Social Risk and Attribution: How Considering the Social Risk of Attributions Can Improve the Performance of Kelley's ANOVA Model in Applied Research

Abstract

Classic models of attribution are increasingly used, despite serious problems with their empirical validation. This study revisits Kelley's (1967) ANOVA model of attribution and argues that it will most usefully predict attributions when attributional processes are socially “safe” and have few social consequences. The results demonstrate that attributions are most likely to be inconsistent with Kelley's predictions when attributional information and the attributions themselves are socially consequential or risky, but are more likely to be made as predicted when they are socially safe. Applications of Kelley's model, therefore, should pay attention to the extent to which attributions and attributional information are socially consequential or risky, particularly when analyzing the use of consensus information.