Ebook Asset Allocation and Portfolio Performance: Evidence from University Endowment Funds
Asset allocation—the process of distributing investment capital across the various asset classes in an allowable universe—is widely regarded as one of the most important decisions an investor faces. The ultimate goal of this process is to construct portfolios that are optimal with respect to some prespecified objectives. According to the paradigm of modern portfolio theory that originated with Markowitz (1952), the exercise of constructing optimal portfolios is ultimately one of balancing expected returns against their contribution to portfolio risk.
In the investment management industry, it is commonly accepted that an investor’s initial strategic asset allocation decision is the most important determinant of the portfolio’s investment performance (see, e.g., Brinson, Hood, and Beebower (1986), Brinson, Singer, and Beebower (1991) and Bogle (1994)). However, empirical evidence on mutual fund and pension fund investment practices seem to cast some shadows on this belief. Both Ibbotson and Kaplan (2000), using data on U.S. mutual and pension funds, and Blake, Lehmann, and Timmermann (1999), using data on U.K. pension funds, conclude that while asset allocation decisions are the major determinant of return variation over time, they are considerably less important in explaining return variation in the cross section.
Our analysis is based on portfolio information and performance statistics for more than 700 public and private university endowment funds collected in two separate and distinctive data sets: (i) a series of annual surveys from 1984 to 2005, administered by the National Association of College and University Business Officers (NACUBO), an advocacy organization devoted to improving management practices in the higher education industry; and (ii) proprietary self-collected quarterly data from 1994 to 2005. To isolate the part of returns originating from the asset allocation decision we follow the methodology proposed by Brinson, Hood, and Beebower (1986) and decompose the total return of each endowment into three components related to: (i) the strategic asset allocation (policy) decision; (ii) the tactical asset allocation (market timing) decision; and (iii) the security selection decision. The strategic asset allocation decision is often referred to as the passive element of a fund manager’s decision-making process while market timing and security selection are the active components of this process.
Using the asset allocation return obtained from such a decomposition, we generalize the tests of Ibbotson and Kaplan (2000) and study the contribution of strategic asset allocation to endowment return variation. Consistent with their findings for U.S. mutual and pension funds and with those of Blake, Lehmann, and Timmermann (1999) for U.K. pension funds, we also find that asset allocation still emerges as the main determinant of return level and variation in the time series. Its contribution to time series return variation is about 75%, somewhat lower than the values documented for other institutional investors. More strikingly, however, the average contribution of an endowment manager’s asset allocation decision to cross-sectional variation in performance is only about 10%, which is once again significantly lower than previously established in other institutional settings. This evidence seems to indicate that the average endowment manager follows a much less passive investment strategy than what appears to be the norm for either mutual fund or pension fund managers.
Having access to a detailed panel of actual portfolio weights of endowments allows us to better understand the nature of this discrepancy in explanatory power of the asset allocation decision in the time series versus the cross section. We demonstrate that the limited amount of cross-sectional explanatory power associated with the policy return component originates from a remarkable lack of variation in the ex-post returns attributable to the strategic asset allocation decision. Conversely, we also show that asset allocation weights vary dramatically across the endowments in our sample. While the first finding is consistent with what Blake, Lehmann, and Timmermann (1999) document for U.K. pension funds, the second is in clear contrast with the homogeneity in asset allocation weights in their sample.
Our findings have interesting implications for the role of active management in the performance of university endowments. The largely invariant sample-wide level of passive risk we document implies that endowments target a common level of volatility for their policy portfolio, thus ending up with very similar passive returns. Given that total returns are the sum of the passive and active return components, a common level of passive return across endowments means that any cross sectional variation in overall performance must come from the active decisions within the portfolio. We therefore investigate how endowments who rely more on security selection (active endowments) fare in comparison to endowments who rely more on asset allocation (passive endowments). Our main finding is that active endowments significantly out-perform passive ones, despite the fact that, as a group, university endowments do not seem to produce significant risk-adjusted returns. The top quartile of active endowments have risk-adjusted returns that are 2.92 to 8.39% larger than those of the bottom active quartile. This suggests that the documented heterogeneity in portfolio weights across funds represents an attempt by endowment managers to select their exposures to broad asset classes based on both their familiarity and selection abilities within that class.
To the best of our knowledge ours is the first study that attempts to quantify the relationship between the asset allocation decision and investment performance for a comprehensive sample of college and university endowment funds. Much of the previous literature in this area has been mainly concerned with understanding the nature of the endowment investment process, with relatively little being known about how these portfolios have actually performed over time. Two more recent studies have also used data from NACUBO as we do. Dimmock (2008) uses one year of data from the NACUBO Endowment Survey to assess the role of background risk (proxied by non investment income volatility) on endowment portfolio choice while Lerner, Schoar, and Wang (2008) rely on similar data to document that Ivy League school endowments have performed much better than non-Ivy league schools in managing their commitments to alternative investments. Finally, Lerner, Schoar, and Wongsunwai (2007) document that endowments have exceptional abilities in selecting the right venture capital partnerships. These last two papers emphasize how some endowments excel in their security selection process. Our paper completes and extends these findings by showing that it is not the returns to a few selected market segments (e.g. alternative assets) that drives the performance of these institutions, but security selection as a whole across the entire asset class universe that is the key determinant of an endowment’s overall success.
The remainder of the paper is organized as follows. In the next section, we describe and summarize the endowment data used in the study. Section 3 rigorously defines the concept of passive asset allocation as part of an endowment’s portfolio while Section 4 relates the variation in passive returns to the variation of total fund returns. Section 5 formally tests the relationship between asset allocation and performance for endowment funds and Section 6 concludes the study. Appendix A contains useful results from the Treynor and Black (1973) model that serves as a basis for some of our tests.
Download
PDF Ebook Asset Allocation and Portfolio Performance: Evidence from University Endowment Funds
Posted in :
