How to match on pre-event characteristics? (PSM)

[Kathrin Me]

Hi all,

I aim at estimating the effect of job loss on the labor supply of the spouse using an event study. Since my panel is quite unbalanced, I would like to use nearest neighbor propensity score matching with replacement. I only want to allow for matches within the same displacement year/ event year. Furthermore, I would like to match on characteristics that refer to the year before the displacement.

So far I tried:

egen matchyear = group(syear)
levels matchyear, local(gr)
foreach j of local gr {
            probit treated $covariates if matchyear == `j' 
            predict double ps`j' if matchyear == `j'
            psmatch2 treated if matchyear == `j', outcome(hh_income) pscore(ps`j') neighbor(15) ai(15) 
      }

, where treated is a dummy, taking the value 1 if one is treated at some point during the observation period and 0 if one is not treated.

My question: How can I ensure that the characteristics on which the units are matched are pre-intervention traits? I defined already an "event"-variable, that takes the value -1 in the period before the event, 0 in the year of the event, 1 in the year after the event, and so on. However, it is only defined for the treated individuals. Therefore, I cannot just write

if event == -1

.

Thanks already now for your help!

Best,
Kathrin

[Lorenzo Bernabei]

Hello,

I know that it is quite late, but I would like to whether you came up with a solution eventually. I am having a similar issue.

Bests,

Lorenzo