*Bounty: 50*

*Bounty: 50*

If I am running a difference in difference as the following:

$y_{i,t}=beta Treat_i x Post_t + Post_t +Treat_i + epsilon_{i,t}$

Where i can index say, state, and t time.

If I wanted to calculate the standard errors of the $beta$ by hand, given that this is just a difference in difference in means, should it no be equivalent to:

$var((bar{y}*{treat,post}-bar{y}*{treat,pre}) – (bar{y}*{control,post}-bar{y}*{control,pre}))$? and and thus this should boil down to:

$var(bar{y}*{treat,post})+var(bar{y}*{treat,pre}) + var(bar{y}*{control,post}) + var(bar{y}*{control,pre})$,

and then taking the square root of this term? I have tried doing this ‘by hand’ in stata and did not receive the same standard errors as the reported ones.