On Thursday morning, a 5–4 Supreme Court decision closed the door on having federal courts police partisan gerrymandering. The Court’s decision left standing a Democratic gerrymander in Maryland’s House districts and a Republican gerrymander in North Carolina’s, both of which were openly partisan. The Court’s decisions are, predictably, reviving the ongoing debate over the centuries-old practice of gerrymandering: state legislatures drawing congressional and state legislative districts to favor their own political party. Partisan gerrymandering is political self-interest in its purest form, and it’s a game played by both parties whenever they can. But for all its mischief, how much does it really affect control of Congress?
There are three ways to analyze the question, depending how far you drill into the numbers. The first is the overall national balance of power: How many seats does each party get in the House, compared with its share of the national popular House vote? I looked at this question after the 2016 election:
Elections analyst Sean Trende has noted that, from 1942 to 1992, “the Democrats had a huge advantage in seats won vs. their popular-vote share, averaging 5 percent,” and that “the discrepancy was less than 3 percent on just five occasions.” Trende reached this conclusion by comparing the Democrats’ share of House seats to the Democratic share of the two-party popular vote. Using Trende’s measurement in today’s House, a five-percentage-point advantage is worth an extra 22 House seats. Today’s Republican majorities regularly enjoy a smaller edge than the Democrats averaged over that half century of dominance. The Republican advantage was 4.9 points (worth an extra 21 seats) in 2016, when Republicans won 55.4 percent of the seats with 50.6 percent of the two-party vote. It was smaller in the three previous elections: 3.8 points (17 seats) in 2014, 4.4 points (19 seats) in 2012, and 2.1 points (9 seats) when the Republican majority was first elected in 2010, before the latest round of redistricting.
As it turns out, in 2018 the Republican “advantage” was negative three seats, as the Democrats won 54 percent of the House seats with 53.4 percent of the popular vote. (For these purposes, I count both the popular vote and the reported outcome in North Carolina’s Ninth District, where a razor-thin Republican victory was later vacated over ballot fraud.) That’s right: The Democrats were so undisadvantaged by gerrymanders that they got more seats than votes. Their three-seat advantage by this measurement was, however, the smallest advantage that a Democratic House majority has enjoyed since 1938:
Now, combining vote totals across districts and states — like any measurement of actual votes cast under one election system, used to measure what would have happened under a different system — is an imperfect measurement. Uncontested races skew the cumulative vote totals, and turnout can vary widely from district to district. Which races are contested or competitive, in turn, can also be affected by the way district boundaries are drawn. All we can do with such measurements is observe the most obvious effects. But that is precisely why we should be skeptical of efforts to use these kinds of measurements to indict the current system.
Of course, House races — like elections for the Senate or the presidency — are not decided by national popular vote, which brings us to the second possible measurement: how far state-by-state caucuses deviate from the statewide popular vote. Even at the state level, congressional races are winner-take-all, rather than giving each party a proportional share. (As Chief Justice Roberts noted in his opinion for the Court, some states in the 19th century took it a step further and had winner-take-all races to elect their entire House slate, instead of geographic districts).
While some countries use proportional-representation systems, the American way has always been to pick one winner for each election. Winner-take-all systems have important merits, as they encourage the building of majorities or broad pluralities rather than just the pursuit of small, dedicated factions. More to the point, even in a world where courts or nonpartisan agencies abolished partisan gerrymanders, winner-take-all elections would still be the American rule. Much of the deviation from statewide popular-vote totals in individual states thus results from factors other than partisan district borders:
- Winner-take-all elections mean that a candidate who wins 50.1 percent of the two-party vote gets 100 percent of the seat.
- Some states have only one seat.
- A party that gets below 40 percent of the statewide vote in a larger state can easily lose 100 percent of the races no matter how the districts are drawn.
- Democratic voters tend to be more geographically concentrated, in urban areas, than Republican voters.
- The Voting Rights Act is sometimes read to require certain districts to be “majority-minority” (i.e. majority non-white) which makes it hard even for nonpartisan districts to be drawn wholly impartially. Some of the same factors would come into play if you were analyzing state legislative districts, rather than congressional districts.
