What Drives Differences in Health Care Demand? The Role of Health Insurance and Selection Bias

This paper employs an econometric model to parse di erences in health care utilization attributable to private health insurance and differences due to self-selection into insurance status, with specifc interest in selection on unobservable traits such as insurance preference or attitude toward health risks. The model has two components, one component to model insurance outcome, the other to model demand for care measured as the annual number of doctor visits and prescriptions filled. Recognizing the endogeneity of health insurance, the model allows for correlated unobserved heterogeneity by assuming a latent factor structure. Values for these latent factors are drawn through simulation and the model is estimated using maximum simulated likelihood methods. For the observable characteristics that predict need for health services we find evidence of adverse selection. However, we also find evidence of advantageous selection on the unobservable characteristics common to insurance choice and utilization. In other words, unobserved heterogeneity that increases the chances of being uninsured isassociated with higher utilization. Given this selection decomposition, there is no inherent conflict in describing the influence of both adverse and advantageous selection in utilization comparisons. After controlling for selection, the insurance incentive effect (ex-post moral hazard) is positive and signifcant. For the average individual, switching from no coverage to full coverage would result in 2 additional visits to the doctor per year (+160%) and 8 additional prescriptions lled (+207%).