Well, Silicon Valley Bank has been on headlines for several days now.
The way a bank works is that it borrows short to lend long. Simplistically, a bank might get its money from demand deposits, checking and savings accounts that customers can withdraw at any time. And the bank might pay, say, 0% interest on those deposits. And then it invests the money in some longer-term assets, loans and bonds that don’t get paid back for years, and that pay, say, 2% interest. The bank earns 2% on its money, pays 0% to depositors for the money, and keeps the spread, the net interest margin, which is 2% in this example.
Sometimes interest rates go up or down, though. Simplistically, short-term interest rates in the US are set by the Federal Reserve, which will raise interest rates to cool the economy if inflation is too high.
This is a risk for a bank’s borrow-short-to-lend-long business model. If the Fed suddenly raises short-term interest rates to, say, 3%, then you have to start paying 3% on your deposits. Meanwhile long-term interest rates have probably also gone up to, say, 5%, but you are still earning the old 2% on your loans and bonds, because they are long-term loans that don’t get paid back for years. Your net interest margin is now negative 1%: You pay 3% on deposits and earn only 2% on loans and bonds.
This is an obvious problem and much of the business of banking is about managing it. Here are some simple things that you can do:
- Your loans and bonds have laddered maturities: Some mature next week, some in 3 years, some in 6. So as rates go up, your loans and bonds are constantly rolling off and you are reinvesting the money in new loans and bonds at higher rates.
- When the Fed raises rates, in actual fact, banks mostly don’t pass along the whole rate increase to depositors. It’s a pain to switch banks, and most depositors don’t pay that much attention to rates. The “deposit beta” — the sensitivity of bank deposit rate increases to Fed interest rate increases — is lower than 1. So if the Fed raises rates from 0% to 3%, you might end up paying only, say, 0.5% on deposits.
- You can try to match your assets and liabilities somewhat. Make some short-term loans, and some floating-rate loans, so that your assets are more sensitive to rising rates. Borrow some money with long-term bonds, so that your liabilities are less sensitive to rising rates. Do interest-rate swaps to hedge.
- You can make money in ways other than net interest margin: Charge overdraft fees or do investment banking or trade securities or whatever.
This is not, however, the biggest problem with borrowing short to lend long. The biggest problem is that your depositors might all ask for their money back tomorrow, and you might not be able to get the money back from your borrowers for years. This problem is known as a “bank run” and is famous from, you know, the Diamond-Dybvig model that won a Nobel Prize last year, or from It’s a Wonderful Life. “The money’s not here. Why, your money’s in Joe’s house,” etc.
Much of the business of banking, and of bank regulation, is about managing this risk. For instance:
- The government can provide deposit insurance on accounts, promising depositors that they’ll get their money back, which will make them less inclined to ask for their money back, since they know it’s safe.
- The government can regulate banks to make them act safely, so that they don’t lose customer money, so that customers are confident and don’t ask for their money back.
- You can keep a lot of cash around, so that if a lot of depositors (though certainly not all of them) ask for their money back, you have some to give to them.
- If a lot of depositors ask for their money back and you run through your cash, you can sell some of your loans or bonds to raise money. In the days of It’s a Wonderful Life it was kind of hard to sell the mortgage on Joe’s house for ready cash, but modern financial markets are much deeper and more liquid, which is good for selling bonds or loans or other bank assets.
- Alternatively, you can borrow against your assets: You can go to a bigger bank, or to a lender of last resort like the Federal Reserve, or a “ lender of next-to-last resort” like the Federal Home Loan Banks, and borrow the cash to pay out depositors. And as collateral for that loan to you, you post your assets, the bonds you own or the loans you made to your customers. This is the most classic way to deal with bank runs, and is famous from Walter Bagehot’s Lombard Street. “Lend freely, at a penalty rate, against good collateral” is the basic advice Bagehot gives central bankers: If a bank has good loans, but it can’t turn them into money right away, the central bank should lend it the money.
