Consumption Functions
Short-sighted

Successful time-series models of aggregate consumption generally find that past consumption plays some role in determining the current level of consumption. The presence of such lagged terms has been attributed to the effects of habits or expectations, whether adaptive or rational.

Evidence from the estimation of complete systems of demand equations suggests that habits play an important role in consumer behaviour. These habits may reflect costs of adjustment, in terms of money or of effort: since households usually follow a complex pattern of interlocked activities, the transactions costs for some of the key decisions such as location of housing, jobs and schooling will be reflected indirectly in the more general consumption pattern conditioned on these decisions. It is clear that some kind of behavioural persistence is essential in modelling consumer behaviour, at least at the level of individual commodities.

In Discussion Paper No. 112, Research Fellow John Muellbauer investigates the implications of habit formation in the aggregate consumption function. He distinguishes between two types of habits, rational and myopic. In both cases past consumption affects current preferences. In the case of rational habits, consumers are aware of the effects that their current consumption decisions will have on their future marginal rates of substitution between goods. If habits are myopic, on the other hand, consumers lack this awareness, so that their behaviour may be inconsistent over time.

The consumption function may be formulated in two ways, as a function in which current consumption depends on asset holdings and on current and expected future incomes, or as an 'Euler equation'. The latter is merely the condition which must be satisfied by any optimal intertemporal consumption pattern, and under rational expectations implied that the marginal utility of consumption follows a random walk with drift. Since the marginal utility of consumption is hypothesized to depend on the current consumption level, consumption itself follows such a random walk, at least approximately. On Hall's theory, lagged consumption is the best predictor of current consumption and it follows that other lagged information should be insignificant in explaining current consumption. If habits are rational, the standard life- cycle model proposed by Modigliani and Brumberg in 1954 is modified into a version of the consumption function incorporating partial adjustment. In the Euler-equation formulation, current changes in consumption depend on past changes. In a cross- section context, rational habit formation implies that estimates of the standard life-cycle model of consumption will suffer from omitted variables bias of a predictable form. Myopic habits have rather similar consequences, but in these models the effects of lagged consumption on current consumption are close to zero at the aggregate level, although this is not the case for consumption of individual commodities. As a result, the myopic habits model is essentially indistinguishable from the standard life-cycle model Muellbauer concludes.

Muellbauer then reviews the empirical evidence and explores the effects of habitual behaviour by estimating the Euler-equation form of the consumption function, using quarterly US data for 1954-1981. He takes into account problems created by the aggregation of data over time and other specification issues such as liquidity constraints on consumers. Muellbauer finds that these estimates are not consistent with rational habit formation. Taken together, the evidence appears, therefore, to favour the hypothesis of myopic habits, a somewhat disturbing result for those given to rationalistic interpretations of consumer behaviour.


Habits, Rationality and Myopia in the Life-Cycle Consumption Function
John Muellbauer

Discussion Paper No. 112, June 1986 (ATE)