In advanced democracies, most government spending is allocated according to criteria approved by a legislature but implemented by the bureaucracy. I ask whether this fact imposes a binding constraint on the ability of legislators to engage in targeted redistribution, by constructing a model in which legislators are constrained to allocate spending by a formula of limited dimension—in contrast to benchmark models where proposers have the flexibility to manipulate the payoffs of individual members directly. The model predicts oversized winning coalitions, positive distributions outside of the winning coalition, and the emergence of persistent voting blocs. I then apply the model to a sample of 31 US federal spending bills, using new data connecting spending outcomes to authorizing legislation. I find that most allocation formulas for spending programs involve five or fewer factors. Formulaic allocation imposes a tight constraint on targeting, eliminating more than 90% of congressional proposers’ degrees of freedom.