In the biennial horse race known as congressional elections, the pollsters who pick winners and losers in advance â€” like candidates â€” also jockey for legitimacy.
The results from Tuesdayâ€™s midterm election should come as no surprise to the political junkies who have been following the polls: Republicans made big gains in Washington, overtaking the House and cutting into Democratsâ€™ sizable majority in the Senate.
But as well as most major polls performed, there were still certain races that caused fits for even the most well-respected prognosticators.
Senate Majority Leader Harry Reidâ€™s successful re-election bid against Republican challenger Sharron Angle in Nevada was among the most missed races. Reid, who was cast as the consummate Washington insider and soft on enforcing immigration law by his opponent, pulled out a victory by a much larger margin than expected (50 percent to 45 percent). Rasmussen Reports favored Angle by 4 percentage points shortly before the election, while others, such as Pollster, FiveThirtyEight and YouGov got it wrong, albeit by slimmer margins than Rasmussen.
The Colorado race between sitting U.S. Sen. Michael Bennet, a Democrat, and Republican Ken Buck was also apparently hard to predict. With 98 percent of precincts reporting, Bennet led Buck on Friday 48 percent to 47 percent. Several major news organizations had called the race for Bennet by late Thursday. But like Reid in Nevada, many pollsters â€” including Larry Sabato, FiveThirtyEight and Rasmussen â€” predicted a win for Bennetâ€™s Republican challenger. Among the polls analyzed, only YouGov, which provides data for â€œThe Economist,â€ got it right, predicting a 1 percentage point victory for the Democrat.
Alaska, where Republican Lisa Murkowski was trying to win a Senate seat as a write-in candidate for the first time in more than half a century, also gave some pollsters problems. Although "write-in candidate" holds a sizable lead in the race, it may be weeks or even months before an official winner is determined. FiveThirtyEight predicted that Murkowskiâ€™s Republican opponent, Joe Miller, would win with 38 percent of the vote to Murkowskiâ€™s 32 percent and Democrat Scott McAdamsâ€™ 28 percent. In reality, it was 41 percent for write-in candidates (presumably almost all of those votes were for Murkowski), 34 percent for Miller and 24 percent for McAdams. Rasmussen also favored Miller, while others â€” like Pollster â€” predicted that Murkowski would win over Miller and McAdams, although by a much closer margin than appears to have actually occurred. (Pollster predicted a narrow victory for Murkowski: 33 percent to 32 percent to 28 percent)
Washington and Illinois
Washington and Illinois also were states that at least one poll got wrong. Rasmussen favored Republican Dino Rossi in Washington, where, in actuality, it was Democrat Patty Murray leading by 2.5 percentage points with 80 percent of precincts reporting Friday. YouGov foresaw Illinois Democrat Alexi Giannoulias beating Republican Mark Steven Kirk by almost 3 percentage points, when it ended up being the opposite. Kirk won with 48 percent of the vote to Giannouliasâ€™ 46 percent.
Predictions for the House of Representatives are more difficult to evaluate. For one, there were many more races to track. And with nine nail-biters still hanging in the balance, itâ€™s still too soon to say that one poll outperformed another. That being said, hereâ€™s a look at some of the major pollsâ€™ projections. The general trend â€” without too many notable exceptions â€” was that pollsters underestimated the Republican wave that swept the country Election Day.
FiveThirtyEight projected that, after Tuesday, Democrats would hold 203 of the Houseâ€™s 435 seats to Republicansâ€™ 232. In reality, the gap between parties looks to be much wider. With nine races still up in the air Friday, according to The New York Times,â€ Republicans held 239 seats to Democratsâ€™ 187.
Ipsos, reporting for Reuters, forecast that Republicans would win 231 seats in the House. Again, it is still too early to determine the validity of that forecast.
YouGov put Democratic losses at 63 seats, which would mean 192 seats total for the now-minority party. With nine House races still undecided, Democratic losses stand at 60 seats, making YouGovâ€™s projections still feasible.
Larry Sabato, who is a professor of political science at the University of Virginia, predicted a net gain of 55 seats for Republicans, a figure that already has been eclipsed by actual results.
Kansas and Missouri
As far as local races in Congress, none of the pollsters analyzed picked the wrong candidate in Kansas, where Republicans Kevin Yoder, Mike Pompeo, Lynn Jenkins and Tim Huelskamp easily won all four House seats and Republican Jerry Moran took Kansasâ€™ one available Senate seat.
In Missouri, several pollsters miscalculated the Missouri 4th Congressional District race between Democrat Ike Skelton and Republican Vicky Hartzler. Sabato gave Skelton a slight edge, whereas FiveThirtyEight gave Hartzler a 52 percent chance of winning. Skelton, who was first elected to Congress in 1976, lost the race with 45 percent of the vote to Hartzlerâ€™s 50 percent.
Political wonks wanting a deeper look behind the numbers and at how different polls collect data can check out Nate Silverâ€™s analysis of his fellow pollsters at FiveThirtyEight.
To reach Joel Walsh, call 816-234-4328 or send e-mail to email@example.com.