Identifying and correcting misclassified South African equity unit trusts using style analysis

Return-based style analysis of domestic targeted absolute and real return unit trust funds in South Africa

“…If a factor lower than 75% with regard to the style benchmark was found, the fund was deemed to be misclassified. In addition, the research of DiBartolomeo and Witkowski (1997), Robertson et al (2000) as well as Sáez and Izquierdo (2000) is lacking with regard to calculating the statistical significance of the constraint regressions coefficients dealt with by Lobosco and DiBartolomeo (1997) later in this study. 39 Buetow, Johnson and Runkle (in Buetow & Rutner 2000:27) emphasise that the results of RBSA can be very volatile for particular types of mutual funds while also acknowledging the risks of multicollinearity.…”

“…An interesting study by Gallo and Lockwood (1999) (1997) and Robertson et al (2000) respectively that misclassification of funds do indeed occur although the misclassification according to Sáez and Izquierdo (2000) is slightly lower than that according to DiBartolomeo and Witkowski (1997).…”

“…According to DiBartolomeo and Witkowski (1997), 40% of the funds in the study were misclassified and 25% of the misclassified funds categorised in a classification lower than the inherent risk of the fund. Robertson et al (2000) applied the methodology employed by DiBartolomeo and Witkowski (1997) to the South Africa general equity unit trust category. Although the sample was much smaller than that used in the DiBartolomeo and Witkowski (1997) 18 "The percentage of funds that change classification each year" (Brown & Goetzmann, 1997:391 Confidence intervals on the individual style weights though are indicative of the quality of fit for individual style weights.…”

“…The researchers concluded that 40% of the funds were misclassified based on the results of the RBSA. Robertson et al (2000) analysed the South African equity unit trust category by means of RBSA. Their initial results showed that more than half of the general equity unit trusts were misclassified.…”

“…weights adding to 1) and style weights being non-negative (Lobosco & DiBartolomeo, 1997;Sharpe, 1992;DiBartolomeo & Witkowski, 1997;Robertson et al, 2000). 6 Regression analysis whereby b is constrained in that the sum of all coefficients has to equal 1.…”