Skip links

Crystal Ball has another strong cycle

Despite a topsy-turvy, hurricane-interrupted election that looked closer a week ago than it actually was on Election Day, the Crystal Ball accurately projected Monday that President Barack Obama would comfortably win a second term in the White House. Our final projection had Obama winning 290 electoral votes. It now appears that he will do better than that, adding Virginia’s 13 electoral votes and probably Florida’s 29 to his column. Still, we’re glad we got 48 of 50 states, and we called some of the swing states long before the campaign’s conclusion. Bellwether Ohio is an example: We moved it to “leans Democratic” in late September and never wavered in that selection, even as the race tightened after the first presidential debate and some experts insisted it had become “too close to call.”

As for the down-ticket races, we correctly projected 31 out of 33 Senate seats and, in all likelihood, 10 of 11 gubernatorial races and roughly 97% of the 435 House seats (some of the races have not been called yet).

Our final projected House gain was Democrats +3 (after having been as high as +7), but the actual Democratic gain might actually tick up to +7 or +8 after an apparently good Democratic night in California, Illinois and the Northeast. Still, Republicans will maintain a solid House majority — it’s likely that Speaker John Boehner (R) will command a caucus bigger than any of the Republican majorities from 1995-2007 (the maximum number of seats Republicans controlled in that period was 232, and Boehner is guaranteed at least 234 seats in 2013, according to the New York Times). We pick all the House races — no toss-ups — and we’re bound to miss some, but we’re pleased with our overall performance as we wait for final results in all races.

The Senate will be 55-45 Democratic, assuming that Sen.-elect Angus King, a Maine independent, caucuses with the Democrats; we thought the Senate would remain 53-47 Democratic. The fact that Republicans lost Senate races in Indiana, Missouri, Montana and North Dakota this year — all states that Mitt Romney won fairly comfortably — is an indictment of the GOP candidates who ran in those states, and a genuine credit to the Democratic victors. Candidates matter.

All in all, we correctly picked about 96% of the combined Electoral College, House, Senate and gubernatorial results this cycle. That’s about equal to our biennial average.

Leading into this election, there was a lot of skepticism — particularly on the right — of the traditional election metrics, such as public polling, in predicting election results. As it turns out, the aggregated polls did quite well: the only state where the final RealClearPolitics average probably erred was in Florida, where polling showed Romney leading by 1.5 points; Obama is currently ahead in the Sunshine State, but it remained too close to call as of this writing. As analysts more interested in data than spin, it came as a relief to us that, for the most part, the data we rely on proved largely accurate. There is no wholesale crisis in the world of polling, and that’s good for everyone’s ability to understand elections.

Over the past two years, we hope we’ve provided Crystal Ball readers with a solid sense of which way the election was going. In the coming weeks and months, we’ll look ahead to the contests of 2013 and 2014 and look back to dissect what happened in 2012. We thank you, our loyal readers, for your continued interest in our predictions and commentary.

The national popular vote

As of Wednesday afternoon, Obama held a 51.2% to 48.8% lead over Romney in the national two-party vote. That tracks pretty well with the projection from Crystal Ball Senior Columnist Alan Abramowitz, whose adjusted “Time for Change” model projected that Obama would get 50.6% of the national two-party popular vote. (Abramowitz wrote about his model here.) While we expect Obama’s national popular vote lead will probably grow as all the votes are counted, we think the Abramowitz model — which uses factors such as presidential approval and GDP growth to project the popular vote — gave our readers a strong indication of the direction of this election over four months ago.

Earlier this fall, we also featured a range of political science election models assembled by University at Buffalo-SUNY professor and friend of the Crystal Ball James Campbell. We reprint them here, below, to see how they performed. As you examine the chart, remember that 51.2% is Obama’s current lead in the national two-party vote. Note that Campbell’s model performed exceptionally well.

Chart 1: Selected 2012 presidential election models

Source: James E. Campbell, “Forecasting the 2012 American National Elections: Editor’s Introduction,” PS: Political Science & Politics, v.45, n.4 (October 2012).


Almost a year ago, a reader — Jim Jolley of Bolingbrook, IL — e-mailed us to offer his Electoral College prediction: 284 for Obama if Romney was his opponent, or over 300 against anyone else. While he didn’t hit the mark, the fact that he was in the ballpark so far away from the election means we’re happy to tip our cap to him.

Today we’ve received reports from many readers that their own Electoral College and popular vote picks were close to, or right on, target. We congratulate you all. Prognostication is for everyone, not just “professionals.” It is a great involvement device for political activism and education — and that is what the Crystal Ball and its parent, the University of Virginia Center for Politics, are all about. When you follow us and talk politics with friends and family members, you promote the mission of civic education. Here’s hoping we all get a perfect score in predictions next time, in the midterm elections coming in 2014.