Sabato's Crystal Ball

Partisan Gerrymandering and the Outlook for the 2018 U.s. House Elections

Alan I. Abramowitz, Senior Columnist, Sabato's Crystal Ball December 14th, 2017


There is a growing sense among political observers that the United States may be heading toward a wave election in 2018. Results of recent special elections, including Doug Jones’ (D) victory in the Alabama Senate race on Tuesday, along with Democratic victories in the New Jersey and Virginia gubernatorial elections and surprisingly large Democratic gains in the Virginia House of Delegates all point toward the likelihood of substantial Democratic gains in next year’s midterm elections, including a real possibility that Democrats could regain control of the U.S. House of Representatives. In addition, results of recent generic ballot polling generally show large Democratic leads.

Despite these signs of an impending Democratic wave, however, many political experts believe that the way House district lines in many key states were drawn by Republicans prior to the 2012 election will make it difficult for Democrats to gain enough seats to take back control of the House. According to this argument, Republican gerrymandering was so effective that Democrats would need to win the national popular vote by a very large margin, perhaps eight points or more, in order to gain at least the 24 seats needed to take back the House.

In order to test this argument and evaluate how large a margin Democrats would need in the popular vote to win a majority of House seats, I conducted a regression analysis of the impact of the Democratic share of the major party vote on the number of Democratic seats in the House for all 36 House elections between 1946 and 2016. In addition to the Democratic share of the national popular vote, I included one additional predictor of the number of Democratic seats — a dummy variable indicating whether the election occurred before or after the 2011 round of redistricting in which Republicans enjoyed complete control of the process in many key states including Florida, Michigan, Ohio, Pennsylvania, Virginia, and Wisconsin. The results of the regression analysis are presented in Table 1.

Table 1: The impact of Democratic vote share on Democratic seats in U.S. House of Representatives, 1946-2016

Source: Data compiled by author

The results in Table 1 show that this simple two-variable model explains almost 90% of the variance in Democratic seats between 1946 and 2016. The Democratic share of the national vote is a very strong predictor of the number of Democratic seats in the House over these 36 elections. For every additional 1% share of the national vote won by Democratic candidates, Democrats gain an average of about 8.3 seats in the House. However, after controlling for the effect of the Democratic share of the major party vote, the results indicate that post-2010 Republican gerrymandering had a significant negative impact on the number of seats won by Democratic candidates. In these three elections, Democrats won an average of 21 fewer seats in the House than they would have won in earlier elections given the same share of the national vote.

Table 2: Predicted Democratic House seats by Democratic share of major party vote

Source: Data compiled by author

The impact of Republican gerrymandering on the number of Democratic seats in the House can be seen very clearly in Table 2. This table compares the number of expected Democratic seats before and after 2011 based on Democrats winning between 46% and 54% of the vote. Based on these results, it is substantially more difficult today for Democrats to win a majority of House seats than it was before 2011. Specifically, Democrats now need to win almost 52% of the national vote in order to win 218 seats compared with just over 49% of the national vote prior to 2011. These results help to explain why Democrats fell well short of regaining control of the House in 2012 despite winning 50.6% of the major-party vote.

What to expect in 2018

It clearly is more difficult now for Democrats to win a majority of seats in the House than it was before 2011. Contrary to what some political observers have claimed, however, they do not need to win the popular vote by a landslide or near-landslide margin in order to regain control of the lower chamber. Based on the results in Table 2, a popular vote margin of between three and four points would be large enough for this purpose. So what are the chances of Democrats winning the popular vote by this large a margin in 2018?

In order to answer this question, we can make use of a simple but accurate model for forecasting the Democratic share of the popular vote for the House of Representatives. This model uses three predictors of the Democratic vote share: the Democratic margin on the generic ballot, whether the election is a midterm with a Republican or a Democratic president, and the number of seats held by Democrats prior to the election. The estimates for this model are displayed in Table 3.

Table 3: Forecasting Democratic share of House vote

Source: Data compiled by author

This simple forecasting model explains about 80% of the variance in the Democratic share of the major-party vote for the U.S. House in 36 elections between 1946 and 2016. The strongest predictor of the Democratic vote share is the generic ballot margin. According to these results, for every one-point increase in margin on the generic ballot, Democrats can expect to receive an additional 0.2% of the vote. There is also a strong effect for the party of the president in midterm elections. Even after controlling for the generic ballot, Democrats can expect to win an additional 1.9% of the vote when there is a Republican in the White House and to lose an additional 1.9% of the vote when there is a Democrat in the White House. That is equivalent to gaining or losing about 15 additional seats in the House. Finally, the number of seats Democrats hold going into an election has a modest impact on the number of seats they can expect to win. However, this incumbency effect appears to be much weaker than that of the other two predictors.

