Why the Banks Failed: Liquidity, Leverage & Losses

Most financial crises are the result of a combination of liquidity mismatch and leverage on balance sheets. The majority of the time the balance sheets are of financial institutions. For example, the recent blowups of several banks such as Silicon Valley Bank and Silvergate Bank, the blowup in the government bond markets of the UK (the gilt markets) in 2022, and the global financial crisis of 2008–09 are all a result of this effect. The cascade of failures among cryptocurrency exchanges and funds was another example. These financial crises also remind of the important role of a liquidity provider of last resort, typically a country’s central bank.

This article will be the first of a series in which we analyze financial crises. We will begin by going over the basic concept of a liquidity mismatch combined with leverage and how that combination is the key to a financial crisis. We will use the recent run on Silvergate Bank and its resulting shutdown to illustrate how this mechanism works. In subsequent articles, we will look at other more complex financial crises to understand the multiple channels through which liquidity-leverage combinations can create losses and lead to crises.

A bank run provides a perfect example of a financial crisis. A highly simplified bank balance sheet for Silvergate Bank on Sept. 30, 2022 is shown below:

This bank’s balance sheet is being used because it is a fine example of a typical bank’s balance sheet but also because Silvergate Bank endured a bank run and went bankrupt soon after. To highlight the main points of this article, we have greatly simplified Silvergate’s balance sheet by aggregating its numerous accounts into just a handful of accounts.

On the right-hand side of the balance sheet is the bank’s financing. We have aggregated the financing to three types: depositors (senior debt providers), other debt providers (junior debt providers), and equity providers.¹ In virtually all banks, the debt providers make up over 90% of the capital structure of the bank. So, the financing structure of Silvergate bank, where deposits/loans make up 92%, is very close to what it would be typically. The senior debt in this bank is very liquid; depositors can arrive at any time at a bank branch or the bank’s website and withdraw their deposits. The bank cannot charge a bid-ask spread at the time of withdrawal, so regardless of the amount of their deposits that depositors ask for, the bank must deliver the full amount asked (there cannot be any haircut as there might be with traded securities). So, the debt is perfectly liquid.

On the left-hand side of the balance sheet, there is a a small amount of the cash proceeds that the bank keeps from the deposits as cash. This is amount is about $1.9 billion in the case of Silvergate.² The bank lends the remainder of the deposit proceeds to companies, individuals, etc. These are accounted for as loans on the left-hand side of the balance sheet, about $12.8 billion. The loans earn the bank a higher interest rate than they must pay on their debt, and so, the bank earns a profit on this interest rate spread. These loans are typically also very illiquid — they are term loans. Contrary to deposits, the bank cannot simply ask for their proceeds back from the entities they have lent to any time they want. Additionally, these loans have much longer maturities than do deposits. Deposits can effectively be thought of as loans to the bank with zero maturity.³

In the case of Silvergate, most of the deposits were from cryptocurrency (crypto) exchanges and traders. Essentially, traders who wished to trade cryptos would deposit dollars at Silvergate. They would then wire some of these dollars to crypto exchanges using the Silvergate Exchange Network (SEN) and leave the rest at Silvergate as a deposit. After the exchanges received the dollars and gave the traders cryptos in exchange for the dollars, the exchanges had extra dollars. The exchanges then took the dollars and deposited them at Silvergate as well using SEN. Of the $13.2 billion in deposits at Silvergate, crypto exchanges and traders represented about $12 billion.

On the asset side, Silvergate invested in fixed-income bonds and loans of all types. Of the $12.8 billion in loans, about $11.4 billion were in publicly traded bonds, and the remaining $1.4 billion was in risky loans, mostly to the same crypto traders and exchanges.

The maturity difference between deposits and loans/bonds at Silvergate created a liquidity mismatch between the bank’s left-hand side of the balance sheet and its right-hand side. The bank recognized the risks of this liquidity mismatch and therefore held some cash to meet deposit withdrawals that did not match up with the maturity of their loans.⁴

The important characteristic to recognize is that a bank like Silvergate is highly leveraged. It is borrowing the vast majority of the proceeds it needs to run its business. Some of these borrowings are coming from individuals’ and companies’ cash deposits, and some are coming from overnight deposits derived from the cash that other banks are lending/depositing to banks and holding on the left-hand sides of their balance sheets. So, banks deposit their extra cash with other banks, therefore not only enhancing leverage in the banking system but also making all banks interconnected.⁵

Some of a bank’s proceeds needed to run its business also comes from bank capital, or equity. Equity is there to protect against loan losses due to any defaults on the loans it has made on the left-hand side of its balance sheet. When there are defaults, the first losses are absorbed by this bank capital, i.e., shareholders. Effectively, this protects depositors by providing a cushion against loan defaults. However, if there are large enough loan losses, these losses will eat through the bank’s capital and then will start eating through the bank’s debt — that’s when depositors start taking losses. So, there is an insurance model of diversification in play in a bank’s business.

