Your browser does not support JavaScript The Battle of the Bases: Negative Partisanship Will Decide Election 2018 - Judy Ford Wason Center for Public Policy - Christopher Newport University The Battle of the Bases: Negative Partisanship Will Decide Election 2018
Skip navigation

Judy Ford Wason Center for Public Policy

Wason Center

November 2, 2018

The Battle of the Bases: Negative Partisanship Will Decide Election 2018

Midterm / National / Elections

Donkey and elephant butting heads

Democrats Poised to Net 45 House Seats

Hostile Senate Map Likely to Lead to Democrats Winning Nevada and Arizona, but Losing North Dakota and Tennessee

Democrats Expected to Have a Banner Day in Gubernatorial Races, Country May See Two Southern States Flip to Democrats 

Just over 4 months ago I posted my forecast for the 2018 congressional midterms. The forecast offers a novel approach to modeling election outcomes based on a theory of voting behavior for the polarized era. My forecast asserts that the results of the 2018 midterm cycle, especially for House and gubernatorial races, have been set in stone since just around 11pm on November 9th, 2016, when Hillary Clinton lost her bid for the presidency unexpectedly to Donald Trump. That loss cemented a considerable enthusiasm advantage for Democrats that will last the length of Trump’s tenure in the White House. The first evidence of this enthusiasm advantage appeared in a series of upset victories by Democrats in state legislative special elections in hostile territory, followed by Democrat Ralph Northam’s 9pt trouncing of his Republican opponent Ed Gillespie in Virginia’s 2017 gubernatorial race. That win also ushered 15 new Democrats into Virginia’s House of Delegates, dramatically altering the partisan division of the state’s lower chamber. Yet, other political analysts expected the election to be close between Gillespie and Northam and were genuinely surprised by Northam’s lopsided victory. They were caught off guard because their understanding of Virginia’s electorate was based entirely on the electorate that existed during the 8 years of the Obama Administration, where strong turnout for Republicans and weaker turnout for Democrats erased much of a growing demographic advantage for Democrats in the ever-expanding northern Virginia area. Democrats went from comprising 37% of the 2013 electorate to 41% of the 2017 electorate and in so doing, increased the vote share for the Democratic Party’s nominee from 48% of the 3 candidate vote in 2013 to 54% of the 3 candidate vote share in 2017.

This sharp change in the partisan distribution of the electorate also drove Conor Lamb’s surprise victory in a special election in Pennsylvania’s 18th congressional district, which at the time held an 11pt partisan advantage for Republicans. According to exit polling, Democratic voter turnout was so strong in that largely rural district that Democrats made up a plurality of the electorate, 46%. It happened again in a special election in Alabama to fill Jeff Sessions’ vacant Senate seat. In a state with a partisan advantage for Republicans of 15pts, Democrats turned out in such large numbers they made up 37% of the electorate, compared to the Republican’s 43%, an impressive feat in a deep red state in the Bible Belt.

What fueled these sharp increases is something called negative partisanship, which I explain in great detail in my original post. In the polarized era, Republicans and Democrats increasingly view each other as existential threats to the survival of the Republic, especially those that the PEW Center calls the “highly engaged.” As such, fear and hatred of the “other party” motivates voters more than appreciation of one’s own party. I argue that this has had a profound effect on voting behavior. During the Obama years, Republicans felt threatened by the Democratic Party’s control of the presidency and responded by increasing their turnout rates in the 2010 and 2014 midterms. The result was two banner cycles for the party where they not only picked up control of Congress but also gained about 1,000 additional state legislative seats and a dozen new state houses, even in blue states. Republican gains were helped by sluggish turnout by Democrats, who, having completed their “Yes, We Can” revolution, scaled back their electoral interest considerably in the midterms. Now the shoe is on the other foot, and Democrats are threatened. This threat response provides a useful framework to anticipate what will happen in the 2018 midterms and beyond.

My Negative Partisanship Model produces an estimated two-party vote share for Democratic Party candidates running in the 2018 cycle. My quantitative model identifies districts/states with demographic characteristics most conducive to a surge in Democratic voter turnout. These factors include the level of partisan competition in the district, the % of the district’s population that is college educated, the level of diversity in the district, and the district’s population density. A full description of my model can be found here. In addition, I implement a fairly simplistic quantitative handicapping system similar to the one used by FiveThirtyEight, which considers contest-specific factors such as incumbency, primary turnout, fundraising, candidate quality, campaign strategy, presence of a competitive statewide race, scandals, and third-party candidates to refine my model’s estimate. For example, in Georgia’s 6th congressional district, where 65% of the population holds a college degree and 40% is non-white, my model is quite bullish on the Democrat’s share of the two-party vote despite the fact that the district broke for the current Republican incumbent Karen Handel by about 4pts in a special election early in 2017. My model’s initial estimate of the Democrat’s two-party vote share has been adjusted down to account for Handel’s incumbency, the strong turnout of Republicans in their regular Republican primary and in the subsequent run-off election, and for Handel’s considerable fundraising advantage, but adjusted up to account for the historic candidacy of Stacey Abrams as the Democratic Party’s nominee for governor, which will almost certainly increase turnout for Democrats, particularly African-Americans, who underperformed their turnout potential in the 2017 special.

