Two Essays on Endowment Portfolio Design

This dissertation examines the design of long-term investment policies for university endowments, with particular emphasis on strategic asset allocation and crisis risk management in portfolios heavily exposed to alternative assets. University endowments differ from conventional investors in two important respects: they allocate substantial capital to illiquid assets such as private equity, venture capital, hedge funds, and real assets, and they operate under spending rules that link annual distributions to reported net asset values (NAVs).
These features raise two central questions. First, how should endowments design strategic asset allocations when return dynamics depart from normality and exhibit time dependence?
Second, how can such portfolios be protected against severe drawdowns, especially when reported NAVs incorporate appraisal-based smoothing and delayed loss recognition?

The first essay proposes a simulation-based approach to strategic asset allocation for university endowments. The method applies a stationary block bootstrap to asset-class returns, preserving time dependence and non-normal distributional features that are especially relevant for portfolios with substantial allocations to illiquid alternatives. Using data across public equities, corporate bonds, hedge funds, private equity, venture capital, real assets, and cash, the analysis shows that the approach leads to meaningfully different portfolio selection relative to multivariate normal simulations commonly used in practitioner work when portfolios are evaluated using endowment-specific risk metrics. In the baseline specification, optimal allocations shift several percentage points out of cash and fixed income and into private equity and venture capital, corresponding to a reallocation of roughly 4–5% of total portfolio value for a representative endowment. Evidence suggests that these differences stem from the ability of block bootstrapping to capture mean reversion in historical returns, a property missed by normality-based assumptions. To highlight this mechanism, the analysis adopts a static policy allocation with annual rebalancing, positioning the contribution as complementary to dynamic models that incorporate liquidity constraints and rebalancing frictions.

The second essay studies crisis hedging for alternative-heavy endowments. Because private asset valuations are appraisal-based and reported quarterly, drawdowns in endowment portfolios emerge with a delay. Risk management must therefore be evaluated on the same reported series that govern payout decisions and institutional responses. A quarterly Synthetic Endowment Index (SEI) is constructed by applying average 1999–2024 endowment policy weights to public- and private-market return series while preserving each sleeve’s reporting cadence. Across four SEI drawdowns exceeding 10%, eleven liquid hedges are compared. Time-series momentum (TSM) delivers crisis payoffs comparable to those obtained when hedging a fully mark-to-market equity benchmark, and in some episodes provides greater drawdown protection. In a simple 90/10 overlay (SEI plus a diversified TSM basket, rebalanced quarterly), peak-to-trough losses fall by roughly 2–5 percentage points in the Dot-com, GFC, and COVID episodes while long-run returns remain essentially unchanged. Under a standard 5% payout rule, the maximum spending shortfall shrinks by about four percentage points. Relative to Treasuries, gold, tail-risk strategies, and global macro, diversified trend-following provides comparatively uniform protection across the historical crises studied and can be implemented as a liquid overlay. 
 

Taken together, the two essays show that both the modeling of return dynamics for strategic allocation and the evaluation of crisis hedges must account explicitly for the time dependence, distributional features, and reporting structure of alternative-heavy endowment portfolios.