Instrumental variable for binary endogenous variable?

I’m trying to figure out how would you implement a Stan instrumental variable model when the endogenous variable is binary. I found @James_Savage post on how to do it when the endogenous variable is continuous. Moreover, mostly harmless econometrics suggest… you should use garden-variety 2SLS (IV) for dummy endogenous variables…”. So, using @James_Savage model would be acceptable? Is there a better alternative? Is anyone aware of a Stan case study that goes over this?


Yep, the standard continuous setting works fine if all you want is an estimator for some effect. You can also use my biprobit code if both the treatment and outcome are binary.


Thanks a lot Jim!