This paper reframes Peck’s (2003) approach to estimating impacts on endogenous subgroups in terms of Frangakis and Rubins’ (2002) notions of principal stratification and principal effects, and shows her estimators can be derived in terms of proxy variables for the (partly) omitted indicator of principal stratum membership.  A specification test is presented, and a method for estimating bounds for principal effects is described.  An alternative estimation approach using multiple imputation is also described.