Inside the Hypothetical World
The Federal Reserve’s 2026 stress test does not attach the word “severe” to a single shock and wait for a final number. It constructs a path through unemployment, output, property prices, interest rates, credit spreads, and equity markets, then carries that path through nine quarters of projected bank performance. This chapter follows that path. The aim is to show why the 2026 scenario is a deliberate counterfactual, how it differs from the prior exercise, and why the capital minimum is a point on a conditional trajectory rather than a forecast of where banks will actually end up.
The timing matters. A bank can face pressure early in the exercise, receive some offset from income later, and still have its most important result defined by the lowest point in between. A scenario is therefore not just a list of bad outcomes. It is an ordered set of conditions whose sequence determines which losses and offsets appear first and how they interact.
A scenario designed to be adverse
The 2026 severely adverse scenario combines a severe global recession with heightened stress in commercial and residential real estate and corporate debt markets. Its selected paths include unemployment rising 5.5 percentage points to 10 percent, real GDP falling 4.6 percent from peak to trough, house prices falling 30 percent, domestic commercial-real-estate prices falling 39 percent, the BBB corporate-bond spread rising 4.7 percentage points to 5.7 percent, and equity prices falling 58 percent. The three elements in each description matter: the variable, the direction of change, and the comparison base.
A statement such as “equity prices fall 58 percent” is not a general prediction about equities. It is a coordinate in the test world. The report describes the projections generated from the exercise as involving an economic outcome more adverse than expected. The scenario’s purpose is to make the banks’ response visible under pressure, not to announce that this combination is the Federal Reserve’s central view of the future.
That distinction also separates a baseline from a stress case. A baseline generally describes a more ordinary or expected path against which other outcomes can be compared. A severely adverse scenario is selected to challenge capital. The fact that a stress scenario is coherent does not make it likely in the statistical sense, and the fact that it is unlikely does not make it useless. Its analytical value comes from the relationships it tests: what happens when employment, property values, borrowing conditions, and market prices move together in an unfavorable direction?
The scenario is not a collection of independent alarms. Its variables can reinforce one another. A recession can weaken borrowers and reduce the income that supports debt service. Lower property prices can affect collateral and portfolios exposed to real estate. Wider corporate spreads can create market losses and signal deteriorating credit conditions. Equity-market stress can affect securities and trading positions. At the same time, interest-rate movements can create an offset in some income channels while worsening conditions in others. The point is not to assign one cause to the final capital ratio; it is to examine the combined path.
The order and duration of the variables matter as much as their endpoints. A quick fall in an asset price can create a market loss at one point in the horizon, while a persistent rise in unemployment can produce credit losses later as borrowers miss payments. A higher rate can first change interest income and securities valuations, then affect repayment capacity or refinancing conditions. The models do not need to treat these as one undifferentiated shock. They can apply the path to the relevant portfolios and allow the timing of projected effects to shape the capital minimum.
This is also why the phrase “severely adverse” should not be used as a substitute for the scenario table. Two paths can both be severe while stressing different parts of a bank. A path with a larger GDP decline may create broad pressure, while a path with a sharper commercial-real-estate decline may concentrate pressure in a narrower but important portfolio. The right question is not which label sounds harsher. It is which exposures are connected to which variables, over what period, and through which model channel.
The Federal Reserve’s 2026 path also reflects the fact that a stress exercise is refreshed over time. The report says that the scenario changes were related to developments in the U.S. economy during 2025 and to the design of the annual hypothetical scenario. That means the scenario is not frozen as a timeless test of an unchanging bank. It is a new application of the framework to a new starting environment.
Why 2025 and 2026 are not interchangeable paths
Comparing the 2025 and 2026 severely adverse scenarios helps reveal what a scenario path does—and what a year-over-year comparison cannot do. In 2025, the scenario raised unemployment by 5.9 percentage points to 10 percent, reduced real GDP by 7.8 percent from peak to trough, lowered house prices by 33 percent, and lowered commercial-real-estate prices by 30 percent. The 10-year Treasury rate fell to 1.0 percent, the BBB spread rose to 5.0 percent, and equity prices fell 50 percent.
In 2026, unemployment still rose to 10 percent, but the increase was 5.5 percentage points and the peak-to-trough GDP decline was 4.6 percent. House prices fell 30 percent, while commercial-real-estate prices fell 39 percent. The 10-year Treasury rate fell by only 1.8 percentage points to 2.3 percent. The BBB spread rose further, to 5.7 percent, and equity prices fell further, by 58 percent.
It would be misleading to compress those differences into a single verdict that one scenario was simply “worse.” Severity is variable-specific. The 2026 path was less severe on the reported GDP decline than the 2025 path, more severe on commercial real estate, corporate spreads, and equities, and different on interest rates. Those differences matter because banks do not hold one generic portfolio. The same change in commercial-real-estate prices can matter more to a bank with a concentrated exposure than to one with little such exposure. The same interest-rate path can produce different income and securities effects depending on business mix and holdings.
The starting population changed too. Twenty-two banks participated in the 2025 exercise, while 32 participated in 2026. Category IV banks generally participate every other year, so the aggregate populations are not fully comparable. A chart showing a smaller aggregate CET1 decline in 2026 can be useful as a description of the two exercises, but it is not a clean experiment in which the same banks face the same model and only the calendar year changes.
