Possible discontinuity in the histograms is evaluated for each of our 12
conditions employing the τ-test used before for the univariate distributions,
which is discussed in the appendix. The τ-statistics confirm concretely what
we see in figure 18.8. The EPS threshold is highly robust. Whether either of
our other two thresholds fails or is met, there is a leap upward in density at
our EPS threshold. Our other two thresholds are less robust to the condi-
tioning of whether another threshold is met or failed; that is, under some
conditions, they are not significant. Specifically, in row 1 of column 2 in fig-
ure 18.8, we observe no clear discontinuity when the EPS earnings are neg-
ative. However, a discontinuity in ∆EPS appears in the other three rows of
column 2. We conclude that the threshold of the previous period’s earnings
is “robust” with respect to the threshold of analysts’ forecast but not to
that of positive earnings.
Similarly, in rows 1 and 3 of column 3 in figure 18.8, we observe no
clear discontinuities in the distribution of FERR when EPS<0 and
∆EPS<0, respectively. The effect of a FERR threshold reasserts itself when
EPS>1 or ∆EPS>0, as seen in rows 2 and 4 of column 3. We conclude
that the threshold of analysts’ forecast matters only if the other thresholds
are met.
Overall, a threshold hierarchy emerges. The positive EPS threshold is the
most important; it prevails regardless of whether or not the other two
thresholds are met. The threshold of previous period’s earnings is second in
importance; it asserts itself only if the positive EPS threshold is met but is
visibly present regardless of whether earnings make the analysts’ forecast.
The threshold of analysts’ forecast is the weakest; it only matters if both the
other thresholds are met.
4.The Effect on Future Earnings from Earnings ManagementThe previous section showed that executives manage earnings to meet
three thresholds. When they do so, as discussed in section 2, current earn-
ings are raised by “borrowing” from future earnings. From an empirical
perspective, firms that barely attain a threshold are suspect of having en-
gaged in upward earnings management. This section examines whether the
future performance of firms barely meeting thresholds is inferior compared
to control groups.
A. The Implication of Earnings Management for Future EarningsEarnings management to meet thresholds in this period will affect next
period’s earnings. Thus, we investigate whether following a period with
likely EM there is any predictable effect on earnings. Our analysis in this
656 DEGEORGE, PATEL, ZECKHAUSER