A spatial econometric investigation into road traffic accessibility and economic growth: insights from the Chengdu-Chongqing twin-city economic circle

The study investigates how road traffic accessibility relates to and influences regional economic growth within the Chengdu-Chongqing twin-city economic circle. Motivated by the broader question of the contribution of transport infrastructure investment to local and regional development beyond travel-time savings, the authors examine 2019 conditions, integrating accessibility indicators into a spatial econometric framework. The purpose is to quantify accessibility’s role alongside economic, investment, and political-institutional factors and to identify spatial spillovers or error-term autocorrelation that could shape observed outcomes. The research is important for informing transport and regional development policy, given uneven development patterns and the potential for accessibility improvements to shape economic performance. The authors also acknowledge a potential bidirectional relationship between economic growth and transport infrastructure, raising endogeneity concerns that are left for future work.

The article situates its inquiry within transport economics and regional science literature linking infrastructure and growth. It notes prior recognition of a bidirectional relationship between transport infrastructure and economic growth (e.g., Hong et al., 2011). In discussing spatial patterns, it references studies showing accessibility’s dependence on administrative level and development degree (e.g., Liu et al., 2009; Bai et al., 2012) and the coupling between accessibility and economic agglomeration (e.g., Yin, 2014). A broader set of related works on transport investment, accessibility metrics, and spatial spillovers is listed in the references, indicating the relevance of spatial econometric approaches to capture interregional effects.

- Study area and period: Chengdu-Chongqing twin-city economic circle, with analysis focused on pre-2019 data to avoid COVID-19 confounding; main econometric analysis uses 2019 variables. Some spatial correlation diagnostics reference 2018 per capita GDP. - Variables: Dependent variable is GDP per capita (GDP_pi). Key explanatory variables include road traffic accessibility (Acc, measured in hours of travel time; lower values indicate better accessibility), urbanization level (Urb), industrial structure (Ind), labor force (Lab), education investment (Edu), and trade openness (Tra). - Spatial weights: A first-order Queen contiguity spatial weight matrix (W) is constructed, reflecting adjacency among county-level units. - Spatial dependence diagnostics: Moran’s I for per capita GDP (2018) is 0.210, significant at the 5% level, indicating positive spatial autocorrelation and motivating spatial econometric modeling. - Model specifications: • Spatial lag model (SAR): Y = ρWy + Xβ + ε, with specific form GDP_pi = ρW(GDP_pi) + β0 + β1Acc_i + β2Urb_i + β3Ind_i + β4Lab_i + β5Edu_i + β6Tra_i + ε_i. • Spatial error model (SEM): Y = Xβ + ε; ε = λWε + v, with specific form GDP_pi = β0 + β1Acc_i + β2Urb_i + β3Ind_i + β4Lab_i + β5Edu_i + β6Tra_i + ε_i. - Model selection and estimation: Lagrange Multiplier (LM) and robust LM tests guide selection between SAR and SEM. Due to spatial correlation, ordinary least squares is biased; maximum likelihood estimation (Anselin) is used for SAR/SEM. Model fit is compared using Log-Likelihood (LogL), Akaike Information Criterion (AIC), Schwarz Criterion (SC), and Likelihood Ratio (LR) tests. - Sample size: 42 spatial units (as per Table 3). - Accessibility analysis: Travel-time based accessibility indicators are computed, revealing a center-periphery pattern. Reported statistics include city-level minimum and maximum travel-time values and averages/standard deviations for Chongqing vs. Sichuan.

- Spatial patterns of accessibility: A pronounced core-periphery structure is observed. Central areas around Chengdu and Chongqing have better (lower travel-time) accessibility than peripheral areas. Suining has the lowest travel-time measure (3.449 h), while Wanzhou and Kaizhou have the highest. Chongqing’s average accessibility is 4.191 h (SD 0.647 h), worse and more variable than Sichuan’s 3.946 h (SD 0.361 h). Outer rings, especially in the northeast, exceed 5.001 h, indicating poor accessibility. - Spatial dependence: Moran’s I for per capita GDP (2018) equals 0.210 (significant at 5%), confirming spatial autocorrelation. - Model comparison (Table 3): The spatial error model (SEM) outperforms OLS and SAR with LogL = -360.631 (vs. OLS -362.992; SAR -362.986), AIC = 735.262 (lower than OLS 739.984; SAR 741.972), and SC = 746.346 (lower than OLS 751.069; SAR 754.641). LR test favors SEM (4.722, significant at 5%). - Spatial error dependence: The SEM’s spatial autocorrelation coefficient (λ) is 0.535, significant at the 1% level, indicating strong spatial dependence in the error term and proximity effects. - Coefficient estimates and significance (SEM): • Acc (accessibility, hours): -5006.560, significant at 10% – lower travel time (better accessibility) associates with higher GDP per capita. • Urb (urbanization rate, %): 315.642, significant at 10% – a 1% increase in urbanization raises GDP per capita by ¥315.642. • Ind (industrial structure, %): 2029.470, significant at 1% – a 1% improvement increases GDP per capita by ¥2,029.470. • Lab (labor force, % employed): 322,575.700, significant at 1% – a 1% rise in employment share increases GDP per capita by ¥322,575.700. • Edu (education investment): 108,469.000, significant at 5% – a ¥10,000 increase in per capita education expenditure increases GDP per capita by ¥108,469.000. • Tra (trade openness): not significant and with an unexpected sign relative to the initial hypothesis. - Overall, accessibility plays a pivotal role but manifests via spatial error dependence; urbanization, industrial structure, education, and labor force positively influence growth; trade openness shows no robust effect.

The findings align with a center-periphery development pattern in the Chengdu-Chongqing region: central cities benefit from earlier planning, stronger economies, and higher-quality highways, yielding better accessibility and higher economic performance. Economic agglomeration patterns mirror accessibility, suggesting coupling between transport access and economic outcomes. The negative coefficient on accessibility measured in hours means improved accessibility (lower travel times) is associated with higher per capita GDP, consistent with the notion that better transport facilitates mobility of people and goods and supports growth. Significant positive effects of urbanization, industrial structure optimization, education investment, and labor force emphasize the importance of complementary economic and human-capital conditions. The strong spatial error dependence indicates unobserved regional factors and spillovers across neighboring areas. Trade openness does not exhibit a significant effect in this setting, contrary to the initial expectation. Policy implications include targeting peripheral areas with transport and supportive economic policies to reduce disparities and leveraging central areas’ network advantages while improving the broader enabling environment.

This study integrates road traffic accessibility indicators into spatial econometric models to examine their relationship with economic growth in the Chengdu-Chongqing twin-city economic circle. The spatial error model provides the best fit, revealing significant spatial error dependence (λ ≈ 0.535). Accessibility improvements (reductions in travel time) are associated with higher per capita GDP, and urbanization, industrial structure, education investment, and labor force contribute positively to growth, while trade openness shows no significant effect. The work underscores the need to strengthen the link between road transportation and economic development and to enhance supporting economic and institutional environments. These insights inform transport and regional development planning in similar urban economic circles and suggest future research should address endogeneity and explore dynamic or instrumental-variable frameworks.

The authors note potential endogeneity due to the bidirectional relationship between economic growth and transport infrastructure; the current model does not instrument for this, which may bias estimates. To avoid COVID-19 confounding, analysis uses data prior to 2019, limiting temporal scope. Spatial diagnostics use 2018 GDP per capita, while main modeling focuses on 2019, and the cross-sectional design precludes dynamic inference.