by Cass Sunstein

More than at any time in American history, we are now engaged in an excellent experiment on the accuracy of a certain kind of market: political prediction markets. The markets now say that Republicans will keep the Senate and lose the House. Actually they say a lot more. According to the Washington Stock Exchange, George Allen is likely to lose in Virginia; Harold Ford will lose in Tennessee; Bob Casey will win in Pennsylvania; Jim Talent will win in Missouri; Joe Lieberman will win in Connecticut; Robert Menendez will win in New Jersey; Jon Kyl will win in Arizona; Rick Santorum is doomed in Pennsylvania. And there's much more; see thewsx.com for all the details. In the past, prediction markets have performed remarkably well. The Iowa Electronic Markets have called presidential elections with great accuracy--usually outperforming polls. Outside of the United States, market predictions have also done extremely well. As far as I am aware, all of the markets now think that the Senate will stay Republican (around a 70 percent probability) and that the House will go Democratic (also around a 70 percent probability). What is perhaps most interesting is the race-by-race projections offered on the Washington Stock Exchange (and a few other places). Note that some of the predictions mentioned above are tentative. Jim Webb, for example, is trading at $57.44 (surprisingly high, I think), and Claire McCaskill is at $45.02. Both of these margins are small enough to suggest that a Webb loss, or a McCaskill victory, would not be a huge surprise. I'm not putting any money on it, but my market-driven belief is that the Senate will stay Republican and the Democrats will win the House--and that on Tuesday, at least six, and probably at least seven, of the foregoing predictions will turn out to be right (unless the markets change in the next few days). Stay tuned?