Donald Trump’s dwindling chances of winning the presidential election have led to a lot of speculation about whether this will hurt Republicans running for the House and Senate. Last week, the speculation reached a fever pitch.
So what does a “real shot” or “on the table” actually mean? We can put hard numbers on the chances of a House Democratic majority, including in the event of a resounding Hillary Clinton victory.
The model is a “fundamentals” forecast that captures crucial factors at both the national and district level:
- The growth in gross domestic product in the first two quarters of the election year.
- Presidential approval as of June of the election year.
- Whether it is a midterm or presidential election.
- The party of the president.
- Whether there is a Republican incumbent, Democratic incumbent, or no incumbent running.
- The presidential vote in the district in the current election (presidential years) or previous election (midterm years).
- Whether the candidates have held elective office, coded to capture whether either candidate has an advantage (that is, has had elective office while the opponent has not).
- The balance of fundraising between the candidates, based on the most current Federal Election Commission data.
As we’ve noted before, this model has its limitations. It comes with uncertainty, which means the model’s predictions have a margin of error. Moreover, the model doesn’t take account of finer-grained information about the candidates — beyond what their experience in elective office or fundraising numbers can tell us.
The model doesn’t take account of polls, which exist only in a small number of districts (although we did use them in 2014). Polls aren’t magic — sometimes a fundamentals-based prediction is more accurate than polls! — but they are helpful. The model tends to give predictions for individual races that compare well to those of handicappers, but where the two diverge right before Election Day, the handicappers are more often correct because they are accounting for more information.
Nevertheless, modeling House elections this way also has its advantages. It gives us a vote forecast for every district and a specific probability of victory, making it more precise and easier to evaluate after the fact. It tends to be even more accurate in the aggregate than it is for individual races — any misses tend to average out — and it did a good job in 2012 and in 2014.
Above all, combining national and district-level factors together in a single model is very effective at testing the sort of “what if” scenarios that have been popular lately. If you want to know what a big Clinton win means for the House, given the competitive dynamics in individual seats, this is an excellent way to do it.
This model currently predicts that the Democrats will control 202 seats after the 2016 election. That is 14 more than they had after the 2014 election. The margin of error associated with that is plus or minus 8 seats. That forecast implies a very small chance — less than 1 percent — that the Democrats could win the 218 or more seats needed for a majority.
But what about Trump? As Robert Erikson noted, the most recent research finds that presidential candidates do have “coattails” down the ballot. If Hillary Clinton does better, so should Democratic candidates for Congress.
Because our model includes the presidential election result in each district, we can simulate what might happen if Clinton beats Trump in the proverbial landslide. This entails two decisions. First, how big is the landslide? Second, what is the exact size of the coattail effect?
Let’s assume that Clinton’s victory is exactly where the Pollster average is right now (Clinton +8 percent). And let’s assume that there is a larger coattail effect than is typical. Erikson estimates that each point for Clinton equals half a point for a House candidate (this agrees with our baseline model). What if instead that effect is about two-thirds of a point — the level we saw in 2012? And what if the influence of candidate factors such as incumbency and spending were weaker accordingly? It’s the “as Trump goes, so goes the nation” scenario.
Even with these favorable assumptions for Democrats, the model predicts just 210 Democratic seats and a 17 percent chance of reaching 218. To get to 218 and even odds of a Democratic takeover, one needs the strong coattails assumption and a Clinton margin of 12 percent.
Even a 17 percent chance is not zero, of course. Moreover, a 12-point win for Clinton no longer seems as crazy as it would have a few weeks ago. But it’s still a stretch.
We are not the only ones to express some skepticism that Trump’s challenges won’t give Democrats a House majority — see, for example, my Post colleague Aaron Blake and the Cook Political Report’s David Wasserman. Our approach considers more factors and offers more precision, but the conclusion is very similar.
For our baseline model that predicts 202 Democratic seats, you can find every single one of the 435 individual House district forecasts here. Not all of them will be right. But taken together, they suggest that the House is likely to be a residual source of GOP power after 2016, even if Clinton trounces Trump.