Recent Speculations on Prediction Markets

Some eminent economists and political scientists believe that prediction-information-decision markets offer significant benefits to society through efficient extraction and consolidation of the knowledge of individuals. They may also offer some insights into the workings of traditional financial markets that have evolved from trading. They could represent a natural progression from increasingly abstract financial derivatives. The summaries below outline potential benefits and shortcomings of prediction markets. Key points are that prediction markets:

• Could provide a framework for systematic collection, synthesis and valuation of knowledge.
• Must be free with regard to setting prices and paying participants.
• Must be interesting (rewarding) enough to attract a critical mass of participants.

Some open questions are:

• Can governments or other organizations effectively command prediction markets (controlling the supply of questions and the demand for participation)?
• Can prediction markets reliably pull information from experts if the monetary stakes for participation are small compared to the economic implications of associated decisions? Are reputation, altruism and self-actualization understood as payoffs?
• Can a robust prediction market framework coexist with a system already set up to allocate power?

Prediction Markets and Policy Design

Could prediction (information) markets really help guide socioeconomically significant decisions? In their September 2004 paper entitled “Using Information Markets to Improve Public Decision Making”, Robert Hahn and Paul Tetlock suggest how the use of information markets could improve the quality of public policy by providing a framework based on market data to address uncertainty in policy outcomes. Unlike traditional cost-benefit analyses, such markets would make payments contingent on these outcomes to those who provide information and implement new projects. They argue that:

• It is generally possible to construct contracts based on different contingencies whose prices will convey useful information on the costs and benefits of a number of policy choices. For example:
  1. The decision maker specifies and monetizes all verifiable benefits of a possible new project.
  2. To establish a benchmark, the decision maker offers a set of information contracts for forecasts of the benefits without the project. This set of contracts become void (payable based on forecast accuracy) if the decision maker implements (does not implement) the new project.
  3. The decision maker auctions a transferable contract to implement the new project and receive the associated monetized incremental social benefits.
  4. If the highest bid for the new project exceeds some specified minimum, the decision maker awards an implementation contract to that bidder. The winner’s profits would equal the project’s monetized incremental social benefits less the actual cost of implementation.
• Benefits considered may be as complex as desired (for example, non-linear or specifying any statistic of an outcome distribution) so long as the decision maker can quantify (set a monetary value on) them.
• The decision maker can augment this process by offering a second set of information contracts for forecasts of the benefits if the new project is implemented. The decision maker can then estimate the incremental benefits of the new project before auctioning implementation rights (and may decide not to conduct the auction). This second set of contracts become payable based on forecast accuracy (void) if the decision maker implements (does not implement) the new project.
• The decision maker can conduct sensitivity analyses on benefits/costs by offering multiple market contracts for a range of possible new contracts.
• The decision maker can rely on incentives, rather than continual monitoring, to control the costs of new project implementation.
• Complete and partial transferability of new project implementation rights allows the most capable/efficient/interested firm at any point in time (including the decision maker via repurchase in the event of a policy change) to control the work. In addition, a controlling firm could hedge by selling shares.

The paper also surveys risks and mitigations associated with this approach and extends it to a broad range of public regulatory and oversight mechanisms.

In summary, prediction markets may make pay for performance (superior knowledge) viable in policy design and implementation.

Decision Market Design

How could those in power use prediction (information) markets to make socioeconomically significant decisions? In their December 2005 paper entitled “Designing Information Markets for Decision Making”, Paul Tetlock, Robert Hahn and Donald Lien model an information market used for decision making in which experts who have interests in the decision participate voluntarily and contractually, either publicly or anonymously. They then examine implications of that model for market liquidity. They conclude that:

• A prediction market with positions tied to significant decisions can efficiently elicit information from interested experts.
• The sum of the profits made by participating experts is equal in magnitude to the losses of the market maker (decision maker). In other words, the market maker pays the experts for participating (according to the ultimate correctness of their individual inputs).
• A decision maker who can discriminate among experts would logically offer better terms for participation (such as lower trading costs) to those most likely to provide good information.
• A decision maker would logically subsidize aggregate market liquidity (take a position opposite to the average expert position), either as a market maker or a noise trader, to stimulate exposure of valuable information. When there is sufficient manipulative trade (seeking to influence the decision for reasons other than trading profits) to motivate informed trade (exploiting the “irrational” positions of manipulators), the decision maker does not need to subsidize liquidity.
• The mere act of tying a decision to a market price will attract participation of experts interested in the decision, thereby enhancing market liquidity. Liquidity is therefore less of a concern in decision markets than in traditional information/asset markets.