Table 4: Predicted Democratic share of House vote and Democratic House seats in 2018

Source: Data compiled by author

Based on the results in Table 3, we can estimate the share of the national vote and, therefore, the number of House seats Democrats would be expected to win depending on the Democratic margin on the generic ballot in early September of 2018. Those results are presented in Table 4. According to the data in this table, Democrats will need a margin of at least four points on the generic ballot in order to win a majority of seats in the House in the 2018 midterm election. In recent weeks, Democrats have been averaging a lead of between eight and 10 points according to RealClearPolitics. According to the results in Table 4, that large a lead on the generic ballot would predict a popular vote margin of around five points and a gain of between 30 and 33 seats in the House — enough to give Democrats a modest but clear majority.

Blue Stars Fell on Alabama

Jones win another confirmation of Democratic momentum, even taking Moore’s horrible candidacy into account

Kyle Kondik and Geoffrey Skelley, Sabato's Crystal Ball December 13th, 2017


Prior to the Alabama special Senate election on Tuesday night, there was an ongoing discussion in the media about whether the Republicans would lose either way in Alabama. But as bad as it would have been for Republicans to have had Roy Moore (R) in their Senate caucus, losing this seat is, in our view, significantly worse. The victory by Sen.-elect Doug Jones (D) will cut the GOP’s Senate majority to a slim 51-49 margin, and it opens the door to an unlikely Democratic Senate takeover next year.

Perhaps more alarmingly for Republicans, the race reinforced several trends we’ve seen in other places this year. Here was another special election where a Democratic candidate ran very heavily ahead of Hillary Clinton’s 2016 showing. Republican turnout was OK, just like in the Virginia gubernatorial race, but Democratic turnout was a lot better, both in urban and suburban parts of Alabama and also in the Black Belt — a rural, heavily African-American part of the state that gets its name from the color of the soil (turnout was exceptionally high throughout this region).

There are obviously a lot of compounding, negative factors for Republicans that led to this result, many having to do with the GOP nominee. But the positive environmental signs for Democrats are a big part of the story, too. Individual special elections are not always a harbinger of the future, but it’s impossible not to compare Jones’ victory to Scott Brown’s (R) January 2010 Senate election upset in Massachusetts, which suggested a terrible Democratic environment that endured through the party’s disastrous 2010 midterm. At the time, Democrats pinned the loss on their own bad candidate, Martha Coakley, but huge electoral upsets don’t have just a single cause, and the Republicans ignore 2017’s warning signs at their own peril.

Meanwhile, Jones should remember Brown, too — his time in the Senate was short-lived, ending in 2012. Jones will have three years in the Senate, but it will be very hard for him to add to that tenure when he presumably seeks a full term in 2020.

But that’s a long time off. Democrats now have miraculously added a Senate seat that, truth be told, they have no business having, and it’s one they do not have to defend next year.

That’s good for the Democrats because they already are defending a lot of territory next year — 26 of 34 seats, including a special election in Minnesota — but none of their incumbents are retiring. Even the most vulnerable Senate Democrats up for election next year — Sens. Joe Donnelly (D-IN) and Claire McCaskill (D-MO) — are probably no worse than 50-50 to win reelection at this point, given the challenges the presidential party historically faces in beating Senate incumbents from the other party (Republicans would counter by arguing that McCaskill, at the very least, looks like a goner — we’ll see). The list of vulnerable Republican seats is much smaller, but Sen. Dean Heller’s (R-NV) reelection bid is a pure Toss-up, as is the open GOP seat in Arizona. The math problem for Democrats prior to Tuesday night was that even if they pulled off an unlikely sweep of all these seats, maintaining the 26 they already hold plus capturing Arizona and Nevada, that would only get them to 50-50 in the Senate, with Vice President Mike Pence breaking ties for the Republicans.

But now, that sweep would get them to a bare 51-49 majority. We are not predicting that to happen, and Republicans are still favored, perhaps significantly, to hold the Senate. But this flub could have real consequences for the GOP.

Democrats will say that they have additional offensive targets, like the open seat in Tennessee that former Gov. Phil Bredesen (D) is now contesting, and perhaps Rep. Beto O’Rourke’s (D) longshot bid against Sen. Ted Cruz (R-TX). We’re holding those races at Likely (though not Safe) Republican for now. We don’t want to overapply the lessons of Alabama to other races: Roy Moore may quite possibly have been the worst Senate candidate in quite a long time, and there have been plenty of duds in recent years (mostly, it must be said, on the Republican side, and these poor choices by their primary electorates have cost them several seats over the past few cycles). However, given what we’ve seen in other special elections this year, and in our home base of Virginia, Moore’s awful candidacy could not have been entirely responsible for turning a 28-point Trump win in Alabama last year into a narrow Democratic Senate victory just 13 months later.

