July 30 marks the 50th anniversary of Medicare and Medicaid, the two essential pillars of public health insurance in America. Medicare, for all of its shortcomings and enormous costs, has become politically unassailable—as dear to Tea Party Republicans as it is to single-payer Democrats. Medicaid has also steadily grown. Including the closely related Children’s Health Insurance Program, Medicaid now covers more than 70 million people. Yet it has never attained the same popularity or political security as Medicare or shed the stigma of “welfare medicine” that it acquired at its founding.
Whatever happens in next year’s presidential election, Medicaid will be the center of a great political fight. If the Democrats win, they should seek to expand the program to cover nearly four million poor and uninsured Americans shut out of the Affordable Care Act’s Medicaid expansion. If the Republicans win, they will likely reduce Medicaid funding, and perhaps convert the whole program into a block grant. Because Medicaid will be a main political battleground, it is important to understand what’s right and what’s wrong with the program.
How Valuable Is Medicaid?
An important recent paper by Amy Finkelstein, Nathaniel Hendren, and Erzo Luttmer, “The Value of Medicaid,” tries to determine how much it is worth, but in a narrow way whose baseline assumptions understate the value of the program. The authors use data from the Oregon Health Insurance Experiment. The experiment worked like this: Oregon had more poor and uninsured people than they could insure under Medicaid. So they held a lottery for who could apply for Medicaid coverage. Finkelstein and her colleagues realized that this lottery was in effect an experiment that would allow them to study the impact of Medicaid coverage on the previously uninsured. Following a reasonable—but contestable—economic methodology, Finkelstein and colleagues estimated an equivalent dollar-value to the improved health observed among lottery winners. The authors then explored what happened to a dollar that was spent on providing Medicaid: Who actually benefited and by how much? And would recipients have been willing to pay for their Medicaid benefits if they had to?
1. Uninsured people who get Medicaid only gained from 20 to 40 cents in value from each dollar spent by the government.
2. A principal reason why the benefit of getting insured was so small is that when uninsured people received care, they typically paid only 20 cents on the dollar for those services. Safety-net providers, state or local government, friends, relatives, or someone else absorbed the remaining costs. When a recipient became insured, Medicaid paid some of these costs at the rate of about 60 cents per dollar.
3. Because a large fraction of Medicaid expenditures financed care that recipients would have received anyway (for example, by leaving bad debt at hospitals), it is unclear whether recipients themselves would have been willing to pay the full costs of Medicaid.
While the Oregon Health Insurance Experiment was a well-designed study, too few of the study participants experienced serious medical conditions to investigate how Medicaid affected their health or quality of life. In addition, the experiment—and thus Finkelstein, Hendren, and Luttmer’s study—did not examine how Medicaid affected recipients’ families or the health-care institutions that care for low-income uninsured patients. Medicaid’s long-term benefits as an investment in infants and children were also beyond the scope of the Oregon experiment.
But perhaps the most important limitation of the study stems from an assumption that many readers would be unlikely to notice. Finkelstein and her colleagues placed a very low value—$25,000—on a year of additional life for Medicaid beneficiaries. The typical threshold used in health services research is much larger, in recent studies far above $100,000 per additional year of (healthy) life. Yet because the median income of the Oregon study participants was about one-fourth of the median income in the United States, the researchers chose to value an additional life-year at about one-fourth of the usual threshold. This assumption powerfully frames everything that follows in this analysis. After all, if you start out by assuming that Medicaid beneficiaries’ lives are worth very little, you will find that it is not worth spending much money to prolong them.
These authors did not use a low value-of-life threshold because they are moral monsters—they are not. Rather, they defined this threshold based on reasonable assumptions about what low-income recipients themselves would have been willing to pay, had they been spending their own money for their Medicaid benefits. Poor people aren’t willing to spend as much as rich people would spend for access to the same health care. Given the choice between a Medicaid benefit that costs $4,000 and $4,000 in simple cash, many or most low-income people might well prefer to take the cash.
Had these authors valued the health of poor patients as highly as health services researchers typically value the health of the average patient, their results would have been quite different. The authors know this and to their credit, they show in an appendix that Medicaid’s health benefits appear much more financially valuable when one employs standard $100,000-plus thresholds used in medical cost-effectiveness research.
Although Finkelstein, Hendren, and Luttmer’s baseline assumptions are methodologically defensible, they have radical implications that are rarely so bluntly applied in other domains of health-policy research. Choices about how to financially value the health of poor people relative to the health of others are inevitably both politically and morally freighted. It strains credulity, for example, to imagine American policymakers using this low a value for life when analyzing mammography, prostate cancer treatment, or implantable cardiac defibrillators for seniors.
So What Have We Learned About the Value of Medicaid?
Finkelstein and her colleagues are right that a dollar of Medicaid does not necessarily provide a full dollar’s worth of benefits to a Medicaid recipient in the strict sense that the recipient might prefer the cash to the insurance coverage. They are also correct that an important share of the value goes to others rather than to the recipient herself. What are the implications?
