OZY's Election Model: Dems a Near Lock to Regain Control of House
WHY YOU SHOULD CARE
The blue wave increasingly looks like it will take the House of Representatives but not dent the Senate.
By Daniel Malloy
For Republicans, the national political picture brightened somewhat last week as the Democrats’ generic congressional ballot advantage dipped down to 8 percentage points and President Donald Trump’s approval rating ticked up slightly. But in the battle for the U.S. House, GOP chances look grimmer than ever — as our exclusive election forecast now gives Dems a better than 90 percent chance of taking over the chamber.
What happened? A deluge of polls in individual races that underscore Republicans’ problem in holding onto the lower chamber. The battleground races are in suburban areas where voters — particularly college-educated women — are fleeing the party in droves. As the brutal Supreme Court confirmation fight over Brett Kavanaugh gets saturation media coverage amid allegations of sexual misconduct, a reversal of fortune looks increasingly unlikely.
In partnership with the Republican technology and data firm 0ptimus, we crunched more than 100 factors that helped predict past elections, with some extra weighting for unique aspects of this political year to produce these forecasts. For more on how these numbers were derived, scroll down to the bottom of this story. For more exclusive election coverage, subscribe to our Midterms in a Minute newsletter.
Pollsters [are] typically finding results more favorable to Democrats than ratings or models would otherwise indicate.
Alex Alduncin, data scientist for 0ptimus
Of the 14 House races that saw the biggest swings this week in our model, 11 went in the Democrats’ favor — including districts in the suburbs of Seattle, Kansas City, Dallas and San Diego. For example, new polls show U.S. Reps. Kevin Yoder, R-Kansas, and Pete Sessions, R-Texas, suddenly finding themselves in toss-up races after winning easily in past years — as both face first-time candidates of color. “This has been a constant theme over the last few weeks, with pollsters typically finding results more favorable to Democrats than ratings or models would otherwise indicate,” says Alex Alduncin, a data scientist for 0ptimus.
Alduncin points out that the national environment, favoring Democrats by 8 percentage points, is about the same as it was at this point in 2006. That year, Democrats picked up 31 seats in the House and six in the Senate. But this year, the Senate map is far more favorable for Republicans: Our model projects Republicans will gain one Senate seat and, with an 86.9 percent chance of capturing at least 50 seats, retain control of the chamber.
That’s a slight improvement for Republicans over last week, given the small shift in national winds in their direction. In individual races, solid polls for Sens. Bill Nelson, D-Florida, and Jon Tester, D-Montana, improved their chances. Nelson now has a 67.5 percent chance of holding the seat, while Tester has a 75.8 percent chance — even in a state Donald Trump won by 20 percentage points. Meanwhile, Sen. Joe Manchin, D-West Virginia, saw his re-election chances decline a bit to 66.3 percent.
Expect the OZY/0ptimus model to start getting more volatile in the coming weeks, when the public gets a look at a host of new fundraising data. Already, Democrats in key races are releasing some mind-boggling money totals for the third quarter: Josh Harder, a challenger in California’s 10th Congressional District, and Amy McGrath, a challenger in Kentucky’s 6th Congressional District, both raised more than $3.5 million for the quarter — unheard-of for House races, and another sign of the energy on the left.
In addition to our extensive on-the-ground coverage of races across the U.S. this year, OZY wanted to build a better product to analyze the national political picture — given the failure of such forecasts in the past. So we decided to team up with Washington-based 0ptimus, a Republican firm that developed an unbiased, nonpartisan prediction model so it could show its clients in the worlds of both politics and finance where the winds were blowing.
0ptimus’ data team created and tested countless models, crunching publicly available data against past results in House and Senate races. They take into account more than 100 variables, including past vote totals, generic ballot surveys of which party voters prefer in Congress, the unemployment rate, fundraising data and public polling. The firm developed an artificial intelligence system to “smartly” average together several models to create the strongest prediction numbers for the 2018 elections, always automatically testing against past elections to assess quality. You can read more about the 0ptimus methodology here.
Given the unique factors of 2018, we asked 0ptimus to tweak its calculations, adding additional weight to:
- The number of small donations — a sign of energy for candidates from Trump to Alexandria Ocasio-Cortez.
- Gender — women are doing exceptionally well this year, and we expect that trend to continue.
- Trump’s approval rating — as he hangs over the political and media scene with a heavier presence than past presidents.
Meanwhile, we asked 0ptimus to reduce the weight for:
- Candidate ideology — their calculations rewarded more moderate candidates, while we think this election year is all about firing up the base.
- Outside money — as OZY has reported, advertising is less persuasive coming from a super PAC than from a candidate.
Because 0ptimus averages several models, for us they introduced a new Bayesian model to the mix that included Trump approval, gender and unitemized donations while removing one that included ideology and independent expenditures. The result is the numbers we update for you each week.