Interpreting coefficient after propensity score matching using teffects psmatch

[Brandon Mathews]

I'm new to propensity score matching and I'm trying to understand the output for my analysis. I've run the following command in Stata to match observations on a variety of preprogram characteristics:

teffects psmatch (recidivism) (compprogram age race gender poffense hhincome offense tcproc)

Recidivism is the outcome variable and compprogram is the treatment variable. Because the outcome variable is dichotomous, I'm a bit confused about how to interpret the coefficient as it pertains to the Average Treatment Effect. I get the interpretation of a continuous variable, however I've been unable to find any literature that specifically discusses the output as it relates to a dichotomous variable.

Any help on interpretation would be much appreciated and I've attached the output for more info.

Brandon

[Sergei Ivanov]

Dear Brandon, I have similar problem with propensity score and binary outcome! Have you solve this problem?

[Rob Wood]

I believe the coefficient for a binary outcome is the absolute difference between the percentages for the two groups. Say your binary outcome is gender (terrible example I know), where 0=female and 1=male. If your treatment effect is say 0.1372, then the difference in the percentages of males between the two groups is 13.7%. So the control group may have 26% males, while the treatment group has 39.7% males.

Hope that makes some sense.