Volatility spillovers across stock returns and user-generated content

This study examines the interdependence between a company’s stock returns and user-generated content (UGC) concerning (the products of) that company. We investigate this interdependence by studying the presence of mean, shock and volatility spillover effects between returns and UGC. The number of positive and negative tweets, blog posts, forum posts and daily searches for ticker symbols in the Google search engine are used as measures of UGC. The UGC data is collected via multiple sources over a six month period. Using a multivariate generalised autoregressive conditional heteroscedasticity - Baba, Engle, Kraft and Kroner (GARCH-BEKK) model we identify the source, magnitude and significance of mean, shock and volatility spillover effects between UGC and returns. We estimate the BEKK model for 27 different combinations of variables and compute averages over those results. The (average) results confirm the presence of spillover effects and show that there is a stronger connection - both in magnitude and in significance - in terms of mean, shock and volatility spillovers from UGC to returns than from returns to UGC. There are significant spillover effects between the various UGC metrics as well and these are larger than the effects from returns to UGC. This indicates that online content is affected more by other online content than by stock returns. Positive and negative content exhibit different spillover effects. Moreover, new product launches explain part of the volatility dynamics in stock returns and UGC.