Consumption
Innovative Models

How is aggregate consumption affected by changes in income prospects? How do interest rates affect savings behaviour? Does the level of government spending on health and other services affect private consumption decisions? These are important questions not only for macroeconomic forecasting, but also in analysing policy questions such as the effect of income taxes on capital formation and the optimal size of the government sector.

Economists commonly base their analysis of private consumption decisions on the 'life-cycle permanent-income' model. In this theoretical framework, individuals are assumed to plan consumption and savings over their entire lifetimes. They base these plans on the income stream they expect over their lifetime and not merely on their income in the current period. This theory must be supplemented, of course, by a description of how individuals form expectations of their future income. One possibility is that expectations are formed rationally, in the sense that agents do not make systematic errors in forecasting their future incomes. In this case, Robert Hall has demonstrated that the behaviour of consumption can be described very simply by a 'random walk'. Tomorrow's consumption depends on consumption today and a random shock which cannot be predicted on the basis of today's information.

Hall's result is, however, strictly true only if real interest rates are constant and if the pattern of spending decisions over time is unaffected by government spending or the amount of leisure consumed. In Discussion Paper No. 54, Research Fellow Charles Bean extends the simple life-cycle model to relax these three assumptions. In empirical tests he finds that this extended model provides a much better description of the US consumption behaviour than the simple model. The estimated parameters are plausible and imply that a temporary one per cent rise in real interest rates would lead to a fall in current consumption of about one-half per cent. A temporary fall in government spending of one per cent with unchanged taxes would lead to a temporary rise in consumption of about one-third per cent.

Bean argues that the overall performance of the extended life- cycle permanent-income model is quite good, even though the restrictions implied by the model are still just rejected on the US data set. In particular he argues that the model can encompass the findings of other researchers who have estimated conventional consumption functions relating spending to current and past values of income and other variables.


The Estimation of 'Surprise' Models and the 'Surprise' Consumption Function
Charles R Bean

Discussion Paper No. 54, February 1985 (ATE)