An important segment of the Smart Beta landscape is that of Risk Minimization, most often represented by Minimum Variance/Minimum Volatility offerings. These strategies are based on a classic tool of asset management and aim to provide a risk mitigation solution. Despite being the most widely used, this framework has certain significant weaknesses. Using the portfolio variance as a risk measure is consistent with the assumption that the returns are normally distributed. Empirically, this is seldom the case with many stocks exhibiting asymmetry and fat-tailness in their return distributions. Consequently, a Minimum Variance portfolio may not offer appropriate protection during periods of market stress when fat-tailed returns are typical. Another undesirable characteristic of the mean-variance framework is that due to the symmetric nature of variance, both the potential upside and downside of the portfolio are penalized.
In this paper we address these drawbacks by proposing a methodology for constructing a Minimum Tail Risk portfolio. We introduce Expected Tail Loss (ETL) as a measure of risk and compare it to other measures used as alternatives to variance. Then, we outline several well-known characteristics of stock returns that cannot be modeled by Gaussian distributions and propose a more capable model of stock returns. Next, we complete the portfolio construction framework of the FactSet Minimum Tail Risk portfolios by describing the minimum ETL optimization problem, before presenting the results based on selection universes including U.S. and Japanese stocks.
Download the Minimum Tail Risk Portfolio Construction white paper to review our methodology.