Because my initial House ratings were released in July, at the time my ratings largely reflected the unadjusted output of my statistical model (although there were a couple of key exceptions, VA 7 for example which received a bump for challenger quality and for Democratic turnout in the primary which allowed me to put this district on the radar far earlier than other analysts). My final ratings reflect both the estimate of the Democratic candidate’s two-party vote share as well as any applicable adjustments from the factors listed above. That being said, there is not much change from my original forecast although my gross seat gain for Democrats has increased to 47 seats. This is largely due to the unique approach of my model, which derives the bulk of its estimate from fixed factors. As I said in my original post my model was designed to closely reflect the seat gain for Democrats 4 months later, on Election Day. Rather than moving my seat count, the last few months have been about filling in the blanks on which specific seats will flip. In my initial release, I had 12 districts coded as “Will Flip” and an additional 12 coded as “Likely to Flip.” I had 3 races rated as “Lean Democrat” and then I had a long list of “Toss-Up” races. Heading into this final update my “Will Flip” list more than doubled, to 26 seats, which locked in the majority for Democrats. I had 11 “Likely to Flip” districts, 7 “Lean D” districts, and 25 “Toss Ups.” This final update identifies 47 seats my analysis suggests Democrats will pick up, offset by one they are likely to lose to the Republicans (MN 8) and one they will lose from the Pennsylvania redistricting (PA 14) for a net gain of 45 seats.

It’s important to remind you that when released, my House model was an outlier. At the time, the generic ballot had been narrowing and the debate was whether Democrats could even win the 23 seats they need to take control of the House. The predictive forecasting models were producing around a 50% probability of a Democratic Party takeover. In my original post, I argued that my then-radical forecast would be in line with other forecasts by Election Day, and I am happy (and relieved) that is exactly what has happened. 5 days out, the FiveThirtyEight model gives Democrats an 85% probability (or 6 in 7 chance) of taking control of the House and a mean seat gain for Democrats of 38 seats (ranging from 19 to 59 seats). The Optimus model shows a 95% probability that Democrats will win control of the House, with a net mean seat gain of 39 seats (ranging from 217-247). At this point, if I am wrong, I will be in very good company.

Before I get into my final race ratings, let’s discuss the evidence we have that supports a Blue Wave, and assess any evidence that contradicts it. While the aggregate forecasting models are primarily concerned with the generic ballot gap, my key data point from survey data is something called the enthusiasm gap. The enthusiasm gap is a measure that compares how excited and tuned in to the election Republicans are compared to Democrats. The larger the advantage for one party, the more likely it is for the party that benefits to turnout out in what are traditional, low turnout elections. Despite all the talk about a so-called Kavanaugh Effect, which purports that the way Democrats handled the Brett Kavanaugh confirmation process erased what has been 24 months of an enthusiasm advantage for Democrats, there is little evidence that suggests that Republican voters are as riled up as their Democratic Party counterparts, who also outnumber them in a great many places. Although some surveys found an uptick in enthusiasm for Republicans during and after the confirmation process my own polling in Virginia revealed no evidence of a suddenly galvanized electorate. We were in the field before, during, and after the confirmation process. In all 3 of our surveys which were in the field for over a week each, the enthusiasm gap for Democrats we measured last spring remained unchanged: 15pts in VA CD 10, 10pts in VA CD 2, and 16pts in VA CD 7. These three districts each contain unique demographics and represent a wide range of voters. Large turnout is all but sure to advantage Democrats aside from in a few key states, which I will discuss below.

Early voting also reveals advantages for Democrats. To be sure, early voting data is noisy, and not that useful for predicting election results. But early voting data is sending us one very clear signal, voters are highly engaged and in most places, turnout is going to be high on both sides, perhaps the highest its been in decades in a midterm cycle. In 7 of 9 states, Republicans have returned more ballots than Democrats, although Democrats have outpaced Republicans overall in Nevada and North Carolina. The early vote returns show signs of strength for Republicans in some of their best-fortified strongholds like Tennessee, but in other red states like Texas and Georgia, Republicans have returned more ballots, yes, but their lead is significantly smaller than it should be, especially in Georgia. Although older voters are outperforming younger voters in early voting thus far, older voters always make up the largest block of early voters and what we’re seeing is strong evidence of more millennial and minority early voting relative to their performance in 2014. In Texas, young people have increased their share of the early voting electorate 5Xs over their 2014 early voting and Democrats, powered by strong turnout rates among African Americans are turning out in big numbers both in North Carolina and Georgia. And perhaps most importantly women are outperforming men in several key states including Georgia, Tennessee, Texas, and Montana. In Georgia, women are beating men in the early vote by double digits.

