Skip to content
Menu
Finabase Learn

Measure and manage portfolio risk Lesson 11 of 14

Portfolio and risk · Managing a portfolio

Stress testing a portfolio: what would a past crisis cost?

How to apply the returns of five past market crises to a portfolio’s current weights, with four example portfolios and the limits of scenario analysis.

On this page
  1. The idea
  2. How it works
  3. Using it
  4. Where it fails or misleads
  5. Compared with its neighbours
  6. How Finabase does it
  7. Frequently asked questions
  8. Summary
  9. Check your understanding
  10. Exercise

A stress test asks a direct question: if a particular bad event happened again, how much would this portfolio lose? It takes the portfolio as it stands today, applies the returns each of its holdings had (or would plausibly have) in a chosen scenario, and adds them up. The answer is a number in dollars for a named event.

That is its advantage over statistical measures. Volatility and Value at risk (VaR): the loss a portfolio should rarely exceed describe ordinary days and assume the future resembles a sample of the past. A stress test makes no statement about probability. It says what the damage would be, and leaves the investor to judge whether it could be borne. Banks have been required to run stress tests since the 2008 crisis. The arithmetic for an individual's portfolio takes a few minutes.

The idea

A portfolio's history can mislead about its present. A fund bought five years ago may have doubled; a position may have grown from 5% to 30%; the mix that sailed through 2020 may not be the mix held now. The portfolio's own past drawdown describes a portfolio that no longer exists.

A stress test fixes this by separating two things: what the assets did in a crisis, which is history, and how much of each is held, which is today.

Four portfolios put through five past sell-offs 60/40 Growth Spread Income −28 −50 −16 −16 2007–09 crisis −8 −15 −2 2 2011 euro scare −7 −16 −4 −2 Late 2018 −11 −18 −7 −6 2020 pandemic −20 −28 −21 −22 2022 rate rise Percent. Today’s weights times each fund’s total return over the scenario’s months (GLD price only).
Return of four example portfolios over five historical episodes, from fixed weights applied to each asset’s total return over the episode. Monthly data.

The figure runs four portfolios through five episodes. No portfolio is safest throughout. The ranking changes with the scenario, and the scenario that was worst for the cautious portfolios was different from the one that was worst for the aggressive one.

How it works

Three steps: list current weights; choose scenarios and find each asset's return in them; multiply and add.

Step one: the scenarios

The data: monthly total returns (GLD price only), within January 2005 to October 2026.

Asset 2007–09 crisis (Nov. 2007 – Feb. 2009) 2011 euro scare (May – Sep. 2011) Late 2018 (Oct. – Dec.) 2020 pandemic (Feb. – Mar.) 2022 rate rise (Jan. – Sep.)
SPY (US large stocks) −50.8% −16.2% −13.5% −19.4% −23.9%
QQQ (Nasdaq-100) −49.7% −10.8% −16.7% −12.9% −32.5%
EFA (foreign developed stocks) −57.4% −23.3% −12.6% −20.8% −27.1%
IWM (US small stocks) −51.6% −25.0% −20.3% −28.4% −25.1%
AGG (US bonds) +6.5% +4.8% +1.8% +1.0% −14.4%
TLT (long Treasuries) +20.4% +30.8% +4.6% +13.4% −29.9%
GLD (gold) +17.8% +3.7% +7.5% −0.9% −9.5%
VNQ (US property trusts) −64.7% −16.2% −6.4% −24.9% −29.3%
Two crises, eight funds: what protected in one did not in the other 2007–09 crisis (Nov. 2007 – Feb. 2009) 2022 rate rise (Jan. – Sep. 2022) −51% −24% SPY −50% −32% QQQ −57% −27% EFA −52% −25% IWM 6% −14% AGG 20% −30% TLT 18% −10% GLD −65% −29% VNQ Total return of each fund over the scenario’s months (GLD price only).
Total return of eight asset-class funds in two episodes: November 2007 to February 2009 and January to September 2022. Monthly data; GLD price only.

The two crises in the second figure were different in kind. In 2007 to 2009 a financial panic sank everything tied to growth and credit, and investors fled to government bonds and gold. In 2022 inflation and rising interest rates lowered the value of nearly every asset at once. A portfolio built to survive the first could be badly placed for the second.

Step two: four portfolios

Portfolio Holdings 2007–09 2011 Late 2018 2020 2022
60/40 SPY 60%, AGG 40% −27.9% −7.8% −7.4% −11.3% −20.1%
Growth QQQ 50%, SPY 30%, IWM 20% −50.4% −15.3% −16.5% −18.0% −28.4%
Spread SPY 30%, EFA 15%, AGG 25%, TLT 15%, GLD 15% −16.5% −2.0% −3.7% −6.8% −20.7%
Income AGG 40%, TLT 20%, VNQ 20%, SPY 20% −16.4% +1.6% −2.3% −5.8% −22.4%

On $500,000 the same results in dollars:

Portfolio 2007–09 2011 Late 2018 2020 2022
60/40 −$139,369 −$38,957 −$36,925 −$56,253 −$100,472
Growth −$252,072 −$76,331 −$82,451 −$89,915 −$142,202
Spread −$82,437 −$9,796 −$18,381 −$34,037 −$103,711
Income −$82,076 +$8,073 −$11,669 −$28,896 −$111,810

Checking one cell by hand: the 60/40 portfolio in 2007 to 2009 is 0.60 × (−50.8%) + 0.40 × (+6.5%) = −30.5% + 2.6% = −27.9%.

