Research on the spatial spillover effect of high-speed railway on the income of urban residents in China

The paper examines how China’s rapid expansion of high-speed rail (HSR) has reshaped regional accessibility, labor mobility, and the spatial distribution of income among urban residents. Following market reforms and the easing of household registration barriers, large numbers of rural laborers moved to cities, forming a more unified national labor market. Despite overall income growth, inter-city income disparities widened from 2000 to 2018. HSR’s development—29,000 km by 2018—has altered spatial connections by reducing travel time and facilitating flows of people, goods, capital, and information, potentially affecting urban labor markets through agglomeration and diffusion. The study’s purpose is to quantify changes in accessibility due to HSR and assess the spatial spillover effects on urban residents’ income, evaluating whether HSR promotes balanced development or exacerbates regional inequality. It highlights the policy importance of leveraging transport infrastructure to restructure spatial economic patterns, reduce regional disparities, and support common prosperity.

Conventional models often assume frictionless spatial movement of factors, implying weak spillovers from transport infrastructure. New Economic Geography (Krugman, 1996; Fujita et al., 2001) emphasizes transport costs and predicts spatial spillovers via agglomeration and diffusion. Transport links can yield positive spillovers (developed areas driving less developed ones) or negative siphon effects (developed areas attracting resources from lagging regions). HSR reduces time costs, expands market access, and can enhance productivity and residents’ income by improving factor allocation, facilitating human capital flows, and knowledge diffusion (Pol, 2003; Chen et al., 2018; Donaldson, 2018). Improved accessibility supports urban scale effects and specialization, raising labor productivity and wages (Chen, 2012; Liu et al., 2022), while declining cities may fall behind (Pan et al., 2020). Spatial dependence typically weakens with distance (first law of geography). Given the network externalities of transportation infrastructure, spatial econometric models that incorporate spatial lags are necessary to avoid bias (Cohen, 2010). The paper therefore focuses on the spatial spillover of accessibility on income across local, adjacent, and broader regions, and its implications for regional balance.

The study measures regional accessibility and evaluates its spatial spillover effects on urban residents’ income using spatial econometric models for 286 prefecture-level-and-above cities in China from 2000–2018. 1) Regional accessibility: A weighted average travel-time approach is used. For city i, Ai = Σj(Tij * Mj) / Σj Mj, where Tij is the minimum inter-city railway travel time (using HSR speeds for cities and years with HSR in operation, and accelerated conventional-rail speeds otherwise), and Mj represents a composite attraction factor (population and GDP of city j). A lower Ai indicates better accessibility. An accessibility coefficient Ai* is defined as the ratio of a city’s accessibility to the mean accessibility across cities; higher Ai* implies better accessibility. 2) Spatial weights: Two matrices are employed. (a) Adjacency matrix Wij* (binary 0/1 for shared borders; isolated cities without neighbors are dropped, yielding 279×279). (b) Inverse distance matrix Wij** = 1/dij for i≠j, 0 otherwise, where dij is great-circle distance computed from latitude/longitude; the 286×286 matrix is row-standardized. Robustness is also checked with a double-weight SDM including both matrices. 3) Spatial autocorrelation: Global Moran’s I evaluates single-variable spatial dependence (accessibility; income). Bivariate Moran’s Ixy assesses the spatial correlation between the spatial lag of accessibility in other cities and local income. 4) Model: A Spatial Durbin Model (SDM) panel with city and time fixed effects: labor_it = α + ρ Σj Wij labor_jt + β1 reg_acc_it + β2 X_it + β3 Σj Wij reg_acc_jt + β4 Σj Wij X_jt + δ_i + μ_t + ε_it, with ε_it possibly spatially correlated via Σj Wij ε_jt. Controls X_it include: industrial structure (share of secondary industry in GDP), consumption scale (retail sales/GDP), government size (fiscal expenditure/GDP), human capital (primary and secondary students per 10,000 people), and resources/environment (per capita green space). Data are log-transformed to mitigate heteroskedasticity and collinearity. 5) Data: 2000–2018 panel for 286 cities assembled from multiple Chinese statistical yearbooks and geographic datasets (city adjacency, latitude/longitude assumed time-invariant). Missing values in some years are linearly interpolated. Balanced panel (N=5434 observations) is formed. 6) Estimation and tests: Baseline OLS LM and robust LM tests reject no-spatial-autocorrelation; Hausman tests favor fixed effects; Wald and LR tests support SDM over SAR/SEM. Results are reported for adjacency and inverse-distance weighting, with direct, indirect, and total effects decomposed. A time-split analysis compares 2000–2011 vs. 2012–2018, reflecting major HSR expansion and labor market turning points.