Let’s talk about Option 4, selling assets to raise money to pay back depositors. Let’s say you loaned a customer $100 for five years at 2% interest. The customer is big and reliable and the loan will definitely be paid back. How much can you sell that loan for? Well, the day after you made the loan, it should be worth about $100 — about what you paid for it. But what if the Fed raises short-term interest rates from 0% to 3%? Then the market interest rate on loans like this will probably go up from 2% to 5%. But your loan still pays only 2%, so it is worth less. The crude intuitive math is that a new market-rate loan would pay $5 of interest per year for 5 years ($25 total), while your loan pays $2 of interest for 5 years ($10 total), so it is worth about $15 less, so people would pay you about $85 for it, though you’d get fired at a bank for doing bond math like that. (Really they’d pay you about $87.)
There is a connection here to the previous problem, the one about your net interest income going down as rates go up. Imagine a super-simple bank: You borrow $100 at 0% and lend $100 at 2% for five years. You also invest $10 of your own capital in the bank, and keep that money in cash, to cover any sudden withdrawals. You make $2 per year in interest, pay $0 per year of interest, and get $2 per year of net interest margin. You spend $1.50 of that on non-interest expense (salaries, rent, etc.), and keep $0.50 as profit. After a year, you have $10.50 of capital: the $10 you put in and your $0.50 of profit.
Then rates suddenly go up, so short-term rates are now 3% and long-term rates are 5%. None of the stuff I said before — about deposit betas and laddered maturities — works for you, for whatever reason; your deposit rates immediately jump to 3% and you keep getting 2% on your bonds. This year, you will have a net loss of $2.50: You pay $3 on your $100 of deposits and get paid $2 on your $100 of bonds, and you also have $1.50 of salaries and rent to pay. (You’ll be down to $8 of your own capital.) Next year, you will also have a net loss, though maybe it’s only $2 because you laid some people off and closed some branches. (Now you’re down to $6 of capital.) Year after, same idea: You will keep bleeding money, because you borrowed short and loaned long and rates went up. But maybe something will change? Maybe rates will go back down. Maybe your borrowers will repay their bonds early and you’ll get to redeploy the money at higher rates. Maybe your employees will find a new line of business to earn some fees. Maybe you’ll reverse the losses. Or not, maybe you’ll keep losing $2.50 per year for five years and run out of capital and have to close the bank. But maybe my assumptions about your high deposit beta and long-term assets are unrealistically pessimistic: This is a simplistic example, and at a real bank the result might be lower margins for a while, not constant losses.
But what if instead of keeping their deposits at your bank and earning 3% interest, the depositors all asked for their money back today? You would have to sell your bonds. You’d get about $87 for them. You would have a $13 loss today. All of the losses that you would have taken over the next five years, from holding long-term bonds at low rates while deposit rates have gone up, you will experience now. And you won’t have any time to reverse them, because you have to sell the bonds now and give everyone their money back. Also you don’t have enough money: Selling the bonds for $87 will raise $87, and you have $10 of capital, for a total of $97. But your depositors want their $100 back, and you only have $97.
If you are a bank that borrowed short to lend long, and rates go up, your portfolio of long-dated assets represents an opportunity cost over time: Instead of earning 5% on your loans, you’re stuck earning 2%. But if you have to sell those assets, you turn that opportunity cost into a cash loss today. You pull the losses forward in time and take them all now.
Now we have to talk about accounting. Oversimplifying, there are two main ways for a bank to account for its assets:
- The traditional way is to account for them at cost, the price you paid for them. If you lend a customer $100, you have a $100 asset — the loan — on your balance sheet. (You deduct something for the possibility they won’t pay you back, but let’s ignore that.)
- The other way is to account for them at their fair market value. If you buy a bond for $100, you can go look at the market price for that bond at the end of the quarter; if it’s $97.75, then you have a $97.75 asset on your balance sheet. The other $2.25 is gone.
