Abstract: Research and development manipulation often occurs in the application of high-tech enterprises. Therefore, a stochastic evolutionary game model of local governments, tripartite declaration enterprises, and intermediary agencies was built to study this phenomenon. The Ito ? stochastic differential equation theory was used to analyze the conditions of game subject behavior stability while computer simulation was used to analyze the directionality and degree of the choice of key variables of the game subject. It is found that as the taxes paid by enterprises are reduced, the probability of speculation of tripartite declaration enterprises for research and development manipulation is reduced, which confirms the positive impact of tax reduction on improving the authenticity of enterprise innovation. The probability of successful research and development manipulation of intermediaries is inversely proportional to the probability of research and development manipulative behavior selection. A low probability of successful manipulation of research and development will increase the willingness of intermediaries to conduct research and development. The research shows that the change in the amount of tax paid by enterprises before the approval of high-tech enterprises and the tax reduction ratio after approval of high-tech enterprises have a difference in direction and degree on the strategic choice of the game subject. Besides, it is confirmed that the improvement in local government innovation performance will weaken its enthusiasm for the supervision of research and development manipulation. In addition, judging from the perspective of regulatory efficiency, the measures taken by the government to increase the cost of research and development and manipulation by enterprises alone are more effective than to increase the cost of intermediaries to manipulate research and development. This paper not only provides practical evidence for corporate taxation, government innovation performance and research and development manipulation, but also provides a theoretical reference for better formulation and implementation of corporate research and development incentives and effective supervision.

Key words: research and development manipulation, random environment, multi-party game, effective regulation

摘要： 围绕高新技术企业申报中存在的研发操纵现象，构建地方政府、申报企业和中介机构的三方随机演化博弈模型，借鉴 随机微分方程理论分析博弈主体行为稳定性的条件，并运用计算机仿真分析了各关键变量对博弈主体策略选择的方向性及程度性影响。研究发现：一方面，企业上缴税收降低，申报企业研发操纵投机概率降低，这确认了企业减费降税对于提高企业创新真实性的积极影响；另一方面，中介机构的研发操纵成功概率与研发操纵行为选择概率成反比，低研发操纵成功概率反而会增强中介机构的研发操纵意愿。研究表明：企业在未获批高新技术企业前上缴税收的额度变化和获批高新技术企业后的减税比例，对博弈主体的策略选择影响均呈现方向性和程度性差异；进一步，确认了地方政府创新绩效提升会削弱其对研发操纵行为监管的积极性。此外，从监管效率看，政府采取举措提高企业单独进行研发操纵成本比提高中介进行研发操纵成本效果更显著。

关键词: 研发操纵, 随机环境, 多方博弈, 有效监管