• Accessibility improved: The maximum accessibility coefficient among cities rose from 1.635 (2000) to 2.347 (2018), with polarization easing and overall accessibility improving, especially after widespread HSR rollout. • Significant spatial dependence: Single-variable Moran’s I for accessibility and income is significantly positive across years (e.g., 2000: accessibility I≈0.889; income I≈0.617), indicating high-high or low-low clustering. Bivariate Moran’s I between the spatial lag of accessibility and local income is significantly positive in many years, strengthening notably after 2009. • Spatial Durbin Model (SDM) results: Spatial dependence of income is strong and positive (spatial autoregressive parameter ρ significant and large; e.g., ρ≈0.63 with inverse-distance weights and ρ≈0.96 with adjacency). Changes in local accessibility generally show a negative direct effect on local income over 2000–2018, and neighboring accessibility often exerts a negative total effect (siphon effect), consistent with more accessible neighboring cities drawing resources away. • Robustness with double-weight SDM: Using both adjacency (W1) and inverse distance (W2), W1(accessibility) has a small positive association with local income (≈0.026, p<0.10), whereas W2(accessibility) is negative (≈−0.136, p<0.05). Spatial lags of income are positive and highly significant in both spaces (W1≈0.504, W2≈0.448; p<0.01), confirming mutual reinforcement of neighboring cities’ income levels. • Time heterogeneity: 2000–2011 shows generally negative or insignificant direct/indirect effects of accessibility on income (e.g., adjacency LR_Total negative and marginally significant). In contrast, 2012–2018 shows significantly positive direct, indirect, and total effects under both weighting schemes, implying that post-network formation, improved accessibility raised local incomes and generated positive spillovers. • Overall, cities with higher accessibility can siphon resources from nearby areas in early phases, but as the HSR network matures and labor market structure shifts (post-2012), positive spillovers dominate and raise incomes more broadly.

The study demonstrates that HSR-driven improvements in regional accessibility reshape inter-city economic linkages and labor market dynamics. Accessibility influences income through direct channels—expanding firms’ effective labor-shed, improving matching, and enabling technology diffusion—and through spatial spillovers due to the network nature of transport infrastructure. Early in the diffusion of HSR and accelerated rail speeds, advantages accrued to already accessible cores, generating siphon effects and negative spillovers for nearby, less competitive cities. As the network expanded and labor market structure shifted after 2012, improved accessibility facilitated wider factor flows, industrial upgrading, and employment growth across broader areas, converting initial centripetal agglomeration into more centrifugal diffusion. Positive spatial dependence of income indicates that cities’ incomes move together, reinforcing regional clustering. The findings underscore that transport networks can both widen and narrow regional disparities over time; policy should aim to harness network effects and complementary labor-market policies to maximize positive spillovers and mitigate siphon effects in lagging regions.

The paper shows that: (1) From 2000 to 2018, HSR introduction and faster regular rail substantially improved urban accessibility and locations. (2) Regional accessibility and labor market income exhibit strong positive spatial correlation; increasing accessibility generally raises income at the regional scale. (3) Neighboring cities with higher accessibility can impose a siphon effect on local areas. (4) Urban labor incomes are mutually reinforcing across cities, with positive spatial dependence. (5) After 2012, coinciding with major HSR expansion and labor market transformation, accessibility produced significant positive spatial spillovers on income. These results suggest leveraging the HSR-centered modern rail network to support balanced regional development, reduce inter-city income disparities, and strengthen national market integration. The work highlights the strategic role of transport infrastructure in overcoming regional gaps and contributing to high-quality growth.

The study assumes city latitude/longitude are time-invariant and applies linear interpolation for missing annual data, which may introduce measurement error. Results depend on the choice of spatial weights (adjacency and inverse distance), although robustness is assessed with a double-weight SDM. The accessibility measure relies on modeled minimum rail travel times (HSR where available, accelerated conventional rail otherwise), which abstracts from other transport modes and potential schedule variability.