large shareholders of firms that own more than 5%. Apart from the name
and the shareholdings, DFGM also characterize the blockholders as officers,
if they were officers of the corresponding firms in the respective year. We
also use the information of ownership and whether or not the blockholder
is also an officer of the firm. We construct portfolios similar to the proce-
dure employed above.13 Since the DFGM data covers a substantially shorter
sample period, we consider two different time periods. The first short time
period is 1996 to 2001, which is also used by DFGM. As findings based on
such a short time period might not be reliable, we also analyze the longer
time period from 1996 to 2004. To do so, we simply assume that ownership
levels stay constant from 2001 to 2004.
Panel D of Table 4 shows results based on this alternative sample. They
are similar to the ones presented above using our sample. For example, the
10% cutoff portfolio now generates abnormal returns of 15.17% p.a. for the
period from 1997 to 2005. This number is even higher at 23.70% p.a. for
the 1997 to 2002 period. Although statistically significant at the 1% level,
this very large number has to be treated with some caution as it is based
on a very short investment horizon. Nevertheless, these results suggest that
the abnormal returns documented before are not due to data problems and
carry over to other data sources.
13There are some necessary adjustments. The data of DFGM are somewhat differently
organized. Each firm of their 1996 sample issues ownership information during the year
1996. As a result, all ownership information of the 1996 firms was public information by
the beginning of the year 1997. Consequently, our portfolios invest in firms of the DFGM
year t at the beginning of year t + 1.
14