Hacker News Reader

HTM accounting isn't the problem, it's one tool among many and has its uses. SVB's mistake was buying long term paper just as the country was heading into a rising rate environment. This has already put them in a hole they probably can never dig themselves out of (selling stock is clearly not gonna work). I think the only non-disastrous path forward here is an acquisition by another bank big enough to just absorb that paper into its balance sheet. If SVB is smart they'll try to get this done while there's still some value left to sell--if they wait too long it'll turn into another WaMu.

Borrowing short and lending long has been the traditional function of banks for hundreds of years. They profit from the interest rate spread between what they pay on short-term liabilities (mainly deposits) and what they receive on long-term assets (usually business or housing loans).

It's hilarious that banks serving crypto and startup ecosystems aren't failing because their crypto and startup loans went bad, they're failing because of the duration risk from holding long-term Treasurys on their balance sheet.

>they're failing because of the duration risk from holding long-term Treasurys on their balance sheet.

That is the number one reason why banks go bankrupt...

It's not hilarious, duration risk is the biggest risk a bank has to deal with and it gets worse the more and more people keep their money as demand deposits. This is why the Fed does QE, the duration of deposits has shrunk so low, that the banks can't buy treasuries as the duration of the treasuries has become too long in relation to instantaneous demand deposits that can switch from bank to bank. The only solution that the Fed came up with, is to do the duration transformation themselves by buying long duration treasuries and giving instantly transferrable central bank reserves with no duration risk. This works because the system as a whole cannot go bankrupt, but individual banks can. If you take your money out of the bank, the CB reserves just get turned into cash. If you transfer between banks, the recipient bank now holds the reserves.

Also, plenty of crypto specific banks did fail because their loans went bad.

There’s duration in the sense of borrowing short and lending long, and there’s duration in the sense of taking on the corresponding interest rate risk. The US model of fixed-rate mortgages is a bit of an oddity — banks will often lend long with adjustable-rate instruments.

If a bank has $100 of deposits and has lent out $90 on a 30-year adjustable rate basis, the bank is exposed to potentially liquidity issues, default risk, and second-order effects like changes in the yield curve. But if the bank lends that $90 as a fixed rate mortgage or buys a 30-year T bond, they are directly exposed to interest rate changes.

(Also, as I understand it, and this is far outside my expertise: banks usually try repackage their debt and sell it to investors. The banks presumably can’t carry those 30 year fixed mortgages on their books without accounting for interest rate risk. Why are T bonds special?)

> If SVB is sitting on a pile of Treasury bonds that mature in 20 years, they can “hold to maturity” and get their principal plus some very low interest rate. But this is useless!

No it isn't. Any other bank would be happy to write a loan backed by US treasury holdings, at no more than a moderate profit. If you're sitting on that, it has value. But it doesn't have value in literal dollars by tomorrow morning to pay out a withdrawal. That's what "liquidity" is all about.

I’m not saying the T bond is useless. I’m saying that the fact that it’s nominally worth a specific amount in the future if I hold it is useless.

This has nothing to do with liquidity. If I had a 0 interest, $100 T bond maturing in 30 years, I cannot sell it today for $100. But anyone who lent me $90, nonrecourse, using it as collateral and asking for only a moderate profit is nuts because this bond is not worth $90 — not even close. Maybe I can get a loan that is based on my own credit-worthiness, but that’s a different story entirely.

If I have this $100 T bond, and I’m a bank, and that T bond is collateral for a $100 savings account paying 4.5% APY, I am in the hole. If my depositor sticks around, I can gamble and hope I can make up my losses (e.g. by interest rates going way down), or I can try to be such an awesome bank going forward that my profits can make up for my losses, and maybe I’ll get away with it, but I don’t really deserve to get away with it.

> this bond is not worth $90

Not to a retail investor looking at short term returns (and irrationally obsessed with Inflation! due to media consumption), but to a bank with regulatory deposit requirements and a longer term outlook? Seems not unreasonable.

Anyway it doesn't have to be worth the full face value. It just needs to be worth enough to back a short term loan big enough to honor current withdrawal demands.

