How does inter-provincial trade promote economic growth? Empirical evidence from Chinese provinces

The study investigates whether and how inter-provincial (domestic) trade promotes economic growth in China amid shifting global conditions and a national strategy emphasizing domestic circulation. It addresses gaps in understanding the mechanisms linking inter-provincial trade to growth and examines regional heterogeneity across eastern, central, and western China. The authors hypothesize: H1) inter-provincial trade positively impacts economic growth; H2) the effect operates via technological improvement and industrial structure upgrading; H3) these mechanisms exhibit regional heterogeneity.

Prior research explores measurement of inter-provincial trade using official reports, VAT invoice data, and model-based approaches (gravity, input–output, location quotient), with gravity-based estimation most common. Empirical findings generally suggest domestic/inter-provincial trade supports growth, market integration, and welfare, though some studies report insignificant or unstable effects. Studies also link interprovincial goods turnover to higher emissions consistent with growth dynamics. However, there is limited consensus on the internal mechanisms (technology, industrial upgrading) and a lack of analyses on regional heterogeneity within China.

Data: Panel of 29 Chinese provinces (excluding Hainan, Tibet, Hong Kong, Macao, Taiwan) from 2000–2021, sourced from China Statistical Yearbook, China Science and Technology Statistical Yearbook, China Transportation Yearbook, and provincial yearbooks. Variables are log-transformed and winsorized. Key variables: - Dependent variable: Economic growth measured by per capita GDP (Inper_GDP). Robustness replaces with gross regional product and per capita disposable income. - Independent variable: Inter-provincial trade (Intrade) computed via (i) transport volume distribution coefficient method (Ihara, 1996) to estimate bilateral trade friction Q_RS from railway freight flows scaled by road/water multipliers, and (ii) an interregional gravity model t_RS = s_R * d_S * Q_RS, where s_R and d_S are provincial supply and demand constructed from GDP minus net exports/outflows (with FOB adjustments and inclusion of services). Post-2017 net outflows are estimated by extrapolating using changes in railway outflows. - Controls: International trade (total imports+exports), technical input (accepted domestic invention patents), investment (total fixed asset investment), human capital (average years of schooling computed from educational attainment shares). Econometric model: Two-way fixed effects panel regression with province and year fixed effects and cluster-robust standard errors: ln(per_GDP_it) = a + a1 ln(Intrade_it) + a2 ln(interna_trade_it) + a3 ln(patents_it) + a4 ln(cap_it) + a5 ln(h_cap_it) + u_i + λ_t + ε_it. Robustness checks: - Alternative dependent variables (GRP, per capita disposable income). - Population migration split (in-migration vs out-migration provinces). - Replace Intrade with total retail sales of consumer goods as a proxy for domestic trade scale. Endogeneity (IV) strategies: - IV1: Lagged Intrade (one period) as instrument; 2SLS with strong first-stage statistics. - IV2: Domestic market convenience (DMC) constructed as reciprocal of river density multiplied by year (time-varying), following Fenske et al. (2023) to introduce time variation. - IV3: Domestic trade constraint (DTC) using topographic relief interacted with RMB–USD exchange rate. Weak-IV and under-identification tests (minimum eigenvalue, Kleibergen–Paap rk LM) indicate strong, valid instruments. Mechanism (mediation) tests and regional heterogeneity: - Technology improvement mediator: total factor productivity (TFP); robustness uses log share of local fiscal S&T expenditure in general budget (from 2007). Mediation model regresses mediator on Intrade and controls. - Industrial structure upgrading mediator: log ratio of tertiary to secondary industry value added. - Regional sub-samples by NBS classification: East (11 units), Middle (9), West (9).

- Baseline effect: Inter-provincial trade significantly promotes economic growth. Preferred specification indicates a 1% increase in Intrade raises per capita GDP by about 0.19% (Table 2, coefficient ≈ 0.191, t≈4.06), confirming H1. Controls for international trade, investment, and human capital are significantly positive. - Robustness: Results hold when using GRP (Intrade ≈ 0.248***), per capita disposable income (≈ 0.067***), splitting provinces by migration status (positive and significant in both in- and out-migration groups), and when using retail sales as an alternative proxy for domestic trade scale (positive effects persist). - Endogeneity: IV estimations confirm causality from inter-provincial trade to growth; instruments are strong (e.g., minimum eigenvalue statistics >10; Kleibergen–Paap rk LM significant). The 2SLS second-stage coefficient on Intrade remains positive and significant (e.g., ≈ 0.221*** in IV1). - Mechanisms and heterogeneity (H2, H3): • Technology improvement (mediator): Nationally positive (Entire sample TFP model: Intrade ≈ 0.074**, Table 5). Established in Central (≈ 0.153**) and West (≈ 0.217***) regions; not established in East (insignificant/negative). Robust to alternative technology measure (S&T expenditure share). • Industrial structure upgrading (mediator): Nationally positive (Entirety ≈ 0.185**). Significant in East (≈ 0.242**) and Middle (≈ 0.401***); not significant in West (≈ −0.112, n.s.), indicating the mechanism does not operate there. Overall, inter-provincial trade fosters growth via technology improvement in Central/West and via industrial upgrading in East/Middle, with clear regional heterogeneity.

The findings validate that expanding inter-provincial trade deepens market integration and division of labor, thereby increasing productivity and growth. The mechanisms differ by development stage and endowments: Eastern provinces, already technologically advanced and highly open internationally, do not gain additional technology spillovers from domestic trade but do channel trade into industrial upgrading. Central provinces benefit from both mechanisms given improving infrastructure and absorptive capacity. Western provinces gain from technology spillovers but face constraints (infrastructure, factor endowment imbalance, industrial transfer of high-emission sectors) that limit industrial upgrading. These results underscore the need for region-specific policies to maximize growth gains from domestic trade.

This paper contributes by constructing province-level inter-provincial trade flows using transport-based frictions and a gravity framework, and by identifying two distinct mechanisms—technology improvement and industrial structure upgrading—through which domestic trade promotes growth, with pronounced regional heterogeneity. Policy implications include: nationally, continue to foster inter-provincial trade alongside stable international trade; in the East, leverage industrial upgrading and strengthen inter-provincial S&T exchange; in the Central region, further enhance technological capability and support both mechanisms; in the West, improve infrastructure and human capital to bolster technology absorption and lay groundwork for upgrading while aligning with carbon goals. Future research should incorporate digital trade, explore heterogeneity under different geographic and policy contexts (e.g., coastal vs inland, border policies), and utilize micro-level (firm/industry) data to unpack channels. Emerging themes such as global value chains, cross-border e-commerce, AI, and green trade merit integration into analyses of domestic trade-growth dynamics.

- Measurement of inter-provincial trade relies on the transport distribution coefficient and an interregional gravity model, which assumes material delivery shares approximate goods and services flows; services may be underrepresented. - Air and pipeline freight are excluded when scaling from railway flows, potentially biasing friction estimates. - Post-2017 net outflows are extrapolated using railway outflow ratios, which may diverge from actual values, affecting provincial demand estimates and trade flows. - Estimation choices and variable constructions can accumulate errors. Future work could use more consistently defined and longer-span datasets, include all transport modes, and leverage VAT invoice data to improve precision.