The authors are careful not to make any normative statements based on these findings, but others such as Michael Cannon, Tyler Cowen, and John Graham have done so. They make two arguments: First, Medicaid is an inefficient way to benefit the poor. If a Medicaid dollar results in only 20 cents in benefit to a previously uninsured person, wouldn’t it be more efficient to simply give that person a dollar? And, second, Medicaid is actually a subsidy for people other than those it ostensibly helps.
We see matters rather differently. Economists have long understood that poor people would prefer cash to subsidized health insurance (especially if they can still get health care for free). So why does every developed country, including the United States, subsidize health insurance for the poor? Part of the reason is that those countries have broader moral and public-health criteria for thinking about health insurance and poor people’s lives. Universal health care expresses a commitment to the well-being of fellow citizens. Everyone should have access to a decent minimum of care; caring for others in distress is a primary expression of human solidarity.
Moreover, such programs, including Medicaid, bring measurable and significant improvements in public health that have wide social—and economic—consequences.
Ben Sommers and his colleagues found that previous expansions of state Medicaid programs significantly reduced all-cause mortality in those states, compared to matched states that did not expand health insurance. By one credible estimate, Massachusetts’s Romneycare insurance expansion prevented about one death per year for every 830 adults newly enrolled in health coverage. If these estimates carry over to the 16 million Americans newly-insured under the Affordable Care Act, health reform is now preventing about 19,000 deaths every year. That’s more than the annual lives saved by seat belts, frontal air bags, motorcycle helmets, child restraints, and minimum drinking age laws—put together.
Moreover, the counterfactual proposal to give every poor person the cash value of Medicaid (but no health insurance) is a valuable thought experiment but an absurd policy proposal. At the margin, many poor people would trade some of their Medicaid benefits for some cash. But such trades are rarely, if ever, actually on offer.
Voters want to provide for the health of their neighbors and to protect these neighbors from the financial ruin that often follows serious illness. When my neighbor requires $50,000 for a life-saving kidney transplant, that need has special urgency, as do other basic necessities such as nutrition and shelter. The moral interest of the community does not hinge on my neighbor’s willingness to pay. We respond to critical needs because we understand the consequences of failing to meet them, not just for the individual in question but for all those whose lives are connected to hers.
Should We Leave Medicaid as Is?
The foregoing arguments help to explain why most citizens of wealthy democracies are committed to universal health insurance, despite its inefficiencies. Of course, these arguments do not require us to support Medicaid as it is currently structured.
The analysis by Finkelstein and colleagues should prompt us to ask how public health insurance could be restructured to deliver better value to both the recipients and society. Proposals to improve these programs include some of the innovations featured in the ACA, such as “medical homes” for patients that attempt to provide better-coordinated care. Medicaid might be extended to require coverage of other services such as dental care, and it could be more effective in promoting beneficial health behaviors. Greater use of comparative effectiveness research in the design of Medicaid benefits could promote individual and population health. We should also consider schemes that would replace Medicaid altogether, including Republican proposals to enroll many low-income people in health insurance exchanges, as long as these approaches guarantee every American access to a decent minimum of care.
What about the finding that much of the value of Medicaid goes to people other than the recipient? The answer to this question depends who is actually getting these dollars, and what they do with them.
When Medicaid relieves burdens on loved ones who are covering the medical bills of an uninsured family member, the program is promoting the security and well-being of American families.
If the transfer is going to community safety-net institutions and other charity-care providers, that transfer may be essential to sustain the health-care institutions that serve the indigent and the uninsured. Medicaid payments for what otherwise would have been uncompensated care benefit the larger community, as well as individual providers.
Financial difficulties of major public hospitals in states that reject ACA’s Medicaid expansion underscore Medicaid’s importance in stabilizing fragile networks of safety-net care. A recent Chicago Tribune story, “Obamacare gives Cook County Health a financial boost,” notes that for the first time, the majority of patients in the county health system were insured, primarily through the Medicaid expansion.
Such coverage expansion also makes possible a properly integrated system of care. Thirty years ago, childbirth was the largest single category of uncompensated care. During the 1980s and early 1990s, Medicaid was expanded to provide a near-universal entitlement to prenatal and neonatal care. This expansion made possible the current system of neonatal intensive care units that has markedly reduced infant mortality.
But not every possible transfer to a non-recipient is benign. In some states, Medicaid funds do not pass directly from the state government to people who provide care. Instead, they pass from the state through third parties such as managed care plans, and then to providers, often at below-market rates. If so, some of the transfer is to such plans, which should be carefully scrutinized to ensure that they deliver substantial benefit to recipients.
How can we be sure that a reformed Medicaid (or an alternative) would provide recipients with more value than the current version? We can't be sure. But here we find the last lesson from Finkelstein and her collaborators, and, more generally, from the Oregon experiment. Despite the limitations to the Oregon data, it is far more valuable to have them than to have none at all. Future Medicaid reform must be designed with powerful data-gathering exercises built into the process, so that we can evaluate program performance against hard benchmarks. Funding mechanisms such as Medicaid provide a key point of leverage for introducing reforms and gathering the data to evaluate those changes. We hope to see many future experiments like the one in Oregon, and many other papers of corresponding high quality that force us to consider what we are really buying within our $2.9 trillion health-care system.