Read the full-text of this article for free

Title: The Relative Performance of Corporate Bond Funds 1991-2003: A Perspective from Pattern Recognition

Authors: James Moore

Addresses: Author address listing can be found in the "About the Authors" section at the end of the article.

Abstract: This study employs a rule induction expert system to search for patterns among characteristics of individual mutual bond funds that could predispose the fund to a specific measure of long-term performance relative to its peer index composite. Individual bond funds are categorized as either 'Outperforming', 'Matching', 'Underperforming' a benchmark index, or as 'No-Longer-Traded'. The system's training period is the nine-year interval from year-end 1991 through year-end 2000. The rule induction results are compared v^iith those of multiple discriminant analysis, with the induced rule structure generating a considerably better classification 'hit-rate' than multiple discriminant analysis in this application. The induced pattern is then used to forecast individual bond fund performance beyond the training cases and beyond the original training period, through year-end 2003. The pattern recognition capability of rule induction detects a substantive role for short-term persistence, portfolio maturity, the percentage change in the fund's asset holdings, and portfolio yield. These factors are operational surrogates for recent historical performance, the activity level of actively managed funds, and asset portfolio characteristics. All such critical factors are readily accessible to the ordinary mutual fund investor

Keywords: Corporate bond funds; pattern recognition; rule induction; long-term performance; mutual funds.

DOI: 10.1504/JBM.2005.141132

Journal of Business and Management, 2005 Vol.11 No.2, pp.7 - 27

Published online: 05 Sep 2024 *