- A new National Bureau of Economic Research (NBER) working paper, co-authored by researchers from Stanford and Columbia, analyses the US banking system’s exposure to rising interest rates.
- Their findings suggest that recent declines in bank assets – around 10% across all banks – have ‘very significantly’ increased exposure to uninsured depositor runs.
- In a scenario where only half of uninsured depositors decide to withdraw, almost 190 banks and $300 billion of insured deposits will be at risk of impairment.
- Nevertheless, few banks have a larger uninsured deposit-to-asset ratio than Silicon Valley Bank (SVB), hopefully implying a minimal risk of future bank runs.
Bank runs are the result of either liquidity issues, solvency concerns, or both. Most are self-fulfilling. Concerns about profitability fuel concerns about deposits, leading to deposit withdrawals requiring the liquidation of long-term assets, which further hurt the bank and finally lead to a run.
In the 1980s and ‘90s, profitability concerns caused by rising interest rates caused roughly a third of all bank runs. Despite stricter banking regulations, SVB was a stark reminder that bank runs can still happen, and that rising interest rates could cause more in the near future.
To this end, a new NBER working paper examines how susceptible the US banking system is to SVB-style bank runs. The results show that bank losses on long-term assets over the last year have significantly increased exposure to uninsured depositor runs. However, reassuringly, few US banks are currently estimated to be unable to meet their uninsured deposit obligations.
Why Do Bank Runs Happen?
The possibility of a bank run is inherent whenever a bank has illiquid assets (loans) and liquid liabilities (deposits). Liquid liabilities give depositors the option of withdrawing early. But since the bank’s assets are illiquid, if all depositors try to withdraw early, the bank will be unable to satisfy all of them.
A simple framework that rationalises the possibility of bank runs is the Diamond and Dybvig (DD) model, published in 1983.
It assumes individuals can deposit their money at a bank in period 0, and then subsequently withdraw their assets either in period 1 (type-A depositors) or period 2 (type-B depositors). The bank can offer a slightly higher return to type-B depositors because it invests deposits into projects, which only yield higher payoffs if they are held over the long-term.
Crucially, the bank does not know how many type-A or type-B depositors there are. They do know there will always be type-A depositors – individuals who withdraw, for example, to pursue unexpected investment opportunities or have unexpected short-term cash troubles – so they need to be able to liquidate projects to meet the demand for early withdrawals.
However, a bank cannot meet the deposit demand if everyone withdraws early. It will return deposits on a first-come, first-served basis, leaving only a fraction of the total depositors able to get their money back. In this scenario, type-B depositors, who are indifferent between withdrawing in period 1 or period 2, will also try to withdraw in period 1 because they are guaranteed no money in period 2 (if uninsured).
There are two equilibrium outcomes in this model.
The first is that only type-A agents withdraw in period 1, and only type-B agents withdraw in period 2. In this scenario, everyone gets both their deposit back and a higher return than what they entered with. Type-A agents get additional value from ‘liquidity’, i.e., being able to access their funds when they need them in period 1. And Type-B agents get additional value through higher returns from the bank’s long-term investments. Both agents also receive interest on their deposits from period 0 to periods 1 or 2.
The second outcome is a bank run. This happens when Type-B agents believe that other long-term deposit holders (other Type-B’s) may start to withdraw. In this scenario, everyone will want a slice of the period 1 returns because, if they get there first, they are guaranteed a higher return than what they put in (remember everyone receives interest). Sadly, only a fraction of depositors then actually get a return.
A Liquidity Crisis in Today’s Context
The bank run described above is a pure liquidity crisis, as opposed to a solvency crisis. Sparked by a lack of confidence in the underlying assets, individuals make a run for their deposits even if there was little chance the bank was going insolvent. It all stems from asymmetric information, on behalf of the bank and long-term deposit holders.
For banks, asymmetric information has arguably become less of a problem since the Global Financial Crisis (GFC). Stricter capital requirements have increased reserve buffers, enabling banks to tolerate a larger fraction of Type-B agents withdrawing in period 1.
For depositors, however, a pure liquidity crisis seems more plausible in today’s society. Most can access their money within seconds and, if they know they can withdraw at the first indication of trouble, they will.
Moreover, this fast access to liquidity (and an element of insurance) reduces the need for depositors to assess the quality of the bank before handing over their money. And, if fewer customers are aware of a bank’s fundaments, they may be less willing to tolerate even small concerns about its profitability, making them more likely to leave.
Insolvency Risk and Bank Runs
Concerns about profitability are more likely to arise when central banks are tightening. As interest rates rise, the value of long-term assets, such as government bonds and mortgages, decline. This can create losses for banks, which then engage in maturity transformation – they finance long maturity assets with short-term liabilities (deposits).
Bank runs in this scenario can occur through two channels. Type-B agents with uninsured deposits worry about their period 2 returns, so withdraw their funds(a liquidity-solvency crisis). Or, a bank’s liabilities end up exceeding the value of its assets (particularly likely for banks which need to increase deposit rates as interest rates rise) and it becomes solvent (a solvency crisis).
How Large Have Bank Losses Been in 2022?
