How to Benchmark Private Capital... and How Not to
Luis O’Shea and Vishv Jeet, Burgiss
Is your private capital benchmark producing numbers that are not meaningful? In fact, is your benchmark unable to answer certain important questions at all?
In this working paper we examine how a private capital benchmark should be constructed. Although we have observed limited partners employing a wide variety of benchmarking methodologies, we feel that there is a single correct way to do so. Furthermore, we feel that the advantages of this approach are not just a matter of opinion.
In our paper we argue that the correct approach is superior because it adheres to well-established guidelines for benchmarks, such as the so-called “Bailey Criteria,” and because alternative approaches may generate performance numbers that are deceptive. Briefly, the right approach is to simply treat the benchmark as a bona fide portfolio, from which all performance measures are computed by pooling.
To set the scene, note that benchmarks come in several flavors. Some are “aspirational,” such as CPI + x%. Others are used for allocation purposes, such as in deciding between public and private equity. In our paper, however, we focus on benchmarks used to understand the relative performance of a private capital portfolio relative to the broader asset class of private capital.
These benchmarks should enjoy a number of characteristics. For example, in order to produce meaningful performance measures, such a benchmark must be replicable; in other words, there must exist some set of commitments which produce the benchmark performance numbers. Or, as the Bailey Criteria author might put it, there must be some way, at least in principle, for an investor to realize the benchmark performance numbers. In addition, such a benchmark must be customizable; not every portfolio will commit in the same ratio to buyout versus venture capital, or follow the same pattern across vintages. And unlike public markets, adjusting these patterns after the fact is difficult.
The requirements in the previous paragraph lead to what we call the Pooling Methodology: “A benchmark should consist of a portfolio of commitments to every fund in a representative universe, in such a way that they mirror the investor's target allocations across fund characteristics”. Thus, a benchmark is a true portfolio, and all benchmark performance measures should be computed by first pooling cash flows or valuations.
Note three things about this approach. First, because it is a portfolio the benchmark is, by definition, replicable. Second, since it is a portfolio, any benchmark measure can be computed (including long-term measures such as the total value to paid-in multiple, or TVPI). Third, there is nothing novel about the approach. In fact, outside of private capital it is universal.
Having focused on pooled measures, the working paper turns to rank measures, such as determining whether a fund is top quartile. These measures are easy to compute for isolated funds. The working paper describes how the same Pooling Methodology leads to a natural way to compute rank measures for portfolios with commitments to multiple funds, meaning LPs can now determine whether their entire portfolio is top quartile.
We hope that the working paper convinces its readers that the Pooling Methodology is the right way to construct a benchmark. But what of other approaches? Generally, these focus on computing a short-term benchmark return (e.g., a quarterly return) by taking a weighted average of the returns of various subgroups, such as vintages or asset subclasses, as opposed to pooling the cash flows.
It turns out that these approaches have a couple of shortcomings. First, they focus exclusively on short-term returns and provide no information about the benchmark from a long-term perspective, such as TVPI, multi-year IRRs, or other measures usually considered critical to understanding private capital portfolios. Second, the returns of these alternative approaches are not replicable.
This lack of replicability results in returns that can be far from the true returns of the asset class. For example, in Figure 1 below the blue line represents the cumulative Time-Weighted Rate of Return (TWRR) computed using the Pooled Methodology. As expected it is similar to the typical behavior of US buyout funds. The red line is the result of employing one of these alternative methodologies. As can be seen it results in a very noticeable distortion to the returns of the asset class.
In short, there are no shortcuts if one wants a benchmark that is both accurate and informative. It must be constructed using the Pooling Methodology. Furthermore, not doing so leads at best to a benchmark which is unable to answer important questions, such as the TVPI of the benchmark, and at worst generates returns that are not an accurate representation of the asset class that is being benchmarked.
Luis O’Shea, PhD, is Head of Applied Research, and Vishv Jeet, PhD, is an associate director at Burgiss.