The potential of large bank failure puts the American taxpayer at risk. Preventing such failure and subsequent bailouts requires understanding the ability of these banks to withstand a large shock, such as a recession. With that understanding will come the motivation to enact government policies that will protect taxpayers.
One of the best tools to assess the condition of banks is a stress test. Providing regular, transparent assessments of the health of the largest banks—and whether they will have sufficient capital to support lending during a future economic shock—will help empower the public to make their own assessments of the strength of large banks. Minneapolis Fed economists use stress test insights to inform their ongoing research on the financial system, including our determination that bank capital remains too low despite weathering the COVID-19 financial shock.
We created the tool below to allow the public to run their own stress tests on the largest banks in America. You can vary the amount of stress that large banks face in this test and see how their financial health (for example, capital buffer) changes. In short, the public will now have the transparency they need to judge the condition of the largest banks.
Results produced from the tool are derived from the model we created that was based heavily on Hirtle, Kovner, Vickery, and Bhanot (2014). You can find output from that original model on the New York Fed website. A description of the key changes in the Minneapolis version can be found here. We continue to evolve the stress testing tool to increase its power and functionality for the public. You can learn more about potential uses of this tool based on the conference we sponsored on the topic, “Empowering the Public to Assess Large Bank Resiliency,” by viewing this video and consulting this article. Feedback and suggestions for additional enhancements to the underlying model are welcome.
Moderate recession | Extreme recession | Rising rates | |
---|---|---|---|
Select all | |||
Real GDP growth | |||
Unemployment | |||
3m T-bill yield | |||
10yr T-note yield | |||
BBB bond yield | |||
CRE price index | |||
House price index | |||
Stock market |
(% of net income paid out in dividends & share buybacks)
Macroeconomic Variables
Resources
Read Frequently Asked Questions about the Minneapolis stress test model.
Notes
- The data shown above in Chart 1 is an extension of the data found in article on stress testing for large banks in the COVID-19 pandemic published on May 12, 2020. This webpage allows users to select different paths for the eight macroeconomic variables used in the Minneapolis stress test tool to generate projection of capital for 22 large domestic banks that are part of the 2022 DFAST exercise.
- The macroeconomic variables shown here are not forecasts but are instead hypothetical scenarios. The scenarios were constructed to help illustrate how large banks would be impacted by the adverse economic conditions associated with severe recessions and rising interest rates.
- The capital projections in the stress scenarios reflect the projected performance of banks under the negative economic conditions along with additional losses resulting from operational risk, counterparty default, and the global market shock to large bank trading portfolios from the official DFAST 2022 results (which can be found here).
- The stress test tool currently assumes that the individual firms hold constant balance sheets over the projection period. Therefore, total assets, liabilities, and risk-weighted assets remain fixed at their starting levels over the nine projection quarters for each firm.
- In 2021, we adjusted the loan loss models that have the unemployment rate as a key input. We made the adjustments to account for the rapid increase in unemployment that occurred in 2020. Historically, there has been a strong positive relationship between losses on these loan types and changes in the unemployment rate (increases in unemployment are associated with higher losses and decreases are associated with lower losses). The 9.3 percentage point increase in unemployment that occurred in the second quarter of 2020 was the largest ever seen since data were first collected in the late 1940s. Moreover, it was over five times as large as the next biggest quarterly change in the series. However, at the same time, there was no corresponding increase in loan losses given the very unusual nature of the shock to the economy and the massive government response. Not adjusting the models for this extreme observation effectively eliminates the relationship between loan losses and unemployment. Accordingly, quarterly dummy variables have been included in the models to denote the second-quarter 2020, third-quarter 2020, and fourth-quarter 2020 time periods. Utilization of dummy variables in this fashion removes the impact from second-quarter 2020 data points on the estimated coefficients and preserves the pre-existing historical relationship. Additional details regarding this approach are available upon request.
- In 2022, we updated our stress test tool to reflect the new stress capital buffer (SCB) framework, which was finalized by the Federal Reserve in 2020 and fully implemented in 2021 (a complete description of the SCB can be found in 12 CFR 217.11, 85 FR 15909, and 85 FR 63423). The red line shown in Chart 1 is the “required amount” of capital that each bank must hold in order to make distributions—such as as dividends, share buybacks, and discretionary bonus payments—with no constraints or limitations. The value for each bank is the sum of the regulatory minimum (4.5%), the countercyclical buffer (currently at 0%), the stress capital buffer for the bank (2.5% or higher) and, if applicable, the global systemically important bank surcharge (1.0% or higher). We hold this amount constant over the projection period at each bank. The value shown for the “22 firm total” is simply the weighted average of the individual bank amounts, where the weights are equal to each firm’s share of the total risk-weighted assets for the group.
- We made the following choices in order to operationalize the SCB framework in the Minneapolis stress test model:
- We define the amount of income that is available to be paid out for dividends and share buybacks in the projection period as being equal to the quarterly average of net income over the previous four quarters. This value is referred to as “eligible retained income” (ERI) in the SCB framework. The official framework specifies that ERI is the maximum of either (i) the average of net income over the prior four quarters or (ii) total net income over the prior four quarters net of any distributions. Thus, our definition of ERI—and the resulting capital ratios—will be lower-bound estimates since banks could, in practice, have a higher ERI under the second definition.
