It’s becoming increasingly clear that Medicaid is the center of the health policy universe.

Medicaid is the largest single insurer in the United States, covering north of 70 million people with low incomes and disabilities, and is responsible for most of the 20 million additional people covered by Affordable Care Act. The ACA expansion established Medicaid as the bedrock of public insurance and public assistance in America. Now, there’s evidence that it not only expanded health insurance coverage, but the electorate itself.

While disgruntlement among people with subsidized private plans on the exchanges over premiums and cost-sharing has provided much of the political ammunition for attempts to repeal Obamacare, Medicaid’s popularity and its guarantee of coverage formed much of the backbone of the opposition to Republican plans to repeal the law. But even before then, Medicaid played a role in the 2016 election, and may have helped reshape the electorate as we know it, increasing turnout among members of both parties.

That finding comes from new research by Jake Haselswerdt at the University of Missouri, which suggests that the ACA’s Medicaid expansion in 2014 increased political participation. The study, which will be published in the Journal of Health Politics, Policy and Law this month, “find[s] that the increases in Medicaid enrollment associated with the expansion are related to considerably higher voter turnout.” More specifically, the research indicates that for every increase in Medicaid enrollment in a district in 2014, there was a corresponding decrease in the “midterm dropoff” in turnout from 2012 to 2014.

The most immediate conclusion is that President Obama and the states that chose to expand Medicaid—perhaps inadvertently—found a way to solve the doldrums of mid-term elections that have bedeviled politicians forever. The problem has been most significant for Democrats, who often face significant turnout dropoffs in midterms and as a result have hemorrhaged legislative seats at every level across the country. Unfortunately for them, although turnout increased in 2014, the data suggest no party gained a clear advantage from it.

Democratic leaders probably never intended a turnout bump, but they did embrace the Medicaid expansion much more broadly and earlier than Republican governors and state legislatures—indeed, all of the states that haven’t yet expanded are run by Republicans—and they implemented reforms and rollouts that were much more likely to increase coverage. Haselswerdt’s paper emphasizes this last point: the reduction in midterm dropoff in a district was not related directly to the state’s decision to expand per se, but to its gusto and effectiveness at getting new people signed up.

It might be surprising then that Haselswerdt did not find that the turnout effects of expanding Medicaid were a boon for liberal candidates. In fact, “there’s no evidence that this effect had any benefit for Democrats at all,” he told me. While he found that the liberal-minded reforms from mostly Democratic officials did increase turnout, the partisan share of the vote did not increase for Democrats, and 2014 was a bloodbath for the party. That means that the Medicaid expansion increased Republican turnout at least as much as it did Democratic turnout.

The paper offers up a compelling theory for this effect. Haselswerdt conjectures that the most obvious partisan effect of the Medicaid expansion was not the mobilization of people to protect their newly-gained coverage, but of Republican-leaning voters who were opposed to it. And the opposing increase in turnout came from Medicaid’s “resource effects,” or its downstream effects on things like time, money, transportation opportunities for voter registration, and ability to participate in civics among a low-income population that skewed towards Democrats.

“I am skeptical of the idea that there was self-political mobilization among recipients,” Haselswerdt says. Rather, he believes the data shows that the increase in turnout among new Medicaid beneficiaries was more indirect.

There are a couple pieces of research that complicate these findings. For one, Haselswerdt cites research noting the long depressive effect that Medicaid has seemed to have had on turnout among beneficiaries. While it’s well known that poor people generally have low turnout, especially in midterm elections, research from Teresa Toguchi Swartz and colleagues at the University of Minnesota found that means-tested, “stigmatizing” forms of social assistance like welfare are associated with depressed civic participation, while non-means-tested “non-stigmatizing” forms of assistance are not. While that research does not include Medicaid in the realm of “stigmatizing” aid, Cornell University’s Jamila Michener found that Medicaid was also associated with decreases in participation through the same mechanism of social stigmatization.

The study from Haselswerdt could be a short-term blip or an outlier against this field of study that shows that expanded Medicaid should decrease turnout, and Haselswerdt told me that he needed more data from more elections to be sure. But it also could indicate that Obamacare’s reforms fundamentally altered the stigma around Medicaid, either by its expansion to a much larger, whiter, and able swathe of mainstream America, or through its injection of the program into the heart of political discourse.

Focus groups conducted with Trump voters by the Kaiser Family Foundation indicate that it might be the former. As my colleague Olga Khazan writes about the research, “many beneficiaries of the Affordable Care Act’s Medicaid expansion didn’t realize their free health insurance was the result of Obamacare.”

Since some state Medicaid expansion programs are branded as unique state products, and many plans are administered by private managed care companies, many new enrollees don’t even know they are on Medicaid. Thus many of the new Medicaid enrollees, though enabled and empowered to vote by the “resource effects” Haselswerdt cites, were not stigmatized by their new receipt of public assistance and participated in politics accordingly.

A final irony is that many of those newly empowered and aided Trump voters did not associate their own coverage with the coverage they voted to repeal and replace. While Haselswerdt doesn’t have data on the 2016 election, it’s possible that the pathways he mentions enabled low-income Republican voters to potentially vote away their own health care.

It’s important not to read into that point as necessarily self-sabotage, because as Khazan notes, health care might not have been a primary election concern for Trump voters, and some may have expected replacement plans to iron out any disruptions in coverage they predicted. And the data for two election cycles, with one coming just months after the Medicaid expansion, cannot be extrapolated far enough to know for sure. But at the least it is very possible that the Medicaid expansion not only motivated, but enabled both defenders and supporters of the law to go to the polls in 2014 and 2016.

If that dynamic holds—and Haselswerdt is interested in pursuing research to discover if it does—future health policy decisions by both parties will have to reckon with the voices of newly-mobilized low-income people.

It’s been more than half a century since Medicaid was passed, and back then it was seen as a bit of a backup player to the massive guarantee of Medicare to elderly people. But the program has come into its own, and in a post-Obamacare world seems to command almost the same kind of party-dogma-bending influence that the more venerable program does. If its current form affects turnout the way Haselswerdt’s research claims, Medicaid might be an unexpectedly empowering agent for people who’ve long struggled to be involved in American politics.