Value-at-Risk Forecasting Using Nonlinear Regression Quantile During the 2008-09 Financial Crisis

• Main Authors: Bo-Kuan Hwang, 黃博寬
• Other Authors: Cathy W. S. Chen
• Format: Others
• Language: en_US
• Published: 2010
• Online Access: http://ndltd.ncl.edu.tw/handle/10485523538432721189

碩士 === 逢甲大學 === 統計與精算所 === 98 === Value-at-Risk (VaR) is a commonly used risk measure of the risk of loss on a specific portfolio of financial assets. VaR becomes important especially during the 2008-09 financial crisis. A well known semi-parametric model, the conditional autoregressive VaR (CAViaR) model, moves the focus of attention from the distribution of returns to the quantiles. We propose nonlinear threshold CAViaR models that incorporate intra-day volatility range. Model stimation and inference are performed using the Bayesian approach via Skewed-Laplace distribution. We examine how risk models performed during the 2008-09 global financial crisis, evaluate how the financial crisis affect the performance of risk models via forecasting VaR. An empirical study is conducted on five Asia-Pacific Economic Cooperation (APEC) stock market indices, including the Standard and Poor’s 500 Index, Nikkei 225, TAIEX, HSI and KOSPI, to forecast VaR from August 2008 to April 2010. We further provide the violation rates, backtesting criteria, market risk charges and the Dynamic Quantile test to measure the forecasting performance of the CAViaR model and a variety of alternative risk models. The performance of the proposed threshold CAViaR models incorporating the range information are shown to forecast VaR more efficiently than using other models.