A Comparison of the Accuracy of ABC and Time-Driven ABC in a Numerical Experiment

Abstract

Kaplan and Anderson (2004, 2007) developed time-driven activity-based costing (TDABC) to provide a simpler system that requires less computation than ABC. However, TDABC is new and the relationship between ABC and TDABC needs to be determined. Our paper clarifies this relationship by showing how the two allocation systems differ, the settings in which each is most accurate, and when the error difference between the two is maximized (minimized). Following Labro and Vanhoucke (2007, 2008), we model costing systems as containing two stages. Stage I maps the resources to activities, while Stage II maps activities to cost objects (products). We find that the key difference between ABC and TDABC is in how the two systems aggregate the resource to activity matrix; that is, how they aggregate the Stage I information. ABC uses resource cost drivers to aggregate first stage information by resource columns. In contrast, TDABC implicitly aggregates first stage information by activity rows. Because of this distinction, ABC and TDABC systems are non-comparable in almost all settings. We ran a numerical experiment to discover where each system performs best. Our experiment starts with the same underlying information for both systems, adds noise, and then uses similar heuristics to generate ABC and TDABC costing systems. We find that the errors in the ABC and TDABC systems are driven by the same underlying characteristics. In particular, errors increase when resources are less traceable, while errors decrease when activities are less traceable. Comparing the two systems, shows that TDABC (ABC) performs better than ABC (TDABC) when both resources and activities are more (less) traceable.