Comment on: Nonparametric Tail Risk, Stock Returns, and the Macroeconomy

We consider the non-parametric measure of tail risk (TR) proposed by Almeida, Ardison, Garcia and Vicente (2017) and illustrate the impact of some of the trade-offs involved when computing it both on real data and in controlled experiments with known dynamics of TR. We also propose an extremely simple to compute conditionally normal benchmark approximation of TR as a baseline to compare against other tail risk measures that aim to capture deviations from conditional normality. Given the empirical challenges associated with identifying and estimating tail risk, we strongly advocate evaluating tail risk measures in simple stylized models with well-understood properties and calibrated to real data as a useful device for understanding the behavior of alternative TR implementations as a function of specific features of the model.