A century of asset allocation crash risk

Abstract

We extend proxies of several popular asset allocation approaches—U.S. and Global 60/40, Diversified Multi-Asset, Risk Parity, Endowment, Factor-Based, and Dynamic asset allocation—using long-run return data for a variety of sub-asset classes and factors to test their long-term performance. We use equity and debt assets, commodities, alternatives, and indices to reconstruct the returns on allocation portfolios from 1926 to the present, the entire period for which comprehensive asset pricing data are available. We contribute to the existing literature by developing a laboratory for testing the performance of popular asset allocation strategies in a wide range of scenarios. We also aim to test the importance of the behavioral aspect of investment decisions for portfolio outcomes. In our framework, Factor-Based portfolios exhibit the best traditionally measured risk-adjusted returns over the long run. However, Dynamic asset allocation is most likely to reduce the risk of abandonment of the strategy by an investor and selling the portfolio in panic when they experience losses over their tolerance threshold, because the dynamic strategy exhibits lower expected drawdowns, even during severe market downturns. Across all strategies, risk-tolerant investors who rely on a longer history to set their expectations, whether based upon actual or extrapolated data, experience significantly better outcomes, particularly if their investment horizon includes times of crisis. This study informs portfolio managers, investment analysts, and advisors, as well as investors themselves, of the impact of information, persistence, and properties of various portfolio allocation methods on investment returns.