Systematic, rules-based investment strategies are where academia and practice are currently interacting strongly. My objective in this editorial is to offer some thoughts on research on systematic investing, including three articles in this issue, that can provide significant practical benefits for academics, practitioners, and investors alike.

Author’s note:

The views expressed in this editorial are my own and do not necessarily reflect the views of Bank of America Merrill Lynch.

One of the most important developments in institutional investing in recent years is the shift to systematic, rules-based investment strategies—of any kind—from purely passive (index investing) to semi-passive (smart beta) to quant (factor-based) and other investment strategies. The Investment Company Institute (2017) recently stated in its annual fact book that from 2007 to 2016, “index domestic equity mutual funds and index-based ETFs [exchange-traded funds] received $1.4 trillion in net new cash and reinvested dividends, while actively managed domestic equity mutual funds experienced a net outflow of $1.1 trillion” (p. 46). In a survey conducted for BlackRock, the Economist Intelligence Unit (2016) reported that for 87% of the 200 institutional investors that participated in the survey (about a third of which had more than $25 billion in assets under management), factors1 play a role in the investment process.

The fact that increasingly more assets are being traded in coordination (i.e., as an index portfolio or as a factor portfolio) has significant implications for asset prices, risk management, and market microstructure. Both academics and practitioners have focused heavily on related research questions in numerous scholarly articles and textbooks2 and at academic/practitioner conferences. This issue of the Financial Analysts Journal includes three articles on factor-related research. Andrew Ang, Ananth Madhavan, and Aleksander Sobczyk (2017b) develop a new model that identifies active funds’ dynamic exposure to standard factors. They document considerable diversity in factor concentrations, with large-cap blend funds showing the lowest concentration and large-cap growth funds the highest (especially in momentum and quality). Timotheos Angelidis and Nikolaos Tessaromatis (2017) create a global factor allocation strategy using country indexes. An interesting practical implication of their strategy is that it is less susceptible to capacity constraints than similar strategies that use single stocks. Malcolm Baker, Ryan Taliaferro, and Terence Burnham (2017) study low-frequency systematic investment strategies and conclude that value, size, quality, and low beta receive almost equal allocations in an optimal portfolio. Evidently, systematic investing is where academia and practice are currently interacting strongly and will almost certainly continue to do so. My objective in this editorial is to offer some thoughts on research that can provide significant practical benefits for academics, practitioners, and investors alike.3

Recent empirical and theoretical work (see, e.g., Ben-David, Franzoni, and Moussawi 2016; Chabakauri and Rytchkov 2016; Baruch and Zhang 2017) has focused on the impact of index investing and ETF trading on the asset prices of index constituents, the underlying investments of these vehicles.5 The evidence in the empirical papers is somewhat mixed: ETFs facilitate efficient pricing of information, but security prices have been shown to become noisier. In addition, published theoretical predictions have yet to be tested extensively with data. In Giamouridis, Neumann, and Steliaros (2017), my co-authors and I explored patterns in trading volumes and equity flows, liquidity measures, and risk in a group of ETF constituent stocks and similar nonconstituents. We documented higher trade commonality in ETF constituent stocks (in down markets), increased commonality in their liquidity/market impact, and less idiosyncratic risk compared with nonconstituent stocks. Taken together, these empirical and theoretical insights have several implications for risk management (e.g., effect on risk diversification), trading and execution (e.g., cross-asset dependence of impact cost), and stock selection (e.g., alpha potential). Future research should further clarify how prices clear, who the marginal investor is, how volatilities and correlations change, the structure of equity risk (i.e., systematic versus nonsystematic), and the likelihood of price deviations from fundamentals and reversions. Such research should cover not only stocks in broad market indexes that are ETF constituents but also specific segments of the equity market (e.g., sector, value, and momentum stocks).

