Subsidized Markets: Introduction
One reason that prediction markets sponsored by TradeSports do not necessarily reflect the predictions that could be made by the most sophisticated models is that TradeSports charges commissions. These commissions, we have seen, are small, but they compound the inherent risk of investing in prediction markets. The best strategy for prediction market investing could go awry in the short term, just as the best card counters in poker might lose money on a bad day or in a bad year. Thus, someone who has created a model and is confident that this model can beat the market by two or three percentage points on average might well decide not to invest in the TradeSports markets. Enough sports fans and gamblers exist to provide ample trading at many of the TradeSports markets anyway, but this is less likely to be the case for markets on esoteric topics aimed at policy analysts. Many policy analysts seem likely to invest only if they can expect to make money. For example, development of an improved account of how clouds work might improve knowledge about global warming, but it could still be risky to invest money in a global warming prediction market. The first step toward increasing incentives for prediction market participation would be to eliminate commissions. Of course, a for-profit institution such as TradeSports seems unlikely to take this step, though an alternative profit model would be to earn money by collecting interest on invested money.6 The nonprofit Iowa Electronic Markets does not charge commissions, and other nonprofit entities might follow suit. Even TradeSports has eliminated the per-transaction commissions charged to a market participant who places an order on the bid and ask queue that another trader later accepts. Presumably, this is because TradeSports recognizes that subsidizing the placement of these orders increases market liquidity and will make others more willing to consider trading. On the other hand, if it turned out that sophisticated players were more likely to place such limit orders, the subsidy of such participation might make rational third parties no more willing to play. So it is not clear how well the TradeSports strategy in general might translate to other contexts. Ultimately, reducing commissions might not be sufficient to encourage participation. Ideally, prediction markets should be subsidized. Policy analysis institutions might provide such subsidies if the government does not. Admittedly, subsidizing prediction markets would leave the policy analysis institutions with less control over the output than would directly subsidizing position papers and scholarship. But sometimes such an institution might genuinely be interested in a particular question and thus might be willing to fund a prediction market in the area. At other times, it might be confident that the market prediction will support its position regarding a particular public policy issue. Suppose, for example, that some politicians have called for limits on oil drilling because they want to make sure that oil will remain available in the distant future. It might then be in the economic interest of the oil industry trade association to persuade the public that the amount of oil produced twenty years from now will be relatively high, but observers will discount the claims and studies of oil industry executives. If the trade association truly believes its position, it might fund a prediction market to show that its position in fact is widely shared. The challenge is to devise a means of subsidizing the market that rewards participants who are acting in a way that enhances market accuracy. The danger is that various techniques might be used to obtain a portion of the subsidy at low risk. For example, a sponsor of a prediction market might promise to distribute a fixed amount of money–say, ten thousand dollars–in proportion to the amount individuals win in the market. A problem with this approach is that it might encourage individuals to enter wash transactions. For example, suppose the bid-ask spread is twenty-eight cents to thirty cents. I might simultaneously enter a large volume of buy and sell orders at 29, using separate accounts. Placing aside the subsidy, and in the absence of transactions fees, there would be no economic upside or downside to the transaction, but the subsidy would mean that one of the accounts is likely to make money. Of course, the result is not only that the subsidy might be distributed to parties who have provided no informational benefit but also that many trades would reflect no information and could distort the market. Distributing a subsidy in proportion to winnings is therefore too simplistic an approach. An alternative approach might be to use some type of market maker. In securities markets, a market maker is a trading firm, often called a specialist. A market maker maintains an inventory of securities of a particular type, posting offers to buy and sell the securities. The bid-ask spread provides the market maker with an opportunity to profit by buying a share at a low price and simultaneously selling a share at a high price. A market maker, however, can lose money on some transactions owing to unexpected changes in prices. In some exchanges, one or more market makers will commit to maintaining a bid-ask spread of no more than a certain size, thus promising to provide liquidity. These market makers receive access to the market as part of the bargain. In many exchanges, traders can execute transactions only with the market maker, rather than with one another, as in a continuous double auction. Independent market makers generally seek to maximize their own profit. In a laboratory experiment, Jan Krahnen and Martin Weber have shown that when there is a single market maker, it keeps a wide bid-ask spread and earns some monopolistic rents, which come at the expense of profits for informed traders.7 When there are competing market makers, informed traders can earn considerably greater profits, and the losses of uninformed traders are reduced. This is especially true if the market makers are just as informed as market participants, but often that is not the case in prediction markets. A continuous double auction design can improve the ability of informed traders to profit on their information by allowing them to enter into transactions with uninformed traders. This helps explain why the prediction markets we have considered so far rely on the continuous double auction without market makers. Uninformed traders may have little interest in some prediction markets, however, especially if they are not generally suitable as investment vehicles. To make it possible for informed traders to profit, the sponsor of the market might wish to add a market maker that is willing to lose money to the continuous double auction, thus subsidizing the market. One approach is for the sponsor of a prediction market to develop an automated market maker, which uses a computer algorithm to offer tradable contracts to buy and sell and provide liquidity to the market. The Hollywood Stock Exchange (HSE) uses such an automated market, and it has received a A disadvantage of the HSE approach is that it may be difficult for the sponsor of the prediction market to determine the extent of the subsidy that it is providing the market by adding liquidity. The subsidy might end up being higher or lower than intended, particularly when trading volume cannot easily be anticipated. The problem is particularly severe in prediction markets involving a very small number of traders, because slight differences in trading volume and in the dispersion of information might have large effects on the subsidy provided. The subsidy, for example, might turn out to be relatively low if all traders have approximately the same estimates and relatively high if a single trader has unexpected information that allows that trader to push the tradable contract price a great distance in the correct direction. In Chapter 4 we will consider the market scoring rule, which can function as an automated market maker that limits the maximum potential loss associated with the market, but even with this rule, the subsidy is not fixed. If the market sponsor wishes to offer a fixed, predictable subsidy, it might agree to pay the full amount of the subsidy to an independent firm that agrees to serve as the market maker. It could choose the firm by holding an auction, agreeing to provide the subsidy to the firm that commits to maintaining at all times the shortest bid-ask spread. That firm might in turn use an automated market maker or human specialists, but it would bear the risk. Placing this high level of risk on a single third party might mean, however, that the best offer the exchange receives will not be as generous as it might prefer.
Leave a Reply