Finally, as a practical data constraint, only three future years’ earnings are
included in the analysis. Thus, revisions in expectations of earnings growth for
periods 4 and beyond are excluded from the analysis, insofar as they are
uncorrelated with the included earnings variables. This also reduces the model’s
explanatory power. Although additional years’ earnings growth could be included,
the further out we go the greater is the errors-in-variables problem
discussed above. For all the reasons just discussed, we believe that the impact of
lack of timeliness is even greater than our models indicate, thus suggesting a still
smaller role for value-irrelevant noise.