Some risks pay you and some just happen to you

Most explanations of compensated and uncompensated risk stop at the point where it starts to matter. This one carries on.

The distinction

Financial theory separates risk that cannot be diversified away from risk that can, and only the first has any reason to be compensated. Market risk is shared by every holder of equities and cannot be removed by holding a larger number of them.

The risk that one particular company fails can be removed almost entirely by holding a great many companies. Because it can be removed at almost no cost, there is no reason for anybody to be paid for carrying it. This is the clearest argument available for diversification, and it does not depend on predicting anything at all.

What that means in practice

Holding a single company exposes you to both kinds of risk while offering an expected return based on only one of them. Concentrated portfolios produce a much wider range of outcomes without a correspondingly wider expected return. Some concentrated holders do extremely well, which is what a wider distribution means rather than evidence of any skill.

The useful part is this: the median outcome for a concentrated portfolio has tended to lag the market even in periods where the mean did not. That gap between median and mean is precisely the shape you would expect from a small number of very large winners.

Where the argument gets contested

Whether characteristics such as company size or valuation represent compensated risk is genuinely disputed among researchers. Some of the evidence has weakened under scrutiny, and several effects have been much smaller in the years after publication.

The debate is unresolved, and anybody presenting it as settled in either direction is overstating what is currently known. Products built on contested characteristics carry higher charges and stretches of underperformance long enough to exhaust conviction. Holding a broad market fund requires taking no position in that debate at all, which is a large part of its appeal.

The uncompensated risks people accept anyway

Home country concentration is uncompensated, since no theory suggests your own market pays more for being yours. Employer share concentration is uncompensated and correlated with your income, which makes it worse than a random single holding.

In practice, sector bets and thematic funds add risk that is diversifiable and therefore has no reason to be rewarded. Currency exposure inside bond holdings is frequently uncompensated and can comfortably dominate the yield being sought.

Each of these can be reduced at low cost, which is more or less the definition of a risk not worth carrying.

Risk that is not in any model

Fraud, platform failure, sanctions and expropriation are risks that no expected return has ever compensated anybody for. These are managed through structure, diversification across institutions and refusing anything that cannot be explained clearly. The most reliable protection is declining to hold what you do not understand, which costs nothing and forgoes very little.

Anything promising returns without a mechanism you could describe to somebody else belongs in this category by default. This is where scepticism earns more than analysis, because analysis quietly assumes the reported numbers are real.

Adjust the size of it until it is something you would actually do tired.

Applying it to a portfolio

Go through each holding and ask whether its distinctive risk could be removed by holding something broader instead. Where the answer is yes and the cost of doing so is low, that risk is not being paid for. Where the answer is no, you are holding market risk, which is the risk the entire exercise exists to take.

This produces a shorter and duller portfolio than most people start with, which is generally the direction of improvement. It also leaves every holding with a reason that does not depend on any forecast being correct.

The takeaway

If a risk could be removed cheaply, nobody is paying you to hold it. That is the whole test.

The version you keep doing is the version that works.