198 /304
投資機會和投資者保護機制對控股股東派息的影響
198
4.3 Additional Tests
We perform following additional tests to verify the documented results. First, the
documented results may be affected by potential economic or political changes across
years. To ensure that the results are free of such biases, we rerun the regressions with year
dummies. We also compute t-statistics that are adjusted for serial correlation using the
Newey and West (1987) procedure. The results remain the same.
If ownership structure, institutional factors, and dividend payouts across countries are
endogenously determined, then the inference from the ordinary least-squares models may
be misleading. To address this concern, we estimate a two-stage least squares (2SLS)
model using the legal origin and wealth of countries as instruments. The first-stage
regressions regress
Diverge
and each of the legal institutions (except
Legal
) on the
dummies for each type of legal origin (English, French, German, and Scandinavian) and a
country’s 10-year average per capita GDP before the test period, in accordance with the
work of Leuz et al. (2003). The 2SLS estimation procedures produce similar results.
7
The financial systems in Japan and Germany are more bank oriented than those of the
other countries (Moerland, 1995; Lehmann and Neuberger, 2001). To ensure that the
regression results are not influenced by the power of banks in determining the dividend
decisions of controlling shareholders, we exclude the observations from Japan and
Germany. The results, which are unreported, are similar.
Although a large control divergence creates an incentive for controlling shareholders
to expropriate, the possession of a high level of cash flow rights may reduce this incentive.
Thus, we exclude 898 observations (approximately 15% of the full sample) in which
Own
is greater than 50%. The results remain almost the same. We also exclude the observations
with the top 5% largest control divergence to ensure that our results are not affected by a
small group of firms with a particularly large control divergence, and the main results
remain unchanged.
7 One should interpret the 2SLS results with caution, because the validity of the results that are
generated depends on how successfully a country’s legal origin and wealth represent the right
instrumental variables in the model.