Empirical models of commodity prices are potentially important aids to decision-makers, especially as the economy has grown more complex. A typical time series of commodity prices exhibits positive autocorrelation, occasional spikes, and random variability, and conceptual models have been developed to explain this behavior. But, the leap from theory to empirical applications is large because of model specification and data quality problems. When modeling price expectations, for example, should a price series be deflated and if so, by what deflator? The choice can have a large effect on empirical results. Nonetheless, it is possible in some applications to obtain relatively stable estimates of structural parameters that are useful for addressing specific problems. This may not happen often, however, because the incentives in academia do not encourage rigorous, in-depth appraisals of empirical results.