GOOGLE SEARCH BASED SENTIMENT INDEXES

As Internet searches are generated by agents’ spontaneous behaviour, these search patterns may possess signalling properties. This study sought to confirm whether Internet search-based data have the potential, both, to reveal populations’ underlying beliefs directly and to affect stock markets of countries – in this case, Portugal. Based on the Internet search volume of several queries related to household concerns, we constructed two Google-based sentiment measures encompassing both positive and negative search terms (i.e., a positive sentiment index [PSI] and a negative sentiment index [NSI]). 

The final list of negative terms comprised   “austerity” ,   “taxes”,   “rents” ,  “Euribor” ,   “crisis” ,  ”debt”,  “finance”,  ”gold price”,  ”unemployment” , and  ”poverty” . The final list of positive terms included  ”stocks” ,  ”consume” ,  ”credit”,  ”GDP”,  ”Lisbon stock market”,  ”dividends”,  ”profits”,  ”investment”,  ”entrepreneurship”, and  ”partnership”.

The results reveal that both measures are correlated with aggregate stock market returns, trading volume and abnormal trading volume. We also found that positive sentiment has a stronger impact on these stock market variables than negative sentiment. In addition, the results show that the proposed sentiment measures are significantly useful when making short-term predictions of market returns and volume. These findings confirm the validity of using sentiment measures based on search queries since they are available for shorter periods than other well-known measures of market sentiment. The results contribute to the literature by highlighting the different roles of positive and negative sentiment in stock market activity.