This Article uncovers a critical disjuncture in our system of providing affordable rental housing. At the federal level, the oldest, fiercest debate in low-income housing policy is between project-based and tenant-based subsidies: should the government help build new affordable housing projects or help renters afford homes on the private market? But at the state and local level, it is as if this debate never took place. The federal government (following most experts) employs both strategies, embracing tenant-based assistance as more cost-effective and offering tenants greater choice and mobility. But this Article shows that state and local housing voucher programs are rare, small, and limited to special populations. States and cities almost exclusively provide project-based rental assistance. They move in lockstep despite disparate market conditions and political demands: project-based spending overwhelmingly predominates in states that are liberal or conservative and high- or low-rent. States have done so across decades of increased spending. This uniform subnational approach suggests an unhealthy federalism: neither efficient nor experimental. This Article further diagnoses why states have made this unusual choice, identifying four primary culprits: 1) fiscally-constrained states use project-based models to minimize painful cuts during recessions; 2) incomplete federal housing subsidies inadvertently incentivize project-based spending; 3) the interest groups involved in financing and constructing affordable housing are relatively more powerful subnationally; and 4) rental assistance's unusual, lottery-like nature elevates the value of visible spending over cost-effectiveness. Finally, this Article points a path towards reform. Taking a federalist perspective allows for a new understanding of federal housing statutes. Better cooperative models could accept states' limitations in providing rental assistance--and exploit their strengths.