By Corey Hoffstein, Newfound Research Co-Founder & CIO
Rob Arnott, founder of Research Affiliates, and Cliff Asness, founder of AQR, have been jousting with one another through blog posts and scientific papers for the better part of the year.
It largely began with Arnott’s broadly published February article, How Can “Smart Beta” Go Horribly Wrong? The article explored the concept of relative valuations – a comparison of cheapness between the short leg and the long leg – within popular equity factors.
The findings were alarming for two reasons. First, the valuation spread between low and high volatility equities was near all-time highs, indicating that the factor was overvalued, and likely subject to dampened forward return expectations. Or, for the more bearish, an outright crash.
Second, Arnott et al found that that once you adjusted for valuation multiple expansion – i.e. remove those returns that are solely due to changes in valuation – factors like low beta and gross profitability lose their edge. In other words, these might not be factors at all: just artifacts of an era of expanding multiples.
Cliff Asness was not convinced, publishing first an opinion-driven article titled The Siren Song of Factor Timing. The argument was that the long-term explanatory power of value spreads, proposed by Arnott, is either overstated or inapplicable.
Arnott et al published a follow-up to their original piece, titled To Win with “Smart Beta” Ask If the Price Is Right. They find that their valuation-spread methodology is highly predictive of forward 5-year returns and, caveat emptor, all 8 smart beta factors they examined were currently historically overvalued.
Arnott and Asness met face-to-face at the Morningstar Investment Conference in June and discussed their differing viewpoints on the topic (the full transcript is available here). While the debate was amicable, there remains a wedge in their philosophy. Arnott believes that valuations are an important guide for factor investing; Asness remains skeptical.
Following up on his first opinion-based piece, in July Asness published an expansive data-driven analysis, titled My Factor Philippic. The basis of Asness’s argument is that timing with valuations is only applicable if the portfolio stays largely constant. For example, knowing what the P/E of the S&P 500 will be 10 years from now provides valuable information largely because we do not expect the composition of the S&P 500 to be meaningfully different. So comparing the P/E in the future versus the P/E today tells us something about whether stocks became more or less expensive over time.
On the other hand, knowing the P/E of a momentum strategy 10 years from now will provide us little information, as the stocks held by the portfolio could be 100% different than the stocks held today.
While to some this debate may seem pedantic, which side is ultimately correct is critical for ETF investors, who have been some of the earliest and largest adopters of “smart beta” factor products. Some questions that hinge upon this debate:
- Has the proliferation of smart beta products caused an increase in valuations for certain factors?
- Do intra-factor valuation spreads imply a time-able premium?
- Are the historical premiums arising from certain factors really only valuation multiple expansion?
- Do valuations even matter?
That last questions seems particularly arrogant against the backdrop of the 2000s, but Asness would argue that in many cases (i.e. high turnover portfolios), they don’t.
In our opinion, this debate has ramifications not only for investors moving towards single-factor solutions, but also those adopting newer multi-factor solutions. To date, product manufacturers have taken two approaches to building these portfolios: integrated and mixed.
The integrated approach seeks to identify the stocks that score highly across all factors simultaneously. Integration, in a way, performs the timing implicitly: a momentum stock is not worth buying if it is overvalued. Therefore, we would expect that this would be the preferred method of Arnott.
The mixed approach relies on building unique portfolios for each factor and combining these portfolios using a sleeve-based methodology. This is the approach we believe Asness would prefer, as he believes that turnover within a momentum portfolio is too high to make valuation a relevant metric.
While determining who is right or wrong in this debate would be nice, we’re sure the real answer probably lies in between. Ironically, while Asness had the final word on the topic during the Morningstar conference, he spent it somewhat defending Arnott’s approach. “It’s very important to view these things in a portfolio context,” he said. “The way Rob does it is allowed; you’re allowed to have a value-first philosophy.”
Perhaps that is the most illuminating takeaway. You’re allowed to invest however you want: just make sure you know what it implies for your investment philosophy.