How finance professors keep moving the goal posts for active management.
This article originally appeared in the October/November 2012 issue of MorningstarAdvisor magazine. To subscribe, please call 1-800-384-4000.
In August, hedge-fund AQR researchers Andrea Frazzini, David Kabiller, and Lasse H. Pedersen published a study arguing that they’ve identified Warren Buffett’s magic formula: borrow lots of money at low or negative interest rates to buy low-volatility, low price/ book, high-quality stocks. They even constructed a mechanical strategy that does a passable job of mimicking the guru’s returns, at least in back tests. It’s a fascinating and well-done piece, conducted by credible researchers with impeccable pedigrees. It’s the bleeding edge of research conducted at the intersection of academic and practitioner finance. But at the surface, the claim stretches one’s credulity.
It’s actually not much of a stretch. Decades of research has firmly established that most of active managers’ returns, even seeming outperformance, can be replicated by a handful of factors, which can be thought of as simple mechanical strategies. Finance professors describe discovering one of these factors as the process of turning alpha into beta. In the argot of finance, beta is performance attributed to factor exposures; alpha is what’s left over, unexplained, often interpreted as evidence of skill. The professors hold the reasonable notion that once they offer convincing evidence that they can replicate a manager’s outperformance using simple, mechanical rules, and they disseminate the rules widely, the manager no longer deserves to command high fees. The operative word here is convincing—anyone can explain away a manager’s outperformance if they look hard enough.
Factor investing may be news to you, but you’ve almost certainly used it. The growthvalue style delineation, popularized by the Morningstar Style Box, is based on the value factor. Splitting the stock market into growth and value wasn’t just to accommodate investors who needed a simple way of sorting active managers. It was to avoid growth stocks and to buy value stocks. Value stocks, usually defined as stocks with high book-to-price, have outperformed growth stocks in nearly every market studied, over suitably long periods. U.S. large-cap value stocks have beaten large-cap growth stocks in six out of the eight decades since 1930. Over the nearly full sample, 1930 to 2010, they won by about 2.9% annualized. Value index funds, then, can be seen as dirt-cheap computerized strategies to replicate the once-secret sauce value managers possessed.
(Note: The performance data above actually starts in 1927, but for simplicity’s sake I chose to stick with round decades. Large-cap value stocks’ outperformance shrinks to 2.3% annualized if you look at the period 1927–2011. The data are from the Kenneth French Data Library.)
Evidence of persistent outperformance beyond factor exposures is rare. The implication is that many fund managers are unknowing members of a cargo cult: The visits to company managements, the poring through financial statements, the chart gazing are useful to the extent that they touch upon these factors. Why not cut out the middle man and own the factors directly? It’s a question some big institutions have begun asking themselves. The Government Pension Fund of Norway, the biggest pension fund in Europe, and CalPERS, the biggest public pension fund in the United States, have embraced the risk-factor based view of the world. Advisors should begin asking themselves whether they should, too.
An Abbreviated Statistical Detour
What exactly is a factor? In its raw form, it is simply a series of returns. Usually, it’s generated by sorting securities by some metric, and owning a set of the highest-scoring and short-selling a set of the lowest-scoring, ideally creating a market-neutral strategy that reflects a spread return. It may seem strange that one can divine whether a manager is really just “loading up” on a factor just with this information and the manager’s returns.
The intuition is simple. Let’s say I have a machine, whose workings are a mystery. Every day, it outputs either “true” or “false” in a seemingly random fashion.