We use futarchy (we call it prognootling) for family decisions sometimes. I gave examples in my talk at Manifest which will supposedly be on YouTube soon.
It points out how a conditional market like "If we take action A, will outcome B happen?" might tell you there's a strong _correlation_ between A and B without telling you that A will cause B. For example, maybe a conditional market says that revenue will nosedive if we fire the CEO. Does that mean we shouldn't fire the CEO? Not necessarily! If we did fire the CEO then probably we're in a universe where the company is imploding. The causation could run the opposite way.
I know you've thought about that issue -- how much it matters and how to mitigate it -- but it would be great to see your direct response to Dynomight.
That's not how it works because the conditional market is *used to decide* whether to fire the CEO. It's not a passive observer of whether the CEO gets fired.
So you're already in the universe where the company is imploding (or not); that's not going to change. The market fires the CEO if E(revenue | did not fire CEO) < E(revenue | fired CEO). If the company is imploding, then E(revenue | fired CEO) will be low (nosediving revenue), but E(revenue | did not fire CEO) would be even lower, so the CEO gets fired.
I am selecting quantified self experiments based on the effect size predicted by a couple of markets on Manifold, more detail at https://niplav.site/platforms. One experiment has finished already, the second is being run.
I don't agree that redistribution is an issue at all. In that case, you've just selected for a crappy metric: dollars invested. If wiping out the existing shareholders is indeed the best way to maximize for new dollars invested, then the futarchy is just doing its job.
If you select instead for share price, redistribution shouldn't be an issue. Diluting the existing shareholders by 60% would make those shares far less valuable. We've had investment proposals to MetaDAO fail because the market deemed the dilution to not be worth the investment dollars.
"In futarchy markets, new information asymmetrically affects bet values. Negative information can nullify the value of “bad outcome” shares by canceling the proposed action, while positive information increases the value of “good outcome” shares without limit. This asymmetry favors “good outcome” shares, skewing futarchy forecasts. This effect is akin to “convexity” in the context of options.
The magnitude of this effect is a function of the amount of new information expected to come to light before the decision deadline. As a result, an attacker could exploit this to increase the value of “good outcome” shares for their desired action, by selecting an action with outsized uncertainty/expected volatility. One naive way of achieving this is to announce that important information relating to the proposal will be announced at some time over the course of the futarchy market."
Out of curiosity, where did you get future markets from? I've always associated it with coming out of the early 90s cypherpunk / financial cryptography crowd but I just don't see you as a Tim May, Jim Bell, Lucky Green sort of guy.
Do you need to account for the interest rates available on cash given that people tie money up in bets until the bet is settled. Given they can get x percent on cash do the probabilities need to account for this in the price to probability function?
Why did you call it futarchy? I mean, I understand that it was a play on the word "future" (which makes it, like automobile, a Latin-Greek mishmash, less than ideal by itself), but did the... alternative reading never visit your head in all these years?
Could you describe a mechanism for using futarchy to set central bank interests rates if our aim were to bring the integrated inflation rate to a target x?
In that paper (Hanson 2013, Shall We Vote on Values, But Bet on Beliefs?*) you say “If the up or the down market had a consistently higher outcome price over most of that week relative to the same market, interest rates would be raised or lowered accordingly.” Can you be more quantitate? Say the price for the bet "inflation will be above inflation target x", was 0.6, and the price for inflation will be below target was 0.38 (not summing to 1 allowing for spread) then what change do you make to current interest rate y. Wouldn’t this suffer from interest rates taking time to filter through to inflation and inflation having other short term causes?
Robert, I am writing a book (and a blog) about the disruption of nations by the Internet, and there is a chapter dedicated to new forms of governance. I am currently working on the section about futarchies, and I am wondering if you have considered the following idea: in a futarchic government, when citizens have to vote for economic programs or parties and their programs, they should invest a percentage of their income, which can be determined by law, in order to vote. This would have the merit of aligning people's incentives with the actual outcomes of these programs.
