What it means
The central insight is about correlation, which measures how closely two investments move together. Two holdings that both return 8% but fall at different times produce a smoother combined result than either one alone, and that smoothing is the free benefit diversification offers.
Risk in this framework is measured as volatility, usually the standard deviation of returns. That is a deliberate simplification: it treats an unexpected gain as risk in the same way as an unexpected loss, which is one of the theory's most frequently made criticisms.
Plotting every possible mix of assets produces a curve called the efficient frontier. Any portfolio on that curve gives the highest expected return available for its level of risk, and any portfolio below it is inefficient because a better combination exists.
The theory also distinguishes two kinds of risk. Specific risk attaches to one company and can be diversified away, while market risk affects everything at once and cannot be, which is why adding a fiftieth stock helps far less than adding the fifth.
Applied outside investing, the same logic explains why concentration is dangerous in a business. A company with one customer, one supplier and one product has a portfolio of correlated exposures, and revenue smoothing works the same way whether the assets are shares or contracts.
The main practical limitation is that the inputs are estimates. Expected returns, volatilities and correlations are all measured from the past, and correlations in particular tend to rise sharply during a crisis, exactly when the diversification benefit is most needed.
In practice
Real-world examples
Example 1
A pension trustee reviews a fund holding only domestic bank shares. Replacing a third of the holding with overseas industrials and index-linked bonds cuts the fund's measured volatility from 19% to 13% while reducing expected return by less than one percentage point.
Example 2
A family office adds an infrastructure fund with an expected return of 7% to a portfolio expecting 9%. The addition lowers overall expected return slightly but reduces volatility more, because infrastructure income has historically moved differently from listed equity.
Example 3
A treasurer applies the same thinking to bank deposits, spreading $12,000,000 across four institutions rather than one. The expected interest income falls by roughly $15,000 a year, which the board accepts as the cost of removing a single point of failure.
Think of it
Imagine packing for a trip where you do not know the weather. Instead of packing only heavy winter coats or only light summer t-shirts, you pack a balanced mix so you are prepared for rain, shine, or snow without carrying unnecessary weight.
Formula
Calculation
Portfolio return = w1 x r1 + w2 x r2
Portfolio variance = w1^2 x sd1^2 + w2^2 x sd2^2 + 2 x w1 x w2 x sd1 x sd2 x correlation
An investor holds 60% in an equity fund with an expected return of 10% and a standard deviation of 18%, and 40% in a bond fund with an expected return of 5% and a standard deviation of 8%. The correlation between them is 0.2.
The expected return is (0.6 x 10%) + (0.4 x 5%) = 6% + 2% = 8%.
The variance is (0.36 x 0.0324) + (0.16 x 0.0064) + (2 x 0.6 x 0.4 x 0.18 x 0.08 x 0.2), which is 0.011664 + 0.001024 + 0.0013824 = 0.0140704. The square root of 0.0140704 is 0.1186, so portfolio risk is about 11.86%.
The weighted average of the two standard deviations would be (0.6 x 18%) + (0.4 x 8%) = 14.0%. The portfolio therefore delivers its 8% expected return at 11.86% risk rather than 14.0%, and that 2.14 percentage point reduction is the diversification benefit.Case study
Seen in the real world
The following is an illustrative and entirely fictional example. Thorncastle Endowment, an invented university fund of $80,000,000, held 90% of its assets in the shares of companies in a single technology cluster because three donors had made their money there.
Reported returns had been excellent for six years, averaging around 14%. A new investment committee measured the fund's volatility at 26% and calculated that a mix holding 55% in a global equity index, 30% in bonds and 15% in property would have produced an expected return of about 8% at roughly 11% volatility.
In this fictional scenario the committee rebalanced over eighteen months. The endowment gave up part of its expected return, but when the technology cluster fell 35% two years later the diversified fund lost around 12%, and the university did not have to cut the $3,200,000 of annual scholarships the endowment funded.
Watch out
Common mistakes
- Treating the number of holdings as the measure of diversification, when twenty shares in the same sector move together and offer very little real spread of risk.
- Assuming that historical correlations will hold, when correlations between risky assets tend to rise towards one during a market crisis.
- Judging a single investment on its own volatility rather than on what it does to the volatility of the whole portfolio.
Questions
People also ask
Does diversification guarantee a lower loss?
No, it reduces the impact of problems specific to one company or sector, but market risk affects everything at once and cannot be diversified away.
Is a portfolio on the efficient frontier the right one to hold?
Only if its risk level matches the investor's tolerance and time horizon, since the frontier shows the best available trade-off rather than a single correct answer.
Why do critics object to using standard deviation as risk?
Because it treats upside surprises as risk, and because real returns are not perfectly bell-shaped, with extreme losses occurring more often than the model implies.