dispersions of variables, such formulas imply a bin width of one penny (the
minimum resolution for our data).
1.“Sustain Recent Performance.”Press reports on corporate earnings
typically compare current results with those from a year ago. Con-
sistent with this practice, we provide evidence that earnings from
one year ago constitute an important threshold for earnings re-
ports, as we posited in our model. The distribution of the change in
earnings, denoted ∆EPS, is simply EPS minus EPS of four quarters
ago. (The appropriate recent available benchmark proves to be the
corresponding quarter from a year ago since earnings exhibit strong
annual seasonal variation.) The distribution of ∆EPS is plotted in
figure 18.5.
Since corporate earnings tend to grow (surely in nominal terms),
we do not expect the central tendency of the distribution to be close
to zero. Indeed, the median and the mode of the distribution of the
overall sample are 3 cents, while the mean is 0.81 cents. It is all the
more remarkable, then, that we observe a large jump in the distribu-
tion at zero. In the region of small negative changes, the distribution
appears to have been “shaved,” with some density mass transferred
to zero or slightly above. This pattern of ∆EPS is consistent with ex-
ecutives’ managing earnings to come in at or above the comparable
figure for four quarters ago.^25650 DEGEORGE, PATEL, ZECKHAUSER
Figure 18.5. Histogram of change in EPS (∆EPS=EPSt−EPSt− 4 ): exploring the
threshold of “sustain recent performance.”
(^25) A qualitatively similar pattern is reported in Burgstahler and Dichev (1997, fig. 1), al-
though, since they deflate earnings, the extreme dip in density just below zero in their distribu-
tion of scaled earnings is most likely spurious (as discussed in 3A above).