Monitoring Cost Changes with Log-Linear Cost Models: Lessons from a Case Study

When studying Worker's Compensation (WC) claim cost experience, researchers often prefer models that relate claim characteristics and other cost drivers to the logarithm of the claim cost, rather than to the dollar cost itself. Linear models based directly on dollars, however, are better suited to decomposing the differences in costs observed over time or between claim populations. Reconciling the two methods within one analysis can be awkward. This led us to a new perspective: one that enables the two approaches to work together while preserving the most desirable features of each. The paper presents a general method for analyzing cost differences. It also illustrates the method in the context from whence it came: monitoring the post-reform experience of WC claim costs. Keywords: Workers' Compensation Insurance, reform, Oxacca decomposition, log-linear model, log-log model, exponential weight.