首页|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

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

扫码查看
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.
Merger of Large and Small Cities and Economic Growth in Boundary Areas:Evaluation by Synthetic Control Based on Machine Learning Algorithms
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 fine-grained 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 resources in its boundary area near Laiwu,which led to a relatively low level of development in this area.Laiwu,with its small city size,tends to proactively approach Jinan to absorb the spillover effects of the economy from Jinan,which made boundary areas located in Laiwu more developed than expected.After the city merger,the administrative division effect is eliminated,and the Jinan boundary area has opportunities for rapid economic growth.Based on the results,this study puts forward three policy recommendations.First,as strengthening the merger between large and small cities helps to weaken the administrative barriers mentioned above,the central government should strengthen the coordination among local governments and encourage the governments of large and small cities to establish a coordination mechanism on their own initiative to promote the rational allocation of resources,thereby realizing the economic development of the areas along the administrative boundaries of cities.Second,the government should consider the level of economic growth and resource endowment of each city and formulate corresponding policies to promote coordinated regional development.Third,governments at the higher level must play a role in coordinating to encourage large cities to abandon the"beggar-thy-neighbor"mentality,open up the broken roads,deepen regional cooperation,and promote the synergistic development of the economies of cities of all sizes.

Administrative BoundariesSatellite Lighting DataSynthetic Control MethodMachine Learning

郭峰、吕斌、熊云军、陶旭辉

展开 >

上海财经大学公共经济与管理学院

西南财经大学经济学院

浙江工商大学公共管理学院

浙江工商大学统计数据工程技术与应用协同创新中心

展开 >

行政边界 卫星灯光数据 合成控制法 机器学习

2024

数量经济技术经济研究
数量经济与技术经济研究所

数量经济技术经济研究

CSTPCDCSSCICHSSCD北大核心
影响因子:1.069
ISSN:1000-3894
年,卷(期):2024.41(9)