Has Diff in Diff Lowered the Price on Synth?
Magic, Pegasuses, Weights, Imputation, and maybe what’s changed
Boy was this fun to write. I’ve had this thing in my drafts for a month, and in my brain for two years. Sort of went on a rant about magic and spells, so reader beware.
A well known result in microeconomics is that when the relative price of one good increases, ignoring income effects, people substitute to some nearby substitute. Watching econometricians writings on difference in differences grow the last few years, and then reading new papers that interact with it, I am inclined to say that the cognitive cost of diff in diff has risen, both in understanding and in practical implementation. What once seemed like the simplest design has now become less so. Estimators that were unbiased under strict exogeneity, they learn, required constant treatment effects — apparently it has always but we’d not quite drawn that connection.
The solutions seemed easy enough though — only use untreated units as a control — but then we were swimming in a half dozen alternatives that did that, just in di…
Keep reading with a 7-day free trial
Subscribe to Scott's Substack to keep reading this post and get 7 days of free access to the full post archives.