Career Dynamics Under Uncertainty: Estimating the Value of Firm Experimentation

IZA Logo
   

IZA Seminar

Place: Schaumburg-Lippe-Str. 9, 53113 Bonn

Date: 13.06.2006, 12:15 - 13:30

   

Presentation by 

Elena Pastorino (University of Minnesota)
   

Abstract:

This paper develops and structurally estimates a dynamic learning model in which a firm can acquire information about a workerís ability by observing his performance on the job. Ability determines both the profitability of a job and the job-dependent distribution of output signals. As a result, the informativeness of performance about individual productivity varies in the job a worker performs. Because of the trade-o between learning and short-run profit maximization, the firmís optimal information acquisition strategy is the solution to an experimentation problem, a multi-armed Bandit problem with dependent and independent arms. Under the firmís optimal employment policy, a worker is assigned to jobs of decreasing degree of informativeness, as measured by the dispersion in posterior beliefs. The purpose of the analysis is to investigate to what extent uncertainty about ability a ects the dynamic pattern of a workerís transition across jobs within a firm, i.e., the timing and job characteristics of a career. To this end, the model is structurally estimated using longitudinal data from a single U.S. firm, on the cohorts of managers entering the firm at the lowest managerial level between 1970 and 1979. Estimation results con- firm that a theoretically restricted learning model can succeed in fitting the dynamic profile of the probability of retention and promotion at the major job positions within the firm. The estimated model is then used to compute the firmís value of information and to evaluate the e ect on this value, the pattern of job assignments, and of turnover of (i) changes in the discount rate, which reflect changes in market interest rates, and (ii) alternative information structures.

   
Download complete paper   
   
For more information, please contact seminar@iza.org