Avoiding the “Fallacy of the Wrong Level”

Abstract

This article illustrates that any bivariate or multiple regression provides inadequate information about the levels of analysis in a data set collected in an organizational setting. As a result, individual-level effects may be incorrectly attributed to the group level, and group-level effects may be incorrectly viewed as being solely individual-level effects. Both of these situations are examples of the “fallacy of the wrong level.” Within and between analysis (WABA) allows levels of analysis to be tested in data. These WABA tests are useful in numerous analytical approaches, including structural equation modeling, hierarchical linear modeling, and various approaches to aggregation. This article provides a decision tree for use in performing tests for multiple alternative levels of analysis in a data set collected in organizations.