中国政策试点中的随机实验:一种方法论的探讨

Randomized Experiments in Chinese Policy Pilots:A Methodological Discussion

中国独特的政策试点机制近年来成为公共政策研究的热点问题。部分研究者将政策试点与发展经济学中广泛开展的随机对照实验进行类比,认为中国的政策试点应当借鉴或采用随机实验的程序与方法,以便对政策效果进行科学的影响评估(Impact Evaluation)或因果推断(Causal Inference)。论文提出了一种不同的观点:尽管政策试点缺乏随机实验中干预的随机分配和精确的测量结果,但是,考虑到中国的政策实施规模与政府执行能力,随机实验评估并不是必要的。理由如下:国家发展是一个宏观的政治经济过程,随机实验评估难以真正实施;即使政策被随机实验验证有效,也可能由于政府部门能力存在差异而难以推广;在科层制下,随机实验方法和实验结果并不一定会被重视与采纳;在局部地区和领域中,得到的实验结果可能无法在更大范围内重现。因此,真正高质量的政策评估,不仅需要准确认识到随机实验的价值与问题,还要充分结合实验与非实验方法,深入理解政策试点对国家治理的重要意义。

China's unique policy pilot mechanism has become a hot research issue in recent years.Some researchers have drawn an analogy between policy pilots and Randomized Controlled Trials,which are widely conducted in developing economies.For them,policy pilots in China should borrow the methods of randomized experiments to evaluate scientifically,and draw causal inferences from,policy effects.This paper proposes a different view.Although policy pilots lack the random assignment and precise outcome measures of interventions in randomized experiments,randomized experimental evaluations are not necessary given the scale of policy implementation and the government's ability to execute in China.The reasons are many.National development is a macro political economy process that makes it difficult to truly implement randomized experimental evaluations.Even policies validated by randomized experiments may be difficult to replicate due to differences in the capacity of government departments.Randomized experimental methods and results are not necessarily valued and adopted under a section system.And,results obtained locally may not be reproducible in larger areas or domains.Therefore,a high-quality policy evaluation requires an understanding of the value and the problems of randomized experiments as well as an in-depth understanding of the significance of policy pilots for national governance that integrates experimental and non-experimental methods.