| 9,597 | 105 | 144 |
| 下载次数 | 被引频次 | 阅读次数 |
本文以“大气十条”政策为例，使用双重差分法和机器学习方法来探讨命令控制型环境规制工具对企业绿色创新水平的影响。基于2007—2017年上市公司绿色专利数据，双重差分法的回归结果显示，“大气十条”政策的实施对企业绿色专利总水平、绿色发明专利水平和绿色实用专利水平均有显著的促进作用；在采用PSM-DID、面板Tobit等方法进行一系列稳健性检验后，该结论依然成立。文中进一步采用广义随机森林模型中因果森林进行异质性分析。在构建森林的过程中发现，企业规模和企业性质是影响绿色创新最重要的两个分裂变量。通过因果森林对连续变量企业规模的异质性分析发现，在“大气十条”政策的影响下，企业规模与绿色专利总水平、绿色实用专利水平先呈“倒U型”关系再呈“U型”关系，与绿色发明专利水平呈“倒U型”关系。这一结果表明，在进行绿色实用专利研发上，小企业和大企业各自发挥了自身的比较优势；在进行绿色发明专利研发上，规模较大的企业激励相对不足。进一步分析表明，为了提高绿色创新水平，企业会加大研发资金和人员的投入，绿色创新的实现可以降低企业的单位排污成本、提高企业的竞争力。本文的研究结果为“波特假说”提供了新的经验证据，为推动我国绿色创新发展提供富有针对性实证参考。
Abstract：Regarding the implementation of Action Plan of Air Pollution Prevention and Control(APAPPC) as a quasi-natural experiment, this paper employs difference-in-differences(DID) method and the machine learning method to explore the impact of command-and-control environmental regulation tools on enterprise green innovation. Based on the green patent data of listed companies from 2007 to 2017, the difference-in-differences estimations show that the implementation of the APAPPC has significantly promoted the level of the total green patents, the green invention patents, and the green utility patents. The conclusion still holds after using the PSM-DID model and panel Tobit model. This paper further uses causal forests in the generalized random forest model to test heterogeneity. In the process of growing the forests, the research finds that scale and ownership of enterprises are the two most important splitting variables affecting green innovation. The heterogeneity analysis on enterprise scale suggests that with the impact of the APAPPC, the relationship between enterprise scale and the proportion of the total green patents, the proportion of the green utility patents is initially inverted U-shaped, and then turns U-shaped, while the relationship between enterprise scale and the proportion of the green invention patents is inverted U-shaped. This result shows that enterprises with small and large scale have their comparative advantages in the research and development of green utility patents, and large-scale enterprises lack motivations in the development of green invention patents compared with small-scale enterprises. Further analysis suggests that enterprises may increase R&D investment to improve green innovation capability, and the realization of green innovation contributes to reducing the unit emission cost and enhancing enterprises' competitiveness. The results of this paper provide new empirical evidence for the “Porter Hypothesis”, and help to form some suggestions for promoting green innovation in China.
[1]范丹,叶昱圻,王维国.空气污染治理与公众健康——来自“大气十条”政策的证据[J].统计研究, 2021, 38(9):60?74.
[2]黄乃静,于明哲.机器学习对经济学研究的影响研究进展[J].经济学动态, 2018(7):115?129.
[3]吉赟,杨青.高铁开通能否促进企业创新:基于准自然实验的研究[J].世界经济, 2020, 43(2):147?166.
[4]寇宗来,刘学悦.中国企业的专利行为:特征事实以及来自创新政策的影响[J].经济研究, 2020, 55(3):83?99.
[5]李树,陈刚.环境管制与生产率增长——以APPCL2000的修订为例[J].经济研究, 2013, 48(1):17?31.
[6]刘生龙,张晓明,杨竺松.互联网使用对农村居民收入的影响[J].数量经济技术经济研究, 2021, 38(4):103?119.
[7]罗知,李浩然.“大气十条”政策的实施对空气质量的影响[J].中国工业经济, 2018(9):136?154.
[8]王馨,王营.绿色信贷政策增进绿色创新研究[J].管理世界, 2021, 37(6):173?188, 11.
[9] Athey S, Imbens G. Recursive Partitioning for Heterogeneous Causal Effects[J]. Proceedings of the National Academy of Sciences of the United States of America, 2016, 113(27):7353?7360.
[10] Athey S, Tibshirani J, Wager S. Generalized Random Forests[J]. The Annals of Statistics, 2019, 47(2):1148?1178.
[11] Athey S, Wager S. Estimating Treatment Effects with Causal Forests:An Application[J]. Observational Studies, 2019, 5(2):37?51.
[12] Cai S, Wang Y, Zhao B, et al. The Impact of the “Air Pollution Prevention and Control Action Plan” on PM2.5 Concentrations in Jing-Jin-Ji Region during 2012—2020[J]. Science of The Total Environment, 2017, 580:197?209.
[13] Davis J M V, Heller S B. Using Causal Forests to Predict Treatment Heterogeneity:An Application to Summer Jobs[J]. American Economic Review, 2017, 107(5):546?550.
[14] Feng Y, Ning M, Lei Y, et al. Defending Blue Sky in China:Effectiveness of the “Air Pollution Prevention and Control Action Plan” on Air Quality Improvements from 2013 to 2017[J]. Journal of Environmental Management, 2019, 252:UNSP109603.
