Estimation of panel data models with binary indicators when treatment effects are not constant over time

We show that two commonly employed estimation procedures to deal with correlated unobserved heterogeneity in panel data models, within-groups and first-differenced OLS, can lead to very different estimates of treatment effects when these are not constant over time and treatment is a state that only changes occasionally. It is therefore important to allow for flexible time varying treatment effects when estimating panel data models with binary indicator variables as is illustrated by an example of the effects of marital status on mental wellbeing.