Thursday, November 6, 2008

And the Winner Is...Prediction Markets!

(Felix Salmon) There were three types of prognosticators that tried to forecast the outcome of year's elections:
  1. Economic models - This group largely relies on economic variables like inflation and GDP growth to make its predictions. Examples here include models from Yale economist Ray Fair, Moody's Economy.com, and Macroeconomic Advisers.

  2. Smart Polls - This group tried to control for a whole host of variables that could potentially introduce bias into individual polls (like gender, income, regions voting history, etc.) Big players here are FiveThirtyEight and Stochastic Democracy.

  3. Prediction Markets - If you're reading this blog, I'm assuming you know what these are.

Before we take a look at how these three groups did last night, there are three caveats:

  • I'm ignoring probability distributions, margins of error, confidence intervals, standard errors and all that, and will only highlight the final raw numbers that the sources themselves advertised except in the case of the Iowa Electronic Markets where I pulled the vote-share number from Nov. 3rd's close.

  • As of the writing of this post, 97 percent of the total vote has been counted, so vote-shares may change. Post will be updated to reflect this.

  • I'm giving Missouri to McCain and North Carolina to Obama.

Here's a table that shows each source's prediction and how far off they were. Not everyone had predictions for both vote-share and electoral votes. For space reasons, only Obama is included:


Vote-Share
Obama
Electoral Votes
Obama
Vote-Share
Error
Electoral
Error
Actual
53.1
364
--
--
BetFair
--
364
--
0
FiveThirtyEight
53.1
349
0
15
Intrade
--
364
--
0
I.E.M.
53.5
--

--
Macro. Advisers
54.3
--
1.2
--
Fair Model
51.9
--
1.2
--
Stochastic Democracy
53.8
353
0.7
11

By and large, everybody got the big picture correct. But Intrade and BetFair were right on the nose in the electoral college category while FiveThirtyEight was identically close in vote-share. So the original title of this post will have to change to It's a Tie! Coming in second was I.E.M. followed by Stochastic Democracy while the economic models pulled up the rear. (Which makes sense since economic models are handicapped by having to use older data.)

To reiterate, this isn't scientific and just for fun, so FiveThirtyEight doesn't have to eat its words just yet.

No comments: