The Diversification Ratio

Everyone says diversification reduces risk, but "how much" almost never gets a number attached to it. A portfolio holding twenty stocks sounds diversified, but if all twenty move together during every market stress, the diversification is mostly cosmetic. The diversification ratio puts a single, precise number on how much genuine risk reduction a portfolio's mixing has actually bought, comparing the risk you would have if the assets moved independently of each other's cancellation effects against the risk you actually have.

The analogy

Picture a group of people carrying a wobbly table together. If everyone pushes and pulls in random, uncoordinated directions, most of their individual efforts cancel out and the table barely moves, a lot of "activity," very little net wobble. If instead everyone happens to push in exactly the same direction at the same time, none of the individual force cancels, and the table swings wildly, matching the sum of everyone's push. The diversification ratio measures where a portfolio sits between these two extremes: are the individual assets' "pushes" (their volatilities) mostly cancelling each other out, or are they all shoving in the same direction?

Building the formula

The weighted average of standalone volatilities is what the portfolio's risk would be if you simply summed everyone's individual wobble with no cancellation credit at all:

$\text{Numerator} = \sum_i w_i \sigma_i.$

In words: take each position's weight, multiply by that asset's own volatility, and add them all up, this is an upper bound on portfolio risk, achieved only if every asset moved in perfect lockstep.

The portfolio's actual volatility, accounting for how the assets really move together, is

$\text{Denominator} = \sigma_p = \sqrt{w^\top \Sigma w}.$

The diversification ratio is simply the first divided by the second:

$\text{DR} = \frac{\sum_i w_i \sigma_i}{\sqrt{w^\top \Sigma w}}.$

In words: how much risk you'd have with zero cancellation, divided by how much risk you actually have. $\text{DR} = 1$ means no diversification benefit at all (typically only when every asset is perfectly correlated, correlation of 1 with everything). $\text{DR} > 1$ means genuine risk reduction from mixing, and the further above 1, the more the portfolio's assets are cancelling each other's wobble rather than reinforcing it. A well-built, low-correlation multi-asset portfolio can post a diversification ratio of 1.5 or more.

Drag the correlation slider down here and watch the frontier bow further to the left. That leftward bow is the diversification ratio increasing in real time, the same weights and volatilities, but a lower correlation means more cancellation, so achievable portfolio volatility drops further below the naive weighted-average line.

Worked example 1: two assets, low correlation

Two assets, equally weighted ($w = 0.5, 0.5$), each with volatility 20%, correlation $\rho = 0.2$.

Numerator: $0.5 \times 20\% + 0.5 \times 20\% = 20\%$.

Portfolio variance: $w^\top \Sigma w = w_1^2\sigma_1^2 + w_2^2\sigma_2^2 + 2w_1w_2\rho\sigma_1\sigma_2 = 0.25(0.04) + 0.25(0.04) + 2(0.25)(0.2)(0.04) = 0.01+0.01+0.004 = 0.024$.

$\sigma_p = \sqrt{0.024} \approx 15.49\%$.

$\text{DR} = 20\%/15.49\% \approx 1.29$. Mixing two equal, moderately-correlated assets bought about a 29% reduction in risk relative to the naive weighted average, purely from the fact that they do not move in perfect lockstep.

Worked example 2: same weights, different correlation

Same two assets, same 20% volatilities and 50/50 weights, but now correlation $\rho = -0.4$ (genuinely offsetting, not just uncorrelated).

Numerator is unchanged at 20% (it does not depend on correlation at all, only on individual volatilities and weights).

Portfolio variance: $0.01 + 0.01 + 2(0.25)(-0.4)(0.04) = 0.02 - 0.008 = 0.012$, so $\sigma_p = \sqrt{0.012} \approx 10.95\%$.

$\text{DR} = 20\%/10.95\% \approx 1.83$. Flipping the correlation from $+0.2$ to $-0.4$, with every other input identical, pushed the diversification ratio from 1.29 to 1.83, a direct, quantitative illustration that the ratio is purely a function of how assets move together, not of how many assets there are or how big each one's individual risk is.

What this means in practice

The diversification ratio is the objective function behind the "maximum diversification portfolio" (see Maximum-Diversification Portfolio), an alternative to mean-variance optimization that sidesteps the need to estimate expected returns entirely, it only needs volatilities and correlations, both of which are far more stable to estimate than expected returns. Risk teams also monitor a book's diversification ratio over time as an early-warning signal: a ratio that has been quietly drifting toward 1 means the book's apparent variety of positions is providing less and less actual risk cancellation, often because correlations across the book have crept up.