Navigating the factor zoo

Posted by Robin Powell on February 8, 2017

 

The next episode of the TEBI Podcast is due to go online tomorrow and it features the guy they call The Swed. As well as being a a well-known investment writer, Larry Swedroe is principal and the director of research for Buckingham Strategic Wealth, and its network of evidence-based advisory firms, the BAM Alliance.

I’m a huge fan of Larry’s, and to whet your appetite for the upcoming podcast, I thought I’d publish (with Larry’s kind permission) the foreword to his latest book, Your Complete to Factor-Based Investing, which he co-authored with Andrew Berkin.

It was Professor John Cochrane who coined the phrase the Factor Zoo, and it’s certainly appropriate. So far more than 600 investment risk factors have been identified by academics and financial practitioners. But there are really only a few that investors need to know about, and I can think of no better guide to help you navigates the zoo and find them than Larry Swedroe.

A great introduction to the book, and indeed this fascinating subject, is the foreword by Cliff Asness, co-founder of AQR Capital Management. It really is a gem. Enjoy. And watch out for tomorrow’s podcast.

Why not subscribe to to the TEBI Podcast and have it automatically download to your phone, tablet, Mac or PC as soon as it goes online?

You can either subscribe on SoundCloud or on iTunes.

 

The following is an extract from Your Complete to Factor-Based Investing by Andrew Berkin and Larry Swedroe and is republished here with the kind permission of the authors.

 

Foreword

by Clifford Asness, Co-founder, AQR Capital Management

At its most basic level, factor-based investing is simply about defining and then systematically following a set of rules that produce diversified portfolios. Imagine I systematically always chose to own a diversified portfolio of stocks with some well-defined, shared characteristic and to avoid, sell, or even short (Larry and Andrew focus on factors that are always as short as they are long) a diversified portfolio with the opposite characteristic.  Once you define a factor, some questions immediately arise: Did it make money in the past? Will it make money, after costs, in the future? Why does it make money? One classic example of a factor is going long a very diversified set of stocks that are “cheap” and short stocks that are “expensive,” where cheap and expensive are measured by comparing a stock’s price to its fundamentals, such as book value, earnings, or sales. Another example is going long stocks with good momentum, meaning they’ve been doing relatively well lately, and short stocks with bad momentum. There are many others, and in fact, that in and of itself can be a bit of a problem.

Professor John Cochrane famously said that financial academics and practitioners have created a factor “zoo.” He didn’t mean it as a compliment. Other researchers have forcibly reminded us recently of the dangers of “data mining.” While in some fields, this may indeed be praised, in finance, it’s generally seen pejoratively. In finance, it means that smart people with even smarter computers can find factors (adding to the “zoo”) that have worked in the past but are not “real.” In other words, they are the product of randomness (together with selection bias) and, thus, their past success won’t repeat in the future. To see this, say there were no real factors, you’d nonetheless find many characteristics that delivered strong historical returns if you tested enough possible candidates (or enough candidate rules for creating diversified factors). But they would have no efficacy going forward, as we started out assuming no factors are “real,” so these were just the result of searching through enough random returns to find factors that accidentally looked good. Such is the power of big data and big computing. The zoo of factors is a result of this fact and, importantly, the powerful incentive, both for academic publication and real-world asset gathering for managers, to find historically powerful factors.

The harm in doing so is real. You might be tempted to think that believing in a factor that isn’t truly real is a neutral event. No harm, no foul. That is not the case. It’s not that you simply fail to make the money the illusory factor randomly made in the past. It’s worse. On top of the factor failing, you still pay trading costs and a management fee to implement it. You still take risk (randomness without a positive expectation of gain is not good!). And because you take risk in this fictitious factor, you likely take less risk elsewhere — perhaps foregoing risk that truly is compensated. So, all in all, data mining over factors is a real problem, and one that’s dangerous to your wealth.

Here’s where Larry and Andrew come in — thankfully. They do, at least, two very important things in this book. First, they offer an extremely useful guide to understanding which of the factors are indeed real. Second, they do it in such a way that a non-super-geek interested reader can really benefit. These are very important, but not easy, tasks — and they do them superbly.

To start, Larry and Andrew show us that the factor zoo isn’t quite as crazy as it seems. They don’t go through every factor on which anyone has written a paper (nobody should do that!). Instead, they recognize that many of the seemingly separate factors are variations on a theme, and these themes make the zoo far easier to navigate. These themes are things like value, size, momentum, carry, and others (not esoteric quantitative jargon but categories that stem from the basic intuitive language of investing). Indeed, that many factors are variations on a theme not unique unto themselves is, in fact, one of the things they insist upon. They correctly look on a theme that works only under one very specific formation and without other reasonable ways to express it, as highly suspect (e.g., if the only value factor that was historically effective was price-to-earnings and not price-to-anything-else, they’d argue — and I’d agree — that our faith in the value effect would be severely shaken).

Next, they give us a wonderfully intuitive and quite specific list of criteria that researchers and practitioners must satisfy to have us believe in, and more importantly invest in, a factor. These are persistence (does the factor historically deliver reasonably reliable returns?), pervasiveness (does it, on average, deliver these returns in a variety of locales and asset classes if such tests apply?), robustness (as just discussed this is the idea that it shouldn’t be dependent on one very specific formulation but fails to work if other related versions are tested), intuitiveness (does it make sense to us or are we just going by historical performance?), and, finally, investability. Notice, aside from investability, all of these come down to some aspect of “do we believe the historical results are real and not just data mining?” Investability is the one that differs, implicitly asking the all-important question: “OK, even if we believe the factor is real, can a practical investor really make money from it after costs?” That’s a very important final question and one Larry and Andrew always make sure to include an answer to.

