Differences in Differences

For Observational data, if we use following methods:

1. Rely on lookalike or regression or machine learning methods. Drawback of this method is selection on unobservables.

2. Rely on pre-post analysis. Drawback of this method is not accounting for potential trends.

So we introduce the DID method: DID methods exploit variation in time (before v.s. after) and across groups (treated v.s. untreated) to uncover causal effects of interest