The bank wants to make sure that loan defaults don’t occur all at once. To ensure this, the bank will lend small amounts to a wide array of customers: a variety of individuals and many companies in a variety of industries. That way, if an individual company defaults, or even if an entire industry gets into financial trouble, it should have no effect on the bank’s depositors. So long as the bank has a well-diversified loan portfolio, a few loan defaults will not be enough to eat through a bank’s capital and threaten its depositors with losses. However, the common element with these loans is that they are all significantly more illiquid than the deposits used to finance them and, most importantly, there is no way to diversify around this liquidity mismatch.

Let’s now consider what happens when there’s a systematic shock to the economy. A systematic shock, such as a recession, affects many companies and individuals in an economy simultaneously. This contrasts with a company-specific shock or an individual-specific shock. With a systematic shock, several of the bank’s loans approach default simultaneously. This is precisely what happened during the Global Financial Crisis when mortgage defaults across the United States increased substantially and all at once. With so many loans in trouble, the possibility of loan losses eating through a bank’s capital becomes a more likely possibility and as depositors of that bank realize that the bank is now facing a higher probability of its capital being used up and therefore depositors taking some losses, some worried depositors start to pull their deposits out of the bank. Because the deposits are perfectly liquid, the bank must fulfill these withdrawal requests — it cannot gate (delay) or haircut the withdrawals. This forces the bank to use up its cash reserves.

As the remaining depositors see both the bank’s capital being used up and its cash reserves beginning to be depleted, they become worried and start asking for their deposits back. At this point, the bank has to generate more cash. Since the bank cannot ask for its money back (due to the long maturity of the loans) from people and companies (borrowers) that have taken out loans, the bank will instead have to sell these loans on a secondary market. However, because these are illiquid bank loans the bank will take a significant haircut when it tries to sell them. This haircut is a loss to the bank, but it at least provides cash proceeds to meet the withdrawal requests coming from depositors. Unfortunately, the haircut is still a loss and consequently will eat into the bank’s capital, causing the bank’s capital to be depleted even faster.

This acceleration causes even more depositors to worry that they will take losses, resulting in an acceleration of withdrawal requests. The withdrawal requests cause more cash needs, forcing the bank to sell more illiquid loans, which causes more loan losses due to illiquidity haircuts, which in turn causes even more losses on bank capital, which creates even more deposit withdrawal requests, and the cycle keeps going on and on. The bank is now in a death spiral, or more commonly known as a “run on the bank.”

In the case of Silvergate Bank, the systematic shock was the plummet in the value of cryptos in 2022. This resulted in the failure of several cryptocurrency exchanges and traders in 2022, mostly occurring in the second half of the year. As the price of cryptos dropped and the exchanges and traders got into trouble, some of Silvergate’s loans to these entities went bad. This resulted in many exchanges and traders pulling their deposits from Silvergate. In the three months from September until the end of December 2022, $8.2 billion in deposits from crypto-related entities were withdrawn.

To come up with the cash needed to meet these withdrawals, Silvergate borrowed $4.8 billion short-term from the Federal Home Loan Bank (FHLB) system and others.⁶ But to maintain the confidence of their remaining depositors, Silvergate had to increase their cash holdings by $2.7 billion. So, they still needed to generate an additional $6.1 billion in cash ($8.2+$2.7–$4.8). To do this, they sold $6.6 billion of the loans and bonds they held and due to illiquidity and interest rate-related losses,⁷ Silvergate lost about $900 million on these sales. These losses ate through some of Silvergate’s equity capital. Unfortunately for Silvergate, this loss of equity capital worried depositors even more and caused even more depositors to pull money.

Surviving exchanges such as Coinbase stopped using SEN and refused to accept payments from SEN. This in turn caused an acceleration of deposits being withdrawn, which in turn caused more losses. A full bank run was on!

Finally, on March 8 Silvergate was forced to shut down. Unlike Silicon Valley Bank and Signature Bank, uninsured depositors (those with accounts greater that the $250K insured by the Federal Deposit Insurance Corp. (FDIC)) did not get their full deposits back. Regulators did not deem Silvergate a “systemically important” bank and any remaining uninsured depositors ended up taking some losses.

The reason a bank run occurs is the mismatch in liquidity between the liabilities that the bank uses to finance itself (deposits) and the assets the bank invests in (private loans). If the deposits were much more illiquid, so that depositors either could not withdraw right away or they took a significant haircut by withdrawing, then the bank run would not occur. Or, if the loans the bank made were more liquid so that the bank could get its loan proceeds back without a haircut, then the bank run also would not occur. It is the combination of having liquid liabilities mismatched with illiquid assets that creates a bank run. And a bank run typically only manifests itself after a systematic economic shock.

