FinUties 101 · Risk · Intermediate
Correlation and its limits
Correlation measures how two assets move together, from -1 to +1. It sets how much diversification helps, but it changes over time and can rise in a crisis, when a portfolio needs it least.
In this lesson
- Read a correlation coefficient between -1 and +1
- Calculate the volatility of a two-asset portfolio for different correlations
- Explain why correlations rise in a crisis and why the figure is unstable
Read first: Volatility explained, Diversification: not keeping everything in one place
Diversification works because assets do not all move together. Correlation puts a number on how much they do. It is the tool behind most portfolio risk figures, and it has limits a careful reader should know.
What the number means
The correlation coefficient runs from -1 to +1.
- +1: the two move in lockstep, up and down together.
- 0: no linear relationship; knowing one tells you little about the other.
- -1: they move in opposite directions in proportion.
Real asset pairs sit in between. Shares of two companies in one industry are often strongly positive. Shares and government bonds have been weakly related or negative in some decades and positive in others.
How it changes portfolio risk
Take two assets, each with a volatility of 20%, held 50/50. The portfolio volatility is 20% x the square root of ((1 + correlation) / 2).
| Correlation | Portfolio volatility |
|---|---|
| +1 | 20% |
| +0.5 | about 17.3% |
| 0 | about 14.1% |
| -0.5 | 10% |
| -1 | 0% |
With perfect correlation, mixing adds nothing. At zero, volatility falls by almost a third. At -1, two risky assets cancel exactly, a case that does not exist in practice for long. Each drop in correlation helps, and the first extra, different asset helps more than the tenth.
Four limits
1. It is not stable. Correlation is estimated from a window of past data and changes with the window. A pair that looked independent for ten years can become tightly linked.
2. It rises in a crisis. When funding is scarce and investors need cash, they sell whatever is liquid, not whatever is cheap. In the global financial crisis of 2008 most risky assets fell together (see the global financial crisis). The diversification a portfolio showed in calm years was smaller than the figure suggested.
3. It misses non-linear links. The coefficient captures straight-line co-movement. Two assets can be unrelated on ordinary days and fall together on the worst ones.
4. A usual hedge can fail. For long periods, government bonds cushioned falling shares. In 2022, when inflation and rates rose together, both fell at once, which the 2022 rates shock describes. The hedge worked because of the cause of the fall, not because of a fixed law.
Correlation is not causation
A high correlation does not mean one asset drives the other. Both may respond to a third factor, such as interest rates or growth. And two series that merely trend in the same direction over time can show a high correlation with no link at all.
Common mistakes
- Treating the long-run average correlation as a promise for the next year.
- Counting positions rather than risk sources. Ten holdings tied to one theme are one bet.
- Using a correlation from rising markets to plan for falling ones.
- Assuming a negative correlation is stable. It can change sign when the cause of the moves changes.
In practice
Check correlations in the past stress periods, not only the full sample, and size positions on the assumption that related assets fall together. Look at how diversification held in the 2008 and 2022 episodes before relying on it.
Check yourself
Two assets have a correlation of 0. Does that mean they never fall together?
No. It means that on average there is no linear tendency to move together. They can still fall on the same day, and the relationship can change.
Each of two assets has a volatility of 20%. Held 50/50 with a correlation of 0.5, what is the portfolio volatility?
About 17.3%. Volatility times the square root of (1 + correlation) / 2: 20 x the square root of 0.75.
Why is a correlation measured in calm years a weak guide to a crisis?
Correlation is not fixed. In a crisis, investors sell what they can, and assets that moved independently often fall together. The calm-period figure understates the overlap.
Key terms
Sources
- Triumph of the Optimists: 101 Years of Global Investment Returns. Princeton University Press (Elroy Dimson, Paul Marsh, Mike Staunton), 2002