Princeton Election Consortium

Innovations in democracy since 2004

Outcome: Biden 306 EV (D+1.2% from toss-up), Senate 50 D (D+1.0%)
Nov 3 polls: Biden 342 EV (D+5.3%), Senate 50-55 D (D+3.9%), House control D+4.6%
Moneyball states: President AZ NE-2 NV, Senate MT ME AK, Legislatures KS TX NC

Can Montana and South Carolina slip off the Presidential coattails?

November 1st, 2020, 9:05am by Sam Wang

In our polarized political age, downticket races are increasingly driven by partisanship. Yet we still pay close attention to individual candidates – in Senate races, Joni Ernst’s inability to name the price of soybeans, or Cal Cunningham’s extramarital affair. Can races be simultaneously nationalized (i.e. track partisan affiliation) and separated from the Presidential race?

Here is evidence that both can happen at once. It points toward the possibility that in close Senate races such as South Carolina, Iowa, and Montana, last-minute phone-banking, canvassing, and even (for Montana) same-day voter registration can be exceptionally valuable in moving a race.

Above is a graph showing poll medians taken from PEC’s public archives for 13 key Senate races which are currently within 10 percentage points. Montana and Mississippi are omitted because there is little data available from summer. Each state is plotted against Presidential data for the same date (“EV” files in the archive).

The fitted red line has a slope of 0.86, which means that for every 1 point of difference between states in Presidential polls, the corresponding Senate poll varied by 0.86 points in the same direction. That’s close to 1, and shows that during the summer, Senate races mostly reflected the Presidential race. The party leading the Presidential race also led the Senate race in all but two states, Iowa (where Greenfield was doing better than Biden) and Georgia (where Perdue was doing better than Trump). So in August, the coattail effect was quite strong.

By the end of October, this effect was much weaker. The slope of the fitted black line is 0.43, only half as much as August. This means that Senate races were much less dependent on the Presidential race.

One might initially conclude that as the fall progressed, voters started paying attention to Senate candidates on their own merits. However, there is evidence that partisanship still matters quite a lot. The leading party for President and Senate is now perfectly matched (though Iowa, at Biden +0.1% and Greenfield +0.5%, is really a tie). This is reflected in two parameters: the correlation coefficient r stayed high, +0.88 in August and +0.81 in October. Also, both fit lines go nearly through the origin.

What does this mean? To me it says that even though Senate and Presidential races are somewhat separated from one another, they are both still driven by a deeper factor: partisan loyalty.

When we react to individual races, we may think we are forming narratives that is fundamentally driven by events. That’s certainly the way political news coverage sounds. For example, it is said that Jon Ossoff (D-GA) gave Perdue (R-GA, incumbent) a thrashing in their debate last week. But to what extent are those stories just making up post-hoc reasons to support our underlying partisan biases? What would a Perdue supporter say? This would also explain why news of Cal Cunningham’s (D-NC) extramarital affair got so little traction. If Cunningham’s party is slightly favored in North Carolina, hotbed of partisan polarization, his texting habits might be beside the point.

Having established that partisanship is key, let’s get back to the question of when more people start tuning in to individual Senate races.

The separation between Senate and Presidential races has been gradual, starting in July. By Labor Day (Julian date 251 this year), the Senate-vs.-Presidential slope had come close to its current value.

Senate front-runners are now underperforming their party’s Presidential nominee in most states. Exceptions are Texas, where Cornyn (R) is safe but Trump is within 1 percentage point of losing, and Arizona, where Kelly (D) is ahead by 7 points but Biden is only ahead by 4 points. So generally it would seem advisable for Senate candidates to tie themselves as closely as possible to their own Presidential candidate. For example, that explains why Senator Perdue (R-GA) chose to attend a Trump rally rather than engage in a second debate with Jon Ossoff (D). The aforementioned public reaction to the first Ossoff-Perdue debate might also have played a role.

All of this is is a fancy way of looking at the fact that in recent weeks, Senate races have narrowed more than Presidential races. It means that there are only a few races where a state might end up electing Senate and Presidential candidates of opposing parties. But there are a few.

Here are the Senate races that are within polling error. Based on past Presidential-year data, Senate poll medians could be off by as much as 3-4 percentage points overall. The states where individual candidates have the best shot at overcoming the partisan leaning of their states are Montana and South Carolina. If you’re phone-banking this weekend, these are great places to focus. And of course Iowa, which is basically a tie.

(And if you care about state legislatures and Redistricting Moneyball, don’t forget Texas and North Carolina!)

Tags: 2020 Election · Senate

7 Comments so far ↓

  • Frances Smith

    Does Hank Gilbert have any chance to beat Louie Gohmart in Texas Dist. 1? If he does, I wil send a donation.

    Fran Smith

    • Sam Wang

      Not a chance. A waste of money.

      If you want to stick it to Gohmert, a better move is to give to Tarrant County Democrats or Harris County Democrats. Bipartisan redistricting in 2021 will at least have some effect on his district.

  • Randall G.

    Which candidates / races do you think might have reverse coattails this week (i.e., VA in 2017 with Northam & downballot Dems for VA House)?

  • Richard Vance

    Alabama Senate is closer than sparse polls indicate. Heavy blue areas are highly motivated I’m in North Alabama with about a 10-2 ratio of Doug Jones yard signs over Tubby yard signs.

    • Sam Wang

      Yard signs? That is not persuasive, though I hope you find comfort in having agreeable neighbors.

      There are three polls in the last two weeks in Alabama, which is not that bad.

  • Jim Strang

    Do you see ANY House movement in Ohio, or is its exquisitely Gerrymandered Velcro going to hold for the GOP?

Leave a Comment