基于时变参数模型的中国研发投入回报率估算

The estimation of the Return to R&D Investment in China Based on Time-varying Parameters Model

研发投入回报率的估算对于科技政策指定具有重要意义。传统的研发回报率的估算一般基于恒定参数模型,恒定参数的假设不符合现实。文章基于傅里叶变换处理时变参数并扩展了Jones and Williams(1998)的研发回报率估算框架。使用我国1995—2013年的省际面板数据,估算我国的研发回报率。研究发现:生产部门生产函数的资本、劳动弹性具有明显的时变性特征;知识部门研发投入对知识生产的贡献也具有明显的时变性特征。在2002—2013年之间,基于时变模型的估算得到我国研发投入的平均回报率为11.8%,而基于恒定参数模型的估计结果为17.1%,即基于恒定参数模型的估算每年高估研发回报率约5%。因此,基于恒定参数模型考虑研发投入的回报率和最优规模可能产生误导性结论。

The return to R＆D investment is crucial for science and technology policy. The traditional estimation of the return to R＆D investment is based on the assumption of constant parameters, which is unreal. This paper employs the Fourier transforma- tion to approximate the time-varying parameters and extends the framework of estimating the return to R＆D investment of Jones and Williams （1998）. We use the panel data from 1995 to 2013 to estimate the return and find that： the parameters of labor and capital in production function and knowledge function have significant time-varying features. From 2002 to 2013, the average re- turn to R＆D based on the time-varying model is 11.8%, while it is 17.1% based on the constant parameter model. Therefore, the estimation of the return to R＆D investment based on the constant-parameter model is overestimated by 4%-6% from 2002 to 2013. We therefore conclude that the estimates of the return and scale to R＆D based on the constant model may be misleading.