The Accuracy of Proportional Cost Models: Evidence from Hospital Service Departments

Using data from hospitals in the state of Washington, we examine the time-series behavior of overhead costs. We find that more accurate predictions of changes in costs are usually generated by assuming a cost will not change at all (except for inflation) than by assuming that the cost will change in proportion to changes in activity. We also find that nearly all of the effect of a change in activity on costs appears to occur in the same year as the change in activity. Finally, using a multi-period regression model we find that the proportion of variable costs in the hospital overhead accounts is apparently very modest. These results suggest that costing systems, such as activity-based costing, that assume costs are proportional to activity, will grossly overstate relevant (i.e., incremental) overhead costs for decision-making and performance evaluation purposes.