Exploring the income impact of rural e-commerce comprehensive demonstration project and determinants of county selection

The study investigates whether participation in China’s National Rural E-commerce Comprehensive Demonstration Project (NRECDP) increases rural residents’ income in national-level key poverty-stricken (NLKPS) counties and what factors drive county selection into the program. It addresses a key gap in prior literature by accounting for self-selection into the program and by explicitly examining the role of digital inclusive finance in both selection and income outcomes. Contextually, rural e-commerce is seen as a transformative force that reduces geographic barriers, lowers transaction costs, and connects rural producers to broader markets, thereby potentially enhancing incomes and supporting poverty alleviation. The paper evaluates the average income impact of NRECDP participation, explores heterogeneity across ethnic versus non-ethnic regions, geographic regions (eastern/central/western), levels of industrial structure, and the designated “three regions and three prefectures,” and examines entrepreneurship as a mechanism linking NRECDP to income growth. The study’s purpose is to provide evidence on policy effectiveness and on the determinants that make some counties more likely to be selected, aiding the design of targeted poverty alleviation and rural revitalization strategies.

The literature identifies rural e-commerce as a key pathway for rural development, poverty alleviation, and narrowing income gaps by overcoming geographical barriers, reducing transaction costs, and integrating rural areas into broader markets. Prior work documents effects on income growth, inequality, and poverty reduction; it also highlights how e-commerce can upgrade supply chains, shift rural industrial structure toward higher value-added services, and modernize agriculture. Government-led demonstrations, including the NRECDP, have been central to promoting rural e-commerce. However, determinants of county participation and the issue of self-selection have been less examined. Digital inclusive finance is emphasized as a new model expanding access to payments, credit, and investment for low-income groups, potentially fostering entrepreneurship, easing financing constraints, improving resource allocation, and stimulating consumption and investment; yet it can face challenges due to limited rural infrastructure, financial literacy, and potential higher costs and risks. The study responds to gaps by considering multi-dimensional determinants of participation (infrastructure, human capital, industry, financial conditions) and explicitly incorporating digital inclusive finance and self-selection into the empirical strategy.

Data cover 592 NLKPS counties from 2014–2020. The outcome variable is the natural log of rural residents’ per capita net income drawn from county statistical bulletins. The key policy variable is participation in the NRECDP (binary), which provides fiscal support for rural e-commerce capacity (logistics, public services, training). The core explanatory variable is the Digital Inclusive Finance Index (Peking University), capturing coverage breadth, usage depth, and digital support services, with sub-indices for payment, insurance, monetary funds, investment, credit investigation, and credit. Controls include education (enrollments in primary and secondary schools), population density, per capita GDP, infrastructure (number of fixed-line subscribers), industrial structure (value-added of tertiary industry relative to GDP), fiscal self-sufficiency (ratio of revenue to expenditure), and e-commerce development (number of Taobao villages). The main empirical approach is an endogenous switching regression (ESR) model to correct for self-selection based on observed and unobserved factors, specifying a selection equation for NRECDP participation and two outcome equations for income under participation and non-participation. An instrumental variable for selection is the interaction of county average slope with year (slope×year), justified as affecting e-commerce feasibility via infrastructure and logistics but not short-term income growth rates. The ESR permits estimation of counterfactual outcomes and the average treatment effect on the treated (ATT). Robustness is checked using an endogenous treatment effect model with similar selection and outcome equations and by replacing the composite digital inclusive finance index with its components (financial coverage breadth and usage depth). Heterogeneity analyses stratify by ethnic versus non-ethnic regions, central versus western regions, levels of industrial structure (above/below mean), and the designated “three regions and three prefectures.” Mechanism analysis examines entrepreneurship as an outcome (log of business registrations plus one) to test whether NRECDP promotes entrepreneurial activity linking e-commerce development to income growth.

