In a careful and thorough empirical study, Christopher Udry (1996) shows convincingly that, in a large sample of West African households, household resource allocations were not Pareto efficient. This paper argues that observation of the Pareto inefficiency of a household resource allocation does not however refute the hypothesis that it chooses this resource allocation as if it maximises some form of household welfare function possessing the Pareto property. To refute that hypothesis it is necessary to show that the observed allocation does not represent a second best optimum. For this it will be necessary to show that the estimated parameters of the model lie in a region of the parameter space for which the second best optimality of the allocation does not hold.
empirical test Pareto efficiency household welfare