We study identification and estimation of treatment effects in common school choice settings, under unrestricted heterogeneity in individual potential outcomes. We propose two notions of identification, corresponding to design- and sampling-based uncertainty, respectively. We characterize the set of causal estimands that are identified for a large variety of school choice mechanisms, including ones that feature both random and non-random tie-breaking; we discuss their policy implications. We also study the asymptotic behavior of nonparametric estimators for these causal estimands. Lastly, we connect our approach to the propensity score approach proposed in Abdulkadiroglu, Angrist, Narita, and Pathak (2017a, forthcoming), and derive the implicit estimands of the latter approach, under fully heterogeneous treatment effects.
翻译:我们研究常见学校选择环境中的治疗效果,在个别潜在结果的无限制差异下,确定和估计其治疗效果;我们提出两种识别概念,分别对应设计和抽样不确定性;我们为大量各类学校选择机制确定的一系列因果估计值,包括随机和非随机断绝的因果估计值;我们讨论其政策影响;我们还研究这些因果估计值的非对数估计值的无症状行为;最后,我们将我们的方法与在Abdulkadiroglu、Angrist、Narita和Pathak(即将出现2017a)提出的常态评分法联系起来,并在完全不同的治疗效果下得出后一种方法的隐含估计值。