Finding alternative sources of identification in generalized selection models using principal stratification1

In this paper we consider two approaches for dealing with “endogenous selection” problems when estimating causal effects, namely selection models (SM) and principal stratification (PS). Our main goal is to highlight similarities and differences of the two approaches, by first investigating the different nature of their parametric hypotheses. In order to support our reasoning, we show their different performances by simulations under both approaches. We argue that principal stratification is able to suggest alternative identification strategies not always easily translated into assumptions of a selection model.