This study tracks American states' policy choices under the No Child Left Behind Act and explores their consequences for student achievement. Using the path analysis of relationships among state-level policy input, context, and outcome variables, the study portrays a Halloween-like "trick-or-treating" game between the federal and state governments in the new ecology of the test-driven education accountability system. States that chose the "trick" path with a calculative policy negotiation and manipulation strategy made significant gains on their own state assessments but not on the national assessment. In contrast, states that followed the "treat" path with a faithful policy implementation for funding strategy have not yet brought about significant gains on either the national or state assessments. The first-generation accountability states with a prior history of high-stakes testing tended to employ both strategies at the same time. However, neither effective illusion nor ineffective implementation serves the goal of long-term, sustainable academic improvement. Implications for research and policy are discussed. (Contains 3 figures and 7 notes.)