Private Insurance, Selection, and Health Care Use

Abstract

Observed differences in medical utilization between the privately insured and uninsured reflect the combined effects of self-selection and insurance incentives (moral hazard). This article provides a Bayesian framework for decomposing the disparity into incentive and selection components. The effect of selfselection in private insurance on the number of doctor visits is estimated using a multiyear sample of the U.S. adult non-Medicare population obtained from the Medical Expenditure Panel Survey. We use a flexible econometric framework based on the "Roy model" and develop a Markov chain Monte Carlo algorithm. We estimate the distribution of treatment effects and find strong evidence indicating selection, which accounts, on average, for 50% or more of the observed disparity in doctor visits.

Partially Identifying Treatment Effects with an Application to Covering the Uninsured

“…They find that private coverage would increase the annual probability by 61%, corresponding to an increase of about 8% across the population as a whole. Estimates in Deb et al (2006) suggest that expanding private insurance to uninsured adults would increase their number of office-based visits by about 46%, corresponding to an increase for the adult population of about 5%. Hadley and Holahan (2003), who do not account for unobserved factors, estimate that public coverage would increase the annual number of office-based visits among the uninsured by 41%, while private insurance would increase visits by 30%.…”

Provider networks and primary‐care signups: do they restrict the use of medical services?

“…Thus, employment characteristics, such as types of plans offered and the size of the firm serve as proxies for plan supply which presumably determines the type of plan an individual chooses, but should have no direct relationship with utilization of medical services. Johnson and Crystal (2000), Olson (2002), Deb and Trivedi (2006a) and Deb et al (2006) also use employment characteristics as instruments in similar contexts.…”

“…Among recent studies, Goldman (1995), Mello et al (2002), Deb and Trivedi (2006a) and Deb et al (2006) explicitly address the issues and find considerable evidence of self-selection. Cardon and Hendel (2001) also model self-selection but do not find substantial evidence of it.…”

Partially Identifying Treatment Effects with an Application to Covering the Uninsured

“…Some evidence of aggregate adverse selection has been reported in parametric studies of the effects of insurance. Using data from the MEPS to estimate a model of private insurance coverage and office-based doctor visits, Deb, Munkin, and Trivedi (2006) find that half of the observed lower healthcare use among uninsured adults compared with privately insured adults can be attributed to self-selection rather than the lack of insurance. Dor, Sudano, and Baker (2006) find similar evidence of selection into private insurance.…”

“…They find that private coverage would increase the annual probability by 61 percent, corresponding to an increase of about 8 percent across the population as a whole. Estimates in Deb, Munkin, and Trivedi (2006) suggest that expanding private insurance to uninsured adults would increase their number of office-based visits by about 46 percent, corresponding to an increase for the adult population of about 5 percent. Hadley and Holahan (2003), who do not account for unobserved factors, estimate that public coverage would increase the annual number of office-based visits among the uninsured by 41 percent, while private insurance would increase visits by 30 percent.…”