Thank you for your interest in my article! I appreciate it.
Your proposal, if I understand it correctly, is this: instead of market participants having to put real money in the system, and then being reimbursed if the bets are off (the naive randomization I and Hanson assumed), you propose that everyone can place unfunded orders at any time. If the bets are off, nothing happen, if they are on, everyone receives a bill for the full amount they ordered, all at once. Alternatively, everyone uses borrowed money to bet on margin, which is essentially the same thing.
You are correct that this does equalize the carry cost between the market manipulator insider and the arbitrageur outsider, which make it possible to profit off arbitrage that are small relative to 1/epsilon (but larger than the carry cost).
The main problem is that removing the requirement to secure the bets basically destroys the epistemic argument underpinning futarchy, which is that real money at play creates skin in the game forcing market participants to trade honestly. Under universal margin, no one has to trade real money to move this price. If the participants can default on their obligations to the platforms after the bets are executed, then skin in the game is largely fictional. If the platform requires some form of collateral, then capital requirements are back on the table.
If the bets are off, nothing happen, if they are on, everyone receives a bill for the full amount they ordered, all at once.
I had in mind something a bit different...
I was not imagining unfunded orders or universal margin, where traders only receive a bill if the randomization occurs. I agree that this would reintroduce default risk unless collateral is posted, in which case the capital-cost problem comes back.
What I had in mind is closer to “lottery collateral.” Suppose there is a lottery-collateral provider that sells claims which pay out $1 only in the randomization branch, say 0.5% of the time. Since each claim only pays in 0.5% of worlds, the fair price should be around $0.005 per $1 claim.
An arbitrageur could then use these claims as collateral for the conditional market without personally locking up the full $1. The collateral provider would also not face the usual insurance-style adverse selection problem, since the payout condition is not privately chosen by the buyer and does not depend on buyer-specific hidden risk. It is just an externally randomized event. The provider could also design the lottery claims to be complementary across many markets, so that different claims tend not to pay out in the same worlds. That would substantially reduce payout variance and reserve requirements, rather than requiring the issuer to hold $1 of idle capital against every $1 face-value claim. In principle, this seems like it could remove most/all of the capital lockup cost without relying on fictional skin in the game or exposing the platform to trader default.
This would directly target the original problem: in naive randomized futarchy, arbitrageurs have to immobilize capital in all worlds even though their trade only matters in the rare world where the decision is randomized. That makes small-but-real mispricings uneconomic to correct, so the market may fail precisely where the epistemic argument needs arbitrage to be cheap. Lottery collateral would instead make the arbitrageur’s capital cost scale with the probability that their collateral is actually needed. If that works, then randomization no longer creates a large asymmetric carry-cost advantage for insiders or manipulators. Outsiders could profitably correct much smaller welfare mispricings, while the system would still have real collateral behind executed bets in the worlds where execution actually occurs.
Thank you for your interest in my article! I appreciate it.
Your proposal, if I understand it correctly, is this: instead of market participants having to put real money in the system, and then being reimbursed if the bets are off (the naive randomization I and Hanson assumed), you propose that everyone can place unfunded orders at any time. If the bets are off, nothing happen, if they are on, everyone receives a bill for the full amount they ordered, all at once. Alternatively, everyone uses borrowed money to bet on margin, which is essentially the same thing.
You are correct that this does equalize the carry cost between the market manipulator insider and the arbitrageur outsider, which make it possible to profit off arbitrage that are small relative to 1/epsilon (but larger than the carry cost).
The main problem is that removing the requirement to secure the bets basically destroys the epistemic argument underpinning futarchy, which is that real money at play creates skin in the game forcing market participants to trade honestly. Under universal margin, no one has to trade real money to move this price. If the participants can default on their obligations to the platforms after the bets are executed, then skin in the game is largely fictional. If the platform requires some form of collateral, then capital requirements are back on the table.
I had in mind something a bit different...
I was not imagining unfunded orders or universal margin, where traders only receive a bill if the randomization occurs. I agree that this would reintroduce default risk unless collateral is posted, in which case the capital-cost problem comes back.
What I had in mind is closer to “lottery collateral.” Suppose there is a lottery-collateral provider that sells claims which pay out $1 only in the randomization branch, say 0.5% of the time. Since each claim only pays in 0.5% of worlds, the fair price should be around $0.005 per $1 claim.
An arbitrageur could then use these claims as collateral for the conditional market without personally locking up the full $1. The collateral provider would also not face the usual insurance-style adverse selection problem, since the payout condition is not privately chosen by the buyer and does not depend on buyer-specific hidden risk. It is just an externally randomized event. The provider could also design the lottery claims to be complementary across many markets, so that different claims tend not to pay out in the same worlds. That would substantially reduce payout variance and reserve requirements, rather than requiring the issuer to hold $1 of idle capital against every $1 face-value claim. In principle, this seems like it could remove most/all of the capital lockup cost without relying on fictional skin in the game or exposing the platform to trader default.
This would directly target the original problem: in naive randomized futarchy, arbitrageurs have to immobilize capital in all worlds even though their trade only matters in the rare world where the decision is randomized. That makes small-but-real mispricings uneconomic to correct, so the market may fail precisely where the epistemic argument needs arbitrage to be cheap. Lottery collateral would instead make the arbitrageur’s capital cost scale with the probability that their collateral is actually needed. If that works, then randomization no longer creates a large asymmetric carry-cost advantage for insiders or manipulators. Outsiders could profitably correct much smaller welfare mispricings, while the system would still have real collateral behind executed bets in the worlds where execution actually occurs.