A Note on the Use of RavenPack in Academic Research

Title: A Note on the Use of RavenPack in Academic Research

Abstract: Using a comprehensive sample of press releases, we analyze the completeness of RavenPack’s (RP) news data. While RP includes 92% of the average firm’s press releases, there is large cross-sectional variation. RP’s coverage is related to (1) prominent drivers of the cross-section of stock returns (e.g., beta, size) and (2) proxies for firms’ information environment (e.g., analyst coverage). This coverage bias may impact research results, which we illustrate by analyzing investors' reaction to earnings announcement press releases. RP’s sentiment score is strongly related to dictionary-based tone measures    (especially negativity), the press release’s topic, and firm characteristics. Most notably, firms with high (low) expected returns have, on average, more positive (negative) RP sentiment scores. We provide several suggestions for avoiding obtaining biased research results when using RP data.