Nonsurvey input—output models, especially those constructed by means of pool and quotient techniques, are suspect when applied to the case of cross-hauling. This is well known. But the US Department of Agriculture (USDA) Forest Service, through its IMPLAN model, and the US Department of Commerce (USDC), through its RIMSII model, encourage the application of pool and quotient nonsurvey models in cases where cross-hauling is likely to occur. In a study of the timber economy of the West-central Idaho Highlands we show the error caused by ignoring cross-hauling in estimates of logger—sawmill trade. We argue that pool and quotient techniques, used in nonsurvey models such as IMPLAN and RIMSII, should not be applied to a single county situation, or to any aggregation of counties that is not, in some sense, a functional economic area. This applies in many cases to full state models. We close by noting that in many applications of pool and quotient nonsurvey input—output methods, technique may be substituting for thought.