Thanks for writing this, very interesting piece. I think your points about the lack of robustness, goodhearting and thin markets are very true. But I’m more skeptical on some other issues:
“the market equilibrium price [would be] an average of beliefs in the trading population, although each belief [would be] weighted according to the average wealth of traders holding that belief, relative to the average wealth level.”[7] (This might happen because the poor are less willing to risk larger amounts of money on futarchy markets, because they have less time available to spend on this, or because they lack necessary knowledge on how to participate, which would all lead to unequal participation in the markets.) This poses an issue if there is no expectation that a wealthy selection of voters will have representative (or somehow better) views than a random selection of voters
This is true, but we absolutely should have an expectation that wealthy voters will have more accurate beliefs! Wealth is positively related to both education and IQ, both of which are linked to having more accurate views. Additionally, wealth plausibly means you are less stressed, which can improve cognitive performance, and gives you access to more information.
I also think you under-state the extent to which current policy is determined by elites under the status quo. My impression is that currently the views of the lower classes have very little impact on policy—see for example Affluence and Influence by Gilens for evidence that US Presidents tend to always massively favour the views of the upper classes over the lower classes (even for George Bush, who had the smallest such bias).
One mechanism by which this occurs seems to be the status of different policies. Politicians want to be high status, and to associate with high status views that are supported by their high status friends, newspapers, academics etc. This is significantly reduced with markets—hedge fund managers in general are perfectly happy to own Walmart’s stock—and even more so if bets are anonymous. I’ve never encountered any credible argument that (wealthy) hedge fund guys have trouble pricing Discount Retailers or Tobacco stocks just because their customers are from a different social class.
Under futarchy, however, if a strange policy were to reliably promise to significantly increase welfare for the next 20 years and then somehow implode, tanking national welfare for the next 100 years, it is quite likely that the policy would pass, as bettors who buy its shares would get paid out on the shorter time-span, and patient bettors are rare, so they would not correct the market. As one example, if you think a market is mispriced by 10%, you’d happily invest in it (and thus correct the market a bit) if the market resolves in a year. But if you think it is mispriced in a way that resolves in 20 years (as in the example above), then it doesn’t make sense to invest; that’s a really poor return on investment.
If the ‘implosion’ is really that bad, then the long-term futures would have a very low fair price, and you would be able to get a decent return on them. It’s true that if the ‘implosion’ was only slightly worse than the short-term benefit, then time value of money might mean that the long term futures were not sufficiently cheap. But then the issue seems to be that your preferred discount rate (0) is different from market interest rates; that’s a common issue that’s not specific to futarchy. So long as the futures market is properly constructed (e.g. daily variation margin with a well-capitalised clearinghouse) the discount rate applied should be essentially the risk free rate.
Moreover, futarchy relies more on self-interest than current systems (which elect or employ decision-makers who are expected to take the interests of those they represent into account), and thus has to be more robust to cold-heartedly self-interested decision-making than modern systems do, so issues in the incentive structure are arguably more significant for futarchy than they are for e.g. representative democracy.
The design of prediction markets (and futarchy specifically) spend a lot of time discussing private interest alignment, which makes it very salient. In contrast, defenders of democracy spend very little time discussing how well aligned the private interests are of the President with the country, because this alignment is clearly very poor. But this potentially creates a misleading impression, conflating compatability with reliance: market can deal with self-interest in a way that politics cannot, but I’m not sure why we would jump from that to assuming they need it. It seems possible for someone to trade out of altruistic motives? Certainly the high participation we see in some prediction markets with either no or very small amounts of money at stake suggest this.
Hanson dismisses the possibility of this manipulation based on the total amount of money that the market can channel to counteract this entity (what he calls the “one rich fool” problem), but Hanson seems to assume that these actions would be visible or at least that the market would be confident enough to notice that manipulation is occurring.
It might be a good idea to link to some of the papers Hanson has on the subject, e.g. here. More substantially, I think your response proves too much. Yes, if no-one can tell a market is mis-priced and no-one can tell there is manipulation, speculators will not be able to correct the market price. But if that is the case then we clearly have a situation of massive ignorance about the policy, in which case I wouldn’t expect democracy or any other system to do well either.
He also does not seem to account for the possibility that the market will follow the rich person’s bets rather than try correcting that person.[11] (The idea that rich people may have extreme influence in such markets is also corroborated by the behavior of cryptocurrency markets in response to the actions of cryptocurrency whales.)
There’s a big difference between currencies (indefinite life) and futures. A cryptocurrency can maintain ‘irrationally’ high valuation for an indefinite period, in the same way that a piece of art can maintain a high value for a long time, even though you can just print out a near perfect copy. But a future has a termination at a specific point in time: it has to converge to the correct price by then, one way or another. This dramatically reduces incentive to follow along with a large irrational player: it is a finitely iterated game with a known end point. AMC’s shares can stay high indefinitely, but a future set to the value of their 2021 revenue has to come down to earth within a short period, so would probably never have spiked in the first place.
