Glenn Reynolds links to the picture below, and states that "Many, many readers have emailed this chart from the Iowa Electronic Markets showing Bush pulling ahead since the convention. I don't know how much these are worth, either."
Maybe not much, but research over a good many elections and election cycles shows that the IEM and other betting markets provide estimates of the probability of winning an election that are tough to beat. Much better than the polls, at any rate.*
These prices are from the "winner take all market." Specificaly, they are the cost of purchasing a contingent claim yielding $1 should that party win the presidential election. Assuming the market is efficient and that 3rd parties have a zero probability of winning, the prices should (i) sum to $1, and (ii) each price represents each candidate's probability of winning. The latest crossing of the lines implies that Bush has taken a slim lead over Kerry in the race. If one attributes this move to events in Boston last week (and not some other factor), this suggests that participants in the IEM viewed the Democratic convention as disappointing, i.e. worse than expected. It is worth noting that in years past, polls would surge in favor of the party holding the convention, while IEM prices would hold steady. Over time, the temporary boost in the polls would reverse to normal. That both the polls and the IEM reacted negatively to the events in Boston is not good news for Kerry, but not damning either. At current prices, the Iowa market still rates the election as a toss-up.
*Footnote: the IEM, run by economists and others at the University of Iowa, has been a useful laboratory for studying how markets assess the likelihood of future events for many years.