Under futarchy, they bet, not vote. If they bet equal amounts on opposite sides, they could bet a lot of money without actually taking a net stake. So it doesn't work to just require a total bet amount per person.
Okay, but surely there would be enough people on each side who would want to bet only on the policies they think will have positive effects, so that people betting on both sides couldn't influence the outcome of the bets?
To solve the redistribution problem, you just need a better outcome metric than total capital invested over 20 years.
Somebody *currently* owns the organization, and futarchy is in the role of a skilled manager working for the current owners. So, as in the case of a human skilled manager, the outcome metric should be a measure of profit to the current owners. The current owners don't care how much capital is invested in the company if they don't see that money themselves, so they would have no reason to approve a redistribution decision if a human manager suggested it, so they would have no reason to approve it if a futarchy market suggested it.
Specifically, for a corporation, the outcome metric could be the sum of (dividends paid to current shareholders) + (profit from the current shareholders selling their shares) + (salaries paid to the current shareholders) + (any other source of income from the corporation to the current shareholders) as a discounted rate over some time period.
That's a good point. You could just let the SEC govern any transfer of shares and dividends in the company between the current owners as it does now; the shares (in the company, not the futarchy market) must be bought and sold normally and any dividends paid equitably according to the number of shares held, the futarchy market cannot reassign company shares at will.
We use futarchy (we call it prognootling) for family decisions sometimes. I gave examples in my talk at Manifest which will supposedly be on YouTube soon.
On a more theoretical note, I've been mulling this thoughtful warning about limits on the application of decision markets: https://dynomight.net/prediction-market-causation/
It points out how a conditional market like "If we take action A, will outcome B happen?" might tell you there's a strong _correlation_ between A and B without telling you that A will cause B. For example, maybe a conditional market says that revenue will nosedive if we fire the CEO. Does that mean we shouldn't fire the CEO? Not necessarily! If we did fire the CEO then probably we're in a universe where the company is imploding. The causation could run the opposite way.
I know you've thought about that issue -- how much it matters and how to mitigate it -- but it would be great to see your direct response to Dynomight.
The two paragraphs starting "But there are many details to consider" are directly on those issues.
That's not how it works because the conditional market is *used to decide* whether to fire the CEO. It's not a passive observer of whether the CEO gets fired.
So you're already in the universe where the company is imploding (or not); that's not going to change. The market fires the CEO if E(revenue | did not fire CEO) < E(revenue | fired CEO). If the company is imploding, then E(revenue | fired CEO) will be low (nosediving revenue), but E(revenue | did not fire CEO) would be even lower, so the CEO gets fired.
I am selecting quantified self experiments based on the effect size predicted by a couple of markets on Manifold, more detail at https://niplav.site/platforms. One experiment has finished already, the second is being run.
I don't agree that redistribution is an issue at all. In that case, you've just selected for a crappy metric: dollars invested. If wiping out the existing shareholders is indeed the best way to maximize for new dollars invested, then the futarchy is just doing its job.
If you select instead for share price, redistribution shouldn't be an issue. Diluting the existing shareholders by 60% would make those shares far less valuable. We've had investment proposals to MetaDAO fail because the market deemed the dilution to not be worth the investment dollars.
Huh, ever read Shockwave Rider? It has prediction markets, sort of.
https://distbit.xyz/correlation-vs-causation-in-futarchy/
key takeaway:
"In futarchy markets, new information asymmetrically affects bet values. Negative information can nullify the value of “bad outcome” shares by canceling the proposed action, while positive information increases the value of “good outcome” shares without limit. This asymmetry favors “good outcome” shares, skewing futarchy forecasts. This effect is akin to “convexity” in the context of options.
The magnitude of this effect is a function of the amount of new information expected to come to light before the decision deadline. As a result, an attacker could exploit this to increase the value of “good outcome” shares for their desired action, by selecting an action with outsized uncertainty/expected volatility. One naive way of achieving this is to announce that important information relating to the proposal will be announced at some time over the course of the futarchy market."
