中国股市投资者情绪指数构建与有效性检验

The Construction and Effectiveness Test of Investor Sentiment Index for China's Stock Market

本文使用了三种不同的数据降维方法——主成分、卡尔曼滤波和拉普拉斯特征映射,选取了城镇居民投资意愿、IPO首日收益率、对数开户比和流通市值加权换手率,建立了三3个从1999年至2015年跨度17年的测度中国股票市场投资者情绪的月度复合指数。同时,基于"孪生股票"现象,使用AH股溢价指数对三个情绪指数进行了有效性检验。结果发现,基于卡尔曼滤波方法的情绪指数最为有效,基于主成分方法的情绪指数次之,而基于拉普拉斯特征映射的情绪指数没有达到实证研究的有效性要求。

This paper uses three composite indexes to measure the investor sentiment of the China stock market during the last seventeen years from 1999 to 2015 through selecting the Chinese households＇ investment preferences, IPOR, logarithm of rate of stock account amount and weighted turnover rate by applying three different methods, namely, principal component analyses, Kalman filtering and Laplacian Eigenmaps. At the same time, the authors use A-H stock premium index to test the effectiveness of three sentiment indexes based on the premium phenomenon of ＂twin stocks＂. The results show that the sentiment index based on Kalman filtering is the most effective, followed by the principal component analysis. However, the sentiment index based on Laplacian Eigenmaps hasn＇t reached the ef- fective requirement of the empirical study.