摘要
从大量的金融资产中提取出的系统风险比基于β系数的单变量方法更为有效,但资产规模的增加会导致"纬数灾难"等问题,难以获得准确估计。本文在将金融资产收益分为公共系统因素和个体特质因素基础上,提出用具有条件异方差形式的动态潜在因子模型(CHDL)估计和预测动态系统因素,用非参数核密度估计系统下跌时的边际期望损失(MES)。本文利用上海证券市场180只样本股进行实证分析,通过IC和Onat检验发现个股和各板块存在显著的系统因子;利用CHDL模型对个股和各板块的系统因子和资产未来收益进行估计和预测,在此基础上计算边际期望损失。Mincer-Zarnowitz回归最优检验法表明,CHDL模型计算的系统风险比常用的市场指数模型具有更高的准确性。
The recent global financial crisis emphasized the importance ot the system risk. Based on the systematic and idiosyncratic factors of assets returns, we propose a conditional heteroskedasticity dynamic latent factor model to measure and forecast the dynamic common {actor. Using the CHDL model, the IVIES can be measured with nonparametric kernel estimation. The empirical application on a set of SSE 180 index stocks finds that the tests and criteria point towards one dynamic common factor driving the co-movements in single stock and sectors. Using the dy- namic common {actor, MES can be estimates. The Mincer-Zarnowitz regression test shows CHDL provides better systematic risk forecast than the methods based on market returns both in single stock and sectors.
出处
《数量经济技术经济研究》
CSSCI
北大核心
2012年第11期102-115,共14页
Journal of Quantitative & Technological Economics
基金
国家社会科学基金项目"中国股市高频数据的金融风险度量与管理研究"(10XTJ0001)资助
