Selection on observables, covariate-specific trends and conditional parallel trends with difference-in-differences: Pedro's Checklist, Step 4(d)

When everyone was first hearing about the “standard” twoway fixed effects specification being biased in differential timing scenarios, it was not uncommon to hear or read authors say something to the effect that since they didn’t have differential timing, they were going to use twoway fixed effects, and then commence to include a variety of covariates. …