1
$\begingroup$

Trying to construct a synthetic control and I've chosen a set of covariates that are correlated with my outcome variable and significant for P>|t| using OLS. Is choosing covariates for synthetic control the same as you might for a multiple regression? Or can I pick covariates even if they are not significant?

$\endgroup$

1 Answer 1

1
$\begingroup$

You need to choose covariates that are confounders, in particular, a sufficient set of variables to remove confounding. This has nothing to do with whether they are significant in a regression of the outcome on the covariates. Do not use that as a criterion to decide which variables to include. Include all variables that need to be balanced in order to be able to claim that the observed differences in the outcome between the treated unit and synthetic control can only be due to treatment.

$\endgroup$
1
  • $\begingroup$ Makes sense, thank you! $\endgroup$
    – prismarine
    Mar 25, 2022 at 7:40

Your Answer

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

Not the answer you're looking for? Browse other questions tagged or ask your own question.