Oregon Health Insurance Experiment (Provided to Single Adults)

Traditionally, Medicaid programs have been provided to children and families with children, or those on social security / disability income. However, recently there have been more expansions of Medicaid programs to low-income adults who are not categorically eligible (e.g. on AFDC). One such expansion occurred in 2008 when Oregon used a budgetary windfall to extend health insurance coverage to low-income single adults who were not categorically eligible for Medicaid (i.e. not on other forms of welfare like SSI). The program was over-subscribed and thus they used a lottery to restrict enrollment.

Finkelstein et al. (2012) use this lottery to generate random variation in Medicaid coverage. They find that Medicaid increased health care use across all categories of care, including outpatient care, preventive care, prescription drugs, hospital admissions, and emergency room visits. They also find that Medicaid improved self-reported health and reduced depression. However, they do not detect any statistically significant impact on mortality or physical health measures. In contrast, they find large reductions in out of pocket medical spending and reduced incidence of unpaid bills.

MVPF SUMMARY

_________

Finkelstein et al. (2019) translate the estimates of the causal effects of the program, combined with measures of the QALY for the health benefits, into a measure of recipients’ willingness to pay for Medicaid and the cost to the government of providing Medicaid. Hendren and Sprung-Keyser (2020) translate estimates from Finkelstein et al. (2019) into their implied MVPF.

Using three different approaches to measure the insurance value of Medicaid, they find a WTP of $793, $1,421, and $1,675 for the average complier enrollee. As discussed in Finkelstein et al. (2019), the analysis above does incorporate the valuation individuals have for any potential long-run impact of insurance on their health. But, it accomplishes this through its measured impact on self-reported health and reduced depression. Since the experiment occurred in 2008, it could of course be the case that future gains have realized. However, it is important to note that the estimates suggest that the program did not have statistically significant improvements in physical health in the short or medium run and seems to have had no statistically distinguishable impact on earnings of the enrollees, although the confidence intervals are quite large.

The cost to the government of covering the medical costs of those who took up the insurance is $3,600. Much of the impact of Medicaid is to reduce the prevalence of unpaid bills: Finkelstein et al. (2012) estimate that the net resource cost of covering these medical costs is $1,440, with the difference comprising transfers from the government to the providers of uncompensated care. These may be transfers to hospitals and other insureds, or other creditors of the uninsured who bear the incidence of unpaid medical debt. Moreover, some of these costs may ultimately be born by the government through reductions in hospital-level Medicaid payments through the DISH program or through reductions in tax deductions. Hendren and Sprung-Keyser (2020) follow Finkelstein et al. (2019) and consider a baseline case in which the cost savings from uncompensated care fall to the government, along with a robustness case in which there are other individuals paying the uncompensated care. In this latter case,

Assuming the ultimate cost of uncompensated care is borne by the government suggests the cost to the government of providing Medicare is $1,440. This is higher than the mechanical increase in expenditure that would have been $1,147 in the absence of behavioral responses. The willingness to pay for the insurance varies depending on the method used to calculate. Their WTP estimates range from $1,675 for the “complete information approach”, $1,421 for the “consumption-based optimization approach”, and $793 for the “health-based optimization approach”. Subsequently, these translate into an MVPF of 1.16, 0.99, and 0.55 respectively. Hendren and Sprung-Keyser (2020) adopt the complete information approach for their baseline specification, leading to an MVPF of 1.16.

Instead of assuming the government pays for the cost of uncompensated care, one can also assume that individuals themselves bear the cost of this care. In this case, the WTP rises to $3,835, $3,581, and $2,953 in the three approaches. But, the cost to the government also increases to the full $3600. This means that the MVPF in the three cases is given by 1.07, 0.99, and 0.82 respectively.