Does Chatter Really Matter? Dynamics of User-Generated Content and Stock Performance

Abstract

User-Generated Content in online platforms or chatter for short provides a valuable source of consumer feedback on market performance of firms. This study examines whether chatter can predict stock market performance, which metric of chatter has the strongest relationship, and what the dynamics of the relationship are. The authors aggregate chatter (in the form of product reviews) from multiple websites over a four year period across six markets and fifteen firms. They derive multiple metrics of chatter (volume, positive chatter, negative chatter, and 5-start ratings) and use multivariate time series models to assess the short and long term relationship between chatter and stock market performance. They use three measures of stock market performance: abnormal returns, risk, and trading volume.

The findings reveal that two metrics of chatter can predict abnormal returns with a lead of a few days. Of four metrics of chatter, volume shows the strongest relationship with returns and trading volume, followed by negative chatter. Whereas negative chatter has a strong effect on returns and trading volume with a short “wearin” and long “wearout,” positive chatter has no effect on these metrics. Negative chatter also increases volatility (risk) in returns.

A portfolio analysis of trading stocks based on their chatter provides a return of 8% over and above normal market returns. In addition to the investing opportunities, the results show managers that chatter is an important metric to follow to gauge the performance of their brands and products. Because chatter is available daily and hourly, it 2 can provide an immediate pulse of performance that is not possible with infrequent sales and earnings reports. The fact that negative chatter is more important than positive, indicates that negatives are more diagnostic than positives. The negatives suggest what aspects of the products managers should focus on.

Keywords: User-Generated Content, stock returns, online word-of-mouth, Vector Auto-Regression (VAR), computational text processing