November 8, 2018

2018 Provides a Midterm Backlash With Continued Geographic Polarization



Public opinion polls and prediction models were largely correct. Democrats won control of the U.S. House of Representatives (netting at least 27 seats) and gained seven governorships and several state legislative chambers—but Republicans retained control of the U.S. Senate (netting up to three seats).

Democrats went into the night facing a very tough map in the Senate. They hung on in deep red Montana and West Virginia (as well as Ohio, Wisconsin, Michigan, and Pennsylvania rather easily) and picked up a seat in Nevada, but lost in North Dakota, Indiana, Missouri, and Florida. These results were part of the ongoing nationalization of American politics, where it is more difficult for Democrats to win in Red States or Republicans to win in Blue States.

But it should not distract from the major story of the night: midterm backlash. Amidst solidifying partisanship and sky-high turnout, Democrats’ House seat share grew by the largest amount since at least 1974 (with votes reasonably distributed for seat gains, unlike in 2012). 

This was the traditional pattern: the president’s party usually loses seats in midterm elections—but on a very large scale. The results suggest both a real swing vote from previous supporters of Trump as well as increasing Democratic House voting among Clinton voters who had previously supported Republicans for Congress. Those kinds of changes should be tough in our era of minimal persuasion. One panel study found that just 2% of voters moved from Trump to the 2018 Democratic House candidate.

The continuation of nationalization was extreme. Nearly all races are now highly-partisan contests with voters judging the performance of the president, even for state offices. For example, the correlation between Florida counties’ Democratic share of the 2012 presidential vote and the 2014 governor’s race was already 0.95. But between the same races in 2016 and 2018, it was nearly fully correlated (0.99/1).

Given these trends, I had been skeptical of the possibility of further shifts among suburban educated women (a constituency that had already disproportionately moved from Romney to Clinton in 2016). But further Democratic gains with the group materialized everywhere outside the South. Given the size of the vote shifts, this likely included increased mobilization by voters who do not usually participate, previously split-ticket Clinton voters supporting Democrats across the board, and even some shift toward the Democrats among Trump-supporting women.

Non-southern suburbs moved overwhelmingly toward the Democrats, led by white college educated women and young voters. Although Democrats made gains with everyone, the shift in preferences was most pronounced among those under 40 and in the Midwest.

The Midwest swung back hard to Democrats. But the divide between suburban and urban areas (on the Democratic side) and rural areas (on the Republican side) solidified. The Republican advantage in rural Midwest counties had risen from a 14% gap in 2012 to a 36% gap in 2016. And geographic polarization continued, with Democratic gains coming more in suburban areas and Republican strength growing in rural areas in some states (like Indiana).

But even in highly-partisan elections driven by a referendum on President Trump, candidate recruitment still mattered. In the Senate, early Republican success with Kevin Cramer, Josh Hawley, Rick Scott, and Martha McSally paid off, as did Jacky Rosen for Democrats. In the House, lots of Democratic challengers did far better in the same districts due to solid credentials and fit with their districts. As prior research suggested, women candidates did not seem to do worse than men in similar districts or states, despite the large increase in Democratic women’s representation (Research also suggests that this could further mobilize women candidates for future elections but also increase the partisan gender gap between party voters).

Ideological positioning also seemed to matter. Liberal insurgent House nominees who beat establishment alternatives in swing districts lost some districts (NE-2, PA-1) whereas establishment picks who defeated liberal alternatives went 3 for 3 (PA-7, KS-3, TX-7). Democrats may still pay an electoral cost for running as an avowed liberal. Some Republicans also paid for extremism (Chris Kobach for Kansas Governor and Scott Walker for Wisconsin Governor, for example). But Republicans won both of the direct gubernatorial conflicts between candidates from the far left and right: Brian Kemp vs. Stacey Abrams in Georgia and Ron DeSantis vs. Andrew Gillum in Florida.

Moving forward, geographic polarization may not be as much of a problem for Democrats in the House or the Electoral College vote for the Presidency. In 2018, Democrats will win about as high a percentage of seats as votes (despite complaints of Republican gerrymandering). They also made state-level gains in advance of the next redistricting process. And all the swing states (even Iowa & Ohio but especially Pennsylvania and Michigan) are back on the table. On the other hand, the Senate looks harder than ever as a long-term prospect for Democrats: without Red State senators, winning and holding the chamber will be near impossible.

Geographic polarization is hard to reverse because many of the same factors affecting individual voting also serve as context and social network effects for voters who do not fit their area’s profile. In two consecutive elections, demographic factors at the geographic level grew in importance at end, suggesting that undecideds swung toward the norm in their geographic area. Even educated voters in rural, less educated areas, for example, might have swung toward the Republicans while even less educated white voters in college towns and suburbs moved toward the Democrats. Our communities are growing apart in self-image and political inclinations, as voters are influenced by their friends and neighbors.

The new Democratic House caucus is much more diverse by gender and age, though not by race. Historically, the Democratic caucus has gotten less diverse when they win more seats because White Christian Men disproportionately represented swing districts, so it is new to see diversification in victory. But Republicans remain overwhelmingly White Christian Conservative Men, and some of the few remaining moderates lost their seats. We should thus expect congressional polarization to continue.

Democratic policies made some gains this year as well. Medicaid expansion won in three out of four red states: Idaho, Nebraska and Utah (Altria successfully killed a Montana expansion based on a tobacco tax). Redistricting reforms passed in Michigan and Colorado, marijuana legalization passed in Michigan, and felon enfranchisement passed in Florida.

The elections did not produce a clear result on the parties’ messaging. Trump decided to ignore his consultants, who wanted to end on a message of economic performance, rather than immigration. A panel study finds no evidence that likely voters changed how much they base their vote choice on their immigration attitudes, despite Trump’s recent fixation; there was also no evidence that focusing on the issue mobilized Republicans. The 2018 campaign overall was surprisingly conventional. Most candidates used traditional ads on health and taxes, even at the end, but there was no clear pattern in what issues were most effective in winning votes.

The announced Democratic House agenda for 2019 looks well-tuned to last night’s results. They are expected to lead with political reform, then move to infrastructure and prescription drugs to try to set up Trump to make a deal or (more likely) oppose initiatives he has said he supports. Nancy Pelosi may even have an easier path now to retaining the speakership and be better able to tamp down on the most left-wing initiatives and the desire for impeachment.

Even though the next election cycle has now begun, this week’s results do not change much for 2020. It is possible that some moderate, standard-issue Democrats like Amy Klobuchar or Sherrod Brown will gain steam as electable Midwest alternatives to Elizabeth Warren and Bernie Sanders. But the historical correlation between midterm results and the president’s re-election performance is not high (and depending on the years included, could actually be negative, with presidents more likely to win after negative midterm verdicts). It looks like we will again fight over the same swing states to a surprising degree. The primary candidates will be back in Iowa in no time, and Democrats and Republicans will again be fighting over close vote counts in Pennsylvania and Florida in the general election.