In contrast to our prognosis for the United States Senate, where the chances of a Republican takeover dropped sharply this week because of the outcome of the Delaware primary, our forecast for the House of Representatives shows relatively little change from last week.
Over the course of 100,000 simulations of our forecast model, Republicans finished with an average 223.4 House seats — down incrementally from 225.3 last week — and Democrats with 211.6 seats. Republican odds of winning the chamber dipped slightly from 67 percent to 63 percent.
As was the case last week, there is some tension between the generic congressional ballot, which remains quite favorable to Republicans, and local indicators like polls and expert forecasts, which are more ambiguous. Although the Democrats received a couple of decent generic ballot numbers this week, like a George Washington University poll that showed the ballot tied among likely voters, there were counter-examples like an Associated Press poll which showed the Republicans ahead by 10 points. Overall, we estimate that Republicans are ahead by 7.5 points on the generic ballot among likely voters, as compared with an advantage of slightly more than eight points last week.
According to one fairly well-regarded formula, a 7.5-point advantage on the generic ballot would translate into a net gain of 55 House seats by Republicans. A gain of this magnitude is quite possible — there is considerable uncertainty in the forecast, and our model assigns about a 30 percent chance to Republicans winning 55 seats or more. However, if you had to settle on a single number, a forecast closer to a 45-seat gain would probably be more prudent.
Why? Because the other indicators that our model evaluates are not quite strong enough for Republicans to support an “over-under line” of a 55-seat gain. For example, we now have at least one local poll in 116 House districts, and on average, the Republicans have a two point advantage in our adjusted polling average in these districts. Moreover, the districts where polls have been conducted are slightly more Republican-leaning than the country as a whole. Thus, the local polling is more in line with the House being about evenly divided, i.e., a Republican gain of around 40 seats, rather than gains somewhere in the mid-50s.
Likewise, the set of four experts whose forecasts we incorporate into our models — these are Cook Political, CQ Politics, the Rothenberg Political Report, and Larry Sabato — are also more in line with a House that would be split about evenly between the two parties (although some, like Sabato, are more aggressive about forecasting Republican gains than others, like CQ).
As time passes, the model will tend to give more weight to local factors like polls and expert forecasts in each district. Particularly in the case of local polls, these indicators tend to become more reliable as the election draws nearer and voters become more familiar with the candidates. Conversely, the model will give less weight to ‘macro’ indicators like the generic ballot. Thus, Republicans will need their apparent advantage in the generic ballot to show up more robustly in the local polls — or in the expert forecasts — in order to maintain or improve their overall position in the House forecast.
Certainly, Republicans have a number of local polls they can point toward that show excellent results for them. For instance, a non-partisan poll released this week showed the Democratic incumbent Chris Carney trailing by 14 points this week in Pennsylvania’s 10th Congressional District, while another showed Gary Peters, the Democratic incumbent in Michigan’s 9th Congressional District, trailing his opponent, Rocky Raczkowski, in a race that was previously thought to favor Mr. Peters.
But another poll released this week in Idaho’s 1st Congressional District had the Democratic incumbent, Walter Minnick, in surprisingly strong position for re-election, with a 30-point advantage over his Republican opponent, Raul Labrador. Meanwhile, polls in Nevada’s 3rd Congressional District, and South Dakota’s At-Large District, showed a rebound by the Democratic incumbents there.
Democrats have also become somewhat more aggressive over the past few weeks in releasing the results of internal polls conducted by their campaigns. This will not necessarily help them in our forecasts, as such polls are subject to heavy adjustments. Nevertheless, it may be a sign of at least modestly improved confidence in some local races, as Democratic campaigns had been conspicuously quiet about disclosing the results of such polls before Labor Day.
It is important to maintain some perspective here: Republicans are poised to make very substantial gains in the House. They are favored to take control the chamber, and have a 40 percent chance of winning a net of at least 50 seats, and about a 20 percent chance of winning at least 60 seats.
Still, as the results of this week’s primaries perhaps suggested, there remains considerable uncertainty — and ambiguity — in the forecast. A 95 percent confidence interval on our projections would encompass everything from a Republican gain of 78 seats to a gain of just 12. Although that interval will narrow some before Election Day, there’s still a lot of campaigning — and poll-watching — left to do.