Hi all,

I'm trying to implement DID (two periods two groups) with Lee bounds by dropping the top/bottom x% outcome first differences from the group with attrition (control or treatment).

1) If attrition rate is higher in the control group but due to missingness in the data (respondents either Didn't know or refused to answer the question), the treatment group is experiencing the greater missingness drop, which reference group should I be using?

2) The DID coefficients differ depending on whether I drop attriters or not - I was wondering why this would be the case and why would attriters be included in the analysis if there is no data available in the second period.

3) Following question 2), manually computing Lee bounds works perfectly when dropping attriters (i.e. I obtain a lower and an upper bound). However, this is not the case when keeping attriters, probably because first differences are computed without second period data available.

Thank you