Difference in Difference

[Alex Lampert]

Hi all,

I am curious about two different output regressions when I specify a diff-in-diff two different ways, where treatment and after are my indicator variables:

1) reg outcome i.treatment#i.after, r
2) reg outcome treatment##after, r

1) gives me three treatment*after coefficients plus the constant. 2) gives me the actual diff-in-diff estimate plus the two coefficients for treatment and after.

2) is more intuitive to me, so I'm looking for any insight into those coefficients for 1). Can anyone explain the output to me?

Alex

[Clyde Schechter]

I think you need to show us the outputs you are looking at and the actual code that generated them. (Use code delimiters so it will be readable--see FAQ #12 if you need to learn how.) The two versions of code you show should give completely identical results since Stata automatically treats the variables around ## as being prefixed with i. unless you explicitly override that with c. So you are asking for an explanation of a difference that does not exist--so I suspect that you have actually done something different from what you are describing here.