Robust Risk Measures in Real Time Settings

最近的金融危机使人们对学术界提出的、金融机构在实践中使用的单方面风险衡量指标的准确性和适当性产生了严重怀疑。风险价值和预期缺口等流行的衡量标准在最需要它们的市场情况下失败了。本文提出的研究项目对风险模型的模型风险研究具有一定的参考价值。它的目的是找出这些故障的来源，并加强实时环境下风险衡量的稳健性。推广标准风险衡量的两个战略将是：(1)开发不同维度(模型选择、市场环境、抽样窗口等)的实时计量方法。为了提高预测的准确性和稳健性，W.r.t.(2)开发实时多变量框架，以考虑风险相互依存的多方面性质及其驱动力，如市场流动性。新方法的优越性将在理论上以及在对广泛的证券、市场和市场状况的样本外预测性能方面进行计算。对经验绩效的评估将基于标准的回溯测试程序、统计拟合优度标准以及经济或财务标准。

The recent financial crisis has cast serious doubts on the accuracy and appropriateness of measures of onesided risks proposed in academia and used by financial institutions in practice. Popular measures like the Value-at-Risk and the Expected Shortfall have failed in market situations when they were needed most urgently. The research project proposed contributes to the research on the model risk of risk models. It aims at identifying the sources of these failures and at enhancing the robustness of risk measurement in real time settings. Two strategies of generalizing standard risk measures will be pursued by (i) developing real time econometric approaches for different dimensions (model choice, market setting, sampling window, etc.) to improve prediction accuracy and robustness w.r.t. changing market conditions, and by (ii) developing real time multivariate frameworks to account for multifaceted nature of risk interdependencies and their driving forces such as market liquidity. The superiority of the new approaches will be worked out theoretically as well as in terms of out-of-sample prediction performance for a wide range of securities, markets and market conditions. Evaluations of the empirical performance will be based on standard backtesting procedures, statistical goodness of fit criteria as well as economic or financial criteria.