The first approach gives you a reasonable picture of a bank that is continuing to operate as a bank: In my example, if you have $100 of bonds and $100 of deposits and the loans pay 2% and the deposits cost 3% and you are losing $2.50 per year, your financial statements will say that. Your balance sheet will say that you have $100 of bonds, so you will ultimately have enough money to pay back your depositors, but your income statement will say you are losing money each year.
The second approach gives you a reasonably accurate picture of a bank that has to shut down today: In my example, your balance sheet will show that you have $87 of bonds at fair market value, so you don’t have enough money to pay back all your depositors if you have to do it today.
Traditionally banks’ assets consisted mostly of loans, there was not much of a market for those loans, and the normal accounting approach was to account for them at cost. Modern banks’ assets include lots of bonds, there is a market for those bonds, and it is reasonably common to account for them at fair market value. Still, banks do like accounting for assets at cost, and so they are allowed to designate bonds as “held to maturity” — meaning that they have no plans to sell them — and account for them at cost. (But then if they sell any of their held-to-maturity bonds, they need to reclassify all of them as “available for sale” and account for them at fair market value, which could mean taking a huge loss all at once.) In general it seems fair to say that the modern trend in financial-institution accounting is toward marking more things to market values, but there remain many exceptions.
From my simple example you can see why banks don’t always like mark-to-market accounting. In my simple example, the bank has $100 of deposits, $100 of bonds and $10 of capital. It is losing money on interest rates, so its capital is going down, slowly. But it is solvent: It has assets of $110 ($100 of bonds plus $10 of cash) and liabilities of only $100 (the deposits).
But if the bank instead put out financial statements showing that it had $87 of bonds (their current market value), it would be insolvent: It would have assets of $97 against liabilities of $100. And if its depositors read those financial statements, they’d say “hey wait this bank is insolvent” and rush to take their money out. And then the bank would have to sell those bonds for $87, and it really would be insolvent. The accounting could cause a run, which would require the assets to be sold at market value, which would result in taking all the losses today, which would result in actual insolvency.
One crude way that the actual US banking system deals with this problem is by letting banks account for their held-to-maturity bonds at cost on their balance sheet, but then making them explain, somewhere in the footnotes to the financial statements, their current market value. This feels like in some ways a fair compromise. On the other hand someone might read the footnotes! I mean, not most people. But someone might, and they might explain the footnotes in a readable way — they might say something like “on a mark-to-market basis, they were broke last quarter” — and people might read that and think “huh that’s bad” and withdraw their deposits. Causing the bonds to be sold, causing insolvency.
But wait! Why sell the bonds? Why not borrow against them? You have $100 of bonds, $10 of capital kept in cash, $100 of deposits. Rates move against you. In a footnote to your financial statements, you confess that the bonds are worth only $87 at current market values. Someone reads the footnote and shows it to someone else. Word gets around. Lots of depositors come and ask for their money back. You have only $10 of cash.
So you go to your lender of last resort, let’s say the Federal Reserve. You say: “Here is a $100 asset that I have. I want to post it as collateral for a loan. You lend me $100 against this collateral, and charge me 3% interest. I will use the $100 to pay off my depositors. I’ll be left with a $100 asset and a $100 liability to you, instead of to my depositors. And then I’ll try to cut costs and earn more fees and otherwise earn enough to pay your interest. I’ll be in the same place as I was before the run. It’s not a great place, necessarily — I’m still earning less interest on my assets than I’m paying on my liabilities — but, you know, markets change and I’ll probably muddle through and anyway the bonds will eventually pay off $100 so you’ll get your money back.”
And then the Fed either says:
- sure, or
- “no, this is not $100 of collateral, this is only $87 of collateral: We will only lend against the market value of this collateral, not against what you paid for it.”
Which answer should the Fed give? On the one hand, the Fed wants to lend against good collateral and not take a lot of risk. Lending you $100 against collateral with a market value of $87 seems kind of stupid for the Fed.
