If Americans elected an entirely new set of senators every two years — as they elect members of the House of Representatives — this November’s Senate contest would look like a stalemate. President Obama remains unpopular; his approval ratings have ticked down a point or two over the past few months. But the Republican Party remains a poor alternative in the eyes of many voters, which means it may not be able to exploit Obama’s unpopularity as much as it otherwise might.
Generic congressional ballot polls — probably the best indicator of the public’s overall mood toward the parties — suggest a relatively neutral partisan environment. Most of those polls show Democrats with a slight lead, but many of them are conducted among registered voters, meaning they can overstate Democrats’ standing as compared with polls of the people most likely to vote. Republicans usually have a turnout advantage, especially in midterm years, and their voters appear to be more enthusiastic about this November’s elections. Still, the gap is not as wide as it was in 2010.
The problem for Democrats is that this year’s Senate races aren’t being fought in neutral territory. Instead, the Class II senators on the ballot this year come from states that gave Obama an average of just 46 percent of the vote in 2012.1
Democrats hold the majority of Class II seats now, but that’s because they were last contested in 2008, one of the best Democratic years of the past half-century. That year, Democrats won the popular vote for the U.S. House by almost 11 percentage points. Imagine if 2008 had been a neutral partisan environment instead. We can approximate this by applying a uniform swing of 11 percentage points toward Republicans in each Senate race. In that case, Democrats would have lost the races in Alaska, Colorado, Louisiana, Minnesota, New Hampshire, North Carolina and Oregon — and Republicans would already hold a 52-48 majority in the Senate.
It therefore shouldn’t be surprising that we continue to see Republicans as slightly more likely than not to win a net of six seats this November and control of the Senate. A lot of it is simply reversion to the mean.2 This may not be a “wave” election as 2010 was, but Republicans don’t need a wave to take over the Senate.
However, I also want to advance a cautionary note. It’s still early, and we should not rule out the possibility that one party could win most or all of the competitive races.
It can be tempting, if you cover politics for a living, to check your calendar, see that it’s already August, and conclude that if there were a wave election coming we would have seen more signs of it by now. But political time is nonlinear and a lot of waves are late-breaking, especially in midterm years. Most forecasts issued at this point in the cycle would have considerably underestimated Republican gains in the House in 1994 or 2010, for instance, or Democratic gains in the Senate in 2006. (These late shifts don’t always work to the benefit of the minority party; in 2012, the Democrats’ standing in Senate races improved considerably after Labor Day.) A late swing toward Republicans this year could result in their winning as many as 10 or 11 Senate seats. Democrats, alternatively, could limit the damage to as few as one or two races. These remain plausible scenarios — not “Black Swan” cases.
Still, the most likely outcome involves the Republicans winning about the six seats they need to take over the Senate, give or take a couple. What follows are probabilistic estimates of each party winning each race. (These forecasts are not the result of a formal model or statistical algorithm — although they’re based on an assessment of the same major factors that our algorithm uses.)
Summing the probabilities of each race yields an estimate of 51 seats3 for Republicans. That makes them very slight favorites — perhaps somewhere in the neighborhood of 60-40 — to take control of the Senate, but also doesn’t leave them much room for error. This bottom line is not much changed from our forecasts in June or in March (or even the one we issued last July).
The outlook in some races has changed — but most of these changes are minor. At this point in the cycle, I’d be suspicious of a large swing in a forecast in the absence of some precipitating event. Most voters are not paying much attention to the campaigns yet — in 2010, Google search traffic for the term “midterm elections” was only about one-sixth as high in August as in October. Furthermore, many important races lack high-quality polling. Apparent shifts may reflect pollster “house effects” — or statistical noise. After Labor Day, polling changes will be more likely to reflect true changes in voter sentiment.
Races where Republican chances have improved
There are shifts working to the GOP’s benefit in five states.4 The largest change is in Montana, where we’d previously given the Democratic Sen. John Walsh, who was appointed to replace the retiring Sen. Max Baucus earlier this year, just a 15 percent chance of retaining his seat. Now we have his chances even lower, at 5 percent. The main reason is the revelation that Walsh plagiarized passages of his master’s thesis at the United States Army War College. Walsh was already well behind his Republican opponent, Rep. Steve Daines, in the polls and was going to need almost everything to break right to come back in the race. This was like an NFL team throwing a pick-six when it was already down 17-3.
Two other changes are the result of Republican primaries. In Mississippi, incumbent Sen. Thad Cochran won his runoff against Chris McDaniel, partly with the help of African-American voters. Most of those African-Americans will wind up voting for the Democrat, Travis Childers, in November. But Mississippi’s white voters represent the majority of its turnout and are overwhelmingly Republican; there aren’t many swing voters in the state and it would take a truly awful Republican nominee to put the party at much risk. Cochran is comfortably ahead in polls since the runoff.
We also have Republican chances slightly improved — to 75 percent from 70 percent — in Georgia, where the party has nominated David Perdue, a former CEO. This is Perdue’s first campaign, and ordinarily there’s reason to be suspicious of candidates who haven’t previously held elected office; our research shows they tend to underperform their early polling. However, this is also the first time the Democratic nominee, Michelle Nunn, has run for office. Furthermore, Perdue is running as a pragmatic, “Main Street” conservative. The bigger risk to Republicans would have been nominating an extremely conservative candidate who might have lost votes in the Atlanta suburbs.
