Publication detail

A semiparametric nonlinear quantile regression model for financial returns

Author(s): Mgr. Krenar Avdulaj Ph.D.,
doc. PhDr. Jozef Baruník Ph.D.,
Type: Articles in journals with impact factor
Year: 2017
Number: 0
ISSN / ISBN:
Published in: Studies in Nonlinear Dynamics & Econometrics, 21(1), pp. 81–97
Publishing place:
Keywords:
JEL codes:
Suggested Citation:
Abstract: Financial institutions use Value at risk (VaR) as the standard measure of market risk. Despite its simplicity, measuring and forecasting it accurately is a challenging task. Recently, quantile regression models have been used successfully to capture the conditional quantiles of returns. We explore further non-linearities in the data, and propose to use realized measures in the nonlinear quantile regression framework to explain and forecast conditional quantiles of financial returns. In addition, we apply the proposed model to a pool of the most liquid U.S. assets across different industries. The nonlinear quantile regression models are implied by copula specifications and allow us to capture possible nonlinearities, and asymmetries in conditional quantiles of financial returns. Using high frequency data covering most liquid U.S. stocks in seven sectors, we provide ample evidence of asymmetric conditional dependence and different level of dependence characteristic for each industry. The backtesting results of estimated Value-at-Risk favour our approach.

Partners

ČSOB
Deloitte
McKinsey & Company

Sponsors

CRIF