Quantitative Stock Selection.pptVIP

Quantitative Stock Selection April 27,2005 Agenda Review of Original Model Industry Rescaling Extension of Dynamic Factors Introduction of Fractile Migration Possible Areas for Further Analysis I. Review of Original Model Factors Subjective Scores Heat Map: EW Returns II. Industry Rescaling Used North American Industry Classification System to Scale by Industry Rescaling increased return, 21.4% vs. 19.3% while decreasing standard deviation, 12.7% vs. 19.9%!! Overlay portfolio consistently produced positive alpha Takeaways about industry rescaling…. For our model Intra-industry effects has historically been larger then inter-industry effect Gets rid of huge anomaly in our out of sample data for 2003 Can a middle ground or new weighting scheme be found to capture both inter and intra effects? III. Extension of Dynamic Variables Dynamic Model With Incremental Term Structure Varying Weights Improved Model Outperforms Over the Period…but Fails in 1999 Overlay portfolio produced positive alpha in fifteen out of seventeen years Takeaways about dynamic factors… Examined a number of interaction variables: Growth and Value Regimes Credit Spreads Term Structure of Interest Rates Incremental scoring adjustments using yield curve information added the most value to our model. Takeaway about dynamic factors… Time varying “dynamic” weights can be very useful in increasing the power of quantitative models: More information Interaction effects Adaptive models However, finding a consistent model can be tricky: Overfitting Lots of noise IV. Introduction of Fractile Migration Preliminary Information In-sample period: 01-Dec-87 thru 31-Dec-98 Out-of-sample period: 01-Jan-99 thru 31-Dec-04 Monthly return data Based upon original model Count of Returns Average of Returns Average of Returns Fractile Stability Turnover Implications Results Conclusions No substantial benefits for long portfolio via analysis of fractile migration … … Short portfolio, however, t