Democrats’ prospects for 2014 do not look rosy. There is little chance that they will retake the House, and a good chance they will lose seats. Even worse, there is a significant chance that they will lose control of the Senate. Our forecasting model said as many as two months ago. That forecast continues to square with the sense of many analysts — even those who mocked the forecast.
Our House and Senate predictions are based on the fundamental factors that affect congressional elections — including presidential approval, the state of the economy, and whether the incumbent is running in each district and state. (See this post for more details.) But as we noted in our earlier posts, it is important to build in additional information when it is available, and especially information about the qualities of individual candidates.
One key piece of information is whether candidates have held an elective office before and, if so, which one. Unsurprisingly, political science research has long shown that candidates who have held elective office and higher levels of office tend to do better on Election Day. They usually run better campaigns and make fewer mistakes, if only because they’ve done it before.
As we have begun to incorporate candidate experience into the model, our initial sense is this: Republicans may have a far better chance of winning control of the Senate than we or other analysts previously thought. Here is a preliminary estimate: The GOP could have as much as a 4 in 5 chance of controlling the chamber.
Better candidates emerge when conditions in the country favor their party. As political scientists Gary Jacobson and Samuel Kernell have argued, strong candidates run when they have a better chance of winning. And in 2014 — as in most midterm election years — the playing field is tilted away from the president’s party. So we should see good Republican candidates emerging.
Anecdotally, that’s what has been happening. In Arkansas, Mark Pryor’s opponent, Tom Cotton, is a House member. The same is true of Sen. Mary Landrieu’s opponent, Bill Cassidy, in Louisiana. And in Colorado and New Hampshire, as The Washington Post’s Chris Cillizza noted.
Here’s how we can turn these anecdotes into data. For each of the known or highly likely general-election Senate candidates in 2014, we put their level of experience in one of five categories. The lowest category is for those who have held no elective office. The highest category is incumbent senators. In between are current or former state legislators, House members, governors and the like. For candidates who have held more than one of these positions, we code only the “highest.” (Thanks here to Phil Stein, a George Washington University undergraduate, for helping gather these data.)
For races in which we don’t yet know who a general-election candidate will be, we assumed that this candidate would have an experience level similar to that of previous Senate candidates in his or her party running under similar circumstances — Republicans seeking open seats, Democrats challenging an incumbent and so on. We’ll update the data after each state’s primary election.
Our forecasting model is thus the same model as before but with one additional factor: the difference in the two opposing candidates’ political experience. As always, the model is essentially a look at what would happen if the election were held today — with current levels of presidential approval (43 percent) and economic growth (1.67 percent in the last two quarters of 2013, non-annualized). Any forecast comes with uncertainty, and our results reflect this. When we report the “percent chance” of some outcome, this is because we’ve run 10,000 hypothetical elections using the model, and the underlying uncertainty means that the forecast is never exactly the same each time.
But the upshot is still very good news for Republicans: As of today, this model suggests that they have more than an 80 percent chance of winning control of the Senate.
It might be helpful to look at the percent chance that the model currently assigns to various scenarios. Below are the percent of the hypothetical elections in which the model predicts that the Democrats will control various numbers of seats.
Democrats need 50 seats to control a majority, given Vice President Biden’s tie-breaking vote. This occurs in only 18 percent of simulations. It is most likely that Democrats will control 46 to 49 seats. Indeed, in nearly one-third of simulations, Democrats control 48 or 49 seats, suggesting that if future events break in their favor — for example, President Obama becomes more popular — their chances of controlling a narrow majority could improve.
This forecast may strike some as too optimistic for Republicans. But it does square with what other analysts are seeing — namely, that the emerging Senate landscape may be more favorable to Republicans than previously thought, a result in part to strong GOP candidates. Two weeks ago, Cillizza wrote:
The Senate playing field has shifted in Republicans’ favor over the last several weeks thanks to recruiting successes in Colorado and New Hampshire, as well as a national political environment that looks increasingly treacherous for Democrats.
Two days later, Jennifer Duffy of the Cook Political Report announced a series of Senate ratings changes that shifted several races in a direction favorable to the GOP.
One response to such a forecast is to point to polls showing close races in places such as Arkansas or Kentucky. I would caution against discounting forecasts in favor of current polls, as Sen. Charles E. Schumer (D-N.Y.) recently did. Early polls tend to have limited predictive value and, in fact, surveys often move in the direction of predictions based on the underlying fundamentals of the election.
At the same time, three caveats are worth emphasizing:
- The model is based only on relatively recent elections: 1980-2012. This probably is a necessity if the model is to include data on candidate fundraising, which are available in digitized form only since 1980. (Although we have obtained Senate fundraising data from 1972-78 — thanks to Phil Stein — so we may be able to incorporate those elections as well.) As we noted previously, basing a forecast on 1980-2012 rather than the full period of our data (1952-2012) produces a forecast that is more favorable to Republicans. This is because in more recent congressional elections the midterm penalty has been greater and presidential approval has been more strongly related to election outcomes. It is entirely possible that 2014 will look more like recent elections than older elections but we can’t be sure.
- Early forecasts can be wrong, too, as Senate Majority Leader Harry Reid (D-Nev.) noted in pushing back against Nate Silver’s recent sense of the Senate landscape. Forecasting presidential elections is usually not that hard but it is difficult to predict the outcome of House and Senate races. Things may change. This is only a forecast based on conditions today.
- Good candidates may prove not so good later on. We have captured only one facet of the candidates thus far. As time goes on, fundraising numbers and polls may reveal different facets, positive or negative, and our model will take that into account. If less electable Republicans emerge from primaries or if some otherwise experienced GOP candidates implode, then perhaps Democratic Senate chances will improve.
In short, this is not a final forecast. It is an estimate based on patterns in previous Senate election years. What the forecast tells us is this: An election year that already was looking good for Republicans looks even better. The Republican candidates who are running have, at least in terms of their experience, probably strengthened the GOP’s chances in November.