Also, as the Maryland and North Carolina cases illustrate, there are partisan gerrymanders on both sides, and at least some of these will cancel each other out nationally. Even if you measure on a state-by-state basis, that means that gerrymandering affects control of the chamber only if just one side does it. That was truer in the 1960s and 1970s, when Republicans controlled very few state legislatures, but it is far from the reality of today.
So, if we look on a state-by state level, do Republicans have an advantage? Yes: but it is only a two-seat advantage, not much different from the three-seat Democratic advantage we see if we use the national popular vote. In seven states, Republicans won between two and four additional seats beyond their share of the statewide popular vote, and in 15 others, they won one additional seat:
That again includes the voided outcome in NC-9. By no means are all of these outcomes the result of partisan gerrymandering: Some of these states have natural district boundaries, and you would have a hard time drawing a Democratic district in Idaho, for example, where Republicans got 64 percent of the statewide vote in a state with two districts. In Michigan, voters in 2018 approved a nonpartisan redistricting commission to take over after the 2020 census.
In both Michigan and Pennsylvania (where a new map was used in 2018 by court order), a pure statewide vote would have produced a majority of Democrats, a fact that (if repeated in 2020) could matter in the unlikely event that the presidential election is decided by the House. In Wisconsin, such a vote would have produced an even 4–4 split.
States with a Republican advantage are only half the story, however. There were also seven states where Democrats gained more than two seats above their share of the statewide popular vote, and ten states where they gained one additional seat:
In California alone, Democrats in 2018 won eleven more House seats than they would have if seats were awarded proportionally by statewide popular vote. Some of these states, such as Maryland and Illinois, have notorious partisan gerrymanders. But in others, such as California, Massachusetts, and New York, the larger issue is simply that the Republican vote share dropped so low that they were bound to lose a lot of winner-take-all races. Notice a pattern: Every one of these states has a Democratic-majority House delegation, while every state where Republicans overperformed has a Republican-majority House delegation, except for Wisconsin, Michigan, and Pennsylvania. Much of the split between popular votes and House delegations is simply a natural outcome of winner-take-all districts.
In eleven states, including evenly divided Florida, the House delegations are the same as if they had been selected by statewide popular two-party vote:
Moreover, the Republicans lost the House after Democrats had claimed for years that it was rigged in the Republicans’ favor. While a new map in Pennsylvania helped, the real lesson is that even the savviest mapmakers cannot always see the future, and most states draw new districts just once a decade. Political coalitions are constantly shifting. Nobody in 2011 was thinking about Donald Trump’s coalition, any more than mapmakers in 2001 were thinking about Barack Obama’s coalition. Unlike the 1954–94 period, when Democrats had unbroken control of the House with majorities that often far exceeded their share of the popular vote, we have now seen the House change party control three times in the last seven election cycles, suggesting that fear of an entrenched majority party is overblown.
That brings us to the third and most complex way of measuring the impact of partisan gerrymanders, which is to move away from numerical generalizations and drill into alternative maps of each state. That’s a much more involved process, fraught with methodological challenges and disputes (such as how to model the requirements of the Voting Rights Act). As I noted in 2016, computer modeling has suggested that most of the Republican advantage in congressional districts can be explained by residential patterns (i.e., Democratic concentration in urban areas) rather than partisan mapmaking. To the extent that progressive critics wanted the courts to get involved not only in preventing partisan gerrymandering but in drawing lines to help Democrats out of a natural disadvantage, it is easy to see why the justices were hesitant to get into that business.
Politicians have been gaming the system to choose their own voters as long as there have been elected legislatures, and it is fair to debate whether there might be a better way for states to draw districts. I confess some skepticism as to whether courts or “nonpartisan” commissions such as the one used in California can really be trusted to draw maps impartially, rather than cloaking their preferences in a veneer of disinterested expertise. But the reality is that the observable impact of partisan gerrymanders on partisan control of the House of Representatives is fairly modest, prone to breaking down over time, and no barrier to frequent changes in control of the chamber. It is not a problem requiring radical solutions, and it is one we can work out without involving the courts.