Tuesday, February 13, 2018 - 12:30pm to 1:30pm
038 Townshend Hall, 1885 Neil Ave
Health Persistence and the Ex Ante Value of Medicaid as a Safety Net
Previous studies that estimate the insurance value of Medicaid typically identify local average treatment effects of crossing an eligibility threshold, such as an income threshold, or winning a lottery conditional on Medicaid eligibility. Although these approaches provide very credible estimates of willingness to pay (WTP) within the population local to the eligibility threshold, they may understate aggregate WTP if people away from the eligibility margin also have meaningful WTP. We estimate the insurance value of Medicaid as a safety net in the full population of Utah. Using a comprehensive new database combining all-payer claims data, hospital discharge data, and administrative Medicaid records for the entire state between 2008-2015, we provide new evidence on reclassification risk by tracing the pathways that lead to Medicaid enrollment, and showing that people away from the eligibility margin benefit from Medicaid. For example, even young, healthy, privately insured people who have never previously enrolled in Medicaid have a 4% probability of enrolling within a 5-year period. We demonstrate that cross-sectional and short-panel estimates understate WTP by failing to capture the persistence of health status over a long time horizon. We discuss how the canonical economic model of social welfare imposes the assumption that insurance against variance-preserving autocorrelation in risks has zero social value, and show that under alternative welfare models the cross-sectional value of Medicaid fails to capture up to 95% of the total insurance value. Finally, if the ex ante value of Medicaid as a safety net, as perceived by a person anticipating a random draw from the population of life-cycles, includes the value of partial insurance against being born with a predisposition for high or low lifetime medical spending, the aggregate insurance value of Medicaid increases further.