We use a linear probability model with interactions and a switching probit model (SPM) to estimate heterogeneous effects of Medicaid expansions on Medicaid take-up, private insurance coverage and crowd-out. Specifically, we estimate: i) LATEs; ii) ATETs for the currently eligible; and iii) ATETs for those made eligible by a non-marginal (counterfactual) expansion in Medicaid eligibility. Both estimation methods can control for observable differences across individuals, while SPM can also control for unobservable differences. For Medicaid take-up and private insurance coverage, the effects are precisely estimated and differ dramatically across demographic groups, but this is less true for the crowd-out estimates.