The great factor debate
Daniel Ben-Ami examines a key question that is too often neglected: why does factor investing work?
At a glance
• It is important to question why factor investing works.
• In broad terms, there are rational and behavioural explanations for the behavioural of particular factors.
• There are good arguments for and against both sets of explanations in relation to different factors.
• Ultimately the investor has to decide what path to follow.
Coming from one of the world’s leading experts on asset price movements, it is a shocking statement. Professor Paul Marsh of London Business School is sceptical of the accounts of why factor investing works.
“There are plenty of explanations,” he says. “The problem is they aren’t very good, theoretically.”
“They’re mostly behavioural-type explanations, it doesn’t mean they’re not true, but it does raise the question that, if this is all about behaviour and everyone starts behaving differently, whether or not these things will continue”.
There is no doubting Marsh’s credentials. He is co-author of the Global Investment Returns Yearbook, currently sponsored by Credit Suisse, which is widely recognised as a global authority on the long-term performance of asset prices. Together with fellow academics Elroy Dimson and Mike Staunton, he was also author of Triumph of the Optimists (Princeton 2002), an authoritative study of 101 years of global investment returns.
Of course, not all the experts in the area accept Marsh’s claim. On the contrary, some are avidly opposed. For example, Andrew Ang, the head of factor investing strategies at BlackRock and a former professor at Columbia Business School, argues that the reasons why factor investing works are well understood. “These factor premiums have been around for decades and they’ve been well studied by academics. Some of them have a rational basis and others have a more behavioural basis.”
In any case, this question deserves closer examination. It is not enough to simply assume that factor investing works. The claims made for its success need to be probed. Otherwise, there is a chance that it will break down in the future.
It should be remembered that the observation that certain factors behave in certain ways came well before any comprehensive explanation. For example, the idea of value investing goes back to Benjamin Graham and David Dodd, both professors at Columbia Business School, in the early 1930s – even though it is only in the past few years that value has been dubbed a ‘factor’. The attempts at a comprehensive explanation of factor behavior came much later.
This point was made by Eugene Fama, a professor of finance at the University of Chicago, in his Nobel prize acceptance speech in 2013. Together with Kenneth French, a professor at Dartmouth College, he wrote a paper published in 1993 which proposed that a three-factor model could explain stock returns: company size, value (as measured by a company’s price-to-book ratio) and market risk. Yet Fama acknowledged in his speech that the observation of patterns of behaviour came before any attempts at explanation: “Empirical asset-pricing models work backward. They take, as given, the patterns in average returns, and propose models to capture them. The three-factor model is designed to capture the relation between average return and size (market capitalisation) and the relation between average return and price ratios like the book-to-market ratio, which were the two well-known patterns in average returns at the time of our 1993 paper.”
Those who assume it is enough to simply observe the existence of such patterns would do well to ponder the experience of Google Flu Trends (GFT). The parallel does not exist but it does show the peril on simply relying on the pragmatic ‘it works’ test.
In 2008 Google launched a service that predicted flu outbreaks by examining internet search terms. Essentially if a lot of people were googling terms related to flu that was taken as a good predictor that a flu outbreak was under way or imminent. For a while the system seemed to work much better and faster than conventional health statistics. However, after a few years it became evident that GFT was inaccurate, often substantially overestimating the prevalence of flu. At present, GFT trend estimates are no longer being published.
It may be the case that a new and improved GFT service is launched in the future. But the experience of the service shows the dangers of relying on correlations as a basis for decision-making. It is necessary to probe deeper to find why particular patterns persist.
In the case of factor investing, many hundreds of factors have been identified since the creation of the Factor-French three-factor model in 1993. In a 2015 paper, the two authors added two new factors (related to profitability and to stocks of companies with high total asset growth) which are generally seen as measures of quality. Other commonly used factors include momentum, liquidity and low volatility.
However, the number of factors that exist, at least theoretically, has grown so large that Professor John Cochrane, then president of the American Finance Association, famously referred to a factor ‘zoo’ in his 2010 presidential address.
