Open Questions About Prediction Markets
The legal status question actually contains Wolfers' and Zitzewitz's "How can markets attract uniformed traders?" to some extent because two of the answers there are hedging and entertainment (gambling). However, hedging markets invite the regulation of the CFTC, and gambling markets of course are subject to general official prohibition in the US. There seem to be at least three schools of thought developing here. One group invites CFTC regulation in varying forms as the most expedient, if awkward and potentially limiting, solution. Others argue that information markets are conceptually distinct from markets designed for risk-sharing, capitalization or entertainment, and that they deserve their own legal status. Still others argue for full-scale legalization/regulation of gambling. Anyone interested in these topics should read Tom W Bell if they haven't done so already. Additionally, as noted here before, the gambling/hedging distinction isn't very strong in theory or practice. Some are even implicitly questioning the CFTC's service of the public interest as it relates to the risks faced by sports-related industries.
What will be the first major corporate PM success story? The recent Marginal Revolution discussion summarized the issues pretty well. The consensus seems to be that it is just a matter of time, but there are worries. Even in cases where managers are open to being second-guessed and corporate cultures are sufficiently free, maybe the benefits of prediction markets won't be material over the existing implicit prediction/reward structures in firms. Perhaps firms with relatively "flat" management hierarchies already sufficiently approximate the "wisdom of the crowd". Securities regulation, time-diversion, bad incentives and accounting for the effects of hedging on prices are further obstacles. Also, even if traders are anonymous, questions relating to project targets and the like focus open scrutiny on specific managers and divisions, potentially creating a poisonous environment.
As Chris Masse suggested, perhaps external corporate PMs will prove to be more useful than strictly internal ventures. This is plausible so long as the outsiders actually have relevant information. It is a known problem that prediction markets tend to fare poorly in projecting outcomes determined by a relatively small and closed group of decision-makers. Enumerating such problematic cases is another open issue.
Related to the corporate PM success question, one wonders how often will information markets actually be directly subsidized? In order to justify setting prediction/information markets aside as a separate class of market, the information they reveal should be valuable enough that it attracts patrons willing to pay for it. Yes, there are examples of this such as the avian flu market subsidized by the Gerson Lehrman Group, but how common will private subsidization actually be going forward? If US laws are in fact holding back a tidal wave of growth in this area, why aren't there more pure information markets (as opposed to gambling and risk-sharing markets) popping-up "in the wild" where they are legal?
Now, government subsidization might raise the kinds of issues seen with PAM, and this indirectly leads to the next question. How "fit" are binary options as opposed to tradable indices? In the aforementioned Marginal Revolution thread, Emile Servan-Shrieber, CEO of NewsFutures, made a slightly heretical remark on the use of binary markets:
But event probabilities are just not actionable. People in a business environment don't know what to do with them. They need specific numbers. Not "What is the probability that I'll hit my deadline or sales target?" but "What will my sales be?" and "When will I be able to ship this?", or "What will my sales be if I include this feature or that one?".In many cases an index market will be more practical, especially if it still implies probabilities. In addition to being more actionable, index markets will be less volatile, more conducive to regulation and less susceptible to controversy. What if it is not possible to create a market in binary form, as in the case of Bin-Laden's capture? Ah, well Bin-Laden is a specific individual, and contracts attached to specific individuals or properties seem to be considered insurance contracts by the CFTC. Insurance is regulated at the State level in the US. The CFTC is very unlikely to authorize a future that refers to a specific individual or property, even if all other criteria (e.g. risk-sharing, price discovery, public interest) are met. Such index contracts that refer to some aggregate measure or count are also less likely to invite controversy. What would have become of PAM if it had only priced indices and made no mention of contracts attached to specific (negative) events?
Other questions like contractability and separating causation from correlation are subtle and interesting, but they are also issues outside of the strict prediction market sphere. They are more general problems relating to contracts and market analysis and don't seem to be very threatening in terms of the immediate future of prediction markets.
Even the "calibration on small probabilities" question that largely captured Charles Manski's objections to the "prediction market" label doesn't seem to worry very many people in the industry. One almost never hears about the potentially very severe favorite-longshot bias caused by the fact that most traders aren't paid interest on their posted margin. By "very severe", we mean that a contract tied to an impossible event that expires in one year should only trade down to r/(1+r) where r is the risk free rate, so the "Pigs Fly" 1yr 100 point contract should be worth around 4.5! Play-money contests with rank-order prizes may be vulnerable to similar distortions.
Finally, the manipulation question is largely reducible to attracting liquidity. The argument could even be made that the problem solves itself. Again though, aside from the ways of attracting liquidity mentioned in relation to the first question, manipulation isn't really an issue particular to prediction markets. A variation on Robin Hanson's experiments on manipulation in which traders took the market price as one of their "clues" would however be interesting. Such scenarios, which would exhibit the kind of momentum/trend-following/herding/feedback trading that one often finds in more established markets, might give more pessimistic results with respect to the potential success of manipulation.
Like Wolfers' and Zitzewitz's list, this wasn't meant to be exhaustive. Please add anything that might be missing, or comment on what might have been overplayed.
4/12/06 Update: In terms of "subsidization" above, this specifically meant paying traders for the information they are revealing, although paying to upkeep a market is also a kind of subsidy.
In the case of HSX, third parties pay the owners/maintainers of the market, who do not pass this on to the traders. This is an interesting model that points to PMs replacing traditional test-marketing groups, a trend flagged here before. Of course, the trading itself has to be entertaining in order to attract uncompensated participants. It also doesn't hurt that 1) HSX has been around since 1996, which enforces the "just wait" answer to the PM success questions, and 2) One suspects that HSX is largely a poll and, in any case, traders don't have to devote time to uncovering information, apart from watching movies.
Likewise, one wonders what kind of creative thinking is going on at WSX, especially in light of the Data Warehouse story.