psmatch2 standard errors

[Renee Edwards]

I am trying to get the correct standard errors for my ATTs after propensity score matching. I am using the most recent version of psmatch2 (Feb 2014) which now appears to have options for calculating standard errors that take into account the estimated nature of the propensity scores. The following is from the help psmatch2 command:

ai(integer) calculate the heteroskedasticity-consistent analytical standard errors proposed by
Abadie and Imbens (2006) by specifying the number of neighbors to be used to calculate the
conditional variance (their formula (14)). With option altvariance one can specify to use the
estimator of Abadie et al. (2004) instead. altvariance When using ai(integer), calculate the conditional variance using the expression in
Abadie et al. (2004, p.303).
I'm not sure whether to use the ai(1) option or altvariance option. I am using psmatch2 instead of teffects because teffects doesn't have the options of noreplacement and common as far as I can tell. My outcome variables are all continuous. I am doing 1 to 1 nearest neighbor matching using a caliper of .25SD of the estimated propensity scores. Can someone suggest whether the Abadie and Imbens (2006) formula or the formula from Abadie et al (2004) would be appropriate in this case?

Thanks so much!
Renee

[Ariel Linden]

Either one is appropriate (as well as using regress with vce(robust)). I suggest that you run it both ways and see how the results differ. I suspect that you won't find a meaningful difference between SE estimators