Abstract:

Industrial policy is an important policy tool for the government to guide the green innovation strategy, accelerate the supply-side structural reform, and promote the adjustment and upgrading of industrial structure. How to improve the total factor productivity of enterprises through industrial policy optimization is an important issue in the construction of a modern economic system. Combining traditional econometric and machine learning methods, this paper systematically explores the effects and trends of industrial policy choices on enterprise total factor productivity from policy content, policy objectives, and policy means. An empirical research of 1238 listed companies from 2007 to 2020 shows that differences in industrial policy content, objectives, and means have different impacts on the total factor productivity of enterprises. Public services, capital investment, target planning, the introduction of policies such as financial support and consumption subsidies can promote the total factor productivity of enterprises, while the introduction of management policies for overseas institutions will inhibit the improvement of total factor production of enterprises. The setting of output targets in policy objectives can inhibit the total factor productivity of enterprises. The energy efficiency target and labor productivity target can promote the total factor productivity of the enterprise. Tax incentives and government subsidies in policy measures can both promote the total factor productivity of the enterprise. A comparison of the linear regression and machine learning methods indicates that the conclusions of the two are basically the same, but the goodness of fit and mean square error of the machine learning method are better than linear regression, which helps to improve the accuracy of the analysis of industrial policy effects and is applicable to industrial policy selection and decision-making scenarios. This paper can provide literature support and analysis tools for improving the intelligent level of government policy decision-making and strengthening government governance capabilities.

Key words:

industrial policy, total factor productivity of enterprises, text analysis, machine learning

摘要：

全要素生产率是经济增长可持续性的重要表征，是经济高质量发展的动力源泉。提升全要素生产率对开启全面建设社会主义现代化国家新征程有着重要的战略意义。产业政策是政府引导绿色创新战略、加快推进供给侧结构性改革、促进产业结构调整升级的重要政策工具，如何通过优化产业政策来提高企业全要素生产率是现代化产业体系建设的重要课题。结合传统计量与机器学习方法，从政策内容、政策目标和政策手段3个角度，系统地探讨产业政策选择对企业全要素生产率的影响效应。基于2007~2020年1 238家上市公司的实证研究表明，产业政策内容、目标和手段的差异对企业全要素生产率有着不同影响。研究发现：政策内容中公共服务、资金投入、目标规划、金融支持、消费补贴等政策的出台能够促进企业全要素生产率，海外机构管理政策的出台则抑制企业全要素生产率的提高；政策目标中产出目标设定对企业全要素生产率有抑制作用，而能效目标和劳动生产率目标则对企业的全要素生产率有促进作用；政策手段中税收优惠和政府补贴对企业全要素生产率均有促进作用。从重要性排序看，政策手段作为对企业的直接干预，对企业全要素生产率的影响排在前列，其中，税收优惠对企业全要素生产率的影响最大，政策补贴次之；再次是目标规划政策的出台，进一步肯定了中国政府制定的发展规划对企业发展的重要影响；随后是公共服务，说明政府对公共服务的重视程度对企业全要素生产率也有着重要影响；排在第5、第6位的是消费补贴和资金投入，随后是海外机构管理；排名第8~10位的依次是能效目标、劳动生产率目标和产出目标。对比线性回归和机器学习方法，两者结论基本一致，但机器学习方法的拟合优度、均方差误均优于线性回归，有助于提高产业政策效应分析的准确性，其适用于产业政策选择决策场景。本研究可为提升政府政策决策智能化水平、强化政府治理能力提供文献支撑和分析工具。

关键词:

产业政策, 企业全要素生产率, 文本分析, 机器学习