Jul 12, 2005
Testing Markets as Knowledge Aggregators
I am not an economist. As I've said before, I am a fan of economics, and as such, I've got a simple question:
Experiments like these would probably provide a lower bound to knowledge aggregation in markets, however, as inarticulate distributed information--"knowledge" in Hayek's knowledge/information dichotomy--would potentially come from sources that even its beneficiaries could not describe.
It strikes me--again, speaking as a fan--that there are at least three distinct types of knowledge that markets are now thought able to detect.
1) Insider trading. When a small number of market actors possesses all of the knowledge relevant to making a good decision, they will consistently act according to what they know. Given sufficient time, markets will notice this and follow suit.
It would be easy to design a series of experiments around this premise that would gauge under what conditions markets tend to recognize insider trading. It strikes me that when, for example, Tradesports offers futures on the identity of the next pope, they can only be expecting inside information to show up.
One other possibility is that the knowledge aggregated by the market is just a restatement of conventional wisdom, which often happens to be correct even about arcane processes like a conclave. A dummy market--say, in one of fifteen names drawn at random--would eliminate conventional wisdom. A small group of insiders would know in advance the name of the winner, be it Joseph Ratzinger or Bede the Venerable, but they would only be allowed to communicate their knowledge through transactions, not through any other venues. Would the market figure things out? Has anyone even tried to find the answer?
2) Articulate distributed information. Picture a market in very large nonprime numbers. If each participant in the market knew only one factor in a prime factorization of the winning number, would the market be able to identify the correct number unaided? The question is easily testable, and again, there would be no conventional wisdom if the market participants were unable to communicate with each other beyond their market transactions alone.
It seems that the outcome of a U.S. Presidential election is analogous to this second case. In an electronic market, each American participant will have some small amount of knowledge about the outcome of the election: Each one knows exactly how he intends to vote. Again, though, conventional wisdom gets in the way, and conventional wisdom is very often accurate.
3) Inarticulate distributed information. This is the"knowledge" that interested Hayek the most. Frustratingly, it's almost by definition impossible to test. Unlike 1) and 2), 3) would seem to depend on conventional wisdom. Arguably it is conventional wisdom. How could we potentially design an experiment around it?
Please, enlighten me.
[Crossposted at Positive Liberty.]
Has anyone ever done empirical, controlled, double-blind, and otherwise scientific tests to examine how, how well, and why markets are able to aggregate knowledge?It would seem that by setting up dummy futures markets, in the style of IEM and Tradesports, albeit with outcomes that were set in advance but unknown to all participants, one might be able to determine the types of knowledge and the degree of knowledge aggregation that markets are able to handle, at least in the abstract.
Experiments like these would probably provide a lower bound to knowledge aggregation in markets, however, as inarticulate distributed information--"knowledge" in Hayek's knowledge/information dichotomy--would potentially come from sources that even its beneficiaries could not describe.
It strikes me--again, speaking as a fan--that there are at least three distinct types of knowledge that markets are now thought able to detect.
1) Insider trading. When a small number of market actors possesses all of the knowledge relevant to making a good decision, they will consistently act according to what they know. Given sufficient time, markets will notice this and follow suit.
It would be easy to design a series of experiments around this premise that would gauge under what conditions markets tend to recognize insider trading. It strikes me that when, for example, Tradesports offers futures on the identity of the next pope, they can only be expecting inside information to show up.
One other possibility is that the knowledge aggregated by the market is just a restatement of conventional wisdom, which often happens to be correct even about arcane processes like a conclave. A dummy market--say, in one of fifteen names drawn at random--would eliminate conventional wisdom. A small group of insiders would know in advance the name of the winner, be it Joseph Ratzinger or Bede the Venerable, but they would only be allowed to communicate their knowledge through transactions, not through any other venues. Would the market figure things out? Has anyone even tried to find the answer?
2) Articulate distributed information. Picture a market in very large nonprime numbers. If each participant in the market knew only one factor in a prime factorization of the winning number, would the market be able to identify the correct number unaided? The question is easily testable, and again, there would be no conventional wisdom if the market participants were unable to communicate with each other beyond their market transactions alone.
It seems that the outcome of a U.S. Presidential election is analogous to this second case. In an electronic market, each American participant will have some small amount of knowledge about the outcome of the election: Each one knows exactly how he intends to vote. Again, though, conventional wisdom gets in the way, and conventional wisdom is very often accurate.
3) Inarticulate distributed information. This is the"knowledge" that interested Hayek the most. Frustratingly, it's almost by definition impossible to test. Unlike 1) and 2), 3) would seem to depend on conventional wisdom. Arguably it is conventional wisdom. How could we potentially design an experiment around it?
Please, enlighten me.
[Crossposted at Positive Liberty.]