Optimal Probation for new Hires

Abstract

This paper examines the use of a probation period for new hires. Probation has traditionally been viewed as a period during which the firm discovers something about a new hire's abilities. The firm follows a Bayesian decision procedure. When its estimate of the new hire's abilities hits an upper critical level it hires the worker into a permanent job, while if its estimate drops below a critical level it fires the worker. While the actual length of probation is random, its expected length is well defined. We propose an alternative explanation of probation periods. In order to explain the use of a probation period one must first look at the principal-agent problem the firm will face after probation has been completed. If we treat the information set in that problem as a choice variable of the firm, then the extent of monitoring will be endogenously determined. If the optimal level of monitoring is less than continuous, then a worker shirking in any period will be detected with a probability less than one. Given less than complete monitoring the firm has an incentive to choose workers with low propensities to shirk (low disutility of effort) and so must develop an appropriate hiring policy. Unfortunately, the propensity to shirk can be hidden by the worker and so cannot be tested for. To overcome this problem the firm chooses to monitor the new worker continously for a period of time after the initial hiring, i.e. to use a probation period. During this period bad workers cannot shirk and must incur the cost of hiding their propensity to shirk. The probation period is chosen to be just long enough to deter bad workers from joining the firm. In short, we view the probation period not as an exercise in Bayesian learning, but rather as an attempt by the firm to screen bad workers in a setting in which such workers can, at a cost, hide their type. In this sense the model is analogous to those of performance bonds. Here, however, the firm incurs a cost (additional to normal monitoring costs) of bonding. This approach to probation relies on the relevant characteristics of * Tedeschi's work was financed by grants from the C. V Starr Center for Applied Economics at New York University and from CNR (Consiglio Nazionale per le Ricerche). Bull's work was financed by NSF grant SES 8409276. The authors wish to thank an anonymous referee. The responsibility for all errors remains that of the authors.