The starting balance sheet changed as well. The Federal Reserve reports that loan balances grew by around 10 percent during 2025, concentrated in credit cards and wholesale loans. More loans can mean more exposure to the same scenario. It can also change the mix of projected revenue and loss. If the portfolio expands in a category with a high stressed loss rate, the dollar amount of loss can rise even when the scenario path itself has not become uniformly more severe.
This is why a year-over-year stress-test comparison needs several questions before it can support a conclusion. Did the participating banks change? Did their exposures change? Did the scenario variables change in direction or timing? Did the models change? Did recent bank performance alter the starting level of projected income? The 2026 report addresses these factors together. Treating the headline decline as a pure trend in resilience would discard the experiment’s changing conditions.
A useful comparison therefore separates three kinds of movement. First, the scenario can move: the 2026 path has a different mix of GDP, property, spread, equity, and interest-rate changes from 2025. Second, the bank population and balance sheets can move: the participating sample and the size and composition of loans are not fixed. Third, the response system can move: model adjustments or recent financial performance can change how the same broad conditions are translated into projected income and losses. A year-over-year result is the product of all three, not a pure observation of one underlying trend.
The qualification is particularly important for a reader who wants to use stress tests as evidence of improvement. A smaller aggregate decline may mean that banks entered the exercise with stronger cushions, that the tested population changed, that income offsets were larger, or that the scenario was less severe for the exposures that dominate the sample. A larger decline may reflect the reverse. None of these explanations should be selected from the headline alone. The report’s decomposition is valuable because it makes the competing explanations visible, even when it cannot turn them into a single unconditional forecast.
From macro path to capital minimum
The 2026 exercise starts with an aggregate CET1 ratio of 12.8 percent at 2025:Q4. Over the nine-quarter projection horizon, the ratio reaches a minimum of 11.2 percent in the fourth projection quarter and then rises to 12.7 percent by 2028:Q1. The state changes are not predictions of actual quarterly capital. They are stages within the counterfactual: starting conditions, adverse path, accumulated pressure, minimum, and modeled recovery.
The minimum is the most important state for a resilience question because it captures peak modeled pressure. An ending ratio alone could conceal a deeper decline earlier in the horizon. Conversely, the minimum should not be detached from the path that produces it. If a bank’s ratio reaches its low point because losses accumulate before revenue offsets appear, the timing tells the reader something about the mechanism. If the ratio recovers because the scenario’s later conditions support income, that recovery is conditional too.
The five states in the path should not be mistaken for five predictions. The starting state is the observed base used by the exercise. The stress-path state is the set of hypothetical macro-financial conditions. The projected-effects state is where those conditions become modeled losses, revenue, expenses, and other capital effects. The minimum state identifies the point of greatest modeled pressure. The conditional-end state shows what the same assumptions imply after that point. Each transition is a calculation under the exercise’s rules, not a claim that the real economy will move through the same sequence.
The Federal Reserve’s decomposition of the 2026 result shows why the path is not a simple response to “the recession.” Projected capital was reduced by greater loan losses associated with loan growth and more severe paths for some variables, including corporate spreads and commercial-real-estate prices. It was also reduced by smaller projected unrealized gains on available-for-sale securities because the 2026 scenario had smaller projected declines in interest rates, leaving interest rates higher through the projection horizon. At the same time, projected capital was supported by higher net interest income, reflecting recent bank financial performance and the path of interest rates in the hypothetical scenario.
The interest-rate example is especially useful because it defeats a one-direction story. Lower interest rates can support the value of some securities, while higher rates can support some interest income. In the 2026 exercise, the smaller projected decline in rates meant smaller unrealized gains on certain available-for-sale portfolios, but higher interest-rate levels also contributed to higher projected net interest income. One scenario variable can therefore push different components in opposite directions. The final capital result records the combined effect, not a simple “rates up is good” or “rates down is good” rule.
The loss total provides another view of the same path. The 32 banks were projected to absorb nearly $708 billion in losses. About $625 billion came from loan losses; $37 billion from trading and counterparty losses; $7 billion from securities losses; and $39.5 billion from other losses or gains. Against this, projected pre-provision net revenue was $719 billion, but cumulative pre-tax net income was only $1.2 billion. The numbers are not a forecast of ordinary earnings or realized losses. They show how large offsetting flows can leave a modest net result under an intentionally severe set of conditions.
The report’s headline capital outcome is consequently a summary of a moving system. Loan balances, scenario severity, market prices, interest rates, income, provisions, securities effects, and other items enter at different points. The aggregate ratio reaches 11.2 percent because these forces combine in the specified model over the specified horizon. It returns to 12.7 percent because the same modeled path later produces enough offsetting effects to lift the ratio. Neither number has meaning outside the conditions that generated it.
That is also why the 2026 result should not be read as a single verdict on all banks. The scenario creates common conditions, but the next question is how each bank’s portfolio and business model receives them. A common recession can produce different loan losses, income cushions, and capital minima when the underlying exposures differ. The next chapter turns from the shared hypothetical world to those differentiated channels.