Factor investing
Factor investing has been very popular for quite a while now. And for good reasons: it makes perfect sense, for an investor, to exploit the underlying common drivers of portfolios’ returns in a systematic way, which also happens to be cheaper. Which is why over the years, with the help of cheap data and ever increasing computing power, academics and practitioners have been hunting high and low for such common drivers of stock returns, thus ‘discovering’ hundreds of equity factors. What’s not to like?
Alas, like anything in life, factor investing is not immune from challenges and potential pitfalls. In the remainder of this short paper, we’ll highlight the three main challenges, and the way we tackle them at Harvest.
The first challenge is that the financial validity of most of the hundreds of equity factors ‘discovered’, is highly questionable. The data mining process might have well discovered statistically significant correlations between stock prices and alleged factors, but what is highly questionable is the economic and financial significance of such correlations. As we know, by ‘torturing data’ for long enough we can find statistically significant correlations between almost any variables, yet investing on such information could prove painful: as an example, a strong correlation has been found, in the past, between the S&P 500 index and production of butter in Bangladesh, yet we would not recommend investing in the former taking cues from the latter.
There is, then, another class of factors whose financial validity is questionable: factors that are indeed backed up by valid economic rationale, yet they can’t be easily exploited because they get quickly arbitraged away after they become known to the investment community.
To prevent falling into the trap of spurious or unexploitable correlations, therefore, we stick to the very few factors that satisfy two requirements: 1) their relationship with stock prices is grounded in solid economic/financial theory (that is, it stands to reason that the factor might well explain the behaviour of the stocks), and 2) the relationship is pervasive, being both persistent over time (that is, it doesn’t only work for a selected period of time) and functioning across geographies (that is, it doesn’t only work for a selected country or area).
The second challenge – after having identified economically sound and financially exploitable factors - is that it’s easy to become enamoured of a factor’s performance and its robust economic rationale, thus overlooking that such performance is highly variable, and a factor can go through spells of poor performance lasting even decades! The ‘low beta’ factor, for example, has recently underperformed the broader market for more than 20 years. If a capital owner held a 20-years’ underperforming ‘low beta’ ETF, how would he know whether the factor is forever ‘broken’, hence he should get out of the position, or whether the underperformance is just transitory, hence he should hang in? Besides, even assuming he is convinced the factor is not ‘broken’ and it will eventually turn around, he might not be in the position to sit on a losing position for more than 20 years. Therefore, factor strategies, like ‘smart beta’ or other mechanical combinations, can be very hard to hold on to for any capital owner.
To obviate the drawback just mentioned, at Harvest we are not wed to any specific factor/strategy, no matter how fashionable it might be at the time. Instead, we start by drawing from the full range of factors/strategies, combining them to maximise our potential sources of return and to diversify the risk of a single factor’s underperformance. Then, at least as importantly, we strive to adapt swiftly to changing market conditions.
The third challenge is, arguably, the most relevant: even the best ‘tried-and-tested’ factors cannot extract all the returns markets offer. As an example, over the last 20 years or so, the broad market factors – such as valuation, growth, quality and momentum – only explained 65% of global equity manager’s relative returns[1]. Hence, focusing only on investment factors leaves to chance the remaining 35% of excess returns, which historically have been attributable to stock selection. Which means, investors may be better off accessing investment factors through stock selection rather than indexation.
Several attributes characterize, and differentiate, Harvest’s approach to active investing. For the purpose of this paper, however, we’ll focus on one specifically; that is, we believe the real power is not in pure quantitative factor investing, but in a combination of two types of analysis: 1) qualitative fundamental (thematic and stock-specific); and 2) quantitative
Traditionally, fundamental and quantitative investing were treated as two separate schools of thought and pretty much mutually exclusive. Fundamental investors used long-established corporate analysis techniques, relying on research and pure human judgement to spot the best investment opportunities. Quantitative investors, on the other hand, constrained the human judgement, by employing innovative computer algorithms to sift through large amounts of data for latent investment opportunities and to make systematic investment decisions.
Naturally, each approach has its own strengths and weaknesses.
The fundamental approach allows for the full expression of human intelligence, creativity and experience. The drawback is in the limitations of the human brain, which: 1) can only process a limited amount of information, relatively slowly and sometimes erroneously; and 2) delivers a variable output, due to the impact of fatigue and emotions.
The quantitative approach offers a solution to human limitations: a computing machine can flash-process a colossal amount of data and its output is not affected by fatigue or emotions. Arguably, however, a machine might prove ineffective when faced with highly unusual situations, the ones requiring ‘lateral thinking’. Please also consider that: 1) markets are, ultimately, made by human beings, hence it can be argued that, in some highly unusual situations, it takes a human to understand a human situation; and 2) computers are built and coded by humans, hence they inevitably suffer from some of the human limitations.
And here is the key attribute of Harvest’s approach to capital management: we believe that the qualitative fundamental and the quantitative approaches should be actively combined, since they ideally complement one another. By blending them, we can exploit both engines of return, qualitative fundamental and quantitative, and diversify our risk whenever the prevailing market environment favors one engine over the other. Perhaps, the biggest benefit of our approach is the discipline the quantitative side of the process enforces upon us, as it shields us from being emotionally attached to particular investment ideas.
We use some of the factors as forward-looking risk indicators and - together with stringent risk controls like position sizing, concentration limits and stop-losses - they represent a key building block of our risk management protocol. Factors may also help us understand why the market has low expectations for a stock’s future growth and profitability. Are the prevailing low expectations for a stock explained by reasons specific to it or are they due to some common factor such as, for example, slow economic growth? If the latter, we will consider investing in the specific stock once we have reasons to believe that investors are becoming more positive to cyclical firms. If the former, we won’t consider buying the stock, not even when investors’ positive change towards economic growth will have made the cyclical attribute of the stock a desirable characteristic.
In conclusion, we strive to harness the full power of ‘Man plus Machine’, in what could be termed a ‘quantamental approach’. Our ultimate goal is robustness in return generation over time.
[1] Morningstar, Rolling 18-month regression for Global Equity Managers, December 2015. Returns are net of fees based on mutual fund NAVs.