研发补贴引致作用研究的Meta分析——异质性、发表偏倚与真实效应

A Meta Regression Analysis on the Effect of Public Subsidy on Enterprises’R&D:Heterogeneity,Publication Bias and Genuine Empirical Effect

中国研发补贴政策能否引致企业研发投入和产出增加,在理论上存在较大争论,实证研究也难以得到一致性的结论。本文收集了中国产业补贴与企业研发关系的122篇中英文实证文献,采用Meta回归分析方法对研究异质性因素进行分解、剔除发表偏倚得到了真实效应值区间。研究发现,异质性来自于样本选择、方法使用以及发表偏倚。在以企业私人研发支出作为被解释变量的研究中,发表偏倚存在30%~50%的高估效应。在研发专利技术作为产出指标的研究中,剥离发表偏倚和实证方法产生的影响,政府补贴对研发产出的影响也是微弱的。剥离发表偏倚后的真实效应值仅支持以研发总支出作为被解释变量的实证效果。细分行业、细分补贴形式的案例研究在未来将更有意义。

It is controversial in both theory and evaluation literature that public subsidies in China can motivate enterprises to invest more in their R&D activities.This paper uses Meta Regression Analysis(MRA)model to analyze 122 evaluation studies and finds out genuine empirical effects after controlling study heterogeneity and publication bias.Studies with additional private R&D fund as dependent variable show a 30%-50%over-valuation is caused by publication bias,and reveal no evidence to support positive conclusion in studies with patent technology as dependent variable.In future,case study on specific policy and representative industry will be more meaningful.