Back in December I wrote:
“[…] the proposed “Policy Analysis Market” (which claims on its website that it’s going to launch in March; sadly there is no currently existing futures market which allows me to bet that it won’t)
Historical note; it didn’t.
BTW, for those who care about that sort of thing, while we’ve expressed plenty of scepticism over the marginal value of election betting numbers in the past, they are probably no worse than polling numbers and available with greater frequency. Bush is currently more or less holding steady on IEM, but weakening on Tradesports. Note that these two figures are not directly comparable, as the IEM contract is for vote share while the Tradesports one is “winner take all”.
{ 18 comments }
justin 04.19.04 at 1:09 pm
Since the funding was pulled in June of last year, its no wonder you won your bet.
But to the point of the PAM, it’s a shame that you won your bet. Futures markets, even with play money, have shown incredible insight into trends. Examples (with a nod to Eli Lehrer):
While it may be offensive to our sensibilities… it has proven very effective in the past.
des 04.19.04 at 1:17 pm
Is there anywhere online I can bet actual cash money on Osama bin Laden being produced, dead or alive, in the month of October?
(Of course, I also need persons willing to stake their dosh that he won’t, but there seems to be little that is beyond the Merkin right.)
justin 04.19.04 at 1:29 pm
Des, Which of the following points best describes your comment: “there seems to be little that is beyond the Merkin right”:
Your comments go to prove my point… the market (or public) assessment of policies and future events can be incredibly productive. The research on PAM showed that press reports, analysis reports, and public political statements were almost half as accurate as a free-market motif.
dsquared 04.19.04 at 1:51 pm
Justin: We’ve been round the houses on CT on the subject of betting markets. I have to say that I regard the claims in the Hollywood Stock Exchange marketing material (including the marketing material in the form of academic papers) to be heavily overstated. I also dispute a number of claims about the IEM; there were also a number of occasions on which it forecast a Gore landslide. I’ve not got time to get into them right now, but if you click on my “past posts” list in the left hand menu bar, you’ll see a number of exchanges between myself and James Surowiecki.
Btw, they were claiming in December that they had new funding (and hence would launch in March). I don’t recall them ever having published information on the reliability of their methods against expert opinion.
Des: Tradesports.com has a number of Osama bin Laden contracts. I’ve put a speculative bet on the Dec ’04 contract myself, because traffic on a conspiracy theory mailing list got me thinking that maybe the Americans captured him ages ago and have got him stored up for a big pre-election news event.
des 04.19.04 at 2:01 pm
Dsquared: Cheers; now all I need is a browser that works with it…
theCoach 04.19.04 at 2:09 pm
At what point to the market-based solutions begin to make fairly solid predictions about events? Is there a time period when they really begin to have predictive value?
dsquared 04.19.04 at 2:16 pm
Well, they’ve got some predictive value right now. About the same as polls right now. Markets aren’t actively evil ways of finding out about the world, they’re just not magic bullets.
des 04.19.04 at 2:16 pm
Who needs smelly old predictions when you can have a cash-weighted averaging Bayesian likelihood estimator?
I know I don’t!
Dan Goodman 04.19.04 at 4:49 pm
Justin is wrong about the Foresight Exchange being dead. And still a place where you can bet that Osami Bin Laden will be elected President,
if you happen to have really, really
hot secret information. (Or if you don’t mind wasting play money on a joke.)
jam 04.19.04 at 5:24 pm
IEM and its friends don’t predict the future. But they do tell you a lot about the present: roughly speaking, if sufficiently actively traded, an IEM market reduces conventional wisdom to a number. To the extent that conventional wisdom is likely to be right about an outcome, an IEM market is likely to “predict” the correct outcome.
Which means they’re more useful than polls, which merely feed conventional wisdom.
Bill Carone 04.19.04 at 5:36 pm
Dsquared,
“Well, they’ve got some predictive value right now. About the same as polls right now.”
The former is about bets on the question “Who do you think will win?” and the latter is about answers to the question “Who will you vote for?”. These are two different kinds of question (and it interests me that they give about the same results).
What about if we polled people and asked “Who do you think will win?” rather than have them bet on it? Would you say that the market or this poll would do better, and in what ways?
des 04.19.04 at 5:43 pm
Weighting by cash is a non-trivial improvement on standard polling, since it allows persons to qualify the strength of their opinion.
Doesn’t stop the results being massively skewed by utter stupidity of course – you only have to look at English bookmakers odds for an English sports team to win something, which are always so short as to be worth shorting as a result of patriotic sentiment about Our Boys.
James Surowiecki 04.19.04 at 6:00 pm
There has been some new work done on the performance of the Hollywood Stock Exchange by professors at Harvard and Penn, although it’s still unpublished. Their conclusion is that the HSX’s price the night, while not fantastically accurate (median error is 16%, I thik), is the single best forecast of a movie’s opening-weekend performance available. That’s roughly the claim — that they’re better than any other available method — that I think can be made for most collective forecasts, whether they’re made via markets, sophisticated voting mechanisms, or a pari-mutuel system.
James Surowiecki 04.19.04 at 6:05 pm
There has been some new work done on the performance of the Hollywood Stock Exchange by professors at Harvard and Penn, although it’s still unpublished. Their conclusion is that the HSX’s price the night, while not fantastically accurate (median error is 16%, I think), is the single best forecast of a movie’s opening-weekend performance available. That’s roughly the claim — that they’re better than any other available method — that I think can be made for most collective forecasts, whether they’re made via markets, sophisticated voting mechanisms, or a pari-mutuel system.
James Surowiecki 04.19.04 at 6:06 pm
There has been some new work done on the performance of the Hollywood Stock Exchange by professors at Harvard and Penn, although it’s still unpublished. Their conclusion is that the HSX’s price the night, while not fantastically accurate (median error is 16%, I think), is the single best forecast of a movie’s opening-weekend performance available. That’s roughly the claim — that they’re better than any other available method — that I think can be made for most collective forecasts, whether they’re made via markets, sophisticated voting mechanisms, or a pari-mutuel system.
dsquared 04.19.04 at 6:36 pm
What about if we polled people and asked “Who do you think will win?†rather than have them bet on it?
The UK used to do this, IIRC and it didn’t make much difference.
James: If I’ve got you right, I already don’t like this work before seeing it. Median error is a silly loss function for measuring the performance of solutions to this kind of forecasting problem. Big errors are much more costly than small ones, so the tails matter much moer than a median error criterion would pick up. Unless that’s a mistype for “mean-squared error”, I would suspect that someone had cherry-picked a criterion by which HSX did well.
Jeremy Osner 04.19.04 at 8:56 pm
The comparison of these markets to polls is an interesting one — finacial markets have no polls to compete against. What if there were a tradition in place of polling randomly selected groups of people on a particular company’s likelihood of making a profit? Would such a poll have about the same odds of making accurate predictions, as the stock market has?
Jam’s point about encapsulating conventional wisdom vs. feeding conventional wisdom falls short a bit when you consider that electoral markets results can be used the same way poll results are, to feed cw.
justin 04.19.04 at 9:48 pm
Here’s a link to PAM’s old website and the overview they provide:
click here
Comments on this entry are closed.