFI begins to reduce poverty. The estimated turning points are 5.646, 5.671, and 5.621 for models (1), (2), and (3) are 5.646, 5.671, and 5.621 respectively (Table 11), all within the observed DFI range of [5.5726, 5.9342] (Table 12).

A formal U-test for the OLS specification using MPI, presented in Table 12, confirms the inverted-U relationship: the slope at the lower bound is positive (0.1200, p = 0.0176), while the slope at the upper bound is negative (−0.4736, p < 0.001), with the extremum falling within the observed interval. Consistent with the estimated threshold, 76.73 % of the observations (8,491 of 11,066) have DFI values above 5.646, indicating a negative marginal effect. This percentage comes from the descriptive statistics. This finding explains the negative coefficient in the baseline linear model, as most observations fall to the right of the turning point where the slope is negative. Figure 3 illustrates the estimated DFI–MPI relationship, with red vertical lines marking the observed support [5.5726, 5.9342] and a green dashed line indicating the turning point at 5.646. These results support Hypothesis 4, demonstrating that the impact of DFI on MP exhibits a nonlinear pattern.

Figure 3:

Inverted-U relationship between digital financial inclusion and the multidimensional poverty index.