While those five are the most important, they also often discuss two additional criteria. Both are relevant even if a factor already is deemed persistent, pervasive, robust, intuitive, and investable. First, they ask whether today is different from the past. It’s always possible that a factor was real, not just the result of data mining, but that its day has just passed. Markets can wise-up to behavioral factors, and risk premia can be priced to lower levels than the past. Larry and Andrew aren’t interested in a factor’s past glory, but its future benefits. Second, they make sure to ask, OK, even if this is real, is it already covered by the other factors we believe in?” An example of both is “low-risk investing.” They make a strong case that low-risk investing is more expensive today than in the past and is historically, perhaps, covered already by some of the other factors (e.g., value and profitability). Note, I don’t totally agree, and I’m more bullish on low-risk investing than they are. While perfect agreement is too much to ask for, credible factor researchers and investors will often agree on far more than they disagree, and that’s certainly the case for me with Larry and Andrew. While I might quibble with their conclusions, this is the right way to think about choosing factors and the right questions to ask. The reader is well served to follow their lead and ask these same questions about factors.

It’s not just low risk that doesn’t quite make their cut. Default risk (often called credit risk in other contexts) doesn’t either. Again, I don’t totally agree. Adapting an old joke, you can put three researchers in a room and get four opinions. Still, I love that they are not just summing up the literature uncritically. Rather, they are willing to make some bold statements. Having five important criteria for choosing factors would have few teeth and little credibility if all factors and factor themes passed all the tests. You are reading great stuff, but more importantly, honest stuff here!

Explaining what the factors are, and how to choose whether you believe in them, forms the bulk of this very useful book. But there is more. They devote a whole chapter to the question of whether publication itself ruins future factor returns (does it still work when the cat is out of the bag?). They explain the issues here and share the important results that although you make somewhat less after publication, the bulk of the factor returns on which they report do survive, to my great relief.

All this and I haven’t even discussed the appendices! You get a couple of free mini-books tacked on here, so please make sure to read them. A great example is their review of the findings that are very similar to factor investing in the world of sports betting. They do this not to encourage sports betting or the starting of profitable book-making operations by their readers (actually, I can’t swear to that). Rather, once again, they are trying to understand if these factor themes are indeed pervasive. When something works in a place you haven’t yet looked, you actually get more confident about your original findings. Remember, there’s always a chance those original findings were random luck, not real, and the result of data mining, not truth. In other words, if value investing wasn’t a real phenomenon, but just data mining, just because it works for stocks wouldn’t imply it works for sports betting (as it wasn’t real — it was random!). On other hand, if you have an “intuitive” theory (one of the five criteria!) for why it works for stocks, and that same intuition applies to sports betting, you would expect it to work there, too. Finding it does is a wonderful confidence-builder, even if you never watch sports, let alone bet on them. That is, after seeing similar intuition hold up for sports betting, you should have somewhat more confidence in that intuition leading to real, not random, returns for stocks.

Vitally, Larry and Andrew always stress that no investing plan, diversifying across good factors included, survives too much impatience. Although they strive to show us factors that work more often than they don’t, and work better and more consistently in combination, Larry and Andrew don’t go down the cheap, easy, and false path of promising an easy road. Rather, they stress that belief in the factors you choose is vital not just to make sure they are real, and not just data mining, but because without well-formed, strongly held beliefs (perhaps stemming from persistence, pervasiveness, robustness, intuition, and investability!), nobody will stick with their investing process through the inevitable tough times. Good and wise advice. It’s not hard to give back more than all the positives of a good investment process by falling down on this last vital hurdle.

Finally, I’d contrast their work with the legion of investing books telling you how to actively pick stocks just like Warren Buffett (actually, Larry and Andrew do this, too, but I’ll let them tell you about it). Those books are, in my view, far easier to write than this one. There are always great stories when it comes to individual companies: fascinating tales of greatness and woe that end wonderfully for the sage stock picker who is the hero of this tale. In contrast, Larry and Andrew have taken on the task of describing an inherently quantitative affair. Factors don’t have made-for-TV endings. Success is measured by less thrilling statements like, “and then the factor had a +1 standard deviation decade,” versus discretionary stock-picking stories that end with “and then the company I invested in invented the iPod!” Of course, while the stories are better in the world of discretionary stock picking, it has one small problem (just a small one, mind you): It generally doesn’t work! This is not the place to dive into such a contentious issue (so I’ll just assert it), but both economic logic and a large body of research has shown that discretionary stock picking is, in broad generality (with obvious exceptions like Mr. Buffett), not the path to riches. So they could’ve gone with the better and easier stories, and written an easier (for them and the reader) and maybe more entertaining book. Instead, they went with stuff that they, and I, believe actually helps. I think they made the right choice.

To sum up, this is a great book. It surveys the increasingly important area of factor investing, suggesting concrete ways to navigate the factor “zoo.” I am confident that you will enjoy your journey through the factor zoo as much as I did.

 

Cliff Asness

 

Robin Powell

Robin is a journalist and campaigner for positive change in global investing. He runs Regis Media, a niche provider of content marketing for financial advice firms with an evidence-based investment philosophy. He also works as a consultant to other disruptive firms in the investing sector.

Read more...

How can tebi help you?