The age-turnover relation documented by the data of table 1 supports the proposition that provisions for mandatory retirement at age 65 affect the process of management change for chief executive officers. The implications of this proposition are twofold. First, it is inconsistent with the proposition that boards of directors can be controlled by a firm's CEO. Those who advance this hypothesis must explain why omnipotent chief executive officers are incapable of overcoming the constraints imposed by mandatory retirement provisions.
Second, the apparent importance of mandatory retirement provisions has implications for empirical testing of propositions about the process of corporate control. Anticipated retirement combined with its associated compensation effects complicates tests for statistical dependence between stock price performance and the rate of change of salary plus bonus. Furthermore, the occurrence of turnover associated with mandatory retirement at age 65 complicates efforts to test for a relation between stock price performance and changes in top management. To avoid the effects of such complications, we divide our test sample into two components. One component consists of CEO's who are at least 64 years old and the other consists of individuals in younger cohorts. Statistical problems associated with the process of mandatory retirement and its associated compensation effects will complicate any efforts to test hypotheses about corporate control in the older cohorts. Such complications, however, should be less severe in the tests conducted using the younger portion of the sample.
The size and pay characteristics of the two subsamples are described in table 2. The sample containing observations on chief executives who are less than 64 years old in the performance year is approximately four times larger than the sample with CEO's in the older cohorts. The younger sample is derived from the records of 218 corporations, while 58 firms are represented in the older sample. The individual performance year samples of older executives are sufficiently small to preclude meaningful tests. Observations for a given CEO for a particular firm can appear in both of the age group samples. For instance, consider an individual who is 62 years old in the 1978 performance year. Observations based on the 1978 and 1979 performance years would be included in the sample of younger individuals. The observation based on the 1980 performance year would be included in the older sample. There are no significant differences in the compensation means for the two samples, and stock price performance is not related to age of the CEO.