When to send an investing model into retirement

Here’s a maxim for life as a famous economist: If your theory wins a Nobel Prize, look out — it might already be running into big problems. When the Black-Scholes-Merton option-pricing model won the prize in 1997, it had already required some major tweaks to fit new data; a decade later, it was being blamed for the financial crisis. By the time the real business cycle theory won in 2004, economists were already showing why it wasn’t a good explanation of recessions.
In 2013, Eugene Fama won the Nobel for his work on the efficient markets theory. Should he be worried?
Efficient markets theory says that the only way to earn higher returns than the market average is by taking on more risk. But how is risk defined? Without a model of risk, the notion of efficient markets is meaningless — any set of eye-popping long-run returns could be justified by shrugging and saying “Well, I guess that investor just took more risk!”
The main type of risk model used by industry practitioners is called the linear factor model. This model says that excess return — the amount by which an investment portfolio can be expected to beat the market on average — is the weighted sum of a bunch of risk factors. The most famous factor model is the capital asset pricing model, which has only one type of risk — the risk that the market will go down. A portfolio’s vulnerability to market risk is called beta.
The CAPM is so popular, even to this day, that beta has become a catch-all buzzword for risk in the finance industry.
When Fama was first arguing for efficient markets in the 1970s, he, like many others, assumed that the CAPM was a good model of risk. But over the next two decades, it became clear that the CAPM didn’t do a good job explaining the data — a few other things, like the size of a company and its ratio of book value to market capitalization, seemed to be very important.
So in 1992, along with his longtime co-author Kenneth French, Fama published a paper declaring that the CAPM should be replaced with a three-factor model that incorporated these two other elements. Some questioned why company size and value are measures of riskiness, to which the authors — and defenders of efficient markets theory more generally — answered that they must be related somehow to conditions in the economy.
Since then, the Fama-French three-factor model — or sometimes a model adding momentum as a fourth factor — has become the industry standard. Asset managers allocate trillions of dollars according to this model. But now that the floodgates have been opened, every finance professor has his or her own preferred factor model.
Private companies sell models that use dozens of factors. With this ever-expanding universe of factors, the three-factor model persists due primarily to the intellectual heft of Fama and French.
In 2013, however, Fama and French once again revised their position on how many factors should be canon. The new model has five factors — the previous three, plus a company’s profitability and its rate of investment. If the historical pattern holds, the Fama-French imprimatur will give this new model enough gravitas to become standard practice throughout much of the asset-management industry.
But is this a good thing? It seems a little odd that the elders of finance keep overhauling a model that has already won a Nobel. Shouldn’t this be settled science by now?
Some practitioners are already questioning the new Fama-French model. In a recent paper, a team from Robeco Asset Management lists a number of reasons for doubting the five-factor model.
First, the Robeco team notes, the new model doesn’t discard any of the old factors. The idea that beta — correlation with the market — leads to higher returns is a staple of conventional wisdom, but it isn’t really supported by the data. Second, the new model still ignores momentum. This is probably because it’s very hard to explain momentum as a source of risk — chasing good performers sounds more like timing the market than taking on risk.
Third, the Robeco authors note that it stretches the bounds of credibility to think of profitability as a source of risk. On average, the stocks of profitable companies tend to do better. But is that because more profitable companies are riskier, on average, than unprofitable ones? That’s what the new Fama-French model is saying. But it makes little sense. Unprofitable companies are more likely to suffer bankruptcy. They’re forced to depend on outside financing, which makes them more vulnerable to macroeconomic shocks and less able to expand to take advantage of new opportunities. And they’re probably more likely to have dysfunctional management.
A more reasonable circumspect explanation for these phenomena is that markets simply aren’t all that efficient. Instead of bending over backward to explain why profitability is risky, it makes more sense to conclude that in the past at least, investors just didn’t pay enough attention to profit.
Robeco’s arguments are persuasive, and they point to an underlying problem with the whole factor-model approach. When a theory needs to be constantly revised by adding new epicycles, it might be time to abandon the theory. If we have to overhaul our notion of risk every two decades in order to preserve the hypothesis that markets are efficient, it seems easier to simply stop assuming that higher returns have to come from taking on more risk.
— Bloomberg

Noah Smith is a Bloomberg View columnist. He was an assistant professor of finance at Stony Brook University, and he blogs at Noahpinion