On the other hand, the Fed is a lender of last resort, and its main job is to lend you money so that you don’t have to sell your assets too hastily. Lending against the market value is kind of the same thing as making you sell your assets: You can only borrow what you’d get by selling them today, not what they’ll eventually pay out. It speeds things up just like selling them would, and the Fed’s job is to give you more time.
You could imagine giving either answer. In the olden days of traditional banking, it might have made sense for a central bank to lend against banks’ loans at their face value. But modern central bankers are plugged into financial markets, and financial markets are good at determining market prices for assets, and so the modern consensus seems to be that central banks should lend against collateral at its market value.
Until yesterday. We talked on Friday about the collapse of Silicon Valley Bank:
- It borrowed very short to lend very long. Specifically, it is funded mostly by deposits, largely from tech companies and venture capitalists who got a lot of money over the last couple of years, and who put that money into SVB deposits that were much bigger than the US deposit insurance cap of $250,000. It plowed that money disproportionately into bonds, quite safe bonds — US Treasuries and agency mortgage-backed securities — but bonds with long duration. It did not do much to hedge its interest-rate risk. Basically it was as reckless as it is possible to be with a business model of “take deposits and invest them in US Treasury bonds.” Which, until recently, might not have seemed that reckless!
- But then rates went up a lot, pretty fast.
- This caused the market value of SVB’s bonds to decline by some $15 billion, to the point that it was more or less insolvent: Its losses on the bonds were enough to wipe out almost all of its equity capital and leave it with assets, at market value, worth only very slightly more than its liabilities.
- It mentioned this fact in footnotes to its financial statements.
- People noticed.
- SVB’s depositors, who again are largely tech firms and venture capitalists, are all on Twitter a lot and move in herds; when they noticed that SVB was insolvent-ish they all pulled their money out at once. (“It turned out that one of the biggest risks to our business model was catering to a very tightly knit group of investors who exhibit herd-like mentalities,” an SVB executive told the Financial Times.) Depositors tried to withdraw $42 billion on Thursday.
- SVB tried to pay them. It used its cash. It sold stuff that was easy to sell. It tried to borrow from the Fed, but “despite attempts from the Bank, with the assistance of regulators, to transfer collateral from various sources, the Bank did not meet its cash letter with the Federal Reserve.”
- It was declared insolvent and seized by the Federal Deposit Insurance Corp. on Friday.
SVB is, I think, an unusually clear case of a bank that
- was almost certainly insolvent, on a mark-to-market basis: By Friday, if not earlier, its assets, at their current market value, were probably worth less than its liabilities; and
- could probably have muddled through and been profitable if people had just kept their money in the bank: Its maturities were laddered, its deposit rates weren’t going up that much, it did have a positive net interest margin even this quarter, it did have various ways to make money, and if people had just kept their money in, the bonds would have matured and been replaced by higher-earning bonds and SVB would have been fine.
And so if SVB had gone to the Fed and been like “hi we have $120 billion of Treasury bonds and stuff like that, will you lend us $120 billion against them,” and if the Fed had said yes, it probably could have muddled through. But that $120 billion of stuff had a market value of only $100 billion.
The FDIC and other banking regulators spent the weekend trying to sell SVB, apparently with no luck. The goal of that sale would be to have some other bank buy SVB’s assets, business, franchise, etc. and assume its deposits, without necessarily paying anything to its shareholders or bondholders. I think the regulators’ preference would have been to leave the depositors intact and zero the shareholders and bondholders. They did not get there yet.
Here is what they came up with instead:
US authorities raced on Sunday to stem jitters about the health of the nation’s financial system, pledging to fully protect all depositors’ money following the collapse of Silicon Valley Bank while also giving any banks squeezed for cash easier terms on short-term loans.
The Treasury Department, Federal Reserve and Federal Deposit Insurance Corp. jointly announced the efforts aimed at strengthening confidence in the banking system after SVB’s failure spurred concern about spillover effects. …
The Fed in a separate statement said it’s creating a new “Bank Term Funding Program” that offers loans to banks under easier terms than are typically provided by the central bank.
