The gerrymandering wars are about to resume. Over the next year, every state in the country will have to redraw its congressional and legislative districts. In anticipation of redistricting, Republicans are eyeing a new tactic: For decades, states have equalized the numbers of people their districts contain. But the GOP is now pushing to equalize districts’ citizen voting-age populations instead. Under this approach, noncitizens and children would be invisible for remapping purposes. Only adult citizens would count.

Republicans think that by omitting non-adult citizens from districts’ populations, they would slash the representation of minorities and Democrats. Their logic goes like this: At present, many districts with non-White or Democratic legislators have relatively low proportions of adult citizens. So these districts would have to grow in size — and shrink in number — to acquire enough adult citizens to hit their new population targets. Conversely, many districts with White or Republican legislators have higher shares of adult citizens. These districts would shed some of their residents into adjacent districts in a world where adult citizens, not people, had to be equalized. These dispersed voters would tilt their new districts in a conservative direction.

But are these forecasts accurate? Remarkably, no one knows. To date, no academic study has examined the consequences of changing the unit of apportionment from all people to adult citizens only. In a forthcoming article, the two of us investigate this topic for the first time. Our findings confirm some — but not all — of progressives’ fears about the effects of switching the apportionment base. Minority representation would drop sharply if states equalized adult citizens rather than people. But the partisan balance of power would be largely unaffected.

Some background: In a series of 1960s cases, the Supreme Court announced the one-person, one-vote rule, which requires districts to have about the same population. However, the court didn’t clarify what exactly has to be equalized — people, adult citizens or something else. A few years ago, a group of conservative plaintiffs returned to the court, arguing that the Constitution compels the equalization of adult citizens. The court unanimously rejected this claim. But several justices suggested that, even if states don’t have to equalize adult citizens, they may do so at their discretion.

The Trump administration did everything it could to induce states to make this choice (though its efforts ultimately fizzled). First, it tried to add a question to the 2020 Census asking respondents if they’re U.S. citizens. But because Commerce Secretary Wilbur Ross dissembled about his motive for adding the question (falsely saying it was necessary to comply with the Voting Rights Act), the justices kept the query off. Next, the Trump administration launched a multiagency initiative to compile citizenship data from existing government records. But President Biden took office before this project could finish — and promptly terminated it.

As a result, states wanting to change their unit of apportionment won’t have access to detailed citizenship information from Washington. They could still forge ahead, though, using the less precise output of the federally administered American Community Survey. Or they could conduct their own state-specific censuses, which could ask about citizenship and thus enable the equalization of adult citizens rather than people.

To find out what would happen if states made the switch, we instructed a computer algorithm to generate millions of statehouse maps for the 10 states with the smallest proportions of adult citizens. The algorithm incorporated line-drawing rules like compactness, respect for county boundaries, and compliance with the Voting Rights Act. But half the maps equalized people, while the other half did conservatives’ bidding by equalizing adult citizens instead.

Our results for minority representation were striking. Across all 10 states, the fraction of districts where minority voters can elect their preferred candidates (usually either Black or Latino, depending on the district’s population) fell by an average of three percentage points when the apportionment base changed from people to adult citizens. In major states such as Arizona, Florida, New York and Texas, this decline exceeded six percentage points. In Texas specifically, roughly 10 minority districts disappeared between the equal-person and the equal-adult-citizen simulations. These districts’ elimination would undo overnight a generation of slow diversification in the Texas Legislature, rendering the body unreflective of the Texas population.

Our partisan findings, however, were considerably less dramatic. Across all 10 states, the share of Republican districts rose by an average of just one percentage point when we switched the unit of apportionment from people to adult citizens. True, Republicans benefited more in a few states, Texas in particular. But in most states we studied, including Arizona, California, Georgia, Illinois and New York, there was effectively no difference in the parties’ fortunes between the equal-person and the equal-adult-citizen simulations. This was because many of the minority districts that vanished from one simulation set to the other remained Democratic. They would typically be represented by non-White Democrats beforehand, and by White Democrats afterward.

Conservatives skeptical of this conclusion might object that our algorithm mimicked the behavior of a nonpartisan mapmaker. Maybe a Republican gerrymanderer would be able to do better (for Republicans). To test this possibility, we reran the algorithm with the additional instruction of maximizing Republican seats. But this extra parameter made no difference. Of course, the maps now included many more Republican districts than before. But the equal-adult-citizen maps still had an average of just one percentage point more Republican districts than the equal-person maps. Even a Republican gerrymanderer, in other words, couldn’t squeeze many more seats out of a different apportionment base.

These results create legal risk for states that choose to equalize adult citizens rather than people. These states should expect their new policies to significantly reduce minority representation — but not the numbers of Democrats elected. In that case, though, courts could easily find that the states changed their unit of apportionment for racially discriminatory reasons. Partisanship, after all, wouldn’t be a plausible explanation for moves with minimal partisan implications.

Fortunately, this scenario isn’t hard to avoid. Nobody would impugn the motives of states that didn’t disrupt the redistricting status quo of the last half-century. Leaving well enough alone, then, would stop one type of litigation in its tracks. It would also preserve the hard-won diversity of those who represent us.