While there are many voting patterns to dig into, we will highlight two trends in particular. As we said in our preview of the Alabama race last week, a Jones win map had to include notably better performances in the big urban and suburban counties of Alabama by way of strong black turnout and improved Democratic performance among college-educated whites. Jones achieved that handsomely. Using the 2012 election for Chief Justice of the Alabama Supreme Court as a baseline — a race Moore won by only about four percentage points, 52%-48% — we can see where Jones improved the most over the Democrat in that election five years ago, Bob Vance. Jones ran between six and seven points ahead of Vance in Jefferson County (Birmingham), Madison County (Huntsville, and in Shelby County, the key Birmingham suburban locality that is solidly Republican. He also ran four points ahead of Vance in Mobile County and two points ahead of Vance in the Black Belt as a whole.

Additionally, if we just look at the college-educated non-Hispanic white percentages of counties and how Moore performed compared to the 2012 race, there was a fairly strong negative relationship between those two variables, as shown in Chart 1 (R-squared = .47). That is, the larger the share of the white population with at least a bachelor’s degree, the worse Moore tended to perform compared to his 2012 contest, and vice versa. Exit polls are imperfect, especially on education questions, but the Alabama exit showed that Jones won 40% of white college graduates and 22% of whites with no college degree (31% of the white vote overall). The improvement for Jones in the metropolitan areas and Moore’s improvement in some of the more rural regions align with those findings. Considering the fact that the 2012 presidential exit poll showed Barack Obama winning just 15% of the white vote, all of this was more than a little remarkable.

Chart 1: Scatterplot of Moore’s change in performance from 2012 to 2017 vs. the percentage of the non-Hispanic white population with at least a bachelor’s degree, by county

For President Trump, Tuesday’s result means he struck out twice in Alabama — the primary by backing appointed Sen. Luther Strange (R) and then the special by supporting Moore — in a state where he arguably should have had as much influence as just about any other across the nation. This brings into question his ability to pull GOP candidates to victory even in places where he is relatively popular. This too should worry Republicans as they look to the midterm election.

Franken Out? How We’d Rate a Minnesota Senate Special

Plus two House ratings changes

Kyle Kondik, Managing Editor, Sabato's Crystal Ball December 7th, 2017


As of Wednesday night, it appeared as though Sen. Al Franken (D-MN) was poised to announce his resignation from the Senate on Thursday morning. Franken has faced several credible accusations of groping women and making unwanted sexual advances, and on Wednesday, the dam finally broke and a slew of his Democratic Senate colleagues began asking for his resignation.

Assuming that Franken does resign — Minnesota Public Radio reported he would late Wednesday afternoon — Gov. Mark Dayton (D-MN) would appoint a replacement for Franken who would serve until a special election in November 2018 to fill out the remainder of Franken’s term. The regular election for this seat would be in 2020, meaning the eventual winner next year would have to turn around and run again in 2020.

If Franken does resign, we will start the Minnesota Senate special election at Leans Democratic, which is also where we have the state’s open gubernatorial race rated. The Democrats have a deeper bench of talent in the state and Dayton will have the choice of several quality candidates to appoint, including state Attorney General Lori Swanson and state Auditor Rebecca Otto (Swanson may be a gubernatorial candidate and Otto already is), as well as Lt. Gov. Tina Smith, among others. One would think Dayton would strongly consider a woman given the number of impressive female Democratic candidates in the state and the fact that Franken is leaving because of his bad behavior toward women. The midterm environment, which should have a Democratic lean with President Trump in the White House, helps Democrats in Minnesota too, and Sen. Amy Klobuchar (D-MN) might provide some coattails from what could be a strong reelection performance for the regularly scheduled Senate election already happening next year.

One could make the argument for Toss-up: Minnesota is actually trending Republican, and while Hillary Clinton carried the state by 1.5 points in 2016, the state’s two-party presidential margin was actually slightly more Republican than the national average for the first time since 1952. If the Republicans can find a strong candidate — former Gov. Tim Pawlenty (R) has been a rumored candidate for his old job and could hypothetically consider this race, too — and/or if Dayton makes a weak appointment or the eventual Democratic nominee in 2018 turns out to have problems, this may end up being a Toss-up.

