revisions following announcements, while a further 19% of revisions are associated with
financial issues (e.g., buybacks and dividend announcements). These proportions have
remained relatively stable over time, suggesting that analysts do not take into account
alternative information at different points in the business cycle, reacting instead
predominantly after quarterly earnings announcements or on guidance news.
Prior research shows that analysts take longer to digest strategic news as the
information is more difficult to map into estimates than accounting-related news.
Bagnoli, Levine, and Watts (2005a, b) show that softer strategic news is initially less
well understood and that it takes time for this information to be interpreted before
analysts react. Such information may only result in a longer term price reaction.
Table 8.1 shows the average duration of each revisions cluster and the percentage of
revisions that occur within two days of the news announcement. We find that around
50% of revisions following earnings and guidance news are made within the subsequent
two days, implying that analysts are quick to respond to such information. On the other
hand, we find that less than one-third of analyst revisions following strategic news occur
within the same time period, confirming that it takes analysts longer to digest the
information and suggesting that clusters are triggered by analysts herding towards a
lead analyst.
Figure 8.8 shows the average change in EPS forecasts from a revisions cluster,
conditioned on good or bad news. When news is good, financial news followed by
earnings news has typically resulted in the largest increase in EPS forecasts. When news
is negative, credit-rating changes and guidance news see the greatest EPS downgrades.
Figure 8.9 illustrates the short-term returns (over a 3-day window) of revision clusters
conditioned on different types of news. Again we find that earnings and financially
The impact of news flow on asset returns: An empirical study 223Figure 8.8.Revisions clusters by news type (source: RavenPack, Factiva, FactSet, Macquarie
Quant Research).