The paper addresses the question whether utility may be viewed as a completely relative concept. In a dynamic setting this means that one has to model both habit formation and utility interdependence. The resulting model contains unobservable variables and requires panel data to be estimated. Using the first two waves of an annual panel in The Netherlands, different specifications of the model are estimated, involving alternative sets of identifying restrictions. It turns out that the data are compatible with the hypothesis that utility is completely relative, but we cannot exclude the possibility that utility is partly relative and partly absolute.