Futarchy and the Transfer Problem
I recently tweeted on the ineffectiveness of Medicine, and thus on waste we could cut. I was scolded for suggesting to cut government waste while Republicans are in charge. I should know they’d only use that to cut taxes, which only helps billionaires.
This highlights the fact that all policy changes have two very different impacts: changing the overall size of a pie, and changing pie shares. And even though most justify their proposals as trying to increase pie size, they are often more strongly motivated by pie share changes.
I think this issue is part of the motivation for allowing amendments to bills in legislative systems. While in principle all changes to a bill could be passed as further bills at the same process level, we instead make it easier to propose and pass amendments to bills, compared to bills themselves. Bill amendments let people say “I like your pie size change, but not your pie share changes, so here’s a different way to change pie shares that works with your pie size change.”
This is a solution to what I’ll call the “transfer problem” in policy-making. A group who can identify a policy that increase the size of a pie should be tempted to bundle that gain with a policy that increases their own group share. While that share change might actually cut the total pie somewhat, if the bundle still increases the pie on net, they might get enough others to support it to get it adopted.
For example, assume total value is 100, and imagine you find a clever way to increase that value by 5, to 105. Further assume that, within some broad limits, you can always transfer 4 to your group at a cost of 5 to the rest, for a net loss of 1. In this case you are tempted to bundle your 5 increase in the total with 4 repeats of the grab 4 for your group at net cost of 1 move.
With this bundle, your group will gain 16, others will lose 20, and the total will increase by 1, from 100 to 101. And if that is enough to get approval for your bundle, you’d be willing to spend at least 15 to search to find such a clever increase of 5 in the total. But doing that would then actually cut total value by 14. To prevent such transfer waste, legislative systems may allow bill amendments as a way to allow decoupling of pie size policy changes from arbitrary added pie share changes.
Futarchy is my proposed form of governance which approves policies when speculative markets estimate that they will increase some ex-post-measure of unit welfare. And I had not until recently given sufficient thought to how futarchy might be subject to this transfer problem.
Now there are some modes of use of futarchy that aren’t subject to the transfer problem. For example, one could just allow futarchy to veto any of the big decisions made by a org CEO. Or there might be a standard set of possible policy changes that are periodically reconsidered, such as raising more funding, or firing key personnel.
A futarchy use mode more vulnerable to the transfer problem is when anyone may pay to win an auction to get their policy proposal considered by a futarchy system. In this case, people may well propose policies that improve welfare overall, but bundle that with costly transfers, but not so costly that the whole bundle is no longer a net gain.
How can we modify futarchy to deal with this transfer problem? If it were possible for independent accountants or other analysts to find a “no transfers added” variation on any policy proposal, we could just make that a third option in any evaluation: futarchy would choose between doing nothing, adopting the proposal as originally submitted, or adopting the no-transfers-added variation.
But what if we can’t reliably find high enough quality no-transfers-added variations? In this case, I suggest that we adopt the amendment solution. First have an auction for what is the next bill to consider. The winner submits a bill, pays their auction price, and will be paid some share of the market estimated value gain if futarchy approves some variation on their bill. But we pause for a short amendment period to let people pay a set fee to propose transfer variations on that original bill. Such variations change the bill by adding transfers from some groups to others.
After that amendment period, all of these proposals are considered by futarchy, which then picks one or reject them all. Under this system, whatever transfers you bundle into a submitted bill can be stripped a way and countered by amendments. While we can’t ensure that there will be no transfers, bills that transfer less are more likely to win when transfers are inefficient. And bill proposers can’t have much confidence in being able to add transfers they like to the ideas they find for overall improvement.
Here’s another approach: if bill variations could be located in some space according to the magnitude of their transfers to different groups, one could also fit the set of futarchy welfare estimates for variations by some simple overall function with a welfare peak, and then actually adopt the policy at that estimated peak position, even if none of the submitted bills actually made that particular suggestion.
Alas these all add some complexity, which I’ve tried hard to avoid in futarchy. But the transfer problem does seem manageable, if we are willing to add some complexity.
This potentially seems related to something I noticed in a post I wrote about futarchies that change their preferences: Research into how to better achieve the goals of the futarchy is something that can potentially be rewarded by the futarchy itself.
In allowing the amendment period to prevent situations where people spend 15 on research to increase the pie from 100 to 105 while transferring 16 to themselves, we simultaneously prevent people from transferring to themselves in cases where research is worth it. In the good situation where someone can spend 1 on research to increase the pie by 5, they now won't do so, since they won't see the gains, even though doing so would have been net good, even if they did bundle 1-3 transfers to themselves. Perhaps we could prevent a maximally destructive transfer in some other way, like by imposing a value on fairness in the futarchy target.
Do you have any evidence that amendments to a bill are more likely to solve the transfer problem as opposed to creating new transfer problems? We often hear about riders on a bill that promote some special interest unrelated to the main bill, and these riders are often attached via amendment.
It's not clear to me what the difference would be between proposing an amendment in futarchy and just proposing a new bill. So your special interest group proposes something that gains +16 for them, -20 for others, and bundles it with something that gives +5 to everybody. If someone notices this, why don't they just, immediately afterwards, propose a bill that reverses the +16 -20 part of the first bill? That proposal would be +4 on net and would pass.