![Show Menu](styles/mobile-menu.png)
![Page Background](./../common/page-substrates/page0023.png)
13
臺大管理論叢
第
27
卷第
4
期
variable QFII_HH is denoted as 1 for a firm which is classified as having both high QFIIs’
ownership and high QFIIs’ shareholdings volatility, and 0 otherwise.
Control Variables:
As the variation of the firm’s earnings increases, an increasing risk of association
related to the violation of debt covenants arises. To avoid a debt covenant violation,
managers of these firms are more likely to adopt aggressive earnings management reporting.
It provides an incentive for a firm to smooth its reported income, which in turn can affect
earnings persistence (Carlson and Bathala, 1997). This study uses leverage (LEV), defined as
total liabilities divided by total assets, to proxy for default risk. We also incorporate market-
to-book ratio (MB), which is calculated by the market value of the common equity to book
value of the common equity at the end of the calendar year into regressions to proxy for
growth opportunities (Collins and Kothari, 1989). Finally, we include firm size (SIZE),
which is calculated by the natural logarithm of the market value of common equity to control
for the potential effects of omitted variables (Becker, DeFond, Jiambalvo, and
Subramanyam, 1998).
3.3 Model Specification
Financial reporting may differ markedly between firms due to unobservable firm-
specific traits (Henderson and Kaplan, 2000). With panel data analysis, especially as the
estimation focuses on within-firm variations, omitted variables bias is avoided, provided it is
reasonable to assume that the omitted variable is constant over the time frame of the study.
However, the balanced panel analysis requires consistently reported data across every year of
the panel, thus, it falls into trouble for the potential sample attrition and survivorship bias.
Consequently, we use the unbalanced-panel regression which controls for the firm and the
year fixed effect in the following examinations.
We follow the earnings informativeness model suggested by Tucker and Zarowin
(2006) to construct the preliminary regression. Namely, this study uses Equation (1) as a
benchmark model to understand the role that managerial income smoothing plays in the
forecasting capability of past and current earnings onto future earnings, or in other words,
earnings informativeness.
R
t
= β
0
+ β
1
X
t-1
+ β
2
X
t
+ β
3
X
t3
+ β
4
R
t3
+ β
5
IS
t
+ β
6
IS
t
*X
t-1
+ β
7
IS
t
*X
t
+ β
8
IS
t
*X
t3
+ β
9
IS
t
*R
t3
+ β
10
LEV
t
+ β
11
MB
t
+ β
12
SIZE
t
+ ε
t
(1)