If you bought one share of each and put in limit sells for 0.67 and 0.47 and they both went through sometime before the election, you’d get 14%. And if one of them went through you could chicken out pretty easy, selling obama for 0.53 or McCain for .33. It looks like there’s a $5 entry fee but no vigorish for making trades.

I see a problem with volume, though. If you put in a limit order for 1000 shares and the price reaches you but there aren’t enough buyers to take the whole 1000, they cancel your limit order and you can’t sell until you make a new one. So it looks like it might be easy to make pennies or nickels on the market, but harder to make enough to feed your family at Burger King.

I don’t have any experience with this sort of thing, do I have the right idea or does it work some other way?

I suspect that Matt’s theory would be orthodox, but the IEM has never really seemed to work on this expectations basis in the past (viz, the big moves in primary races). I guess people must have been selling into the Hillary speech and are now covering.

(Here’s some pundosphere stuff, reflecting the institutional masochism that’s common among Democrats.)

The bet is for who will get the larger popular vote, not for who will win the election. One of the biases here is that the enthusiasm for risking 60 cents to win 40 may be much less than that to risk 40 to win 60.

Also, there may not be a market-maker here who tries to maximise the number of trades. About 18000 shares sold on the republican side since August 1. If every one of them sold to the same man, that would cost him around $8000. If he was selling his shares for what he could get and then buying them back higher, it could cost considerably less. So if somebody wants to manipulate the price for political purposes they could do so easily. What’s harder is to manipulate the price and make a profit doing it.

For whatever it’s worth, here’s the graph for the 2004 WTA IEM market. That year, the Democratic Convention ran from July 26-29 and the Republican Convention from August 30 to September 2 (9/11!).

As of August 1, there seemed to be no convention bounce for Kerry, or a small Democratic convention bounce for Bush. This meant that Kerry was significantly underperforming historical expectations. However, by August 3, a modest, if still historically small, bounce had appeared.

In September, Bush’s bounce, was also small (around 2 points), though as far as I can tell nobody argued that Kerry had gotten a bounce out of Bush’s convention.

The IEM graph linked above seems to match the way these bounces compared to expectations. Kerry’s IEM numbers declining pretty steadily during the last week of July (when the convention occurred and no bounce emerged. Starting on August 3, as Kerry’s small bounce appeared, his numbers took a notable turn for the better. Bush’s numbers, on the other hand, seem to be pick up pretty rapidly right at the time of the Republican Convention (when he again took a statistically significant lead in the polls for the first time since Kerry’s bounce). Perhaps because Kerry had already had lower-than-historically-expected bounce results, Bush seems not to have been punished for his lower-than-historically-expected bounce. The gallup article that I link to above on Bush’s RNC bounce puts forward the general notion that, given the intensive coverage of modern campaigns before the convention, bounces might simply be smaller as a general rule than they were in the past. Perhaps the IEM had already assimilated this (potential) fact.

Of course an additional factor in the mix is whether those trading shares in IEM are playing the market (i.e. trying to buy low and sell high) or predicting the winner. I’m obviously not an economist, but my guess is that their behavior might be different depending on what they’re trying to do.

Good point. But I suspect people who play this particular market really are playing to pick the winner. It is a somewhat irrational goal for this kind of market, I agree.

As for Obama’s actual chances in November, I’m still going with fivethirtyeight.com’s Nate Silver’s predictions, which seem based on the most sober, thorough, and objective assessment of the polling data. According to Silver’s algorithm, Obama’s chances have certainly decreased (from a high of around 70% to a low of around 50% sometime last week). But today they are back up to 58%. And that sounds about right to me given where the campaign is at the moment.

Bounce-schmounce, so what? He started 5-6 points ahead, and after a relatively steady slide he is now 2 points behind. If that’s not a good reason to cut your losses, I don’t know what is.

Yesterday’s Gallup tracking poll was taken before the first night of the convention. It provides no information whatsoever about a convention bounce. In addition, in past years, the bounce doesn’t appear until the third day of the convention.

Here’s another interesting series of data points. Nate Silver over on the indispensable fivethirtyeight.com calculated that the average convention bounce was nine points. The McCain campaign has announced that it “expects” Obama to get a fifteen point bounce. Presumably they are trying to build expectations for their opponent which will then be dashed.

Both Silver’s calculation and McCain’s spin suggest that we might reasonably expect something like a ten-point bounce for Obama.