Abstract: The scale and intensity of a country’s R&D activities reflect a country’s scientific and technological strength and core competitiveness at the macro level, and reflect the company’s innovation capabilities and development speed at the micro level. Therefore, more and more countries have begun to attach importance to R&D investment, and are rushing to introduce various policies to promote enterprise R&D and increase the proportion of R&D. In April 2008, our government promulgated the " Administrative Measures for Determination of High and New tech Enterprises". The recognized high and new tech enterprises can receive tax reductions and exemptions, the tax rate is reduced by 10 points, and the tax amount is reduced by 40%. In addition, these companies also enjoy many preferential policies such as direct rewards, brand enhancement, and asset depreciation. But if a company want to get this title, he need some indicators to exceed the prescribed threshold, the most important of which is the intensity of research and development. Manufacturing enterprises can meet the policy threshold through R&D manipulation and obtain policy preferences such as tax relief, but deliberate R&D manipulation does not necessarily lead to the improvement of enterprise productivity. Based on the micro data of Chinese enterprises from 2008 to 2018, the nonparametric quantile model based on B-spline basis function expansion can capture more truly the non-linear heterogeneous effect of R&D intensity on enterprise productivity. The results show that: (1) R&D manipulation has a nonlinear effect on enterprise productivity, that is, with the continuous increase of R&D manipulation, It will inhibit the high-quality development of enterprises. (2) R&D manipulation has significant heterogeneous effects at different quantiles, and the sensitivity to R&D manipulation is higher at the low quantile, that is, the lower the enterprise productivity level, the more significant the negative impact of R&D manipulation. (3) Non state owned enterprises are more motivated to carry out R&D manipulation, and have a more significant negative impact on productivity. In addition, R&D manipulation has different effects on the productivity of enterprises in different regions or technology types, especially on high-tech types and central and western regions. (4) R&D manipulation has a significant lag effect on enterprise productivity. Chinese enterprises should not blindly manipulate R&D, but also closely integrate with their own enterprise productivity. It can provide decision-making basis for improving enterprise productivity by selecting reasonable R &D intensity.