基于技术分析指标的投资者情绪指数有效性研究

Validation of Investor Sentiment Index Based on Technical Analysis Indicators

投资者情绪对资产定价有广泛而持久的影响,综合情绪指数是当前使用最为广泛的情绪度量方式,但其偏宏观和总体,反映市场所有投资者的综合情绪,很难在行业和地域等层面,特别是个股层面进一步细分,且受部分代理变量可观测频率的限制,综合情绪指数多为年度或月度指数,进一步提高该指数的频率比较困难。为克服综合情绪指数的不足,基于行为金融和股票分析技术,以人类行为模式重复可预测为假设的思想,使用股票技术分析指标作为代理变量,构建能够同时适用于市场和个股、能以较高频率度量投资者情绪的情绪指数。基于股票价格和成交量变化反映投资者决策过程、投资者决策过程直接反映投资者情绪这一逻辑,沿用综合情绪指数的构建方法,使用换手率、相对强弱指标、乖离率、人气指标和成交量比率5个技术指标,通过主成分分析法构建技术情绪指数,并以1999年至2015年为样本期,通过多元回归对技术情绪指数在市场和个股层面的有效性进行验证。研究结果表明,市场层面的技术情绪指数与综合情绪指数存在显著相关性,说明二者存在显著的相关性和近似性,间接验证了技术情绪指数在市场层面的有效性。市场和(或)个股技术情绪指数差异对交叉上市公司的股票价差现象有较好的解释能力,这一解释能力满足有效的情绪指数应具备的特征要求。在横截面上和时间序列上个股技术情绪指数均与个股收益率显著负相关,买入个股技术情绪指数值最高组、卖出最低组的零投资组合策略能获得显著的超额收益。同时,流通市值越高、上市时间越短、账面市值比越低、盈利能力越强、营业收入增长率越高的公司对投资者情绪的敏感程度越高,验证了技术情绪指数度量个股情绪的有效性。研究结果验证了使用技术分析指标度量投资者情绪的可行性,为将股票技术分析实践与投资者情绪学术研究关联在一起做出一定贡献;提供了一个在市场和个股层面都适用的投资者情绪度量方法,为将投资者情绪应用于个股层面资产定价和公司财务等领域研究提供了有益启发;技术情绪指数在时间上的可拓展性也为未来探索以更高频率度量投资者情绪提供了可参考思路。

Many studies show investor sentiment affects assets price market widely and consistently. BW index firstly introducedby Baker ＆ Wurgler （2006） is the most widely used investor sentiment proxy. However, there are disadvantages of BW index. First, BW index is constructed with market-wide indicators. It is not suitable to measure investor sentiment at industry or firm- specific level. Second, because of the unavailability of certain indicators, BW index can hardly be constructed at high frequency. For example, IPOs are not daily routines. There may be no IPO events for weeks or even months. Meanwhile, in emerging mar- kets like China, regulators may suspend IPOs as reactions to unfavorable market condition. Either way, sentiment indicators such as numbers of IPOs and IPO first-day returns will be missing for months thereby hindering the construction of BW index. Based on the conception that technical analysis and behavior finance share the same notion that human behavior is repetitive and predictable, we try to construct investor sentiment composite index with technical analysis indicators so as to overcome the disadvantages of BW index mentioned above. Changes of stock prices and volumes reflect investors＇ trading pattern, and inves- tors＇ trading pattern reflect investor sentiment directly. Therefore, technical analysis indicators based on prices and volumes are naturally proxies for investor sentiment. For this consideration, we construct Technical Investor Sentiment index （TS index in short） with 5 technical analysis indicators by principal component analysis, and verify its validity at both market-wide and firm- specific level by multiple regressions analysis with a sample of stocks from 1999 to 2015. First, we find that the correlation coefficient between annual （monthly） TS index and BW index is 0. 873 （0. 376） and is significant at 1%, indicating that there is significant correlation and similarity between them, which in turn illustrates the validity of TS index on measuring market-wide investor sentiment. Second, we find that differences of TS index explain price differences of cross-listed stocks, such as China A- and H-shares, and A- and B-shares. According to Baker ＆ Wurgler （2012） , this also proves the validity of TS index on measuring sentiment. Third, we find firm specific TS indices are negatively related with indi- vidual stock returns at both cross-sectional and time-series level. A hedge portfolio that longs stocks with lowest firm-specific TS indices and shorts stocks with highest indices gains significant excess returns. Meanwhile, firm-specific TS indices are signifi- candy related with firm characteristics. Firms with larger scale, shorter history, lower book/market ratio, higher profitability and higher revenue growth are more sensitive to investor sentiment. These findings are consistent with those of Baker ＆ Wurgler （2006） and Firth （2015） and thereby prove the validities of TS indices on measuring firm-specific investor sentiment. This study makes several contributions to the investor sentiment literature. First, we illustrate that technical analysis indica- tors are suitable to measure investor sentiment and therefore contribute to combine investment practice with academic investor sen- timent research. Second, we verify TS indices as valid proxies on measuring investor sentiment at both market-wide and firm-spe- cific level, thereby offering useful ways to study the effect of investor sentiment on asset prices or corporate finance decisions at firm-specific level. Third, since technical analysis indicators can be obtained at high frequency, the way we construct TS indices provides insights to measure investor sentiment in short terms such as weeks or days.