本論文分別以報酬波動性和代表投機股特徵之財務比率兩種方式來判別個股的投機程度,並依此標準建構投資組合進行研究。除了採用多個分別代表不同投資族群之單一情緒指標外,並參考Brown and Cliff (2004)之作法,將所有情緒指標之第一主成份的因素負荷量建構一個「多變數複合情緒指標」,探討情緒對投機程度不同之股票當期報酬影響與未來報酬預測。最後,參考Baker and Wurgler (2006)所建構之「高-低特徵投資組合報酬差異之迴歸預測模型」,檢驗當投資人情緒高(低)時,具有投機特徵的股票是否會有較低(高)的未來報酬表現。實證結果顯示:第一、選用的情緒指標即使是代表同一族群投資人,其對股票當期和未來報酬的影響都沒有一致性的結果,無法比較不同族群投資人情緒對股票報酬的影響方向。第二、若以9個代表投資人情緒指標之第一主成份的因素負荷量來建構「多變數複合情緒指標」,則出現情緒所造成之當期偏離,下期修正的現象。第三、整體而言,投資人情緒對投機性愈大之股票之當期和未來報酬的影響愈大。第四、在三因子為控制變項下,未能提供顯著且一致性的結論證實當情緒高(低)時,具有投機特徵的股票會有較低(高)的未來報酬表現。第五、大部份的單一情緒指標和本論文所建構之多變數複合情緒指標對股票未來報酬的預測效果均不佳;若將情緒水準區分為高、低時期,低情緒相對於高情緒有較佳的預測效果。
In this paper, I cite foreign reference to determine stocks’ speculative level according to the volatility of stock returns and the financial ratios, then design portfolios on these two ways. I adopt several single sentiment indexes of three different investor groups and form a complex sentiment index by loadings of first principal component of all the single sentiment indexes to discuss the impact of investor sentiment on current and future returns of stocks. Finally, I use the model of ‘’Predictive Regression for Long–Short Portfolios’’ (Baker and Wurgler, 2006) to discuss when investor sentiment is high(low), would future returns be relatively low(high) for the stocks with speculative characteristics. The empirical results show that : (1) Different investor sentiment indexes have different impact on returns of stocks, even they are on behalf of the same investor group. Therefore, it is hard to compare the difference of the impact of investor sentiment of three groups on stock returns. (2) I form a complex sentiment index by loadings of first principal component of the 9 sentiment index, then I find that the deviation of stocks’ price caused by investor sentiment have been corrected on next month. (3) Overall, speculative stocks’ returns are most affected by investor sentiment. (4) If the Predictive Regression Model controlled with the variables of market risk premium(RMRF) and Fama-French factors (SMB and HML), there shows no strong evidence to prove that when investor sentiment is high(low), future returns would be relatively low(high) on stocks with speculative characteristics. (5) There is no strong significance of predictability on mostly single sentiment indexes and the complex index;If I divide sentiment level into high and low sentiment period, the predictability of low sentiment period is better than high sentiment period.