要素禀赋、地方工业行业发展与行业选择

Factor Endowment, Local Industrial Development and Selecting Industries

本文利用中国省际层面和行业层面的26种投入要素,使用机器学习方法揭示地方工业行业的非充分发展和受要素约束状态,进而基于各地区要素禀赋特征揭示各地区代表性工业行业的选择和发展路径。研究发现:(1)各地区工业行业发展具有明显的地区差异性。东部省份有较多的充分发展行业和要素约束行业,东北三省和中西部地区有较多的非充分发展行业。对于非充分发展的工业行业,其发展路径是提高要素的使用效率或改善要素结构;对于充分发展和要素约束的行业,其发展路径是增加要素投入和改进生产技术。(2)影响不同行业发展的最重要投入要素不尽相同,但整体而言,就业人数、城市工业用地、科技禀赋、物质资本、高速公路里程数是工业行业最重要的投入要素。影响先进制造业最重要的投入要素是就业量、城市工业用地、科技禀赋、物质资本、国内专利申请授权数。(3)不同先进制造业存在不同的要素组合最优区间,各地区应根据自身投入要素处于最优区间的位置而选择优先发展行业和发展路径。广东、江苏、浙江、山东等省份较多的重要投入要素已进入先进制造业的最优组合区间,可以方便地依据本省特征选择发展路径。贵州、云南等西部欠发达地区先进制造业的重要投入要素都远离要素组合最优区间,这些地区发展先进制造业将相对困难。

Supply-side industrial reform is limited by the abundance and structure of factor endowments.The key to China s transformation of its industrial structure and development from a large industrialized country to a powerful one lies in breaking through the constraints of factor endowments.This paper comprehensively considers the characteristics of factor endowments in every region and industry and uses machine learning methods to reveal whether regional industries are fully developed from the perspective of factor endowments.Furthermore,based on the structure of factor endowments,the development paths of industries in each region are explored.This paper provides not only a new perspective on supply-side industrial reform but also a basis for regional industrial development and selection.Despite researchers introduction of more factor endowments that affect output in recent years,single(multiple)regression equations and VAR models used in the literature obviously limit the number of independent variables(Debaere&Demiroglu,2003;He,2014).To address these deficiencies,this paper uses the regression trees model in machine learning,as it allows for the inclusion of more independent variables,helps avoid possible multicollinearity,decreases the degrees of freedom of traditional econometric models,and relieves endogeneity from omitted variables.This paper offers three main contributions.First,it offers a new perspective to the research.Because of its critical practical significance,supply-side structural reform and the full development of industries have been widely discussed in academia and the government.However,no empirical studies concerning partially developed regional industries and industrial selection have yet been carried out from the viewpoint of factor endowments.Second,using a machine learning model,this paper considers abundant input factors,thereby avoiding the limitation in previous literature of selecting only a few main factors and omitting other important ones.Finally,from the view of research findings,partially developed industries and factor endowment constrained industries in each region are demonstrated,and the selection of advanced manufacturing industries and optimized paths of the factor structure are revealed.These research conclusions are different from those in the literature.Based on economic growth theory and the Heckscher-Ohlin-Vanek model and following the research of Feenstra(2011),this paper selects 26 input factors from natural resources,human resources,physical capital,science and technology,and economic structures and institutions,and collects data from different provinces and industries from 2006 to 2016.Using the random forest algorithm,the following is found.(1)Industrial development differs in each region.Eastern Chinese provinces have more fully developed and factor-constrained industries,whereas the northeast,center,and western provinces have more industries that are not fully developed.For these partially developed industries,development lies in enhancing the efficiency of factor usage or improving factor structure.For fully developed and factor-constrained industries,the development path is through increasing factor inputs and improving production technology.(2)The most important input factors affecting development differ between regions.As a whole,employment,urban industrial land,science and technology,capital stock,and highway mileage are the most important.For advanced manufacturing industries,employment,urban industrial land,science and technology,physical capital,and the number of authorized domestic patent applications are the most important factors.(3)As the optimal combination of factor endowments differs between advanced manufacturing industries,regions must prioritize industries and development paths according to the position of their input factors in the optimal range.With a number of important input factors in the optimal combination of advanced manufacturing industries,Guangdong,Jiangsu,Zhejiang,and Shandong provinces can easily choose a development path based on their factor characteristics.