NBER Working Paper 23668: Subsidizing Health Insurance for Low-Income Adults: Evidence from Massachusetts
By Amy Finkelstein, Nathaniel Hendren, and Mark Shepard
National Bureau of Economic Research, August 2017
How much are low-income individuals willing to pay for health insurance, and what are the implications for insurance markets? Using administrative data from Massachusetts’ subsidized insurance exchange, we exploit discontinuities in the subsidy schedule to estimate willingness to pay and costs of insurance among low-income adults. As subsidies decline, insurance take-up falls rapidly, dropping about 25% for each $40 increase in monthly enrollee premiums. Marginal enrollees tend to be lower-cost, consistent with adverse selection into insurance. But across the entire distribution we can observe – approximately the bottom 70% of the willingness to pay distribution – enrollee willingness to pay is always less than half of own expected costs. As a result, we estimate that take-up will be highly incomplete even with generous subsidies: if enrollee premiums were 25% of insurers’ average costs, at most half of potential enrollees would buy insurance; even premiums subsidized to 10% of average costs would still leave at least 20% uninsured. We suggest an important role for uncompensated care for the uninsured in explaining these findings and explore normative implications.
From the Introduction
Governments spend an enormous amount of money on health insurance for low-income individuals. For instance, the U.S. Medicaid program (at $550 billion in 2015) dwarfs the size of the next largest means-tested programs – food stamps and the EITC ($70 billion each). Perhaps because of these high and rising costs, public programs increasingly offer partial subsidies for health insurance, requiring enrollees to pay premiums to help cover costs. Partial subsidies are a key feature of market-based programs such as Medicare Part D and the Affordable Care Act (ACA) exchanges, and even traditional low-income programs like Medicaid and the Children’s Health Insurance Program (CHIP) increasingly require premiums for some enrollees. Partial subsidies are also a textbook policy response to adverse selection if a full coverage mandate may not be efficient.
In this paper, we estimate low-income individuals’ willingness to pay (WTP) for health insurance, assess how it compares to the cost they impose on the insurer, and discuss the positive and normative implications for subsidized health insurance programs. We do so in the context of Massachusetts’ pioneer health insurance exchange for low-income individuals, known as “Commonwealth Care” or “CommCare.” Established in the state’s 2006 health care reform, CommCare offered heavily-subsidized private plans to non-elderly adults below 300% of poverty who did not have access to insurance through an employer or another public program.
We first document two main descriptive patterns. First, enrollee demand is highly sensitive to premiums. With each discrete increase in enrollee premiums, enrollment in CommCare falls by about 25%, or a 20-24 percentage point fall in the take-up rate. Second, we find that despite the presence of a coverage mandate, the market is characterized by adverse selection: as enrollee premiums rise, lower-cost enrollees disproportionately drop out, raising the average cost of the remaining insured population.
We use a simple model to analyze the implications of these descriptive patterns.
The model allows us to translate the descriptive patterns into two main results about willingness to pay and costs for these plans. First, even large insurance subsidies are insufficient to generate near-complete take-up of insurance by low-income adults.
These findings suggest that even modest enrollee premiums can be a major deterrent to universal coverage among low-income people. This deterrent is likely to be even larger in the ACA exchanges, in which income-specific premiums are significantly higher than in CommCare.
Second, although adverse selection exists, it is not the primary driver of low take-up. The cost of marginal consumers who enroll when premiums decline is less than the average costs of those already enrolled, implying that plans are adversely selected. But across the in-sample distribution – which spans the 6th to the 70th percentile of the willingness to pay distribution – the willingness to pay of marginal enrollees still lies far below their own expected costs imposed on insurers for either the H or L plans. For example, for the median willingness to pay individual, the gap between the costs of the marginal enrollee and average costs of enrollees explains only one-third of the $300 gap between willingness to pay and average costs. This suggests individuals would not enroll even if prices were subsidized to reflect the expected cost of marginal enrollees. Enrollment is low not simply because of adverse selection, but because people are not willing to pay their own cost imposed on the insurer.
In the final section of the paper, we briefly explore potential explanations for our findings and analyze their normative implications. Back-of-the-envelope calculations using estimates from the prior literature suggest that moral hazard effects of health insurance cannot explain much of the gap between willingness to pay and costs. However, an estimate of the uncompensated care available to the uninsured suggests that this can account for nearly all of our estimated gap between willingness to pay and costs of insurance.
If we interpret our willingness to pay estimates normatively as individuals’ value of insurance, they suggest that most low-income enrollees would prefer being uninsured to having to pay the cost they impose on the insurer. This suggests that subsidies in this market cannot be justified simply as a response to adverse selection. Our results suggest two potential such justifications for subsidies: as an offset to the externalities resulting from uncompensated care (i.e., the Samaritan’s dilemma (Buchanan, 1975; Coate, 1995)), or as a means of redistribution to low-income households.
5.1 Why Is WTP so Low?
We estimate that willingness to pay is less than one-third of average costs. This cannot be fully explained by adverse selection driving a wedge between average costs and costs for marginal enrollees. We estimate that willingness to pay is also always less than half of marginal enrollees’ own expected costs. This stylized fact runs counter to a standard assumption in textbook models of insurance demand – that willingness to pay for insurance equals expected costs plus a value of risk protection.
