Using High-dimensional Corporate Governance Variables to Predict Firm Performance

The purpose of this paper is to empirically identify and quantify correlations between corporate governance practices of firms and their future financial performance. LASSO estimation technique was used on a comprehensive set of corporate governance-related variables provided by The Board Director Training Institute of Japan (BDTI) and compared to firms’ total shareholder returns (TSR) as well as other performance measures for the listed firms in Japan. Through LASSO, we find the following: First, a number of corporate governance policies or attributes that relate to external monitoring have positive correlations with future TSR as expected. Second, somewhat unexpectedly, only a few variables associated with internal monitoring and incentive practices show correlations with future TSR. Third, such unconditional associations between specific corporate governance practices and TSR are affected by other governance practices. After confirming the stability of these results through OLS estimation, we constructed a prediction model of firms’ future TSR and further show that the investment strategy based on the model’s predictions could generate non-negligible improvement in returns by including the corporate governance-related variables in the predictors. These results suggest that high-dimensional corporate governance variables contain more informative signals associated with future firm performance than simple reliance on purely financial data can provide.