Quantile Regression Based Risk Measurement in the Chinese Stock Markets
Zhao Shao-juan · Journal of China University of Mining and Technology · 2008
Financial time series are modeled using quartile regression theory.These models are compared to a least squares regression model in terms of risk measurement.Three conditional quantile regression models were used on dynamically forecast Values at Risk(VaR).The three models used lagged daily returns,weekly virtual variables and lagged means and deviation of return as the predictor variables.The data were taken from Chinese stock market records over the 1996 to 2004 period.In the empirical research the week effect on VaR in the Chinese stock market is investigated.The results show that daily market risk presents typical non-uniformity in a week,and that the quantile regression model is superior to the APARCH model during back-testing.The quantile regression approach estimates VaR for heavy-tailed financial time series at a high significance level.Quantile regression also proved to be an effective semi-parametric approach for measuring market risk and for exploring complex patterns.