Managing Editor’s Letter

In the lead article, “It’s All About Data: How to Make Good Decisions in a World Awash with Information,” Mehrzad Mahdavi and Hossein Kazemi explain how, in the financial sector, the rise of big and alternative data has resulted in significant new opportunities in all areas of the financial industry, including risk management, portfolio construction, investment banking, and insurance. To build trusted outcomes in artificial intelligence/machine learning initiatives, the roles of financial professionals are critical. Because there are many nuances in using big data, the authors explain why there is a need for vetted protocols and methods in the selection of data sets and algorithms. There are best practices and guidelines in effectively reducing the risks of using artificial intelligence and machine learning, including overfitting, lack of interpretability, biased inputs, and unethical use of data. Mahdavi and Kazemi emphasize that a key success factor for the deployment of artificial intelligence initiatives is talent and bridging the skills gap. Given the major shortage of talent in this space, practical training of employees and continued education are key to ensuring sustainable growth in the fast-moving field of trusted data-driven decision making.

Machine learning can help asset managers in most portfolio construction tasks, such as idea generation, alpha factor design, asset allocation, weight optimization, position sizing, and the testing of strategies. Derek Snow’s article, “Machine Learning in Asset Management—Part 2: Portfolio Construction—Weight Optimization,” is the second in a series of articles dealing with machine learning in asset management focusing on weight optimization strategies. Specifically, Snow explains the different weight optimization methods that can be used for supervised, unsupervised, and reinforcement learning. Within the supervised learning framework, three techniques are explained. The first is a more traditional linear approach using OLS, Ridge, and LASSO regressions. The second is a nonlinear deep learning approach using autoencoders. The third is a Bayesian sentiment method. Three methods for unsupervised methods are principal component analysis, hierarchical clustering analysis, and network graphs. Finally, Snow explains portfolio weight optimization for the deep reinforcement learning approach.

Zihao Zhang, Stefan Zohren, and Stephen Roberts adopt deep reinforcement learning algorithms to design trading strategies for continuous futures contracts in their article, “Deep Reinforcement Learning for Trading.” They begin by discussing the connection between modern portfolio theory and the reinforcement learning reward hypothesis, showing that they are equivalent if a linear utility function is used. Both discrete and continuous action spaces are considered, and volatility scaling is incorporated to create reward functions that scale trade positions based on market volatility. They backtest their proposed algorithms on 50 very liquid futures contracts for the period 2011 to 2019 and investigate how performance varies across different asset classes (commodities, equity indexes, fixed income, and foreign exchange markets). When their algorithms are compared to classical time-series momentum strategies, Zhang, Zohren, and Roberts find that their method outperforms baseline models, delivering positive profits despite heavy transaction costs. The authors conclude that their proposed algorithms can follow large market trends without changing positions and can also scale down, or hold, through consolidation periods.

Quantitative asset management firms typically rely on a multitude of investment signals to formulate models to drive alpha. Ideally, each signal will provide information that is both predictively useful and complementary to the other signals feeding into an investment process. For this reason, a critical component of quantitative investing is model validation. A balancing of informative power and simplicity is required in the construction of any model employed by quantitative asset management firms. In “Modular Machine Learning for Model Validation: An Application to the Fundamental Law of Active Management,” Joseph Simonian introduces an approach to model validation that he calls “modular machine learning” and applies the model to build a methodology that can be used to evaluate investment signals within the conceptual scheme provided by Grinold and Kahn’s fundamental law of active management. The author’s framework is modular in two respects: (1) it is composed of independent computational components, each using the output of another as its input and (2) it is characterized by the distinct role played by traditional econometric and data science methodologies. An econometric model in the first module is used to classify data in an economically intuitive way. To demonstrate how modular machine learning works, Simonian provides an example using the well-known Fama–French factors as test signals.

In “Interactions and Interconnectedness Shape Financial Market Research,” Otto Loistl and Gueorgui S. Konstantinov investigate two issues relevant to financial data science. The first issue covers the entire production chain “from orders to prices” by modeling the microstructure of stock exchanges. The authors explain how data-driven research can model decisions to place orders and to generate prices by matching orders accordingly. The second issue relates to the price interconnectedness of markets by networks. Network analysis for factors and assets and centrality-based solutions in portfolio management are advantageous for capturing nonlinearity in data. The authors show that interactions shape a market’s performance and provide two examples that market participants interact, learn, and trade.

Financial modeling and analysis have traditionally relied primarily on financial econometrics, which applies statistical methods to the problems of finance. This has changed in recent years as advances in analytical and algorithmic methods, powerful technological solutions, and the expansion of the financial data ecosystem now offer novel and robust ways to solve existing and new problems in finance. Tamer Khraisha, in “A Holistic Approach to Financial Data Science: Data, Technology, and Analytics,” discusses in detail the financial data ecosystem, illustrating the types and structures of financial data and their distinguishing features. The author proposes a comprehensive, financial-domain–oriented approach to financial data science management that can be used by researchers and practitioners to exploit the opportunities arising from the evolving finance landscape.

Principal component analysis (PCA) is a useful tool when trying to construct factor models from historical asset returns. Using PCA, Marco Avellaneda, Brian Healy, Andrew Papanicolaou, and George Papanicolaou demonstrate that a relatively small number of factors can account for most of the variation in the collective movements of the implied volatilities derived from US equity options. In their article, “PCA for Implied Volatility Surfaces,” they show that this market factor is the index resulting from the daily compounding of a weighted average of implied-volatility returns, with weights based on the options’ open interest and Vega. Based on their analysis of the singular vectors derived from the tensor structure of the implied volatilities of S&P 500 constituents, the authors also find evidence that some type of open interest- and Vega-weighted index should be one of at least two significant factors in this market.

Asset management and custody banks are the two primary subindustries within the capital market industry. In their article “Evaluating Cost Efficiencies in Asset Management and Custody Banks Using Data Envelopment Analysis,” D. K. Malhotra, Rashmi Malhotra, and Robert Nydick use a data envelopment analysis (DEA) model to evaluate the economies of scale in the asset management and custody bank subindustry for the period of 2010 to 2017. Measuring company size using a firm’s total assets under management and total revenue generated, the authors evaluate cost efficiencies with cost measured in terms of the two important components of total expenses (total distribution expenses and general administrative and selling expenses). Using a returns-to-scale DEA model, they find that whenever asset management and custody banks reach their most productive scale size and begin to experience decreasing returns to scale, they make structural changes by adopting new technologies and/or changes in the processes, so that their total operating expenses decline and they can grow again without a proportionate increase in operating costs.

Francesco A. Fabozzi

Managing Editor

• © 2020 Pageant Media Ltd