Wednesday 6 January 2010

Futarchy and its discontents

Chris Masse might well call me a Hanson fanboy for it, but I am a fan of Hanson's Futarchy idea. I like it so much, I usually lecture on it in the closing week of my public choice class.

The mechanism, in a nutshell:
  1. We vote democratically on some aggregate measure of social welfare, which Hanson calls GDP+. Take GDP and whatever else you think it part of "good stuff", assign all components a weight by democratic voting, and voila! You have an aggregate measure of whatever we're trying to maximize. Scale it to some sensible interval.
  2. Open sets of futures markets that pay out based on the value of GDP+ at future dates stretching far into the future.
  3. Open sets of policy markets that pay out based on whether particular policies are implemented. Anyone can propose a policy, though there'd be a charge for proposing a policy as setting up the sets of markets isn't free.
  4. Open sets of contingent contracts for future prices on GDP+ conditional on the policy being implemented or not being implemented.
  5. If a particular policy improves GDP+, as measured by price differences across enough of the futures markets for a long enough period of time, then that policy is implemented.
Hanson poses and answers a long series of objections in the original paper (linked to here), including the potential for market manipulation.

My main worry about futarchy has been the definition of GDP+. Right now, politicians can dissemble relatively easily about the amount of weight they put on various objectives. Many of these objectives are important political symbols to a lot of folks, but they only really want their symbols to be shown respect and attention; they pay little attention to whether effective policies are actually implemented. So protectionists and trade unionists can be bought off with a "Buy New Zealand Made" advertising campaign that's utterly ineffectual in doing anything but showing those groups that the government cares about them and values their views. If politicians don't want to completely kill the economy (and thereby reduce tax revenues), they're well advised to offer up cheap symbolic policies in lieu of growth-killers.

By contrast, GDP+ is completely transparent. There's no reason to expect that the same biases that plague regular voting wouldn't also plague voting on constitutional matters like the definition of GDP+. But it could well be worse because politicians would have far less slack to work around silly definitions of GDP+ than they have to work around silly pandering election promises.

That worry aside, I like Futarchy and I love the idea of contingent markets on the effects of policies as a useful input for policy evaluation.

Paul Hewitt is less keen. He winds up, in the comments, asking Robin three questions.

First and second are whether it's possible for long term markets to be accurate and whether traders would be able to understand GDP+, given that economists have a hard time forecasting simple GDP for medium horizons. I'll expect Robin to answer that so long as whatever inaccuracies there are in such forecasting are constant across the sets of markets for any period of time, then the price differences across such markets are going to be meaningful, and it's the price differences across markets that really matter. Suppose that we have a market on what the value of GDP will be in 20 years time. I have no clue what that number will be; at best I'd take the current value and assume a 2% real growth rate. But that doesn't matter. If there's one market that pays out on GDP in 20 years conditional on no change in policy and the other pays out conditional on adopting a $50/hr minimum wage, I'd short the latter if its price were high relative to the former.

Third, he asks whether the high market listing fee isn't anti-democratic. I'll here expect Robin to reply "relative to what?" I have a very very hard time imagining that it would ever be more expensive to get a policy listed on Hanson markets than it is under the current system to get a policy considered by Parliament.

I'll look forward to seeing Hanson's debate with Moldbug later this month on Futarchy...I trust there'll eventually be a webcast....

Update: Hanson discusses here.


  1. Very interesting idea. One potential issue might be the difficulty making sure that you have a broad/balanced range of citizens & interests represented, otherwise both the policy setting and the pricing may be skewed towards the objectives of specific groups.

  2. @Stompin: Check Hanson's paper; I'm pretty sure that's one of his raised (and answered) objections. So long as GDP+ is sound and there are liquid traders who care about profiting on their trades, then it doesn't much matter whether traders aren't like other folks. We only need the marginal traders to be sensible.

  3. You make a good point that the cost of policy proposals may not be much different from that in existing political systems. However, in many existing systems, when the stench of bad proposals gets high enough, the citizens will protest (even Canadians!), revolt or perhaps vote. The "costs" are not quite the same. Under futarchy, the citizens have to write a cheque. Today, they can spend their time and effort to make their policy preferences known.

    You seem to think that just because the GDP+ definition is known, it is "completely transparent". I'm afraid I have to disagree. My point is that as the definition becomes more complex (as described by Robin Hanson), it becomes less "transparent". Few, if any, will know how it "works" or what it "means". In this sense, revealing the calculation doesn't mean "transparency".

    The problem with GDP+ is that it is far too simplistic, even in its most complex form. It hard-wires relationships among, and tradeoffs between, values. Not only does it specify these for the present day, but it requires that the relationships and tradeoffs not change for a very long period of time. Further, the same relationships and tradeoffs are deemed to apply to all policy proposals presently being considered. It is a gross simplification of decision-making, which has little chance for success.

    You commented that Robin's paper adequately addresses the manipulation issue. It does not. It merely assumes the problem away. My post on futarchy contains a number of links to other posts that show this clearly.

    As for long-term prediction markets, there is no proof, whatsoever, that they are accurate or "calibrated". They may be calibrated, but even if they are, they display a very high level of uncertainty about the outcome, making them useless for decision-making. The markets are merely telling us that we need more (a lot more) information.

    I'm afraid that StompinRhino would be very disappointed with futarchy, because prediction markets have nothing to do with a "balanced" range of citizens and interests. These markets do seek diversity, but the "informed" feast on the chimps in order to make the markets work.

    I'm sure I could like futarchy "so much" too, if it worked "as advertised". While the benefits of such an institution might be extremely significant, applying a cost-benefit analysis to the further investigation of futarchy, we would find the costs far outweigh the expected benefits (i.e. very high potential benefit times an extremely low likelihood = very small expected benefit).

  4. @TorontoPM: I'm about to be without internet access for a couple days, so apologies if I don't get back to you quickly. But I'd make a few points:
    1. When was the last time there was a revolution in any country that might plausibly start moving towards futarchy?
    2. I'm more worried about the weights being influenced too strongly by expressive voting and being fixed than I am that they're not entirely transparent to most voters: all we need is that the marginal trader gets it.
    3. I'll take the point on hard-wiring, but only because I don't think that we can get the hard wired numbers right to start with.
    4. I've seen your critiques on manipulation, but am still rather of the view that with markets where traders aren't bounded by liquidity constraints, manipulation won't be a big issue (or at least no bigger an issue than similar problems under democracy)
    5. Again, the markets may well have high variance over long time periods. But so long as they get the difference across markets right, it's all good. So long as most of the uncertainty washes out in the fixed effects, all's great! And even if not, we still don't need to achieve perfection, just beat the status quo.