A new attempt at measuring the value of health insurance

The value of government-provided health insurance in the United States is surprisingly hard to calculate. Researchers can’t use market prices as a measurement of the value of health insurance to individual recipients because Medicare and Medicaid—the two main government health care programs—aren’t available on the open market. A new working paper, however, takes advantage of a unique experiment at the state level to expand Medicaid coverage to value Medicaid as a source of income. The results have broad and important implications for our measurements of income and its distribution.

The current methods for measuring the value of Medicaid as income are at two ends of a spectrum—either the value of the program is ignored when talking about income trends or the value of the program to each recipient is assumed to be its average overall cost to the federal government and state governments. (Medicaid is a joint federal-state program that provides health coverage to low-income and disabled Americans.) The Congressional Budget Office, for example, chooses the second option in its analysis of income trends.

But assuming that every dollar spent on the program is valuable to every recipient equally seems hasty without any evidence. Ignoring the value of the program altogether seems even more obviously unwise. Enter the new paper by economists Amy Finkelstein, Nathaniel Hendren, and Erzo F.P. Luttmer. The economists from the Massachusetts Institute of Technology, Harvard University, and Dartmouth College, respectively, use data from the Oregon Health Insurance Experiment to try to value Medicaid as a source of income.

This experiment grew out of the 2008 decision by Oregon to expand its Medicaid program, which the state couldn’t do for everyone who wanted coverage. So it held a lottery. The state gave data about the winners and losers of the lottery to several economists, including Finkelstein—access that enabled them to evaluate the income benefits of expanded Medicaid insurance.

Finkelstein, Hendren, and Luttmer used the data from the experiment to figure out how households that received insurance change their consumption and out-out-pocket spending on health care. By seeing how these behaviors change for households that received Medicaid through the lottery relative to those who did not, the economists derive estimates for the income value of Medicaid for the recipient families.

In every single estimate, they find that the benefit of Medicaid to one of the recipients is less than the government’s cost of providing the insurance. Specifically, they estimate that for every dollar that the federal government and Oregon spend on Medicaid, recipients receive $0.40 or $0.20, depending on the specific valuation method used, with those who would have provided health care the absence of insurance (an emergency room, for example) reaping the rest of every dollar.

In short, valuing Medicaid as income for individual recipients at the average cost of the program is very much off the mark.

These differences could have a significant impact on how incomes are measured by government agencies. Take the Congressional Budget Office. Remember, in the data used by CBO Medicaid is currently valued at average cost. Medicaid was about 8.7 percent of post-tax-and-transfer income for the average household in the bottom 20 percent of households in 2011. Using the estimates from Finkelstein, Hendren, and Luttmer, the average income of these households would drop by between 5.7 and 7.6 percent.

Moreover, this calculation assumes that Medicare, which was 9.4 percent of the average income for the bottom 20 percent of income earners in that year, is still valuated at average cost. If the value of Medicare is below average cost as well, which seems plausible, then the incomes for those at the bottom would be even lower.

This new paper will very likely start a broader debate about how to value Medicaid, Medicare, and other government health insurance programs. Resolving that debate is very much needed if we actually want to fully understand the U.S. distribution of income among many other important topics.