• No se han encontrado resultados

Instrumento Notarial Helmasperger 1

In document Agradecimientos xi (página 110-117)

1.4.3 ¿Planeo Gutenberg la impresión de un misal?

1.4.4 Instrumento Notarial Helmasperger 1

The starting point is the multivariate probit estimation of CEO depar-ture given the simultaneous and lagged board turbulence. I perform re-gressions for the three definitions of CEO turnover, that is, all turnover, CEO turnover simultaneous with chairman turnover, and forced turnover.

Stock return and ROA are the measures of firm performance in each turn-over definition. In this section, estimations of busy directors instead of the network variable are undertaken.

Table 4.3 shows results of regressions of CEO turnover, with the differ-ent definitions, and the two specifications of firm performance.

Table 4.3

The Pseudo-R2shows satisfactory results in all regressions, indicating relevant model selection. Notice that all explanatory variables are lagged one period, unless otherwise indicated.

The simultaneous board turbulence is positively and significantly as-sociated with CEO turnover in all regressions, while the lagged turbulence is significant for the all CEO turnover category only. This pattern is com-mon to both the stock return and ROA. In the all CEO turnover category, the economic significance of the simultaneous is higher than the lagged, since the contemporaneous average partial effects (APE) is 0.25 and the lagged is 0.09 for the stock return, and 0.23 and 0.09 for ROA. Thus, CEO turnover comes with a higher rate of substitution of board members and relative change in board size. This indicates that the CEO-board relation-ship is of the joint control type. The friendly board (Adams and Ferreira, 2007) type comes to mind.

I run an exclusion test for the board turbulence variables. The null hy-pothesis that the coefficients of the two variables are zero is rejected. This further strengthens my conclusion that CEO turnover and board turbu-lence are related, and that joint control is the most common control type.

At the same time, this is also a powerful test of the shareholder control hypothesis that CEO turnover and board turbulence are unrelated. The results of the exclusion test mean rejection of the shareholder control hy-pothesis.

Now look at the other explanations of CEO turnover. The negative ef-fect of firm performance conforms to the evidence found from the earliest (Coughlan and Schmidt, 1985; Weisbach, 1988; and Warner et al., 1988) and to the latest (Parrino, 1997; Huson et al., 2001; Dahya et al., 2002; and Fich and Shivdasani, 2006) studies. Thus, the general result is that higher performance reduces the likelihood of CEO turnover. The result confirms the prediction common to all board control types. However the effect is economically rather weak, with both coefficient values and APEs close to zero, and not always significant. But again agreeing with former stud-ies, the effect is stronger in regressions of forced dismissal than in all CEO turnover regressions. This is to be expected. The forced dismissals are likely to appear more often in crisis situation than the all CEO turnover category, even though we have noticed the difficulty of drawing a line be-tween the two. The results hold for both stock market and ROA, contrary to the Hermalin and Weisbach (1998) argument that accounting informa-tion is the more reliable.

Outside ownership concentration is significant in only one regression and its sign is not consistent in the regressions. This confirms the expecta-tion for all control types.

The results of board independence are contrary to expectations for all board control types and also at variance with the extant literature. It has a negative sign in all regressions, indicating that higher board independence leads to a lower likelihood of CEO turnover. This applies to all turnover as well as to forced dismissals. Its APE is -0.01 in the all CEO turnover in both the stock return and the ROA regressions, and closer still to zero in the forced turnover regressions. Thus, the longer the average board tenure relative to the CEO’s tenure, the less likely the board is to discharge the CEO. Denis and Denis (1995) find that board independence has less importance when controlling for blockholder pressure, takeover attempts, financial distress, and shareholder lawsuits. It is possible that a similar ef-fect appears when including board turbulence in the regressions. The odd result for independence may of course be due to the fact that the indepen-dence measure here is the tenure difference between the board on average and the CEO.

Among the CEO entrenchment variables only the imported CEO is

consistently significant in all regressions, and with a sign consistent with an information interpretation rather than a colluding board interpreta-tion. An increase in the fraction of CEOs from other companies means a markedly higher likelihood of CEO turnover. Looking at the stock re-turn case, the APE is 0.33 for the regression for all CEO re-turnovers. Thus, having friends on the board in the Gilson and Kraakman (1991) sense does not seem to make the board friendly towards the CEO. This refutes the hy-potheses for entrenchment made under the assumption of CEO control.

I find that the CEO’s age has an impact in all turnover regression, but not in forced dismissals. However, the APE is close to zero in both. This indicates that resignations are to some extent voluntary.

My results for the network variable have no counterpart in the litera-ture. The higher the board network measure, the less likely the board will be to dismiss the CEO. The APE for both the stock return and the ROA case is -0.23 in the regression using all CEO turnovers, and -0.10 in forced turnover. This is contrary to the shareholder control prediction that a bet-ter networked board is betbet-ter able to compare potential CEO candidates.

Fich and Shivdasani (2006) obtained a similar result for busy directors. I make a more detailed comparison below, see table 4.4.

Are busy directors less likely to dismiss the CEO? A main result in Fich and Shivdasani (2006) answers in the positive. My network variable is not directly comparable, since it takes account of both direct and indirect director connections to other firms, but the results in table 4.3 point in the busy director direction. The Fich and Shivdasani (2006) measure is a dummy variable showing 1 if the board has more than half busy directors.

A director is termed busy if he or she holds three or more board positions in other listed firms. In order to compare, I have constructed busy director variables in the same manner, but I have set two more cut-off points, one at zero and the other at 1 outside directorships in addition to the cut-off at three. The results are shown table 4.4.

Table 4.4

Only coefficients for the busy director dummy variables are reported.

Overall statistics and other explanatory variables’ coefficient values and their significance remain highly similar to the results in table 4.3. Table 4.4 shows that the dummy variables replacing the network variable are only significant at the cut-off of three other directorships. The sign and significance is the same as for the network variable in the three or more directorships category. Here, the APE is -0.35 for both stock return and re-turn on assets for the busy director specification, compared to -0.23 for the

network variable. Thus, a board filled with busy directors is less likely to dismiss the CEO, and the conclusion is very much in line with the results for the network variable. This confirms the Fich and Shivdasani (2006) result that really busy directors are less likely to terminate the CEO’s con-tract.

Together, the joint board control type finds support in the regressions on CEO turnover. The conclusion rests on the confirmation of the simul-taneity of CEO turnover with board turbulence joint control type predic-tion, and it is further confirmed by the entrenchment variables’ low sig-nificance or reverse sign. Thus, the CEO and the board work as a team, as modelled in Adams and Ferreira (2007). The conclusion is confirmed in re-gressions with definitions of CEO turnover ranging from the all turnover to forced turnover, and with definitions of firm performance being either the stock return or ROA.

In document Agradecimientos xi (página 110-117)