CFM held its annual Research Retreat in the French countryside in May. During our retreats we take the opportunity to invite eminent members from both academia and industry as guest speakers. This year, we had the pleasure of welcoming Prof. Dr. Thorsten Hens from the University of Zurich who presented his work on Evolutionary Portfolio Theory and how it applies to factor investing. We sat down with him to discuss his take on some of the most topical debates in the world of factor (or smart beta) investing, and in what direction he sees research, and the industry going.
|CFM:||One may safely claim that factor investing has become a mainstay of the asset management industry. Why do you think it has taken the better part of two decades after the now canonical Fama-French (1992) paper for factor investing to come into vogue?|
|TH:||When Fama and French published their seminal paper the finance industry was still selling alpha generated from discretionary asset management. Massive evidence showed that there is no value in those traditional active funds and eventually investors switched to index products e.g. in the form of ETFs. Finally, after the indexing-hype investors were ready to buy systematic strategies generating alternative betas.|
|CFM:||Capitalising on the current popularity of factor investing, many, if not most passive managers have launched their own version of factor tilted ETFs – in a countless number of guises. Do you harbour any concerns about excessive ‘risk premia harvesting’ or crowding? What does this mean for the capacity of factor-tilted investments?|
|TH:||Yes, I share the concerns and recommend that every factor investor should reflect on the impact a lot of factor investing has on his strategy. That question is at the centre of our evolutionary model. We analyse the capacity and the cross impact of factors on factors. It turns out that those effects can be negative (crowding out) but sometimes they are also positive (crowing in). In general the interaction of factors is dynamic, generating a cyclical performance of any single factor.|
|CFM:||The impressive growth of ETFs has also stoked many contradictory opinions on whether these investment vehicles should be considered as ‘passive’ or ‘active’. Where do you fall on this debate?|
|TH:||The meaning of the acronym ETF is ‘exchange traded fund’. One can trade active and passive funds. So in principle ETFs are both – even though originally ETFs started as passive index funds.|
|CFM:||In your research, you distinguish between the naïve and reflective factor investor. Could you briefly explain this distinction?|
|TH:||The naïve factor investor beliefs he acts in a huge market whose fluctuations are determined exogenously, e.g. by business cycles, monetary policy or skill of managers running companies. The reflective investor understands that the market is composed of a handful of factor strategies that dynamically interact generating endogenous fluctuations. When he targets an excess return over the market – knowing that the market return has to be the average return of all investors – he will understand ‘who is on the other side’, i.e. who is paying for his returns by falling short to the market.|
|CFM:||Academic literature is littered with what some have called a ‘Factor Zoo’. Your research has revealed that there might only be as many as 10 unique factors. Would you be surprised if any amongst the hundreds of proposed factors show real statistical promise?|
|TH:||Yes, 10 factors are sufficient – fortunately. To show this one computes the variance-covariance matrix of all stocks in the market (e.g. a matrix with 500 rows and columns in the case of the S&P 500) and then one does a statistical test called principal component analysis, PCA. The result is that at most 10 genuine factors are needed to model the dynamics of the market. But of course there are a million different ways to represent these genuine factors by concrete factors which can be invested in based on simple indicators like book-to-market, size, momentum etc.|
|CFM:||Data-mining has become an ever cheaper computational exercise. Looking towards the future, what role do you think machine learning, artificial intelligence, or big data may play in combination with exploiting risk premia?|
|TH:||Those powerful techniques are not needed to identify the 10 factors – a simple PCA can do that. But once you have identified those factors one needs to study their dynamic interaction. You can do this without a structural model by data-mining. But I would suggest to use the insights from evolutionary finance to get some structure into the search for the dynamic interaction. Our evolutionary portfolio theory structures the dynamics in three steps: First identifying a few genuine factors, secondly assessing the market ecology composed of those factors, i.e. estimating the assets under management of those factors and finally modelling the evolution of the market ecology by e.g. flow of funds functions that are well known from the hedge fund and the mutual fund industry. Even if the machine learning approach ultimately discovers the same dynamics our approach would still be superior because one gets a better understanding of the driving forces behind the dynamics.|
|CFM:||We couldn’t let you go without asking your opinion on factor timing. Does your research convey some credence to such an endeavour, or is it a fool’s errand?|
|TH:||Sure – factor timing is the ultimate goal of our research. So far all attempts have failed because the dynamic interaction of factors was not understood. This understanding is impossible with the standard finance models like the CAPM or APT because they are routed in equilibrium theory. But if anything a financial market is a dynamic stochastic game and methods of evolution are better suited to understanding its dynamics. To give an example. A famous paper in the Journal of Finance claims that the two factors of value and momentum are complementary but that besides holding a 50:50 exposure to both you cannot do anything. In contrast we can show that the dynamics between value and momentum is like the predator-prey dynamics between foxes and rabbits that is well known in evolutionary biology. Thus one can time value and momentum by estimating the relative importance of these factors for asset returns and combining it with the (cross) impact of these two strategies. Or to say it more bluntly, if you see too many foxes be careful when you are a rabbit.|
Any description or information involving investment process or allocations is provided for illustrations purposes only. Any statements regarding correlations or modes or other similar statements constitute only subjective views, are based upon expectations or beliefs, should not be relied on, are subject to change due to a variety of factors, including fluctuating market conditions, and involve inherent risks and uncertainties, both general and specific, many of which cannot be predicted or quantified and are beyond capital fund management’s control. Future evidence and actual results could differ materially from those set forth, contemplated by or underlying these statements. There can be no assurance that these statements are or will prove to be accurate or complete in any way. All figures are unaudited.