Nonparametric Approaches to Empirical Welfare Analysis
综述了基于截面微观数据计算政策干预福利效应和净损失的非参数方法，避免对个人偏好施加统计和函数形式限制，使福利估计更可信，但需要更大的样本内数据变异。
Welfare analysis of policy interventions is ubiquitous in economic research. It plays an important role in merger analysis and antitrust litigation, design of tax and subsidies, and informs the current debate on a universal basic income. This paper provides a survey of existing empirical methods, based on cross-sectional microdata, for calculating welfare effects and deadweight loss resulting from realized or hypothetical policy change. We briefly outline classical parametric methods that are computationally tractable, then discuss recent nonparametric approaches that avoid making statistical and functional-form restrictions on individual preferences. This makes the welfare estimates theoretically more credible, and clarifies exactly what welfare-relevant information is contained in demand distribution in various choice settings. However, these methods also demand greater in-sample variation in the data for practical implementation than classical parametric approaches. We then cover settings with externalities. The above results are theoretical, and take the demand function as known; therefore, we briefly discuss empirical problems around demand estimation. We conclude by suggesting areas for future research.