Behavioural finance tells us that investors may be characterised by a lack of singular introspection, which causes them to act and think in the same way as those around them. The question asked is whether social media can reflect this behaviour and if so, can networks such as Twitter be correlated to, or even predictive of, market returns?
This thesis aims to explore how sentiment reflects earnings season events on attention-grabbing stocks. Twitter – a microblogging platform – is examined as a proxy for social mood with regards to its potential in predicting investor sentiment around market events. Ten major technology- and service-based stocks are analysed during Q4 2016.
Together with Google App Scripts and natural language processing, sentiment indicators are extracted and converted into quantifiable values. The predictive power of investor sentiment is examined to determine whether ex-ante Twitter sentiment can explain ex-post stock price and volume movements.
The results show high correlations between financial market instruments and sentiment indicators. Further, using regression analysis, it is validated that the movement of stock prices and volume can be modelled up to four days in advance. Of the stocks measured, 80% indicated predictability through sentiment models. Moreover, influencer weights were 1.9 times more significant when compared to collective sentiment.
Finally, the results indicate that the accuracy of security predictions can be significantly improved by the inclusion of the Weighted TIS sentiment dimension. The results are consistent with existing theories of social mood and are supportive of the investor sentiment hypothesis in behavioural finance.
|Educations||MSc in Applied Economics and Finance, (Graduate Programme) Final Thesis|
|Number of pages||74|