Hi community,
I would like to test for the common trend assumption of the difference-in-difference estimator. I use a dataset on yearly sales of cars for different years in Europe. In more detail, I take the German Reunification as the natural experiment and, thus, Germany as the treated group after 1990, while I'd use Belgium as the control group as the trends in different car segments look similar on charts. I would like to verify this trend assumption empirically using a test. Any ideas?
Thank you a lot,
Mert
You can visually test the trend assumption by graphing the trends before the treatment to get a first indication if the treatment and control group follow the same trend.
In addition, you can assign placebo-treatments at e.g. 2 years before the actual treatment, to make sure you are not finding significant effects before the reunification in your case.
A third option (see https://economics.harvard.edu/files/economics/files/ms31543.pdf?m=1604679900 , p16 and figure IV) is to measure that the slopes are the same in your pre-treatment period.
- Badges
- Science topic
- Similar topics
- Publishing