What Is a Gerrymander?

Over the past 15 years or so, the practice of gerrymandering has moved from the sidelines of contemporary political discussions to take center stage. Most of the debate focuses on political gerrymandering. But with the Supreme Court’s recent decision in Louisiana v. Phillip Callais—which ruled that Louisiana’s congressional map was an unconstitutional racial gerrymandering has moved increasingly into the public mind.

Lost amid all the sound and fury is an important question: What exactly is a gerrymander? We may call it something akin to the “manipulation of electoral district boundaries to give one political party or interest an unfair advantage in elections,” but this only bounces the question back a level to “okay, but what does ‘unfair’ mean?”

Getting to the heart of this issue means first understanding the different conceptions of a gerrymander and the core questions of representation and fairness dominating the debate today. While attempts to identify gerrymandering are not, as Chief Justice John Roberts infamously put it, “sociological gobbledygook,” neither are they something that, as Justice Elena Kagan less-infamously described, “pretty scientific by this point.” Two broad concepts of gerrymandering are anchoring today’s debate. Examining them illustrates why the Supreme Court was on to something when it concluded that gerrymandering, while distasteful, was beyond the judiciary’s reach.

The Fundamental Questions

A few questions help to illustrate the complications. First, was California’s congressional map from 2020 to 2026 a gerrymander? After all, the map gave Democrats between 75 and 83 percent of the seats in a state where they usually win around 60 percent of the vote. At the same time, it was drawn by a nonpartisan redistricting commission, which was expressly prohibited from relying on political information. In other words, insofar as it was a gerrymander, it was so either by choice or because the state’s political geography made it so.

Second is Massachusetts, where Republicans routinely win around 40 percent of the vote but haven’t won a Congressional race since 1994. Is it a gerrymander? The issue is that the party’s vote in this state is extremely inefficiently distributed. This makes it difficult, though not impossible, to draw a map offering a Republican seat in Massachusetts, and it is extremely difficult to draw a second Republican seat there. In other words, to help secure Republican representation, you likely have to take politics into account. It can be done, but there are real trade-offs involved.

In a third case, Texas, Republicans won around 66 percent of the seats under the prior map, and they often win statewide races by low double digits these days. Is it a gerrymander? While it is clear that Republicans relied on political data, these data produced less of an advantage for Texas Republicans in terms of seats than the California map did for Democrats.

Finally, consider Indiana. The Indiana map is, at face value, a beautiful map, with compact, regularly shaped districts that represent the various regions and urban areas of the state. Yet Democrats, who frequently earn around 40 percent of the vote in the state, are reduced to just 23 percent of the representation. In this way, Indiana’s “normal” map yields disproportionate political outcomes. Is it a gerrymander?

Gerrymandering as an Intentional Act

If figuring out which, if any, of the above counts as a gerrymander seems straightforward, good for you! Trying to find a principled through line here is difficult for many, because people often have conflicting views of what “unfairness” means. This is true in the political science community as well, where opinion on what constitutes a gerrymander splinters into a couple of broad camps.

If a redistricting body draws a map that just happens to favor one party or another, that’s just the vagaries of redistricting or the impact of the state’s political geography.

The first broad camp would somewhat refine the definition of a gerrymander presented above, calling it something akin to the intentional manipulation of districting lines to favor one party or another. Under this conceptualization of a gerrymander, to prove a map was a partisan gerrymander you would have to prove that the legislature drew lines the way that it did because it wanted to favor one party or the other. If a redistricting body draws a map that just happens to favor one party or another, that’s just the vagaries of redistricting or the impact of the state’s political geography. In the examples above, California’s old commission-drawn map probably would not be a gerrymander, while Texas’s map probably would be one because of the evidence of political manipulation in the latter and the lack thereof in the former. Whether the maps in Indiana and Massachusetts count as gerrymanders would depend on the vagaries of litigation and just how strongly one felt about the intent standard.

The type of proof one would employ here is different from what we would use with the other vision of gerrymandering. Obviously, a direct statement by a mapmaker could help prove intent. But sometimes map drawers are more careful. In these instances, we would need indirect proof of intent. For this, we typically employ the “ensemble” approach to redistricting. Here, a computer program is asked to simulate nonpartisan redistricting by drawing many maps without partisan information. After the maps are drawn, they are evaluated for partisanship. If the partisanship of the ensemble reflects the partisanship of the enacted maps, we would not conclude that the map, as drawn, was a gerrymander. On the other hand, if the partisanship of the ensemble looks very different from the partisanship of the enacted map, we may conclude that the enacted map is a gerrymander.

There are limitations with this method of gauging whether a map is a partisan gerrymander. First, the ensemble approach assumes that mapmakers think about redistricting in the same way as a computer. But this isn’t necessarily the case. A mapmaker may have a much broader concept of compactness they would consider, or they may be wedded to keeping certain communities together. Deeply personal considerations—such as keeping a particular county in an influential incumbent’s district or keeping an incumbent’s business in his district—can influence the partisanship of a district in a way that is difficult for a computer to capture. In other words, for the ensemble approach to work, the computer has to take into account all of a mapmaker’s legitimate redistricting considerations. This is difficult to do and sometimes easy to end run by using a legitimate consideration as a pretext for an illegitimate one.

