Consumption and Interest
Young spendthrifts?

If the interest rate is interpreted as the relative price of present versus future consumption, the relationship between them should depend on the extent of their substitutability, which a number of researchers have sought to estimate using aggregate consumption and interest rate data. The estimation of such intertemporal substitution is important for the assessment of the effects on savings of changes in taxation or of the burden of the budget deficit. Some argue on theoretical grounds that if consumers are rational and enjoy free access to credit markets then consumption growth should depend on the interest rate alone. Most empirical studies indicate, however, that aggregate consumption growth is strongly related to growth in disposable income, while its relation with interest rates is weak.

In Discussion Paper No. 476, Research Fellow Orazio Attanasio and Guglielmo Weber question whether aggregate data are suitable for estimating the elasticity of intertemporal substitution, and they use a long time-series of cross-sections to compare the results from using macro and micro data sets. They compare the estimates obtained from UK national accounts data over 17 years with those from the equivalent data constructed by aggregating the Family Expenditure Survey data set. They find that all their subsequent results derive not from a peculiarity of the micro data set but rather from aggregation. They estimate the elasticities of intertemporal substitution for households whose heads were born during 1930-40 to be 0.74 (using micro data) and 0.30 (using aggregate data). Further, by controlling for demographic and occupational status and using micro data they obtain estimates consistent with the theory: consumption growth is not related to income or productivity growth.

Attanasio and Weber then construct an overlapping generations model in which aggregate consumption growth in equilibrium is determined by productivity growth, while at the micro level it depends on the elasticity of substitution, the time discount parameter and the interest rate. Consumption can therefore be simultaneously decreasing at the micro level and increasing in the aggregate, because younger generations have more spending power than their parents, even though their consumption age- profile may be downward sloping.

The authors argue that the estimation bias introduced by using aggregate rather than micro data is therefore qualitatively very large and negative, but they warn against too literal an interpretation of these theoretical results: the bias introduced into the simple model is very large because half the population dies in each period. They also find, however, that the evidence from models in which people live for 3 and 55 periods shows that this bias can remain sizeable.

Consumption, Productivity Growth and the Interest Rate
Orazio Attanasio and Guglielmo Weber

Discussion Paper No. 476, November 1990 (IM)