Presidential prediction 2012 (Election Eve)

November 6th, 2012, 1:47am by Sam Wang

If state polls perform as well as they did in 2004 and 2008, most aggregators should get within +/-15 electoral votes and 48/51 races correct. The question is how to squeeze a bit more out of the data.

The topline listed just below the title of this website is not our prediction, but the automatically-generated snapshot. It will fluctuate as the last few polls trickle in. The last 3 updates are at 8:00am, 10:00am, and noon. Then we freeze it.

I make two electoral predictions.

ELECTORAL PREDICTION (mode): Barack Obama 332 EV, Mitt Romney 206 EV. The mode is the single most frequent value on the histogram. It corresponds to the map below, and has a chance of being exactly correct.

ELECTORAL PREDICTION (median): Obama 309 EV, Romney 229 EV, Popular Vote Meta-Margin Obama +2.34%. This is the automatically-generated snapshot for November 5th 8:00pm. This prediction is almost guaranteed to be off, since 309 EV is not a common combination. (Note: I’ll edit this with the last few updates.)

ALL-STATE PREDICTION (binary outcomes):


Most outcomes arose clearly from the median of the last week of polling. The exception was Florida, for which I used polls with at least half of their respondents on Nov. 1 or after. This gave a median margin of Obama +0.5 +/- 0.8% (n=10), for a win probability of 72%. I will be unsurprised for it to go the other way (outcome Obama 303 EV, Romney 235 EV). In Florida, a recount is triggered by a margin of 0.5% or less (recount rules, Brennan Center for Justice). There is about a 50-50 chance that we will see that happen.

The next-closest states are Colorado, Virginia, and North Carolina, with margins of less than 2.0%. Excluding these three states and Florida still leaves Obama 281 EV, Romney 191 EV.

In addition to all-or-none outcomes, later today I will provide final polling medians in graphical form in the Geek’s Guide. I hope it will be useful as you follow the returns.