Futarchy vs. Predictocracy

In 2000 I proposed “futarchy“, a government by prediction markets where we would “vote on values but bet on beliefs.”  In his 2008 book, Michael Abramowicz summarizes:

At the heart of Hanson’s proposal is the use of conditional markets to estimate the effect of proposed policies on a measure of national welfare. … The existing legislature … could pass legislation to change the welfare measure, producing what he calls GDP+ … Any policy that, according to a prediction market, would clearly improve GDP+ would be automatically adopted.

In contrast, Abramowicz prefers “predictocracy”:

The essence of predictocracy is that decisions are made on the basis of the market’s anticipation of whether they will … be approved of later. … Individuals in this legislature might be asked to vote not on values that others would then seek to achieve but instead on concrete statutory proposals, as existing legislatures do today. These legislators’ decisions, however, would have no immediate effect but would serve only to discipline the text-authoring market, whose decisions would determine whether particular amendments were allowed and particular bills were enacted.

Why is this better?

[Predictocracy] can evaluate small as well as large decisions. Perhaps the most important legislative changes would be those that would have a large anticipatable impact on the measure of national welfare, but a legislature ought to be able to make smaller changes as well. At the same time, the market-based legislature would not need to rely on proxies for the national welfare but could directly forecast ex post decision makers’ assessments of whether particular proposals would increase or reduce welfare. Those decision makers … would apply their own intuitions and judgment about issues that elude easy measurement.

The issue here is how to make our most basic decisions; we agree that each approach, as a basic rule, could authorize the other approach to make many other decisions.  If we can estimate costs and error rates for different decision rules, then for the biggest decisions we should prefer the least error-prone rules, even if they have high costs, such as being slow or troublesome.  For smaller decisions, we may prefer “fast and frugal,” even if less accurate, rules.  So I am not very concerned about futarchy giving unclear answers to small questions, but I am very concerned about futarchy having higher decision errors.

So let us catalog the errors, starting with common ones.  Futarchy, predictocracy, and standard legislative democracy today all suffer from misjudgments and fluctuations in who is allowed to vote at what cost, who votes in what district, who runs for office, who is how informed about candidate inclinations, and so on.  All these systems all also suffer from errors in whatever agenda process suggests bills for consideration, and errors in the court and police processes that enforce provisions of approved bills.

Democracy today suffers from enormous errors regarding estimates of policy consequences, i.e., of passing particular bills.  Voters have serious illusions and misconceptions that sway their minds on election day, when they have little expertise and only mild motivations to attend to their task.  Candidates have strong expertise and incentives to attend to their task, but that task is largely to pander to voter illusions.  Under futarchy, the task of estimating policy consequences would be transferred to self-selected market speculators, with very strong incentives to make accurate estimates on very large decisions. Large decisions give large manipulation incentives which give large market subsidies.

Futarchy would require legislators to explicitly define national welfare, and definition errors would arise for the same reasons ordinary legislatures today make errors in all bills.  In addition, the very process of trying to define an explicit measure would introduce deviations from the values that would otherwise have been expressed regarding case-specific decisions.  Whether these deviations increase errors via neglect of detail or reduce errors via “smoothing” mistakes depends on how errors in value judgments are distributed.

Predictocracy would instead replace legislative votes with market predictions about the verdicts of random individual legislators at some future date.  Large decisions would again give large incentives to speculators to estimate well, but what they would be estimating would be the decision of a legislator trying to pander to voter illusions. Yes, those illusions would be expressed later in time after nature has revealed some info, and speculators would have strong incentives to persuade voters.  But I can’t be very optimistic that this will give large accuracy gains in estimating policy consequences.  Illusions and misconceptions of weakly motivated amateur voters can plague retrospective evaluations of the effectiveness of past policies nearly as easily as prospective evaluations of future policies.

Given all the other errors we expect in values expressed by legislatures, and since it is not clear if smoothing values via an explicit welfare function increases or decreases errors in values expressed in case-specific decisions, it seems to me that, relative to predictocracy, futarchy risks only a small increase in value errors, and offers an enormous decrease in policy consequence errors.  Seems an easy choice to me.

Added 2Feb: Michael responds over at Volokh Conspiracy.