We’ll have real votes to talk about soon enough, but here’s a final look at where the polls ended up in the 2014 Senate races.
Republicans are almost certain to pick up seats in South Dakota, Montana, West Virginia, and Arkansas, where they are leading their opponents by seven percentage points or more, according to the Real Clear Politics averages. Going back to 2004 (the first year for which RCP polling averages are available for Senate races), there’s only one race* where a U.S. Senate candidate led by more than 3 points in the final RCP average but lost the election.
There are, however, plenty of examples of candidates with leads of three points or less losing the election. And there are seven races today where candidates are separated by three points or less in the RCP polling averages:
Iowa: Ernst (R) leads Braley (D) by 2.3 points.
Colorado: Cardner (R) leads Udall (D) by 2.5 points.
Alaska: Sullivan (R) leads Begich (D) by 2.4 points.
Georgia: Perdue (R) leads Nunn (D) by 2.8 points.
Kansas: Orman (I) leads Roberts (R) by 0.8 points.
North Carolina: Hagan (D) leads Tillis (R) by 0.7 points.
New Hampshire: Shaheen (D) leads Brown (R) by 0.8 points.
Republicans are also favored to pick up a seat in Louisiana’s likely December runoff, so they have a very good shot at netting the 6 seats total needed to take the majority in the Senate. But a quick look at the last decade of tight races shows there’s a lot of uncertainty about the outcomes of the seven closest races in 2014. They call them “toss-ups” for a reason after all.
In 2004 there were four races where candidates were separated by 3 points or less: The polling averages correctly predicted the winner in two of those states (North Carolina and South Dakota), but not in Florida, where Republican Mel Martinez trailed by 0.5 in the RCP average but won by 1.1. RCP didn’t average the polls in Alaska (presumably due to small number of polls conducted), but just one (partisan) poll showed Republican incumbent Murkowski winning, which she did.
In 2006, the polling averages went four for four in predicting the winner of the closest races (Montana, Virginia, Missouri, and Rhode Island).
In 2008, Minnesota Democrat Al Franken was down 2.8 points per RCP; he won by a handful of votes after a recount. There weren’t any other races that year where the candidates were separated by three points or less in the polling average.
In 2010, Colorado Democratic incumbent Michael Bennet was down 3.0 points; he won by 1.7 points. In Nevada, Democratic incumbent Harry Reid was down 2.7 points; he won by 5.6 points.
Alaska’s three-way Senate race was such a mess that year–incumbent Lisa Murkowski waged a write-in campaign against the Republican and Democratic nominees–and there was so little polling that RCP didn’t try to average the polls. Just one (partisan) poll showed Murkowski with a lead; she won the race by four points. In 2010, the polling averages predicted the outcome of just one race where the candidates were separated by less than three points: Washington state’s Democratic senator Patty Murray led by 0.3 points in RCP; she won by 3.8 points.
In 2012, Montana’s Democratic incumbent Jon Tester was down by 0.4 in RCP; he won by 3.7. The RCP poll leader won in three states–Wisconsin, Virginia, Massachusetts–where the candidates were separated by 3 points or less on Election Day.
*North Dakota in 2012 is the only race in the past decade where a candidate led by more than 3 points in the RCP average but lost the election: Democrat Heidi Heitkamp trailed by 5.6 points in the polling average, but won by 0.9. You’ll notice that very few public polls were conducted in October, and the only poll taken during the final two weeks of the campaign showed Heitkamp down just two points.
We definitely don’t have enough data to draw any firm conclusions about the predictive quality of the Senate polling averages. Polls have tended to be biased against Democrats in recent close elections, but Nate Silver pointed out recently that “polls are just about as likely to be biased against Republicans, in which case the GOP could win more seats than expected.” What the errors of the past decade tell us is that there’s a significant amount of uncertainty still about the 2014 Senate elections, which is why Silver gives Democrats about a 24 percent chance to hold the Senate but also gives Republicans a 24 percent chance to hold 54 or more seats.