Fed officials said on a briefing call that the facility will be big enough to protect uninsured deposits in the wider US banking system. It was invoked under the Fed’s emergency authority allowing for the establishment of a broad-based program under “unusual and exigent circumstances,” which requires Treasury approval. …
Under the new program, which provides loans of up to one year, collateral will be valued at par, or 100 cents on the dollar. That means banks can get bigger loans than usual for securities that are worth less than that — such as Treasuries that have declined in value as the Fed raised interest rates.
Here is the joint statement:
After receiving a recommendation from the boards of the FDIC and the Federal Reserve, and consulting with the President, Secretary Yellen approved actions enabling the FDIC to complete its resolution of Silicon Valley Bank, Santa Clara, California, in a manner that fully protects all depositors. Depositors will have access to all of their money starting Monday, March 13. No losses associated with the resolution of Silicon Valley Bank will be borne by the taxpayer.
We are also announcing a similar systemic risk exception for Signature Bank, New York, New York, which was closed today by its state chartering authority. All depositors of this institution will be made whole. As with the resolution of Silicon Valley Bank, no losses will be borne by the taxpayer.
Shareholders and certain unsecured debtholders will not be protected. Senior management has also been removed. Any losses to the Deposit Insurance Fund to support uninsured depositors will be recovered by a special assessment on banks, as required by law.
And here is the Fed statement:
The additional funding will be made available through the creation of a new Bank Term Funding Program (BTFP), offering loans of up to one year in length to banks, savings associations, credit unions, and other eligible depository institutions pledging U.S. Treasuries, agency debt and mortgage-backed securities, and other qualifying assets as collateral. These assets will be valued at par. The BTFP will be an additional source of liquidity against high-quality securities, eliminating an institution’s need to quickly sell those securities in times of stress.
I think the main point here is just “these assets will be valued at par.” Silicon Valley Bank is gone, but the next bank with bonds that it bought for $100 and that are now worth $80 can go to the Fed and borrow $100 against them.
One way to look at this crisis is:
- Banks borrowed short to lend long, as they generally do.
- Rates went up far and fast, causing the market value of the banks’ long-lived assets to decline a lot.
- Unlike in It’s a Wonderful Life, this decline was very noticeable, because it is relatively well accounted for. At least in the footnotes of the banks’ financial statements, you can see how much market value their assets have lost. You can make a chart! It might show hundreds of billions of dollars of mark-to-market losses on investment securities, spread among US banks.
- Also unlike in It’s a Wonderful Life, there is Twitter and people can tweet about this. There is online banking and people can pull their money from their banks on their phones while they are on the bus to the ski resort. Runs can happen fast; panics can spread fast.
- The problem is a fairly pure one of interest-rate sensitivity. The banks did not make bad loans to bad borrowers; they are mostly still profitable day-to-day. It’s just that if they had to sell all their bonds today, some of them would be in bad shape.
- The basic job of the Fed, as lender of last resort, is to make it so the banks don’t have to sell their bonds today.
- By lending the banks money against those bonds.
- And by valuing those bonds at par — by ignoring the mark-to-market accounting that has gotten everyone else worried.
The Fed can do that! The Fed can certainly hold the assets to maturity; it can’t lose its funding and be forced to sell everything in a panic.
One way to think of this is that US banks — especially SVB, but not only SVB — have had huge mark-to-market losses on their bond portfolios as interest rates go up, but it is traditional for banks to ignore those losses. In traditional banking, rising interest rates are a matter of opportunity costs and net interest margin, not of large mark-to-market losses.
But in the modern world — of more pervasive financial markets and more sophisticated accounting and faster-moving information — the banks and their customers were unable to ignore those losses. So the Fed stepped in and said: Look, we are best positioned to ignore those losses, so we will. The service that the Fed is providing to the banking system here is ignoring that rates went up when it values banks’ bonds. That service is incredibly valuable. Historically banks’ retail depositors provided it, but now only the Fed can.