Diversification is the one principle nearly everyone agrees on. The mechanism behind it, and the conditions under which it stops working, receive far less attention.
The actual mechanism
Combining assets that do not move together produces a portfolio whose volatility is lower than the weighted average of the individual volatilities.
Which is a mathematical result rather than a heuristic, and it depends entirely on correlation being less than one.
If two assets moved identically, combining them would achieve nothing. The benefit comes precisely from the extent to which they differ.
The diminishing return
Adding a second holding to a single holding reduces risk substantially.
Adding a fiftieth to forty-nine changes very little.
Studies of how many holdings are needed to capture most of the available benefit generally land at figures far lower than most portfolios contain, with the exact number depending on how correlated the holdings are.
Which means very large numbers of holdings deliver limited additional diversification and considerable complexity.
What cannot be diversified away
Risk divides into the specific and the systematic.
Specific risk — a company loses a contract, a factory burns, a management failure — affects one holding and is reduced by holding many.
Systematic risk — a recession, a rate shock, a war — affects everything and is not reduced by holding more of the same market.
Which is why a portfolio of two hundred shares in one market still falls substantially in a downturn, and why people who thought they were diversified discover they were not.
Correlations change
The most important practical limitation.
Correlations measured over calm periods are frequently poor guides to correlations during stress.
Assets that behaved independently for years have repeatedly moved together during crises, as participants sell whatever can be sold to raise cash.
Which means the diversification benefit is smallest exactly when it is needed most.
This has been observed in enough crises to be considered a general property rather than a coincidence.
The bond and equity relationship
The most consequential example.
For an extended period, government bonds rose when equities fell, which made a mixed portfolio far less volatile than either component.
That relationship depended on the dominant shock being growth-related, where weak growth hurts equities and helps bonds.
When the dominant shock became inflation, both fell together, and mixed portfolios delivered their worst results in decades.
Which was not a failure of diversification. It was a change in what was driving markets, and the relationship was never a law.
Geographic diversification
Less effective than it once was, since markets have become more integrated and large companies operate globally regardless of where they are listed.
Which means a company's revenue geography frequently matters more than its listing location, and portfolios that look internationally diversified by listing may not be by exposure.
Currency adds a further layer — a foreign holding carries currency exposure that may add to or offset the underlying movement, and hedging that exposure is a separate decision with its own cost.
Diversification across time
A different dimension and a real one.
Investing gradually rather than all at once diversifies across entry points, reducing the consequence of a single bad moment.
The counterargument is that markets rise more often than they fall, so delaying investment has an expected cost.
Both are correct. The choice is between a better expected outcome and a narrower range of outcomes, which is a preference rather than a calculation.
The honest summary
Diversification reduces the risk of catastrophic single-holding loss, which is its main and sufficient justification.
It does not protect against broad market declines, does not work reliably in crises, and depends on relationships that change.
Understanding those limits is what prevents the disappointment of discovering them during an event.
This describes a concept and is not advice about how anyone should construct a portfolio, which depends on circumstances best discussed with a regulated adviser.
Factor exposure
A dimension of diversification that holdings counts do not capture.
Research has identified characteristics — company size, valuation, profitability, price momentum — that explain a substantial share of return differences across shares.
Which means a portfolio of many holdings that all share a characteristic is concentrated in that characteristic, regardless of how many names it contains.
A portfolio of fifty fast-growing technology companies is not diversified in any meaningful sense, and the number of holdings actively conceals that.
Examining what the holdings have in common is more informative than counting them.
Correlation of the household
The exposure most often overlooked entirely.
Human capital — the value of future earnings — is generally the largest asset a working person holds, and it is undiversified by definition, concentrated in one employer and one industry.
Holding shares in that employer, or heavily in that industry, compounds a concentration that already exists.
Which is the argument against large holdings of employer stock, and the reason employee share schemes deserve more scrutiny than their favourable terms suggest.
The same logic extends to property, where a home in the same region as the employer links two large exposures to one local economy.