The paper examines whether, among inpatient psychiatric admissions in California, for-profit (FP) hospitals engage in cream skimming, i.e., choosing patients for some characteristic(s) other than their need for care, which enhances the profitability of the provider. We propose a novel approach to identify cream skimming using cost outcomes. Naïve treatment effect estimates of hospital ownership type consist of the impact of differential patient case mix (selection) and hospital cost containment strategies (execution). In contrast, an instrumental variable (IV) approach can control for case mix and establish the causal effects of ownership type due to its execution. We interpret the difference in naïve and IV treatment effects to be driven by FP hospitals’ selection (cream skimming) based on unobserved patient case mix. We find that FP specialty hospitals are more likely to treat high-cost patients than their not-for-profit (NFP) counterparts, providing no evidence of cream skimming by FP hospitals. Our findings may alleviate concerns about the recent proliferation of FP psychiatric hospitals, particularly regarding cream skimming.