I am writing my thesis right now and have had many troubles trying to export the results of the user-written command psmatch2. Eststo and esttab generally only report the results from the unmatched sample instead of the average treatment effect on the treated (ATT) and the associated standard error. I have seen at least 5 different forum posts, all unsolved, with this issue. After much experimentation, I have found a solution.

Let's examine the problem. Run the psmatch2 command with eststo, and then run "return list" to see the scalars stored as a result of psmatch2:

Code:
eststo: psmatch2 dependent_var independent_var1 independent_var2, out(outcome_var) 
return list
You will see that the ATT is stored in r(att), and the standard error of the ATT is stored in r(seatt).

However, eststo does not store r(att) and r(seatt) automatically as an active estimation result. Therefore, running esttab after simply gives you the results from the unmatched sample.

The solution here is to use estadd to store r(att) and r(seatt), and then run esttab. Estadd will store r(att) and r(seatt) as estimation results, which esttab can then access.

Here is my solution, in which I also calculate the t statistic and two-tailed p-value. Obviously, substitute the dependent, independent, and outcome variables for your own variables.

Code:
eststo clear
eststo: psmatch2 dependent_var independent_var1 independent_var2, out(outcome_var) 
estadd scalar r(att)
estadd scalar r(seatt)
estadd scalar t_stat = (r(att)/r(seatt))
estadd scalar p_value = (2 * ttail(e(df_r), abs(e(t_stat))))
esttab, scalars(att seatt t_stat p_value) drop(_treated _cons) obslast
Under the esttab options, I drop _treated and _cons because these are the results on the unmatched sample. I use obslast so that the number of observations will appear at the bottom of the table instead of the top. Of course, you can modify the code to specify if you want to export the results to HTML, pdf, or whatever other format is possible with esttab.

This is still an incomplete solution as I cannot find a way to automatically star the p-values, but I believe it is helpful nonetheless.