Finally a very pedantic comment:
As a caricature of an example, if a significant part of the welfare measure is GDP, one law that would pass if proposed is “people are required to spend an hour a day giving each other $100 bills” (since every transfer of a $100 bill increases GDP by $100).
Actually this would not increase GDP, because transfers are not counted, only consumption, investment etc. But there are closely related failure modes—e.g. the GDP benefit of women entering the workforce is overstated because previously they were doing useful work that was simply unwaged and hence excluded—so your underlying concern is correct.
This is significantly reduced with markets—hedge fund managers in general are perfectly happy to own Walmart’s stock—and even more so if bets are anonymous. I’ve never encountered any credible argument that (wealthy) hedge fund guys have trouble pricing Discount Retailers or Tobacco stocks just because their customers are from a different social class.
Is this a good analogy for prediction markets, though? The stock prices don’t feed directly into policy decisions about taxes and regulations, but prediction market prices would under futarchy (unless you limit their scope to exclude specific policies), from my understanding. Betters have a better path to influence policy in their favour through these prediction markets than through stock markets. I don’t know how large these effects would be, but this is discussed in the section “(5) Buying policies: manipulation through extreme wealth”.
Of course, in many countries they can directly lobby governments or make large campaign contributions, but those could be strictly limited separately.
Thanks for writing this, very interesting piece. I think your points about the lack of robustness, goodhearting and thin markets are very true. But I’m more skeptical on some other issues:
This is true, but we absolutely should have an expectation that wealthy voters will have more accurate beliefs! Wealth is positively related to both education and IQ, both of which are linked to having more accurate views. Additionally, wealth plausibly means you are less stressed, which can improve cognitive performance, and gives you access to more information.
I also think you under-state the extent to which current policy is determined by elites under the status quo. My impression is that currently the views of the lower classes have very little impact on policy—see for example Affluence and Influence by Gilens for evidence that US Presidents tend to always massively favour the views of the upper classes over the lower classes (even for George Bush, who had the smallest such bias).
One mechanism by which this occurs seems to be the status of different policies. Politicians want to be high status, and to associate with high status views that are supported by their high status friends, newspapers, academics etc. This is significantly reduced with markets—hedge fund managers in general are perfectly happy to own Walmart’s stock—and even more so if bets are anonymous. I’ve never encountered any credible argument that (wealthy) hedge fund guys have trouble pricing Discount Retailers or Tobacco stocks just because their customers are from a different social class.
If the ‘implosion’ is really that bad, then the long-term futures would have a very low fair price, and you would be able to get a decent return on them. It’s true that if the ‘implosion’ was only slightly worse than the short-term benefit, then time value of money might mean that the long term futures were not sufficiently cheap. But then the issue seems to be that your preferred discount rate (0) is different from market interest rates; that’s a common issue that’s not specific to futarchy. So long as the futures market is properly constructed (e.g. daily variation margin with a well-capitalised clearinghouse) the discount rate applied should be essentially the risk free rate.
The design of prediction markets (and futarchy specifically) spend a lot of time discussing private interest alignment, which makes it very salient. In contrast, defenders of democracy spend very little time discussing how well aligned the private interests are of the President with the country, because this alignment is clearly very poor. But this potentially creates a misleading impression, conflating compatability with reliance: market can deal with self-interest in a way that politics cannot, but I’m not sure why we would jump from that to assuming they need it. It seems possible for someone to trade out of altruistic motives? Certainly the high participation we see in some prediction markets with either no or very small amounts of money at stake suggest this.
It might be a good idea to link to some of the papers Hanson has on the subject, e.g. here. More substantially, I think your response proves too much. Yes, if no-one can tell a market is mis-priced and no-one can tell there is manipulation, speculators will not be able to correct the market price. But if that is the case then we clearly have a situation of massive ignorance about the policy, in which case I wouldn’t expect democracy or any other system to do well either.
There’s a big difference between currencies (indefinite life) and futures. A cryptocurrency can maintain ‘irrationally’ high valuation for an indefinite period, in the same way that a piece of art can maintain a high value for a long time, even though you can just print out a near perfect copy. But a future has a termination at a specific point in time: it has to converge to the correct price by then, one way or another. This dramatically reduces incentive to follow along with a large irrational player: it is a finitely iterated game with a known end point. AMC’s shares can stay high indefinitely, but a future set to the value of their 2021 revenue has to come down to earth within a short period, so would probably never have spiked in the first place.
Finally a very pedantic comment:
Actually this would not increase GDP, because transfers are not counted, only consumption, investment etc. But there are closely related failure modes—e.g. the GDP benefit of women entering the workforce is overstated because previously they were doing useful work that was simply unwaged and hence excluded—so your underlying concern is correct.
Is this a good analogy for prediction markets, though? The stock prices don’t feed directly into policy decisions about taxes and regulations, but prediction market prices would under futarchy (unless you limit their scope to exclude specific policies), from my understanding. Betters have a better path to influence policy in their favour through these prediction markets than through stock markets. I don’t know how large these effects would be, but this is discussed in the section “(5) Buying policies: manipulation through extreme wealth”.
Of course, in many countries they can directly lobby governments or make large campaign contributions, but those could be strictly limited separately.