An Alternative Measure of Corporate Governance Using Discrete Principal Component Analysis

How is corporate governance to be measured? Much past research has focused on forming an index but this approach ignores the correlations between the variables forming the index and may also carry redundant variables. Principal Component Analysis (PCA) is a possible solution to these issues by distilling components from a Pearson correlation matrix. However, the estimated correlation matrix may be biased when discrete governance variables are involved. We use an alternative method known as discrete Principal Component Analysis which involves applying PCA to a modified correlation matrix which takes into account the underlying distribution of the governance variables. Using a sample of 760 firms in 2002, we find mixed support for prior research. Larger boards are found to negatively impact future ROA of firms whilst the proportion of old directors on the board has a negative relation with abnormal accruals and abnormal returns