First, a bit of background. The Senate has 53 Democrats, 45 Republicans, and two independents who caucus with the Democrats (Bernard Sanders of Vermont and Angus King of Maine), so for the purposes of analyzing party control there are 55 Democrats and 45 Republicans. There are 36 Senate elections in November that involve 21 Democratic seats and 15 Republican seats. (Included in the 36 are three special elections, in Hawaii, Oklahoma and South Carolina.) This leaves 34 Democratic seats and 30 GOP seats that are not up in 2014. To control the Senate, the Democrats need 50 seats and the Republicans need 51 seats (because Vice President Biden can break a tie). So, the critical question is whether the Democrats win 16 (or more) of the 36 elections in 2014. If they do, then they will maintain control of the Senate.
Our forecasting model resembles the House model introduced previously. The Senate forecasting model — developed by political scientists Eric McGhee, Ben Highton and me — is based on all Senate elections between 1952 and 2012. It includes several features of the national political and economic landscape: the president’s approval rating as of June of the election year, the change in gross domestic product in the first two quarters of the election year, and whether it is a midterm or presidential election year. The model also includes three key features of each Senate race: the partisanship of the state (measured by the presidential vote), whether the incumbent is running, and the outcome of the previous race for that Senate seat.
All of these things are related to Senate races in intuitive ways. Candidates from the president’s party tend to do better when the president is popular, but somewhat worse in midterm years. A lot hinges on how strongly Democratic or Republican the state is. And incumbency advantage, although not as strong as in the House, is still a powerful force in senators’ favor.
Based on the 60 years of Senate races from 1952 to 2012, current presidential approval and current economic growth, we estimate that the Democrats have only a 56 percent chance of retaining a Senate majority. That is to say, in 56 percent of the simulations from our model, Democrats won 16 or more seats out of the 36 currently up for election, or enough to give them the majority. This forecast is in line with others, perhaps most notably from Larry Sabato’s Crystal Ball, whose terminology — “coin flip” — corresponds closely to the 56 percent to 44 percent split in our simulations.
We can look at some specific races, too. This model gives Sen. Mark Pryor a 75 percent chance of retaining his seat in Arkansas. This is probably more bullish than the polls and other analysts, such as the Cook Political Report, suggest. The model is even more confident in Sen. Mitch McConnell’s reelection in Kentucky, giving Allison Grimes only a 3 percent chance of winning that race. McConnell, the minority leader, appears to have a narrow lead in the polls. The model is also confident in Sen. Mark R. Warner’s reelection in Virginia, giving him an 86 percent chance. This accords with most analyses of that race. For example, Nathan Gonzales, though noting that Republican Ed Gillespie is a “credible contender,” still sees the race tilted very much in Warner’s favor.
But this forecast might actually be too rosy for Democrats. One question that confronts all forecasters and analysts is whether new trends make it misleading to rely on history — as we’re doing by basing our forecast on 1952-2012. Of course, it’s typically valuable to have more data rather than less. But are Senate elections in the current era similar enough to those 50 or 60 years ago to use those earlier elections in a 2014 forecast?
In some respects, Senate elections in more recent years are different. If we compare elections from 1952-78 to those from 1980-2012, two differences stand out, and neither work to Democrats’ advantage. First, the midterm “penalty” for the president’s party is larger in the more recent period. Second, the relationship between presidential approval and Senate elections is stronger in the more recent period, which could hurt Democrats in 2014 given that President Obama is more unpopular than popular. (See Sean Trende for more on Obama’s approval numbers and what they mean.)
If we produce a midterm forecast solely based on Senate elections from 1980 to 2012, the forecast is much more favorable to Republicans: They now have a 64 percent chance of taking the Senate. And an incumbent such as Pryor is in more trouble. The model based on these recent elections gives him a 60 percent chance of winning — much closer to a pure toss-up race.
As with our House forecast, there are important caveats. First, we don’t know for sure whether 2014 will look so much more like recent Senate elections that leaning on earlier ones for extra information will lead us astray. Thus, the odds of a Republican takeover could really be closer to 44 percent than 64 percent.
Second, just as in House elections, it is harder to predict seats than votes. In fact, predicting seats is more difficult in the Senate than the House for reasons we will explore in a future post. This is why we rely on simulations and estimates of the “percentage chance” that a party will win a seat or a chamber majority, rather than trying to forecast the exact number of seats or which seats each party will control.
Third, the dynamics of the campaign will certainly matter for Senate elections, maybe more so than in the House. That means our Senate predictions will ultimately rely more on polling and other measures of competitiveness such as campaign spending. None of those factors are currently part of the forecast. We will build them in as the campaign unfolds.
Finally, the election is, of course, months away and things may change. Trends in the economy or presidential approval, as well as shifts in spending and the polls, will alter our forecast. What we are reporting today is simply where things stand right now.