In recent years, portfolio optimization and construction methodologies have become an increasingly critical ingredient of asset and fund management while at the same time portfolio risk assessment has become an essential ingredient in risk management, and this trend will only accelerate in the coming years. Unfortunately there is a large gap between the limited treatment of portfolio construction methods that are presented in most university courses with relatively little hands-on experience and limited computing tools, and the rich and varied aspects of portfolio construction that are used in practice in the finance industry. Current practice demands the use of modern methods of portfolio construction that go well beyond the classical Markowitz mean-variance optimality theory, and require the use of powerful scalable numerical optimization methods. This book fills the gap between current university instruction and current industry practice by providing a comprehensive treatment of modern portfolio optimization and construction methods illustrated by using the powerful NUOPT for S-PLUS optimizer and the S-PLUS computing environment for financial analytics on a wide variety of examples.NUOPT is used because it is one of the leading highly scalable commercial optimizers that includes a variety of mature methods for solving LP and QP problems, including interior point methods, simplex methods, active set methods and mixed integer programming, and because it has a rich language (SIMPLE) for setting up portfolio optimization problems. The modern optimization methodologies covered include a wide range of attractive alternatives to standard deviation as risk measure objective function, such as expected utility function maximization via scenario based optimization, use of semi-variance and lower partial moments, and use of the increasingly popular and important conditional-value-at-risk (CVaR) downside risk measure. The optimization methodologies include treatment of a wide variety of portfolio weights constraints encountered in practice, treatment of transaction costs, and choosing a best subset of assets via NUOPT mixed -integer programming. A number of modern statistical methodologies in support of portfolio construction are treated, including resampling methods, robust statistical methods for dealing with outliers, and Bayesian methods that allow incorporation of prior information. Bootstrap type resampling methods allow one to assess the oft-neglected variability in efficient frontiers and Sharpe ratios. Robust methods serve as useful diagnostic for determining whether or not one or a small number of outliers unduly influence a portfolio choice, prompting the portfolio manager to look more carefully at the allocation choice in making a final decision. Bayesian methods allow a portfolio manager to use his/her prior information as an input to the portfolio construction process. Alternatively, the manager may wish to use an empirical Bayes method such as the Bayes-Stein method for estimating mean returns, or a hierarchical Bayes model that constructs priors through the use of cross-sectional information and Markov Chain Monte Carlo methods. Collectively, the above methodologies promise to improve the practice of portfolio construction.Finally, the book provides a large number of examples in the form of NUOPT SIMPLE and S-PLUS code, both short code segments and longer scripts and functions, applied to a variety of artificial and real financial returns.
(0 Comentarios)
