The 2-century investment mystery

Factors work, but the explanation why remains elusive

John Rekenthaler 1 April, 2019 | 2:00PM

A Common Practice
First and foremost, most investment researchers seek positive (or negative) conclusions. Confirming the null hypothesis does no good at all. The aim is either to be published or to arrive at a new investment scheme. Neither goal is achieved by learning that the factors are unrelated. What matters are results!

No surprise, then, that investment analysts are adept at discovering "anomalies"--investment oddities that can be discussed in a paper, used as an exchange-traded-fund strategy, or both. (If such anomalies did not exist, one might say, it would be necessary to invent them.) Analysts scour for data that support their views, but they tend to be much less diligent with disconfirming evidence.

This enterprising mindset has its merits. It can identify temporary conditions. Its speculations lead to hypotheses that might otherwise go unexplored. And often, its findings are robust, to use the statistical lingo. For measuring powerful effects, such as the relationship between fund expenses and future returns, the back of the envelope works just as well as econometrics' calculations.

The Big Picture
There are times, however, when one desires something more complete: a study that might be faulted for using outmoded data, or for bludgeoning what could be caught with a butterfly net, but that will never be accused of being perfunctory. There is no need to test how its results perform out of sample, because there is no out of sample. The study encompasses all.

Such, with only modest exaggeration, is "Global Factor Premiums," or GFP, a Dutch collaboration between a university and asset manager. (Guido Baltussen and Laurens Swinkels of Erasmus University of Rotterdam and Robeco Asset Management; Pim van Vliet of Robeco.) As the authors note, most investment papers "start typically around 1980," often focusing "on a single asset class, typically U.S equities." Their ambitions are, to understate the matter, greater.

Specifically, they examine six potential investment factors, across 68 markets, for time periods that can exceed 200 years. The markets consist of equities (for example, the United Kingdom, beginning in 1799); bonds (Norway 10-year, 1822); commodities (cocoa, 1979); and currencies (Australian dollar, as expressed in U.S. dollar terms, 1834).

(The study, it should be emphasized, evaluates markets and not individual securities. I initially made the mistake of thinking that the authors' assessment of "value" strategies for equities would emulate the well-known research by Eugene Fama and Ken French. Not so. Fama & French compare one stock against another. CVS Health (CVS) goes into the value bucket, while Amazon.com (AMZN) belongs in growth. GFP's authors compare countries. Italy is value, while the United States is growth.)

The paper offers the broadest of perspectives. It obscures most of the effects of time periods, countries, and asset classes. This is by no means unambiguously good. If a factor was strongly significant from 1850 to 1950 in European countries, for stocks, bonds, and currencies, but not commodities, and insignificant under every other situation, then it will likely appear as weakly significant overall. A steak that is charred on the outside and blue on the inside is not "medium."

(The authors, of course, recognize this issue and run various sensitivity tests. There are only so many nuances that they can uncover, though, in a study so expansive.)

6 Factors
But what a view! Here are the authors' six factors:

1. Trend: The trend-following strategy is simple indeed. Invest in the asset if the market's 12-month trailing return is positive. Short the asset if it is not.

2. Momentum: Momentum is the relative version of trend. Invest most heavily in the asset that has performed the best over the past 12 months. Place progressively less into assets as their relative performance declines. (The authors short assets that have negative trend scores, because trend is absolute, but for the other factors, which are relative, they use the scores to scale the positions.) 

3. Value: Value can't be measured in the same fashion across the asset classes. For stocks and bonds, the authors define value as being those markets with the highest yields (expressed in real terms for bonds). Value commodities are those that have suffered the worst five-year returns, and value currencies are those that have the highest purchasing power. 

4. Carry: Carry is the cost (or benefit) of holding an asset. A positive carry trade delivers profits if market prices remain unchanged. In contrast, a negative carry trade loses money, if all things stay the same. (Presumably, those holding negative carry trades do not expect such an outcome.) Invest most heavily into assets that have the highest carry.

5. Seasonality: The authors define seasonality as the asset's average performance during that month, over the past 20 years. Invest most heavily into that asset's best month, and least into its worst. 

6. Betting Against Beta: BAB, as the authors call it, is the newest of the six factors. It involves taking long positions in the less-volatile assets, short positions in the more-volatile assets, and then adjusting the portfolio's overall risk level that it matches a standard. Effectively, it rewards assets for having milder volatilities than one might expect.

Building a Mystery
The upshot? Everything works! Not for all six factors, for all four asset classes, but the direction is uniform. Only seasonality has nonpositive results, for bonds and currency. (The authors state in the text that they did not test those two combinations, and yet their appendix carries the numbers…well, it's a working paper.) Of the 22 remaining combinations, 10 are statistically significant at the 1% level, and three others at the 5% level.

I cannot explain why these factors have been successful across the centuries, countries, and asset classes. Owning securities that have high carry scores, relatively low betas, or strong trend returns would not seem to be riskier than doing the reverse. (Never mind seasonality, which smacks of voodoo.) So why, over all those years and markets, should investors earn a premium for doing so?

The authors have no answer either. "We find no supporting evidence for (unified) explanations of the global factor returns based on market, downside, or macroeconomic risk…the global return factors bear basically no statistically or economically significant relation to common global macroeconomic factors."

In other words, the existing theories of investment behavior--which, to be sure, are plenty useful, being (for example) the underpinning for the development of market-index funds--are inadequate. That is an odd conclusion to derive for a subject that has been studied so extensively, for so long, but GFP does not make its case lightly.


Securities Mentioned in Article

Security NamePriceChange (%)Morningstar Rating
Amazon.com Inc1,792.21 USD0.91
CVS Health Corp60.28 USD2.45

About Author

John Rekenthaler

John Rekenthaler  John Rekenthaler is Vice President of Research for Morningstar.