Quote: (and clearly they calculated incorrectly if they did)
I am less confident that, if an amoral person applied cost-benefit analysis properly here, it would lead to “no fraud” as opposed to “safer amounts of fraud.” The risk of getting busted from less extreme or less risky fraud would seem considerably less.
Hypothetically, say SBF misused customer funds to buy stocks and bonds, and limited the amount he misused to 40 percent of customer assets. He’d need a catastrophic stock/bond market crash, plus almost all depositors wanting out, to be unable to honor withdrawals. I guess there is still the risk of a leak.
I don’t think we disagree much if any here—I think pointing out that cost-benefit analysis doesn’t necessarily lead to the “no fraud” result underscores the critical importance of side constraints!
He’d need a catastrophic stock/bond market crash, plus almost all depositors wanting out, to be unable to honor withdrawals.
I think this significantly under-estimates the likelihood of “bank run”-type scenarios. It is not uncommon for financial institutions with backing for a substantial fraction of their deposits to still get run out due a simple loss of confidence snowballing.
Could you say more about that? I suggest that “substantial fraction” may mean something quite different in the context of a bank than here. In the scenario I described, the hypothetical exchange would need to see 80-90% of deposits demanded back in a world where the stocks/bonds had to be sold at a 25-50% loss. It could be higher if the exchange had come up with an opt-in lending program that provided adequate cover for not returning (say) 10-15% of the customers’ funds on demand.
I’d also suggest that the “simple loss of confidence snowballing” in modern bank runs is often justified based on publicly-known (or discernable) information. I don’t think it was a secret that SVB had bought a bunch of long-term Treasuries that sank in value as interest rates increased, and thus that it did not have the asset value to honor 100% of withdrawals. It wasn’t a secret in ~2008 that banks’ ability to honor 100% withdrawals was based on highly overstated values for mortgage-backed securities.
In contrast, as long as the secret stock/bond purchases remained unknown to outsiders, a massive demand for deposits back would have to occur in the absence of that kind of information. Unlike the traditional banking sector, other places to hold crypto carry risks as well—even self-custody, which poses risks from hacking, hardware failure, forgetting information, etc. So people aren’t going to withdraw unless, at a minimum, convinced that they had a safer place to hold their assets.
Finally, in conducting the cost/benefit analysis, the hypothetical SBF would consider that the potential failure mode only existed in scenarios where 80-90%+ of deposits had been demanded back. Conditional on that having happened, the exchange’s value would likely be largely lost anyway. So the difference in those scenarios would be between ~0 and the negative effects of a smaller-scale fraud. If the hypothetical SBF thought the 80-90%+ scenario was pretty unlikely . . . .
(Again, all of this does not include the risk of the fraud leaking out or being discovered.)
Okay yes, I agree that a driver of bank runs is the knowledge that the bank usually can’t cover all deposits, by design. So as long as you keep that fact secret you’re much less likely to face a run.
I am now unsure how to reason about the likelihood of a run-like scenario in this case.
Quote: (and clearly they calculated incorrectly if they did)
I am less confident that, if an amoral person applied cost-benefit analysis properly here, it would lead to “no fraud” as opposed to “safer amounts of fraud.” The risk of getting busted from less extreme or less risky fraud would seem considerably less.
Hypothetically, say SBF misused customer funds to buy stocks and bonds, and limited the amount he misused to 40 percent of customer assets. He’d need a catastrophic stock/bond market crash, plus almost all depositors wanting out, to be unable to honor withdrawals. I guess there is still the risk of a leak.
I don’t think we disagree much if any here—I think pointing out that cost-benefit analysis doesn’t necessarily lead to the “no fraud” result underscores the critical importance of side constraints!
I think this significantly under-estimates the likelihood of “bank run”-type scenarios. It is not uncommon for financial institutions with backing for a substantial fraction of their deposits to still get run out due a simple loss of confidence snowballing.
Could you say more about that? I suggest that “substantial fraction” may mean something quite different in the context of a bank than here. In the scenario I described, the hypothetical exchange would need to see 80-90% of deposits demanded back in a world where the stocks/bonds had to be sold at a 25-50% loss. It could be higher if the exchange had come up with an opt-in lending program that provided adequate cover for not returning (say) 10-15% of the customers’ funds on demand.
I’d also suggest that the “simple loss of confidence snowballing” in modern bank runs is often justified based on publicly-known (or discernable) information. I don’t think it was a secret that SVB had bought a bunch of long-term Treasuries that sank in value as interest rates increased, and thus that it did not have the asset value to honor 100% of withdrawals. It wasn’t a secret in ~2008 that banks’ ability to honor 100% withdrawals was based on highly overstated values for mortgage-backed securities.
In contrast, as long as the secret stock/bond purchases remained unknown to outsiders, a massive demand for deposits back would have to occur in the absence of that kind of information. Unlike the traditional banking sector, other places to hold crypto carry risks as well—even self-custody, which poses risks from hacking, hardware failure, forgetting information, etc. So people aren’t going to withdraw unless, at a minimum, convinced that they had a safer place to hold their assets.
Finally, in conducting the cost/benefit analysis, the hypothetical SBF would consider that the potential failure mode only existed in scenarios where 80-90%+ of deposits had been demanded back. Conditional on that having happened, the exchange’s value would likely be largely lost anyway. So the difference in those scenarios would be between ~0 and the negative effects of a smaller-scale fraud. If the hypothetical SBF thought the 80-90%+ scenario was pretty unlikely . . . .
(Again, all of this does not include the risk of the fraud leaking out or being discovered.)
Okay yes, I agree that a driver of bank runs is the knowledge that the bank usually can’t cover all deposits, by design. So as long as you keep that fact secret you’re much less likely to face a run.
I am now unsure how to reason about the likelihood of a run-like scenario in this case.