The Washington PostDemocracy Dies in Darkness

The 8 most pivotal Senate seats in 2014

U.S. Sen. Mark Pryor, D-Ark., holds a sign given to him by his father, former senator David Pryor, during a news conference at his Little Rock, Ark., campaign headquarters Friday, Feb. 28, 2014. (AP Photo/Danny Johnston)

This is a guest post by political scientist Ben Highton of the University of California, Davis.

Yesterday we updated our Senate forecast and reported that the chances of a Republican takeover had improved from our initial forecast.  Assessments of the 36 individual elections and our model’s estimates of the probabilities of the parties winning and losing each one of them underlie the prediction.  We can examine the 36 elections to map out the most likely path to a Republican takeover and first did so on Feb. 6.  With this update, although we see a very similar road to Republican control, there are  some interesting amendments and elaborations.

First, a quick refresher.  The Republicans hold 30 seats that are not up for election in 2014.  To gain control of the Senate, they need 51 because Vice President Biden would cast tie-breaking votes in a 50-50 Senate.  Thus the Republicans need to win 21 (of the 36) elections this fall.

Last time we identified 17 elections where the Republicans chances of winning were quite good.  According to our model updates, those same 17 elections still appear to be very good bets for the Republicans.  (I will discuss one of them, Georgia, below, because some analysts see this race as notably more competitive than we do.)

If the Republicans win those 17, then they need four more to take control of the Senate.  Last time we identified the elections in Alaska, Iowa, Louisiana, and Montana as the next most likely possibilities for the Republicans, with the elections in Michigan, Arkansas, and North Carolina as the next three.  This time, the top four elections to put the Republicans over the top appear to be Montana, Louisiana, Iowa, and Michigan, followed by Arkansas, Alaska, and a tie between Colorado and New Hampshire.  Here’s a graph:

Here are state-by-state rundown:

Michigan. In our updated predictions, Michigan enters the top four because our model now takes account of “candidate quality.” While the Democratic nominee will likely be Rep. Gary Peters, the Republicans also have an apparently strong candidate with former secretary of state Terry Lynn Land.  In combination with the seat being open and the state being only modestly more Democratic in presidential voting than the nation overall, our model currently gives the Republicans a 58 percent chance of winning this seat.

Montana and Louisiana. The model gives the Republicans about a 3 in 4 chance of winning each of the Montana and Louisiana elections.  In Montana John Walsh – the appointed Democratic incumbent and former lieutenant governor – will likely face Rep. Steve Daines.  As we noted earlier, appointed senators do not typically receive the electoral benefits of incumbency that accrue to those who have previously been elected to the Senate.  Yet, Walsh certainly qualifies as a “quality” candidate.  But so does Daines, especially because his House constituency covers the whole state since Montana has just a single House member.  In a situation like this, we expect the partisan balance in the state, the midterm penalty, and Obama’s low approval ratings to decide the outcome.  Hence the substantial advantage to Republicans.

The same is true in Louisiana, where Republican House member Bill Cassidy is challenging the Democratic incumbent, Mary Landrieu.  Unlike Walsh, Landrieu is a senator by virtue of having been previously elected to the institution, but our model also takes into account the previous margin of victory, and in Landrieu’s case it’s small.

Iowa. The last election in the top four is Iowa, which our model currently gives the Republicans a 65 percent chance of winning.  Here, if we had to guess, we’d say the model is overly bullish for the Republicans.  The reason is that the Democrats have a good candidate in Rep. Bruce Braley, and the Republicans may not be able to recruit a top-tier candidate.  But, our model does not have this information about what’s going on with the Republicans in Iowa.  It assumes that the Republicans will have a “typical” candidate in terms of quality for an open-seat race, which does not appear very likely right now.

Arkansas. Mark Pryor’s seat in Arkansas looks quite vulnerable, though not as vulnerable as Landrieu’s seat in Louisiana.  The reason for the difference with Landrieu is that no Republican challenged Pryor in his last election, and we have found that an incumbent’s previous electoral margin – even when adjusted for uncontested races – is significantly correlated with current electoral performance.  That said, if the lack of a Republican challenger to Pryor in 2008 was fluke, then the Republican chances are even better than they appear.  But, if the lack of a Republican challenger was related to Pryor’s electoral strength, then there is good reason to believe that just as he benefited from it in 2008 there is a benefit to be gained from it in 2014, even though he is facing a quality challenger in Representative Tom Cotton.

Alaska. Our model gives Democratic incumbent Mark Begich a  49 percent chance of retaining his seat, making the election a true toss-up.  If the Republicans recruit a high quality candidate to challenge him, his reelection odds will go down.  But, if they fail to do so, then our model will predict a narrow victory for him.

Colorado and New Hampshire. The appearance of these two states is due to the fact that Republican candidate recruitment has gone well in those states.  In Colorado, Rep. Cory Gardner has decided to challenge Democratic incumbent Mark Udall.  In New Hampshire, former Massachusetts senator Scott Brown appears very likely to seek the Republican nomination to challenge Democratic incumbent Jeanne Shaheen.  In both elections our model now gives the Republicans about a 4 in 10 chance of winning.

Two other races deserve mention for not being on the current list of eight races that our model currently predicts will determine who controls the Senate.  First, there’s Georgia.  Our model gives the Republicans a 94 percent chance of winning the seat, placing it safely in the set of 17 elections where the Republican chances are very good.  In contrast, Larry Sabato rates the election “leans Republican,” Charlie Cook calls it a “toss-up,” and Nate Silver gives the Republicans a 70 percent chance of winning.  The reason our model views the election as safe for the Republicans is that Georgia is a very Republican state in a midterm election year with an unpopular Democratic president.  This is a basic recipe for a Republican win.  On top of that, the Republicans are likely to nominate a strong candidate, though which one remains to be seen.  The likely Democratic nominee is Michelle Nunn.  She is the daughter of one of Georgia’s most beloved politicians, former senator Sam Nunn. But, Michelle Nunn has no previous experience in elective office, and when our model judges “candidate quality” it is on the basis of previous elective office experience.

The second state that is not on this list is North Carolina.  The model only gives the Republicans a 1 in 4 chance of beating Democratic incumbent Kay Hagan.  In contrast, Sabato, Cook, and Silver all see the race as a toss-up.  In comparison to Hagan’s fellow Democratic incumbents seeking election from former Confederate states (Mary Landrieu in Louisiana and Mark Pryor in Arkansas), her state is closer to a toss-up in presidential elections, and while Landrieu and Pryor will likely face U.S. House members in their elections, a top-tier challenger does not appear likely to emerge in North Carolina.  (The current leading contender appears to be the Speaker of the state House, Thom Tillis.).  We therefore see a significantly greater difference between North Carolina on the one hand and Arkansas and Louisiana on the other.

Moving forward, what would change our model’s predictions?  Significant changes in economic performance and presidential approval could alter them considerably.  So, too, could fundraising and polling data, which our model does not yet include.  If for example, trial heat polls between John Walsh and Steve Daines in Montana show Walsh doing well, then our model will become more optimistic about Walsh’s chances.  But, based on the factors that are already in the model, we would expect future polls to look good for Daines and other Republican candidates in key states.