Evaluations of complex interventions such as sentencing guidelines provide an opportunity to understand the mechanisms by which policies and programs can impact intermediate and long-term outcomes. There is limited previous discussion of the underlying frameworks by which sentencing guidelines can impact outcomes such as crime rates. Guided by a realist evaluation framework, this article examines the impact of linkages of sentencing policy to resource capacity--a cost-control paradigm under which a few states created guidelines to control rising prison populations and expenditures. Additionally, we argue that the moderating influence of this linkage will depend on the severity of the crime. A key conclusion is that in addition to social science theory, evaluation theory is needed to understand how programs work; there is a greater need for identifying conditions under which policies work or do not work. We find the realist approach as a promising approach to build such knowledge. (Contains 10 tables, 2 figures and 7 notes.)