| The research analyzes consumer decisions over multiple durable and nondurable goods. The multi-good utility function is semi-nonparametric and is based on a multi-variate polynomial expansion. The semi-nonparametric utility function has the constant relative risk aversion function as the leading term of the series expansion and the higher order terms determine the departure from the constant relative risk aversion function. The Euler equations derived from the semi-nonparametric utility function are jointly estimated. The specification of the utility function, i.e. the degree of the polynomial expansion, is determined empirically by the nonlinearity of the data.;The annual and quarterly private consumption expenditure data are reconstructed by replacing expenditure and the price of durables by the stock of durables and the user cost for durables. The user cost for a durable good is initially calculated assuming perfect foresight. Both data sets pass the generalized axiom of revealed preference over two sub-periods with the rejections coinciding with the recession beginning at the end of 1981 and ending in 1982. When calculating user costs based on static expectations, both data sets pass generalized axiom of revealed preference over the entire sample.;Goods are then grouped into categories and the implied weak separability assumption is empirically evaluated. The traditional method of labeling goods into two composite commodities labelled durables and nondurables plus services is rejected. Motor vehicles, other autos, other durables and nondurables plus services are found to be weakly separable from each other. The groups that pass the test for weak separability are then aggregated using simple-sums and the Divisia index.;The Euler equations from the semi-nonparametric utility function is estimated using the generalized method of moments. There is overwhelming evidence against the CRRA function. The SNP utility function does not seem to be misspecified as the overidentifying restrictions fail to be rejected. The results are sensitive to the choice of instruments, starting values, number of Euler equations estimated, frequency of the data, and if the 1980s are included or excluded from the regressions. Durability of goods and nonseparability between expenditures on goods and services have an important impact on how the consumer allocates his/her income over time. |
