194 /304
投資機會和投資者保護機制對控股股東派息的影響
194
As the three legal institutions (
Legal, Right
, and
Disclosure
) are highly correlated
with each other, we examine the potential effects of multicollinearity among them by
developing an aggregate legal institutional measure (
Aggregate
) using the first principle
component of the three legal factors. The regression results, which are reported in Table 6,
are similar to those in Table 5. The coefficients of
Aggregate
are positively and
significantly associated with all of the dividend payout ratios except
IADvd/E
. More
importantly, the coefficients of
G*Diverge*Aggregate
are negatively and significantly
associated with all four dividend payout ratios. This suggests that the positive association
between dividend payouts and control divergence that is documented by Faccio et al.
(2001) becomes less pronounced in countries with better investor protection, especially
when the investment opportunities are good.
6
4.2 Sensitivity Tests
We perform three sensitivity tests to verify the documented relationships and report
the results in Table 6. First, as 29% of the observations in our sample are from the United
Kingdom and 17% are from Japan, the regression results may be driven by these two
countries. However, the deletion of the observations from the United Kingdom and Japan
does not change our main results (shown in column (1) of Table 6).
Second, since our sample period covers 1990 to 2000, the regression results may be
affected by the Asian financial crisis that occurred in 1997-1998. To eliminate the effects
of the financial crisis on the dividend decisions of controlling shareholders, we exclude
the observations from 1997 and 1998 from Asia and the results remain unchanged as
shown in column (2) of Table 6.
The fact that our sample includes multiple firm-years for the same firm may bias the
regression results. To address this issue, we use the Fama-MacBeth method to calculate the
annual regressions, and the results remain unchanged as shown in column (3) of Table 6.
6 The (unreported) regression results on the dividend payout ratios before adjustment for industry effects
with
Aggregate
produce similar results.