高频金融数据在风险价值VaR度量和预测方面的价值已经引起了学术界和业界的广泛兴趣。计算和预测VaR的方法广义上可以分为两大类:间接法和直接法,在有了高频数据后,两类方法均可行,尤其是由于高频数据导出的"已实现"波动率的出现,使得间接法有明显改进。本文将从间接法中选取基于"已实现"波动率的ARFI模型与从直接法中选取的两个CAViaR模型进行比较,采用沪深300、上证指数、深证成指的5分钟高频数据,根据多种在评价VaR预测模型表现时广泛使用的后验测试,对各模型进行实证检验,结果表明基于CAViaR模型的预测表现优于基于"已实现"波动率的ARFI模型,这对风险管理从业者有一定的参考意义。
The models for calculating and forecasting VaR can be classified into two broad categories: indirect-VaR and direct-VaR approaches.The VaR forecast performances between models taken from these two approaches respectively are compared. One is ARFI-VaR forecast model the indirect-VaR approach based on "realized volatility" obtained by high frequency data. Another two are CAViaR-based models, which are the representatives of direct-VaR approach. By the various backtests that are extensively used for VaR performance evaluation, using 5-min high frequency data of CSI 300 Index, SSE Composite Index and SZSE Component Index, the empirical evidence shows the CAViaR-based models perform better than realized volatility-based ARFI model.
[1] Artzner P, Delbaen F, Eber J, et al. Coherence measures of risk [J]. Mathematical Finance, 1999, 9(3): 203-228.
[2] 林宇, 卫贵武, 魏宇, 等. 基于Skew-t-FIAPARCH的金融市场动态风险VaR测度研究[J]. 中国管理科学, 2009, 17(6): 17-24.
[3] 林宇. 基于双曲线记忆HYGARCH模型的动态风险VaR测度能力研究[J]. 中国管理科学,2011, 19(6): 15-24.
[4] Engle R, Manganelli S. CAViaR: Conditional autoregressive value at risk by regression quantiles [J]. Journal of Business and Economic Statistics, 2004, 22(4): 367-381.
[5] Huang Dashan, Yu Baimin, Lu Zudi, et al. Index-Exciting CAViaR: A new empirical time-varying risk model [J]. Studies in Nonlinear Dynamics & Econometrics, 2010, 14(2): 1-24.
[6] 刘新华, 黄大山. 中国股市风险CAViaR方法的稳定性分析及其时变建模[J]. 系统工程理论与实践, 2005, 25(3): 1-6.
[7] Andersen T, Bollerslev T, Diebold F, et al. The distribution of realized stock return volatility [J]. Journal of Financial Economics, 2001, 61(1): 43-76.
[8] Andersen T, Bollerslev T, Diebold F, et al. The distribution of realized exchange rate volatility [J]. Journal of the American Statistical Association, 2001, 96 (453): 42-55.
[9] Andersen T, Bollerslev T, Diebold F, et al. Modeling and forecasting realized volatility [J]. Econometrica, 2003, 71(2): 579-625.
[10] Barndor-Nielsen O, Shephard N. Non-Gaussian Ornstein-Uhlenbeck based models and some of their uses in financial economics [J]. Journal of the Royal Statistical Society, 2001, 63(2)67-241.
[11] Barndor-Nielsen O, Shephard N. Econometric analysis of realized volatility and its use in estimating stochastic volatility models [J]. Journal of the Royal Statistical Society, 2002, 64(1): 253-280.
[12] 徐正国, 张世英. 调整"已实现"波动率与GARCH 及SV 模型对波动的预测能力的比较研究[J]. 系统工程, 2004, 22(8): 60-63.
[13] 魏宇, 余怒涛. 中国股票市场的波动率预测模型及其SPA检验[J]. 金融研究, 2007, (7): 138-150.
[14] 房振明, 王春峰, 付臣余. 基于高频数据的股指风险价值预测[J]. 统计与决策, 2007, (18): 89-91.
[15] 邵锡栋, 殷炼乾. 基于实现极差和实现波动率的中国金融市场风险测度研究[J]. 金融研究, 2008,(6): 109-121.
[16] 陈功, 程希骏, 马利军. 基于CAViaR的DCC模型及其对中国股市的实证研究[J]. 数学的实践与认识, 2009, 39(4): 75-81.
[17] 王新宇, 宋学峰, 吴瑞明. 基于AAVS-CAViaR模型的股市风险测量研究[J]. 系统工程学报, 2010,25(3): 327-333.
[18] Huang Dashan, Yu Baimin, Fabozzi F, et al. CAViaR-based forecast for oil price risk [J]. Energy Economics, 2009, 31(4): 511-518.
[19] Kupiec P. Techniques for verifying the accuracy of risk measurement models[J]. Journal of Derivatives, 1995, 3(2): 73-84.
[20] Christoersen P. Evaluating interval forecasts [J]. International Economic Review, 1998, 39(4): 841-862.
[21] Geweke J, Porter-Hudak S. The estimation and application of long memory time series models[J]. Journal of Time Series Analysis, 1983, 4(4): 221-238.
[22] Koenker R, Bassett G. Regression quantiles[J]. Econometrica, 1978, 46(1): 33-50.
[23] Bao Yong, Lee T, Salto glu B. Evaluating the predictive performance of value-at-risk models in emerging markets: A reality check [J]. Journal of Forecasting, 2006, 25(2): 101-128.
