1. That any Democrat except Dean would give Bush a close run;
2. That Dean would get clobbered; and
3. That Dean has about two chances in three of being the nominee.
Those bettors could be wrong, of course. (I think they’re overestimating the solidity of Dean’s lead in the race for the nomination.) But it’s more likely they’re right. And Dean is clearly the outlier in terms of the vote-share betting.
Several readers found the exposition above too elliptical, so here it is in full:
The Iowa site has two markets on the curent Presidential race. The one linked to above is the vote-share market for the election in November. If you own “Clark” you get a penny for every percentage point of the (two-party) vote Clark gets if he’s the nominee; if he’s not the nominee, you get zero. If you own “Bu/Clark” you get a penny for every percentage point of the vote Bush gets if he’s running against Clark; if Clark isn’t nominated, you get zero.
Since the vote-shares have to sum to one, the sum of the prices of Candidate X and Bu/Candidate X ought to be the probability of X’s nomination. This can be checked, or arbitraged, against a different Iowa futures market, a winner-take-all on the nominee. If you own “Dean” in that market and Dean is the nominee you get a dollar per share, else zero.
In the vote-share market, the sum of Dean and Bu/Dean trades at about 66 cents (taking the midpoints), while in the winner-take-all nomination market Dean trades at 63.5. So the betting is that Dean has almost two chances in three of being the nominee. The Clark vote-share prices sum to about 19 cents, and Clark’s nomination share is worth about 20 cents. (The numbers change all the time, so what you see if you click through may not be exactly what I see as I write this.)
These numbers are pretty close. The thinness of the market makes arbitrage hard, and the $500 upper limit on your initial stake makes it only marginally worthwhile.
If we look at the vote-share numbers, we see that in most cases Bush is a little bit ahead, but within the bid-asked spread: Clark trades at 9.2 bid, 9.7 asked, while Bu/Clark is 9.4 to 9.9, suggesting something like a 1% edge in expected popular vote for Bush over Clark. (About .2% difference divided by about 20% total.) That would be a 50.5%-49.5% election, a little bigger than the margin by which President Gore beat Mr. Bush but still a cliffhanger.
By contrast, the Bush-Dean difference is bigger and outside the spread. Dean trades at 30.1-32.3, while Bu/Dean trades at 36.0-37.8. That suggests that Bush has an expected votes-share advantage of almost 6%/66% = about 9%, which would be an almost Dukakis-sized blowout.
Of course the bettors aren’t prophets, but that’s how they’re betting. Anyone who claims that all of the current Democratic candidates would be expected to do about equally well against Bush, or that Dean would be stronger than the others, needs to reckon with that fact.
[Tradesports has a nomination market in which Dean is trading at 65 cents and “field” (which includes Clark) at 25. It also has a market on Bush to be re-elected; he’s trading at 74.]
Brainster published the same analysis almost a month ago, but I hadn’t seen it.
Second update A reader challenges the analysis above, arguing that the mechanics of the market tend to push the vote-share prices artificially close together for candidates other than the front-runner. That may be a factor; thin markets are hard to interpret.