Developing more accurate costing systems
For decades, business schools taught costing heuristics to management accounting students. But those “rules of thumb” went untested until accounting professor Eva Labro took on the task. Her findings give practical guidance on developing accurate costing systems.
“We’re testing heuristics that people have simply taken to be true before,” Labro says. “Management accountants have to make down-to-earth decisions when they design costing systems. They want to provide the most accurate information in a cost-efficient way. The reaction I get to my research is that this is the kind of thing they are waiting to learn.”
To offer ways to translate vague guidance into implementable methods, Labro collaborated with Ramji Balakrishnan of the University of Iowa and Stephen Hansen of George Washington University. They published their findings in “Evaluating Heuristics Used When Designing Product Costing Systems” in Management Science.
Getting the product costs “right” is important to firms, Labro says. Companies allocate fixed costs to products to:
Labro’s research focuses on the product-costing role because businesses make product-planning, capacity planning and pricing decisions based on product costs. Such product-cost approximations based on allocations inevitably produce estimates that differ from a product’s true marginal cost.
Cost systems usually comprise two stages in which the company allocates: resource costs to form “cost pools” and costs from the cost pools to products. Management accounts can group resources in cost pools by their size (size-based) or by how “like” they are (correlation-based).
To test the rules of thumb taught about costing systems, Labro ran computer simulations of a large variety of true production environments, called benchmark systems, to learn when certain heuristics perform well.
She varied the parameters of the environment, such as whether costs were dispersed across many costs pools or concentrated in a few large ones. Then she simulated costing system approximations of the true benchmark that follow the rules of thumb under study. She calibrated the simulations using data from case studies. “Running the simulations produced generalizable results, so we can teach our students in a much more refined way when these heuristics will hold and when they won’t,” Labro said.
Labro’s simulations revealed that different environments call for different rules. Size-based rules work best for companies with more than 70 percent of their costs falling into 20 percent of their cost pools. For instance, a firm in the service sector where staff resources make up a huge part of total cost would do well to follow size-based rules.
In contrast, a distribution firm would be better off following correlation-based rules. It would have a variety of products brought in from different vendors and use multiple resources (such as packaging materials, vehicles and gas) without one standing out as the main resource.
Labro’s research takes into account the information demands of different heuristics. Implementing a size-based rule requires data only on resource costs (usually available in accounting records). Correlation-based rules, in contrast, are information intensive because they require information on resource consumption patterns — information that might be costly, if not impossible, to collect.
However, Labro found that correlation-based rules perform well even when the precision of available correlation information is low. Crude estimates of correlations in consumption patterns (for example, merely knowing whether the correlation is greater than 0.4) appear to be sufficient to implement correlation-based rules effectively.
After separating out the highest cost categories, management accountants need to decide what to do with the remaining small value resources. Sometimes a small-value resource is placed in the same pool as a large-value resource, based on the rule of thumb that both resources are consumed in a similar way by product. For example, labor supervision might be sufficiently similar to labor to be pooled in that category. Alternatively, it could remain in a miscellaneous cost pool together with other small-value resources.
Labro found that management accountants worried unnecessarily about having a large miscellaneous pool of small-value resources. Her simulations revealed that having a large miscellaneous overhead cost pool outperformed adding small resources to the larger pools (as when adding labor supervision to the labor cost pool), even when the miscellaneous overhead accounted for 50 percent of the total costs. In addition, pooling all small-value resources together in a miscellaneous pool is considerably less time- and labor-intensive than sorting through the miscellaneous pool and re-categorizing the small-value resources.
That finding alone can save businesses significant spending on consulting. “More information always seems better, but it comes at a cost,” Labro says. “It’s much harder to figure out which resource could potentially go with which large pool rather than throwing them all together. Consultants charge a lot of money to do those kinds of detailed costing-system designs. But it turns out there are some simple ways of designing systems that are pretty accurate.”
“If a simple heuristic performs better or not a lot worse than something highly complicated and costly,” Labro said, “then it may be better to go with the simple heuristic.”
Labro’s next project is to determine how often costing systems need to be updated. A costing system is a snapshot of what resource consumption looks like at a point in time. But management decisions about reducing costs and altering the product or service mix turn product costs into a continually evolving picture. Making updates costs time, talent and, often, consulting fees. Labro will look at which changes are important enough to warrant an update and how much impact a minor change has on the information used to make management decisions.
Labro wants to make sure that when her students learn a rule of thumb that they can count on its accuracy.