The choice of investment performance benchmark can make an investment manager look bad or good, regardless of skill. This is particularly true for 1999 numbers, when the spread between value and growth was 4,500 basis points.
Most would acknowledge that the Standard & Poor's 500 stock index is an inappropriate benchmark for most managers for 1999, but most have made the comparison anyway. Because of the importance in investment performance evaluation, Pensions and Investments recently has published two "Other Views" commentaries on the subject of style benchmarking, which certainly is preferable to the one-size-fits-all S&P 500 approach.
In the April 17 issue of P&I, Bob Barkley documents some deficiencies in the S&P/BARRA and Russell indexes. Mr. Barkley argues that 35% of the securities in the value indexes of these firms should be classified as growth. The reasons for these misclassifications are definitional and computational. Mr. Barkley defines high price/earnings and high price/book as growth, while S&P and Russell use additional measures.
But the real culprit is probably computational, and it deals with the frequency of rebalancing. June is an event month for the Russell indexes, when the annual rebalancing occurs. Some stocks whose prices have risen are reassigned to growth from value, and/or to a larger size category. In the current volatile market environment, infrequent rebalancing will result in frequent misclassifications. Today's cheap Internet idea can become a "success" overnight.
In the May 1 issue, Tom Richards and Sandra Weiskirch properly admonish "evaluate skill, not style." They contend that all style indexes, particularly the S&P/BARRA and Russell style indexes, are inappropriate for performance evaluation because most managers have their own unique definitions of stock styles. Accordingly, a value manager could beat a value index simply because his definition of value is different from the index provider's definition. In 1999, so-called "relative value" managers beat the value benchmarks because they held securities that were in the growth indexes. Mr. Richards and Ms. Weiskirch recommend custom benchmarks, or normal portfolios, as a solution.
An alternative exists to custom benchmarks that, while less precise, is easily constructed and, if done properly, reasonably accurate: returns-based style analysis.
RBSA regresses a manager's returns against a family of style indexes to determine the combination of these indexes that best tracks the manager's performance. Several recent studies have used RBSA to show that most managers employ some blend of styles, so no one style index is appropriate for a particular manager.
The style profile produced by RBSA can be viewed as a poor man's normal portfolio. It's not as robust as a carefully constructed custom benchmark, but generally far better than picking a single generic style index. The manager's benchmark is its RBSA style profile.
Some concerns about style indexes go away with RBSA, but you still want to set your style palette as best you can. Without getting into the science of regression analysis, it is well understood that the family of style indexes should be mutually exclusive and exhaustive. This means that no one stock should be in more than one style category and the collection of securities classified should comprise the entire market. An additional desirable characteristic is fairly frequent rebalancing.
None of the widely used generic style indexes meet the desired criteria for sound RBSA. All discard data, violating the exhaustive criterion. Many classify stocks into more than one category, violating the mutually exclusive criterion. All rebalance infrequently. All, that is, except style indexes I developed eight years ago, now called Surz indexes.
My style indexes meet all of the desirable requirements, yet they are not often used because my name is not well known. Two service providers of RBSA, Mobius Group Inc. and Zephyr Associates Inc., have conducted extensive research to identify the best family of indexes for RBSA. These tests focus on which indexes give the most accurate and consistent style profile -- the best fit. Mine have won these contests, but both of these service providers default to Russell indexes because the name is well recognized, although they do make mine available. One criticism of my indexes has been the frequency of their availability, which is quarterly. For most applications this isn't an issue, but nevertheless Surz indexes will be available daily by year end.
The problems with benchmarks can be solved. RBSA and custom benchmarks are two such solutions, but care must be taken in implementing either. For a good RBSA you'll need state-of-the-art regression technology and a sound style palette. For a good custom benchmark you'll need extensive research of your manager's process and philosophy, and a rigorous review of past portfolio holdings.
You may wonder if it's worth all of the effort. The answer depends on how costly mistakes can be in performance evaluation. What do you think?