Correlation Management

A portfolio can look diversified on paper and still behave like a single position. If most holdings rise and fall together, a bad day for one is a bad day for all. Correlation management means measuring how positions move together, understanding the combined risk and keeping it within limits. It matters for long term investors balancing asset classes and just as much for active traders, who often hold several trades that are really the same idea in different tickers.

Measuring correlation#

Always use returns, not prices, and choose a window that matches your holding period.

Combining correlated risks#

If you hold N positions, each risking the same amount r, the combined risk depends on their average correlation rho:

Combined risk ≈ r × Square root of (N + N × (N - 1) × rho)

Crisis correlations#

Correlations are not stable. In calm markets, stocks in different sectors may have moderate correlations; in a crash, many jump toward 1 as investors sell broadly. Stock and bond correlations have also changed sign over the decades: often negative from around 2000 to 2020, and positive in periods of high inflation such as the 1970s and 2022. Planning only with calm period correlations understates crisis risk. See Systemic Risk and Stress Testing and Scenario Analysis.

Practical controls#

Correlation is not the whole story#

Correlation measures linear co movement on average. Two assets can have low average correlation but crash together in rare events, a property called tail dependence. Examine joint behaviour in the worst historical periods, not only the full sample correlation. See Fat Tails and Expected Shortfall (CVaR).

Correlation between strategies#

Portfolio managers who run several strategies watch correlations between strategy returns, not just between assets. Two strategies trading different instruments may still be correlated if both are, for example, short volatility or long momentum. See Factor Timing, Crowding and Crashes and Combining Signals.

Common mistakes#

1. Counting positions instead of measuring correlation.
2. Using too short a window, producing noisy estimates.
3. Using calm period correlations for risk limits.
4. Ignoring common factors such as the US dollar or interest rates that link seemingly different trades.
5. Correlating prices instead of returns, which produces spurious results. See Stationarity, Differencing and Unit Roots.

Frequently asked questions#

What is correlation management?#

Measuring and controlling how much portfolio positions move together, so risk is not secretly concentrated in one common driver.

Why do correlations rise in a crisis?#

Investors sell risky assets broadly to raise cash and reduce risk, so assets that usually move independently fall together.

How do I reduce correlation in my portfolio?#

Add assets or strategies driven by different factors, limit exposure to common drivers, and size correlated positions smaller.

Next, compare ways to weight positions in Equal, Value and Volatility Weighting.