Counterfactuals

[imccart]

• Consider counterfactual where all PCPs see all the outcomes of specialists and not just a single PCP (public vs private full information)

[sethrs-jhu]

Another counterfactual, but not worth doing given the small effects: randomize some referrals to learn about new doctors

[imccart]

Priority Counterfactuals:

1. All PCPs see all outcomes of specialists (not just their own patients)
2. All PCPs have perfect information about specialist quality

Consider each counterfactual with and without familiarity

[sethrs-jhu]

try to reconcile small effects from logit model with larger effects in event study