Evaluating predictive performance of value-at-risk models in Chinese stock markets

  • Jianshe OU

Student thesis: MPhil Thesis (Lingnan)


Risk can be defined as the volatility of unexpected outcomes, generally for values of assets and liabilities. Financial risk, risk refer to possible losses in financial markets, includes markets risk, credit risk, liquidity risk, operational risk, and legal risk. This MPhil thesis is specializing on market risk, which involves the uncertainty of earnings or losses resulting from changes in market conditions such as asset prices, interest rates, and market liquidity.

The primary tool to evaluate market risk is VaR that is a method of assessing risk through standard statistical techniques. VaR is defined a measure for the worst expected loss over a given time interval under normal market conditions at a given confidence level. The greatest benefit of VaR for an asset manager lies in the imposition of a structured methodology for critically thinking about risk. Institutions applying VaR are forced to confront their exposure to market risk.

There are three methods to calculate VaR, parametric, nonparametric and semi-parametric. Parametric method includes The Equally Weighted Moving Average (EqWMA), The Exponentially Weighted Moving Average (EWMA), GARCH, Monte Carlo Simulation (MCS) approaches. Parametric method includes The Historical Simulation approach (HS), and semi-parametric method includes filtered historical simulation (FHS), extreme value theory (EVT) approaches.

At present stage, Chinese asset managers apply RiskMetrics approach, i.e. EWMA, proposed by J.P. Morgan to calculate VaR. But this approach assumes that error term is conditionally normally distributed. However, there has been criticism that the VaR is based on assumptions that do not hold in times when the financial markets are experiencing stress, and that the normal distribution does not make a good job in predicting the distribution of outcomes. Financial returns experience fat tails, skewness and kurtosis, which implies that the normal distribution works well in predicting frequent outcomes but is not a good estimator to predict extreme events. In addition, when applying EWMA approach, Chinese asset managers often use the decay factor proposed by J.P. Morgan instead of obtaining it on the basis of China’s financial markets’ data.

The purpose of this MPhil thesis is to compare the applicability of different parametric VaR methods for Chinese equity portfolios. We will also analyze whether equity market cap has any impact on the VaR methods. To assess whether VaR can be considered as a reliable and stable risk measurement tool for Chinese equity portfolios, we have performed an empirical study. The study covers four VaR approaches at the 95% and 99% confidence levels. Moreover, in order to describe skewness and kurtosis, we propose EWMA approach with a mixture of normal distributions. Based on these results we discuss the implications of VaR for asset managers.

Our conclusion is that GARCH-normal is superior to Riskmetrics approach at both 95% and 99% confidence levels. The LOG-MLE (maximum Likelihood Estimation) can be improved when GARCH-t approach is used to replace GARCH-normal. However, GARCH-t is more conservative than GARCH-normal at 95% confidence level. At the same time, EWMA with mixed normal distributions is superior to RiskMetrics approach at 99% confidence level, but it is too conservative at 95% confidence level. For EWMA with mixed normal distributions and GARCH-type models, the former is better at 99% confidence level and the latter perform better at 95% confidence level. Due to this fact we recommend EWMA with mixed normal distributions and GARCH-t at 99% confidence level. The performance of GARCH-normal and EWMA is fairly good at 95% confidence level.
Date of Award2007
Original languageEnglish
Awarding Institution
  • Department of Finance and Insurance
SupervisorZiyou YU (Supervisor)

Cite this