摘要
分位数回归方法因为考虑了分布函数的各局部信息而比只考虑条件期望的普通最小二乘回归方法更具有优势,特别是在具有厚尾分布的金融数据分析方面,提供了更详尽的信息。本文通过分位数回归方法重新审视中国股市截面收益率的共同风险因子,查看是否存在规模效应与帐面市值比效应。结果发现,分位数回归结果与普通最小二乘结果显著不同,不同分位数下回归系数及其统计显著性都存在巨大差异。股票收益率与规模正相关的规模效应显著,且高收益率部分的正规模效应更加强烈。帐面市值比效应在低收益率部分正相关,高收益率阶段负相关,中间部分不显著。
OLS (Ordinary Least Square) regression merely reports the conditional expectation. However, QR (Quantile Regression) makes use of local information of the entire distribution, thus is superior in most applications, especially for fattailed financial data. In this paper, by utilizing beth OLS and QR, we investigate systematic risk factors, such as market betas, size and book-to market ratio, for cross-sectional returns in China's stock market. As expected, the QR method tells more stories. There is a positive size effect, which becomes stronger for higher quantiles. With respect to book-to-market effect, we find positive effect in lower quantile, yet negative effect in higher quantiles.
出处
《南方经济》
北大核心
2006年第1期61-71,共11页
South China Journal of Economics