Crystal balls

[mmcirvin]

The Iowa Electronic Markets, a toy (but real-money) trading market on the chances of various future events, ran several different kinds of share in the California gubernatorial recall election, including recall yes/no vote shares, Gray Davis in/out probability shares, and popular vote shares. As it turned out, the latter predicted the actual popular vote pretty well by the end of the campaign, better than the hunches of many individual pundits. The track record of these markets at predicting stuff is supposedly pretty good.

It's been fascinating watching their Democratic convention and presidential vote-share markets, though they were somewhat handicapped early on by belated recognition of the Democratic-market-leading Dean and Clark campaigns. (I think Hillary Clinton is way overvalued.) From the point of view of an election prognosticator, though, it's unfortunate that the presidential market is just a popular vote-share market, with no "who wins" shares for sale as in the California election. As we saw in 2000, the one doesn't necessarily predict the other. As it is, the traders currently seem to have Bush pretty close to neck-and-neck with most of the Democratic hopefuls in popular vote share (anticipating the recent narrowing of "electability" results in the polls as the names become more recognizable). But the popular vote usually isn't all that far from fifty-fifty even in landslide elections, and to me, at least, Bush seems to have a significant (though certainly beatable) advantage on the electoral-vote map right now. Maybe they ought to start offering electoral victory shares around convention time to see if the traders think this holds.