The forecasting model, which political scientist Eric McGhee and I developed, is centered on what we might call the “fundamentals” of House elections. It is based on elections from 1952-2012. It takes advantage of key indicators at the national level: the popularity of the president as of June of the election year, growth or decline in the economy (gross domestic product) in the first two quarters of the election year, and whether it is a presidential or midterm election year. (Similar factors went into a presidential forecasting model that I helped develop for Wonkblog in 2012.) The model also builds in key indicators at the district level: the partisanship of the district as measured by the presidential vote, and whether a Democratic, Republican incumbent, or no incumbent is running in the district. Thus, the model provides a forecast not just of how many seats each party will have in the House as a whole, but of the outcome in each district. We first deployed this model in 2012, and it proved quite accurate.
To generate a forecast for 2014, we assumed three things: that Obama’s approval in June 2014 will be what it was in November 2013 (42 percent), that the economy will grow at the same rate in the first two quarters of 2014 as in the second and third quarters of 2013 (0.7 percent, non-annualized), and that this Roll Call “casualty list” tells us which incumbent will and won’t be running in 2014. Obviously, all of these things can change, and if they do, so will the model’s baseline prediction. Obama’s approval could be lower in 2014, for example. Certainly it is on that trajectory. The number of open seats will also grow. So we treat this only as a preliminary forecast.
At this moment, the model predicts that Democrats will win approximately 48 percent of the national popular vote for the House. The model also predicts that Democrats will win 196 seats, for a loss of 5 seats.
That may seem like a small loss for the Democrats — perhaps “too small,” given the rocky rollout of Obamacare and the loss that the president’s party typically sustains in midterm elections. But there aren’t that many Democratic seats for the taking, thanks to the Republicans’ huge gains in 2010. There are currently 24 seats held by Democrats where Obama received less vote share than his national average in 2012. In 2010, there were 69 such seats.
As we were in 2012, we will be transparent about potential shortcomings of the model. First, no forecast is certain, and we can quantify the uncertainty in our forecast. If we were to construct something akin to a poll’s margin of error, it is large: plus or minus about 20 seats. That’s a lot! We plan to reduce this uncertainty throughout 2014 by incorporating candidate fundraising data and polling data from both the generic ballot and from polls in individual House districts. Second, this baseline model is explicitly designed to see how far only “the fundamentals” can take us. We think that they can take us pretty far, but clearly other factors matter — such as the qualities of individual candidates. This is a potential reason that the model tends to underestimate the magnitude of partisan waves when they happen, and another reason we will build in fundraising and polling data.
Third, it’s still early. The forecast could change, and in fact our model will be explicitly designed to change as new information comes in, such as from polls. Indeed, it would be odd for the forecast not to change if conditions in the country or in particular districts changed. However, our model also has the benefit of not overreacting to momentary blips, as would have happened if we had relied only on the generic ballot throughout these last two months, when the Democrat went from near-parity to a large lead and back to near-parity.
But for the moment, I think the model’s forecast nicely accords with some other prognostications based on different methodologies, which are very valuable in their own right. For example, handicappers like Charlie Cook, the Rothenberg Political Report and Larry Sabato’s Crystal Ball see relatively few toss-up races, which are balanced between Democratic and Republican seats. Similarly, political scientist Michael McDonald has argued that the most likely outcome in 2014 is no shift in control of the House. Political scientist Alan Abramowitz’s forecasting model also suggests only a small seat shift given the rough parity in the generic ballot today.
As Lynn Vavreck and I wrote two months ago, the metaphor of “cross-currents” seems more appropriate in 2014 than does “a wave.” This is likely to leave the House under a Republican majority not much smaller or larger than it is now.