大小城市合并与行政边界地区经济增长:基于机器学习算法的合成控制评估  

Merger of Large and Small Cities and Economic Growth in Boundary Areas:Evaluation by Synthetic Control Based on Machine Learning Algorithms

作　　者：郭峰[1] 吕斌 熊云军 陶旭辉 GUO Feng;LV Bin;XIONG Yunjun;TAO Xuhui(School of Public Economics and Management,Shanghai University of Finance and Economics;School of Economics,Southwest University of Finance and Economics;School of Public Administration,Zhejiang Gongshang University;Collaborative Innovation Center of Statistical Data Engineering Technology&Application,Zhejiang Gongshang University)

机构地区：[1]上海财经大学公共经济与管理学院 [2]西南财经大学经济学院 [3]浙江工商大学公共管理学院 [4]浙江工商大学统计数据工程技术与应用协同创新中心

出　　处：《数量经济技术经济研究》2024年第9期26-48,共23页Journal of Quantitative & Technological Economics

基　　金：上海市哲学社会科学规划课题(2023ZJB006);浙江省教育厅科研项目(Y202353407)的资助。

摘　　要：城市行政边界地区在经济发展中处于弱势地位,行政区划调整则为边界地区经济增长带来新的契机,但政策效应在边界两侧可能是不对称的。本文以“莱芜并入济南”作为一次自然实验,利用细颗粒度的卫星灯光数据和基于机器学习算法的合成控制法,评估了大、小两个城市合并对原边界地区经济增长产生的影响。结果显示,济南莱芜合并对原边界地区经济增长产生正向影响,边界地区灯光亮度的实际值比基于机器学习算法预测的反事实合成值平均高出10.8%。这一结论在更换了边界地区度量方式、对照组选取方式、机器学习算法等后依旧保持稳健。进一步研究还发现,城市合并的正向效应主要体现在核心城市济南市边界一侧,而被合并的小城市莱芜市边界一侧则受益甚微。另外,本文也通过预留样本、考察其他案例的方法验证了核心结论的外部有效性,表明基于机器学习算法的合成控制法有较为广泛的适用性。The slow economic growth in areas located on administrative boundaries due to factors such as the division of administrative power and local protectionism is one of the major obstacles to achieving the great circulation of the national economy and the high-quality development of the regional economy.Adjusting administrative divisions makes it possible to accelerate the economic development of administrative boundary areas.At present,there have been many studies on changing counties into districts,the merger of two districts,and the establishment of development zones,but there has been a lack of detailed research on whether the adjustment of administrative divisions,especially the merger of small cities by large cities,can reduce the poverty of the boundary areas.Using the Laiwu merger into Jinan as a natural experiment,this study investigates the impact of the small cities merged into large cities on the economic growth in the original boundary area using finegrained satellite lighting data and a synthetic control method based on machine learning.The results reveal that the merger of Jinan and Laiwu has had a positive impact on the economic growth of the boundary area—the actual value of the light brightness of the boundary area is on average 10.8%higher than the predicted counterfactual synthetic value.The conclusions remain robust after replacing the boundary area metric,changing the control group,and using the machine learning method.Further,the study examines the treatment heterogeneity effect on both sides of the boundary located in large and small cities.The results reveal that the positive effects of mergers are mainly on the boundary located in the core city of Jinan,while the boundary located in the small city of Laiwu benefits very little from mergers.The likely reason for this is the competition for economic growth between local governments in China,which relies on the tournament model of relative ordering at the regional level.Before the merger,the Jinan government lacked incentives to invest more r

关 键 词：行政边界 卫星灯光数据 合成控制法 机器学习 

分 类 号：F224.0[经济管理—国民经济]