Step three: reading the result

The dollar figure is the point. "A 50% decline" is abstract. "$252,072 of $500,000" is a figure that can be set against a mortgage, a retirement date or a tuition bill.

Diversification helped in four scenarios and not in the fifth. The Spread portfolio lost 16.5% in 2007 to 2009 against 27.9% for 60/40. In 2022 it lost 20.7% against 20.1%. Its bonds and gold, the source of its earlier protection, fell with its stocks.

The cautious portfolio had the highest loss in one scenario relative to its own risk. The Income portfolio's long-run volatility was 8.7%, the second lowest of the four. Its 2022 loss, 22.4%, was greater than the 60/40 portfolio's. Low measured volatility did not mean a low loss in the scenario that struck its particular holdings.

These figures hold weights fixed through each episode and so differ slightly from what a rebalanced or untouched portfolio would have experienced. Over the full period, rebalanced monthly, the four portfolios' deepest actual falls were 32.3%, 50.2%, 22.3% and 23.1%.

Using it

  1. Use today's weights, including holdings in every account. Drift is what a stress test is best at exposing.
  2. Map each holding to an asset with a long history. A fund launched in 2019 has no 2008 return. Use the nearest index fund that existed: a technology fund maps to QQQ, a corporate bond fund to something riskier than AGG. Individual stocks need care, since a single company can do far worse than its sector.
  3. Run several scenarios of different kinds: a financial crisis, a rate shock, a fast crash, a regional scare. The worst one for a given portfolio is the informative one.
  4. Add hypothetical scenarios. History is a short list. Ask what would happen if stocks fell 40% while bonds also fell 15%, or if the portfolio's largest holding halved. The formula works the same with invented returns.
  5. Convert to money and to time. Compare the dollar loss with spending needs over the next few years, and remember that the 2007 to 2009 fall took about four years to recover in a stock index.
  6. Decide beforehand what would be done. If the answer to the worst scenario is "I would have to sell", the allocation is too aggressive now. That is the cheapest time to learn it.

Where it fails or misleads

  • The next crisis will not be on the list. Each past scenario had its own cause. A stress test built from them can miss what is new: before 2022, no scenario in most investors' memory had stocks and bonds falling 15% to 25% together.
  • Month-end figures understate. The episodes here are measured between month-ends. The low on the worst day was lower, and an investor watching daily saw a larger loss.
  • Proxies hide specific risk. Mapping a portfolio of twelve stocks to SPY assumes they behave like the index. A concentrated portfolio can lose much more; see Concentration risk: when one holding dominates a portfolio.
  • Returns are not the only stress. A scenario also brings wider spreads, margin calls, job losses and withdrawals. A test of prices alone leaves out Liquidity risk: the cost of selling when you have to and the possibility of being forced to sell.
  • It gives no probability. A stress test cannot say whether a scenario is a once-in-five-years or once-in-fifty-years event. It complements statistical measures and does not replace them.
  • Sixteen months is treated as one step. Applying fixed weights across a long episode ignores what the investor does in the middle, which in practice is where most of the harm or good is done.

Compared with its neighbours

Method Question Gives a probability? Uses current weights?
Stress test What would a named event cost? No Yes
Maximum drawdown What was the worst fall this portfolio actually had? No No, uses its history
Value at risk What loss is exceeded only rarely on ordinary days? Yes Yes
Monte Carlo simulation What range of outcomes do assumed returns produce? Yes, given the assumptions Yes
Sensitivity analysis What does a one-unit move in one factor do? No Yes

How Finabase does it

Finabase does not run stress tests, store scenarios or apply historical returns to a portfolio.

It supplies both inputs. The portfolio page lists every position with its market value, from which current weights follow. For scenario returns, a symbol's chart at the five-year or maximum range, or its History tab, gives prices at the start and end of any past episode, and the percent change between them is the asset's return (before dividends). The comparison view shows several asset-class funds as percent changes over a common range, which is a quick way to see which holdings fell together in a recent episode; its ranges extend to two years, so older crises need the chart or the history table. Multiplying weights by returns is then a few lines in a spreadsheet.

Frequently asked questions

What is a portfolio stress test?

A calculation of what a portfolio, as currently held, would lose in a specified scenario, such as a repeat of the 2008 financial crisis or a sharp rise in interest rates.

How is a stress test different from value at risk?

