Full Citation

Title: Roy-Model Bounds on the Group Differences in Treatment Effects: Theory with an Application to the Great Migration

Citation Type: Miscellaneous

Publication Year: 2013

ISBN:

ISSN:

DOI:

NSFID:

PMCID:

PMID:

Abstract: I study the conditions under which information about the causal effect of a treatment can be identified by applying difference-in-differences regression to two groups that both self-select into treatment. I establish that, in many cases when selection and counterfactual outcomes can be described by a Roy model, differences-in-differences provide a lower bound on group differences in the average effect of the treatment on the treated. This group difference in causal effects is particularly informative in cases where treatment effect heterogeneity is of direct interest or when it is reasonable to assume that the average treatment effect is nonnegative for both groups. Furthermore, because the requirements for identification are relatively weak, this group difference may provide a framework for understanding treated-untreated comparisons in causal terms in the absence of a credibly-exogenous source of variation in the propensity to be treated. I use the identification results to interpret North-South wage differentials in terms of black-white differences in the causal effect of Northward migration on wages, finding that migration increased wages for black migrants by at least 24% as much as for white migrants.

User Submitted?: No

Authors: Gardner, John

Publisher: Carnegie Mellon University

Data Collections: IPUMS USA

Topics: Methodology and Data Collection

Countries: