Princeton Election Consortium

A first draft of electoral history. Since 2004

The wisdom of crowds of crowds

November 4th, 2008, 4:27pm by Sam Wang


Is it time for Meta-Meta-Analysis? A consistent answer emerges!

The Meta-Analysis’s final EV Median is Obama 352 EV. My mode prediction, in which I force all leaner states to a single outcome, is 353 EV. I favor this outcome.

Over at 3BlueDudes.com, the, um, three dudes have a compilation of 45 projections. Not counting my own, the median is Obama 353 EV (68% confidence interval 337-367 EV).

The Great Even-Numbered Satan gives a median of 349 EV and a mode prediction of 353 EV.

DailyKos readers give an average of 352 EV (rounded).

On Bob Vanderbei’s comment thread I asked for your guesses. The median was 364 EV (68% CI 337-390 EV). This was identical to Bob’s estimate, and equal to my “Obama +1%” case.

What to do with all this information? Why, calculate the median of course:

Obama 353 EV, McCain 185 EV.

It’s also the mode.

Tags: 2008 Election

11 Comments so far ↓

  • Ms. Jay Sheckley

    I am in your archives, trying to understand what date and time we’ll stop getting updates in the header.
    :C
    I doubt I’ll be back to this thread…
    Maybe you answered my questions in PEC in the news comments…
    Sign me
    The Lost Kvetchman

  • Sam Wang

    Readers, we cannot rule out the hypothesis that as a group, you are discerning and rational evaluators of polling data. Your median (and my prediction) win!

  • alistair Connor

    I’m picking a popular vote margin of 6 to 7 %, and 375 electoral votes.

    Though my favourite outcome includes all of IN, MO, AZ,GA : tempted to add ND, but 400 is such a nice round number.

  • Gene Glass

    Prof. Wang:
    Thanks for your efforts. A beacon of light in a world of darkness.

  • Frank

    “My mode prediction, in which I force all leaner states to a single outcome, is 353 EV. I favor this outcome.”–Sam Wang

    How did you get 353 in that way? States with a positive median margin total 338. You can’t get the remaining 15 all from the ties: IN (11), MO (11), ND (3). My pick was Of the states with negative margins, the closest is NC, worth 15. This was actually my pick: 353 = Kerry states + IA, NM, NV, CO, OH, VA, FL, NC; with the ties in your data being resolved in McCain’s favor.

  • Glenn

    Already the shenanigans are starting at the polls. Prof. Wang: Have you figured all known voting schemes as variables into your analysis?

  • bjorn akerman

    The outlier explanation I can see. Thanks.

  • Mark

    Thanks for your analysis, Prof. Wang. You have turned the dial up to 11 with the release of your new cheat sheet.

    I am hoping you are wrong about the MN senate race. The strong independent candidate (Barkley) could make this one difficult to predict.

  • bjorn akerman

    Secondly, I can’t stop worrying about these long lines at the voting booths. In 2004 more than 100 000 voters turned back only in Ohio (according to the Guardian) and did not cast their vote because the lines were too long. This year it will only be worse. What’s that gonna do to your medians?

  • Sam Wang

    Rejects outliers, and in the case of the Electoral College is usually a real outcome.

  • bjorn akerman

    Why this faith in the median? What is your motivatition as a statician?
    Pure and simple question from Sweden