Markov Processes International

Bowdoin FY2021: How to Replicate a Brilliant CIO

In our first article in this series, we covered the $20.5B UPenn endowment’s strong returns for FY 2021, and asked what risks the future might hold for endowments concentrating more of their allocations in equities and illiquid investments.

It’s one of many questions Nyles Bryant faces as he takes over as SVP and Chief Investment Officer for the Bowdoin College Endowment. Outgoing CIO Paula Volent stepped down in June to take up the CIO position at Rockefeller University, and leaves behind a strong record of success.

One of her famously good picks was her 2005 investment in Sequoia Capital’s China-focused venture fund, but there are many more examples. As The New York Times points out, Bowdoin’s 8.8% average annual return through the end of FY 2018 outperformed all Ivy League endowments for the previous ten years.

Clearly Bowdoin’s approach paid off again this year. The endowment posted a 57.4% return for the FY ended June 30, 2021.

And that means Bowdoin has been outperforming all Ivies on a 10-year basis since 2015, with its latest FY2021 result bringing it to 14.4%, an almost impossible number to beat.

But investors seeking to confirm Bowdoin’s allocations and understand its returns are left with the same analysis problem as with any endowment – how to analyze endowments using only their annual returns?

We’ll be taking the same approach as we did with UPenn’s endowment: using our Dynamic Style Analysis (DSA) model to analyze annual endowment returns, and the returns of benchmarks and indices representing each asset class, to develop an initial view of Bowdoin’s dynamic asset class exposures (below).

We can immediately see here Bowdoin’s ongoing tilt towards private assets, with venture capital taking the lead. Our analysis shows this coming at the cost of hedge funds (absolute return strategies), which Bowdoin confirms in their annual reports.  In the chart below, we show the DSA estimates of where this leaves Bowdoin’s exposures specifically in their recent fiscal year.

In reviewing our analysis, we note that – while actual allocations are substantial – our initial analysis overestimates exposures to private equity (including VC) and underestimates exposures to hedge funds and public equities.

Bowdoin College Endowment Historical Allocations by Year

That said, the in-sample style/tracking portfolio we created out of benchmarks and indices does a very good job of mimicking the cumulative performance of the endowment itself:

This approach is using in-sample estimates, meaning that we use all of the available information to produce the analysis (in this case, the annual returns for both the endowment itself, and for the strategy indices in our tracking portfolio). And as we saw in our FY21 analysis of UPenn, diversification does wonders: once an endowment reaches a certain size (both in AUM and the number of investments), asset allocation is doing a pretty good job of explaining the variation of the endowment’s annual results, leaving very little to individual manager success or failures. We’ve seen this before in endowments (with some occasional aberrations, such as Brown) and also with large pensions, such as $1.4T Norges.

But we’re frequently asked how we know it’s the allocation (rather than alpha) that contributes to the success of the fund. It’s a reasonable question, and the best way to answer it is to perform an out-of-sample test.

We start by estimating exposures through FY2016 without “looking into the future”. We then take DSA-estimated exposures and use observed index returns from the following year (FY2017) to provide a forecast for FY2017. We repeat this year-by-year, expanding our estimation window (shown here) next to the realized endowment returns themselves:

The quality of replication is striking, especially given the fact that we’re working with annual data scraped from public endowment reports. There is one caveat – some of the factors are not investible, as opposed to MPI Hedge Fund Trackers, which replicate portfolios of top hedge funds with liquid ETFs.

Now that we know our model is doing a good job with the endowment’s performance – without prior knowledge of next-period returns – we find ourselves back to the same question: why is there a discrepancy between our original estimate and the posted allocations?

The answer may simply be this: our estimates are a victim of Bowdoin’s own success. Let’s look at the contribution by asset class to our tracking portfolio’s performance, both for FY 2021 and for the last several fiscal years:

Public equity, private equity, hedge funds – their strategies and mechanics may differ, but this is still fundamentally one asset class (equities), and correlations can be high. Working with annual data points, and the endowment’s stellar performance, it seems reasonable that a model will over-estimate allocations to the best performing strategies having higher beta to the respective market benchmark. Thus, if VC allocations of Bowdoin (such as in Sequoia Capital’s fund, mentioned earlier) are more aggressive than the VC index, and US equities are less aggressive than the S&P500 Index, their respective exposures will differ from reported allocations.

Our clients face similar questions every day when analyzing mutual funds or alternatives. The goal of Sharpe’s RBSA approach utilized in our dynamic DSA is not simply to confirm reported allocations. (If you want to do that, and create a reasonable tracking model, our approach above is already doing the job.) It’s the potential discrepancies between reported and effective (in Sharpe’s RBSA vernacular circa 1992) allocations that is the focus of such an analysis pointing into potential risks and cutting into the managers’ “alpha.”

To better understand the fund’s behavior – and, as with the Ivies, how much risk is being taken on – we’ll look forward to the data points to come in the years ahead.

We don’t imagine we’ll be the only one watching to see what shifts, if any, Mr. Bryant brings to the fund’s allocations.

Factors used in DSA Analysis:

DISCLAIMER: MPI conducts performance-based analyses and, beyond any public information, does not claim to know or insinuate what the actual strategy, positions or holdings of the funds discussed are, nor are we commenting on the quality or merits of the strategies. This analysis is purely returns-based and does not reflect actual holdings. Deviations between our analysis and the actual holdings and/or management decisions made by funds are expected and inherent in any quantitative analysis. MPI makes no warranties or guarantees as to the accuracy of this statistical analysis, nor does it take any responsibility for investment decisions made by any parties based on this analysis.

Exit mobile version