With respect to factor investing, there is an increased concern that perhaps too many investors are chasing too few factors. This concern is typically associated with the events of the week of 6 August 2007, when a number of quantitative equity long–short funds experienced significant losses. Khandani and Lo (2011) concluded that these temporary losses were due to the combined effects of coordinated factor-based portfolio deleveraging and a temporary withdrawal of market-making capital. Since then, crowding has frequently been the subject of discussions among investment professionals but very often in a rather qualitative way. More research on what makes a sensible ex ante measure of crowding is needed, and my view is that the microstructure and trading of factor portfolios should be the focus of future research. I argue that bottom-up measures of factor constituents’ liquidity and market impact, as well as investors’ propensity to trade factors over time and cross-sectionally, are important indicators for timely warnings. Measures based on holdings and positioning of the market should also provide good approximations. Moreover, we need to clarify how such measures are associated with the performance of factor portfolios and under what market conditions. Crowding strategies alone—that is, without any exogenous event that triggers fast liquidation—may not have a negative impact on their performance.

One focus of many quantitative asset managers is how to achieve the best execution of their portfolio strategies. Depending on the investment strategy’s turnover, the duration of the associated investment signals, and the sizing of optimized portfolios, execution costs can become significant—even of the same order of magnitude as the strategy’s potential return—materially affecting the strategy’s capacity. Optimized portfolio executions can significantly reduce execution costs, increasing investment returns and thus the corresponding capacity of the strategy. Systematic investment strategies naturally lead to coordinated portfolio trade lists—natural order flow. The market impact from executing these portfolio trade lists increases the correlations of intraday stock returns and affects the covariance of stock returns. The coordinated order flow also suggests an increase in the covariance of intraday traded volume, which could itself be a source of the variability of execution costs. Benzaquen, Mastromatteo, Eisler, and Bouchaud (2017) suggested that stocks’ order flows exhibit correlations, the disregard of which leads to an underestimation of trading costs and thus to suboptimal execution programs. Finally, the increased concentration of assets managed under such investment strategies is likely to affect the availability and type of discretionary and latent liquidity, which, in turn, should change both the nature and the cross-asset dependence of impact costs.

Factor returns are rewards for risk taking.6 Although factor exposure is compensated in the long run, factor portfolios may exhibit losses in interim periods. These days, investors are increasingly interested in whether timing good times is a plausible exercise. I usually like to answer the complementary question: Is timing bad times plausible? A number of asset managers argue that timing is quite challenging and that simply diversifying across weakly correlated factors is a sensible strategy for avoiding bad times. But academic evidence (for the most comprehensive study, see Moreira and Muir 2017) suggests that an easily implemented strategy that takes less risk when volatility is high can be a very effective timing strategy. In Miller, Li, Zhou, and Giamouridis (2015), my co-authors and I likewise concluded that a risk-oriented factor-timing model that uses such machine-learning-type algorithms as classification and regression trees (CART) is compelling. I expect factor-timing models to attract significant interest from practitioners. Market positioning and trading of factor portfolios, exposure of factors to macroeconomic risks, and factor portfolio risk concentration and fragility would be interesting research angles to pursue.

Capturing a theoretical risk premium (or anomaly return) is more easily said than done. Building a portfolio of such payoffs may be even more complicated. Institutional investors can currently access risk premiums and factor returns externally—through an asset manager, listed ETFs, or investment bank products. These products differ in many ways, including their investment objectives, which can vary widely—for example, to seek maximum factor exposure or controlled factor exposure; how they utilize the cross section of the investable universe (e.g., implemented as long only [buying positive-factor-exposure firms] with a market hedge or long–short [buying positive-factor-exposure firms and selling negative-factor-exposure firms]); and how they approach multifactor investing (e.g., bottom up [mix of factor exposures at the stock level] versus top down [mix of factor portfolios]). Although I have seen some research addressing multifactor investing (see, e.g., Clarke, de Silva, and Thorley 2016), there seems to be a demand for empirical research regarding the other objectives as well as related implementation issues. Obvious additional questions are how a product’s performance compares with what it is meant to achieve (transparency is essential in this regard) and how what it is meant to achieve compares with the actual premium (proper or feasible benchmarking). For example, are all value products equally good at capturing the value premium? And how is the value premium defined for an applied setting in the first place?

Many investors are concerned about the capacity of quant strategies, particularly factor-based strategies, and whether premiums and factor returns are likely to persist. Capacity is largely a function of turnover and execution costs. Two excellent studies in this area are Frazzini, Israel, and Moskowitz (2012) and Novy-Marx and Velikov (2016). Although they are not focused on capacity, Beck, Hsu, Kalesnik, and Kostka (2016) study the impact of execution costs on factor returns. Turnover is dictated by the signal decay (i.e., the duration of the signal predictability), and perhaps there are limits to what portfolio managers can do to control it. With respect to execution costs, earlier I highlighted the merits of taking a portfolio approach to trading—that is, accounting for interactions with peer stocks, both in terms of risk and market impact modeling and in terms of trade scheduling. The timing of implementation has also been found to be relevant (Bogousslavsky 2017). The other side of the equation is the actual factor return. Is it likely to persist? To answer this question, one needs to reconcile who is going to pay the return—that is, who is on the other side of the factor-based strategy (Ilmanen 2016)?7 We should be seeing more research in this direction as more detailed data (e.g., investor type, timely equity flow data) become available. Ang, Madhavan, and Sobczyk (2017a) presented a different, top-down approach to capacity estimation. Their estimate is based on the dollar amount that would be required to neutralize the current exposure of active managers to a particular factor. The capacity of systematic investment strategies ultimately depends on their impact on order flow, available liquidity, the resulting market microstructure, and, finally, their joint effect on equilibrium execution costs.

Over the past five years, machine-learning models and big data have attracted significant interest.8 More recently, investors pursuing systematic strategies have started to consider how to adapt to what some might deem a new investment paradigm. In 2017, I had the privilege of attending Professor David Hand’s presentation on this subject. My prior belief was that quants have always used big data and have always used statistical techniques,9 which form the basis of what is now called machine learning. Professor Hand pointed out that interest in machine learning has increased because of (1) growth in computer memory, (2) faster computers, and (3) automatic data capture. I know many firms that are keen to obtain and test “any dataset” for stock selection. Trading-algorithm developers advocate using machine-learning methods to identify the most appropriate trading algorithm for an order.10 At the conference where I heard Professor Hand, a presenter originating from one of the largest risk system providers did not rule out the possibility that machine learning could be used to identify fund-specific benchmark portfolios in the context of risk and performance attribution. In his conclusion, Professor Hand quoted Eric Schmidt, the former CEO of Google, who told the 2017 SkyBridge Alternatives (SALT) hedge fund conference that he was “looking forward to the startups that are formed to do machine learning against trading, to see if . . . pattern recognition . . . can do better than the traditional linear regression algorithms of the quants.” Professor Hand argued that there are reasons why pattern recognition may not do better than the traditional linear regression algorithms—namely, nonstationarity, small signal/noise, and overfitting. Up to this point, we have considered many interesting ideas, some of which are backed by economic intuition, but we have yet to see a significant amount of empirical evidence. Future research should focus on actionable ideas regarding machine learning and big data in the entire spectrum of the investment process—that is, alpha, beta, risk management, and execution and trading.

I have identified some key areas for research that could have a high practical impact. My assessment is based not only on interactions with asset managers and investors who pursue systematic investment strategies but also on conversations with discretionary, active (i.e., fundamental/macro) investors who recognize that the growth of assets in systematic investments affects their performance. Professionals with quant training11 have never been in higher demand—they appeal to all types of investors! So, I would recommend that any young or prospective investment professional thinking about career development seriously consider quant training. The CFA® designation provides an outstanding background for this career path. But perhaps that is only the starting point. You should also be able to read and comprehend academic papers. The Financial Analysts Journal is an excellent publication that offers papers full of relevant and useful information for practitioners. There are a number of other places where you can find practical research, such as the CFA Digest and similar publications from investment banks that highlight topical academic research. Finally, and most importantly, you should be able to think how you can tackle your day-to-day tasks and longer-term projects diligently, rigorously, and scientifically.

Notes

1 I use the term factor in a generic sense. A factor is a characteristic of a firm or asset class that is believed to drive its returns and that is typically used to create a rules-based portfolio. For example, a firm’s P/B can be used to construct a portfolio of low-P/B firms.

2 See, for example, Andrew Ang, Asset Management: A Systematic Approach to Factor Investing (New York: Oxford University Press, 2014); Lasse Pedersen, Efficiently Inefficient: How Smart Money Invests and Market Prices Are Determined (Princeton, NJ: Princeton University Press, 2015); and Antti Ilmanen, Expected Returns: An Investor’s Guide to Harvesting Market Rewards (Chichester, UK: John Wiley & Sons, 2011).

3 The focus of this editorial is equities, because systematic strategies were first developed in that asset class. But the majority of my assertions also apply to other asset classes to the extent that systematic strategies are widely implemented. For example, the fixed-income space has seen tremendous growth in ETF assets in the last couple of years, factor-based strategies have become popular in all asset classes, and numerous investment products offer access to cross-asset risk premiums.

4 I define coordinated investing as buying stocks in “baskets.” For example, when an investment in an ETF is made, the ETF sponsor buys the entire basket of constituent stocks at the same time.

5 For a detailed discussion of ETFs as products and how they are used in investment strategies, see Hill, Nadig, and Hougan (2015).

6 This particular interpretation of factor returns explains why returns to some factors have been very persistent. An alternative interpretation is that some factor returns are either behavioral or friction-induced market anomalies and so timing them should be associated with the ebbs and flows of the underlying behavioral tendency or market friction. A third interpretation is that some factor returns can be the result of extensive data mining. When I refer to factor returns here, I am referring only to the former categories. A separate stream of research is focused on establishing criteria for factor return significance and determining a sensible set of factors (see, e.g., Harvey, Liu, and Zhu 2016).

7 Here, I am referring only to factor returns in the two categories I mentioned earlier (i.e., rewards for risk taking or for a behavioral tendency) and not those that are the result of data mining.

8 On Google Trends, for example, interest in “machine learning” is stable over 2004–2012 and then increases monotonically to its current peak level.

9 Standard techniques, including CART, LASSO (least absolute shrinkage and selection operator), principal component analysis (PCA), and countless variants, are found in widely used machine-learning libraries (e.g., Python’s scikit and MATLAB®’s Statistics and Machine Learning Toolbox).

10 A trading algorithm determines and implements a schedule of trade execution—that is, the volume and timing, during a day or multiple days, of a stock’s purchase or sale.

11 “Quant training” can mean many things, but skills in basic empirical asset pricing, risk management and portfolio construction, market microstructure, quantitative methods, and programming are essential.

