中国的高速公路、市场准入与产业发展: 基于城墙面板最小生成树工具变量的研究
Expressways, market access, and industrial development in China: Using walled-city panel instrumental variables of minimum spanning tree
Journal of Development Economics · 2025 · [{"name": "Kecen Jing", "affiliation": ["Macau University of Science and Technology"]}, {"name": "Wen‐Chi Liao", "affiliation": ["National University of Singapore", "Massachusetts Institute of Technology"]}]
中文摘要
中国等新兴经济体大力投资高速公路建设，以改善市场可达性。这种交通网络扩张可能影响区域工业发展，但因果推断颇具挑战性。现有研究通常仅关注较长期效应，或将经济上重要的地区排除在外，缺乏一种既能应对快速交通发展、又能覆盖全国整体经济地理的识别策略。为填补这一空白，我们引入准随机的“城墙城市”最小生成树面板工具变量，对市场可达性进行工具变量估计，并考察中国快速扩张的高速公路网络如何影响制造业发展。这些工具变量通过模拟假想网络的动态扩张路径生成，这些网络最终连接由1820年城墙特征所预测的全部目标地点。研究从静态和动态两方面验证了工具变量的正交性，以及纳入限制和排他性限制。研究在截面工具变量与面板工具变量之间、以及使用与不使用历史城墙数据的工具变量之间进行了比较检验。当交通发展与经济增长均十分迅速时，面板工具变量策略尤为有效。无需排除任何地点。县级层面的分析表明，2000—2009年间，市场可达性每提高1%，制造业GDP和企业数量分别增加0.28%和0.07%；制造业企业的平均产出、资本和劳动分别增加0.08%、0.06%和0.03%。该过程还对区域分权、产业升级以及制造业的地理集中产生了影响。本文还估计了贸易弹性。文中构建的市场可达性模型推导出可估计方程，并在居住用地和工业用地供给无弹性、资本与劳动力可自由流动的假设下，对其经济机制进行了合理化说明。
Abstract
Emerging countries like China have heavily invested in motorways to improve market access. Such transport network expansion can affect regional industrial development, but causal inference is challenging. Existing studies typically focus on longer-term effects or exclude economically important locations, lacking an identification strategy that can address rapid transport development and cover the country's entire economic geography. To fill this gap, we introduce quasi-random Walled-City MST Panel IVs to instrument market access and examine how China's rapidly expanding expressway network affects manufacturing development. The IVs are generated by simulating dynamic expansion paths of hypothetical networks that ultimately connect all target locations predicted by 1820 city-wall characteristics. Research verifies IV orthogonality and inclusion and exclusion restrictions in both static and dynamic aspects. It conducts horse races between cross-sectional and panel IVs, and between IVs with or without historical city-wall data. The panel IV strategy is useful when both transportation development and economic growth are rapid. Excluding locations is unnecessary. County-level analysis shows that during 2000–2009, when market access improved by 1 %, manufacturing GDP and firm count increased by 0.28 % and 0.07 %. Manufacturers' average output, capital, and labor increased by 0.08 %, 0.06 %, and 0.03 %. There were impacts on regional decentralization, industrial upgrade, and geographic concentration of manufacturers. Trade elasticity is estimated. A model of market access derives estimable equations and rationalizes the economics, with the assumptions of inelastic residential and industrial land supply and mobile capital and labor.