Peter Levin’s Rethinking Markets

August 30, 2007

When Information Markets Should Work

Filed under: Daily — Peter @ 9:58 am

Information markets are all the rage. You can dip your feet in the waters with this article by James Surowiecki, whose Wisdom of Crowds has catapulted the idea into the public imagination. Robert Hanson has done lots of work here as well, and his fingerprints are on the FutureMAP project. Crookedtimber has had a number of posts over the last several years about them as well. In thinking through the relationship between information markets and other speculative markets, I have some thoughts on them too.

1) Information markets are rarely the same as other futures markets. The basis for a futures market - and we’ll take them on their own terms here - is that there is a future, calculable risk that one group doesn’t want and that another group is willing to take in return for profits. Farmers worry about future prices, banks worry about future interest rates, Mattel worries about the future costs of plastic (and Euros). They hedge these risks, laying them off to speculators, who are willing to take on some risks in return for profit. Given enough speculators and hedgers (ie a liquid market), both risk and profit are re-allocated so as to shift the risk to someone who will take it and profit to someone willing to take risk for it.

In information markets, there is no risk, there are no hedgers. This is not like other futures markets. It is an aggregation of opinions, and profits accrue to those who have the best information, ability to analyze future events, or who simply are best at guessing future informational states of the world. This is powerful, per the wisdom of crowds, but it shares a name with markets without being quite the same thing.

2) Information markets should work best when they elicit adverse information. That is, as someone like Hayek would argue, markets work because they induce participants to give up private information in return for profit. Making this information public, in the form of market prices, makes the market more accurately reflect the totality of information in the world. So in some cases, you want insiders trading, since they have the most valuable information. (I’m going to leave aside market design here, since there’s the issue of whether others will continue to participate if they are being effectively suckered by insiders with better information).

3) Information markets should be outstanding at disseminating informational certainty as soon as it is possible to do so. If an event becomes inevitable (or close to it), we should see information markets rush very quickly to profit from that. Here, it’s a speed game. As soon as ‘real’ news comes out, it should flow into market prices like a hot knife through butter.

4) Information markets are only as good as their designated contract outcomes. They don’t tell about the world per se, they tell about the particular contract. They lack discretion beyond their design.

5) Prediction markets should be as bad at predicting non-straight-line futures as anyone else. There is no theoretical reason why information markets should be any better than anyone else at predicting events that are not obvious or that don’t flow from the current state of the world.

6) Particularly in winner-take-all information markets, we have no good way of knowing whether they are right or not - that is, they are a blunt instrument. Let’s say there is a 60% chance that X will happen in six months, according the pricing in an information market. And let’s say it happens. Does that mean there really was a 60% chance of it happening since months earlier? There may have been a 1% chance - or a 100% chance. This is simply unknown, and more interestingly, unmeasurable.

7) Information markets will indeed replace independent judgement, and we will be worse off for it. We know not nearly enough about expertise, discretion, judgement, and the ways that these things intersect with information.

2 Responses to “When Information Markets Should Work”

1. Alex Soojung-Kim Pang Says:
   August 31st, 2007 at 12:35 am

   Interesting points. Futurists, not surprisingly, have started to be interested in prediction/information markets, but these experiments are still in their infancy, and we still have a long way to go before our understanding of their theoretical and practical strengths and limitations equals that of, say, scenarios. Certainly many kinds of futures are either difficult to reduce to yes/no bets, or happen so far out in the future you can’t get useful feedback from the results. Figuring out where they do work well– and where they work better than other forecasting tools– is the next thing we’ve got to figure out.

2. Peter Says:
   August 31st, 2007 at 11:14 am

   Yes, I’ve spoken with Andrew about this for a while. The problem as I see it is, interesting, almost a subset of the data-mining stuff. Information markets seem to work under lots of circumstances, and there’s good stuff on design - he pointed me to HP’s research on small-group forecasting in imperfect information markets.

   But there’s an almost surreal sense of trial-and-error in these design studies. Maybe the market metaphor is getting in the way here - I can not for the life of me figure out why some designs should necessarily be better. For example in Wisdom of Crowds we learn that, yes, independence of ideas and diversity help, but really why? I mean, theoretically, why? Does this mean that market prices in corn are more accurate when there is independence of ideas and diversity? Is this something about markets, or about information-inducing betting schema?

   It reminds me of the FCC auction stuff, when the game theorists met the experimental market design people - and the latter group mopped them up (this is Nik-Khah and Mirowski’s paper, in Do Economists Make Markets now, I think. Just tweak, test, tweak again. That this bears passing but not thorough relationship to other kinds of markets is a decreasing problem. As long as the market metaphor continues, it works.

   Ahh, cynicism.

Leave a Reply