- NRECDP participation significantly raises rural income in NLKPS counties. The ESR-based ATT is 0.122 log points, corresponding to an approximate increase of 12.97% in rural residents’ net income (Table 4). A related counterfactual impact reported for the control group is about 9.20%. - Determinants of county selection into NRECDP (selection equation, Table 3): higher digital inclusive finance (coef 3.183, p<0.01), better infrastructure (0.114, p<0.05), stronger industrial structure (1.139, p<0.01), higher per capita GDP (0.224, p<0.05), and more Taobao villages (0.718, p<0.01) increase the likelihood of selection; terrain slope×year is negative and significant (−0.320, p<0.01), indicating flatter areas are more likely to be selected. Education and fiscal self-sufficiency are not significant predictors; population density is negative but not significant. - Income determinants (outcome equations, Table 3): digital inclusive finance is positive and significant for both participants (0.401, p<0.01) and non-participants (0.292, p<0.01), with a larger marginal effect among participants. Industrial structure is positive and significant in both regimes (0.283 and 0.369, p<0.01). Education is positive and significant in both (0.175 and 0.067), and per capita GDP is significant among participants (0.216, p<0.01) but not among non-participants. Population density is strongly positive among participants (2.819, p<0.01) and modestly positive among non-participants (0.053, p<0.05). Infrastructure and fiscal self-sufficiency are not significant in outcome equations. - Instrument validity: slope×year is significantly negative in the selection equation (<1% level), supporting relevance; LR tests indicate endogenous selection (ESR ρ parameters significant; LR test of independence 14.46, p<0.01). - Robustness: An endogenous treatment effect model confirms positive and significant effects of NRECDP on income and similar roles for digital inclusive finance, e-commerce development, and industrial structure (Table 5). Replacing digital inclusive finance with coverage breadth and usage depth yields comparable treatment effects (~20.32% and ~18.05%, Table 6). - Heterogeneity (Table 7): Effects are larger in non-ethnic regions (~25.23%) than ethnic regions (~13.88%); central region (~23.12%) exceeds western (~11.85%); regions with higher industrial structure levels show somewhat larger effects (~14.45%) than lower levels (~13.43%); “three regions and three prefectures” exhibit the largest gain (~37.98%) compared to non-designated areas (~8.11%). - Mechanism (Table 8): NRECDP significantly increases entrepreneurship (ATT 0.111 in logs; ~11.74% increase), indicating entrepreneurship as a channel linking e-commerce support to income growth.

The findings demonstrate that the NRECDP effectively raises rural incomes among NLKPS counties, addressing the core research question on policy impact while correcting for self-selection. The positive and larger effect of digital inclusive finance among participating counties suggests complementarities between program support and local digital financial ecosystems in translating e-commerce development into income gains. Selection into the program aligns with existing e-commerce readiness and stronger economic and industrial bases, consistent with policy design intent to leverage counties with solid foundations for scalable demonstrations. Heterogeneity results indicate that while gains are widespread, they are more substantial in non-ethnic and central regions and in areas with stronger tertiary industries, reflecting lower logistical frictions and better market access. Yet the very large effect in the “three regions and three prefectures” underscores the policy’s capacity to deliver impactful gains even in severely impoverished areas when support is targeted. The mechanism analysis supports entrepreneurship as a pathway through which e-commerce support policies convert into higher rural incomes, highlighting the role of training, finance, and infrastructure in lowering startup barriers and enabling non-agricultural income opportunities. Collectively, the results validate rural e-commerce development—augmented by digital inclusive finance—as an effective instrument for poverty alleviation and rural revitalization.

The study provides causal evidence, accounting for self-selection, that participation in the NRECDP increases rural residents’ income in NLKPS counties by approximately 12.97%. County selection into the program is shaped by economic development, e-commerce readiness, infrastructure, industrial structure, and digital inclusive finance, while income levels are further influenced by education and population density. Effects vary across regions, with notably high gains in the “three regions and three prefectures,” supporting the value of targeted intervention. Entrepreneurship emerges as a key mechanism linking policy support to income growth. Policy recommendations include: continuing and intensifying NRECDP implementation with targeted support for severely impoverished areas; improving logistics and digital infrastructure and expanding market linkages, including international channels for local agricultural products; and accelerating digital inclusive finance alongside training to enhance digital literacy and entrepreneurial skills. Future research could refine infrastructure measures (e.g., internet users, road mileage), better account for changes in county classifications over time, and explore additional mechanisms and spillover effects.

The study’s infrastructure proxy (fixed-line telephone subscriptions) may not fully capture aspects most relevant to rural e-commerce, such as broadband penetration or road connectivity, due to data availability constraints. Changes in the NLKPS county lists after 2014 were not fully controlled for, which may introduce classification inconsistencies over time. These factors could affect the precision of estimated relationships and the generalizability of results.