Should factors be a factor?

Factor research is about seeking a deeper understanding of markets rather than racking up silver bullets

Readers who have followed me over the years will know that I like to keep up with as much academic literature as time permits, but general references aside, I have never specifically written about factor investing.

Factors are numerical attributes of shares believed to lead to better performance prospects. Examples would include low p:es (value), high six-month to annual returns (momentum), low debt to equity (quality) and small firm size (size).

In the early days, the finance industry clearly thought the academics were not worthy of their rarefied airspace, and ignored efficient market theory, perhaps with some justification. Then along came some factor research, which was ignored for a long time until it seemed to show real persistence and generate interesting results.

Of course, financial markets have a habit of swinging from one extreme to another, and there was huge adoption of all sorts of stylebased strategies.

The finance industry, greedy bastard that it is, didn’t seem to appreciate that factor research was always about seeking a deeper understanding of markets rather than racking up silver bullets. After all, anything that is going to create higher performance is an odds-on candidate to become a victim of its own success, to be annihilated as soon as it’s discovered.

This is beautifully summarised in a research paper led by Rob Arnott of Research Affiliates, titled Alice’s Adventures in Factorland: Three Blunders That Plague Factor Investing. The years since this paper was written in 2019 have continued to support its validity, as far as I’m concerned.

First, factor expectations seem to have proved to be unrealistic because of mining the data (searching to validate your opinions), crowding of the trade and unrealistic costs. Second, investors used naive risk management models and failed to appreciate just how vicious tail risk downside shocks can be. Third, they believed their portfolios were diversified because of the factor exposures, but this diversity vanished under market stress conditions when factors started autocorrelating.

Originally, the expectation was that there would be a verifiable reason that a factor might “work”. If we take the case of the value factor, which is evidenced by low p:es, low price-to-book ratios and high dividend yields, there were two basic theories as to why it would work.

The first was that the stock in question had been subject to some kind of shock or surprise that put pressure on the price, and therefore this required a return premium. The second was a mispricing explanation which suggested that the value share was mispriced for behavioural reasons, and that a patient visionary leaning against the market tide would realise a return as the share normalised, whatever that might entail.

Now, it is certainly true that in the days of yore — from the 1940s to the 1990s, say — shares that had high repeated and apparently persistent rates of earnings growth would become highly prized. The reason was simple: all you had to do to keep your highly paid position was buy the steadily growing chestnuts and if it bombed out, well, you were in good company.

The issue, of course, was that the earnings growth rate tended to get overestimated, and if the company strategy bombed out it got very ugly. The result was that all you had to do to beat the market was go against the prevailing dogma and outlast the tough patches. This was fine, right up until researchers started highlighting the value premium.

Once the search for value became widespread and widely exploited, the bettercalibre value stocks were seized upon and rerated upwards, leaving the newlook range of lower-p:e stocks a ragbag of value traps.

The real proof of the pudding is that this has happened in just about all the factors that have been “discovered”. Several papers in 2016 demonstrated that point, and it should hardly be a surprise because markets are always evolving.

I’ve noted market participants grumpily saying “style is broken”, but in reality, style was never a promise of any description. As a matter of fact, to define a style you first have to demonstrate that it had a statistically relevant return, historically independent of the market. This meant that a factor was actually a risk descriptor. There was no free lunch on offer.

Many hedge funds discovered this to their great cost in 2008. Fortunately for them, most have since had the good sense to move beyond style.

Financial markets have a habit of swinging from one extreme to another