We use Principal Component Analysis (PCA) to
combine CG attributes. In prior research, CG attributes
are usually combined using equally weighted or
binary approaches. Gompers et al.’s (2003) governance
index, for example, consists of 24 variables, all
of which are binary. They calculate the index using
the equal weighted sum of the variables. Others, such
as Credit Lyonnais Securities Asia (CLSA) (2001)
Patel et al. (2002), and Garay and Gonza´ lez (2008),
use similar methods to obtain their indices. Although
this method creates data that are more reproducible,
it omits important information that continuous CG
variables yield, and it subjectively sets the thresholds
of binary variables. CG attributes should have
different degrees of importance, and thus an equal
weighting of index components makes the comparison
more difficult. The challenge here is to find the
appropriate weighting of CG attributes. In our
model, we first apply the PCA, and then use the
largest eigenvector to combine the CG attributes.
This method captures the commonality of both
binary and continuous variables, and thus the indices
account for the contributions of individual attributes
to the CG mechanisms.
We use Principal Component Analysis (PCA) tocombine CG attributes. In prior research, CG attributesare usually combined using equally weighted orbinary approaches. Gompers et al.’s (2003) governanceindex, for example, consists of 24 variables, allof which are binary. They calculate the index usingthe equal weighted sum of the variables. Others, suchas Credit Lyonnais Securities Asia (CLSA) (2001)Patel et al. (2002), and Garay and Gonza´ lez (2008),use similar methods to obtain their indices. Althoughthis method creates data that are more reproducible,it omits important information that continuous CGvariables yield, and it subjectively sets the thresholdsof binary variables. CG attributes should havedifferent degrees of importance, and thus an equalweighting of index components makes the comparisonmore difficult. The challenge here is to find theappropriate weighting of CG attributes. In ourmodel, we first apply the PCA, and then use thelargest eigenvector to combine the CG attributes.This method captures the commonality of bothbinary and continuous variables, and thus the indicesaccount for the contributions of individual attributesto the CG mechanisms.
การแปล กรุณารอสักครู่..
