Matt Levine is a finance writer with a very entertaining free newsletter, also available on Bloomberg to subscribers. Today's newsletter struck me as a fairly remarkable failure analysis of a very expensive failure, in which Credit Suisse lost $5.5billion dollars when the hedge fund Archegos collapsed. That doesn't usually happen, and banks are, of course, very incentivized to avoid it. When it happened, Credit Suisse commissioned a very thorough investigation into what went wrong.

Some background: Archegos was a hedge fund, founded in 2013, that defaulted spectacularly this spring. The Wall Street Journal estimates that they lost $8billion in 10 days. Levine wrote at the time:

The basic story of Archegos is that it extracted as much leverage as possible from a half dozen Wall Street banks to buy a concentrated portfolio of tech and media stocks (apparently partially hedged with short index positions[2]), and those stocks went up a lot, before going down a lot.

If you merely own some stocks, and they go way up and then way down, you'll end with approximately the money you started with and everything will be fine. But if you have taken out loans to buy stocks, then when they go up your wealth has increased. And if you then use your increased wealth to borrow lots more money and buy more stocks, then when they go down you will lose $8billion in ten days.

None of this is unknown to bankers, so it's confusing that the bankers let Archegos do this. In the immediate aftermath, there was a lot of theorizing about how the banks might have had inaccurate or incomplete information about how heavily leveraged Archegos was. Levine:

When the Archegos story came out this spring, there was a sense, from the outside, that the banks had missed something, that there was some structural component of Archegos’s trades that caused the banks to underestimate the risks they were taking. For instance, there was a widespread theory that, because Archegos did most of its trades in the form of total return swaps (rather than owning stocks directly), it didn’t have to disclose its positions publicly, and because it did those swaps with multiple banks, none of the banks knew how big and concentrated Archegos’s total positions were, so they didn’t know how bad it would be if Archegos defaulted.

But, nope, absolutely not, Credit Suisse was entirely plugged in to Archegos’s strategy and how much trading it was doing with other banks, and focused clearly on this risk.

So what went wrong? According to the report Credit Suisse commissioned from a law firm on the whole mess, what went wrong is that Credit Suisse determined that Archegos was overleveraged, and that they needed more collateral, and they called Archegos to that effect, and Archegos responded "hey sorry I've been swamped this week, can we talk later?" and that was that.

No, really, that's pretty much it.

The report:

On February 23, 2021, the PSR analyst covering Archegos reached out to Archegos’s Accounting Manager and asked to speak about dynamic margining. Archegos’s Accounting Manager said he would not have time that day, but could speak the next day. The following day, he again put off the discussion, but agreed to review the proposed framework, which PSR sent over that day. Archegos did not respond to the proposal and, a week-and-a-half later, on March 4, 2021, the PSR analyst followed up to ask whether Archegos “had any thoughts on the proposal.” His contact at Archegos said he “hadn’t had a chance to take a look yet,” but was hoping to look “today or tomorrow.”

Of course, when your counterparty is refusing to give you more collateral, you can pull all their loans. But Credit Suisse was kind of reluctant to pull that lever given that it wasn't like Archegos was bankrupt yet, or refusing to pay, they were just having a super busy week and kept apologetically cancelling calls at the last minute. So they just kept rescheduling the calls and sending concerned emails internally.

Levine: 

The report is … look, tastes will vary, and I concede that I am a weird guy, but it is so, so good? It is thrilling reading, as good as anything you will ever read about the management and sociology and processes of big investment banks. And that, it turns out, is the story. There is not some fancy finance thing that went wrong, some clever trick that Credit Suisse missed, some interesting problem in the math or the legal regime that caused Credit Suisse to lose so much money. Nor is this a story of individual stupidity or greed or recklessness; people generally had the right facts and were trying to do the right thing and kept each other in the loop. It’s just that they sort of kept each other in the loop as a substitute for actually doing anything. The processes were all moving along nicely, which gave everyone a false sense of security that they would produce the right result. Unfortunately the processes were slow and Archegos blew up quickly.

and 

This report is not a bunch of lawyers identifying a bunch of problems and characterizing them, in hindsight, as “red flags.” Everyone saw all the problems here, evaluated them reasonably, came up with sensible solutions and then didn’t do them. A ghostly recurring character in the report is the CPOC, the Counterparty Oversight Committee, a fancy committee that Credit Suisse started after losing a bunch of money on swaps with another hedge fund, Malachite Capital Management, in 2020.[3] CPOC grew out of “‘Project Copper,’ an initiative to ‘improve [CS’s] ability to identify early warning signs of a default event,’ and ‘enhance [CS’s] controls and escalation framework across functions during periods of stress’” (page 15). And at the very first meeting of CPOC, in September 2020, it discussed Archegos, identified the key problems, agreed on good solutions, and then forgot about them forever:

I am for a lot of reasons very interested in how big organizations do risk management. Banks in particular are interesting because they have pretty clear incentives: they make more money if they're better at risk management, and when they fail it's very conspicuous. People have relatively tight feedback loops on their decisions. The stakes are high enough that lots of smart people are thinking full-time about how to do this well, but low enough that a meaningful retrospective is much easier to come by than for, say, the Covid pandemic.

 I could imagine the above failure story for how an unsafe project of much greater concern made it to release despite smart people all agreeing that they had a lot of reservations that needed to be resolved first, and despite lots of emails that accurately characterize the exact risk and the steps that should be taken about it. Appreciation of the high stakes does motivate smart people to do something, but it might not motivate them to do anything confrontational enough to actually solve the problem.