ings as cutoff. However, the statistical significance is slightly reduced to the
5% level. With abnormal returns of 12.0% p.a. and 8.8% p.a. for the rank
1-100 and rank 1-250 portfolios, respectively, these results are still highly
economically significant.
3.3 Multivariate Analysis
Brennan, Chordia, and Subrahmanyam (1998) report several individual firm
characteristics that can drive returns and are not captured by the four factor
model employed above. In order to explore whether such firm characteristics
might drive returns of firms with high managerial ownership and thus explain
our results, we also run multivariate Fama and MacBeth (1973) regressions.
In these regressions we relate monthly raw and industry-adjusted returns of
firms to managerial ownership and further firm-specific characteristics. We
estimate the following regression separately for each month in our sample of
S&P 1500 firms:
Ri,t = αi + βi,ι ∙ Shrown + βi,2 ∙ D(10%) + βi,3 ∙ Fi,t + εi,t, (2)
where Ri,t denotes the return (raw or industry adjusted) of firm i in
month t, Shrown is the maximum share of the company’s stock owned by
an officer, D(10%) is a dummy variable that takes on the value 1, if Shrown
is larger than 10%, and Fi,t is a vector of firm characteristics. It includes
the firm characteristics examined in Brennan, Chordia, and Subrahmanyam
(1998), and additionally the Gompers, Ishii, and Metrick (2003) G-index,
five-year sales growth, and S&P 500 inclusion.15 Final parameter values are
15These variables are also used in Gompers, Ishii, and Metrick (2003). A detailed de-
scription is given in their Appendix 2. Data on the governance index G are taken from
Andrew Metrick’s Web page http://finance.wharton.upenn.edu/ metrick/data.htm.
17