Ang, A., A. Madhavan, and A. Sobczyk (2017). "Crowding, Capacity, and Valuation of Minimum Volatility Strategies." Journal of Index Investing, a Vol. 7, No. 4 (Spring): 41-50. Google Scholar
Ang, A., A. Madhavan, and A. Sobczyk (2017b). "Estimating Time-Varying Factor Exposures." Financial Analysts Journal, Vol. 73, No. 4 (Fourth Quarter): 41-54. Google Scholar
Angelidis, T., and N. Tessaromatis (2017). "Global Equity Country Allocation: An Application of Factor Investing." Financial Analysts Journal, Vol. 73, No. 4 (Fourth Quarter): 55-73. Google Scholar
Baker, M., R. Taliaferro, and T. Burnham (2017). "Optimal Tilts: Combining Persistent Characteristic Portfolios." Financial Analysts Journal, Vol. 73, No. 4 (Fourth Quarter): 75-89. Google Scholar
Baruch, S., and X. Zhang (2017). "Is Index Trading Benign?" Working paper (June): https://ssrn.com/abstract=2990283. Google Scholar
Beck, N., J. Hsu, V. Kalesnik, and H. Kostka (2016). "Will Your Factor Deliver? An Examination of Factor Robustness and Implementation Costs." Financial Analysts Journal, Vol. 72, No. 5 (September/October): 58-82. Google Scholar
Ben-David, I., F. Franzoni, and R. Moussawi (2016). "Exchange Traded Funds (ETFs)" Charles A. Dice Center Working Paper 2016-22, Fisher College of Business Working Paper 2016-03-022, Swiss Finance Institute Research Paper 16-64 (November): https://ssrn.com/abstract=2865734. Google Scholar
Benzaquen, M., I. Mastromatteo, Z. Eisler, and J.-P. Bouchaud (2017). "Dissecting Cross-Impact on Stock Markets: An Empirical Analysis." Journal of Statistical Mechanics: Theory and Experiment, Vol. 2017, No. 2 (February). Google Scholar
Bogousslavsky, Vincent (2017). "The Cross-Section of Intraday and Overnight Returns." Working paper (June): https://ssrn.com/abstract=2869624. Google Scholar
Chabakauri, G., and O. Rytchkov (2016). "Asset Pricing with Index Investing." Fox School of Business Research Paper 15-051 (July): https://ssrn.com/abstract=2531753. Google Scholar
Clarke, R.G., H. de Silva, and S. Thorley (2016). "Fundamentals of Efficient Factor Investing." Financial Analysts Journal, Vol. 72, No. 6 (November/December): 9-26. Google Scholar
Economist Intelligence Unit (2016). "The Rise of Factor Investing." (April). Google Scholar
Frazzini, A., R. Israel, and T. Moskowitz (2012). "Trading Costs of Asset Pricing Anomalies." Fama-Miller Working Paper/Chicago Booth Research Paper 14-05 (December): https://ssrn.com/abstract=2294498. Google Scholar
Giamouridis, D., M. Neumann, and M. Steliaros (2017). "The Impact of ETFs in Financial Markets." Institutional Investor Journals, Article Collection. Google Scholar
Hand, David (2017). "Data Science and Machine Learning." Bank of America Merrill Lynch Global Quant Conference, London (15 June). Google Scholar
Harvey, C.R., Y. Liu, and H. Zhu (2016). ". . . and the Cross-Section of Expected Returns." Review of Financial Studies, Vol. 29, No. 1 (January): 5-68. Google Scholar
Hill, J.M., D. Nadig, and M. Hougan (2015). A Comprehensive Guide to Exchange-Traded Funds (ETFs) Charlottesville, VA: CFA Institute Research Foundation. Google Scholar
Ilmanen, Antti (2016). "Who Is on the Other Side?" Q Group Fall Conference, Scottsdale, AZ (October). Google Scholar
Investment Company Institute (2017). Investment Company Fact Book: A Review of Trends and Activities in the Investment Company Industry. 57th ed. Washington, DC: Investment Company Institute. Google Scholar
Khandani, A.E., and A.W. Lo (2011). "What Happened to the Quants in August 2007? Evidence from Factors and Transactions Data." Journal of Financial Markets, Vol. 14, No. 1 (February): 1-46. Google Scholar
Miller, K.L., H. Li, T.G. Zhou, and D. Giamouridis (2015). "A Risk-Oriented Model for Factor Timing Decisions." Journal of Portfolio Management, Vol. 41, No. 3 (Spring): 46-58. Google Scholar
Moreira, A., and T. Muir (2017). "Volatility-Managed Portfolios." Journal of Finance, Vol. 72, No. 4 (August): 1611-1644. Google Scholar
Novy-Marx, R., and M. Velikov (2016). "A Taxonomy of Anomalies and Their Trading Costs." Review of Financial Studies, Vol. 29, No. 1 (January): 104-147. Google Scholar

Author Information

Daniel Giamouridis is global head of scientific implementation at Bank of America Merrill Lynch, London and co-editor of the Financial Analysts Journal.

Related Topics

Users who read this article also read


Rate and Share