Lee's (2009) Bounds

[Joshua Smith]

Hi,

I need to compute the annual coefficients in a diff-in-diff regression of the form:

y_it=a_i+b_t+d_t*sum_t(treatment*year)+e_it

where each d_t corresponds to the treatment effect in year t.

Since I have attrition over time in i, I would like to compute the lee's (2009) bounds for each coefficient. How can I do that? the standard lee bounds command in stata doesn't seem to allow that.

Thanks a lot for your attention.

[paulvonhippel]

I think Lee's bounds were designed to compare the treatment and control group in an RCT. I don't think they are designed for use in a nonrandomized diff-in-diff analysis using panel data.

[Joshua Smith]

Thank you for your answer. Don't you think there is a way to implement this procedure in other settings, even with some caveats?