Spreadsheet with relevant example: https://docs.google.com/spreadsheets/d/1TNM85DoQqOvlQFZJQ20yEaK6QsS33ZIo3fcyws5X124/edit
Out of curiosity, where did you get future markets from? I've always associated it with coming out of the early 90s cypherpunk / financial cryptography crowd but I just don't see you as a Tim May, Jim Bell, Lucky Green sort of guy.
Do you need to account for the interest rates available on cash given that people tie money up in bets until the bet is settled. Given they can get x percent on cash do the probabilities need to account for this in the price to probability function?
In my prior work I make it clear that I prefer people to bet assets that appreciate in value, like stock or bond index funds.
Why did you call it futarchy? I mean, I understand that it was a play on the word "future" (which makes it, like automobile, a Latin-Greek mishmash, less than ideal by itself), but did the... alternative reading never visit your head in all these years?
No alternative meaning occurred to me in 1999.
Could you describe a mechanism for using futarchy to set central bank interests rates if our aim were to bring the integrated inflation rate to a target x?
I use that as an example in my main futarchy paper.
In that paper (Hanson 2013, Shall We Vote on Values, But Bet on Beliefs?*) you say “If the up or the down market had a consistently higher outcome price over most of that week relative to the same market, interest rates would be raised or lowered accordingly.” Can you be more quantitate? Say the price for the bet "inflation will be above inflation target x", was 0.6, and the price for inflation will be below target was 0.38 (not summing to 1 allowing for spread) then what change do you make to current interest rate y. Wouldn’t this suffer from interest rates taking time to filter through to inflation and inflation having other short term causes?
Those aren't the sort of market estimates I propose using.
Robert, I am writing a book (and a blog) about the disruption of nations by the Internet, and there is a chapter dedicated to new forms of governance. I am currently working on the section about futarchies, and I am wondering if you have considered the following idea: in a futarchic government, when citizens have to vote for economic programs or parties and their programs, they should invest a percentage of their income, which can be determined by law, in order to vote. This would have the merit of aligning people's incentives with the actual outcomes of these programs.
What do you think of this ?
Under futarchy, they bet, not vote. If they bet equal amounts on opposite sides, they could bet a lot of money without actually taking a net stake. So it doesn't work to just require a total bet amount per person.
Okay, but surely there would be enough people on each side who would want to bet only on the policies they think will have positive effects, so that people betting on both sides couldn't influence the outcome of the bets?
You are proposing a constraint and I'm saying the constraint doesn't actually constrain.
To solve the redistribution problem, you just need a better outcome metric than total capital invested over 20 years.
Somebody *currently* owns the organization, and futarchy is in the role of a skilled manager working for the current owners. So, as in the case of a human skilled manager, the outcome metric should be a measure of profit to the current owners. The current owners don't care how much capital is invested in the company if they don't see that money themselves, so they would have no reason to approve a redistribution decision if a human manager suggested it, so they would have no reason to approve it if a futarchy market suggested it.
Specifically, for a corporation, the outcome metric could be the sum of (dividends paid to current shareholders) + (profit from the current shareholders selling their shares) + (salaries paid to the current shareholders) + (any other source of income from the corporation to the current shareholders) as a discounted rate over some time period.
But even the metric of average return to recent owners allows for redistribution between such owners.
That's a good point. You could just let the SEC govern any transfer of shares and dividends in the company between the current owners as it does now; the shares (in the company, not the futarchy market) must be bought and sold normally and any dividends paid equitably according to the number of shares held, the futarchy market cannot reassign company shares at will.
Or you could go the full "vote values" route and let the outcome metric be voted on regularly by all current shareholders. But in that case, beware of the king! This fascinating logic puzzle shows how a king with the ability to propose wealth redistributions to be voted on, can ultimately reassign almost every piece of gold to himself. https://puzzling.stackexchange.com/questions/127242/what-s-the-highest-salary-the-greedy-king-can-arrange-for-himself
> Those who paid for approved proposals should then a fraction (maybe a third) of the market estimate of the proposal’s value
I think you're missing a verb there around "fraction".
Fixed; thanks.