The response of expenditures to anticipated income changes : panel data estimates

Standard models of intertemporal allocation predict that the time path of expenditures should beindependent of the time path of income. Recently two papers, Parker (1999) and Souleles (1999)have suggested that U.S. households have a high marginal propensity to spend within yearanticipated income changes. We use an expenditure survey panel from Spain to re-examine thisissue. We exploit two important features of the Spanish data. First, we have quarterly panel datathat follows households for more than four quarters. Second, we use the fact that workers areexogenously sorted into one of two payment schemes: some receive the same amount eachmonth of the year and others receive an extra payment in June and December. The extra paymentis large and predictable. We examine the detailed pattern of expenditures over the year to seewhether they differ between the two groups. We fail to find even weak differences. Wecomplement this with a conventional Euler equation analysis of excess sensitivity. Our predictingequation for (quarterly) earnings growth is much better than usual and is likely to give a powerfultest of the hypothesis that predictable changes in income do not lead to changes in expenditurepatterns. The results of this analysis confirm the graphical analysis: we find no evidence ofexcess sensitivity. We conclude that households in normal times do smooth consumption overthe year. We suggest a reconciliation of our results with those of Parker and Souleles.