Labor Day marks the time to get more serious about predicting the outcome of the November national elections. Political scientists — including this one — have recently returned from the American Political Science Association Annual Meeting. In even-numbered years, there is invariably one or more sessions on projecting the outcome based on past trends and mid-to-late summer poll results.
In an off-year election, the most critical question is which party will have a majority in the U.S. House of Representatives and the U.S. Senate. Senate contests are more difficult to project on a national scale because only one-third are on the ballot and each has its own dynamic. By contrast, all 435 House seats are potentially in play although redistricting aimed at protecting incumbents has sharply reduced the number that are truly competitive. As of now, Democrats hold 255 seats, Republicans have 178, and two are vacant.
In the prevailing prediction models for House elections, two variables dominate: the generic ballot poll results and which party is not in the White House. The generic ballot question is: “If the election for Congress were being held today, which party’s candidate would you vote for in your district: the Democratic Party’s candidate or the Republican Party’s candidate?”
As measured by Gallup’s tracking poll, from March to early June, the two parties were essentially tied. That was followed in June and July with fluctuating results, first a five point lead by the Republicans, then a similar advantage for the Democrats. Since early August, the GOP has pulled ahead, initially by six points and then, in the last week of August, by ten points. With a Democratic president, the Republicans also have a plus on that factor in the prediction model.
Since the value of the presidential occupancy variable is fixed and will remain the same over the next two months, the only value that can change is the generic ballot result. According to the leading off-year model developed by Joseph Bafumi (Dartmouth College), Robert Erikson (Columbia University), and Christopher Wlezien (Temple University), a six-point generic ballot advantage translates into the Republicans gaining about fifty seats, giving them roughly a 230-205 margin. If the GOP stays at its late August ten-point lead, the model predicts a net of 60 seats going from blue to red. A similar model developed by Alan Abramowitz (Emory University) forecasts about the same results: a 49-seat net gain for the Republicans.
Other political scientists regard using polling data as a form of social science cheating since, in effect, the survey question is the equivalent of having the election before November. The purer approach, they argue, is to use variables other than vote preference. The most prominent example of this approach is a model developed by Michael Lewis-Beck (University of Iowa) and Charles Tien (Hunter College). Using primarily two factors—how optimistic/pessimistic voters are about the economy (the more pessimism, the better the out-party will do) and job approval ratings of the current administration (the lower the ratings, the better for the party not occupying the White House), they project a Republican increase of 22 seats, not enough for the GOP to achieve control of the House.
In addition to the top-down models devised by political scientists, there is a cottage industry for estimating Congressional election outcomes from the bottom-up, district by district. For many years, the Cook Political Report and the Rothenberg Political Report have provided ongoing analysis of each of the 435 House districts. Although Cook and Rothenberg keep one eye on the generic ballot figures, for the one hundred or so seats that conceivably could go either way, they also examine candidate quality, campaign fundraising, and within-district issues.
Where are Cook and Rothenberg as of Labor Day? Cook’s web site indicates his “current outlook is for a Republican net gain of at least 35 seats” while Rothenberg’s site forecasts “GOP gains of 28-33 seats seem likely though considerably larger gains in excess of 40 seats are quite possible.”
Terry Jones is professor of political science at UM- St. Louis.