Many critical goods and services in modern-day economies are produced and distributed through complex institutional arrangements. Agent-based computational economics (ACE) modeling tools are capable of handling this degree of complexity. In concrete support of this claim, this study presents an ACE test bed designed to permit the exploratory study of restructured U.S. wholesale power markets with transmission grid congestion managed by locational marginal prices (LMPs). Illustrative findings are presented showing how spatial LMP cross-correlation patterns vary systematically in response to changes in the price responsiveness of wholesale power demand when wholesale power sellers have learning capabilities. These findings highlight several distinctive features of ACE modeling: namely, an emphasis on process rather than on equilibrium; an ability to capture complicated structural, institutional, and behavioral real-world aspects (micro-validation); and an ability to study the effects of changes in these aspects on spatial and temporal outcome distributions.