Automatic Enrollment in Health Insurance
Introduction
Shepard and Wagner (2022) study a change in the automatic enrollment policy for CommCare, a free public health insurance program for low-income adults in Massachusetts. Beneficiaries were adults age 19-64 with family incomes below 100% of the Federal Poverty Level who did not have access to any other form of health insurance, so were otherwise uninsured. The paper uses individual data on program enrollment and health care spending and outcomes while enrolled coupled with a policy change in mid-2009 that suspended the auto-enrollment policy, which had been in place from 2007-2009. The paper uses a difference-in-difference design, comparing enrollment trends for affected groups who lost auto-enrollment (enrollees with incomes below poverty) vs. a control group of higher-income enrollees (100-300% of poverty) who were not subject to auto-enrollment throughout.
Shepard and Wagner (2022) use this design to infer the marginal enrollees who gain health insurance through the auto-enrollment policy — and who therefore lose health insurance when the policy is suspended. The paper reports the share of enrollees affected, their health status and costs, and model estimates of their value of insurance. The paper also estimates changes in government subsidy spending and uncompensated care in order to calculate net costs for marginal enrollees.
MVPF = 1.0
Net Cost
The paper calculates the net public costs per person-month enrolled in subsidized public health insurance because of auto-enrollment. Net public costs are calculated as:
where:
C_{gross} = Gross government spending = Insurance-paid health care spending for marginal enrollees due to the policy. These costs are entirely government funded (enrollees pay $0 in premiums and copays). The authors observe this directly in their health insurance claims data. These gross costs are $228.32 per enrollee-month.
FE = Estimated government savings on uncompensated care that would have been incurred by the individuals had they been uninsured. The paper estimates these based on two estimates from other research:
- Finkelstein, Hendren, and Luttmer (2019): This paper uses the Oregon Health Insurance Experiment to estimate the causal impact of health insurance on uncompensated care (among other outcomes). Based on their estimates, total uncompensated care costs for low-income individuals are about 59% of gross medical spending while insured.
- Garthwaite, Gross, and Notowidigdo (2018): This paper estimates that about 63.5% of uncompensated care costs are borne by the government, with the remainder borne by private hospitals.
Together, these results imply that:
Putting these estimates together yields Net Cost = Gross Cost – FE = $228 – $86 = $142 per enrollee-month ($1,704 per person-year).
Willingness to Pay
The paper calculates total WTP for health insurance = WTP for enrollees + spillover benefits to private hospitals (via uncompensated care savings)
(1) WTP for enrollees: The paper estimates this using demand estimates calculated for higher-income enrollees (100-300% of poverty) in the same market. The paper uses the fact that subsidies vary discretely with income for an RD strategy to estimate a demand curve for insurance (see Finkelstein, Hendren, Shepard (2019)). Shepard and Wagner (2022) estimate demand separately by age, sex, and health risk groups, and then project these demand estimates onto the lower-income population of marginal enrollees due to auto-enrollment. The paper’s estimates suggest WTP for enrollees = $93 per person-month.
(2) Spillover benefits (uncompensated care savings): The paper estimates savings on uncompensated care when people gain insurance (as in the net cost estimates) and uses the 36.5% share borne by private hospitals as a private spillover benefit. See the net cost section for sources for these estimates.
Therefore: Net WTP = $93 (enrollee WTP) + $49 (spillover benefits) = $143 per enrollee-month ($1,716 per enrollee-year)
MVPF Summary
The MVPF then is Total WTP / Net Costs = $143 / $142 = 1.00 (using non-rounded estimates).
The confidence interval for MVPF is calculated with the point estimate of WTP divided by the lower and upper CI for net costs.
For additional details and assumptions of the calculation, see Table 3 of the paper, in column (2) for “passive enrollees” in the row for “Value-Cost Ratio”.
References
Finkelstein, Amy, Nathaniel Hendren, and Erzo FP Luttmer (2019). “The Value of Medicaid: Interpreting Results from the Oregon Health Insurance Experiment.” Journal of Political Economy, 127(6): 2836-2874. DOI: https://doi.org/10.1086/702238
Garthwaite, Craig, Tal Gross, and Matthew J. Notowidigdo (2018). “Hospitals as Insurers of Last Resort.” American Economic Journal: Applied Economics, 10(1): 1-39. https://www.aeaweb.org/articles?id=10.1257/app.20150581
Shepard, Mark, and Myles Wagner (2022). “Do Ordeals Work for Selection Markets? Evidence from Health Insurance Auto-Enrollment.” NBER Working Paper w30781. https://www.nber.org/papers/w30781
Policy Features
- Category
- Social Insurance
- Sub-Category
- Health Adult
- Beneficiary Type(s)
- Adults
- Average Age
- 34.2
- Average Income
- 26% of FPL in 2009 (about $2,800 to $6,700 depending on family size for FPL. We observe income/FPL but not family size.)
- Country of Implementation
- United States
- Year of Implementation
- 2009
- Empirical Method
- Difference in Differences
- Research Type
- Primary
- Peer Reviewed
- No
- MVPF Publication Link
- www.nber.org/papers/w30781