The landmark United States healthcare reform law – the Affordable Care Act – provides an opportunity to study the dynamics of implementation for complex, politically contentious policies. Matland’s Ambiguity-Conflict Model suggests that bottom-up models will dominate in such cases. I exploit variation across states in the implementation of online health insurance marketplaces to test whether the federal- (top-down) or state-managed (bottom-up) implementation model produced better outcomes. Specifically, the study examines if state, federal or partnership exchanges were most effective at offering generous plans for consumers based on premiums, deductibles and copayments in 2014, the first year of operation. The results unambiguously indicate that state exchanges were most successful. The findings provide evidence for what Matland suspected – that bottom-up models, by providing more discretion to local implementers to adapt to contexts and build coalitions, are superior for high-conflict, high-ambiguity policies.