Team Performance: Nature and Antecedents of Nonnormal Distributions

Abstract

Team research typically assumes that team performance is normally distributed: teams cluster around average performance, performance variability is not substantial, and few teams inhabit the upper range of the distribution. Ironically, although most team research and methodological practices rely on the normality assumption, many theories actually imply nonnormality (e.g., performance spirals, team composition, team learning, punctuated equilibrium). Accordingly, we investigated the nature and antecedents of team performance distributions by relying on 274 performance distributions including 200,825 teams (e.g., sports, politics, firefighters, information technology, customer service) and more than 500,000 workers. First, regarding their overall nature, only 11% of the distributions were normal, star teams are much more prevalent than predicted by normality, the power law with an exponential cutoff is the most dominant distribution among nonnormal distributions (i.e., 73%), and incremental differentiation (i.e., differential performance trajectories across teams) is the best explanation for the emergence of these distributions. Second, this conclusion remained unchanged after examining theory-based boundary conditions (i.e., tournament versus nontournament contexts, performance as aggregation of individual-level performance versus performance as a team-level construct, performance assessed with versus without a hard left-tail zero, and more versus less sample homogeneity). Third, we used the team learning curve literature as a conceptual framework to test hypotheses and found that authority differentiation and lower temporal stability are associated with distributions with larger performance variability (i.e., a greater proportion of star teams). We discuss implications for existing theory, future research directions, and methodological practices (e.g., need to check for nonnormality, Bayesian analysis, outlier management).