We show that, in a sufficiently large population satisfying certain statistical regularities, it is often possible to accurately estimate the utilitarian social welfare function, even if we only have very noisy data about individual utility functions and interpersonal utility comparisons. In particular, we show that it is often possible to identify an optimal or close-to-optimal utilitarian social choice using voting rules such as the Borda rule, approval voting, relative utilitarianism, or any Condorcetconsistent rule.