Your browser does not support JavaScript What Would a Partisan Surge from Negative Partisanship Look Like? - Judy Ford Wason Center for Public Policy - Christopher Newport University What Would a Partisan Surge from Negative Partisanship Look Like?
Skip navigation

Judy Ford Wason Center for Public Policy

Wason Center

November 6, 2018

What Would a Partisan Surge from Negative Partisanship Look Like?

Midterm / National / Elections

3D US map broken into random fragments

My negative partisanship model argues we should expect to see a significant surge in turnout for Democrats in today’s election over their 2010 and 2014 turnout rates. My model is partly based on the 2017 elections in Virginia, which many analysts expected to be highly competitive between the Republican Party nominee and the Democratic Party nominee. Instead, that election produced a blow out win for Democrat Ralph Northam and dragged 15 new Democrats in the state’s House of Delegates with him, a feat that should have been impossible due to the state’s Republican partisan gerrymandering. Heading into that fall’s election, I expected an uptick in Democratic Party enthusiasm, and by extension, their turnout, due to backlash from Donald Trump via a mechanism called negative partisanship. Negative partisanship is the recent phenomena that finds that Republicans and Democrats are more motivated by their fear of the other party than appreciation for the benefits ascribed to them by their own party. As such, the party locked out of power in the White House is highly motivated to show up in subsequent elections in order to fight back against their opposition party “enemies.” Based on this new understanding of the electorate I attempted to account for a change in the demographic composition of the electorate in the 2017 Virginia elections. Looking back at the 2013 gubernatorial election the partisan split of the electorate was 37% Democrat, 32% Republican, and 31% Independents which produced a 2.6pt win for the Democrat, Terry McAullife. Having lost Independents by 9pts McAullife’s win was made possible by the changing demographics of northern Virginia and the Republican Party’s nomination of a controversial ideologue which allowed the Democratic Party to disrupt a decades-long trend of electing a governor from the party opposite the winner of the previous year’s presidential election.

I thought we might see as much as 40% of the 2017 electorate identify as Democrat. This increase in Democratic identifiers would likely produce a win for their nominee in excess of 3pts and perhaps even higher than 5pts. We ran a total of 4 surveys on that race and in each of them, Republicans kept pace with Democrats reasonably well. The natural mix of our data on average produced an electorate that was predicted to be 34% Republican, 36% Democrat, and 28% Independent. And even under this partisan breakdown, Northam had a strong lead outside of the margin or error. Our first survey produced a 7pt advantage for Ralph Northam and the remaining three surveys produced 6pt leads. Despite being technically correct (our 6pt win was buffered by a 3.8pt margin of error) we ended up undercutting Northam’s margin precisely because our survey underestimated the proportion of the electorate that would be made up of Democrats. Exactly as I expected, actual Democratic voter participation surged and Democrats ended up making up 41% of the electorate. Independents were almost evenly split, with 50% voting for the Gillespie and 47% for Northam. That additional 5pt share of the electorate was a big difference maker. Not only did it drive Northam’s unexpectedly robust win, it nearly flipped control of the House of Delegates to its long-suffering minority party.

Perhaps one of the most underappreciated aspects of contemporary elections in general, and election polling specifically, is how powerful a predictor of vote choice partisanship has become in the polarized era. For most contests, in most places, 90% of Republicans vote for their party’s nominee and 90% of Democrats do the same. This was true even in the highly controversial 2016 presidential election, which created the unprecedented “Never Trump” movement. It was also true in the special Senate election last year, where 90% of Republicans cast ballots for a man credibly accused of child molestation to protect their partisan interests. Heading into the ballot counting tonight I thought it might be useful to illustrate just how large an effect a shift in the composition of the electorate can have on election outcomes by demonstrating its power on the surveys used to predict them. This cycle, in lieu of a competitive statewide senate race in Virginia, the Wason Center polled three of the state’s most competitive congressional districts. Despite expecting a surge in Democratic Party turnout we did impose party weights on our data. Instead, we produced two likely voter models: one that included respondents that indicated they would probably or definitely vote and a second we called a “committed” voter model which limited our results only to voters who indicated they would definitely vote.

In the 2nd district, both of our models we ended up with fairly equal splits between Republicans and Democrats which in itself may be capturing evidence of Democratic voter surge due to the fact that aside from the 10th district, the 2nd and 7th districts are majority Republican. In the 2nd district our partisan breakdown for both versions was 36% Republican, 33% Democrat, and 29% Independent and our survey found a 7pt lead for Republican incumbent Scott Taylor over his Democratic Party challenger Elaine Luria in our regular model and a 6pt Taylor lead in our “committed” voter model. Our survey in the 10th district had a partisan breakdown of 32% Republican, 33% Democrat, and 32% Independent which produced a 9pt advantage for Democrat Jennifer Wexton. The “committed” voter model in the 10th did not change the partisan composition much, increasing the Democrat percent to 34%. That model produced a 12pt advantage for Wexton. And in CD 7 survey our partisan breakdown ended up at 34% Republican, 34% Democrat, and 32% Independent in our regular model and 33% Republican, 35% Democrat, and 32% Independent in our “committed” voter model. In that race, our regular model produced a 1pt margin between Spanberger and Brat, a statistical tie, and in our “committed” voter model, Spanberger leads by 3pts, still inside the margin of error. The chart below shows the impact a partisan surge would have on these three races. The partisan split imposed on the data reflects the statewide partisan split from the 2017 election.

Will we see a partisan surge for Democrats in today’s midterm elections? The potential is certainly there given that 30 million people have participated in early voting this cycle, putting 2018 turnout on track to more closely mirror presidential cycle turnout than turnout in recent midterms. If the partisan surge happens, I will not be surprised to see many races that polled within the margin of error produce decisive victories for Democrats and would expect Democrats to come in somewhere between 30 and 50 seats. A partisan surge is also required to pull Democrats running in hostile territory over the finish line in places like Georgia, Tennessee, and Texas, even if Independents break in their favor. If Democrats have a big night, look for the partisan breakdown in the exit polling. It will reveal a significant surge in the Democratic share of the electorate. If the Democrats underperform expectations it will almost certainly be because the partisan surge didn’t manifest. Under that scenario, Democrats may still have enough juice to flip the 23 seats they need for a majority, but the large wave will not happen. If the Democrats struggle to win their House majority, this will also mean a much tougher night in the Senate elections. Right now, the consensus is that Republicans may gain 1 or 2 Senate seats and if Democrats surge enough to pick up 40 or more seats, they may actually take control of the Senate too. Stay tuned!

Report a problem
Version 3.4