Benchmark misfit is calculated as the difference between the return of the asset class benchmark and the weighted average return of all benchmark mandates assigned to individual asset managers. In other words, misfit exists when the sum of the assigned parts does not equal the desired whole.

Benchmark misfit can be decomposed into two categories: (1) gaps and overlaps; and (2) allocation misfit.

Misfit #1: Gaps And Overlaps

Gaps and overlaps occur when the subasset class benchmarks are mixed and matched among different index providers. For example, many in the investment community use the S&P 500 Index and the Russell 2000 Index as the standard benchmarks for large-cap and small-cap domestic equities, respectively. Figure 1 illustrates an example of the gaps in market coverage when the large-cap and small-cap benchmarks are set to the S&P 500 Index and Russell 2000 Index. The asset class benchmark has been set to the Dow Jones U.S. Total Stock Market Index, which covers the entire opportunity set of all U.S. equity securities with readily available prices.

As of Jan. 1, 2009, the gap in market coverage of the S&P 500–Russell 2000 combination results in 2,165 missing constituents, which is nearly half of all available constituents. This equity gap leaves over $1 trillion, or 11 percent, of the U.S. stock market unaccounted for. The median market capitalization of the missing 2,165 stocks is $28 million, and the median market capitalization of the top quartile of those stocks is $1.51 billion, and includes such names as Genentech ($38.44 billion), Visa ($22.53 billion) and Accenture ($18.03 billion). No reasonable investor should use a benchmark that excludes the top 11 percent of the U.S. equity market, the bottom 11 percent or any other 11 percent.

Misfit #2: Allocation Misfit

Allocation misfit exists when asset allocations deviate from the actual market coverage of the asset class benchmark. Figure 2 illustrates an example of allocation misfit. A hypothetical investor has decided to benchmark 85 percent of their portfolio to the large-cap subasset class, 15 percent to small-caps and no exposure to micro-caps. The actual market coverage of the policy benchmark is listed as of Jan. 1, 2009. The result of these decisions (i.e., underweighting large-cap, overweighting small-cap and underweighting micro-caps) is approximately 9 basis points in benchmark misfit. Those 9 basis points represent a performance mismatch that can be directly attributed to the allocation decisions exclusive of manager performance.

Compounding The Problem

Figure 3 illustrates an example of the compounding effect of the allocation misfit along with gaps in market coverage. Using the benchmarks from Figure 1 and the subasset allocation decisions from Figure 2, the result of these decisions (i.e., ignoring ~500 mid-caps and 1,665 micro-caps) is, after one year, approximately 104 basis points in benchmark misfit. Stated differently, the decision to deviate from the actual market coverage of the asset class benchmark has, in this case, “cost” the investor 104 basis points. Note that the decision is on the part of the investor, who has responsibility for assigning the individual benchmarks and weights. The misfit excludes any “alpha” which, whether positive or negative, is generated by the aggregate decisions made by the managers. This distinction is critical.

Achieving A Zero Misfit Portfolio

Figure 4 illustrates a comprehensive asset class benchmark. The subasset class benchmarks and weights are specified in advance and properly aligned with it.

Under this scenario, investment performance can be directly attributed to each mandate decision without any benchmark misfit. Any mandate that deviates from this baseline opportunity set will then not allow an investor to properly measure the impact of their own decisions on their own portfolio for a given asset class. The benchmark misfit analysis must precede—and remain separate from—the active versus passive decision as well as the decisions on the part of their selected managers in attempting to generate alpha.

The Cost Of Inappropriate Benchmarking

Benchmark misfit is an investment decision that leads to uncompensated risk. Prudent investors do not take uncompensated risks because they do not receive additional return for doing so. Therefore, benchmark misfit needs to be properly measured and monitored.

Using an inappropriate benchmark can result in misleading information being provided to all parties involved in the investment process. If a manager is outperforming the assigned benchmark, which is not an appropriate benchmark, an investor may believe the manager is successful, when in fact they may be under-performing by a material amount. In such cases, an investor may be paying for less-than-expected performance. The investor has spent more and received less while being led to believe by the benchmark misfit that the manager is doing well.

This can also occur in more-subjective benchmark assignments, especially when managers are ranked relative to a peer group. Unfortunately, peer group analysis falls short of being a good benchmark because the average is neither investable nor specified in advance. Further, the peers may actually be a poor proxy for the asset class or subasset class. Finally, the practical implementation of gathering peer group data of sufficient scale and relevance is quite complicated and often incorporates significant biases.

From One To Many

In practice, investors hold more than just equities or any single asset class in their portfolios. They more typically assemble a portfolio with a number of diversifying asset classes (even if just two). However, benchmarking such multi-asset class portfolios is not a straightforward exercise since it is not simply an extension of the single asset class approach. Although the decision to use a given benchmark in both single and multi-asset class portfolios may be driven by similar purposes and mechanics, the process for establishing a multi-asset class benchmark is fundamentally different than that for a single asset class benchmark.

Many single asset classes have a known universe of securities or investment options that constitute “the market” and can be used to establish a nearly definitive single asset class benchmark. Other single asset classes, such as commodities, do not have such a defined market, so reasonable proxies are established and marketed with one or two emerging as the standard.

In either case, any single asset class benchmark is designed to identify the aggregate performance of the class and allow for standardized comparisons between and among market participants and managers within that class. Advisers can then measure the implementation skill, or “alpha,” behind a specific investment assignment.

However, a multi-asset class benchmark is generally designed to measure something else—the total alpha of all such investment assignments plus the alpha, if any, of a policy portfolio itself.

What is a policy portfolio in a multi-asset class context? It is the mix of asset classes and weights derived from the set of short- and long-term investment objectives and capital market assumptions for an investor or investment plan. This mix may change over time in certain ways to reflect changing objectives (e.g., liabilities) or assumptions (an attempt to capture excess returns from short-term asset class price fluctuations in (market timing).2

Establishing a policy portfolio is not an easy task. Numerous challenges and complexities arise almost immediately.

• How are these decisions made in practice?
• How are specific asset classes included?
• How are the weights set?
• Where do the capital market assumptions come from?
• How do these decisions evolve over time?

For example, multi-asset class efficient frontiers are based on, and highly sensitive to, their inputs and may contain volatile or inaccurate data (private equity and venture capital performance data come to mind), making it difficult to gain agreement on what that efficient frontier looks like. This is further complicated by the fact that there is no standing consensus on the asset-class palette; after all, the simple decision to invest strictly in domestic equities (or any other set of asset classes) creates the opportunity cost of not investing in some other market, such as international equities, commodities, real estate or bonds, to name just a few.

Establishing a policy portfolio is not a trivial exercise either. Analysis shows that the asset allocation decision explains about 90 percent of the variability of a fund’s returns over time, and on average across funds, explains approximately 100 percent of the level of returns.3 The constraints and decisions that drive the policy portfolio, whatever they may be, clearly have a big impact and must be evaluated appropriately.

Of course, an investor (and their adviser or consultant) must construct a policy portfolio even in the absence of specific guidance on these matters. After all, establishing an investment policy, delineating responsibility and measuring the performance contribution of those activities form the core of the investment management process.2

Yet benchmark identification is routinely dismissed as a lesser element of the process; since there is no single definitive approach, the ultimate decision is viewed as relatively inconsequential.

This perception is far from accurate. The need for benchmarking is heightened, not diminished, in a multi-asset class environment given the much wider spectrum of required decision making and investment opportunities.

The Case Of Target Date

In recent years, the individual retirement market has witnessed an explosion of “target date” products. These are multi-asset class products that dynamically and automatically evolve an investor’s “policy portfolio”—their asset class mix and weights—at each stage of an investor’s working life.

The benchmark process in this realm is very challenging. How should such evolving policy portfolios be measured against the finite horizons of individual investors who have real-world and often conflicting needs that may vary with age, existing wealth, living expenses and personal circumstances?

Given existing tools and capabilities, there may be a reasonable approach to benchmarking. In general, the proper design of a portfolio involves at least four decisions 2: (1) the asset-class palette; (2) strategic asset allocations; (3) tactical asset allocations (market timing); and (4) security selection. The first two decisions form the core of the policy portfolio, and should be the focus of benchmarking in a multi-asset class environment.

Ultimately, though, there will be ongoing debate about the objective nature and best practices of benchmarking such multi-asset class portfolios. The different approaches will each bring their own strengths and weakness, but the growing competition can only help benefit the market drive toward the optimal approach.

Jeffrey Fernandez, project manager, Analytics & Research, Dow Jones Indexes, contributed to this article.

References

1 Bailey, Jeffery V., “Are Manager Universes Acceptable Performance Benchmarks?” Journal of Performance Management, Spring 1992.

2 Brinson, Gary P., L. Randolph Hood, and Gilbert L. Beebower, “Determinants of Portfolio Performance,” Financial Analysts Journal, July/August 1986.
http://www.cfapubs.org/doi/pdfplus/10.2469/faj.v51.n1.1869

3 Ibbotson, Roger G. and Paul D. Kaplan, “Does Asset Allocation Policy Explain 40, 90, or 100 Percent of Performance?” Financial Analysts Journal, January/February 2000. http://corporate.morningstar.com/ib/documents/MethodologyDocuments/IBBAssociates/AssetAllocationExplain.pdf