Diversification

The mathematics of diversification rests on correlation. If two investments each have the same expected return and the same volatility, a portfolio split between them has the same expected return as either, but its volatility depends on how their returns move together.

If they move in perfect step, the portfolio is as volatile as each; if they move independently, the portfolio is considerably less volatile, because one's bad year is as likely as not to coincide with the other's good year; if they move in opposite directions, the portfolio can be much less volatile still. Adding more holdings with imperfect correlation continues to reduce volatility, with diminishing effect, until only the risk common to all of them remains.

That common risk, the market risk or systematic risk, cannot be diversified away; the risk specific to each holding, the unsystematic risk, can be, and a portfolio of twenty to thirty well-chosen holdings eliminates most of it. This has a consequence for how risk is priced.

Since any investor can eliminate specific risk by diversifying, the market does not pay a premium for bearing it; only systematic risk, measured by beta, commands a higher expected return. An investor who holds a concentrated portfolio is bearing risk for which no reward is offered.

The same logic applies within a company's project portfolio and, with qualifications, to the company itself: shareholders can diversify across companies more cheaply than a company can diversify across businesses, so corporate diversification is justified by operating benefits, not by risk reduction alone. Corporate diversification takes several forms.

Related diversification moves into products or markets that share customers, technology, channels or capabilities with the existing business: a food company adding a new category, a software company adding an adjacent product, a manufacturer entering a neighbouring country. It offers synergies, the ability to use existing assets and knowledge, and it is the form that most often succeeds.

Unrelated or conglomerate diversification moves into businesses with no operational connection, on the argument that a portfolio of unrelated businesses smooths the group's earnings and that a strong central management can run anything. Its record is poor: the smoothing is worth little to shareholders who can diversify themselves, the central management usually cannot run everything, and conglomerates have tended to trade at a discount to the sum of their parts.

Vertical diversification, into suppliers or customers, is a form of integration with its own logic. The operational case for diversification in a company is about concentration risk.

A business that depends on one customer, one product, one supplier, one market or one technology is exposed to the loss of that one thing, and the exposure may be fatal. Reducing the dependence, by developing other customers and products or entering other markets, is a form of insurance whose cost is the investment and management attention the new activities require and whose benefit is survival when the concentration fails.

Boards should know their concentrations, in revenue, in profit, in supply and in people, and should have a view on which are acceptable. Diversification has costs and limits.

For an investor, over-diversification into hundreds of holdings dilutes the effect of good judgement and adds costs while removing no more risk. For a company, each new activity draws on management, capital and focus, and a business spread across activities it does not understand is weaker, not safer.

Diversification also fails when the correlations it relies on break down, which they tend to do in crises, when everything falls together; and it provides no protection against risks that affect the whole system. The right degree of diversification is the one that removes the specific risks that can be removed cheaply, without giving up the concentration of effort and expertise that produces returns in the first place.