Election models predict victory for Democrats in House for 2018 midterms

The academics have crunched the numbers, and their models say Tuesday should leave Republicans blue.

One model by Joseph Bafumi, an associate government professor at Dartmouth College who specializes in election behavior and campaigns, says Democrats will end up with at least 221 seats in the House, or slightly more than the 218 needed to claim control in the new Congress next year.

“It could be more — not wildly more, but more,” Mr. Bafumi said.

Another model, created by University of Iowa emeritus professor Michael S. Lewis-Beck, a pioneer in the field, puts the likely Democratic gain in the House at between 30 and 44 seats.

James E. Campbell at the University of Buffalo runs two different scenarios, one of them showing Democrats gaining at least 44 seats and the other 62 seats, giving them a total of 257 seats.

“House seat changes are in an era of feast or famine. Based on the history of seat changes, the real question is not which party will gain seats, but whether Democratic House seat gains will be small or large?” Mr. Campbell wrote in introduction to a recent Cambridge University Press collection of 2018 model predictions.

The calculations are an alchemy, looking at anything from polling and presidential approval ratings to economic data.

For instance, Mr. Lewis-Beck’s basic calculation uses the per capita disposable income of a voter and the president’s job approval rating in a Gallup poll, then factors in the fact midterms are historically bad for the incumbent majority. That produced the 30-seat swing.

Adding in more granular factors such as state and local data or polling grows that advantage to 44 seats, Mr. Lewis-Beck calculates.

“Our model is theoretical, what you might call a referendum model,” he said, noting that one of the undergirding principles is midterm elections tend to be the moment at which voters reward or punish the party in power in Washington.

“It’s pretty much an iron law of political science the party with a majority in the House loses seats,” Mr. Lewis-Beck said.

The ivory tower calculations predicting Democratic control of the House dovetail with what might be called the modeling for the masses at statistician Nate Silver’s FiveThirtyEight.

There, the current odds in favor of Democrats seizing control of the House stood at 85.5 percent Wednesday, with an 80 percent chance the party picks up between 20 and 60 seats.

One thing to watch on election night, according to FiveThirtyEight and others, is the national vote total. Should the Democrats win the popular vote by at least 5.6 percentage points, they also will win the House majority, FiveThirtyEight predicts, and Mr. Bafumi said the “partisan preference” tally is one poll his models also take into account.

As of Thursday, the generic congressional vote had Democrats with a 7.5 percentage-point advantage in the Real Clear Politics polling average.

That generic vote is crucial to the model of Alan I. Abramowitz of Emory University. Should it stretch to +10 for the Democrats, Mr. Abramowitz’s figures show them gaining 34 House seats, and even if that polling figure is a wash between the Democrats and the GOP, the model predicts a pick-up of 16 seats.

While that generic vote figure has fluctuated over the summer and fall, it has always favored Democrats in 2018, although the same was true for Hillary Clinton before the 2016 presidential election and Republicans insist trends have gone more in their favor since Democratic senators tried to destroy Justice Brett M. Kavanaugh’s nomination to the Supreme Court in September.

The calculus is less grim for the GOP in the Senate for a variety of reasons, according to the experts. One is the Democrats are six years removed for an unusually good year, another is that several of them are defending seats in states President Trump carried two years ago.

“We see no change in the Senate,” Mr. Lewis-Beck said. “Or maybe plus one for the Democrats, but they’re not going to take it.”

Mr. Bafumi has a slightly different calculus on the Senate, where he believes it is possible the GOP could marginally improve their thin, two-seat majority. There’s less confidence in Senate predictions largely because modelers have much smaller sample sizes to work with, he said.

“In net, I’d say pretty close to where they are now, although the Republicans might pick up a seat,” Mr. Bafumi said.

Of course, these election Rosetta stones don’t always provide exact translations of the eventual outcome, but they have a respectable track record in terms of general returns.

Mr. Campbell’s “Seats in Trouble” models, for example, didn’t have the Democrats losing 64 seats in their 2010 “shellacking,” as then-President Barack Obama put it, but it did predict a loss of more than 50. Two years later, the Democratic gain of eight seats fell right in the middle of Mr. Campbell’s model estimate of their gain.

Similarly, in the 2014 midterms, when Republicans wound up with their largest congressional majority in a century, Mr. Lewis-Beck’s “Structure X” model was within two of the GOP net gain of 13 seats.