Diversification Benefit Versus Average Standalone Risk

What Does “Diversification Benefit” Compare?

A portfolio combines assets whose returns may move together or differently. Even when the holdings and their weights are fixed, the portfolio’s variance can be lower than the weighted average of the assets’ individual variances. The difference measures a specific diversification benefit: it compares portfolio variance with the portfolio-weighted average of standalone variances, not with an unweighted average unless the weights happen to be equal.

Throughout, consider returns over the same period. Let \(R_i\) be the return on asset \(i\), \(w_i\) its portfolio weight, and \(R_p=\sum_{i=1}^n w_iR_i\) the portfolio return. Assume the weights are fixed, sum to one, and are nonnegative, and that all returns have finite variances. Write \(\sigma_i^2=\operatorname{Var}(R_i)\). No particular probability distribution is required; the calculations use only variances and covariances.

This comparison is meaningful when the weights are nonnegative: each standalone variance then receives a nonnegative weight. It is not the same as comparing the portfolio’s variance with the unweighted average of asset variances, and it is not itself a comparison of standard deviations. Variance is measured in squared-return units; standard deviation is measured in return units.

The Pairwise Variance Identity

The gap can be expressed entirely in terms of pairwise return differences. This form explains why imperfect co-movement can reduce portfolio variance.

To justify the identity, expand the variance of a weighted sum:

Subtract this expression from \(\sum_i w_i\operatorname{Var}(R_i)\). Since the weights sum to one, \(w_i-w_i^2=w_i\sum_{j\ne i}w_j\). Thus each asset’s remaining variance term can be allocated across its pairs with the other assets:

The last equality follows from the variance-of-a-difference formula. With nonnegative weights, every term on the right is nonnegative. Therefore the portfolio variance cannot exceed the portfolio-weighted average standalone variance. It is strictly lower if at least one pair with positive weights has a return difference with positive variance.

The equality condition deserves care. The gap is zero exactly when \(\operatorname{Var}(R_i-R_j)=0\) for every pair with positive weights. A variance of zero means the difference is constant with probability one; it does not require the two returns to be identical. For example, if one return is always 1 percentage point higher than another, their difference is constant and has zero variance.

How Correlation Determines the Pairwise Benefit

For assets with positive standard deviations, define their correlation as \(\rho_{ij}=\operatorname{Cov}(R_i,R_j)/(\sigma_i\sigma_j)\). Then:

If \(\rho_{ij}<1\), this pairwise difference variance is positive whenever both standard deviations are positive. Consequently, if two positively weighted, risky assets have correlation below one, their pair contributes a strictly positive amount to the variance gap. Correlation alone does not determine the size of the benefit: the weights and the assets’ volatility levels also matter. If an asset has zero standard deviation, its correlation with another asset is undefined, so use the variance-of-the-difference formula directly.

A separate but related result compares standard deviations. For nonnegative weights, covariance cannot exceed the product of standard deviations, so:

This is a comparison with the weighted average of standalone standard deviations. It is not obtained by taking the square root of the weighted average variance: in general, \(\sqrt{\sum_i w_i\sigma_i^2}\) and \(\sum_i w_i\sigma_i\) are different quantities.

Several Assets: Add the Pairwise Contributions

With more than two assets, each pair contributes \(w_iw_j\operatorname{Var}(R_i-R_j)\) to the total gap. A pair with a small weight product may contribute little even if its returns differ substantially. Conversely, a heavily weighted pair with materially different movements can account for a large part of the reduction.

When the Comparison Can Fail

The nonnegative-weight condition matters. If short selling is permitted, a weight can be negative, and some products \(w_iw_j\) in the pairwise identity are negative. The identity remains algebraically true for weights summing to one, but it no longer guarantees that the portfolio variance is below the weighted sum of standalone variances. That weighted sum may also be a poor description of an “average” when some weights are negative.

Practical Interpretation and Exam Traps

The identity is useful because it separates the size of the variance reduction from the portfolio’s expected return. Expected returns matter for evaluating a portfolio’s reward, but the gap here is determined by weights and the variability of pairwise return differences. A lower correlation generally increases that difference variance when asset volatilities are positive, but it does not by itself establish which of two portfolios has the better return-risk tradeoff.

When interpreting a result, check the comparison and units before drawing a conclusion:

• Use the portfolio weights. The benchmark is \(\sum_i w_i\sigma_i^2\), not automatically the simple average of the variances.
• Keep variances and standard deviations distinct. A variance gap is in squared-return units. If reporting standard deviations, take the square root of the portfolio variance only after calculating it.
• Check the correlation-to-covariance conversion. Use \(\operatorname{Cov}(R_i,R_j)=\rho_{ij}\sigma_i\sigma_j\).
• Check the weights’ signs. Nonnegative weights make every pairwise term nonnegative; short positions can reverse the inequality.
• State the equality condition precisely. A zero gap requires every positively weighted pair’s return difference to have zero variance, not necessarily identical returns.

For a long-only portfolio, the pairwise identity provides both a proof and a calculation check. Calculate the portfolio variance by the covariance formula, calculate the gap by summing pairwise difference variances, and confirm that the two approaches agree. This algebraic view of diversification is aligned with the SOA VEE Accounting and Finance syllabus.