Another very good sign for Democrats is that Democratic Party nominees have outraised their Republican opponents. In just the last two weeks, Democratic House candidates raised $65 million dollars, outraising their Republican opponents by more than $30 million dollarsSurvey and early voting data also show potential for the sharp increase in turnout among millennials, Latinos, and women, key to producing a seat gain large enough to take control of the House and to win key Senate races. Largely thanks to the efforts of the New York Times Upshot/Sienna College polls, which conducted nearly 100 polls of House races, the 2018 cycle has been the most polled midterm in history. The collective story from their polling is that most competitive House races are within the margin of error and contain enough undecided voters to break either way, assuming the weights used end up representing each district’s demographic composition accurately. My own polling of 3 of Virginia’s competitive elections reveals just how bad the Republican Party’s problems are in suburban districts and that Democrats are competitive in districts normally inhospitable to Democrats such as VA CD 7. Democrats have also maintained an average generic ballot lead of about 8pts, which produces about a 30 seat gain in political science’s most reliable forecasting model.

It's also important to keep in mind two things that are only visible from 30,000 feet. First, we saw a historic number of Republican retirements leaving Republicans forced to defend 40 open seats. Although a handful were from members who left to serve in the Trump Administration, the majority left due to strategic retirement, which is when a member chooses to retire because they expect to lose their reelection. In 2010 it was Democrats that fled Congress in droves and the end result of those open races produced almost a total sweep for Republicans. In the case of the 23 vacancies created by strategic retirement, I expect Democrats to pick up the vast majority of these seats. Another long-view that gets lost in analyzing the midterms at the molecular level is the fact that our Senate map is focused on states like Texas and Arizona and NOT on states like Ohio, Michigan, and Pennsylvania. The fact that none of these three races is competitive and that Scott Walker is facing the race of his life trying to win his reelection as Wisconsin’s governor reminds us just how hostile the political terrain really is for the GOP in 2018.

While by no means conclusive, up to this point, there is nothing I am seeing that gives me a reason to reassess my forecast and plenty of things I needed to see to support a stronger than expected electoral performance for Democrats on Tuesday. So, with no further delay, I will give you my final race ratings for the House, Senate, and governor races. However, a word of warning on the Senate races. Like many other forecasts, my model relies strongly on the level of partisan competitive in a race and this Senate map strongly favors Republicans. No matter how large the Blue Wave is, it is going to run up against a heavily fortified Republican levee in the Senate. It's hard to know which force will prevail until the size of the storm surge is known. If Democrats pick up less than 30 House seats, Republicans are likely to net 2 or even 3 new Senate seats, but if Democrats pick up 40 or more House seats the overall net gain for Republicans in the Senate may be 0 or even negative. I’m not alone in this uncertainty. Although FiveThirtyEight’s model can tell us with pretty solid confidence that Republicans will maintain control of the Senate, most of the key races are at, or close to, a 50/50 split. When you are handicapping races with a 50% probability you need to make an educated guess as to which way the race breaks. As I discussed in my original Senate post Democrats running in states with manageable PVI scores are in a better position than those running in states with PVI scores in excess of 10pts, with the exception of Joe Manchin who is running in a state that still engages in meaningful levels of split-ticket balloting. Missouri, Arizona, Nevada, Texas, Indiana, and Florida all have large metro areas and/or have populations that are heavily concentrated in large, urban areas. These cities, and their surrounding suburbs, are the key source of voters for a Blue Wave. North Dakota and Montana are more Republican now then they were when their Democratic incumbents won election in 2012 AND 2012 was a presidential election year with optimal turnout for Democrats to win close races. And both won narrow victories. Still, we have two things to consider. First, 38 out of 38 Senate incumbents have won reelection when their party is on the right side of a midterm effect. Even one loss by a Senate Democratic incumbent would be a significant departure from historical trends. Second, my analysis of polling in the last two midterms reveals two things. In general, even though aggregate leads are not statistically significant, the candidate with the lead, no matter how narrow, tends to prevail. Second, statistical ties tend to break in favor of the party benefiting from the midterm effect. I keep these two things in mind in my final Senate ratings. Finally, as I discussed in my original post, the governor race map is the opposite of the Senate map. In general, Democrats have a structural advantage under this map and when that advantage meets the enthusiasm gap, it should produce a good day in state-level races for Democrats.

Report a problem
Version 3.4