Normative Prediction Market
A normative prediction market can provide assessments not tainted by the ideology of current officials concerning questions that are purely normative. Such a market predicts a subjective assessment of an issue to be made at some point in the future. That point could be relatively near if there is some method for identifying a relatively representative group of potential decision makers. For example, if what is most relevant is the view of the hypothetical average informed voter, then one might select at random a single voter, provide the voter with information about the relevant issues, and then ask for that voter’s opinion. That voter would not decide the relevant issue–that would be a system that embraces second-order diversity to an extreme–but the voter’s conclusion would determine payouts to the participants in the normative prediction market. Such a market would provide an assessment of the way an average member of the population would be expected to treat a particular issue. This is a relatively populist approach, but for those who believe that administrative agency officials have expertise that voters could not harness, even with a full presentation, an alternative is to select a decision maker from a pool of experts. The challenge is in identifying a suitable pool, given that there may be legitimate disagreement, itself politically charged, about who should count as an expert. Those who are skeptical about environmental issues, for example, will generally not be willing to count any self-proclaimed environmentalist as a neutral expert, and those skeptical about corporate There is, however, a relatively simple solution that anchors the random selection to the political process while avoiding the randomness and variance associated with individual electoral outcomes. That solution is to require that the decision be made by a particular administrative official such as the head of the agency some number of years in the future. The key is for that future point to be far enough away that it will be difficult to forecast who the decision maker might be. Of course, sometimes it is possible to infer that in ten years some anticipated demographic change will shift the country in one direction or another, but in general, the average expected official will be a close approximation of the average current expert opinion about a particular issue. After a decade the political crisis of the moment would be unlikely to provide much guidance about whether a Democrat or a Republican is in the White House, and so a prediction market will weight the possibility of each outcome at roughly 50 percent. Delay thus serves as a cloaking device, disguising the identities of the eventual decision makers from prediction market participants so that they make their predictions based on some idealized pool. That does not mean, however, that the decision to adopt a normative prediction market itself is politically neutral. Often it will be apparent that adoption of such a market in some context will lead to a predictable deviation from the status quo. Whether one embraces a normative prediction market in a particular context will depend on whether one approves of this expected deviation. This will, of course, make such markets more difficult to enact, because if there is strong support for a particular shift in policy, a majority would not need a prediction market to achieve it, and if there is not, then the majority will not want the prediction market. Those who oppose adoption of normative prediction markets, moreover, are not necessarily wrong. Individual policies that such markets might produce matter, and it may be legitimate to oppose a procedural device on substantive grounds. Moreover, there might be legitimate grounds for debate about who should constitute the pool of ex post decision makers and for opposing any particular proposal for a normative prediction market on that basis. Ultimately, a normative prediction market is a device for making policy conform more closely to what a specified pool of individuals could be expected to decide if they took the time to study an issue thoroughly. Even as a matter of political theory across a wide range of issues, it may be debatable whether the expected views of any particular defined pool of decision makers should be more legitimate than the actual decisions that emerge from any more traditionally structured democratic process. An objection to normative prediction markets is that they do not really weigh preferences but instead anticipate preferences. Sure, but so what? Sometimes anticipated preferences might turn out not to forecast actual preferences correctly, for example, because market participants have insufficient information. But this problem can be reduced by greater market subsidy. More important, however, republican government itself is a rather imprecise tool for aggregating preferences. As in all prediction market proposals, the question is how the market-based institution would fare in comparison to the existing institution. It seems plausible that “anticipated preferences” might come closer to meaningfully combining the public’s actual preferences or the public’s hypothetical informed preferences than does the existing legal system. Agencies’ decisions in many cases are quite far from any plausible aggregation of the public’s views as a result of ideology. Perhaps there are many decisions for which the processes of republican government help determine a reasonable aggregation of the public’s preferences. With sufficient subsidy, however, prediction market participants might select random decision makers, provide them with information, and measure their reactions. Of course, government could continue to use existing processes for gathering data and arguments, although prediction markets might be designed to accomplish much the same thing. We have seen, for example, that a deliberative prediction market might be used as an alternative to the traditional process of notice-and-comment rule making (see Chapter 4), but predicting the evaluation of a random decision maker rather than of current agency decision makers might further improve the process. A benefit of the traditional notice-and-comment process is that it forces agency decision makers actually to consider and respond to comments. A designer of a normative prediction market might develop a more formal process by which the randomly selected ex post decision maker would consider the relevant arguments. For example, rules might provide some form of adversarial presentation, both to ensure that all views are presented and that the decision maker actually takes the time to hear them. Advocates might be selected (preferably ex post, so that the anticipated quality of the advocacy does not affect the prediction market) who have some financial incentive to convince the randomly selected decision maker. This would provide market participants with incentives to identify all relevant arguments, not simply those that would be appealing to the current leadership of the administrative agency. It also might be desirable to require or expect that the ex post decision makers would give detailed analysis. They otherwise might shirk, and to the extent that shirking leads to predictable errors, that would harm policy. Granting power to decision makers may be one way of reducing shirking, but it should be possible to develop effective procedural substitutes.
Leave a Reply