The Neglect of Base Rate Data by Human Resources Managers in Employee Selection

Article excerpt


This study investigated the use of base rate information in decision making by individual and groups of human resources (HR) managers. A 2 x 2 (base rate x decision entity) mixed factorial design was used. Data were collected from 91 managers who were subsequently placed into 30 groups. The managers were presented with a written scenario describing a selection decision. The scenario contained both the prior probability that a job candidate had partnership potential and the candidate's score from a structured selection interview. Analysis of variance (ANOVA) confirmed that HR managers, both individually and collectively, were insensitive to the base rate data and relied too heavily on the interview scores. The HR managers as a result made inaccurate judgments about a job candidate's potential. Implications for the decision-making and selection literatures are discussed.


Cette etude a pour objet l'emploi de taux de probabilite de base (tablis sur un ichantillonage prealable) dans les prises de decision par des gestionnaires en ressources humaines, pour des decisions effectuees tant sur un plan individuel que collectif. L'etude emploie un plan experimental factoriel de 2 x 2 (probabilite de base x agents) de decision). Les donnees furent recueillies aupres de 91 gestionnaires, lesquels furent ensuite repartis en 30 groupes. Une situation fictive decrivant un choix d'embauche leur fut presente. Le cas fictif indiquait la probabilite qu'avait un candidat de devenir eventuellement un associe d'une societe, de meme que le score du candidat dans une entrevue de selection evaluative. L'analyse de la variance (Angl. ANOVA) confirme que les gestionnaires en ressources humaines, individuellement de meme qu'en groupe, privilegiaient les resultats de Ventrevue aux depens des probabilites de base. Par consequent, les gestionnaires evaluaient asset imprecisement le potentiel du candidat. L'auteur discute ensuite les consequences de ces resultats pour les prises de decision en general, ainsi que pour l'etude des strategies devaluation du personnel.

What is the probability or likelihood that a particular object, event, or person belongs to a specific class or category? Examples of such judgments under uncertainty include determinations about the likelihood that (a) a new product will succeed in the marketplace, (b) a given project will be completed on time or on budget, or (c) a particular job candidate will flourish if hired. According to Bayes rule (Press, 1989), these judgments appropriately involve considerations of both distributional and individuating information (Kahneman & Tversky, 1972). The former type of information refers to the base rate frequency or prior probability of class or category membership. The latter type refers to data that are descriptive of and specific to the particular entity under evaluation.

Considerable research has focused on the processes by which individuals make judgments under uncertainty. Much of this research concerns heuristics such as representativeness, anchoring and adjustment, and availability. According to the representativeness heuristic, people when given individuating information assess the probability that a person, event, or object belongs to a particular category or class on the basis of the extent to which the person, event, or object is perceived to be similar to or representative of the category or class. Reliance on the representativeness heuristic, however, implies both the underutilization of base rate data in favour of individuating information and the violation of Bayes rule (Tversky & Kahneman, 1974).

In one type of problem that has received considerable research attention (e.g., Argote, Devadas, & Melone, 1990; Argote, Seabright, & Dyer, 1986; Kahneman & Tversky, 1972), base rate information is provided to subjects regarding how many members of two classifications, engineers and lawyers, are in a group. …