The literature on the quality of Medicaid has mixed findings—some shows that having Medicaid is better than being uninsured; some shows the opposite. But virtually all of these studies suffer from a statistical issue that makes it impossible to tell whether or not it’s Medicaid or something else driving the results.

This is because there are lots of reasons why people may enroll in Medicaid or decide to remain uninsured—and some of the reasons are completely unknowable to the researcher. For instance, developing a chronic condition that’s expensive to treat might drive some of the uninsured who qualify for Medicaid to the program. Therefore, the benefits or non-benefits of Medicaid might actually be a reflection of the underlying conditions that drove people to apply for the program in the first place.

A group of prominent health economists just released a paper that purports to avoid the same sort of complications with proving causation by randomly assigning people to Oregon’s Medicaid program based on a lottery. However, this should be viewed with caution, because the study’s main conclusions are not based on random assignment into Medicaid, but on self-selection into the Medicaid program.

Here are the basics of the Oregon study. In 2008, 10,000 slots became available for Oregon’s Medicaid program. However, since the program had been closed to new enrollment since 2004, nearly 90,000 people had signed up for a waiting list. To allocate people from the list to Medicaid, the state held a series of lottery drawings. If your name came up, you had 45 days to submit the appropriate paperwork and—voila!—you’re on Medicaid.

Since the point of the study was to explore the impact of having Medicaid on certain outcomes, in an ideal world all (or at least nearly all) individuals initially selected off the waitlist would enroll in Medicaid. Then you’d be able to say something meaningful about Medicaid based on random assignment into the program by asking, “What is the likelihood of having some outcome based on selection by the lottery?”

But that’s not what happened. Instead, 40 percent of the people who were selected never sent back their paperwork. An additional 30 percent of everyone selected didn’t qualify for Medicaid any more, because their incomes were too high. In the end, it took eight lotteries and more than 35,000 drawings to fill the 10,000 slots.

The low take-up of Medicaid caused the authors to split their research question into two: What is the likelihood of having some outcome based on the estimated probability of enrolling in Medicaid? And, what is the likelihood of enrolling in Medicaid based on being selected by the lottery? The question of being selected by the lottery to have the ability to enroll in Medicaid is random. The part about the outcomes being based on enrollment into Medicaid is not random. This is because people enrolling in Medicaid are still self-selecting into Medicaid.

If the sickest of those selected by the lottery are more likely to enroll, I would hope Medicaid could improve access and self-reported health more than being uninsured. However, this debate needs to shift entirely away from comparing those on Medicaid to the uninsured. Instead, we need to know whether people on Medicaid have better or worse outcomes relative to those with private insurance coverage and whether the benefits are worth the ever-higher cost of Medicaid.