Value at risk estimates a loss that is exceeded with a given small probability under normal conditions. A stress test assumes an extreme event has happened and measures the loss, without saying how likely the event is.

Which scenarios should I use?

At least one deep stock-market crisis (2007 to 2009), one in which bonds fell as well (2022), and one fast crash (2020). Add a hypothetical in which your largest holding falls by half.

My portfolio did fine in 2020. Is that a stress test?

Only of the portfolio you held then. If the weights have changed, apply the 2020 returns to today's weights. And 2020 was one kind of crisis, brief and followed by a rapid recovery.

What should I do if the result is worse than I can accept?

The result is information about the allocation, not a prediction. The usual responses are to hold a smaller share of the riskiest assets, to diversify across assets that fail in different scenarios, or to hold more cash for near-term needs.

Summary

A stress test multiplies a portfolio's current weights by each asset's return in a chosen scenario and reports the total, best expressed in dollars. Run on four example portfolios, it shows a 60/40 mix losing 27.9% in 2007 to 2009 and an all-stock growth mix 50.4%, and it shows the diversified and income portfolios, which held up in four episodes, losing more than 20% in 2022. Use today's weights, several different kinds of scenario and at least one invented one. The test cannot say what will happen. It shows what the portfolio is exposed to while there is still time to change it.

Check your understanding

5 questions on this lesson. Each answer comes with the reason, and nothing is scored.

  1. A portfolio is 70% stocks and 30% bonds. In a scenario, stocks return −40% and bonds +5%. What is the portfolio’s return in the scenario?

    1. −35.0%
    2. −26.5%
    3. −17.5%
    Show the answer

    B. −26.5% 0.70 × (−40%) + 0.30 × (+5%) = −28% + 1.5% = −26.5%.

  2. Why does a stress test use the portfolio’s current weights instead of its own past performance?

    1. Because past performance is not allowed in calculations
    2. Because the holdings may have drifted or changed, so the past drawdown describes a different portfolio
    3. Because current weights are always more conservative
    Show the answer

    B. Because the holdings may have drifted or changed, so the past drawdown describes a different portfolio A position that has grown, or a fund added recently, changes the exposure. Applying historical asset returns to today’s weights measures the risk actually held now.

  3. The diversified "Spread" portfolio lost far less than 60/40 in 2007–09 and slightly more in 2022. What does that show?

    1. Diversification never works
    2. Protection depends on the scenario: bonds and gold rose in a financial panic and fell in a rate shock
    3. The calculation was wrong
    Show the answer

    B. Protection depends on the scenario: bonds and gold rose in a financial panic and fell in a rate shock Different crises hit different assets. Running several kinds of scenario reveals which one a given portfolio is most exposed to.

  4. What can a stress test not tell you?

    1. How much the portfolio would lose in the scenario
    2. Which holdings contribute most to the loss
    3. How likely the scenario is to happen
    Show the answer

    C. How likely the scenario is to happen A stress test assumes the event and measures the damage. It assigns no probability, which is why it is used alongside statistical measures.

  5. A fund in the portfolio was launched in 2019. How can it be included in a 2008 scenario?

    1. Leave it out
    2. Assign it the return of the closest index or fund that existed then, noting the approximation
    3. Assume it would not have fallen
    Show the answer

    B. Assign it the return of the closest index or fund that existed then, noting the approximation Mapping to a proxy with a longer history is standard practice. It is an approximation, and a specialized fund may behave worse than its proxy.

Exercise

One task to do in Finabase. Nothing is graded or saved.

Open the comparison of SPY, AGG, TLT and GLD and read the 6-month return row. Treat the last six months as a scenario and apply the lesson’s formula to two portfolios: 60% SPY with 40% AGG, and 25% in each of the four funds. Multiply each weight by its fund’s return and add. Turn both results into dollars on $500,000. Which portfolio came out ahead, and did the bonds and gold offset the stocks or move with them?

Compare SPY, AGG, TLT and GLD
What you should find

The answer depends on the half-year, and that is the point: no mix is safest in every scenario. If bonds and gold rose while SPY fell, the four-way mix lost less, as in 2007 to 2009. If all four fell together, spreading helped little, as in 2022. The figures are price changes without dividends, and one recent scenario says nothing about what the next one will look like.

Sources and further reading

  1. Basel Committee on Banking Supervision, Stress Testing Principles (Bank for International Settlements, 2018)
  2. Board of Governors of the Federal Reserve System, Supervisory Stress Test Methodology
  3. Philippe Jorion, Value at Risk: The New Benchmark for Managing Financial Risk, 3rd ed. (McGraw-Hill, 2007) — chapter on stress testing
  4. Carmen M. Reinhart and Kenneth S. Rogoff, This Time Is Different: Eight Centuries of Financial Folly (Princeton University Press, 2009)
  5. Finabase chart data — monthly bars and dividends of SPY, QQQ, EFA, IWM, AGG, TLT and VNQ, and monthly bars of GLD, January 2005 to October 2026; statistics computed by Finabase