New health insurance insights

A new analysis of a randomized health insurance program in Oregon sheds light on the value the program has for enrollees and providers alike.

The study, by MIT economist Amy Finkelstein and two co-authors, suggests that adults with low incomes value Medicaid at only about 20 cents to 50 cents per dollar of medical spending paid on their behalf.

“The value of Medicaid for most low-income adults is much lower than the medical expenditures paid by the insurance,” says Finkelstein, the John and Jennie S. MacDonald Professor at MIT and a leading health care economist.

That finding reinforces the results of another, separate study that Finkelstein and multiple co-authors conducted in Massachusetts. In that case, across 70 percent of people in the Massachusetts state health insurance program for low-income adults, their valuation of the program was equal to less than 50 percent of their expected insurance costs. 

While it might seem puzzling that recipients value health insurance at less than the covered medical expenditures, the study also offers an explanation for this: Low-income individuals who do not have insurance still only pay a fraction of their medical costs. In the Oregon data, this figure was roughly 20 percent of medical costs; prior studies have found similar results nationwide. The remainder of the spending on the low-income uninsured comes from a variety of sources, including charity care from nonprofit hospitals, publicly funded health clinics that offer free care, state funding to hospitals for uncompensated care, and unpaid medical debt.

“The nominally uninsured have a fair amount of implicit insurance,” Finkelstein says. “Once you put it in that light, it becomes a lot less surprising that Medicaid spending is valued by them at a lot less than dollar for dollar.”

One further implication of the findings is that a significant portion of public spending on health insurance for low-income individuals effectively acts as a subsidy for health care providers and state programs that cover the costs of uninsured patients.

The new paper, “The Value of Medicaid: Interpreting Results from the Oregon Health Experiment,” appears in the December issue of the Journal of Political Economy. Its co-authors are Finkelstein; Nathan Hendren PhD ’12, a professor of economics at Harvard University; and Erzo F.P. Luttmer, a professor of economics at Dartmouth College.

The previous paper, “Subsidizing Health Insurance for Low-Income Adults: Evidence from Massachusetts,” was published last spring in the American Economic Review. Its co-authors are Finkelstein; Hendren; and Mark Shepard, an assistant professor at the Harvard Kennedy School of Government.

A random walk in Oregon

The latest paper examines a distinctive Medicaid policy that Oregon implemented in 2008. With funding to cover only about 10,000 of eligible adults, Oregon conducted a lottery to decide who would be eligible to apply for Medicaid.

That random assigment of slots using a lottery allowed the researchers to develop a study comparing two otherwise similar groups of Oregon residents: those who had obtained Medicaid coverage via lottery and those who entered the lottery but did not gain coverage. In effect, Oregon had developed a randomized controlled trial, which the scholars used for their research.

Medicaid eligibility regulations and administrative practices can vary by state. In Oregon, adults and children generally qualify for Medicaid when they live in a household with income no greater than 133 percent of the poverty level defined by the U.S. federal government; in 2016, in the 48 contiguous states, that was $11,800 for a single person and $24,300 for a family of four.

Previous studies of the Oregon experiment that Finkelstein has led have shown that, among other things, emergency room use increases among Medicaid recipients, contrary to expectations of many experts.

Being covered by Medicaid also increases patient visits to doctors, prescription drug use, and hospital admissions, while reducing out-of-pocket medical expenses and lowering unpaid medical debt for recipients. Medicaid coverage also appears to lower the incidence of depression, although it does not seem to change the available measures of physical health.

The current study uses data from the prior Oregon studies, as well as state Medicaid records, and survey data from individuals who applied for Oregon’s lottery. The survey data show how much people used health care, including prescription drugs, outpatient visits, emergency-room visits, and hospital visits.

In line with previous studies, the current paper shows that having Medicaid increases total spending on health care — about $3,600 reimbursed to providers annually on behalf of each Medicaid enrollee, compared to $2,721 annually for each low-income uninsured individual. Of that $2,721, the low-income uninsured paid about $569 in annual out-of-pocket costs — the source of the paper’s estimate that uninsured individuals pay about 20 percent of charged costs.

Using this data, the researchers also estimated an annual net cost of Medicaid in Oregon of $1,448 per recipient. This is the average annual increase in health care spending by Medicaid recipients, plus their average annual decrease in out-of-pocket spending. Thus moving a low-income uninsured individual in Oregon onto Medicaid results in a $1,448 increase in insured health care spending on behalf of that person.

Because the Oregon Medicaid program’s reimbursements to health care providers are an average of $3,600 annually per recipient, the researchers estimate that about 40 percent of Medicaid spending underwrites costs incurred by enrollees. The other 60 percent is, as they write in the paper, “best conceived of as … a monetary transfer to external parties who would otherwise subsidize the medical care for the low-income uninsured.”

Simultaneously, the researchers refined their “willingness to pay” metric by using multiple methods to estimate how much having health insurance affects consumer spending generally. These methods yielded three estimates ranging from $793 to $1,675 in annual health care spending for low-income individuals. This is the source of the paper’s conclusion that people value Medicaid at 20 percent to 50 percent of charged costs.

Two approaches, similar results

Significantly, the two studies use different methodological approaches to study different programs in different states, and arrive at similar conclusions. In Massachusetts, the scholars used data from the state’s health insurance program — a forerunner of the federal Affordable Care Act — to see how the share of eligible individuals who signed up for insurance changed as their subsidy level changed.

“Despite a different design and different setting, even though it’s Massachusetts and not Oregon, and different method, we got pretty much the same result,” Finkelstein observes.

Overall, Finkelstein says, it will be valuable to keep learning about the care obtained by uninsured people, as well as the ultimate destination of Medicaid funding, including the 60 percent that is routed to other parties that subsidize care for the low-income uninsured. Understanding who ultimately gets those transfers, she notes, could help illuminate how redistributive Medicaid actually is, as a program intended to benefit lower-income Americans.

Moreover, Finkelstein says, more research will be needed to study how best to provide health care for lower-income Americans.

“Right now we have an implicit, informal insurance system that likely reduces demand for formal insurance but provides a sort of patchwork of care that may not be very good,” Finkelstein says.

Funding for the two studies was provided by the National Institute of Aging, the National Science Foundation, and the Harvard University Lab for Economic Applications and Policy.