基于ICA模型的投资组合稳健VaR方法研究

Portfolios Robust VaR Procedure Based on Independent Component Analysis

针对参数VaR方法在测度庞大复杂的投资组合风险时,存在模型参数多估计困难和需要设定风险因子联合分布容易使风险计量产生较大偏差等问题,本文提出了基于独立成分分析(ICA)技术的半参数IC-SP-VaR模型,给出了模型的参数估计方法,并对模型进行了模拟研究和实证分析。模拟研究表明新方法在不同情形下的估计都是有效的,在非线性经济序列中优势尤其明显。实证分析验证了新方法能够提高资产组合风险计量模型的稳定性和准确性。

There are two problems when using parametric VaR procedure to measure the portfolios risk:one is the model includes too many parameters to be estimated,the other is the model specification error which affects the measurement precision when fitting the distribution of risk factors.According to these problems,the paper proposes a semi-parametric IC-SP-VaR model based on independent component analysis(ICA).We develop parameters estimation method and conduct Monte Carlo simulation and empirical study for the model.Simulation study shows our procedure is useful in different type of data,especially in the non-linear time series.An empirical study tests the new method can improve stability and accuracy of portfolios risk measure model.