I was just scanning through InTrade and noticed the following probabilities listed for the winner of the Iowa Straw Poll taking place in August:
Bachmann - 50% Pawlenty - 24% Paul - 20% <several other candidates> - 17.5% (cumulative) Any Other Candidate - 15%
That makes a total of 126.5%! That means the contracts are collectively overvalued by about 27%. Neglecting market frictions (finite trades, transaction costs), you could sell all the contracts short and make a guaranteed 27% profit in mid-August!
In reality: 1) InTrade's 5% per contract fee erases all profit, 2) the relative low-volume of the market inherently limits position size thereby limiting profit size and 3) there may be a bid/ask spread that isn't reflected in the top-line prices. However, it's interesting to note just how far off opinion markets can be.
At a quick check, other political markets such as Republican VP Nominee and Democratic Presidential Nominee are also significantly overvalued, although the Ames straw poll market is the most over-inflated.
Posted by Samprimary (Member # 8561) on :
Intrade contracts on a straw poll. Yeah, I wouldn't touch that with anything. Though I'm still wondering if there isn't money to be made shorting people's bizarrely irrational ideas that certain candidates could win the general election. You could bet against people like palin, paul, bachmann, and get some cash. I think. Haven't looked yet.
Posted by SenojRetep (Member # 8614) on :
The value of current unlikely Presidential contracts isn't large enough to cover expenses:
Bachmann - 2.7 Paul - 1.9 Palin - 1.8 Cain - 1.3
No money there. InTrade should offer bundled contracts, given their margin is so high.
Posted by fugu13 (Member # 2859) on :
Intrade doesn't have a per contract fee. There's a flat rate monthly fee, and no trading fees at all. The problem is, you won't find anyone to take the other side of the contracts at any volume worth speaking of.
For instance, there are a ton of people looking to short Santorum right now, but few people looking to buy. You could short 50 shares at ten cents a share and 50 at 11 cents, for a potential profit of all of $10.50. The volume is just uninteresting.
There's no one at all willing to buy Sarah Palin for Iowa, even at a big discount, it looks like there's about $20 of shorting available in the Ron Paul market, you get the idea. And to play for such small numbers, you'd have to pay $5/month through then.
Posted by Samprimary (Member # 8561) on :
Yeah it just looks flat-out dry. Doo doo doo. I guess I'll sit and wait for the general elections to have a bit of a clearer picture.
Posted by SenojRetep (Member # 8614) on :
Hunh. I thought they took 5% on each contract. I must have mistaken the $5/month membership fee for a contract fee.
My main interest is the theoretical problem of using low volume markets to give estimates of probability. This seems like a case where it failed pretty badly.
Posted by fugu13 (Member # 2859) on :
For the low probability events, sure, but in politics those probabilities are basically inestimable anyways (too much noise). I think you'll find intrade's track record pretty good on the mid to high probability events, even when the volume isn't huge. Also, for political contests this far in the future, it's sometimes more helpful to think of them as measures of public sentiment weighted by the context (such as the Iowa straw poll) than probabilities. Which is really what they are all the time, it's just that such measures can be interpreted as probabilities due to those who are better at being right being willing to put more money where their mouth is.
If you're interested in more discussion of prediction markets in general, I'd be happy to talk more.
Posted by 0Megabyte (Member # 8624) on :
...this entire concept seems completely idiotic.
By which I mean the buying and selling of futures based on the popularity of people.
Posted by fugu13 (Member # 2859) on :
0Megabyte: that's not what a prediction market is. While in this case people are the focus, ultimately what's always being traded is bets for or against the occurrence of particular, measurable events. And political events are extremely good candidates for that.
People buy or sell them because they think they have a better estimation of the probabilities involved (or think they have found an arbitrage opportunity), and if they turn out to be correct, they'll make money in the long run. This will lead the odds of any event of sufficient interest (such as Bachmann winning the Iowa straw poll) to be estimated quite accurately, even using much less information than polls do (because polls become incorporated into the knowledge of people bidding on the events). Indeed, Intrade for larger political events is about as accurate as the best polls, and often more quickly.
Posted by BlackBlade (Member # 8376) on :
Part of me is mad I didn't invest some money in Obama back in 2008, I had enough hard data then to essentially know he was going to be elected no matter what other folks were saying.
Posted by SenojRetep (Member # 8614) on :
The "2012 Presidential Election Winner (by Party)" is a higher volume market. Each of the three contracts is trading on the order of 100 units per day.
Currently the market is slightly overinflated at: D - 57% R - 43% O - 1.7%
If we wanted to short all three, the current highest ask is D - 57 R - 41 O - 1.5
which is just short of an arbitrage. It seems somehwat likely that at some point, even given the realistic constraints, some small arb opportunities will show up, even in higher volume markets.
Last year, if my data from the time is correct (which, since you've pointed out my error on the InTrade revenue model, I doubt to some degree), I found two contracts trading in separate markets that were logical complements. Even as the election approached and volumes got quite large, they still showed upwards of a 10% arbitrage.
IEM on the other hand almost never has a significant arbitrage. Comparing similar contracts for InTrade and IEM make me wonder if IEM doesn't manipulate their numbers, because the level of incoherence in contracts is so small.