Back to previous page


Do special elections predict general elections?

By Ezra Klein,

Democrats suffered two nasty losses in special elections last night. One in a Nevada House district where John McCain and Barack Obama had split the vote in 2008 and another in a New York district that Obama had won by 10 points. But do these special elections have anything to tell us about the upcoming election in 2012? Yes, say political scientists. The elections might be able to tell us quite a bit. And what they say is not good for the Democrats.

Bob Turner, center, joined by his wife Peggy, right, and family smiles as he delivers his victory speech during an election night party, Wednesday, Sept. 14, 2011 in New York. Turner says his shocking win in a heavily Democratic New York City district is a "loud and clear" message to Washington. (AP Photo/Mary Altaffer)

Political scientists David Smith and Thomas Brunell looked at (pdf) special elections going back to 1900 and found that “those special elections that result in a change in partisan control do have predictive power for the general election.” How much predictive power, and through what mechanism? That’s where it gets complicated.

Smith and Brunell find that only one sort of special election has substantial predictive power: special elections in which the partisan control of a seat flips. And a simple look at the data suggests they have quite a bit of predictive power. “When the Republicans have a net gain in special elections they also tend to win seats in the following general election (66.7 percent of the time). For the Democrats, the relationship is even stronger as that they take seats away from the Republicans in the special elections they follow that up with a seat gain in the general election 82.35 percent of the time.”

But as Nate Silver notes, this relationship can break down quite spectacularly. “In advance of the 2010 elections,” he writes, “Democrats won seven of the nine special elections to the House, with an even score as far as seats changing parties. (Democrats lost Hawaii’s First Congressional District to Republicans, but won New York’s 23rd, both under somewhat unusual circumstances with multicandidate fields.) Then they lost 63 seats in November.”

The other question is whether these elections are demonstrating predictive power or actual power. It’s possible, as Smith and Brunell notes, that special elections exert influence over scheduled elections by boosting the morale of the winners and dimming the hopes of the losers. If that happens, “high quality candidates” on the winning side who were thinking about running for office might jump into the race, while similarly able candidates on the losing side might sit the next election out. There’s not enough evidence to say with any certainty whether that’s happening, but it’s certainly possible.

Bottom line? Special elections do have some relationship to scheduled elections, and they could even help the winners win the next election by affecting recruiting efforts. All in all, last night was bad news for the Democrats.

© The Washington Post Company