How Much Will Gerrymandering Impact the 2026 Midterms?
Research finds that redistricting and geographic sorting are making elections less competitive
- Computer simulations of nonpartisan district maps allow researchers to separate the effects of gerrymandering from political geography; the research finds widespread partisan gerrymandering but only a small net national seat advantage.
- Geographic polarization and gerrymandering are making congressional elections less competitive, with the number of highly competitive districts declining substantially over time.
- Weakening Voting Rights Act protections can reduce Black voting power in parts of the South, while nonpolitical redistricting commissions can reduce partisan bias.
As the 2026 US midterm elections approach, voters across the political spectrum are concerned about the impact of gerrymandering, the partisan manipulation of electoral boundaries. Since the 2024 presidential election, California, Florida, Texas, Missouri, North Carolina, Tennessee, and Ohio have redrawn their congressional districts to the advantage of each state’s ruling political party. (Districts in three more states—Louisiana, Alabama, and Utah—were redrawn due to litigation.) In the more than 50 years prior, only two states had voluntarily redistricted mid-decade, between the decennial US census.
As founder of the Algorithm-Assisted Redistricting Methodology (ALARM) Project, Edith and Benjamin Geisinger Professor of Government and of Statistics Kosuke Imai, PhD ’03, dissects gerrymandering through the use of computational methods and also helps to develop tools election officials can use to draw districts fairly. He spoke with Harvard Griffin GSAS Communications about his recently co-authored research that found a steep decline in competitive congressional districts due to both gerrymandering and partisan geographic sorting, new work that measures the impact of weakening the Voting Rights Act, and how both are shaping the upcoming midterm elections.
In 2023, you and a group of social scientists from Yale, Princeton, the University of Michigan, and elsewhere published a paper on how gerrymandering and geographic polarization have reduced electoral competition. What was your method and what did you find?
We wanted to estimate how biased the redistricting plans enacted after the 2020 Census were. The tricky thing is that you cannot simply compare plans across states because states’ geography is different, the number of seats is different, and even the rules of redistricting are different. You also cannot simply compare a new plan with the previous plan because people move, the electoral environment changes, and sometimes the rules of redistricting change. Those comparisons are not a valid way of figuring out how biased a particular redistricting plan is.
The way we tackled these problems was to use a computer algorithm to simulate alternative redistricting plans under a set of nonpartisan criteria. Most states are supposed to follow a certain set of rules, so we generate a representative sample of redistricting plans that could be adopted under the rules that each state is supposed to follow. Once you have a large set of those plans, you can compare that nonpartisan baseline with the enacted plan in terms of, say, the expected number of Democratic or Republican seats. That allows you to measure partisan bias. You can also measure the degree to which minority voters are represented under the enacted plan versus the nonpartisan alternatives.
What do you mean “nonpartisan alternatives”?
We do not use any electoral or partisan information. We do not use data from past elections, and we do not use information about who is registered with which party. We follow the traditional criteria that many states are supposed to observe, such as equal population, compact districts, and minimizing the splitting of administrative boundaries. We also impose state-specific criteria such as complex county-splitting rules used in Ohio. There were also protections for minority voting rights under the Voting Rights Act, though this requirement was significantly weakened by a recent decision by the Supreme Court. Those are criteria that are not based on partisan information, so that is what we call nonpartisan.
So you first apply rules such as equal population, compact districts, and protections for minority voting rights, and only afterward examine the partisan makeup of the resulting districts?
Yes. At the end, once you generate the simulated districts, you bring in the partisan measure and evaluate whether the enacted plan is biased in a particular direction relative to the simulated plans.
That does not mean political geography itself cannot favor one party over another. For example, in Massachusetts, about one-third of voters are Republican, but there is no Republican representative in Congress. One reason is that Republican voters are spread across the state rather than concentrated in a particular area. If you try to create compact districts, it is difficult to draw a district with a Republican majority.
What did you find about how the enacted plans deviated from the nonpartisan alternatives?
We found that both parties engage in partisan gerrymandering. There are states such as Texas where Republicans are favored, and there are states such as Illinois where Democrats are favored. Overall, those effects largely cancel each other out. We estimated a net gain of about two seats for Republicans, but there is a great deal of cancellation because some states favor Democrats and others favor Republicans.
We observed that the number of . . . safe districts has increased dramatically over the last 10 or 20 years as political geography has become more polarized—by around 25 percent.
You mentioned proportional representation. Looking at the total Democratic and Republican votes cast in 2020, 2022, and 2024, was representation in Congress roughly proportional to the two parties’ vote totals?
At the aggregate national level, we found that it takes somewhat more votes for Democrats to obtain 50 percent of the seats. Our estimate was about 1 percentage point: Democrats would need roughly 51 percent of the votes to obtain 50 percent of the seats. We estimated that Democrats would win eight fewer than half of the seats in Congress if they received 50 percent of the vote. There was therefore some bias against Democrats.
To what extent is that also a product of partisan sorting by geography, with Republicans increasingly living near Republicans and Democrats near Democrats?
One thing we have been doing is extending this exercise back about four decades, to 1990, to examine trends in political geography and redistricting. What we have found, particularly over the last couple of decades, is that residential segregation by partisanship has worsened. That does not necessarily mean people are physically moving; it may also mean people are changing their political preferences. It is probably a combination of the two.
Rural areas are becoming more Republican, while cities are becoming more Democratic. That creates geographic polarization at the national scale. The changes are particularly striking in large states such as Texas and California, as well as in parts of the Midwest. Geographic polarization also creates opportunities for gerrymandering because, once voters are concentrated, it is easier either to pack them into one district or crack them across multiple districts.
