Hacker News Reader

Comment by pg_bot | Hacker News Reader

It's not a straw man to call this a prisoner's dilemma. The bank went from well respected institution to receivership in approximately 24 hours. 42 billion dollars worth of deposits was taken out of the bank. In the amount of time it would take to coordinate everyone (assuming they would even cooperate), someone else would've beaten you to the punch and removed their money. Every additional person that has made that decision puts you at risk of holding the bag. I do not fault any of the VCs who requested that their portfolio companies pull their money out. They saw the writing on the wall and were trying to protect themselves, the companies they invested in, and their LPs.

View on HN

It was a modified prisoners dilemma where 100 prisoners are faced with steal or share, and have every possible chance to communicate but just can change their choice after making it. And if just 20 had shared, then everyone wins, even those who chose steal. But the panic and the irresponsible VCs lack of communication with the bank and each other meant that in a heartbeat 85 prisoners had chosen steal, before even looking up at each other or listening to the rules being fully explained, and the remaining 15 where left perplexed that there wasn't even 5 amongst the panicked bunch left to help resolve the situation calmly to everyones benefit.

You can allow as many phone calls as you want but with thousands of players, you're going to get defect every single time you run the experiment.

There's no "steal" here. It's your money, at a bank, and if you're late on the defect button you might not get it back. The rational action is always to defect.

> The rational action is always to defect.

If that was true, then every bank every day would see runs on their holdings. Obviously that doesn't happen, because most traditional banks have larger institutional customers who understand that the market-to-market value on bonds might dip a bit, but that as long as the bank is able to negotiation loans or raise capital to provide needed liquidity then it's not a big issue. You don't see runs on traditional banks like this, because in the traditional banking sector people pick up the phone and talk to each other instead of taking to twitter and screaming doom and destruction because they read a blog post that they didn't understand but which said something about market-to-market insolvent and sounded scary.

Oh, you sweet summer child.

You don't see runs on banks because people believe their money is safe. Full stop. As soon as that belief is broken, a run is guaranteed. This is why we have deposit insurance, and why the fed just effectively made it unlimited. If they hadn't, every small bank in the country would now be experiencing a run, including those in "the traditional banking sector".

What changed late last week is not any fundamental of the economy or banking sector, but people's belief in the safety of banks.

In your assertion here are you assuming that the bank was solvent and just had a liquidity problem, not a solvency problem?

At the root of the issue was that they where not solvent when bonds where valued at market-to-market, but you don't really care about that when you are intending to hold the bonds to maturity. They where solvent when bonds where valued at maturation rates. The latter valuation is what you traditionally calculate and use in banking, the prior only becomes relevant if there is a run on the bank and you have to liquidate immediately to cover, which is of cause what happened.

The great comedy of the situation is that if everyone had been calm and said "hmm, they need to raise some capital lets see what happens" SVB might have had a bad quarter or two, but they wouldn't have gone bankrupt. They only went down because the startup community reaction to hearing "we need to raise capital to cover day to day liquidity requirements" was to initiate an immediate flash run on the bank, which meant they where forced to sell all those bonds at a discounted market-to-market rate instead of holding them to maturity as was the expectation.

>> but you don't really care about that when you are intending to hold the bonds to maturity. They where solvent when bonds where valued at maturation rates

Actually if you're a bank whose business model is predicated on interest spread between your assets and liabilities you do care. This is particularly true for banks like SVB whose duration of liabilities (deposits) is short because suddenly cost of deposits rises while longer-duration 'held to maturity' assets still generate the same (low) rate of interest. Suddenly the bank is loss-making.

People are talking about mark to market like it's irrelevant but it's actually a very important market signal. It is not just relevant for balance sheet valuation. It is also a signal of future income statement profitability unless duration of assets is well matched with duration of liabilities.

Remember that we know a lot more about what was happening now than they did at the time. At the time they knew that 1. SVB had liquidity issues and were forced to sell underwater bonds. 2. Needed to raise more capital. 3. Transactions were starting to have issues going through.

So it’s like your modified prisoners dilemma, but also, the prisoners don’t know if there is enough for 1 person or 100 people. That changes the calculation.

Prisoner's Dilemma:

Two members, A and B, are being driven around Silicon Valley and possesses every modern means of communication with each other. The bank holding their funds will collapse, but only if both of them send a globally accessible message advising everyone in the world to remove their funds from the bank.

I've run some simulations in Python-- the best I can get is for one of the members to learn to refrain from resending the first member's global message 32% of the time. (But be careful with this-- once B accidentally escaped the simulation and shot out a tweet that caused a small ruckus for an investment bank in Omaha.)

You’re assuming homogeneous agents. In other words, no leaders; no whales. The real world doesn’t work like this.

It was a mismanaged crisis, I think on that point we agree.

In theory a restaurant that is over capacity and catches on fire should be able to have everyone exit without death or injury. In practice, people panic and act in their own self interest leading to disaster.

Except often enough in such a situation everyone does exit orderly. People are generally predisposed to cooperate; it's what makes us human. The situations where panic prevails are the exception.

What's the saying... success has many fathers, but failure is an orphan?

> everyone does exit orderly.

this only happens when there's leadership present to which the rest get directed, and the crowd trusts the leadership to act with their interest in mind. E.g., a uniform wearing fire-warden, is often enough.

If there's no such leadership, or they are acting (clearly) not in the interest of the "crowd", you will see panic and a failure to cooperate. This, not cooperation, is what makes humans truly humans.

You also can't yell fire in a crowded theater when there isn't one. You could make the argument that someone did.

You totally can do that, it's not illegal! Schenck v. United States is a pretty odious court case I wouldn't want to be on the censor's side in that one.

https://www.theatlantic.com/national/archive/2012/11/its-tim...

No you could not. LegalEagle did a video on this in the last week talking about that ruling that was overturned and doesn't apply to the vast majority of case law anyway.

Reply on news.ycombinator.com