To implement IPTW, we first estimate a probit model predicting the probability of a firm having a high or low level of CEO-CFO LSM. Following Almandoz and Tilcsik (2016), we use the median value of CEO-CFO LSM to code whether a CEO-CFO dyad has a high or low level of LSM. We include all the firm-level and CEO-CFO dyadic level variables used in predicting CEO-CFO LSM in the first-stage probit regression. In addition, we include the ratio of the total number of words spoken by the CFO during the Q&A section to the total number of words spoken by the CEO in a firm year as an additional predictor. If the CFO attempts to ingratiate the CEO, the CEO may delegate more speaking opportunities to the CFO. Yet, we would not expect the ratio of words spoken by the CFO to the words spoken by the CEO to directly influence our outcome variables. We use the inverse of the propensity calculated from the first-stage probit regression as a weight in the second-stage regressions. Results from the IPTW methodology are reported in Table 6.