Using ln(Y+1) as dependent variable in diff-in-diff -- bad idea

[George Ford]

Recently came across this paper, forthcoming in QJE.

Details the problems (and solutions) when Y has zero values.

https://www.jonathandroth.com/assets/files/LogUniqueHOD0_Draft_Accepted.pdf

[Joao Santos Silva]

It is generally a bad idea, not only in diff-in-diff; see for example

https://onlinelibrary.wiley.com/doi/full/10.1111/obes.12583