In this paper we generalize the allocative efficiency test of Banker et al. (2007) to the situation where relative input prices vary across firms. It is shown that when firms face different input prices, using actual expenses for measuring aggregated inputs (as is a common practice) would lead to misleading conclusions about allocative efficiency. We describe the appropriate procedure for constructing the aggregate input when the relative prices are known and they vary across firms. The parallel between input aggregation in DEA and parameter restriction in regression models is also pointed out. The proposed aggregation procedure needs to be followed, irrespective of whether the subsequent statistical test is parametric, standard nonparametric, or based on bootstrap.