20 /274
公司盈餘平穩化行為與盈餘資訊性之關係－合格境外機構投資者角色之檢測
20
(Greene, 2004) to examine the influences QFIIs on the earnings informativeness for firms
with income smoothing.
This study firstly examines the regression suggested by the Tucker and Zarowin (2006)
as a benchmark model. Table 5 reports the benchmark regression results (hereafter, we
denoted them as the IS model). From Table 5, it is found that the coefficient of IS
t
*X
t3
is
0.045 (
t
= 2.90), which is statistically significant at the 1% level in the IS model. This result
is consistent with the findings documented by Tucker and Zarowin (2006). Thus, income
smoothing is one way for managers to communicate private information and supports the
notion that income smoothing could enhance earnings informativeness. Definitely, the model
specification of Tucker and Zarowin (2006) provides this study with a unique setting to
further examine the role of QFIIs in regard to earnings informativeness for firms with
income smoothing using the case of Taiwan stock market.
Incorporating the role of QFIIs into consideration, the coefficient of QFII
t
*IS
t
*X
t3
is
-0.004 (
t
= -1.93) and statistically significant at the 10% level in the IS_QFII model. This
result suggests that the more that income smoothing associates with higher QFIIs’
ownership, the more that the earnings informativeness of income smoothing is reduced. This
empirical result falls in line with the opportunistic role of QFIIs in earnings informativeness
of income smoothing and supports the hypothesis. Thus, QFIIs are likely to access the
private information and adopt self-interested behavior, which in turn, reduces the informative
component of earnings for firms with income smoothing in the Taiwan stock market. The
coefficient of QFII
t
*X
t3
is 0.004 (
t
= 2.20) and it is statistically significant at the 5% level. It
implies that QFIIs enhance the informativeness of past and current earnings about future
earnings for firms without income smoothing. The coefficients of IS
t
*X
t3
, IS
t
*X
t
and IS
t
*R
t3
are 0.061 (
t
= 2.83), -0.056 (
t
= -2.58), and -0.067 (
t
= -2.02) in the IS_QFII model, which
are statistically significant at the 1%, 1% and 5% levels, respectively. These results are
approximately the same as the benchmark IS model.
4
4 This study further divides observations with positive discretionary accruals and negative discretionary
accruals and examines whether the earnings informativeness of income smoothing is different between these
two distinct types of manage earnings up and/or down. The coefficients of IS
t
*X
t3
are 0.016 (
t
= 1.80) and
0.028 (
t
= 1.68) in the positive and negative discretionary accruals subsamples, and both statistically
significant at the 10% level in the IS model. The coefficients of QFII
t
*IS
t
*X
t3
are -0.004 (
t
= -2.19) and -0.005
(
t
= -2.39) in the positive and negative discretionary accruals subsamples, and both are statistically significant
at the 5% level in the IS_QFII model. It is fair to conclude that the empirical results are unlikely to confound
by the discretionary accruals between IS and QFIIs on the earnings informativeness.