摘要
基于沪深A股上市公司的样本数据,将2018年加计扣除比例提升的政策激励作为一项准自然实验,通过构建双重差分模型进行实证分析,从创新投入、创新产出以及创新效率三个角度研究探索提高加计扣除比例的政策效果。实证结果表明,加计扣除比例的提高,不但能够促进企业加大其创新投入,增加企业创新产出数量,还对企业的创新效率起到了显著的促进作用。在作用机制方面,融资约束充当起了中介桥梁作用。同时发现,加计扣除比例的提高更能激励国有企业、成长期企业以及非高新技术企业的创新产出;并且该政策在不同地区所产生的政策效应也不同,表现为在创新环境较好、财政政策扶持力度大以及执法力度大的地区,加计扣除比例的提高所激发的创新激励作用更为显著。基于此,建议应深化研发费用加计扣除政策改革、优化配套政策、加强法制保护。
Based on the sample data of Shanghai and Shenzhen A-share listed companies,the policy impact of the increased deduction ratio in 2018 is taken as a quasi-natural experiment,and empirical analysis is carried out by constructing a double differential model,from the perspective of innovation input,innovation output and innovation efficiency three angles of study to explore the policy effect of increasing the percentage of super deduction.The empirical results show that the increase in the percentage of super deduction can not only promote enterprises to increase their innovation input and increase the number of innovation output of enterprises,but also play a significant role in promoting the innovation efficiency of enterprises.In terms of mechanism,financing constraints act as an intermediary bridge.At the same time,it is found that the increase in the percentage of super deduction can more stimulate the innovation output of state-owned enterprises,growth-stage enterprises,and non-high-tech enterprises;and the policy effects produced by this policy in different regions are also different,which is reflected in a better innovation environment,In areas where fiscal policy support is strong and law enforcement is strong,the increase in the percentage of super deductions stimulates more significant innovation incentives.Based on this,it is recommended to deepen the policy reform of R&D expenses deduction,optimize supporting policies,and strengthen legal protection.
作者
石绍宾
李敏
Shi Shaobin;Li Min
出处
《公共财政研究》
2021年第3期4-28,共25页
Public Finance Research Journal
关键词
研发费用加计扣除
企业创新
双重差分模型
中介效应
Pre-Tax Super Deduction of Research and Development(R&D)Expense
Enterprise Innovation
Dual Difference Model
Mediation Effect
