Calculations of risk are supposed to be stable over time, particularly if one uses use an interval of five years or longer. Unfortunately, they are not stable.
Standard deviations vary from less than 10% to more than 25%, depending on the starting date. However, if we can accurately estimate the style of the manager with William Sharpe's returns-based style analysis, we can get a reliable estimate of the risk associated with a manager's style. Some multiple of this risk is subtracted from the one- and five-year returns to produce a risk-adjusted return. This is the procedure used to calculate the risk-adjusted returns. in this report. The style beta indicates the degree to which a manager takes more or less risk than the passive style associated with that manager. For further details, visit Pensions & Investments' Web site: www.pionline.com/whitepaper/-sortino.
This quarter, I am using mutual fund data from LCG Associates, Atlanta, instead of Morningstar Inc., Chicago. I believe the data is more precise due to fewer rounding errors. Also, indexes used to calculate the styles are now the Frank Russell style indexes. Russell indexes weren't used before because they did not go back far enough to give reliable downside risk calculations. Gary Miller, chief financial officer at LCG, has corrected this shortcoming by using statistical procedures to fill in the missing data to 1970. The result is a substantial improvement over other indexes we have tried. Only one fund has an R squared less than 0.75 (Merrill Lynch Growth was 0.72). Only funds with seven-year performance histories were used to calculate the style betas, because it has been seven years since the last down market.
This analysis is intended to assist the reader in understanding the statistics in the accompanying risk-adjusted return table. A passive style benchmark was generated for each fund. The benchmark represents an alternative strategy to active management. AIM Constellation Fund took 97% as much risk as the benchmark (style beta of 0.97). For the last year, the risk-adjusted return was negative. For the five years ended March 31, the risk-adjusted return was 10.3%, as opposed to the nominal return of 15%. In other words, 470 basis points were subtracted.
Nevertheless, on a risk-adjusted basis, the fund produced 460 basis points more than the passive benchmark. An investor could have replicated this fund's style with a high degree of accuracy (89%), but the fund beat the benchmark by a wide margin, after adjusting for risk.
Frank Sortino is director of the Pension Research Institute, Menlo Park, Calif.