Quick, what are the most racially charged elements of voting? There is a familiar roster of complaints: felon disenfranchisement, which, given the racial disparities in the justice system, disproportionately affects minorities. Voter-ID laws and other restrictions on voting hours. Gerrymandering, which concentrates minority voters into districts.

But what about mortality rates? It’s not something that enters the political discussion much, but a new paper, “Black lives matter: Differential mortality and the racial composition of the U.S. electorate, 1970-2004,” argues that the racial gap in mortality rates could have a major impact on national politics.

The premise of the paper, by researchers at Mathematica Policy Research, the University of Michigan, Stanford, and Oxford, is simple: Unless you live in Chicago, you can’t vote when you’re dead. Since overall black health outcomes are worse than white ones, and life expectancies are lower for black Americans, that must have an effect on the results of elections. The team crunched the numbers from 1970 to 2004 and calculated “excess deaths” among African Americans:

The total number of black deaths would have been reduced from 8.5 million to 5.8 million if blacks faced the same mortality schedules as whites. Thus, 1 out of every 3 black deaths occurring within this time period was an excess death.

Of the 2.7 million black excess deaths, we project a total of 1.87 million hypothetical survivors to 2004, 1.74 million of voting age, about 1 million of whom would have been voters.

For comparison’s sake, and to see how significant that is, 1 million is also the number of African Americans who have served time but remain disenfranchised, according to a 2010 analysis.

What effect would that have on elections? The researchers assumed those 1 million extra voters would have behaved similarly to the surviving black population in terms of both turnout and partisanship, controlling for gender, age, and geography. Their votes wouldn’t have changed the outcome of the 2004 presidential election between George W. Bush and John Kerry—although Kerry would have netted more than 86 percent of the votes they cast. (As Andrew Prokop notes, but the authors do not, the 2000 election was close enough that the missing black voters would likely have swung the outcome in Florida and, with it, the national electoral vote.)

But it would have helped Democrats beat Republicans in seven close Senate elections and 11 close gubernatorial elections between 1970 and 2004. That would have handed Democrats continuous control of the Senate between 1986 and 2002—a pretty large impact.

One of the interesting things here is who the missing voters are. Although media attention has rightly focused on the deaths of young black men like Freddie Gray, the big difference in mortality has nothing to do with violent deaths from street violence, police brutality, or drugs. “In fact, the predominant and persistent driving force behind US black/white mortality disparities is the unequal distribution of chronic morbidity among young through middle aged adults,” the paper notes:

Age of Death by Race

Rodriguez et al.

The 40-65 range highlighted in the graph above also happens to match up with the peak age of voting. (Naturally, the longer someone lives, the more opportunities they have to vote.)

There’s a public-policy catch-22 at work here. The authors note that there is a range of factors that help determine racial disparities in mortality. Residential segregation and lower levels of public investment in black neighborhoods reduce lifespans. Black neighborhoods tend to have greater exposure to pollution, are less pedestrian-friendly, have higher food costs, and see more tobacco and alcohol marketing. Access to healthcare is also sharply divided by race. In short, there is a range of problems that could be addressed by public policy. But altering public policy requires voting, and the present policies have combined to remove a chunk of the voting-age population that would otherwise have benefited from their repeal.

Of course, there’s an element of parlor game in this study—you, too, can play along at home, as long as you’ve got access to the General Social Survey, exit polls, and Microsoft Excel. There are all sorts of ways to design a different electorate. What if the voting age returned to 21? Or what about differences in mortality rates based purely on socioeconomics? Better-educated and wealthier people tend to live long. What if mortality rates among (more conservative) poor southern white voters were the same as those for (more liberal) wealthy Manhattanites?

But black votes are a particularly interesting and important area of focus, both because of the long history of efforts to prevent or suppress black voting, and because of the partisan homogeneity of African Americans. Electoral impact isn’t the only reason to worry about racial disparities in health outcomes, much less the best one, but the vicious public-policy cycle described here makes a compelling case that mortality rates deserve more attention in conversations about race and voting.