|
Creating and Managing Custom Benchmarks – A Practitioner’s Guide Authors: Stephen Campisi Source: Journal of Performance Measurement Date: Summer 2002 |
Abstract
The choice of benchmark is essential in portfolio performance measurement, in order to properly evaluate the results of portfolio managers. This benchmark must be suited to the strategy followed by the portfolio. In this article, the author first identifies the problems caused by the use of broad benchmarks. More specifically considering bond management, he then describes how to construct and manage custom benchmarks.
The author first enumerates the desirable properties of a benchmark. A benchmark should have the same systematic risk as the portfolio strategy, the same style characteristics and it should be highly correlated with the portfolio. Most of the time, the benchmarks used to evaluate the performance of portfolios are not representative of the portfolios. For US stocks, the most widely used benchmark is the S&P 500 Index and for bonds, it is the Lehman Brothers Aggregate Index. These indexes are broad and are not representative of the investment strategies, risks and style biases of most active investment managers. This causes misevaluation of portfolio managers’ performance. The author stresses that beta is not a good measure of risk for the portfolio, if the benchmark does not properly match the portfolio. Using bond and stock examples, he demonstrates that a portfolio with a beta equal to one relative to its benchmark may have a greater risk than the benchmark, and that a strategy with a beta of less than one can produce the same return. This can be explained by the inclusion in the portfolio of risky sectors that are not part of the benchmark.
Portfolio managers wish to evaluate whether the excess return produced by their portfolio is the result of their skill, or whether it is due to differences in risk and style between their portfolio and the benchmark. By analysing the investment process and the objectives and constraints of investors in detail, the author explains how to create suitable benchmarks for performance analysis, using a combination of readily available indexes. Two methodologies are presented. The first method proposes to create a benchmark using a quantitative, data-oriented method. The second one is based on an intuitive method, proceeding from the understanding of the client’s risk tolerance.
Campisi first gives some general rules for developing benchmarks. The best way to build a benchmark is to use indexes that have no overlapping securities. This enables the correlations between indexes to be limited and gives a better representation of the different styles. To form a bond benchmark, it is preferable to use indexes instead of individual issues, as the quality of the data will be better. For certain equity strategies, it may be necessary to use a select set of individual issues. In that case, it is best to limit the selection set to the most liquid and actively traded issues.
The data-oriented method for constructing benchmarks relies on the use of a simple optimizer with a set of historical index returns. It is nonetheless possible to derive the optimal allocation between the indexes. This allocation delivers the highest return for a given level of risk and enables an efficient benchmark to be obtained. It is possible to constrain this optimization so that the client’s risk tolerance is respected in each sector. The intuitive approach is based on an analysis of the objectives and constraints specified by the client in each investment sector. The intuitive allocation in each sector can then be refined by performing a constraint optimization.
According to the author, the intuitive method for deriving a custom benchmark gives a more accurate result, compared with the mathematical solution provided by the optimization. This may seem surprising, but Camprisi points out that the intuitive method is superior because it guarantees that the solution falls within appropriate ranges, whereas the data-oriented allocation may be more unstable and can lead to significant changes in sector weights to obtain a small additional return.
The last part of the article describes how to validate the derived custom benchmark. The tests performed show a high correlation between the portfolio and its benchmark. Using custom benchmarks, it is thus possible to analyse the performance of a portfolio properly.