At the very least, the Democrats now have to play defense in yet another competitive seat next year, thanks to Franken’s indiscretions, in a year (2018) where they already were defending 25 of the 33 seats being contested. Now that looks like 26 of 34, and no party has had to defend a more lopsided Senate map in a postwar midterm. The fact that Democrats have so much territory to defend — 10 of their Senate states were carried by Trump in November 2016 — means the Democrats will be hard-pressed to net seats even if national conditions otherwise favor them next year.

One bright spot for Democrats on what was otherwise a trying day was a report from the Nashville Post that former Tennessee Gov. Phil Bredesen (D) has decided to run for the open Senate seat in Tennessee. We moved Tennessee from Safe Republican to Likely Republican last month in anticipation of Bredesen’s candidacy, so we’re not going to move the race any further for now. In our analysis last month, we noted the similarities between Alabama and Tennessee, and we want to see how the Alabama Senate special plays out next week before further upgrading Democratic chances in another dark red southern state. But at the very least, the Bredesen candidacy does make the Republicans play a little bit more defense next year when they would prefer to be playing offense, and it could develop into something more threatening if his campaign and national conditions allow.

House ratings changes

Last week, we made a slew of new House ratings changes and argued that the race for the House next year is basically a coin flip. Next week, Crystal Ball Senior Columnist Alan Abramowitz will analyze how big of a margin Democrats might need in the national House vote to win back the lower chamber. In the meantime, we announced two House ratings changes earlier this week that we wanted to explain in more detail. One concerns a Democratic recruiting victory in a red-leaning district, and the other concerns the question marks surrounding a freshman Democrat who has been rocked by sexual harassment allegations.

KY-6: Rep. Andy Barr (R, KY-6) has held this district since 2012, when he defeated then-Rep. Ben Chandler (D) after coming up just short against him in the 2010 Republican wave. He then captured about 60% in both 2014 and 2016 in noncompetitive reelections. But the Lexington-based district has some Democratic DNA despite its recent rightward shift, and Lexington Mayor Jim Gray (D) narrowly carried the district as the Democratic nominee in his unsuccessful challenge to Sen. Rand Paul (R-KY). While retired Marine pilot Amy McGrath (D) generated a lot of national buzz and fundraising for a strong debut ad, some national Democrats were holding out hope that Gray would run for the seat, and he announced his bid earlier this week. Those same national Democrats angling for a Gray bid argue that the popular Lexington mayor makes this race a Toss-up (or even better) for them, but we’re not going to go that far given Barr’s proven performance the past few elections and the generic lean of the district. Still, this could be a real Democratic opportunity next year, and we’re moving the rating from Likely Republican to Leans Republican.

NV-4: After defeating then-Rep. Cresent Hardy (R) in a close race last year, Rep. Ruben Kihuen (D, NV-4) seemed well-positioned to solidify his hold on this bluish purple district, which Hillary Clinton won by five points in 2016 (down from Barack Obama’s 10-point win in 2012). But he now finds himself caught up in the wave of sexual harassment allegations after Buzzfeed reported late last week that Kihuen made sexual advances toward a woman who worked on his 2016 campaign. House Minority Leader Nancy Pelosi (D, CA-12) and Democratic Congressional Campaign Committee Chairman Ben Ray Luján (D, NM-3) quickly called for his resignation. It’s unclear what Kihuen will do: On one hand, Kihuen’s chief of staff is apparently circulating the resumes of Kihuen’s staffers in an apparent effort to find them jobs, an indication that Kihuen might be close to resigning. On the other hand, Kihuen says he won’t resign and claims that Pelosi and Luján knew about these allegations last year: “They looked into them. They didn’t find anything, and they continued investing millions of dollars in my campaign. They went out there and campaigned for me.” Pelosi and Luján denied this. Kihuen said he plans to make a statement in the next few days; if he eventually does resign, there will be a special election, and the two parties will pick the candidates. It’s better for Democrats if this becomes a vacant or open seat. National Republicans like Las Vegas City Councilman Stavros Anthony (R), who was running for the seat prior to the Kihuen reports emerging, but if there’s a special, Jon Ralston of the Nevada Independent suggests that the seat’s two previous members could end up as the respective party nominees: Steven Horsford (D), who won the seat in 2012 after it was first created because of Nevada’s population growth, and Hardy (R), who beat Horsford in a big upset in 2014 and then lost to Kihuen in 2016. We’re moving this district from Likely Democratic to Leans Democratic as we wait to see whether Kihuen remains in Congress and, if he doesn’t, what the special election field looks like. This is another seat (like the very likely Minnesota special Senate election discussed above) where the generically pro-Democratic environment prompts us to give the Democrats the benefit of the doubt.

Table 1: Crystal Ball House ratings changes