In summary, prediction markets publicly tied to significant decisions may be an effective/efficient substitute for buying advice directly from a small group of experts.

Why Prediction Markets Cover What They Do

Why do prediction (information) markets tend to focus on the frivolous and sensational rather than the meaningful and pragmatic? In their 2006 article entitled “Markets for Markets: Origins and Subjects of Information Markets”, Miriam Cherry and Robert Rogers examine why information markets cover certain subject areas, sometimes of minor importance, while neglecting others of greater significance. They conclude that:

• To date, most information markets cluster in politics and entertainment. More than half of the current operating markets deal with these two subject areas.
• The stories of information market founders offer no clear themes on why information markets have emerged for some subject areas and not others.
• Private for-profit firms run most information markets,with founding entrepreneurs seeming to target subject areas that they believe will attract the most participants. These are also areas that participants find “easy” to understand (gather information) and intrinsically “fun.” This bootstrapping interplay has led to a “random walk” evolution in the areas covered by information markets.
• Neither private associations (for competitive reasons) nor government agencies (due to funding constraints and political risk) are likely to bring order to this random evolution.

The article also includes extensive background on information markets around the world, including results of interviews with some founders and a listing of active markets.

In summary, it is impossible to predict when information markets will start producing predictions with significant socioeconomic value.

Five Obstacles to Effective Political Prediction Markets

What is holding prediction markets back? In their February 2006 paper entitled “Five Open Questions About Prediction Markets”, Justin Wolfers and Eric Zitzewitz identify and discuss five issues to be resolved before prediction markets can perform a serious role in forecasting, decision-making and risk management in both the public and private sectors. They argue that prediction markets must:

1. Find incentives that attract uninformed (noise) traders to keep orders flowing even while rational traders maintain staked-out positions? Even with low transaction costs, abstract policy issues will likely attract little volume and therefore offer poor liquidity.
2. Balance content such that predictions are both interesting enough to attract a reasonably broad set of potential players and specific enough to be contractable. Complex policy issues are often difficult to express as simply stated binary choices.
3. Suppress manipulation when the economic value of a decision that might be derived from prediction market data far exceeds the stakes of the market. Prediction market stakes must be kept low to address regulatory constraints. Stake limits prevent single players from strongly moving the market (turning the market into an auction), but multiple players from a single interest group could do so (turning the market into a skewed poll).
4. Calibrate prediction data for very low probabilities, which behavioral research indicates people judge poorly. Prediction markets may have to avoid “black swan” event betting.
5. Clearly distinguish between correlation and causation to facilitate application of prediction market outcomes to real-world decisions. In general, design of betting options that reasonably establish causality results in complex, abstract propositions that aggravate issues 1-4.

In summary, prediction markets as presently constituted have significant limitations that affect both the accuracy of their predictions and the portability of those predictions to real-world use.

Enabling Useful Prediction Markets

In their May 2007 “Statement on Prediction Markets”, a prestigious group of economists and political scientists state that the U.S. government should stimulate innovative design and use of prediction markets by relaxing regulatory constraints. They argue that:

• A system such as a prediction market that incorporates broad reach, stake holding and the profit motive encourages people to seek and thoughtfully share information that is widely dispersed in society.
• Prediction markets therefore can provide more accurate forecasts of future events than other methods, thereby substantially improving decision making and risk management in the private and public sectors.
• The Commodity Futures Trading Commission (CFTC) should establish a safe harbor for certain types of small stakes markets. This safe harbor should be available to not-for-profit research institutions (universities and think tanks) and government agencies, and to private firms for internal use only. Such markets should be operated on a not-for-profit basis to price economically meaningful risks or uncertainties (not sports), with individual capital stakes limited to perhaps $2,000.
• The CFTC should allow operators of such markets to experiment with fees, liquidity enhancements, manipulation countermeasures and other design factors that affect market performance.
• Congress should support the CFTC with any required funding and, if necessary, with enabling legislation.

In summary, many eminent economists and political scientists believe that prediction markets could offer significant benefits to society and that government should remove barriers to their productive use.