The author employs his 15 years of experience in global multi-assets investment industry and 6 years of teaching at the University of Illinois at Chicago to present a comprehensive textbook on quantitative investment theory and practice, providing explanations on financial markets and econometric modeling needed for identifying proper strategies based on macroeconomics and data analysis. The book is structured in nine chapters divided into multiple sections and subsections, each covering a special investment topic, finance theories, fundamental insights on the asset and portfolio performance, statistical models, and R programming to analyze real-word data.
Chapter 1 introduces major indices of the stock market, such as S&P 500 and other Dow Jones indices, and defines return on securities for a portfolio of selected stocks. By the real data for mutual, institutional equity, and fixed income funds it shows that a portfolio can outperform the market benchmarks, discusses challenges to beat the market, and considers conditions for persistent outperformance. Quantitative investment theories are described, starting from B. Graham and D. Dodd on the security analysis, H. Markowitz on the portfolio theory, W. Sharp on the risk modeling, E. Fama on the return forecasting, and many others. The quantitative process includes a strategy with risk and return targets, an alpha model for generating forecasts of future returns on securities, portfolio construction with constraints for risk control, portfolio rebalancing and trading, and performance attribution. The needed information together with basics on the R software are described.
Chapter 2 present the univariate statistical analysis of stocks. It traces development of the S&P 500 index and its daily behavior in the years 1950–2018 with its return (relative change) and volatility or risk (standard deviation), and shows how to calculate the main measures of asset performance, such as annualized return and volatility, and their quotient known as the Sharpe ratio. The data of 1918–1949 years, including the Great Depression and Black Tuesday time (October 29, 1929), business cycles, the formulae for security value and margin of safety, and the market-level price-to-earnings (P/E) ratio are discussed. Stock price as a random variable is considered by its mean, standard deviation, skewedness, kurtosis, density function, and cumulative distribution function (CDF). The uniform, normal, and t-distributions are described, with hypothesis testing by Students t-criterion, confidence intervals, type I and II errors, and checking P/E measure for stability. Asset returns and distributions, detection of outliers and their treatment are shown in examples.
Chapter 3 explores the bivariate statistical analysis for studying relationship between two random variables on examples of the U.S. and Chinese stock markets, S&P 500 and CSI 300, respectively. Top 10 markets of different countries are considered by percentage of global stock market capitalization in 2008 and 2018. Developed, emerging, and frontiers markets are described, focusing on the rapid growth Chinese market and its state-owned public companies and individual investors. Behavior of the Shanghai Stock Exchange (SSE) and its Composite Index (SHCI) are studied in 1991–2019, and China Security Regulatory Commission (CSRC) laws are discussed, including stock halting, caps on price changes, day trading ban, maximum order size, and short selling. SSE annual return and risk are traced by years, and the Buffett factor (quotient of a real economic value to the stock market value) is used. The Pearson pair correlation and Spearman rank correlation are applied to the S&P 500 and CSI 300 data on the daily return, risk, and Sharpe values in 2005–2019. Impact of different events and information decay between the markets are described via the lagged correlations in R examples.
Chapter 4 presents the ordinary least squares (OLS) approach to building multiple linear regressions for a stock selection strategy. The index return of a taken portfolio serves as the dependent variable and various market and economic indicators are employed as predictors (or factors) in the model. The capital asset pricing model (CAPM) proposed by Sharpe, long- versus short-run investments, market efficiency and sources of inefficiency are discussed. The OLS model and its properties are described in detail, from a pairwise to multiple regressions formulated in algebraic and matrix forms, with characteristics of parameters’ significance and overall fit. Examples include the so-called alpha model of the future stock return in its dependence on the factors of profitability, earning quality, management quality, value, price momentum, and market sentiment. These variables are explicitly constructed and used for modeling by different data, including the large companies in the Russell 1000, with R functions and numerical results.
Chapter 5 continues with the weighted least squares (WLS) which permits to incorporate risk into a multiple alpha model. Arbitrage pricing theory (APT) is explained in a multiple regression of the risk by the macroeconomic variables, and the known in econometrics two-stage least squares (2SLS) estimator is formulated. Detection and treatment of heteroscedasticity, and the WLS modeling are considered for the risk-adjusted alpha models. Nonlinearity analysis and nonparametric diagnostics with bootstrapping estimations are illustrated on numerical examples shown in plots built by R scripts.
Chapter 6 deals with time series models for forecasting commodity prices by periodical data containing components of the level, trend, seasonality, and noise. Autoregressive (AR) models and a unit root for spuriously correlated variables are described and examples on it are given. Crude oil prices for the top 10 countries of most oil reserves are described, with oil world production and consumption in 1987–2017. Petrodollar transaction, exporters and importers, pricing regimes, OPEC and oil market are considered. Pricing models by GDP and other macroeconomic factors are described, and cointegration of two time series of the price of oil and price of gold indicates a possible unit root, so dependence on time. Cointegration of two non-stationary series is explained on examples of two drunk men or a drunk man with a dog, and a pair trading strategy is discussed in illustrations.
Chapter 7 investigates portfolio construction with main strategies for the long-only and long-short market neutral, using market capitalization weights for example of stocks in the S&P 500. Modern portfolio theory (MPT) by H. Markowitz and efficient frontier are described, and the mean-variance portfolio optimization is formulated as the least square with a Lagrange restriction solved for the optimal weights. Estimation of systematic and unsystematic parts of the total risk, minimum volatility (minvol) in the smart beta and low beta modeling for portfolio selection in different sectors, Lasso and shrinkage regularization techniques, portfolio backtesting, performance attribution, and global portfolio are considered in many data examples.
Chapter 8 employes the modern quantile regression (QR) for estimating the asset returns’ tail behavior and incorporating the percentile information into portfolios. Empirical distributions of assets such as equity, commodity, and currency are non-normal in different stock markets, but they have long tails. The Kolmogorov-Smirnov test and interquantile ratio (IQR) measure deviations of a distribution from normal, and the tail behavior is characterized by the value at risk (VaR) and conditional value at risk (cVaR). The VaR is defined as the quantile function of the CDF and measures the amount which an investment might lose at a time period, and cVaR is defined as a mean value of integral of VaR during this period, so it is a more realistic measure of losing a range of amount of money. Introduced by R. Koenker, the QR technique is described in detail, the linear programming building of the pure location-shift and scale-shift models is illustrated in multiple examples, with interpretation of quantile effects, description of finite and asymptotic properties, and t-statistics inference for checking the parameters’ significance. Portfolio construction in QR approach is presented in real data examples.
Chapter 9 finalizes the book with the so-called quantamental investment approach which encompasses the fundamental and quantitative analyses and has been developed after the financial crisis of 2008. The total return in this approach consists of four parts related to the financial market, industry, factor, and company, where the last two components comprise the alpha model of quantamental analysis. Alpha sources and model efficacy, risk control, portfolio construction, and their characteristics for company specifics expressed by factors are considered in detail and illustrated by numerical examples on economic growth and development, stagnation after 1990, and dynamics and chaos in financial markets. Quantamental stock selection portfolio and its performance with actual constraints are discussed.
Each chapter gives a brief history of the discipline with pictures of its founders and developers, proposes problems for investigation, examples for calculation, and a list of references. Multiple R codes with explanations and real applications help to learn this software in depth. There is a supplemental material at the Springer site Extra Materials Archive Search (springer.com). The book presents a compendium of classical and modern methods of studying stock markets, and can be useful to advanced students and practitioners in finance and investments. The presented methods and techniques could be applied in other areas of statistical data analysis as well.
Minneapolis, MN
[email protected]