Risk Markets And Politics

Friday, November 25, 2005

The No-Trade Theorem: Looking Back to Look Forward

The intuition underlying Milgrom and Stokey's venerable "no-trade" theorem is that the willingness of a trader to make a trade indicates to the market that they have new private information, and thus no-one will be willing to take the other side of their trade. As with many mathematical results, this theorem requires some assumptions which likely do not hold in real-world markets. Nonetheless, it is a valuable concept to keep in mind when envisioning possible future markets.

This preliminary paper joins the work detailing exceptions to the no-trade theorem; it shows that even if the standard assumptions for the theorem are met, trade may still materialize when the final value of the instrument is dependent on the actions its eventual owner may take, as when the controlling interest in a corporation is in play. (These situations can also throw real-world traders for a loop, as evidenced by Kerkorkian's GM maneuver in May; people tend to neglect the use-value of equity.)

Last week also brought an update on the Ladbrokes/New Scientist prediction market which, for two weeks in August 2004, allowed bets to be placed on 5 scientific breakthroughs occurring before 2010. This story suggests another exception to the theorem. Who in their right mind would take the other side of leading researchers' (long) trades on science claims? It seems that Ladbrokes accepted bets on these markets mainly as a function of publicity. (I for one had not heard of this Hilton subsidiary before this story.)

Meanwhile, at Agoraphilia, Tom Bell has been running ahead with Robin Hanson's far-seeing work. He has proposed using prediction markets on research claims as a new form of intellectual property, which he feels is superior to traditional mechanisms such as patents and copyrights with respect to funding, collaboration and hedging. I think this idea is exciting and has potential, but in my comments I broached what I called the "insider problem": "when less-informed participants actually lose money to researchers influencing the resolution of the contract, they are going to be something less than happy, and I'm not sure that's a tenable situation." While this sort of thing happens all the time in commodity futures markets, the impact on price that individual producers and consumers can have in those cases is far less material than one might find in a binary prediction market (prone to sudden and irreversible massive returns), and this seems to be a qualitative difference. Furthermore, in a standing market, if a participant's actions are seen as rapacious, they might encounter friction going forward if they cannot disguise their trades. There is no such recourse/check in a one-time market. Chris Hibbert is, however, sanguine:
Most of the proponents will argue (as I do) that the value of disseminating information outweighs considerations of making a fair playing field. As long as everyone knows that insider trading is encouraged (though malfeasance or dereliction of duty remains actionable) they should be able to anticipate and prepare for the actions of people with better access to information. The idea that the market is a lottery with only ignoramuses allowed to participate is a very bad one.

But will this produce a viable market? I will let the masters of prediction market form, Justin Wolfers and Eric Zitzewitz, answer. From the Spring 2004 "Prediction Markets" paper:
Ambiguous public information may be better in motivating trade than private information, especially if the private information is concentrated, since a cadre of highly informed traders can easily drive out the partly informed, repressing trade to the point that the market barely exists. Indeed, attempts to set up markets on topics where insiders are likely to possess substantial information advantages have typically failed.

"How to attract uninformed traders?" was later listed as the team's #1 open question about prediction markets, when they wrote:
Counterintuitively, the problem for most prediction markets is attracting sufficient uninformed order flow. Markets need uninformed order flow in order to function; with only rational traders trading whose only trading motivation is expected returns, the No Trade Theorem binds and the market unravels. Uninformed order flow can have a variety of motivations (entertainment, overconfidence, hedging), but with the exception of hedging, these motivations are usually non-economic, putting economists at a comparative disadvantage in predicting which markets will succeed.

There is clearly a tension between using markets to aggregate and reveal information, and ensuring that these markets are viable and remain liquid. Designing markets to serve a hedging need goes a long way towards making them sustainable, but to the extent that markets are used for risk-sharing (or entertainment, another sort of "use-value"), prices may be less usefully interpreted.

It is important to always bear in mind the formal aspects of markets as they impact liquidity. Companies like Tradesports might take a cue from the CME, which has the potential to become the US powerhouse of prediction exchanges. Not all markets are best served by the continuous double auction process. Brainchild of pioneer Robert Shiller, the CME/Goldman Sachs Economic Derivatives Markets, for example, operate as sporadic pari-mutuel auctions, thus providing liquidity to traders even if there is no explicit bid-offer match. The markets can operate even if there are no offers whatsoever, since bids on a given strike/indicator range/state are implicit offers on all other states.

11/28/05 Update: David Pennock, who will be presenting for Yahoo! at the Prediction Market Summit, has devised an interesting hybrid of continuous double auctions and pari-mutuel auctions, detailed here. The result combines the guaranteed liquidity of pari-mutuels with the flexibility of being able to exit trades before the final outcome is determined. The latter property allows more information to be revealed during the process of the auction.

11/29/05 Update: Robin Hanson will be out with a new paper addressing some of these topics soon. Meanwhile, here is an existing work of his on the "thin market" (no-trade) problem. He succinctly asks us to:
Consider the case where a single person knows something about an event, and everyone else knows that they know nothing about that event. In this case the standard information markets based on that event simply cannot acquire this person's information.

Dr. Hanson's design uses market scoring rules to ensure possible trading on many combinatoric outcomes, but relies on "patrons" to subsidize this liquidity. To whom would the information be valuable enough to subsidize such a market? For instance, would it be worthwhile for disease-sufferers to pool money to ascertain the probability that their affliction will be cured, or, more likely, pharmaceutical companies, essentially outsourcing early stages of research? This is an interesting and worthwhile line of thinking. Generally, providing trading subsidies (or negative transaction costs) is a much more attractive solution when the providers can be identified by non-monetary or indirect monetary interests. Tax-breaks on such patronage would help as well.

Robin Hanson's original Could Gambling Save Science? paper is, of course, the backdrop of this entire discussion. I note one apparent divergence between Hanson and Tom Bell. In this paper, it seems that Hanson would agree with me on my Cold Fusion example:
The only similar problem in idea futures is when a research lab is trying to keep a result temporarily secret before trading on it, and an employee sneaks out and trades first. This can be dealt with exactly as if it were stock insider trading, through private trading records accessible to criminal investigators.

As I asked Tom Bell, "Where exactly is the line drawn?" What if there is technically no definitive result yet, but the researchers are darn sure that they have cracked the problem? I say it is an "apparent" disagreement between Bell and Hanson because, to Hanson's great credit, the "Gambling" paper was written 15 years ago.

Finally, some (including myself) may have overstated the ability of such markets to provide funding to researchers. After all, they are not paid until the discovery occurs or some progress has been made, by which point they would have garnered success and influence anyway. Individual researchers are unlikely to be in a position to take substantial bets, and to the extent that they do they are incurring more risk for themselves. To the extent that they instead hedge their success, the market yields less information.


  • Possibly I misread the last quote that was included, and it does not indicate a disagreement between Bell and Hanson. I originally took the quote to mean that the keeping of the secret before trading was the problem, but on a subsequent reading, it seems that the "problem" referred to the information being used by someone who does not technically own it. Nonetheless, all of my other points stand.

    By Blogger Jason Ruspini, at 2:38 PM  

Post a Comment

<< Home