Changes in the I/B/E/S Database and Their Effect on the Observed Properties of Analyst Forecasts

Abstract

I/B/E/S is a common source of analyst forecast data, and the consistency of the data is important for both practice and academic research. We document substantial changes in the historical contents of I/B/E/S over time. We compare two versions of the I/B/E/S EPS detail file: one available in 2009 and the other available in 2015. Examining a common sample period (1993 through 2007), we find that 12 percent of the observations in the earlier version of I/B/E/S are omitted from the later version, whereas 6 percent are added to the later version as new observations that did not exist in the earlier version, and 9 percent of the observations in both versions report different forecasts of earnings. These discrepancies are systematically associated with various firm and analyst characteristics. We find that earnings forecasts in the later version are significantly more accurate, less biased, and bolder than the forecasts in the earlier version. We also find substantial differences in the firms identified as meeting or just beating analysts’ expectations of earnings. Finally, these discrepancies are less pronounced when comparing separate versions of the EPS summary file, but are more pronounced when examining individual analyst quarterly forecasts and annual cash flow forecasts.