股票外汇技术分析英文版 (2).pptVIP

April 27, 2005 ? Gertsch Wachob Fuqua Investment Analytics Stefan D. Gertsch Brian Wachob Purpose of Study Build a quantitative stock selection model to guide portfolio management at a long/short hedge fund. Adopted perspective of a fund with $100M in assets under management This fund size estimate guided investable universe definition. Assumed goal of maximizing returns with negligible correlation to other asset classes In practice, we did consider variance as well, but it is of much lesser importance if correlations with other asset classes are indeed negligible. Overview Definitions: Universe, Methodology, Conventions Cross-Sectional (Time Invariant) Factor Research Identification and Evaluation Factor Refinement Negative Factor Values and Counter-Intuitive Resultant Rankings These concerns are often overlooked by others and arise among very common factor definitions (such as forward earnings yield) Industry Normalization Optimal Fractile Resolution and Clustering/Groupings/Aggregation Integration of Factor Portfolios into a Multivariate Model Based on Mean-Variance Portfolio Optimization with the Imposition of Custom Constraints Multivariate Model’s Out-of-Sample Performance Forecasting Factor Portfolio Returns Demonstration of dynamic factor weightings based on factor performance forecasts Universe Definition U.S. stocks only Time-scaled min. market cap threshold $54M in 1987, rises by 7% annually to $200M in 2005 Time-scaled min. estimated mean daily dollar volume threshold $84K in 1987, rises by 10.4% annually to $500K in 2005 Excluded ETFs and other investment funds trading as stocks Excluded 3 instances of likely erroneous Compustat returns data Number of stocks in universe grows with time 998 on 1/31/1987 2,649 on 11/31/2001 ≈3,336 today (4/27/2005) Methodology / Conventions In-sample period: February 1987 – December 2001 Out-of-sample period: January 2002 – March 2005* FactSet Alpha Testing fractile sorts Monthly rebalancing, 1 month holding period