Researchers using fiscal choice models have had limited success predicting fund diversion in federal grant programs. The application of a principal-agent framework to questions of fiscal federalism offered a potentially valuable alternative approach, but the traditional model employed by Chubb (1985) neglected potential variability in the degree of goal conflict between principals and agents. This article proposes an expanded framework, which incorporates the possibility of variation in goal conflict between participants in intergovernmental aid programs. The theory suggests that the level of policy congruence between recipient jurisdictions and the national government will determine the amount of grant funding diverted away from targeted policy areas. Findings from analyses of grant programs in two distinct policy areas support the hypothesis that grant effectiveness is partially a function of goal congruence. The relationship between intergovernmental partners is interactive, with the degree of policy agreement determining fund diversion in subnational jurisdictions, as well as the effectiveness of federal oversight activities. The findings have important theoretical implications for understanding both fiscal federalism and principal-agent relationships more generally.