The Non-Normality of Firm Performance and the Consequences for Strategic Management Theory

Abstract

Firm performance represents an integral outcome for most strategic management scholarship and corresponding theoretical frameworks (e.g., agency theory, RBV, competitive dynamics, and many more). Despite the inherent idea across strategy research that firm performance is heterogeneous, we have little knowledge about just how heterogeneous it is across samples of firms and how this might impact empirical tests of theories used by strategy scholars. We investigate performance heterogeneity by examining the distributions of firm performance measures (e.g., ROA, ROS, EPS, etc.). We find that extreme levels of both skewness and kurtosis vary substantively across different measures, samples, transformations, and years. We create simulations to mimic these distributions and find that such non-normality negatively impacts the efficiency of OLS, robust regression, and OLS with winsorized values. In contrast, we illustrate that quantile regression is more appropriate for modeling dependent variables following these types of extremely non-normal distributions and represents an attractive approach for researchers examining firm performance. The primary implication of our research is that the extreme non-normality of performance measures makes it difficult for researchers to find support for theoretical frameworks when using models that focus on average relationships (e.g., OLS, multilevel models, etc.). Instead, we suggest strategy theories could benefit from re-examining theorized relationships using the types of non-parametric techniques that are more appropriate for the non-normal distributions of firm performance.