Long-Run Investment Horizons and Implications for Mixed-Asset Portfolio Allocations

Abstract

When different asset classes display varying degrees of serial correlation, the investment horizon may substantially alter optimized mixed-asset portfolio allocations. Private-market assets (such as commercial real estate and private equity) often display much higher levels of autocorrelation than their public-market counterparts. Consequently, the one-year returns typically used in mixed-asset portfolio optimization procedures often generate excessive allocations to private-market asset classes. To counteract these excessive loadings, many researchers and practitioners advocate either ad hoc rules designed to constrain the maximum allocation to private-market vehicles and/or the use of some de-smoothing procedure to inject additional volatility in the observed return series. This paper takes a different approach by examining the three components in the standard portfolio optimization technique as the investment horizon lengthens; the auto-correlated nature of the observed returns of private-market assets implies that (annualized) long-horizon volatility decays less slowly – as compared to those public-market asset classes exhibiting the random walk – and long-horizon correlation (with most public-market alternatives) increases. In re-running the mixed-asset portfolio optimization with long-horizon returns, the allocations to commercial real estate – as one example of private-market assets – are much reduced and more consistent with allocation levels generally found in large institutional pension plans.