We take CG
attributes, CG variables, and control-variable data
from the Annual Reports of HK-listed companies for
financial years 2001, 2003, 2005, 2007 and 2009.
We use the data on all stocks listed on the main board
for HK companies, except those companies incorporated
in mainland China (i.e. H shares and Red
Chips6), because their CG structure is fundamentally
different from that of local firms. We also exclude
newly listed companies, because of possible bias
resulting from the fact that companies need to have a
good track record to satisfy listing requirements. This
exclusion yields a more balanced panel of data. We
omit foreign companies from our dataset, because
such companies use significantly different layouts in
their annual reports than those used by local companies.
We obtain accounting and firm characteristics
data from Datastream International. We use company
annual reports to fill in any gaps caused by
missing data. To reduce the effect of outliers on our
regressions and index results, we winsorize the
Tobin’s q, market to book value (MTBV), and
nonbinary variables in our model at the 5% and
95% percentiles. The order of the statistics is thus
unaffected, and we can use all the available data.
Table 3 shows the unwinsorized and winsorized
summary statistics, in which the medians remain
unchanged but the distributions of the variables are
less skewed.