CFE 2019: Start Registration
View Submission - CFE
A1586
Title: A resampling method for dynamic quantile models of asset returns Authors:  Richard Luger - Laval University (Canada) [presenting]
Abstract: Suppose the joint distribution of daily returns is symmetric and a consistent point estimator is available for the parameters of a dynamic quantile model of the asset's multi-day returns. The considered class of dynamic quantile models includes linear and non-linear autoregressive specifications. In this setting, a simple and general resampling method is proposed to obtain the distribution of parameter estimates, which may be constrained to avoid the crossing problem when several quantile levels are fitted. With large sample sizes, the resampling distribution allows the construction of simultaneous confidence intervals for continuous functions of the model parameters. The usefulness of this non-parametric inference procedure is illustrated by means of a simulation study and with an empirical application featuring a conditional autoregressive value-at-risk (CAViaR) model for daily returns and a quantile autoregression (QAR) model for longer horizons.