Conceptual and Practical Obstacles to Futarchy

In his comments on my post last week, Robin Hanson asked about the conceptual work still needed to advance the cause of prediction markets as tools for governance. To my knowledge, Hanson’s definitive statement of futarchy can be found here:

When an approved betting market clearly estimates that a proposed policy would increase expected GDP+ (E[W|N] > E[W|Q]), that proposal immediately becomes law. [“GDP+” is the Gross Domestic Product plus unspecified “measures of lifespan, leisure, environmental quality, cultural prowess, and happiness.” “(E[W|N] > E[W|Q])” is the formula for deciding whether the  proposed policy has a probability-adjusted benefit greater than the status quo: “When the market estimate of E[W|N] is clearly greater than E[W|Q], speculators are saying that this new policy is expected to increase national welfare.”]
Hanson proposes to give certain kinds of “ideas markets” legislative authority. Conceptually, Hanson is building on some of the research into public ignorance to seek better mechanisms for informed policy-making, which is an undeniably laudable goal. Yet here are some conceptual challenges I believe he must still address:
  • How are we going to measure GDP+?

We don’t currently do ordinary regulatory cost-benefit analysis using anything approaching a full-fledged social welfare function. What reason do we have to think that futures contracts would be enforced using a fully-inclusive measure of the Gross National Happiness? This suggests that metric-design is one of the most important conceptual and practical obstacles to futarchy.

  • How many contracts would there be?

There are many more possible law and rule changes than the number of equities that are regularly traded in the stock market. A healthy IPO market might have 360 new stocks offered each year. In contrast, there were 3503 new rules reported in the 2009 Federal Register. That’s 3503 actual rules made, not proposed or hypothetical rules: a prediction market would have to consider many different variations on each of those rules. Could we adequately capitalize that many different markets each year?

Of course, it’s important to recognize that most of these new rules are fairly inconsequential, sometimes internal to a specific agency. A better guide is the “major” or “economically significant” rulemakings, which cause at least $100 million of effects. Of such major rules, there were only 23 in 2009, which is a much more manageable number. (Cite) However, we still face the problem that prediction market contracts must track specific policies, and so we would have to multiply those 23 major rules by the number of substantial variations in those rules. For such a rule with $100 million impact, it’s easy to imagine one hundred variations, or 2300 possible contracts.

What’s more, a prediction market would end up consider many more rules than those proposed by relatively limited federal agencies directed by the executive branch. All the old major rules would also be on the table for possible revision. There’s currently about $1.1 trillion dollars in federal regulation, or about 11,000 $100 million dollar rules. And this is only the federal government: we’d have to also identify the regulations at the state and local level that would be subject to futarchy. Again, let’s say there are at least one hundred possible changes to each of those major rules. Each of these markets would need to be capitalized at a fairly large ratio of the expected impact of the policy being put into effect: how large should the market for a rule that could cost or benefit $1 million be? Now we’re back to astronomical numbers… or following Richard Feynman, perhaps we should say economical numbers.

A functioning prediction market for each of those regulations would have to be  orders of magnitude larger than the size of the entire “real” economy it governed. Yet I think  that Hanson could revise his futarchy by allowing agencies to limit the number of major rules and variations of those rules under consideration at any given time. If the regulatory state only submitted 23 new contracts each year for a simple up-or-down market vote, (i.e. a clear margin of acceptance or veto by the predictions markets) we could easily have a manageable prediction market on each of them.

  • Are markets a coherent decision procedure?

One of the most important findings in decision theory is Kenneth Arrow’s impossibility theorem. Arrow argued that the fairly plausible expectations of democratic decision-making do not actually allow us to coherently aggregate our preferences. There have been similar impossibility results in markets as well, notably Grossman and Stiglitz’s paper “On the Impossibility of Informationally Efficient Markets.” In general, impossibility results are a conceptual concern that needs to be addressed by proponents of futarchy, just as we have tried to address them in deliberative democracy circles under the title “Discursive Dilemma.” Of course, we usually ignore this problem in democracy, just as we ignore it in deliberation and in the currently extant markets.

At the conceptual level, however, there remains the possibility of interaction between the Grossman and Stiglitz result and Arrow’s Paradox. This would primarily result from the violation of the “non-imposition” criterion, which dictates that the decision procedure ought always to allow consideration of all possible policy preferences, or in this case, to allow markets for all possible policy-effect predictions. That said, I’ve already suggested that we would need to violate non-imposition by leaving the rulemaking up to the regulatory proposal process.

  • Will I be freer or more dominated in a futarchy?

On its own terms, futarchy doesn’t need to worry about the liberal impulse to be free from domination. We are all slaves of ignorance, after all. Yet I would have thought that Hanson would understand that sometimes we’d rather be free and wrong than coerced and right. This is one of the larger concerns generated by the manipulation literature: once it’s possible that we’ll do whatever it takes to expand GDP+, it’s likely that some very restrictive regulations may ultimately lead to greater GDP+ than the status quo. In response to related hypothetical criticisms, Hanson allows for constitutional limitations on futarchy’s legislative authority, including perhaps liberal rights. This is yet another violation of non-imposition, with many possible unintended or market-information-impairing consequences. (It’s my understanding that Hanson would prefer to develop a social welfare function that includes rights for this reason, rather than allowing the existence of overpowerful judicial vetoes and legislative trump cards that stymie our best policies.)

  • What kinds of market manipulation will be possible under futarchy?

Hanson has done some small experiments with undergraduates that suggest that we shouldn’t worry about market manipulation so much. But I continue to worry about the impact of outsized or overleveraged actors who have two ways to win in a futarchical prediction market: either they can change law or they can levy a hedging tax on those who oppose them. This is what hedge funds like Long-Term Capital Management did in the nineties, and even Hanson acknowledges that during transition times there can be wild swings as market actors try to determine whether a manipulator has better information or is simply trying to cash in. I think there’s ample room for some experiments on prediction markets where some participants have significantly more capital than others, or else have multiple strategies for victory beyond simply getting paid under the prediction contract.

In the comments to the last post, Hanson defended futarchy from this criticism that it hasn’t properly been tested or conceptualized by questioning the kinds of rigorous testing that has been performed on small group deliberations like juries. This is a valid criticism. Obviously both markets and juries are quite common in the wild, but just because we’ve had ample experience with an institution doesn’t necessarily mean that we understand the mechanisms by which that institution functions or thrives. Yet both academics and law firms run hundreds of mock juries every year to “game” possible manipulation scenarios. Markets receive much less direct experimental attention.

In his paper on futarchy, Hanson styles himself an “engineer” rather than a “scientist” in the sense that he’s less concerned with always being right than with advancing promising projects. In this regard, however, he has the monikers backwards. It’s the engineers who best understand that effective product development occasionally requires you to respond charitably to what you believe are uninformed or nonsensical criticisms. Recall how Thomas Edison discredited his rival’s alternating current design for electrical transmission by advocating its use in executions. All I’m suggesting is for some experimental models that respond to a popular concern. Perhaps, in the process, we will come to understand that outsized actors have less impact on markets than we generally assume. Yet it’s also possible we will learn that overleveraged or dual-strategy market participants are a matter of real concern. And that, too, would be good to know.

Second Opinions