Senate Minority Leader Charles E. Schumer (D-N.Y.) and House Minority Leader Nancy Pelosi (D-Calif.), right, speak with the media on Jan. 4. (Kevin Lamarque/Reuters)

On Sunday, the Trump administration signaled its intention to convert Medicaid to a block-grant program, giving states more flexibility in how they finance health care for low-income residents. If implemented as part of an Obamacare repeal, the change would likely result in overall less funding for the states.

Although the details of an overhaul would determine where and how large any cuts would be, President Trump may have reason to worry about the electoral effects of a Medicaid rollback.

Our research shows that a significant portion of Trump’s support in 2016 came from low-income areas that would likely be harmed by cuts to Medicaid. And even though those voters did not abandon Trump during the campaign because of his opposition to Obamacare, an actual reduction in benefits is easier said than done.

The politics of Medicaid expansion

Medicaid has been a major political issue since its creation in the 1960s, but it became even more contentious after the passage of the Affordable Care Act in 2010.

Under the ACA, the program was expanded to include all non-disabled adults whose Modified Adjusted Gross Income (MAGI) is below 138 percent of the federal poverty level. After a 2012 Supreme Court decision, states had the choice to implement the new eligibility standards in exchange for additional federal funds, or to opt out altogether. As of today, 31 states and the District of Columbia have adopted the Medicaid expansion.

Given Trump’s campaign pledge to repeal Obamacare, one might have expected him to perform poorly in states where the ACA’s expansion of Medicaid gave low-income Americans better access to health care. But our analysis suggests that Trump did not lose support among low-income voters in Medicaid expansion states.

How Medicaid expansion played out (or didn’t) in the election

The Medicaid expansion was implemented in January 2014, so we examined Trump’s performance relative to that of 2012 Republican nominee Mitt Romney. In particular, we compared the president’s gains in counties where Medicaid has not been expanded to his showing in counties where more adults are now eligible to benefit from the program. By taking the difference in vote share between Trump and Romney, we tried to capture Republican voters’ sensitivity to the Obama administration’s health-care policies.

We also collected demographic and financial data from IPUMS-CPS, an integrated set of individual and household-level variables in the United States. Following IRS guidelines, we estimated the national share of non-disabled adults whose MAGI is below 138 percent of the federal poverty level, the threshold for Medicaid expansion eligibility.

Then, we weighted these shares using a county-level indicator of poverty, which produced a measure of the degree of potential eligibility for the Medicaid expansion in each county, shown in the map below.


Data: IPUMS-CPS and ACS. Figure: Cerrato, Ferrara and Ruggieri.

Low-income households are concentrated in the Southeast (Mississippi, Louisiana, Arkansas, Alabama, Georgia, Florida, South Carolina, North Carolina), in the Southwest (New Mexico, Arizona, California), and in the Northwest (Oregon, Washington, Montana). Thus, darker counties are located in both traditionally Democratic and Republican states.

Finally, we performed a regression analysis to estimate the effect of extended Medicaid eligibility on the shift in the Republican vote share between 2012 and 2016. We weighted the observations for each county’s population and added controls for ethnic and racial composition and educational attainment, as well as state fixed effects.

Perhaps surprisingly, Trump’s gains were uniform across counties with more low-income households, regardless of whether they were in Medicaid expansion states. Indeed, on average Trump outperformed Romney in traditionally Democratic states that extended health-care eligibility.

This counterintuitive result corroborates one of the main trends of the 2016 presidential election. Overall, Trump performed better than any other Republican candidate in the recent past among low-income voters. His opposition to Obamacare had a negligible effect in areas that one would expect to be affected by Medicaid expansion.

Why Medicaid cutbacks could be risky

On the one hand, this could suggest that Trump has little to worry about if the GOP converts Medicaid to a block grant, effectively reducing the size of the entitlement program. After all, if low-income voters were not concerned about Trump’s opposition to Obamacare during the campaign, why would they be now?

But two factors suggest caution might be in order. First, a pledge to roll back a welfare benefit may not have the same impact as its actual repeal. As political scientist Paul Pierson has argued, “frontal assaults on the welfare state carry tremendous electoral risks.”

One reason is that interest groups and voters often oppose direct threats to welfare programs. And already, the specter of a reduction in health-care benefits appears to have mobilized unhappy constituents in some parts of the country.

Second, to the extent that a reduction in Medicaid benefits weakens Trump’s support among low-income voters, their shifting allegiances could prove pivotal, either in the 2018 midterms or in the 2020 presidential election. This is especially true in light of his narrow margin of victory in key battleground states.

The Trump administration has not yet revealed the details of its health-care policy, nor is it clear what Republican leaders in Congress will pursue. But even though Trump’s opposition to Obamacare during the campaign seems not to have hurt him among some of its main beneficiaries, welfare state retrenchment is tougher to execute than to announce.

Andrea Cerrato and Francesco Ruggieri are research professionals at the University of Chicago Booth School of Business.

Federico Maria Ferrara is a PhD candidate in political science at the University of Geneva.