Almost two decades of research on the relationship between socioeconomic variables and the incarceration rate has produced divergent results. Some of these inconsistencies may be due to the various models specified (some use total crime rather than violent crime, and some control for system-level variables while others do not). Virtually all of the previous research has focused on the direct effect only. In the current study, models that use arrest rate data rather than crime rates were developed to capture the effect of the recent "war on drugs" on imprisonment, which is not included in the Index Crime rate data. These data show that the percentage of nonwhite defendants processed through the courts had a significant and direct effect on imprisonment levels across the Nation, but less of a direct effect in nonsouthern States. The data also show that percentage of nonwhite defendants had a substantial indirect effect. The indirect effect of economic inequality was greater than the direct effect. Finally, there was no variation in the effects of extralegal variables between those models that controlled for system level data and those that did not. 4 tables and a 48-item bibliography