> regulatory deposit requirements

I wonder if this is the problem.

> to a bank with regulatory deposit requirements and a longer term outlook?

Maybe with some generally accepted accounting principle, but not by any sensible business standard.

If that bond trades for $80, no one would buy it for $90. It’s worth $80. Similarly, if you already own it, it’s not magically worth more.

And if you are a bank with a long term outlook, you expect interest rates to hold near current levels, and you pay 4.5% APY to depositors, your long term outlook of paying 4.5% to deposits where that deposit money is locked up in a very safe bond earning 1.8% APY, you are losing a lot of money, very safely, in the long term. Almost exactly as much as you would lose by booking the loss right away and investing in something else.

Other than tax or regulatory arbitrage, complex accounting is no substitute for actual profits and losses :)

You’re confusing liquidity with solvency.

For most banks, it is fine to assume a 20 year deposit window because deposits are fungible and for most of recent history deposit bases have gone up.

SVB was wrong for not assuming that the 2021 deposit spike was (in hindsight obviously) a short term blip, but you can look at their loan book on page 19 of [0].

It's not immediately clear to me that there's some sort of systemic risk in VCs/PE firms not paying back their loans, but given how circular the tech ecosystem is, maybe we get there.

[0]https://s201.q4cdn.com/589201576/files/doc_financials/2022/q...

> For most banks, it is fine to assume a 20 year deposit window

Let me try again. You're still thinking about liquidity -- assuming a 20 year deposit window seems okay to me.

But the problem is solvency. It's not that the assets are illiquid -- it's that they insufficient. If you run a bank, assume a 20 year deposit window, and invest those deposits in safe assets that carry similar interest rate risk to the deposits themselves, then you are solvent. If depositors leave faster than expected, you may be forced to pay a spread or other haircut to sell your long term assets early, but that's a small effect and can be managed gradually as long as you stay on top of it.

This isn't SVB's problem AFAICT. SVB bought assets that were perfectly liquid (T bonds!) and (if I read the filings right) assets that are still fairly liquid (MBSes), but they had interest rate exposure that did not match the deposits. Savings account interest rates (at SVB!) were up to 4.5%, and those T bonds had much lower fixed interest rates. [0]

If depositors held their money at SVB for 20 years and SVB didn't have a bank run right now, SVB would still be in trouble: SVB would be paying more interest on those deposits than they would receive on their investments, and their portfolio would slowly go negative.

[0] Whether you think about them as paying low interest for a long time or as being marked to market at a loss and then paying current interest rates at maturity is immaterial. You end up with the same number of actual dollars at the same times.

>It's not that the assets are illiquid -- it's that they insufficient.

You are confusing price today with the actual characteristics of the instrument.

A T-bill pays back it's face value always, it just trades below face when interest rates are higher than when it was purchased.

That is the definition of a liquidity problem. You'll get the face value back eventually (it is sufficient), but not if you sell today.

A liquid market is not the only characteristic of liquidity. There's a liquid market for anything at a low enough price...

>If depositors held their money at SVB for 20 years and SVB didn't have a bank run right now, SVB would still be in trouble: SVB would be paying more interest on those deposits than they would receive on their investments, and their portfolio would slowly go negative.

This is a true statement people are making that has nothing to do with what happened here. The point of a bank is interest rate arbitrage, so I agree SVB would have to do their job better over those 20 years, but this thing that happened over 3 months is a liquidity crisis. They weren't running out of money to pay interest on deposits, they were running out of money to give those deposits back and were veering into the problem discussed above - selling things below value to create liquidity.

> In 20 years, 4.5% multiplies money by 2.4, those T-bonds will not multiply by 2.4, and SVB will slowly but surely end up in the hole.

But they wouldn't continue to pay 4.5% interest for 20 years would they? This is just a brief moment of high interest rates, so I think that's the main detractor from this argument. In other words, it's not like 4.5% APY on savings is anywhere close to the norm.

> This is just a brief moment of high interest rates

You are welcome to make that bet for your personal finances or on behalf of your company. You should not make that bet with someone else’s money unless they have signed up for it.