Vitali Kalesnik, the head of equity research at Research Affiliates, points out that some factors that appear to exist are “spurious”. That is, they appear to appear on paper, or at least in the electronic data, but they do not exist in reality. They are just the result of researchers discovering patterns that exist by chance when they engage in mining vast amounts of data.
In broad terms, there are two, or possibly three reasons why genuine factors can exist. The main distinction is between rational or fundamental factors and behavioural ones. Ang also favours a third class of explanations he refers to as structural impediments. These can arise from leverage constraints, regulations or benchmark rules. For instance, institutional investors with limitations on how much they can borrow could pre fer high-volatility stocks.
On Marsh’s reading of the evidence, there is little rational basis for many of the best-known factors. In relation to value, for example, he says: “The problem is that when you look at the standards of risk it is extremely hard to demonstrate value stocks are riskier. If anything, it’s the other way round. Growth stocks are riskier. And yet growth stocks give inferior returns.” He also rejects the view that higher risk during a recession justifies a greater risk premium for value investment.
Instead, he argues that behavioural factors explain the excess performance of value investing. “People fall too much in love with growth stocks. These are great stories. They’re typically exciting companies. And so the behavioural argument is that people pay too much for growth because they fall too much in love with them.”
For him, the evidence for behavioural explanations is even clearer in relation to momentum. “You can contort yourself stupid but you still can’t get a risk argument to work on momentum.”
The problem, from his perspective, is that behavioural advantages can be arbitraged away when investors become aware of them. Increased investment in value stocks would tend to make them more expensive which would, in turn, decrease their attractions.
However, the problem is that others can draw substantially different conclusions from the same evidence. That is particularly the case when the evidence is so plentiful and complex.
So, for Ang, the rational case for value investment is clear. “Value stocks are riskier than growth stocks. Value tends to underperform during economic troughs. It tends to outperform during expansionary growth periods like the one we’re experiencing today.” His conclusion is that the excess returns to value investment are “a reward for bearing risk. Value firms tend to underperform in bad times.”
The nature of value and growth firms helps to explain this discrepancy. Typically, value firms are tied down by large amounts of physical capital, whereas growth firms are more reliant on human capital. This means that “unless we have a Star Trek holodeck – that allows you to create anything you like out of thin air – value is going to be more inflexible than growth and the value premium is going to continue to exist”, says Ang.
Noël Amenc, professor of finance at the EDHEC-Risk Institute in Nice, is closer to Ang’s view. “I am more favourable towards rational explanations,” he says. “Smart beta [an alternative term for factor investing] works because you have serious sources of return, well-documented by academia. The first one is obviously the exposure to rewarded factors. These are backed by Nobel prize-winning work. Such returns from factors are not an anomaly. They are really a reward for taking long-term risks.”
That does not mean that Amenc rules out behavioural explanations entirely. On the contrary, EDHEC has produced a useful summary of the main explanations, rational and behavioural for some of the best-known factors (see Economic mechanisms behind main factors).
Kalesnik of Research Affiliates is probably closer to Marsh. He accepts, for example, that value companies are not more volatile than growth companies. “I mostly believe the behavioural story. That story looks the most compelling”.
However, he argues that behavourial anomalies are so deeply rooted that they will not be easily arbitraged away. “For every educated investor there are tons of uneducated investors on the street,” he says. “There is a portion of market participants who are not fully rational.”
One area where Marsh is close to making an exception is the size effect. “With size, you might rationally expect a premium if smaller companies were riskier and you might rationally exert a premium of smaller companies were less liquid.”
But he goes on to partly negate his own argument by saying that “individual small companies are risky but portfolios of smaller companies are not”. So, in his view, the risk argument does not work except, perhaps, in that smaller stocks are less liquid than larger ones.
Ang rejects the claim that portfolios of smaller companies are not riskier. He points to Marsh and Dimson’s own dataset to make his case.
If there is anything concluded from such disagreements it is that, ultimately, investors must make their own choices. The most eminent and well-qualified experts can sometimes draw strikingly different conclusions from the same datasets. No doubt there is much to be gained by weighing up their arguments but in the end it is the person putting up the money who needs to make the final choice.