What other effects on electoral competition did you find?
Geographic polarization reduces electoral competition even in the absence of gerrymandering. If Democrats are concentrated in cities and Republicans in rural areas, then under a system of compact, single-member districts you naturally create more safe seats: districts that are overwhelmingly Democratic and districts that are overwhelmingly Republican. We observed that the number of those safe districts has increased dramatically over the last 10 or 20 years as political geography has become more polarized—by around 25 percent.
Does that create a vicious cycle in which primaries matter more, the most ideological voters participate, and politicians become less responsive to the broader electorate?
That is my concern. Moderate voters may feel that they do not have much influence. If you are an independent and do not vote in a party primary, and you live in an overwhelmingly Democratic or Republican district, you may feel that you have little influence over the outcome. That can discourage political participation. Politicians then have less incentive to respond to the broader electorate because they only need to satisfy a relatively small fraction of voters, often the more ideologically extreme members of their own party.
Mid-decade redistricting has happened before. What is unusual is the intensity of the current wave and the explicit partisan goal behind much of it.
Earlier this year, you released another study with Princeton’s Christopher Kenny, PhD ’25, the University of Michigan’s Tyler Simko, PhD ’24, and Harvard College student Brian Zhou that looked at the impact on minority representation of the weakening of the Voting Rights Act. What did you find?
In the South, Black voters lose voting power. In states such as Alabama, Louisiana, and South Carolina, Black voters are highly concentrated, including in the Black Belt. Voting Rights Act protections helped guarantee representation for those voters, and many of the resulting seats have been held by Black members of Congress. Those seats can be endangered when those protections are weakened, so Black voters in the South are more likely to lose the ability to elect their preferred candidates.
In other areas, however, the opposite can happen. Minority voters may live near white voters with similar ideological preferences, especially in cities. Even without drawing a majority-minority district, they may still be able to elect candidates of their choice. In those places, minority voting power can increase slightly. At the national level, the partisan consequence of the Voting Rights Act appears to be quite small.
Are you saying that minority communities may lose representation in the South but could actually gain some ability to elect preferred candidates elsewhere? Where?
Yes. In parts of the South, minority representation is likely to be lost. In other places, including Washington State, Michigan, and parts of California, minority voters may do as well or better under compact districts drawn without explicit consideration of race. In those places, simple compact districts can create what we sometimes call opportunity districts, where minority voters are not a majority but are joined by enough white voters with similar ideological preferences to elect the candidates they prefer. So, the national net effect is close to a wash, perhaps with a small loss overall.
That does not mean the change is harmless. The impact depends on where you are. Some groups, particularly in the South, are harmed, while groups elsewhere may gain. A small national net effect can conceal important local effects.
Your research on gerrymandering came out before the latest wave of mid-decade redistricting. Historically, how unusual is it to have this many states redrawing congressional lines between censuses? And which party stands to gain more from it?
Mid-decade redistricting has happened before. What is unusual is the intensity of the current wave and the explicit partisan goal behind much of it. I think Republicans may gain a few seats on net. In some states it is difficult to gerrymander because redistricting goes through a commission or because other provisions limit what legislators can do.
That said, I think the media focus too much on the net gain or loss in seats. The more important consequence is reduced electoral competition. Some estimates put the number of toss-up congressional seats at only around 30 to 35 out of 435 nationwide. That means most seats are effectively decided before the general election, and only a relatively small number of competitive seats may determine which party wins the majority in Congress.
In 2024, Representative Marjorie Taylor Greene, a Republican, won reelection in Georgia’s 14th district by about 29 points over Democrat Shawn Harris. In the 2026 special election after Greene’s retirement, Harris still lost, but by only 12 points—a swing of 17 points. If a party moves voters out of one district considered “safe” to make another more favorable, could it end up putting both at risk in a wave election? In other words, could gerrymandering backfire?
When parties gerrymander, they generally leave a cushion. Incumbents do not want to risk losing reelection, so parties usually do not design a supposedly safe seat to be only two or three percentage points ahead. They try to leave a comfortable margin. In a very large wave, though, that cushion may not be enough, and seats thought to be safe can be lost. So, yes, if you have a 15- or 20-percentage-point swing, gerrymandering can backfire. But by historical national standards, gerrymandering still reduces electoral competition. If the districts had not been gerrymandered, there would generally be many more competitive seats that could flip in response to a large national swing.
When parties gerrymander, they generally leave a cushion. . . . In a very large wave, though, that cushion may not be enough, and seats thought to be safe can be lost. So, yes, if you have a 15- or 20-percentage-point swing, gerrymandering can backfire.
Finally, it seems like we’ve entered a time of “tit for tat” gerrymandering, where blue and red states consolidate control at home and counter redistricting by each other to keep power on the national level. Moreover, on the rare occasion that control changes, a party newly in power has an incentive to roll back previous gerrymanders and institute its own. In either case, politicians are choosing their voters, rather than the other way around. Is there a way out of this back-and-forth battle?
The fundamental problem in the United States is that politicians themselves often draw the districts. Some states, including Michigan and California, use commissions or other nonpolitical actors to draw the lines. We have studied these systems and shown that this type of commission reduces partisan bias.
I think the best solution, then, is to take the power of redistricting away from politicians. In many other countries that is standard. Countries in Europe, Japan, and elsewhere in Asia use nonpolitical bodies to draw districts according to criteria such as equal population, respect for administrative boundaries, and compactness. We are working with the Irish nonpartisan electoral commission to help with redistricting there. Bureaucrats, rather than politicians, draw the lines. That is the type of system I would like to see more often in the US.
