« The Strange Case of One Laptop per Child | Main | Today's Availability Cascade »

January 08, 2008


Feed You can follow this conversation by subscribing to the comment feed for this post.


How impressive are the markets really?




Yglesias doesn't provide particularly good evidence that the markets don't work, but it is at least suggestive that he sees them as lagging indicators. Meanwhile, they can't be all that accurate or we wouldn't see the arbitrage opportunities noted by Bainbridge.

Todd Henderson

First, the experience with prediction markets generally is extremely positive. The successes are well documented in numerous areas. They consistently outperform all other forecasting methods.

Second, it may be that some new, niche markets do not generate sufficient trading at first to be very informative, but this says nothing about the theory or their potential.

Third, many firms are using these conditional prediction markets with great success.

Fourth, no markets are perfect. These are just another tool that helps us forecast. So the relevant comparison should not be "are these markets perfect?" but "how do they do compared with other ways of estimating X?".

Finally, these markets are a work in progress, and there is much learning and experimenting to do. The potential is clearly there for tremendous gains from these markets. Insofar as government or litigation overhang is retarding their deployment, we should remove it.


Right, I'm sympathetic with the overall point. It's just odd that political prediction markets are so small and flawed (the markets establish widely varying "prices" for different outcomes, implying a wider margin of error than many polls). Perhaps the problem is that we have a bunch of fragmented markets, many of which have sharp limits on the amount you can bet and/or high transaction costs.

Steven Levitt argues that prediction markets should not operate under a legal cloud, but he also argues that they shouldn't be distinguished from gambling and that all gambling should be legal:


It strikes me that there's a big difference between playing the slots and betting on whether 3M will hit its revenue targets next quarter, but I guess it might be hard to prevent a lot of pure gambling from taking place on prediction markets.

Todd Henderson

"It might be hard to prevent a lot of pure gambling from taking place on prediction markets."

Yes. But this proves too much. How much "pure gambling" goes on in the stock market? Currency markets? Commodity markets? These are legal not because they do not involve gambling but because, on average, skill predominates over chance and (mostly) because they serve a very important social function. Here is what I wrote about this in a section of my paper with Abramowicz that was left on the cutting room floor:

As for gaming laws in general, many if not all prediction markets would seem to fall outside the reach of both federal and state bans on gambling. Although the rules vary from state to state, “gambling” consists of three elements: prize, chance, and consideration. While prediction markets will clearly involve a “prize” of some kind, and perhaps some form of “consideration”, it is doubtful that the “chance” element will be satisfied in well-designed prediction markets.

First, let’s tackle the consideration issue. In the gambling context, “consideration” is typically interpreted as requiring individual participants to put up money, as when a gambler places a monetary bet on the outcome of a horse race. Clever designers can avoid this problem by providing prizes that are not tied to the participants placing a wager of their own funds. Bingo games and corporate-sponsored cash giveaways are two examples of non-gambling events that look a lot like gambling. Accordingly, the Google design whereby participants are competing for points redeemable in a lottery of prizes would likely not satisfy the “consideration” element.

This issue gets slightly more complicated if the market operator provides traders with cash subsidies to participate in a real-money market. This would be akin to the racetrack giving each participant a $10 chit that could be used to place a bet on a horse race. It is hard to imagine that this would not be considered gambling, even in cases where an individual did nothing more than wager their $10 chit. The question then becomes one of defining the contours of value given in consideration by the trader in the market. For example, if the participant is forsaking salary to participate in the market, then this looks like a contribution of money that would satisfy the gambling consideration element, whereas tying performance in firm-operated prediction markets to an annual bonus scheme looks more like routine firm decisions about how to judge individual employment performance. These are not easy lines to draw. Thankfully, the third element of the test will likely be dispositive in these close cases.

To qualify as a game of chance, elements of chance must predominate over elements of skill. While every non-gambling market—from art to stocks to appliances—has some degree of chance, and every gambling market—from horse racing to poker—has some degree of skill, it is obvious that participation in a prediction market, which is designed precisely to extract information from trades, looks more like the former than the latter, which is purely profit seeking. The most detailed assessment of the legality of prediction markets to date concludes that prediction markets of the kind we imagine here “should not be construed as gambling.”

The analogy to stock markets is apt here as well. Many participants in stock markets are making “bets” on the direction a stock will take in a way that is closely related to a bet on a horse or the White Sox, but equity markets are not “gambling” because skill predominates over chance for most price-setting participants. Another reason why equity markets, and thus prediction markets, are not gambling is because of the public purpose served by the market. In other words, even if chance plays a big role for some participants in equity or prediction markets, the social usefulness of the activity—be it raising capital and allocating scarce resources through the price mechanism for equity markets, or helping reduce agency and information costs for firms for prediction markets—countervails any chance element. While the externalities of gambling—addiction, crime, penury—are highly negative, capital markets and prediction markets have (primarily) positive externalities. Moreover, the chance-based trades of uninformed participants are an essential element of creating market liquidity, which in turn attracts the skill-based (or information-based) trades of informed participants. Thus stock markets (and prediction markets) could be technically considered gambling, but are not because for the market as a whole, skill predominates over chance (in the long run) and the benefits of these markets clearly outweigh any conceivable negatives.

Kimball Corson

Speaking of predictions for markets, namely those of the US economy, Martin Feldstein, a top economist at Harvard and President of the National Bureau of Economic Research, says the odds now favor a US recession.

It can only be avoided, he argues, by quick and aggressive fiscal and monetary policies, but that is not likely to happen with the Fed and this Government (read Bush, who has just conceded that the economy does have some problems along with its great strengths).

What is the chance, therefore, that Bush will react quickly and decisively? We should not hold our collective breath on that one, I suggest. But maybe, just maybe, better heads can prevail on him. We'll see.

Kimball Corson

The DOW has just broken through its 12750 floor. Sell short. Sell short and run with Chicken Little.

The comments to this entry are closed.