439 followers
@NeumarkDN For those interested, these are some of the points David makes. Key idea: the harder you try to get good causal identification, the more you're restricting your sample, so the higher your standard errors will be. (much more at the link) https://