Cost hierarchies and the pattern of product cost cross-subsidization: Extending a computational model of costing system design

Abstract

Cost information is critical to ease managers’ decisions in daily business, but its provision is informationally demanding and error prone. Effective design choices for costing systems that can reduce errors are the subject of a growing body of research. The computational model by Anand, Balakrishnan, and Labro (2019) collates previous research in a unifying framework, turning it into a potential standard for future studies. This paper uses this framework and aims to investigate the mechanism behind the well-documented empirical pattern of product cost cross-subsidization in a large-scale simulation experiment. According to this pattern, volume-based costing systems bias the costs of high-volume products upward and of low-volume products downward. Although this pattern has important implications for firms and is discussed extensively in the literature, it has not yet been investigated with computational models. As the first objective of this paper, we replicate the original model by following a pattern-oriented model replication approach. The second objective is to study the mechanism behind the pattern of product cost cross-subsidization. We are unable to reproduce it systematically with the original model. However, the pattern emerges when we extend the model to include a simple cost hierarchy with distinct resource consumption types and volume-based cost drivers. This allows us to specify the likely mechanism behind it. Building on these results, we further extend the model with empirical and theory-based ABC cost hierarchies and assess their effect on product cost cross-subsidization. Our results suggest that production environments underpin more diverse cost hierarchies in practice than previously implemented in the model. Overall, we argue that our extension provides relevant insights into the pattern of product cost cross-subsidization, while our replication and extension strengthen the models’ credibility and usability for future research.