Résumé : We consider the computational problem of testing whether observed household consumption behavior satisfies the Collective Axiom of Revealed Preferences (CARP). We propose a graph such that the existence of a node-partitioning giving rise to two induced subgraphs that are acyclic implies that the data satisfy CARP. Furthermore, we propose and implement heuristics that are quite fast, that can be used to check reasonably large datasets for CARP and that can be of particular interest when used prior to computationally demanding approaches. Finally, from the computational results we conclude that these heuristics can be effective in testing CARP.