Perdue has also pulled slightly ahead of Nunn in the polls. It’s a dubious bunch of surveys, full of partisan polls and “robopolls.” In the absence of high-quality polling, one should default toward placing more weight on the “fundamentals” of the race. In our view, those don’t favor Democrats in a midterm year. President Obama wasn’t that far from winning Georgia in 2008, but he came close because of votes from African-Americans and college students — groups that don’t turn out as reliably in the midterms.
In Arkansas, we have have Republican Rep. Tom Cotton with a 60 percent chance of defeating the Democratic incumbent, Sen. Mark Pryor — up from 55 percent in June. A few other statistical forecasts have the race shifting more in Cotton’s direction, while “subjective” forecasts like that by the Cook Political Report still have the race as a true toss-up.
The reason we’re hedging is because of the low quality of the polling. A number of recent polls have shown Cotton ahead — but almost all of them use non-traditional methodologies, are partisan polls, or both. The last “gold standard” poll in the state came way back in May, from Marist College, and had Pryor with an 11-point lead.
Pryor almost certainly isn’t that far ahead. His campaign’s internal poll, released last week, put him six points in front. But the polls a campaign releases to the media are usually badly biased in favor of its candidate; on average, campaigns exaggerate their candidate’s standing by about six percentage points. That would imply that Pryor’s poll sees the race as a toss-up once “unskewed.”
If the polls of the race are confusing, shouldn’t we turn to the fundamentals? The problem is that they’re also hard to pin down. Arkansas has become an extremely red state, but Pryor is a moderate Democrat while Cotton is a conservative Republican who is sometimes associated with the Tea Party. Pryor won re-election without drawing a Republican challenger in 2008 — but his former Democratic colleague in the Senate, Blanche Lincoln, lost by 21 percentage points in 2010. Arkansas is a race that calls for some caution until we get some better polling.
Iowa is another tricky case. There, we have Republican Joni Ernst with a 45 percent chance (up from 40 percent in June) of defeating Democratic Rep. Bruce Braley. The polling in Iowa has been more consistent than in Arkansas and has the race virtually tied.
Our model will view the fundamentals of the race as slightly favoring Braley. The candidate-quality measures it evaluates all come out in his favor: He rates as being slightly closer to the center of the electorate than Ernst, he’s been elected to a higher office, and he’s raised considerably more money. Iowa is normally as purple as purple states get — the sort of state where candidate quality can make a difference.
But Braley lost ground in the polls after referring to Iowa Republican Sen. Chuck Grassley as a “farmer from Iowa who never went to law school.” And President Obama’s approval ratings have been conspicuously low in Iowa. It’s hard to say why — we haven’t observed a similar pattern in demographically similar states like Minnesota and Wisconsin — and it may be a statistical fluke.5
There are some very tricky races this year and perhaps we’ll see more disagreement between forecasts than we did in 2010 or 2012, depending on what factors they emphasize.
Races where Democratic chances have improved
There are also a couple of changes that favor Democrats. One is in New Hampshire, where we have Sen. Jeanne Shaheen’s chances of holding her seat at 90 percent, up from 80 percent. In Minnesota, meanwhile, we have Sen. Al Franken as a 95 percent favorite, up from 90 percent before.
In theory, both Franken and Shaheen are somewhat vulnerable. New Hampshire is an extraordinarily “swingy” state and Shaheen’s opponent, the former Massachusetts Sen. Scott Brown, is well credentialed. Franken won his election by the narrowest possible margin in 2008 and has a very liberal voting record in Minnesota, which is not as much of a blue state as it once was.
But both also have reasonably good approval ratings. And they’ve held fairly consistent leads in head-to-head polls — by about 12 points in Franken’s case and eight points in Shaheen’s. Furthermore, both have often been at or above 50 percent of the vote in polls of likely voters — incumbents who achieve that distinction very rarely lose.
Minnesota and New Hampshire may be cases that speak to the differences between the 2010 and 2014 political environments. In 2010, a Republican wave year, the Democratic incumbent Sen. Russ Feingold lost re-election despite his passable approval ratings. This year, Shaheen and Franken seem pretty safe.
The other change favoring Democrats is very minor: In Kansas, we give them a 5 percent chance of winning, up from 1 percent in June. Those slim chances probably depend upon Milton Wolf, an extremely conservative candidate, upsetting the incumbent Sen. Pat Roberts in the Republican primary this week. (Roberts is favored in the primary, but polling in congressional primaries can be erratic.) More likely, however, Republicans’ larger concern in Kansas will be Gov. Sam Brownback, whose approval ratings are underwater and who is in a close race against Democrat Paul Davis.
There are also a couple of races where other forecasters have shown a change in favor of Democrats but where we don’t think there’s enough information to merit a shift and have the races at 50-50 instead. In North Carolina, the majority of recent polling has shown Democratic Sen. Kay Hagan slightly ahead of Republican Thom Tillis. But the polls are of dubious quality — and some are of registered voters, a big deal in North Carolina, where midterm-year turnout looks a lot different from turnout in presidential years. Some polls also give a fair amount of the vote to the Libertarian candidate, Sean Haugh — with most of those votes presumably taken from Tillis — but third-party candidates often see their polling fade down the stretch.
Meanwhile, in Alaska — which has a track record of inaccurate polling — some models now perceive a slight advantage for the incumbent, Democratic Sen. Mark Begich. We think the polling is too thin and too inconsistent to warrant that prediction, particularly given that the GOP has not yet held its Aug. 19 primary.