While our variables include ratio variables such as debt-to equity
ratio and current ratio, some observations represent
unacceptably high values in the ratio variables if only because
the denominator is infinitesimal. To reduce the effect from
such outliers, we first ran the regression and calculated standardized
residuals. Then, we dropped the observations having
standardized residuals greater than 5.7 Further, given
that our estimation method requires at least two consecutive
years of data, some observations were additionally dropped.
After this procedure, our data consisted of 121 publicly traded
firms from 35 Chaebols. Since we constructed unbalanced
panel data from 2003 to 2010, our final data consisted of 584
affiliate firm-year observations.