Among the major trends in the higher education (HE) sector, the development of rankings as a policy and managerial tool is of particular relevance. However, despite the diffusion of these instruments, it is still not clear how they relate with traditional performance measures, like unit costs and efficiency scores. In this paper, we estimate a variety of models to evaluate costs in US higher education institutions. A particularly innovative feature of our approach involves the estimation of latent class and random parameter stochastic frontier models of a multiproduct cost function: this allows us fully to accommodate both the heterogeneity across institutions and the presence of technical inefficiencies. Such methodological strategy is essential in analyzing the US HE system, which is characterized by a strong internal differentiation. Our main findings are two. First, on a public policy ground, the estimates suggest that global economies could be achieved by effecting a reduction in the number of institutions providing undergraduate instruction, while increasing the number of institutions engaged in postgraduate activity. Second, the current existing rankings turn out as coherent with ratings provided by the calculation of efficiency scores.