AbstractContribution to online communities, such as stock message boards, may be considered as “paradoxical” (Rafaeli and LaRose, 1993), in that a “free-riding tendency” (Olson 1985) by rational participants would result in a lack of original content. However, online financial message boards continue to thrive as a platform for investors to share tips, express opinion and to disseminate (mis)information. The high interactivity and low communication costs of online message boards result in a potential for social networks to amplify the effects of behavioural biases (Zhang, 2014; Russ, 2007). However, previous studies on the informational content of message board postings, and the power of online communications in predicting abnormal returns and trading volumes produce conflicting results (Nardo et al., 2015).
Prior studies focusing on small-capitalisation firms have so far produced stronger associations (Leung and Ton, 2015), but academic study in the field is still considered to be in its relative infancy (Zhang, 2014). This thesis contributes to existing literature examining relationships between message board postings and trading activity, by focusing on 854 securities listed on the comparatively lightly-regulated AIM submarket of the London Stock Exchange. Sentiment analysis techniques are applied to a dataset of 245,200 message board postings to calculate novel measures of posting sentiment and agreement, alongside posting frequency.
The first empirical chapter of this thesis uses the event study method to analyse intraday abnormal returns and trade volumes surrounding increases in message board posting activity. Positive cumulative average abnormal returns are identified prior to – and immediately following – instances of abnormal posting frequencies, before drifting downwards. However, these returns are found to be economically small and statistically insignificant in many cases. Disaggregation of posting events by firm-specific characteristics finds only sporadic statistical significance in returns. However, abnormal trade volumes do appear to differ according to company type and the level of online sentiment.
The second empirical chapter tests for causality between message board posting and share trading variables. Lagged correlation coefficients find evidence that posting behaviour impacts upon trading activity, and this is further supported by results of Granger Causality and Impulse Response testing. To the contrary, little evidence is found that message board postings impact upon share returns. Indeed, stronger evidence suggests the reverse to be true; share returns inform message board postings.
The final empirical chapter of this thesis focuses on an individual instance during which an AIM-listed firm was subjected to a damning research note by a short-seller employing the online medium. Behaviour of the online (and offline) investment community following publication of the research is consistent with prior theories of social contagion and online herding. When considered in isolation, high-strength “shock” online content can possess the potential to impact upon online investor behaviour and the price-discovery process.
This research ultimately adds to previous studies of online financial communities, and the findings of this these are discussed in the context of both traditional and behavioural finance.
|Date of Award||2019|
|Supervisor||Bruce Burton (Supervisor) & David Power (Supervisor)|