Historical stress scenarios: reading the paths and contributions
This guide uses 60% SPY + 40% AGG at the beginning of a historical window, with unchanged quantities afterwards. The example uses real price history and illustrates the method only. Example portfolio, not a recommendation. The comparison line is SPY, an exchange-traded fund (ETF) that tracks the S&P 500. An ETF is a fund traded on an exchange; here it represents a broad market rather than an individual company's history.
What the tool answers
The tool asks how today's holdings would have moved through a past market shock. Each window begins with today's value weights, then lets prices change the mix without rebalancing it. This keeps the starting composition comparable across episodes. It does not reconstruct the transactions made at the time, or assume that today's holdings existed in your account then. The result is a replay of the past, not a forecast.
One episode, two curves
This example shows the 2025 tariff shock only. The Jade card offers all five episodes; the four other buttons here are greyed out and cannot be selected. Both curves use the same dates and have zero at the S&P 500 peak. The example's line and SPY's line describe different mixtures over the same market episode, so a gap between them reflects that difference in composition.
- 60% SPY + 40% AGG
- SPY
The shock in percentages and dollars
The example's peak-to-trough return is -11.40%, compared with -18.76% for SPY. Applied to USD 100,000, the example change is -$11,395. That dollar figure gives the percentage a scale; it is not an account balance or a separate calculation of loss. The table also reports the largest drawdown across the entire window, alongside its own high and low dates. All figures describe this fixed example.
| Metric | 60% SPY + 40% AGG | SPY |
|---|---|---|
| Peak to trough | -11.40% -$11,395 | -18.76% |
| Largest drawdown in the window | -11.40% Feb 19, 2025 – Apr 8, 2025 | -18.76% Feb 19, 2025 – Apr 8, 2025 |
How the window is defined
A window runs from six months before the S&P 500 closing peak to six months after its closing trough. Here that is Aug 19, 2024 to Oct 8, 2025. The dashed vertical line marks Feb 19, 2025, and the dotted line marks Apr 8, 2025. Zero on the chart means the value at the peak, not the value at the beginning of the window. Values before the peak can therefore be below zero even before the named shock begins.
Peak to trough and largest drawdown
Peak to trough fixes the dates using the S&P 500 episode. Largest drawdown searches for the deepest fall from any earlier high to a later low in the portfolio's own window. These are different questions. In this example both identify Feb 19, 2025 to Apr 8, 2025, and the example's largest drawdown is -11.40%. In other episodes, a portfolio concentrated in another market can reach its high or low on other days.
Reading the contribution bars
The contributions explain the total by asset class. They use each class's price change and its weight at the peak, rather than simply multiplying a loss by the starting weight. This matters because weights already drifted before the shock. The equity and bond contributions are -11.73 and +0.33 percentage points. Their sum equals the example's peak-to-trough return, apart from display rounding.
- The center line is zero. A bar extending left pulled the portfolio down; one extending right added to its value.
- A bar reflects both the asset class's movement and its share of the portfolio. SPY was 62.5% of the example at the peak.
- Lengths are scaled to the largest absolute contribution. The longest bar fills half the track; a shorter bar is a fraction of that length, not an absolute percentage of the track.
- The numbers add up to -11.40%. The bond contribution partly offsets the equity contribution in this episode; the bars describe that arithmetic without rating either asset class.
| Contribution to the peak-to-trough return, by asset class | Contribution (percentage points) |
|---|---|
| US broad-market equity | -11.73 |
| Bond fund | +0.33 |
| Total | -11.40% |
Proxies and beta
A holding without its own history at the window start follows its replay proxy ETF until its own history begins. Beta is the estimated strength of its movement relative to that ETF, and scales the proxy's returns. An index proxy loses the holding's individual swings, so a scenario can understate its losses. The card states the approximated share and shows an amber note when that share is 66% or more. This example uses the two funds' own histories.
Earlier episodes and substitute history
The 2000–2002 and 2008 episodes reach back before some ETFs existed. A fixed substitute chain supplies the first available candidate with the required dates, using mutual funds or indexes. Some substitutes are price indexes that exclude dividends, rather than total-return series with dividends reinvested. Exchange rates before 2021-09 come from a one-time FRED export; FRED is the Federal Reserve Economic Data service. The amber note above the table appears for an approximated share of at least 66%, any substitute, or any price index.
What is not shown
There is no recovery metric. The curve after the trough shows the following six months, but that segment does not establish when a portfolio recovered its earlier high. There is also no questionnaire reference line. The historical paths and contributions answer what happened to a defined mixture through an episode; they do not compare a loss with a person's questionnaire answers.
Limits of five past episodes
Only five past episodes are represented. Other market events, including events unlike any of these, are absent. Today's holdings also carry hindsight: their presence today can reflect what became known later. Prices, proxies and substitute series set further limits on what the calculation describes. History does not repeat in the same form, and a fixed historical path supplies no probability for the next shock.
References
Basel Committee on Banking Supervision, Principles for Sound Stress Testing Practices and Supervision (2009).
Berkowitz, J., A Coherent Framework for Stress-Testing, Journal of Risk 2(2), 2000.