This paper argues that the two most serious structural shortcomings of the current system of low-income housing assistance are (1) its excessive reliance on unit-based assistance and (2) its failure to provide housing assistance to all of the poorest eligible families who ask for help. Evidence on the performance of housing programs indicates that unit-based assistance has a much greater cost than tenant-based assistance for providing equally good housing, and it needlessly restricts recipient choice. Unit-based assistance has no advantage over tenant-based assistance to offset these disadvantages. The nonentitlement nature of the current system is inconsistent with plausible assumptions about taxpayer preferences. The paper argues for a transition to an entitlement housing assistance program that relies exclusively on tenant-based assistance. It describes concrete actions that would achieve this result without spending additional money, and it shows that the major objections to these proposals are inconsistent with the evidence on program performance. The proposed transition would benefit most current recipients of housing assistance, and the reforms would give those taxpayers who want to help low-income families with their housing more for their money. After the transition is complete, millions of additional families would receive housing assistance that enables them to occupy better housing in nicer neighborhoods, and to consume more of other goods. Millions of other families that would have received unit-based assistance with the continuation of the current system would live in housing, neighborhoods, and locations that they prefer to their units in subsidized projects.