Correlation

Why correlation matters

Correlation connects individual asset risk to total portfolio risk. When imperfectly correlated assets are combined, gains in one can partly offset losses in another, creating diversification. Managers use it to evaluate new holdings, hedges, factor overlap, and concentration across mandates. The same expected return and volatility assumptions can produce very different portfolio outcomes when the correlation estimate changes, so it deserves the same scrutiny as return forecasts.

How it is applied

Analysts calculate correlation from paired returns over a chosen frequency and period. The choice should match the decision: daily data may help with liquid trading risk, while monthly data may suit strategic allocation. Rolling correlations show whether relationships are stable. Scenario and stressed correlations supplement historical estimates. For a portfolio, the correlation matrix is combined with weights and volatilities to calculate variance and to identify exposures that appear different by name but behave similarly.

Formula

Portfolio example

Suppose global equities and government bonds each have positive expected returns, with a measured correlation of 0.10. Combining them may produce lower volatility than the weighted average of their standalone volatility. If inflation unexpectedly rises, both could fall together and their short-term correlation might become positive. A risk report should therefore show the benefit under normal estimates and the loss under a less favorable stressed relationship.

How to interpret it

A low correlation does not automatically make an investment attractive. Expected return, volatility, liquidity, fees, and position size still matter. Correlation also measures linear co-movement, not causation. A value of 0.8 is strong positive co-movement in the sample, while -0.3 indicates a partial tendency to move oppositely, not a dependable hedge. Changes across windows can be more informative than one long-term average, especially around recessions or liquidity shocks.

Limitations and common misconceptions

Correlation is sensitive to the observation period, return frequency, outliers, currency, and valuation method. It can change abruptly and often becomes less favorable during crises. Smoothed private-market prices may understate true correlation with public markets. Nonlinear relationships and tail dependence can be missed entirely. Estimating many pairwise correlations from limited data also creates noise. Investors should combine statistical evidence with an economic explanation for why return drivers differ.