But the bigger drawback is that, once again, the ensemble simply pushes back a step the question of what constitutes a gerrymander. The definition of a map that looks “very different” from the ensemble is contested. We may employ the traditional political science metric of 95 percent confidence: If a map is more extreme than 95 percent of the drawn maps, then it is a gerrymander. But the 95 percent standard doesn’t necessarily map neatly onto redistricting for complex reasons.

Moreover, this leads to potentially unsatisfying conclusions. For example, in the absence of direct statements from legislators of partisan intention, the Indiana map would not count as a partisan gerrymander, because a computer will generate maps that are similar to what the Indiana legislature drew. In our Massachusetts example, it is possible to draw a map with a Trump-won district in Massachusetts, and a computer will draw some districts like this. But the map doesn’t appear to be outside the 95 percent interval we would encounter in Massachusetts. Even so, could a legislature choose from any range of partisanship within the 95 percent confidence interval?

We may not expect to see a district that Trump wins outright in Massachusetts, but a computer will draw a lot of maps with districts that he lost by one or two points, where a Republican would have a good chance in a good GOP year.

Wouldn’t a legislature be expected to draw a map that is as favorable as reasonably possible for a disadvantaged party, within the universe of computer-generated maps? We may not expect to see a district that Trump wins outright in Massachusetts, but a computer will draw a lot of maps with districts that he lost by one or two points, where a Republican would have a good chance in a good GOP year. If you also believe in fairness, that may be the upshot here. The upshot is including intent concepts in a definition of a gerrymander smuggles in real normative considerations. It also doesn’t really answer the question of what a gerrymander is.

Gerrymandering as an Accidental Act

This leads to the second conception of a gerrymander: the partisan fairness measures. This idea of a gerrymander is the one that initially made its way to the Supreme Court, with the “efficiency gap” at the center of litigation. The efficiency gap, however, is just one of a broader family of “partisan fairness” measures. It argues that gerrymandering is measured by comparing the “wasted” votes of one party or another. That is, if a party has a bunch of votes in districts where it only receives, say, 45 percent or 46 percent of the vote, combined with a bunch of votes in districts where it wins by large margins, it is likely the victim of packing (concentrating opposition voters into a few districts) and cracking (splitting opposition voters across many districts). Declination, partisan bias, and mean–median are all ultimately variations on this theme.

What this approach is really saying, though, is that intent isn’t as important as what you end up with. In other words, even if you accidentally disfavor a party in a state while drawing without partisan information available, you have still drawn a gerrymander. This approach is less concerned with the political geography of a state and focuses more on the resulting partisan balance. Under this approach, all the maps described above would likely be considered gerrymanders, since they produce unequal outcomes that are avoidable.

There are problems here, as well. First, what exactly counts as a “win” or “loss”? We can look at actual results in congressional elections in the district, but that requires us to use a potentially gerrymandered map at least once. Moreover, it’s subject to the vagaries of the political process. Republicans might lose a Republican district simply because it’s a bad Republican year, or because there’s a bad Republican candidate, or some combination of the two, which may skew the resulting efficiency gap. In other words, this approach often tells us more about a given election than the actual map.

Finally, we once again confront the question of “how much gerrymandering is too much?” No one really knows.

We can also look at previous election results, but this introduces a host of potentially contradictory outcomes. In particular, the question of whether a map qualifies as a gerrymander can be sensitive to the elections selected. A map that uses former Massachusetts Governor Charlie Baker as its baseline is going to look different from one that uses a typical statewide Republican. We might average multiple elections, but then we are comparing the map to an electoral baseline that never actually happened. These measures also tend to be sensitive to random tweaks in election outcomes. A 49 percent Republican district is treated as substantially different from a 51 percent Republican district; from an electoral standpoint, they’re quite different. Additionally, some states simply have naturally large partisan fairness measures, like Kentucky (where you must work to produce more than one Democratic district) and Massachusetts (where you must work to produce more than one Republican district). Finally, we once again confront the question of “how much gerrymandering is too much?” No one really knows.

The Best Way Forward

From my point of view, the intent standard is clearly the best one. The choice to elect representatives via single-member districts is a meaningful one, which tells us that place matters. If one wants partisan fairness, there exists a system of elections used all over the world that produces it: proportional representation. But hopefully this helps illustrate a broader truth: Defining a gerrymander is not easy, nor is it scientific. It’s a question that involves deep theoretical questions and policy questions, and voters should be wary of anyone who oversimplifies it. The Supreme Court could not have resolved this debate without making the type of arbitrary choice that it is ill-equipped to make, but that legislatures handle well.

Sean Trende is a nonresident fellow at the American Enterprise Institute, where he works on elections, American political trends, voting patterns, and demographics. He is also the senior elections analyst for RealClearPolitics.