[15] Gilder G. The Revitalization of Everything:The Law of the Microcosm[J]. Harvard Business Review, 1988, 66(2):49–61.
[16] Gulyas A, Pytka K. Understanding the Sources of Earnings Losses After Job Displacement:A Machine-learning Approach[C]. Beitr?ge zur Jahrestagung des Vereins für Socialpolitik 2021:Climate Economics, ZBW-Leibniz Information Centre for Economics, Kiel, Hamburg.
[17] Huang J, Pan X, Guo X, et al. Health Impact of China's Air Pollution Prevention and Control Action Plan:An Analysis of National Air Quality Monitoring and Mortality Data[J]. The Lancet Planetary Health, 2018, 2(7):e313?e323.
[18] Knittel C R, Stolper S. Machine Learning about Treatment Effect Heterogeneity:The Case of Household Energy Use[J]. AEA Papers and Proceedings, 2021, 111:440?444.
[19] Li P, Lu Y, Wang J. Does Flattening Government Improve Economic Performance? Evidence from China[J]. Journal of Development Economics,2016, 123:18?37.
[20] Pavitt K, Robson M, Townsend J. The Size Distribution of Innovating Firms in the UK:1945—1983[J]. The Journal of Industrial Economics,1987, 35(3):297?316.
[21] Scherer F M. Size of Firm, Oligopoly, and Research:A Comment[J]. The Canadian Journal of Economics and Political Science, 1965, 31(2):256?266.
[22] Wager S, Athey S. Estimation and Inference of Heterogeneous Treatment Effects Using Random Forests[J]. Journal of the American Statistical Association, 2018, 113(523):1228?1242.
[23] Yin X, Zuscovitch E. Is Firm Size Conducive to R&D Choice? A Strategic Analysis of Product and Process Innovations[J]. Journal of Economic Behavior&Organization, 1998, 35(2):243?262.
[24] Zhang J, Jiang H, Zhang W, et al. Cost-benefit Analysis of China's Action Plan for Air Pollution Prevention and Control[J]. Frontiers of Engineering Management, 2019, 6(4):524?537.
[25] Zhang X, Xu X, Ding Y, et al. The Impact of Meteorological Changes from 2013 to 2017 on PM2.5 Mass Reduction in Key Regions in China[J].Science China Earth Sciences, 2019, 62(12):1885?1902.
(1)具体目标为到2017年，全国地级及以上城市可吸入颗粒物浓度比2012年下降10%以上，优良天数逐年提高；京津冀、长三角、珠三角等区域细颗粒物浓度分别下降25%、20%、15%左右，北京市细颗粒物年均浓度控制在60微克/立方米左右。
(1)本文从计量经济学中条件均值、潜在结果框架下的因果效应等基本概念出发，逐步介绍作为机器学习方法的随机森林模型如何被应用于估计处理效应。因篇幅所限，该部分内容以附录1展示，见《统计研究》网站所列附件。下同。
(2)本文绿色专利范围包含了能源节约类、替代能源生产类、废弃物管理类三类。根据《联合国气候变化框架公约》推出的检索环境友好型技术专利的在线工具，纳入绿色专利范围的包括7类：能源节约类、替代能源生产类、废弃物管理类、交通运输类、农林类、核电类和行政监管与设计类。本文也考虑了将这7类作为绿色创新活动的衡量范围，结论一致，结果留存备索。
(1)因篇幅所限，描述性统计结果以附表1展示。
(1)2007—2012年各个工业行业单位产值废气排放量是0.49亿立方米/亿元，超过均值的行业视为空气污染密集行业，具体包括的行业：08 黑色金属矿采选业，10非金属矿采选业，22造纸及纸制品业，25石油加工、炼焦及核燃料加工业，26化学原料及化学制品制造业，28化学纤维制造业，31非金属矿物制品业，32黑色金属冶炼及压延加工业，33有色金属冶炼及压延加工业，44电力、热力的生产和供应业。
(2)“波特假说”认为，适宜的环境规制可以引导企业进行创新活动，这在短期内可能会增加生产成本，长期则会提高企业生产效率、增强企业市场竞争力。
(1)因篇幅所限，绿色专利平行趋势以附图1展示。
(2)因篇幅所限，回归结果以附表2展示。
(1)因篇幅所限，安慰剂检验以附图2展示。
(2)当因变量分别是绿色专利总水平、绿色发明专利水平和绿色实用专利水平时，安慰剂检验中系数估计值的均值分别是-0.0004、-0.0012和-0.0007，远小于表2中的真实估计值。
(3)因篇幅所限，匹配平衡性检验结果以附表3展示。
(1)当企业规模大于25时，曲线走势稍微发生变化，考虑到分布的样本点较少，不作详细讨论。
(1)因篇幅所限，回归结果以附表4展示。
(1)为增强系数可读性，文中将排污费与营业收入的比值乘以10进行回归。
基本信息:
DOI：10.19343/j.cnki.11-1302/c.2022.12.004
中图分类号:X32;F273.1;F832.51
引用信息:
[1]唐礼智,周林,杨梦俊.环境规制与企业绿色创新*——基于“大气十条”政策的实证研究[J].统计研究,2022,39(12):55-68.DOI:10.19343/j.cnki.11-1302/c.2022.12.004.
基金信息:
国家社会科学基金一般项目“连续处理效应的异质性分析及其在政策评估中应用研究”（21BTJ037）