臺大管理論叢 NTU Management Review VOL.28 NO.3

Does CEO Career Horizon Lead to Corporate Misconduct? Evidence of Taiwanese Semiconductor Firms 156 3.4 Control Variables In addition to the independent and moderating variables, we also included several levels of control variables (i.e., firm-, board- and CEO-level control variables), which fall outside the purview of our inquiry but may have effect on corporate misconduct. This study first considers two firm-level variables, i.e., firm size and performance, because they have been proven to be related to corporate misconduct (Baucus and Near, 1991; Gabbioneta et al., 2013). We measured firm size as the natural logarithmic transformation of the number of total assets. Considering the different meanings of various performance indicators, the decision-making processes of CEOs are likely influenced by different performance metrics, suggesting the necessity of controlling their possible impacts (Elyasiani and Zhang, 2015; Shani and Westphal, 2016). Each firm’s ROA and ROE during the study period were thus included as control variables. In addition to board size and board functional diversity, we also controlled for several board-level variables, some of which were shown to predict a firm’s level of corporate ethics and social responsibility (Harjoto and Jo, 2011; Liao, 2010), including board independence, national diversity and ownership. Board independence, as the ratio of outside directors was calculated by dividing the number of outside directors by the total number of directors for each firm during the study period. We were conservative in our identification of outside directors, and we did not consider those directors who currently work for the firm in some capacity, or have previously worked for the firm, or have family or professional ties with the CEO. Board nationality was calculated by dividing the number of foreign directors by the total number of directors for each firm, and board ownership was measured as the number of shares owned by the board members multiplied by the stock price per share on the last day of the stock market (Barker and Mueller, 2002). Regarding CEO-level control variables, CEO duality, CEO overconfidence, CEO ownership and CEO tenure were controlled for because they have been identified in prior studies as influencing illegal activity within firms. CEO duality (i.e., the separation of the CEO and chairman positions) was measured as a binary variable which was coded 1 when the CEO was not the board chair and 0 when both positions were held by the same individual. About the measure of CEO overconfidence, we used a dummy variable which was coded 1 if the CEO was a net buyer of the company stock for a longer period than when they were a net seller during the first five years in which the company appeared in their sample; whereas if on the contrary, a dummy variable coded 0 was used (Malmendier