Cliff Asness "My Factor Philippic"

Cliff Asness – My Factor Philippic – My latest on factor timing (why you shouldn’t do much of it) and on more general factor investing disputations

Abstract:

Arnott, Beck, Kalesnik, and West (2016) (ABKW) study smart beta or factor-based strategies and come to the following conclusions: (1) Aside from value, most popular factor strategies currently look expensive. (2) These expensive factor valuations portend lower future returns and a strong possibility of a future “factor crash” in which they go “horribly wrong.” And (3) many of these non-value factors were never real to start with because their historical performance was due to factor richening. That is, researchers mistook the one-time returns from factor richening for truly repeatable “structural alpha.” ABKW’s implied bottom line (their many protestations to only making modest recommendations aside): stick with value, dump the other factors. This essay elaborates on my response in Asness (2016). In summary: (1) I find non-value factor valuations moderately expensive, but not as expensive as ABKW. (2) I argue that ABKW exaggerate the power of factor timing by improperly using long-horizon regression techniques. More proper short-horizon regressions suggest some weak factor timing ability and given this predictability, I construct value-based tactical factor timing strategies to test them. Unfortunately, these strategies add little to portfolios that are already invested in the value factor. It turns out that this “newly” discovered timing tool is, yet again, mostly just a version of regular old value investing. And (3) I examine ABKW’s claim that factor richening drives much of non-value long-term factor performance and find that this very serious allegation about other researchers’ work is totally without merit. Overall, these results suggest that one should be wary of aggressive factor timing. Instead, investors are better off identifying factors they believe in, and staying diversified across them, unless we see far more extreme pricing than we do today.

I. Introduction2
There is no question that smart beta or factor-based strategies have become increasingly popular in recent years.3 In light of this, Arnott, Beck, Kalesnik, and West (2016) (henceforth ABKW) examine these strategies and come to provocative conclusions which, if true, have important implications for smart beta investors.4,5,6,7 To facilitate this discussion let’s break their essay into three parts:

(1) They start by examining recent valuations of a handful of popular smart beta and/or factor strategies. Valuations are a natural starting point in understanding whether these strategies have become too popular or crowded. Presumably, if this is the case, we would see the footprint of investor flows in the form of expensive factor valuations or narrow factor “value spreads.”8 They find that, with the exception of the book-to-price factor, that these factors are currently expensive.

(2) Given that, they turn to the question of whether expensive valuations matter to future performance. To answer this, they examine regressions of future long-term factor returns on past factor valuation levels and find a significant relation. Their regressions suggest that when factor valuations are expensive, future factor returns tend to be lower ? bad news for current factor investors. Furthermore, they argue by analogy that we’ve seen something like this before. In the late 1990’s the world fell in love with the stock market and growth/technology stocks. At that time we saw extremely expensive stock market valuations and we all know what happened from there right? Is a factor crash coming? ABKW clearly think yes.

(3) Finally, and most provocatively, they argue that some of these so-called factors never really even had true positive expected returns (or “structural alpha” as ABKW call it) to start with. Why? Because the realized return to any investment can be influenced by valuation changes. To the extent valuation changes are not expected to repeat (or even worse, could reverse), any return that is due to these valuation changes shouldn’t count as part of one’s assessment of a factor’s true expected return going forward. They argue that once you remove the effect of these valuation changes from historical factor returns, for many of the non-value factors, there is little left over. In other words, both investors and academics have failed to realize this important insight and as a result have unknowingly flocked into zero expected return strategies that have simply richened over time.

The implications of this work are seemingly clear. If you believe (1) and (2) above, then the expected return to non-value factor strategies are lower going forward and could even crash. Clearly, investors in these strategies should reduce their allocations. If you believe (3), investors should never have invested in the first place and certainly should run with their ill-gotten gains now! In this case, many non-value factor strategies have no true positive expected return and most likely no role in one’s portfolio. In the case of (3) it’s really not a matter of timing or valuation or “lightening up” as these factors just are not real to begin with.

Full PDF below

SSRN-id2799441