- No capital is assumed to be paid out for discretionary bonuses. The only distributions that banks make in the projection period are for dividends and share buybacks. This choice means that capital projections from the Minneapolis stress test tool would be lower in cases where banks would have made other discretionary expenditures.
- We initially assume that dividends and buybacks in the projection period will be equal to the average amount that a bank paid out for these items in the four quarters before the projection period begins. We further limit this amount so that it does not exceed the income earned by the bank over that period. We then adjust the payout for dividends and share buybacks in accordance with the SCB framework during the projection period when the capital ratio falls below the “required amount.”
- The “payout policy” switch allows users to determine how capital projections change if banks are not allowed to pay dividends and share buybacks. The projected capital for a bank will be higher when the bank does not pay dividends and conduct share buybacks.
- The “remove SCB” option in the Minneapolis stress test tool allows users to determine how capital projections change if the SCB limitations on paying dividends no longer apply. In this option, we assume that banks cannot buy back shares if their projected capital falls below the “required amount.” We also assume that banks can continue to make dividend payments at the same level as banks were at the beginning of the projection period. These assumptions are generally similar to the situation in 2020 when banks stopped making buybacks voluntarily followed by the Federal Reserve halting those buybacks later in 2020. Banks continued paying dividends during this period. Removing the SCB results in lower capital projections for banks as they can make more payouts.
- The stress test tool was also modified in 2022 to incorporate changes in the value of available-for-sale (AFS) securities due to movements in interest rates on accumulated other comprehensive income (AOCI). Since AOCI is a component of bank capital, gains or losses on AFS securities will now directly impact the projected capital of a bank in our framework. Note that this treatment only applies to banks that are classified as “advanced approaches” institutions (and for banks that have chosen to opt into this treatment). Capital projections will incorporate this impact for the following ten banks in the current version of the tool:
American Express Bank of America Bank of New York Mellon
Citigroup Goldman Sachs JP Morgan Chase
Morgan Stanley Northern Trust State Street
Wells Fargo
Our projections for these impacts are based on the publicly reported data available in the standard regulatory reports filed by the banks, as are all of the items included in the stress test tool. The framework we use to model gains and losses on AFS securities is similar in spirit to what occurs in the official DFAST exercises. However, we do not have access to the detailed security level data used in those tests and our projections are therefore far simpler. Importantly, we are not taking into account any valuation changes that occur for reasons other than interest rate movements, nor are we controlling for interest-rate hedges or credit losses.
The specific process we use to model gains and losses on AFS securities is detailed below:
- We utilize information from the Call Reports of individual banks within a given organization to measure the amount of AFS securities and held-to-maturity (HTM) securities that are in the following three groups:
Group 1 – Treasury securities, Agency securities, municipal securities, pass-through securities not backed by 1-4 family residential loans, asset-backed securities, structured products, US corporate debt, and foreign debt securities.
Group 2- Pass-through securities backed by 1-4 family residential loans
Group 3- Other mortgage-backed securities (including CMOs, REMICs, and stripped MBS) - We also use the Call Report data to measure the total amount of securities (both AFS and HTM) in the three groups listed above that have a remaining maturity or next repricing data in the following categories:
Group 1 (six categories) – Less than 3 months, 3 – 12 months, 1 – 3 years, 3 – 5 years, 5 – 15 years, greater than 15 years
Group 2 (six categories) – Less than 3 months, 3 – 12 months, 1 – 3 years, 3 – 5 years, 5 – 15 years, greater than 15 years
Group 3 (two categories) – Less than 3 years, greater than 3 years - We calculate the share of total securities in each group that are AFS securities based on the data in [a] for an organization.
- Next, we multiple the maturity category totals in [b] by the corresponding AFS percentage found in [c] to obtain an estimated balance amount of AFS securities by remaining maturity/repricing category for the organization.
- We assume all securities in a given maturity/repricing category have the same effective duration, based on the following formula:
Duration
= 0 if the maturity/repricing category is “less than 3 months”
= 7.5 if the maturity/repricing category is “greater than 15 years”
= 4.5 if the maturity/repricing category is “greater than 3 years”
= [0.5] * [mid-point of the maturity/repricing range] for all other categories - We calculate the quarterly gain or loss in AFS securities in a given maturity/repricing category from interest rate movements as the following:
Gain/loss = [-1] * [quarterly change in the interpolated interest rate that corresponds to the duration] * [AFS securities balance]
The interpolated interest rate is found using the 3-month and 10-year Treasury rates from a given scenario in the respective quarter.
- The impact on CET1 capital in a given quarter is equal to the tax-adjusted sum of the eight individual maturity/repricing categories listed in [b]. Increases in interest rates will result in capital losses while decreases will result in capital gains.
- We utilize information from the Call Reports of individual banks within a given organization to measure the amount of AFS securities and held-to-maturity (HTM) securities that are in the following three groups: