Active vs. Passive Debate Aside, Ivy Endowment Allocations to Private Markets Appear Justified

The endowment model, and active management in general, has come under increased scrutiny, while indexed, or passive, products have grown in popularity and number. Regardless of where you stand on that debate, it's hard to deny that the Ivies approach to asset allocation has been very good.

December 14, 2018

Pioneered by Yale endowment CIO David Swensen, the endowment model is characterized by long-term investment horizons and increased risk budgets, with significant allocations to alternative asset classes, including private and illiquid investments. As such, allocations by Ivy endowments to private equity, venture capital and real estate have increased significantly in recent years.

Over that period of time, the endowment model, and active management in general, have come under scrutiny, while indexed, or passive, products have grown in popularity and number. Our own analysis of Ivy performance showed that every Ivy endowment underperformed a passive 60-40 Equity/Bond portfolio (a common endowment benchmark) over the past 10 fiscal years (FY2009-FY2018).

This leads to an obvious question: is the endowment model, with its focus on private and niche investments, justified?

To answer this question, we compared the asset allocations of Ivy endowments to the most efficient mix of the same assets with “hindsight 20/20” historical inputs (risk, returns, correlations) for the major asset classes utilized by endowments. In other words, given perfect foresight on the performance of asset classes over the last 10 years, what would be the optimal mix for a specific risk budget pertaining to each endowment portfolio?

Regardless of where you stand on the active vs. passive debate, there’s one thing that is quite clear about the Ivies when viewed over the long-term: their approach to asset allocation has been good. Between fiscal years 2009 and 2018,1 the Ivies managed, on average, to assume exposures that were very close to optimal, when analyzed against the efficient frontier and using our patented resampling technique.23

The below chart puts the risk and return characteristics of each Ivy endowment in context.4 We see that some Ivies (Penn, Columbia and Princeton) are closer to the efficient frontier line than others. The 60-40 portfolio (dark blue) is the closest to the frontier and has higher returns than all of the Ivies, as per our recent study.  At the same time, however, endowment portfolios are taking on much higher risk—50 percent or more than that of the 60-40 portfolio. Consequently, they are scattered to the right of the 60-40 portfolio on the below efficient frontier diagram.

In the below chart, we show allocations of efficient portfolios along the frontier corresponding to each level of risk (X-axis). It’s interesting to note that this starts with a 50/50 allocation to bonds and equity (public + private) as the lowest risk allocation and increases allocations to private investments and public equity as risk budget increases. We highlight the most efficient portfolio with a StdDev of 12 percent (labeled Ivy Efficient), which is the StdDev of a portfolio consisting of the average Ivy exposures between fiscal years 2009 and 2018.5

In the following chart, we drill down to show how the allocations of the average Ivy portfolio compare with those of the most efficient portfolio corresponding to the same level of risk (with a StdDev of 12 percent corresponding to the yellow point on the above efficient frontier chart). It’s worth noting that the hindsight 20/20 “Ivy Efficient” portfolio has more than 65 percent allocated to private investments (private equity, real estate, venture capital), which is very much in line with the allocations of most of the Ivies. This signals that even after the inherent risks of private investments were taken into consideration by our approach, the returns and diversification effect of this investment class are still super attractive—in retrospect.

We observe that allocation to hedge funds and real estate was higher in the average Ivy portfolio, which is expected given that implementation constraints are much more elaborate and stricter than the naive 50-percent maximum we used for our study. But even taking this into consideration, the average and efficient Ivy portfolios are strikingly similar. And, despite criticism of the endowment model, the large allocations to private assets such as private equity and venture capital were seemingly fully justified by our retrospective ex-post analysis—even for this very challenging decade. That said, these latest findings underscore the fact that the endowment model is not for the average investor, due to the high level of exposures to illiquid assets, which magnifies risk and poses significant illiquidity concerns.

Footnotes

  • 1The Ivy endowment fiscal year is July 1 to June 30.
  • 2We use resampling to take into account any uncertainty in the parameters passed into the optimizer, namely the returns and covariances. We then resample both returns and covariances multiple times and then map them back to the frontier based on a proprietary MPI approach. We used a max 50-percent allocation constraint for each asset class as a realistic boundary.
  • 3To demonstrate this, we explore what the allocation of a purely data-driven portfolio construction approach would have been under perfect hindsight: defined as using historical estimates of risk, return and correlations among asset classes based on realized investment returns. Our analysis therefore is purely in-sample, aiming at an historical evaluation on how well the endowment model worked. It is based on a mean-variance optimization with resampling, using the factors that Ivies invest in. For private equity, venture capital and real estate, we apply a de-smoothing approach as described in our recently published paper in order to obtain proper risk and correlation estimates, while making sure the overall annualized return of the de-smoothed series is the same as the one of the original series.
  • 4For this period, which uses quarterly returns that span fiscal years 2009 to 2018, we also plot the Ivy factor mimicking portfolios we previously used to estimate Ivy Sharpe ratios.
  • 5This allows us to view the optimal portfolio exposures with a StdDev of 12 percent. It also enables us to compare the Ivy Average and Ivy Efficient portfolios on the same risk basis, while also being able to see what’s inside the efficient frontier portfolio, along the entire frontier.
Tags
Comments
Leave a Reply
Fields to complete: