Time Stamp Errors and the Stock Price Reaction to Analyst Recommendation and Forecasts Revisions

This paper investigates the problem of time stamp errors in the IBES database. We show that IBES did not store the original announcement date of both recommendations and forecasts on U.S. stocks until 2001 and even later for other countries. The announcement date in IBES is often effectively the date on which the information was recorded by IBES and, therefore, systematically delayed. Using event study analysis and comparing IBES to alternative data sources, we show that the announcement day effect is underestimated in IBES while pre-announcement returns are overestimated as they often include the effective announcement day. We also show that time stamp errors in IBES are not randomly distributed in the cross-section but differ significantly across several firm, broker, and analyst characteristics. A consequence is that cross-sectional differences in announcement returns may be driven by cross-sectional differences in time stamp errors. We also show how existing research is affected by time stamp errors in IBES. Finally, we suggest three alternative ways to mitigate the time stamp error problem in daily IBES data.