Testing the equality of DiD coefficients

[lloydheng]

Talk through whether this approach sounds right:

Example testing equality between taco and burrito DiD coefficients:

1. Get monthly DiD estimates for taco and burrito.
2. Do a paired t-test using estimates in the first year, extracting the p-value. This tells us if the pair-wise monthly coefficients in the first year are significantly different between tacos and burritos.
3. Repeat for the second-year estimates.
4. Adopt some kind of correction for the numerous pair-wise comparisons/hypotheses we're testing. Bonferroni correction = divide the significant threshold by the number of within-category comparisons we're making. So, if we're comparing 5 item categories, the threshold is now $${0.05 \over 5} = 0.01 $$

This, versus just looking at the CIs give us quite different conclusions.

[lloydheng]

Testing z-test instead.