A simple bank run can evolve into an economy-wide financial crisis in two ways. First, with a systematic shock, such as economy-wide mortgage defaults, it isn’t just one bank getting hit with loan losses — all banks are getting hit with loan losses. This means that the bank run scenario we just described is hitting all banks simultaneously, with some hit harder than others. However, as the more desperate banks start liquidating illiquid loans on secondary markets, the values of those loans plummet, creating more losses for banks. This results in banks having to sell even more illiquid loans to generate the cash that they need. This has two effects. First, the selling of more loans at deeper discounts results in even greater losses at banks, and therefore an even faster pace of withdrawal requests. Second, banks anticipate that other banks might be in forced liquidation mode to meet deposit withdrawals, and so banks will race other banks to liquidate first before the market values of these loans drop too much — so the bank run scenario occurs even faster than one might expect.

In summary, a systematic shock has transformed the state of banks and markets. Banks are now frozen: as they are facing rapid withdrawal requests and are rapidly liquidating loans to raise cash, they certainly are not making new loans. With enough banks not making loans, the whole economy starts to suffer. As a result, new systematic shocks start to form in the economy. These new systematic shocks will eventually have the same effect as and compound the original shock. Soon, the whole economy becomes caught in a death spiral.

The second channel a bank run evolves into a macro-financial crisis is if financial institutions have exposures to each other, which is the norm. A simple way to think about it is that one bank lends to (or deposits with) another bank. When the deposit-taking bank (Bank A) gets into a financial crisis, then a bank run occurs amongst depositors of Bank A, and the depositing bank (Bank B) is now in line with all of the other depositors. Of course, then the depositors of Bank B start to worry whether Bank B will get all of its money back, and so the depositors of Bank B start pulling their deposits out of Bank B. And now, the bank run that originally started at Bank A has now spread to Bank B. And so on. This is commonly known as financial contagion.

Essentially, financial contagion causes a bank run to spread from one bank to another. The financial crisis evolves sequentially (and quickly) from bank to bank as the crisis at each bank gets several other banks in trouble. Financial contagion is a more complex mechanism by which a bank run can spread economy wide. We will discuss this concept in more detail in subsequent articles focusing on the recent cryptocurrency crisis as well as the global financial crisis that began in 2008.

The way to break a bank run is to have a liquidity provider of last resort who is willing to step in and take the place of depositors who want to withdraw their money. As we discussed earlier, depositors are merely lenders, who are loaning money to the bank with perfect liquidity, i.e., they can call their loans back (withdraw their deposits) at any time. When a bank run starts, more of these loans are being called back by depositors than the bank has cash. What the bank needs to do is generate cash, and the best way to do that is to find a lender who will provide that cash and take the place of depositors (who are also lenders) who want out. This lender in most countries is the central bank. In the US, the Federal Reserve (Fed) serves as the country’s central bank.

Through the Fed’s discount lending window, banks can borrow as much as they need.⁸ This allows banks to not only survive bank runs but also to prevent bank runs in the first place — if depositors know that a bank has ready access to a central liquidity pool, they are far less likely to withdraw deposits in a panic and start a bank run. So, a liquidity provider of last resort creates stable financial institutions and markets by not only actively providing liquidity to stop bank runs but also setting investors’ expectations that it will step in and do so whenever needed.

In the case of Silvergate Bank, they turned to the FHLB (Federal Home Loans Bank), sometimes known as the lender of second-to-last resort. This should in theory have stopped the bank run. However, the FHLB only lent a total of $4.3 billion to Silvergate and then stopped. This allowed the bank run to continue. The question of why the FHLB did not go further in their loans to Silvergate or why the Federal Reserve didn’t step in and lend to Silvergate through their discount window is an interesting question.

Some speculate that (perhaps due to political pressure) the bank regulators’ views about the crypto industry had changed. They did not want to be perceived as supporting the industry, and specifically they did not want to be seen using taxpayer dollars to support the industry. So, the regulators became more willing to allow banks who had worked with the crypto industry to fail.

Footnotes

[1] In a typical commercial bank, depositors make up a smaller slice of the set of lenders to the bank. However, depositors have one of the highest priorities in terms of claims on cashflows, so we focus on them here.

[2] Most of this “cash” is actually either deposited as bank reserves with the Federal Reserve (the central bank of the US) and earns the Fed’s interest on reserve balances, or it is lent overnight to other banks and earns the overnight lending rate.

[3] Deposits have zero maturity because they can be demanded back by depositors at anytime.

[4] It also does so because it is legally required to, but this legal requirement is there simply to ensure the bank does indeed hold on to some cash in order to meet withdrawals.

[5] This is isomorphic to the statement that banks borrow overnight from other banks, thus making all banks interconnected.

[6] $3.6 billion was from the FHLB, and another $1.2 billion was from an increase in non crypto-related deposits.

[7] The interest rate losses occurred because of the dramatic increase in interest rates in 2022 as the Federal Reserve engaged in Quantitative Tightening (QT) and raising the Fed Funds rate.

[8] The banks have to put up some assets as collateral for these borrowings. However, they do not have to liquidate these assets, so they do not experience a loss which would eat into their equity capital. Thus, they both get cash and preserve equity capital by borrowing from the Fed.