Wednesday, November 03, 2004

Time to Retire the Fair Model

The die-hards of the liberal press still refuse to call Ohio for Bush, so I'm not quite ready to claim my Nostradamus Award. But I most respectfully suggest that it is time for Yale economist Ray Fair -- whose model of presidential election outcomes I have discussed here -- to abandon his econometric prediction model and go with the betting markets.

Fair issued his final prediction for the 2004 election on October 29. He said that Bush would get 57.70 percent of the two-party popular vote. As it turns out, Bush will probably get something close to 51.5 percent of the two-party popular vote.

My own prediction (issued at 11:23 a.m. CST on November 1) was that Bush would get 51 percent of the two-party popular vote. I based that prediction on the state of play at Iowa Electronic Markets on October 31, where the average price on Bush's two-party popular-vote share was equivalent to a bet of 51.7 percent -- which is about as close as you can get.* I shaved my prediction to 51 percent because Rasmussen's presidential tracking poll (as of November 1) gave Bush 50.7 percent of the two-party popular vote.

It's only fair that Fair concede defeat and heed the wisdom of the markets -- as a good economist would.
* I ignored the trading on Monday, November 1, because the betting markets had by then begun to show signs of last-minute specualtive volatility. The Bush popular-vote-share contract at Iowa Electronic Markets, for example, hit a low of 48 percent on Monday. That was followed by a low of 47.7 percent on Tuesday, and an average price of 48.9 percent. Yesterday's last price (51.9 percent at midnight last night) was close to the mark -- but that's like calling the winner of a horse race when he's near the wire with a three-length lead.