Strategic portfolio allocation decisions are the main determinants of investment risk and returns for institutional investors. But what is the optimal method for constructing and evaluating strategic allocations, incorporating risk metrics that span beyond standard deviation? The importance of evaluating multiple risk metrics cannot be overstated, especially when considering that many multi-asset portfolios and the asset classes comprising them exhibit non-normal return distributions, negative skewness and considerable drawdown risk beyond what can be summarized with the conventional metrics of return and standard deviation. To consider these issues, we evaluate a range of available portfolio construction options, considering different risk metrics. To better assess potential return paths for different allocations we also conduct forward-looking simulations that incorporate various investment regimes that are historically more representative of negatively skewed asset classes contained in multi-asset portfolios.
The shortcomings of portfolio construction methods that assume all asset returns are normally distributed have been widely documented. Multiple studies have identified significant negative skewness and excess kurtosis in equity returns. Studies have further found that investors are averse to negative skewness and will seek hedging assets in longer-horizon portfolios on the expectation of occasional periods of adverse investment climates.
The assumption of elliptical multivariate normal distributions for asset class returns contradicts their behavior. Because these methods often place no weighting on asymmetric return distribution or serial correlations, they are not an ideal guide for choosing among portfolios. While several models that evaluate single-period negative outcomes demonstrate the probability of experiencing losses, some fail to determine the magnitude of losses, making them flawed choices for optimizing risk metrics.
We consider several risk metrics used in portfolio construction, which broadly fall into two categories: those that measure individual period risk (such as portfolio variance) and those that measure cumulative risk (such as maximum drawdown). Since portfolio return processes with high variance tend to have higher drawdowns and vice versa, choosing the appropriate metric is difficult and requires a deeper analysis of the asset return process.
We therefore examine a range of options for building strategic multi-asset portfolio allocations, considering multi-dimensional risk parameters. We illustrate how portfolio construction incorporates managing risk across three dimensions:
- What are the outcomes under Mean-Variance Optimization (MVO) and alternative risk models that consider equal risk contribution, historical drawdowns, historical shortfall to target and portfolio higher moments?
- How ex-ante constraints on asset classes and asset class groups impact those outcomes?
- How increasing the opportunity set with a defensive equity allocation impacts outcomes?
Given the higher volatility and skewness present in equities, we evaluate the impact on strategic portfolio outcomes of including a defensive equity allocation to help to manage higher dimensional risk. We examine PGIM Quantitative Solutions’ US Market Participation Strategy (MPS), which allows upside participation when the US equity market advances, while reducing downside risk. The strategy utilizes long-dated S&P 500 call options in combination with US Treasury bonds. The call options provide upside participation, while US Treasury bonds serve as a safe haven during turbulent market conditions and provide downside protection. Using a disciplined process, exposures (market, volatility and duration) are actively managed in response to the changing market environment using a rules-based framework.
We show that adding MPS to the asset universe meaningfully improves portfolio outcomes for all evaluated risk models. The resultant portfolios have lower volatility, reduced drawdowns and a substantial increase in skewness to positive and significant levels in some cases. Sharpe ratios of all risk model portfolios with added defensive equity allocation are improved with no reduction in annualized returns.
Finally, to better evaluate the potential future path of strategic portfolio outcomes in the presence of negative skewness, we evaluate a regime-aware simulation methodology that seeks to replicate the historical characteristics of asset classes and portfolio outcomes. This analysis sheds additional light onto the important task of building long-horizon strategic portfolios to meet long-term funding requirements that incorporate risk beyond the conventional metrics of mean and variance.