A1493
Title: Semiparametric ultra-high dimensional model averaging of expected shortfall for nonlinear dynamic time series
Authors: Jiangtao Wang - Huazhong Normal University (China) [presenting]
Abstract: A semiparametric model averaging scheme is proposed for expected shortfall (ES) in a nonlinear dynamic time series setting with a very large number of covariates, including exogenous regressors and autoregressive lags. The objective is to obtain more accurate estimates and forecasts of ES via using a large number of conditioning variables in a nonparametric way. The proposed scheme consists of two steps. The first step is to forecast ES in a nonparametric manner based on a few variables. And the second step is to obtain the final forecast of ES by a combination regression based on the obtained prediction of ES derived from the first step. Asymptotic properties for the proposed scheme are investigated under some regularity conditions. Numerical studies including both simulation and an empirical application are given to illustrate the proposed methodology.
