In a matching diff-in-diff, which pre-treatment variables can be legitimately used in the matching procedure? I'm reading this paper, which uses propensity score matching to estimate the effect of military interventions on homicides in Mexico. They compare homicide rates one year before the intervention to homicides one year after. Since the interventions started in 2007, they match on pre-treatment covariates in 2006 (including the 2006 homicide rate $H_{2006}$), assigning two control municipalities to each treated municipality.

Now, three interventions occurred in 2007. For these cases, they compare $H_{2008}$ to $H_{2006}$, while also matching on $H_{2006}$. Is this kosher? Can you match on the same variable that is used to compute the diff-in-diff?



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.