Advocates for universal health coverage frequently claim that they’re out to improve health and save lives. Just one problem: Despite the trillions of dollars we’ve spent on public health insurance programs, there’s very little strong evidence to suggest that subsidized health insurance actually improves health. Indeed, the push for universal coverage may be preventing other, more effective health measures. So here’s the question: Is the drive for universal health insurance actually making people worse off?

That’s the alarming possibility raised by Michael Cannon, the health policy director at the Cato Institute, at a forum hosted by the Kaiser Family Foundation in Washington, D.C.

Cannon’s comments came in a discussion of a recently released study attempting to gauge the effects of Medicaid—the joint federal/state health insurance program for the poor and disabled—on low-income adults in Oregon. Researchers from Harvard took advantage of the state’s recent Medicaid expansion to run a rare randomized trial experiment that would allow them to determine with a high degree of reliability just what the effects of Medicaid actually are on health, financial security, and utilization of care.

Because here’s the thing: Prior to this study, we didn’t know. We didn’t know despite the fact that Medicaid has existed since 1965, and despite the fact that last year’s health care overhaul expanded Medicaid coverage to all individuals up to 133 percent of the poverty line—with the biggest increase concentrated in childless, low-income adults. Indeed, despite the fact that the program already costs taxpayers roughly $400 billion each year, no one has ever never run an experiment like this before. As the website for the Oregon Health Insurance Experiment happily declares, it is “the first randomized controlled experiment to examine the causal effects of having some type of insurance coverage versus having no insurance at all.”

What Cannon suggests is that the fact that we are only studying this now “shows that the push for universal health insurance is not about improving health and security.” If it were, he says, we’d have seen some of the funding that now pays for coverage used to determine whether that coverage is actually producing the desired effects. And we’d have seen a far more rigorous attempt to spend public money on programs and benefits that have actually been proven to improve health outcomes and save lives. Instead, we’ve sacrificed the possibility of better care for the nebulous, uncertain effects of expanded coverage.

The Oregon study helps us understand what those effects really are. The clearest benefit is improved financial security. This is not a huge surprise; health insurance insulates individuals from large, health-related financial shocks. Even still, that benefit may be counteracted somewhat by downward pressure on savings created by the program. Researchers have found that larger Medicaid benefits correlate with lower personal asset levels; tying benefits to asset tests makes the effect even larger.

Meanwhile, Medicaid’s health benefits are far less certain. On the only objective measure of health the study has yet to report—mortality—there was no significant improvement. (A report on a wider array of objective health measures is due out next year.)

The study did report a substantial increase in self-reported health status. But the correlation of self-reported improvements to objectively better health, or even to increased health care access and utilization, is less clear. Two-thirds of the increase in self-reported health status came almost immediately, after individuals had been provided coverage but before they’d sought significant medical attention.

Meanwhile, despite the study’s randomized design, it was only able to test the effects of Medicaid on a particular population: relatively poor adults—with annual incomes in the range of $13,000—most of whom were quite unhealthy compared to the rest of the population, reporting an average of 10 sick days each month. Last year’s health care law, in contrast, will expand Medicaid to an estimated 16 million individuals below 133 percent of the poverty line, many of whom will not be quite so sick, or even so poor. The population examined in Oregon, says Cannon, is the “most likely to benefit from coverage.” Which he argues probably means that “whatever results you see are likely to be the high water mark” in terms of health improvements when expanding coverage to various populations.

The Oregon Health Insurance Experiment is now the gold standard in studies on the effects of Medicaid. Even still, it has limits. Medicaid’s systemic effects on employment and assets, on health innovation and administration, are harder to measure, if not impossible. Testing these broader effects would require comparing a world with Medicaid to a world without.

Nor has the study answered the question of whether Medicaid’s health coverage actually improves health. “The effect of expanding health insurance on health is a priori ambiguous,” says Cannon, though we still need to wait to see what the objective measures say.

What it does highlight, however, is how much we don’t know about Medicaid’s effects—and, despite having spent trillions on the program, how much we’ve never bothered to find out.

Peter Suderman is an associate editor at Reason magazine.