Congratulations! You’ve completed a lesson on the fundamentals of difference in differences for causal inference. Through applied exercises you learned:
- DID uses the trend of a control group as an approximate counterfactual trend for the treatment group.
- DID requires that we have pre-treatment data for both groups and that we meet the parallel trends assumption.
- We can compute the ATT by taking mean differences or through linear regression.
You may be wondering why we would use linear regression at all when we can just use means. Regression allows us to control for confounding variables as well as get estimated standard errors. We explored simple models in this lesson to get a strong foundation, but more complex DID analyses benefit from the added flexibility of regression modeling.
Check out the output in the learning environment. In this DID analysis, we found that the tax in Sydney reduced average ticket sales by 2.25. What does each coefficient in the model represent for the ticket sales problem?