10 /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
10
Hypothesis 2a: Ce t er i s par ibus , the negat i ve as soc i at i on be tween earnings
informativeness of smoothed firms and high QFIIs’ ownership is
supported in high shareholdings volatility case.
Hypothesis 2b: Ce t er i s par ibus , the negat i ve as soc i at i on be tween earnings
informativeness of smoothed firms and high QFIIs’ ownership is
mitigated in low shareholdings volatility case.
3. Research Design
3.1 Data and Sample Selection
The years 1997~2007 are chosen as the observation periods. The year of 1997 is chosen
as the starting year because the data of QFIIs ownership is available and we require six years
of financial data from which to calculate the income smoothing measure.
2
The sample period ends in 2007 because we need the future three years’ earnings and
stock returns to examine the earnings informativeness. The empirical data are retrieved from
the
Taiwan Economic Journal
(TEJ) database. Table 1 reports the sample selection process in
the study.
The total observations on the TEJ database from 1997~2007 are 14,437 firms/year.
Consistent with extant literature, finance-related institutions (Code 28) are excluded since
they are subject to different disclosing requirements. We also exclude 679 observations
which are classified into other industries (Code 99) for their diverse characteristics and 131
observations for belonging to regulated industries (Code 97). To compute discretionary
accruals, industries with fewer than ten observations in a year are excluded (Dechow, Sloan,
and Sweeney, 1995; Jones, 1991) in the coefficients estimating of the non-discretionary
accruals model. This study further deletes 5,588 observations for data unavailable in
calculating the income smoothing measure and 1,188 observations for data unavailable for
the QFIIs related variables. Further exclusions include 844 observations with data
unavailability for other control variables. The final sample consists of 5,766 observations in
empirical analysis.
2 Following the study of Tucker and Zarowin (2006), we need six years to compute the continuous five years
correlation between the change in discretionary accruals and the change in pre-discretionary income using the
current year’s and past four years’ observations. For example, we use years from 1992~1997 to calculate the
discretionary accruals variable, and then use the five years change in discretionary accruals (1993~1997) to
calculate the income smoothing measure in the year 1997.