Inference for Joint Quantile and Expected Shortfall Regression

Abstract: Quantiles and expected shortfalls are commonly used risk measures in financial risk management. The two measurements are correlated while have distinguished features. In this project, our primary goal is to develop stable and practical inference method for conditional expected shortfall. To facilitate the statistical inference procedure, we consider the joint modeling of conditional quantile and expected shortfall. While the regression coefficients can be estimated jointly by minimizing a class of strictly consistent joint loss functions, the computation is challenging especially when the dimension of parameters is large since the loss functions are neither differentiable nor convex. To reduce the computational effort, we propose a two-step estimation procedure by first estimating the quantile regression parameters with standard quantile regression. We show that the two-step estimator has the same asymptotic properties as the joint estimator, but the former is numerically more efficient. We further develop a score-type inference method for hypothesis testing and confidence interval construction. Compared to the Wald-type method, the score method is robust against heterogeneity and is superior in finite samples, especially for cases with a large number of confounding factors. We demonstrate the advantages of the proposed methods over existing approaches through numerical studies.