We focus on the revealed preference conditions that characterize the collection of nite data sets that are consistent with the maximization of a weakly separable utility function. From a theoretical perspective, we show that verifying these revealed preference conditions is a difficult problem, i.e. it is np-complete. From a practical perspective, we present an integer programming approach that can verify the revealed preference conditions in a straightforward way, which is particularly attractive in view of empirical analysis. We demonstrate the versatility of this integer programming approach by showing that it also allows for testing homothetic separability and weak separability of the indirect utility function. We illustrate the practical usefulness of the approach by an empirical application to Spanish household consumption data. In this application we also include two statistical tests in which we account for measurement error.
JEL Classi cation: C14, C60, D01, D10.Keywords: weak separability, revealed preference, integer programming * We thank Ian Crawford for generously providing us the data of the Encuesta Continua de Presupestos Familiares (ECPF), which we use in our empirical study. We are also grateful to Adrian Fleissig, James Swofford and Gerald Whitney for helpful comments. is paper is a merger of two working papers, Cherchye, Demuynck, and De Rock (2011c) and Hjertstrand (2011