882 JOURNAL OF POLITICAL TABLE I MANAGER SELECTION EFFECTS Sample Size Model I: Yi=a+yXi+ u Teri (2.014 Security deposit 395,4* Model II: P[;=1X,]=G(a+y'Xi 645 Demotion NoTE-See sec. B of the Appendix for detailed specifcations. Xu is relative hrm performance. for our degrees of freedom is 2.71: for a 5 percent level, 3.84: for a I percent leve,6.6/cen level . Significant at the I percent level characteristics of the new managers' contracts and the fate of the previous manager on the one hand and the performance of the firm prior to the new contract on the other. Specifically, we considered four dependent variables: (1)the term of the new managerial con- tract,(2)the size of the security deposit, (3)the use(or not) of an auction for selecting the firm,s top management position, and(4) the demotion(or not) of the firms previous manager. Each of these variables was regressed on a measure of relative firm performance over the period immediately prior to the new contract. (See secs. B and C of the Appendix for a description of the estimated models and procedures. )The results for all four dependent variables are given in table 1, 6 Table I shows the effect on managerial choice of poor performance (measured by the firm's output per worker relative to the industry) hat is observable ex ante. Low performance of the firm prior to a change in manager is significant in explaining the use of an auction select a new manager. Not only were poorly performing firms disproportionately subject to auctions, they were also associated with a larger security deposit and also with a shorter management con- tract. Thus the industrial bureaus were more willing to allow poorly ample period and for which the of the new manager's contract