This November, members of Congress will be running in new districts based on the 2020 Census data. So how might the new district maps influence the midterm elections — and perhaps more important, which party wins control of the House?
Of course, many issues will affect that result, from the fact that the president’s party usually loses seats in the midterms, to the Supreme Court Dobbs decision, inflation, and the Trump investigations, and any surprises between now and November. We can’t offer any predictions on those factors. But our research finds that this round of gerrymandering hurt Democrats less than the maps in place during the 2020 elections for the House. Here’s how we know.
Don’t miss any of TMC’s smart analysis! Sign up for our newsletter.
Using both geography and election results to detect gerrymandering
How can we calculate and compare the impact of gerrymandering across the whole country? By using a new metric that can give an idea of how many districts each party is winning due to gerrymandering in each state. We call it the “GEO metric,” for Geography and Election Outcomes.
When mapmakers draw a partisan gerrymander, they must use two key pieces of information. First is the partisan data of whether a set of voters leans toward Democrats or Republicans, which the mapmakers infer from such sources as the percentage of votes that went to a Democrat or Republican in the last presidential election. Second is the geographic data of where those voters are located on the map.
Surprisingly, researchers measuring partisan gerrymandering in the past have not used both those pieces of information.
For instance, consider two widely used older metrics: the Polsby-Popper, introduced in the early ’90s, and the Reock ratio, introduced in the ’60s. Both use only the irregularity of a district’s shape — in other words, information about the map — to detect gerrymandering. But that can’t tell us whether an irregular shape is drawn because of natural boundaries like coastlines and mountains or because mapmakers are trying for partisan advantage.
Since 2015, researchers have used more modern metrics, like the Efficiency Gap and Mean-Median, which use only the partisan makeup of each district. But these metrics cannot tell whether voters of different parties are being separated because that’s what the mapmaker is trying to do or because that’s how the state’s geography falls.
However, it’s necessary to use both the partisan data and map data. For example, looking at the newly released district maps, the Efficiency Gap concludes that Connecticut has a much more extreme gerrymander than Illinois, which is gerrymandered to favor the Democrats, or than Florida, gerrymandered to favor the Republicans. Our new metric uses both election and geographic data. With that, we can see that Democratic support is distributed across Connecticut consistently enough throughout the state that Republicans are unlikely to win a congressional seat no matter how maps are drawn — something that the Efficiency Gap could not detect. Thus, while our GEO metric appropriately flags Illinois and Florida as gerrymandered, it does not inappropriately flag Connecticut.
If you’re curious how this can happen, our paper introducing the GEO metric gives a clear example with a pair of small, easy-to-understand fictional states.
State judges tend to favor their own party’s district maps — especially Republican-appointed judges.
GEO metric shows that this year’s maps favor Democrats more than maps used in 2020
For most states, the GEO metric finds that, compared with the last round of districts, this round’s districts offer roughly equal room for improvement for both Republicans and Democrats’ chances at taking seats. For example, in Washington state, data from Dave’s Redistricting App predicts Democrats will win seven seats and Republicans will win three. The GEO metric gives a score of 2 to the Democratic Party and 3 to the Republican Party in Washington, meaning that if the map were adjusted slightly, Democrats could gain another two seats, and other small adjustments could give Republicans another three. Since two and three are very close, the map is largely fair.
The GEO metric gives a score of 7 to Republicans and 1 to Democrats in Illinois, meaning that if there were changes to the map, Republicans could gain a lot but Democrats couldn’t gain much — which suggests it’s a Democratic gerrymander — which is predictable in a state where Democrats control the legislature. On the other hand, the GEO metric gives a score of 4 to Republicans and 10 to Democrats in Florida, indicating that while a changed map could help Democrats a lot, changes would benefit the Republicans much less — which suggests it’s a Republican gerrymander, which, similarly, isn’t surprising with a Republican-dominated legislature.
When we add up all of the GEO scores across all states for the maps that will be used in 2022, we get an accumulated score of 82 for Democrats and 92 for Republicans — meaning that Republicans are at a slight disadvantage compared to where they could be. For the maps used in 2020, the Democrats had an accumulated score of 91 and the Republicans had a score of 82 — meaning that the Democrats were at a slight disadvantage at that point. In other words, the 2022 maps give Democrats a bit more of an advantage than the 2020 maps did.
GEO metric gives an interpretable count
That’s one benefit of the GEO metric: It offers a count of the number of additional districts a party could have won with small changes to the map. This number is very understandable and interpretable. What’s more, the GEO metric tells us exactly which districts could have become competitive — giving anyone drafting lawsuits to challenge a district the relevant ammunition. And the GEO metric reveals exactly which districts were gerrymandered, which previous metrics couldn’t do.
We can clearly see that a number of state legislatures are gerrymandering districts for partisan gain. We hope that the GEO metric will give state courts a better tool to detect gerrymanders, so that by the time we’re discussing what might affect the 2032 elections, gerrymandering won’t be in the list.
Professors, check out TMC’s newly indexed and improved list of classroom topic guides.
Marion Campisi is an associate professor of mathematics at San José State University.
Tommy Ratliff is a professor of mathematics at Wheaton College in Norton, Mass.
Stephanie Somersille is a math consultant specializing in the areas of gerrymandering and math education.
Ellen Veomett is a professor of mathematics at Saint Mary’s College of California.