In principle, moral hazard offers a potential explanation. If the provision of health insurance leads individuals to consume care they wouldn’t have consumed when uninsured (“moral hazard”), their willingness to pay for this additional care will be less than the cost they would have had to pay for it when uninsured. In practice, however, moral hazard is unlikely to be large enough to explain the results. Insurance would have to increase costs by a factor of at least 200% to explain the estimated gap between willingness to pay (Wj) and own costs (Cj). This seems well outside the plausible range of estimates.
By contrast, uncompensated care for the low-income uninsured is large enough to rationalize their low willingness to pay for CommCare. Uncompensated care for the uninsured reduces willingness to pay relative to the gross cost of this insurance, since some share of those costs pays for care that, if uninsured, would have been paid for by third parties. Estimates suggest that the uninsured pay only about 20% to 35% of their cost of care, which is remarkably similar to our estimated ratio of WTP to own costs for the H plan. Not surprisingly therefore, a simple back-of-the-envelope calculation suggests that the magnitude of uncompensated care for the low-income uninsured could close most of the gap between willingness to pay (WH) and own expected cost (CH).
A large role for uncompensated care also can potentially explain differential take-up findings for low- vs. high-income populations. The low willingness to pay we find for the low-income population in Massachusetts contrasts with the findings for higher-income individuals in Massachusetts: Hackmann, Kolstad, and Kowalski (2015) estimate that individuals above 300% of the federal poverty line in Massachusetts are willing to pay the (gross) cost they impose on the insurer. One parsimonious way to rationalize these findings is that only low-income individuals are able to obtain substantial uncompensated care, and this uncompensated care when uninsured reduces the willingness to pay for formal insurance.
From the Conclusion
This paper estimates willingness to pay and costs for health insurance among low-income adults using data from Massachusetts’ pioneer subsidized insurance exchange. For at least 70% of the low-income eligible population, we find that willingness to pay for insurance is far below insurers’ average costs. Adverse selection exists, despite the presence of the coverage mandate, but is not the driving force behind low take up. We estimate that willingness to pay is less than half of enrollees’ own expected costs; thus, even if insurers could set prices conditional on an individual’s willingness to pay, at least 70% of the market would be uninsured.
From a positive economics perspective, our results point to substantial challenges in getting to universal coverage via partially subsidized insurance programs like the ACA’s exchanges. For example, we estimate that even subsidizing premiums down to 10% of insurer costs would generate only 80% coverage. This reality may underlie the incomplete take-up of insurance under the ACA, despite a coverage mandate and generous subsidies. We provide suggestive evidence of an important role for uncompensated care in explaining why willingness to pay is substantially below (gross) insurance costs. This in turn suggests a potential economic rationale for subsidies as a response to the “Samaritan’s dilemma” to offset the implicit tax that uncompensated care imposes on formal insurance (Coate, 1995).
Altruism, the Samaritan’s Dilemma, and Government Transfer Policy
By Stephen Coate
American Economic Review, 1995, vol. 85, issue 1, 46-57
This paper shows that altruism provides an efficiency rationale for public provision of insurance to the poor. The framework is one in which there are rich altruists and risk-averse poor who face some possibility of loss. The government represents the rich and makes transfers on their behalf. With unconditional transfers, the poor may forgo insurance and rely on private charity to bail them out in the event of loss. This reliance on private charity has adverse efficiency effects. These may be avoided if the government makes in-kind transfers of insurance.
Distribution of the Nonelderly Uninsured by Federal Poverty Level (FPL) in 2015
Kaiser Family Foundation
26% – Under 100%
27% – 100%-199%
28% – 200%-399%
19% – 400%+
By Don McCanne, M.D.
Wading through this technical 56 page paper, the following understated observation may be the most important: “our results point to substantial challenges in getting to universal coverage via partially subsidized insurance programs like the ACA’s exchanges.” That is, using income related premium tax credits for plans in the ACA exchanges, even with an individual mandate, is a grave design defect because it can never get us to universal coverage.
This study looked at low-income individuals and showed that, not only did adverse selection result in individuals remaining uninsured, but a much greater factor was that the willingness to pay for a percentage of the premium is far below the insurers’ average costs. Even with 90% of the premium subsidized, 20% of the individuals would remain uninsured, and the percent insured dropped off rapidly as the individuals’ share increased.
The authors cite a reference that indicates that higher-income individuals are willing to pay the cost they impose on insurers. But lest we think that avoiding the purchase of insurance due to cost is limited to lower income individuals, that is refuted by the Kaiser Foundation report that shows that the uninsured are distributed throughout all income levels – 19% of them having incomes over 400% of the federal poverty level. Any proposed policy correction should not be limited to lower-income individuals.
The authors suggest that “only low-income individuals are able to obtain substantial uncompensated care, and this uncompensated care when uninsured reduces the willingness to pay for formal insurance.” The authors cite Coate in regard to the “Samaritan’s dilemma” as a rationale for subsidies, but Coate states, “The government represents the rich and makes transfers on their behalf. With unconditional transfers, the poor may forgo insurance and rely on private charity to bail them out in the event of loss. This reliance on private charity has adverse efficiency effects. These may be avoided if the government makes in-kind transfers of insurance.”
Providing highly technical studies such as this one from NBER advances the science on which we base our policy decisions, but it does not take an MIT economist to come to the conclusion that when a policy is not working, you need to change it. Even generous premium tax credits for ACA exchange plans can never get us to universal coverage, yet government making in-kind transfers of insurance will, providing we include all income levels and fund the program with income transfers through progressive taxation.
Again, subsidizing health insurance cannot ever get us to universal coverage. Single payer will.
Stay informed! Visit www.pnhp.org/qotd to sign up for daily email updates.