CMStatistics 2016: Start Registration
View Submission - CMStatistics
B0728
Title: Quantile difference in differences Authors:  Brantly Callaway - University of Georgia (United States) [presenting]
Tong Li - Vanderbilt University (United States)
Tatsushi Oka - National University of Singapore (Singapore)
Abstract: The effect of a binary treatment on the quantiles of some outcome is identified and estimated under a Conditional Quantile Difference in Differences assumption. This assumption says that, for individuals with similar covariates, the path of the $p$-th quantile ($p\in (0,1)$) is the same as the observed path for untreated observations. We show that this assumption is valid in common panel data-type models with unobserved heterogeneity. Our estimates of unconditional quantile treatment effects converge at the parametric rate. Moreover, we provide simple estimators for quantile treatment effects that are easy to implement in practice, provide results on the validity of the bootstrap for computing standard errors, and consider the empirically relevant case with more than two time periods and where treatment can occur in different periods for different individuals. Finally, we consider an application on the effect of union membership at different points in the skill distribution.