IMPORTACIÓN

In this section, I estimate the regression model to test Hypothesis 3 and report the results. To test Hypothesis 3, I need to compare the difference in future firm performance of the treated firms before and after the promotion, relative to the same difference in the future firm performance of the control firms, given that the manager is retained. For future firm performance, I calculate the average ROA across years

t+ 1 to year t+ 3 for every firm i and year t.

I examine the effect of a promotion of a director on ex-post performance using the following linear regression model:

ROAit+1,t+3=ηt+ηi+η1CEO Retainedit+η2P ostit+η3CEO Retainedit×P ostit +η4T reatit+η5T reatit×CEO Retainedit+η6T reatit×P ostit +η7T reatit×P ostit×CEO Retainedit+η8Xit+it , (1.3)

where the dependent variableROAit+1,t+3 is the average ROA from yearst+1 tot+3.

fixed effects, and ηjt represents industry-times-year fixed effects. CEO Retainedit is an indicator variable that takes the value of one if the CEO of firm i is retained in year t. T reatit is an indicator variable that takes the value of one for firm i from year t −4 to year t+ 3, where the promotion happened in year t, and takes the value of zero for the matching control firm during the years t−4 to t + 3. P ostit is an indicator variable that takes the value of one for both control and treatment firms from year t to year t+ 3, where t is the year of a promotion. Xit represents a vector of controls including CEO age, CEO tenure, board size, firm size (measured by logarithm of Sales), book leverage, board’s busyness, fraction of non-executive directors on the board, and directors’ past experience. All variables are defined in Table 1.1 in the appendix. The main coefficient of interest is η7 which estimates the

difference in average future profitability of the retained manager for the treated firms over a four-year period before and after the promotion, relative to the same difference for the control firms where no such promotion occurs.

Table 1.8 presents the results of this regression. In column (1), the coefficient

η1 is positive and significant at the 1% level. It shows that relative to the years in

which the CEO is fired, firm profitability is higher in the years when the CEO is retained. In column (2), I present the results of regression equation 1.3 by comparing the average future performance of the treated and control firms. The coefficient

η5 on T reatit × CEO Retainedit is insignificant. It suggests that if the CEO is retained, the future performance of the treated firms in which a promotion occurs is not statistically different from the future performance of the control firms. In column (3), I present the results of the overall regression model. The coefficient η3 is

statistically insignificant. It suggests the future performance of the retained manager for control firms does not change after the year of the promotion in treated firms. The coefficient on the triple-interaction term (η7) is negative and significant at the

declines after the promotion for the treated firms relative to the matching control firms. In terms of economic significance, a promotion in the treatment firms, relative to control firms, is associated with a 0.89% decline in the average future profitability over a four-year period.

The results in column (1), (2), and (3) include firm fixed effects, so the effect of promotions on future firm performance is relative to the average firm performance across the sample period. However, the average firm performance across the sample period includes the performance by the retained CEO as well as the performance of other CEOs who worked with that firm. Therefore, in column (4), instead of firm fixed effects, I control for CEO fixed effects, which enables me to capture the effect of promotions on future firm performance relative to the average firm performance by the same CEO. The results remain the same.

1.3.7 Other Results

In this section, I study the effect of reputational concerns on turnover-performance sensitivity using alternative ways to capture the reputational concerns of the board of directors. First, I use the equity compensation of the directors as a proxy for how much they care about shareholder value. The motivation for using this measure is based on the agency view: shareholders provide incentives in the form of equity compensation to align the interests of the board with those of the shareholders. Fol- lowing the assumption about the board’s payoff in the model, one minus the fraction of the directors’ equity compensation captures the director’s reputational concerns (parameter β in the model). The equity compensation of the directors for a firm i

in year t is calculated using the average of the proportion of equity compensation to total compensation across all directors on the board.

I study the relation between the equity compensation of the directors and turnover- performance sensitivity by estimating the linear probability regression model specified

in equation 1.1. Table 1.9 presents the results of this regression. The results suggests the effect of performance on CEO dismissal is weaker when the board’s equity com- pensation is high. In terms of economic significance, a one standard deviation change in equity compensation is associated with a 65% change in turnover-performance sen- sitivity. The results are robust to including year, industry, industry-times-year, and firm fixed effects.

Second, I use the tenure on the board as an inverse proxy for a director’s repu- tational concerns. The motivation for using this proxy is that in the early years of their tenure on the board, directors have greater reputational concerns than in the later years. The reason is that the board’s decisions in the early years of their tenure have a greater impact on their future payoffs through additional board seats and pos- sibly more compensation on those board seats. Board tenure is calculated using the average tenure across all directors on the board.

I use the linear probability regression model in equation 1.1 to study the relation between board tenure and turnover-performance sensitivity. Table 1.10 presents the results of this regression. The results suggests that in the later years of the direc- tor’s tenure, turnover-performance sensitivity becomes weaker. In terms of economic significance, a one standard deviation change in the director’s tenure is associated with an 83% change in turnover-performance sensitivity. The results are robust to including year, industry, industry-times-year, and firm fixed effects.

Directors who are new to the board or new to the role might be selected into the board because the current directors have stronger ties with the CEO and have become entrenched. Hence, the arrival of new directors decreases the average board entrench- ment. In that case, in the early years of a director’s tenure, turnover-performance sensitivity is stronger because the board’s entrenchment is low. I correct for this bias using director deaths as an exogenous shock to directors’ tenure on the board. After a director’s death, the average tenure of directors decreases. The identifying

assumption is that the decrease in average board tenure is caused by a director death and not by the selection of new directors to an entrenched board.

I analyze how turnover-performance sensitivity changes from five years before a director’s death to five years after a director’s death.7 _{Suppose at least one director}

death occurs in firm i in year t. I construct an indicator variable Post Death that equals one for firmi from yeart+ 1 to year t+ 5 and equals 0 from year t−4 to year

t. Using this variable, I estimate the following linear probability regression model:

CEO T urnoverit =ηt+ηi+η1ROAit+η2P ost Deathit

+η3P ost Deathit×ROAit+η4Xit+it, (1.4)

Table 1.11 presents the results of this regression. In column (1), the coefficient on the interaction term is negative and significant at the 5% level. The results suggest turnover-performance sensitivity increases after a director’s death. The results are robust to including year, industry, and industry-times-year fixed effects.