Over the past four decades, academic researchers have documented hundreds of different stock market anomalies and their related investment strategies. But while most of these strategies look compelling on paper, the results often end up being much less convincing in practice, and are sometimes even downright disappointing.
One of the key reasons for this mismatch is that, in their research, academics generally analyze the returns generated by a given strategy without taking into account transactions costs, management fees and other real-life investment constraints. They also often fail to support their empirical findings with proper out-of-sample confirmation tests. This is also frequently the case for the usual backtests and simulations put forward by some product providers in their brochures and presentations.
In this context, for asset owners considering factor allocation, but who are still unfamiliar with the associated practical consequences, keeping implementation costs down is often perceived to be a major challenge. A FTSE Russell survey carried out in 2016 showed that controlling implementation costs actually ranked eighth among investor concerns when it comes to factor oriented allocations.
This issue is actually increasingly in the spotlight. Over the past few years, a growing number of academic papers1 delving further into this question have been published. Meanwhile, various prominent asset managers and index providers, including Robeco, have voiced their concerns on the matter.
Indeed, transaction costs and other practical hurdles can be a major drag on performance. This is especially true since factor-based strategies tend to generate higher turnover than passive market-weighted ones, due to the fact that the portfolio must be rebalanced regularly in order to maintain exposures to the different factor premiums.
As we saw in a previous article of this series that covered the major challenges identified by investors considering factor investing, it is possible to keep turnover within reasonable bounds, while ensuring the appropriate factor exposure. But this usually requires a more sophisticated approach than those used for the popular products – mostly ETFs – that is based on generic ‘smart beta’ indices. For example, we found that the low volatility anomaly can be harvested with a turnover of less than 30% per annum2.
However, potentially excessive turnover is only one of the pitfalls. Other aspects, such as liquidity issues, potential ‘mistrades’, inefficient portfolio construction processes and investment constraints can also have a serious impact on performance. The cost of switching from a traditional asset allocation framework, based on asset classes, geographic areas and business sectors, should not be overlooked, either.
When assessing a factor-based strategy, it is essential to take all these implementation costs into account. Unfortunately, this is easier said than done and estimating transaction costs remains a challenge. In a recent note3, EDHEC-Risk Institute researchers argued that the costs incurred, when replicating a number of generic smart beta indices, can vary dramatically depending simply on the size of the investment universe.
Interestingly, however, they also noted that transaction costs could be significantly reduced by applying common sense implementation rules, such as controlling the total index weight invested in one stock relative to the stock size. Other examples included imposing limits on turnover or on the frequency of rebalancing.
At Robeco we acknowledge that implementation costs are of paramount importance. This explains why we focus on only four of the hundreds of factor premiums reported in the academic literature: value, momentum, low volatility and quality. These meet the required rigorous academic criteria and can be put to work efficiently in real-life conditions. Moreover, our portfolio construction models have been designed to minimize implementation costs, both in equity and fixed income markets.
Our research, and also our real-life experience with managing factor strategies, has shown that ensuring strategies are properly designed by focusing on the well-established factors is clearly worth the effort. Targeting these factors can add value, even after taxes, trading costs and restrictions.
This has also been found to be true for mutual funds. In 2014, Eduard van Gelderen and Robeco’s head of Factor Investing research, Joop Huij, studied4 the returns of US mutual funds adopting common factor-based equity strategies over the 1990-2010 period. They found that managers targeting low beta, small cap, and value either outperformed or, in the case of low beta, achieved significantly lower risk. For momentum and reversal strategies the evidence was more mixed, but few funds seemed to be specifically targeting these factors. In all cases, the dispersion in performance was large though, which underscores the need for well-designed strategies.
1See for example: ‘Trading Costs of Asset Pricing Anomalies’, Andrea Frazzini, Ronen Israel and Tobias Moskowitz, 2014. Or: ‘A Taxonomy of Anomalies and their Trading Costs’, Robert Novy-Marx and Mihail Velikov, NBER Working Paper No. 20721, 2014. Or: ‘Will Your Factor Deliver? An Examination of Factor Robustness and Implementation Costs’, Noah Beck, Jason Hsu, Vitali Kalesnik, and Helge Kostka, Financial Analysts Journal, 2016.
2‘Low Volatility Needs Little Trading’, Pim van Vliet, SSRN working paper no. 2612790.
3‘Smart Beta Replication Costs’, Mikkel Esakia, Felix Goltz, Sivagaminathan Sivasubramanian and Jakub Ulahel, EDHEC-Risk Institute, 2017.
4‘Academic Knowledge Dissemination in the Mutual Fund Industry: Can Mutual Funds Successfully Adopt Factor Investing Strategies?’, Eduard Van Gelderen and Joop Huij, The Journal of Portfolio Management, 2014.
Esta serie de artículos se propone aclarar algunas de las cuestiones clave a que se enfrentan los inversores cuando adoptan estrategias de factor investing.