Portfolio risk and the correlation trap

Portfolio Management·July 30, 2026·4 min read

A practice question gives two assets with volatilities of 20% and 10%, equal weights of 0.5, and a correlation of 0. A candidate averages the two volatilities and selects 15.00%, because the familiar expected-return method feels transferable. The correct portfolio standard deviation is 11.18%. The arithmetic is not difficult; the lost point comes from overlooking the one input that describes how the assets move together.

Expected return and risk use different machinery

Expected portfolio return is built from each asset's expected return and its weight. No interaction term is needed, since the calculation asks how much each holding contributes to the portfolio's average outcome.

E(Rₚ) = w₁E(R₁) + w₂E(R₂)

Risk is different because two volatile assets can rise and fall at the same time, move in opposite directions, or behave somewhere between those extremes. Their individual standard deviations therefore tell only part of the story.

σₚ = √(w₁²σ₁² + w₂²σ₂² + 2w₁w₂σ₁σ₂ρ₁₂)

The placement of the weights matters. Each weight is squared inside portfolio variance, and the square root is taken only after the variance terms have been combined. A simple average of standard deviations skips that structure entirely. It can look tidy on a calculator screen while answering a calculation that the portfolio formula never asked for.

The final term captures the interaction. Its correlation coefficient, ρ₁₂, links the assets' movements; leaving that term out changes the economic question rather than merely simplifying the calculation.

Correlation creates the diversification benefit

For a long-only portfolio in which both weights and both volatilities are positive, ρ₁₂ = 1 is the sole case where σₚ equals w₁σ₁ + w₂σ₂. The assets move together perfectly, so combining them provides no reduction relative to that weighted-volatility benchmark.

Whenever ρ₁₂ is below 1 under those conditions, portfolio standard deviation is lower than the benchmark. That gap is the diversification benefit. A correlation of 0 still produces one: it removes the correlation term from the variance calculation, but it does not turn portfolio risk into a direct average of the two standard deviations.

Working the zero-correlation portfolio

Correlation belongs in the setup even when its numerical effect is zero. That distinction is easy to miss: a zero-valued term has been considered and resolved, whereas an omitted term means the relationship between the assets was never evaluated.

The question's equal weights produce squared weights of 0.25. The 20% and 10% volatilities enter the variance calculation as 0.04 and 0.01, while ρ = 0 makes the final term zero.

σₚ = √(0.25 × 0.04 + 0.25 × 0.01 + 0) = √0.0125 = 11.18%

The 15.00% choice comes from 0.5 × 20% + 0.5 × 10%. It is a plausible distractor because every number has been used and the weighted-average structure looks legitimate, yet the supplied correlation has disappeared. Its absence is the warning sign.

Quick reference

The equality at ρ = 1 depends on the conditions attached to it: a long-only portfolio with positive weights and positive volatilities. Exam questions often hide the real test in those words. A statement that drops the conditions can be false even when its equation looks familiar, while a question that supplies correlation or covariance is signaling that the relationship between holdings belongs in the calculation.

This has the same shape as the after-tax cost of debt trap: both punish a familiar weighted average when one input requires special treatment before aggregation.

Charter5m is an independent study tool and is not affiliated with CFA Institute. CFA Institute does not endorse, promote, or warrant the accuracy or quality of this content. CFA® is a registered trademark of CFA Institute.