bcallaway11/ife: Treatment Effects in Interactive Fixed Effects Models

This is code to estimate treatment effects in interactive fixed effects models where the number of available time periods is relatively small. It implements the approaches proposed in Callaway and Karami (2023) and in Callaway and Tsyawo (2023). For CK2023, the basic idea is to use identify/estimate the parameters of an interactive fixed effects model for untreated potential outcomes using the group of units that do not participate in the treatment by exploiting that the effect of at least one particular covariate does not change over time (this condition may not be met in all applications). See the paper for more details. For CT2023, the idea is to exploit staggered treatment adoption to be able to recover the parameters of the interactive fixed effects model. This setup is likely to be attractive in applications where the distribution of time invariant unobservables is different between the treated group and untreated group and when the effect of these time invariant unobservables can vary over time. This situation would cause alternative identification strategies such as difference-in-differences or individual-specific linear trends to generally fail.

install.packages("remotes")
remotes::install_github("bcallaway11/ife")