The authors begin by estimating the losses banks have sustained due to rising interest rates. Since a substantial portion of bank portfolios, specifically loans held to maturity, are not marked-to-market, they use exchange-traded funds (ETFs) across various asset classes to calculate an estimated mark-to-market loss.
They focus mainly on real estate loans and securities linked to real estate, which make up over half of the average bank’s assets (66%). For this they use traded indexes in real estate and treasuries to impute the market value of real estate loans held on bank balance sheets. For longer duration fixed income assets, they adjust across maturities using treasury prices.
Marking the value of real estate loans, government bonds, and other securities, shows a significant decline in bank assets over the last year (Chart 1). The median value of banks’ unrealised losses was around 9% after marking to market, and >20% for the worst 5% of banks affected.
For the largest banks (Global Systemically Important Banks – G-SIBs), the losses were smaller and arose mainly from residential mortgage-backed securities. For smaller banks, the loses were more severe and occurred due to non-real estate loans. Finally, SVB’s losses were bad (bottom 15th percentile), but roughly 500 other banks suffered greater losses.
According to the authors, the average bank in their analysis funds 10% of their assets with equity, 63% with insured deposits, and 23% with uninsured debt comprising uninsured deposits and other debt funding.
Banks differ significantly in the share of funding they obtain from uninsured sources. The 5th percentile bank uses 6% of uninsured debt. For this bank, 94% of funding is not run prone comprising equity and deposits.
Whereas, the 95th percentile bank uses 52% of uninsured debt. For this bank, even if only half of uninsured depositors panic, this leads to a withdrawal of one quarter of the bank’s total marked-to-market value.
SVB was in the 1st percentile of distribution insured leverage. Over 78% of its assets were funded by uninsured deposits. This suggests that uninsured deposits played a critical role in SVB’s failure.
Chance of a Run?
Using the mark-to-market losses, the authors calculate whether banks in Q1 2023 have enough assets to cover their uninsured deposit obligations if all uninsured depositors run.
Thankfully, most do (Chart 2). Only two banks have an uninsured deposit-to-asset ratio over 100% after accounting for mark-to-market losses on their Q1 2022 balance sheets. For SVB (the dotted line), this was 92.5%.
So, if the Federal Deposit Insurance Corporation (FDIC) does not intervene to protect the deposit insurance fund, or if the liquidation of the assets does not cause large enough fire sales, there may be no reason for uninsured depositors to run (if the run is only determined by whether a bank can cover its obligations).
Has Monetary Policy Tightening Increased the Run Risk?
Next, the authors look at whether insured depositors would be impaired by uninsured depositors running.
Before Federal Reserve (Fed) interest rate increases, US banks were all solvent regardless of whether uninsured depositors ran or not. In other words, even if all uninsured deposits would have been withdrawn, the remaining assets would have been sufficient to cover insured deposits.
However, with the substantial losses obtained because of Fed tightening, this is no longer the case (Chart 3 – grey bars). If all uninsured depositors withdrew funds from US banks, 1619 would have negative insured deposit coverage, leading to insolvency and impaired insured deposits.
While the average bank in the sample is small, with assets of $0.3bn, the aggregate losses would be large, and would involve $2.6tn of aggregate deposits, and a shortfall for the deposit insurance fund of $300bn. In this scenario, the FDIC would need to intervene.
Even if only half of all uninsured deposits were withdrawn, the authors estimated that the losses to the deposit insurance fund would total approximately $10bn and affect 186 US banks.
The paper finishes with an example of the above scenario.
Assume a bank has $100bn in assets (10% cash, 90% treasuries), $80bn in deposits at the deposit cost of 3% and $10bn in long-term debt at a fixed rate of 3%. The current risk-free rate is 3%.
The central bank increases the risk-free rate by 100bps. This decreases the value of long-term assets by 25% to $67.5bn ($90bn*3%/4%), leaving total bank assets at $77.5bn.
Assume 50% of depositors are uninsured (as is the case for >5% of US banks) and they all run.
The bank then needs to pay uninsured depositors $40bn, with the remaining marked-to-market value of the bank’s assets $37.5bn. This is less than the face value of the remaining insured deposits ($40bn), so the FDIC would close the bank – it is insolvent.
Bank runs are still possible, even with the emergence of tighter regulations. Banks with a large share of uninsured deposits are especially susceptible, and that risk is amplified in a rising interest rate environment.
The paper shows that the losses sustained over the last year have very significantly increased exposure to uninsured depositor runs. Whether those runs happen is another question. In theory, most US banks can cover their uninsured deposit obligations. However, new technology is making it easier for depositors to withdraw, and making it less likely depositors will stay in times of stress.
If runs happen, and banks lose 30-50% of their deposits in days, the paper shows how dramatic that could be for the US banking system. Many would not cope and large losses would ensue. If a lender of last resort steps in, this would also increase inflation, as new money is created to support banks.
In response, regulators may insure all deposits – a possibly uncredible move. Instead and in theory, the incentive should be just enough to encourage Type-B depositors to stay, which must be higher than Type-A depositors. This could, for example, involve discounting early withdrawals.
Sam van de Schootbrugge is a Macro Research Analyst at Macro Hive, currently completing his PhD in international finance. He has a master’s degree in economic research from the University of Cambridge and has worked in research roles for over 3 years in both the public and private sector.
Photo Credit: depositphotos.com