Prevailing economic models of consumer behavior completely ignore the well-documented link between context and evaluation. We propose and test a theory that explicitly incorporates this link. Changes in one group's spending shift the frame of reference that defines consumption standards for others just below them on the income scale, giving rise to expenditure cascades. Our model, a descendant of James Duesenberry's relative income hypothesis, predicts the observed ways in which individual savings rates respond to changes in both own and others' permanent income, as well as numerous other stylized fact patterns that are difficult to reconcile with prevailing models.