4. Risk or behavior? The two explanations

Level 9: Factors

This module brings order to the whole level. So far you have seen four characteristics that in the past separated winners from losers. Now comes the question that decides what to do with them: why did they pay? There are two answers, they have been at odds for forty years and they are not interchangeable, because they imply different futures.

Two stories that draw the same chart

The first answer is that the factor compensates a real risk. Cheap companies are cheap because they are fragile: they have more debt, thinner margins and they suffer more when the economy turns. Whoever buys them takes on a genuine danger and gets paid for doing so. Fama and French defended this reading, and it fits with all the preceding theory: the extra return is not a gift, it is the price of sleeping worse.

The second answer is that the factor exploits a systematic error. People extrapolate: they take a few good years and project them indefinitely, thereby overpaying for the brilliant and underpaying for the boring. Josef Lakonishok, Andrei Shleifer and Robert Vishny argued this with an observation that is hard to dodge: cheap shares not only returned more, they also did not appear to suffer especially in the worst stretches, which is exactly what the risk explanation requires.

The crucial detail is this: both stories produce exactly the same chart of past returns. No track record, however long, distinguishes between them on its own. You have to look at other things.

If it is risk, it has to hurt, and hurt when it suits you least

A risk premium has a signature it cannot fake: it pays badly precisely when you need the money most. If it never hurt, nobody would demand compensation for bearing it and the premium would have evaporated decades ago through sheer competition.

Out of that comes a practical criterion that rules out many products: if you are sold a structural risk premium and its track record has no genuinely painful stretch, either the risk is hidden where you cannot see it, or the data is not what it looks like. The pain is not a defect in the strategy. It is the product.

And there is an encouraging consequence. If the payment comes from a real risk, it will still be there as long as people prefer not to take that risk on. Everyone knowing about it does not destroy it: knowing about a risk does not eliminate it. That is why this kind of factor can be held for twenty years, provided you survive its bad stretches without selling.

If it is behavior, somebody has to be getting it wrong and be able to keep getting it wrong

The other branch requires two conditions at once, and almost nobody checks the second.

• That the error exists. Somebody has to be systematically overpaying. If you cannot say who is on the other side and why they accept losing, you do not have an explanation, you have a wish.
• That the error can persist. A known, easily exploited error disappears. It only survives if something prevents money from eliminating it: costs, execution risk, constraints, the short horizons of those who manage money. It is exactly the material of level 2, the limits of arbitrage.

If this is the explanation, your factor is not an asset you buy and forget: it is a position that requires you to monitor whether the error is still there and whether the obstacle protecting it is still standing. When that obstacle falls, the factor switches off, and level 4 deals with how that switching off happens.

The third possibility: that it never existed

Before choosing between risk and behavior it is worth considering the option that there is nothing to explain. Hundreds of factors have been published, and when you search that many characteristics over the same data, a few will look extraordinary through pure chance and those are precisely the ones that will get published. Campbell Harvey and his co-authors argue for this reason that the bar of evidence demanded of any new factor should be raised considerably.

How to tell them apart without fooling yourself

No single test is decisive on its own. These five questions, answered in writing and before you invest, do almost all the work:

• When does it lose? If its worst stretches coincide with recessions and credit crises, that points to risk. If it loses at macroeconomically irrelevant moments, that points to behavior.
• Can I write the risk story in advance? A story invented after seeing the data is worth nothing. Write it beforehand and check whether it holds.
• Does it show up outside its sample? Other countries, other asset classes, other eras.
• Who is on the other side and why? Without an identifiable loser there is no sustainable gain.
• What protects the effect from being eliminated? If there is no obstacle at all, do not count on it over twenty years.

Antti Ilmanen sums up the state of the debate in a way worth adopting: in most cases the honest answer is that there is some of both, and the sensible thing is not to win the argument but to size the position in line with that uncertainty. Note too a case where there is no dispute: low volatility cannot be a risk premium, because it pays you for taking on less risk. Its survival depends entirely on the constraints and biases that produce it still being there.

Real case: value's crossing, read with this criterion

Go back to value's thirteen bad years. Under the risk explanation, you would expect the punishment to be concentrated in a crisis, with fragile companies collapsing because their fragility materializes. That is not what happened in the worst stretch. Between 2018 and 2020, in a calm economy with no credit crisis, the valuation gap between cheap and expensive shares widened to its most extreme recorded level, above even the year 2000, according to AQR's analyses. A good part of the punishment came from the cheap getting cheaper still, not from their businesses collapsing relative to those of the expensive ones.

That detail does not settle the argument, and it would be dishonest to claim otherwise: a market can legitimately reassess the risk of a group of companies and express it in the price. But it does change the question, which stops being whether value has died and becomes how much of the punishment was price and how much was business. When in 2022 the gap partly closed and cheap shares beat expensive ones by on the order of twenty points in a year, it became clear that what had moved before was above all the price.