To examine this prediction, we classify our sample into three groups (i.e., high, middle, and low) based on terciles of manipulation firms’ CEO_SENSITIVITY, and estimate the conditional logistic regression of the likelihood of manipulations separately for the low and high subgroups. Results are tabulated in Panel B of Table 4. Consistent with our prediction, the CEO power variables are generally positively and significantly associated with accounting manipulations for the high CEO_SENSITIVITY group, but the coefficient estimates for the CEO power variables are not significant for the low CEO_SENSITIVITY group. This finding suggests that CEOs are more likely to engage in accounting manipulations when both CEO equity incentives and CEO power are high, consistent with the CFO pressured explanation of accounting manipulations.26