利用2015-2020年A股上市公司数据，实证检验了加计扣除优惠对研发支出会计政策选择的影响。通过二次项函数模型，研究发现：加计扣除优惠与研发支出费用化比例呈倒U型关系，其中企业关注重点从税收规避逐步到盈余管理，是加计扣除优惠对研发支出费用化比例的正向影响开始降低甚至可能为负的一个重要因素。进一步检验发现，加计扣除优惠与研发支出费用化比例的倒U型关系由管理者操纵所致，且这种操纵行为集中在内部代理成本更高、管理者权力更大的企业中。本文建议，政府应提高加计扣除优惠资格的审核质量和研发信息操纵的违规成本，同时依据企业内部治理水平实施差别化的加计扣除政策。

   In order to encourage enterprises to participate in R&D activities, the Chinese government has been increasing the additional deduction in recent years, first increasing the percentage of additional deduction for technology-based SMEs from the original policy of 50% to 75% in May 2017, then to 75% for all enterprises in September 2018 as well, and further to 100% for manufacturing enterprises in March 2021. It has been shown that the bonus deduction concessions have a significant incentive effect on enterprises′ R&D activities (Ren and Song, 2017; Bao and Zhao, 2019; Liu et al., 2021; Liu and Lin, 2021), but they also induce a large number of low-quality enterprises to inflate their R&D expenditures and fraudulently obtain tax incentives through accounting account adjustments and actual business manipulation, resulting in a distorted incentive effect of the policy. In fact, compared with the manipulation of the total amount of R&D expenditure, the implicit space for the choice of accounting policy for R&D expenditure is greater and the supply of information is less, thus the manipulation behavior existing in it may be more common. According to our statistics, the average expense ratio of R&D expenditure is as high as 83.3%, which is obviously contrary to the fact that the number of patents of listed enterprises has increased rapidly in recent years. For example, Quanchai Power (600218) has expended all the R&D expenditures of 662 million yuan accumulated in past five years, while the company applied for 411 patents in these years, including 116 applications of invention patents with high innovation quality. Thus, we focus on the question: How does the additional deduction policy affect the accounting policy selection strategy for R&D expenditure?There has been a wealth of research on the effects of implementation of the policy of additional deduction, mostly qualitative research on the implementation of the policy, problems and countermeasures in the early years, and a large number of quantitative studies in recent years, but the conclusion of whether there is an incentive effect of the policy implementation is still controversial. The main reason for the controversy is that while incentive′s enterprise innovation, the bonus deduction preference may also induce adverse selection behavior of low-quality enterprises, leading to uncertainty in the effect of policy implementation. However, prior research has only focused on the manipulation of the total R&D expenditures induced by the additional deduction, while ignoring the manipulation in the R&D expenditure accounting policy selection. In fact, a large number of research has explored the adverse selection behavior in the R&D expenditure accounting policy selection from the perspectives of earnings management motivations or tax avoidance motivations, respectively. However, the scenarios of these research cannot incorporate both types of motives into the analytical framework. In view of the shortcomings of existing research, we analyzed the impact of the additional deduction on R&D expenditure accounting policy selection by considering both earnings management and tax avoidance motivation. Using a quadratic term design and 5943 observations of A-share listed enterprises during the period 2015-2020, we found that the relationship between add-on deduction preferences and the proportion of R&D expenditure expended is inverted U-shaped, in which the gradual shift of corporate focus from tax avoidance to surplus management is an important factor for the positive impact of add-on deduction preferences on the proportion of R&D expenditure expensed to begin to decrease or possibly even become negative. Further, we showed that the inverted U-shaped relationship is due to management manipulations, and these manipulations were concentrated in the enterprises with higher agency costs and more powerful managers.Our findings have two research contributions: (1) it provides a new interpretation perspective for the research topic of "why the incentive effect of additional deduction is distorted". There are controversies about whether the additional deduction has incentive effect, but it is more important to answer the question of why the policy incentive effect is distorted. Prior research tries to explain the manipulative behavior from the total R&D expenditures (Wan et al., 2020; Chen and Lian, 2020), we focus on the manipulative behavior in R&D expenditure accounting policy selection and finds that the additional deduction also induces enterprise to adjust their earnings and tax through R&D expenditure accounting policy selection. (2) it provides a research scenario in which two types of motivations for R&D expenditure accounting policy selection coexist. Prior research analysis the adverse selection behavior in R&D expenditure accounting policies from earnings management or tax avoidance, respectively. We investigate the dynamic selection process of R&D expenditure accounting policies made by enterprises under the premise of taking into account both motives, based on the scenario of increased additional deduction preference.Our findings have two policy implications: (1) government might improve the audit quality of eligibility for additional deduction and the violation cost for R&D information manipulation. (2) government might implement differentiated additional deduction policies according to the internal governance level of the enterprises.