矫偏条件波动及其在VaR估计中的应用——以一般正态VaR估计为例

Bias-corrected Conditional Volatility and Its Application in Estimating VaR: The Example of Normal-VaR

反映金融市场条件波动时变性的模型都只是对资产真实数据生成过程的一种近似拟合,许多模型严格来说存在设定偏误,这样可能导致由它们所预测的条件波动相对真实值存在偏差。本文用一回归模型检验一般正态VaR估计中的简单平均方法估计的条件波动是否存在偏差,另一方面用它对这一条件波动估计值进行矫正,并以此为基础估计相应的VaR。通过对上证指数和深成指数1995-2005年日收益率数据的分析,我们发现,用简单平均方法对我国股市波动进行估计存在非常大的条件偏差,并发现波动矫偏后所预测的VaR值的准确性和有效性都明显提高。

Models to capture time-varying conditional volatility in financial markets are only approximations of the true data-generating process for financial asset returns, and many are seriously misspecified. A mis-specified conditional volatility model will generate biased forecasts of the true conditional volatility. Using a regression model this paper, on one hand, tests whether the conditional volatility the simple average method, estimates in normal-VaR are conditionally biased, on the other hand, and corrects the conditional volatility forecasts, based on which VaR are estimated. By analyzing the daily return of Chinese stock market in 1995-2005, we find that the simple average method systematically generates forecasts of Chinese stock market＇s conditional volatility that are conditionally biased, and that applying bias-corrected volatility forecasts to estimate one-day VaR substantial improves both performance and precision of the forecasts.