Estimating Causal Effect in Observational Studies Issue

[James Clarkeson]

Hi there,

I am studying the effects of a subsidy on sustainable investment for a large sample of firms, using cross-sectional data. The subsidy was awarded to certain firms based on the awarding authority’s own assignment mechanism.

I am currently using PSM methods, estimating pscores with a set of covariates. However, one of the important covariates (firm size) is likely impacted by the subsidy itself, but is also an important predictor of the grant assignment - by economic intuition. What can I do to deal with this endogeneity problem? I can achieve an estimate by including firm size as a covariate in the binary choice model (probit), but might a structural equations approach work here instead?

Any help is much appreciated.

[Dimitriy V. Masterov]

Is firm size measured after the grant decisions are announced or before?

[James Clarkeson]

Is firm size measured after the grant decisions are announced or before?

It is measured after the grant was awarded. Sorry, I forgot to include this important information.

Which would suggest that firm size is removed from the BCM, however I feel like it is a very good predictor of the assignment mechanism, and would be more important to keep in. Data is only for observations 1 year post-grant award.

[Dimitriy V. Masterov]

In that case, I don't see any way of solving that with the data you have.