Each year, over 1.5 million Americans rely on homeless programs for overnight shelter. De- spite robust federal funding for this critical part of the social safety net, more than 200,000 remain unsheltered on any given night. In this paper, I quantify behavioral responses to pro- gram generosity to study the tradeoffs inherent in expanding homeless assistance. I utilize a new, national dataset on sheltered and unsheltered homeless populations and exploit differential distribution of federal homeless assistance grants across communities. An outdated formula sets each region's funding eligibility, inadvertently generating exogenous variation in homeless assistance. Program providers use the resulting marginal funds to add beds in both individual and family programs. Homeless individuals and families, however, have very different characteristics and behavioral patterns. I find that greater individual program generosity reduces unsheltered homelessness without drawing others into the local homeless population. A permanent $100,000 annual increase in homeless assistance decreases the size of the unsheltered population by 35 individuals, and all of the individuals who utilize marginal beds would otherwise be unsheltered. The effects of family program expansions are quite different. More generous funding helps house otherwise unsheltered families while also drawing in a larger homeless family population (73 additional people in families for every $100,000). I show that this increase is primarily driven by homeless families migrating to communities with greater funding. These results illuminate the policy responsiveness of homeless populations and shed light on the efficacy of homeless assistance funding.