Learning from Coarse Information: Biased Contests and Career Profiles

Abstract

An organization's promotion decision between two workers is modelled as a problem of boundedly-rational learning about ability. The decision-maker can bias noisy rank-order contests sequentially, thereby changing the information they convey. The optimal final-period bias favours the leader, reinforcing his likely ability advantage. When optimally biased rank-order information is a sufficient statistic for cardinal information, the leader is favoured in every period. In other environments, bias in early periods may (i) favour the early loser, (ii) be optimal even when the workers are equally rated, and (iii) reduce the favoured worker's promotion chances. This paper presents a model of boundedly-rational learning by an organization. The aims are to represent in a simple form the costs of gathering, processing, or transmitting information and to analyse their implications for the procedures selected to handle information, as well as for the decisions actually made. Like other recent work on the internal organization of firms, this work explores to what extent organizational structure and behaviour, as well as organizational performance, can be explained by adaptation to the costs associated with information.' In the learning model we formulate, the decision-maker not only selects a rule specifying what action to take as a function of his observations but also chooses, sequentially, the information partitions generating those observations. This general problem is interpreted in the context of an organization which needs to make an important promotion decision and which attempts to maximize the value of the information generated during the observation periods that precede the decision. One of the motivations for this analysis was the finding by organizational sociologists that earnings and promotions in the later stages of a worker's career are strongly correlated with earnings and promotions in the early stages (Kanter (1977), Rosenbaum (1984)). Later success is positively associated with early success, even when one controls for the effect of observable characteristics likely to affect performance, such as education. One obvious source of this correlation is differences in ability that persist over time and that are not captured by the observable covariates. In this paper, we focus on differences in ability which are initially unobservable by organizations and their workers, as well as by researchers. We present a simple model of an organization trying to learn about differences in ability and show