Model Misspecification in the Sticky Cost Literature

Abstract

Beginning with Anderson et al., multiple studies have used a common model to investigate "sticky costs." This model regresses the log change in SG&A on the log change in revenues. However, Balakrishnan et al. assert that the finding of sticky costs is the result of model misspecification, and using an alternative model that regresses the change in costs scaled by lagged revenues on the changes in revenues scaled by lagged revenues, they find no evidence of sticky costs. I assert that their model also suffers from misspecification, and I propose a new model for measuring sticky costs that addresses misspecification in both prior models. Using this model, I again find evidence of sticky costs.