Finance education, free and open.An independent learning project
THEFINANCE ARCHIVES
Finance foundationsEquitiesFixed incomeDerivativesEconomicsFinancial statementsPortfolio management
Home / Portfolio management / Correlation Breakdown and What Changes Under Stress
Portfolio management

Correlation Breakdown and What Changes Under Stress

In this chapter

Intermediate

A correlation breakdown is a material change in how investments move together compared with the relationship you used to plan. Two funds may usually offset each other and then move in the same direction during a funding shock. The term can also describe a previously strong relationship weakening. Its importance comes from what that change does to the portfolio.

Put the ordinary and stressed relationships side by side

The Diversification chapter shows why it helps to hold investments that respond differently. For a stress test, pick a specific event and ask what could pull their returns in the same direction. For example, a stock fund and a corporate bond fund can both be exposed to worries about companies’ ability to borrow or repay debt. They may respond differently in other conditions.

The inputs below are scenarios, with no probability attached. They describe hypothetical one-year nominal USD simple total returns including income, before fees and taxes. The portfolio holds 60% stock fund and 40% corporate bond fund, with fixed opening weights and no borrowing, shorting or outside flows.

Worked case. Calculate risk in ordinary conditions

Suppose the ordinary estimates are 18 percentage points (pp) of stock volatility, 8 pp of corporate bond volatility, and a correlation of −0.20. Multiply each asset’s spread by its weight. Call those weighted spreads \(a\) and \(b\). Let \(\rho\) be correlation and \(\sigma_p^2\) portfolio variance in pp².

\[ \sigma_p^2=a^2+b^2+2ab\rho \]

Here \(a=0.60(18)=10.8\) pp and \(b=0.40(8)=3.2\) pp. The cross term is \(2(10.8)(3.2)(-0.20)=-13.824\) pp². Adding \(10.8^2\) and \(3.2^2\) gives a variance of \(113.056\) pp² after the cross term. Its square root is 10.63 pp. The negative co-movement assumption reduces the spread below the value you’d get at zero correlation.

Separate the loss of offset from wider individual swings

Worked case. A funding shock hits both funds

Run two passes. First change only correlation to +0.70. The weighted spreads stay at 10.8 pp and 3.2 pp, and the cross term becomes +48.384 pp². Variance rises to 175.264 pp² and volatility reaches 13.24 pp. That increase is the effect of co-movement at the original spreads.

Now allow the funding shock to raise the two individual volatilities to 24 pp and 12 pp. The weighted spreads become 14.4 pp and 4.8 pp. Keeping +0.70, variance is \(14.4^2+4.8^2+2(14.4)(4.8)(0.70)=327.168\) pp². Volatility reaches 18.09 pp.

AssumptionsStock volatility (pp)Bond volatility (pp)CorrelationPortfolio volatility (pp)
Ordinary188−0.2010.63
Correlation stress only188+0.7013.24
Correlation and volatility stress2412+0.7018.09

If you were testing a 14 pp risk limit, the first stress would still be inside it, while the second would exceed it. The two passes show why changing the correlation alone can give a different decision. The difference between the stress rows measures the effect of wider individual swings at the stressed correlation. With the new individual volatilities, even zero correlation would leave variance of \(14.4^2+4.8^2=230.40\) pp², or about 15.18 pp. Staying below the 14 pp limit would now require an offset. So a recovery from +0.70 to zero would still leave this particular allocation outside its limit.

Decide what a short stress sample can tell you

One week when both funds lost money shows they fell together that week. Correlation needs variation across multiple pairs of observations. Even two distinct pairs with variation in both series give a perfect +1 or −1 mechanically. A short window can be dominated by one or two dates, by changing volatility, or by how you selected the window.

When using actual data, align both return intervals, keep the frequency and currency consistent, and state the number of pairs in each regime. A small stress sample can reveal a scenario worth testing. Use a range of plausible co-movements and volatilities for the decision. Claiming a permanent shift or a particular cause calls for more evidence.

Practice questions

Choose one answer per question. Round only the final result where needed.

1. Read a changed relationship

Ordinary correlation was −0.40. Several relevant stress tests use +0.60 with the original positive weights and volatilities. What happens to portfolio variance in those tests?

Question 1 choices

B is correct

  • A. The sign matters. A move from negative to positive raises the cross term.
  • B. A more positive correlation adds to the variance at fixed positive weights.
  • C. The own terms stay fixed while the cross term changes.

2. Translate a stress scenario

Two assets have weighted volatilities of 6 and 4 pp. If stressed correlation is +0.50, what is portfolio variance?

Question 2 choices

C is correct

  • A. This includes the two squared contributions only.
  • B. This uses a negative correlation of −0.50.
  • C. The variance is \(36+16+2(6)(4)(0.50)=76\).

3. Separate the inputs

Which comparison isolates the effect of changing correlation?

Question 3 choices

A is correct

  • A. Only the co-movement assumption changes.
  • B. This tests the effect of changing the asset spreads at a fixed correlation.
  • C. The combined effect would be measurable, but the individual causes would be mixed.

4. Interpret one stressful day

Both assets fell on one date. What can you conclude?

Question 4 choices

B is correct

  • A. One pair lacks the variation required to calculate Pearson correlation.
  • B. The date can motivate a scenario without measuring a correlation.
  • C. That conclusion needs a relevant set of matched observations and analysis.

5. Check a two-point reading

Two distinct observations show variation in both series and give a correlation of −1. How should you read it?

Question 5 choices

C is correct

  • A. Two observations give very little evidence about other periods.
  • B. Two points with nonzero variation lie exactly on a straight line.
  • C. The calculation can be correct even though the sample is too thin for that broader claim.

6. Compare with a risk limit

The volatility limit is 11 pp. A correlation-only stress gives 10.8 pp, and a joint stress gives 12.6 pp. What do these tests show?

Question 6 choices

A is correct

  • A. Compare each volatility directly with 11 pp.
  • B. Weights can stay fixed while joint return behaviour changes.
  • C. The 1.8-point difference is in volatility. Neither result supplies an expected return.

Find a finance concept

Press Enter to see all matching chapters.