PUBLISHED Mar. 18, 2019 at 10:18 AM

One Way To Spot A Partisan Gerrymander

Will it convince the Supreme Court?

On March 26, challengers to North Carolina’s congressional map will argue before the Supreme Court that it is a partisan gerrymander — that is, the district boundaries were drawn to benefit one political party, the GOP, in a way that violates the Constitution. The challengers are using a variety of quantitative tools to make their argument, including a metric called “partisan bias” that tries to evaluate how skewed a map is by looking at the number of seats a party would have won in a hypothetical election in which the vote was evenly split between Republicans and Democrats. Here’s how that metric works and what it says about congressional maps going back a few decades.

1Our analysis is based on the two-party vote share, which considers only votes that were cast for Republican or Democratic U.S. House candidates. All races won by independent candidates were excluded from this analysis. For races that did not have both a Democrat and a Republican running, a statistical model was used to determine the likely two-party vote share. The average Democratic two-party vote share is the average of the results for every individual district, not the Democratic share of the national popular vote.
2North Carolina’s 9th Congressional District is not included in this analysis because the State Board of Elections recently voted to hold a new election after evidence of fraud was uncovered.
3For simplicity, we’re assuming that an equal number of percentage points are lost in each district even though that’s only a rough approximation of how voting patterns change. An expert for the challengers in the North Carolina case also makes this assumption when calculating partisan bias.
Let’s start by comparing the average share of the vote that Democratic candidates received1 in the 2018 U.S. House elections with the share of seats the Democrats won.
Democrats won about 55 percent of the average two-party vote share and about 54 percent of U.S. House seats.2 That seems pretty good, but to determine a map’s level of partisan bias, we need to figure out how many seats Democrats would have won if they’d received just 50 percent of the average district’s votes.
Let’s start by imagining that Democrats win a smaller and smaller percentage of the votes in each race3 and calculating how that affects their seat share.
For example, if the average 2018 Democratic vote share dropped about 3 percentage points — to 52 percent — the Democrats would have won fewer than half the seats despite still receiving more than half the vote.
And in a scenario in which Democrats received 50 percent of the vote, on average, they would do even worse, winning 45 percent of seats.
That gap is the “partisan bias” — the difference between an even split in seats and the share of seats a party would get if the vote were evenly divided between Republicans and Democrats. In 2018, the partisan bias was about 4.8 percentage points, or roughly 21 fewer Democratic seats (which also means 21 more seats for Republicans).
If we extend this line, we get 2018’s “seats-votes curve,” which shows possible seat outcomes given a wide range of vote scenarios. An election with little partisan bias would have a seats-votes curve that passes close to the 50-50 point — which, as you can see, 2018’s does not.
But it’s hard to know, without more information, whether the partisan bias measured in 2018 is unusual. So let’s compare 2018 to all House elections since 1992, a period that covers the last three rounds of congressional redistricting.
2018’s partisan bias is the fourth-largest (favoring either party) among the 14 elections that have taken place since 1992. That’s not great. And it’s part of a trend. The four U.S. House elections that have taken place since 2012 have the highest levels of partisan bias among all elections since 1992.

We can also draw seats-votes curves for individual states, which typically redo their congressional and legislative maps every 10 years, after the decennial census. In the North Carolina case, however, the congressional map that is being challenged was enacted in 2016 after a 2011 map was declared unconstitutional (in that case, because the map makers relied too heavily on race). To support their argument that the 2016 map is a partisan gerrymander that disadvantages Democrats, the challengers use an expert’s finding that the map’s partisan bias in the 2016 election was the largest in the state since at least 1972. In the chart below, which looks at the state’s congressional elections since 1992, you can see that the partisan bias against Democrats increased dramatically in this decade.

Republicans in North Carolina seem to have gained an advantage this decade
Seats-votes curves for North Carolina’s congressional elections since 1992 by decade
1992-2000
2002-10
2012-18

The 2016 map’s defenders argue, however, that metrics such as partisan bias may incorrectly classify a district map as favoring Republicans because Democrats tend to cluster in cities. In response, the challengers say that a computational method that takes clustering into account also reveals the 2016 map’s unusually high degree of partisan skew.

But Republicans are not the only ones who critics say have unfairly drawn maps to their advantage. The Supreme Court is also considering a case out of Maryland, in which Democrats are accused of partisan gerrymandering. The challengers aren’t citing the partisan bias metric in their case, but they are arguing that the map makers diluted Republican votes with the intention of creating an additional Democratic district.

Sometimes challenges to maps result in court orders that the maps be redrawn. Last year, Pennsylvania’s congressional district boundaries were redrawn after the state’s Supreme Court ruled that the map used from 2012 to 2016 was an unconstitutional partisan gerrymander. And according to the partisan bias metric, it seems as though the new map is indeed fairer. As you can see in the charts below, the 2018 map had a partisan bias of 0, while the map that was used for the three other elections this decade had a partisan bias of roughly 22 points against Democrats.

Pennsylvania’s redrawn 2018 map has a partisan bias of zero
Pennsylvania’s seats-votes curves for elections since 1992 by decade
1992-2000
2002-10
2012-18

Remember that partisan bias is just one of several ways to measure whether a map is gerrymandered and that it’s only a piece of the argument being made by the challengers to North Carolina’s map. So far, the Supreme Court has shied away from offering a clear decision on the constitutionality of partisan gerrymanders. Previous attempts to use metrics to persuade the court have not seen much success. Whether it will prove persuasive this time around remains to be seen.

Comments