This paper addresses whether households save enough for their retirement. For successive date-of-birth cohorts we analyze income and expenditure patterns around the time of retirement. We find a fall in consumption as household heads retire which cannot be fully explained by a forward-looking consumption-smoothing model that accounts for expected demographic changes and mortality risk. Controlling for labor-market participation explains part, but not all, of this dip. We argue that the only way to reconcile fully the fall in consumption with the life-cycle hypothesis is with the systematic arrival of unexpected adverse information.