Diversification
Diversification only works if the components don't behave the same way. Ten stocks in the same industry are ten positions, but hardly any diversification.
Learning objective: After this lesson, you can judge whether a portfolio is genuinely diversified or just made up of many similar positions.
The point of diversifying is that not everything falls at once. So it's not about the number of positions, it's about how different they are.
Ten tech stocks fall together when the sector falls. That feels like diversification and isn't. Real diversification spans industries, countries, asset classes, and currencies.
The effect has a limit. The first ten to twenty well-spread positions deliver most of the benefit, after which the curve flattens sharply. Fifty individual stocks is mostly just work.
Important to know: in severe crises, many investments suddenly drop together. Diversification protects against the failure of individual holdings, not against a general downturn. Expect otherwise, and you'll be disappointed.
The variance of an equally weighted portfolio of n holdings is σ²/n + ((n−1)/n) · c, with σ² the average individual variance and c the average covariance. As n grows, the first term vanishes, and the second converges toward the average covariance. That means holding-specific risk can be fully diversified away, while shared market risk can't.
The marginal benefit falls fast. By far the largest share of achievable variance reduction happens in the first few positions, provided they're sufficiently uncorrelated. With high correlation, say within one industry, the effect is correspondingly smaller, since c then sits close to σ².
Empirically problematic is the instability of correlations. During stress periods, pairwise correlations between risk assets rise sharply, an effect known as correlation breakdown. Diversification is therefore weakest exactly when it's needed most. What typically remains effective in such phases are structurally different holdings, like safe government bonds or cash.
Summary
- What matters isn't the count, it's how dissimilar the positions are.
- After roughly twenty well-spread positions, adding more barely helps.
- In crises, correlations rise, exactly when diversification works worst.
Did you get it?
Which risk can be diversified away, and which can't?
Holding-specific risk vanishes as the count rises; shared market risk remains.
Why does ten stocks in the same industry provide little diversification?
Because their covariance sits close to their individual variance. They move together.
What is correlation breakdown?
During stress periods, correlations between risk assets rise, which is exactly when diversification works worst.
Check your understanding
Sources and further reading
- The classic study by Evans and Archer (1968) in the Journal of Finance empirically examined how a stock portfolio's dispersion falls as the number of randomly chosen holdings grows, founding the debate over how many positions are actually needed. View source ↗
Related
- CorrelationStage 4
- The one-percent ruleStage 2
- Win rate and expected valueStage 2
- Matching questions for this stageQuestions