This paper seeks to determine whether the underlying assumption of utility maximization can be empirically validated for the local public sector. In addition, three alternative functional forms of a system of demands are compared on the basis of their theoretical restrictions as well as their predictive performance. Based upon a likelihood ratio test we find that the restrictions implied by utility maximization are too restrictive to be consistent with the unrestricted estimates. Once integrability restrictions are imposed, however, both CES and Cobb-Douglas utility functions are consistent with the data. The unrestricted estimates had the best predictive performance based on information inaccuracy.