Household energy demand analysis: An empirical application of the closure test principle

In this paper a set of ten different single-equation models of household energy demand is being analyzed. These simple models are being derived by the imposition of linear parameter restrictions on a fairly general autoregressive distributed lag (ADL) model in log-linear form. Household energy consumption is assumed to be explainable by movements in households' real disposable income, real price of energy, and the temperature variable 'heating degree days'. In the empirical application, Austrian annual data for the period 1970 to 1992 are used. The overall significance level alpha of the tests is being reduced by applying the closure test principle, introduced by Marcus, Peritz, and Gabriel (1976). The application illustrates nicely how one can, by defining a closed system of hypotheses, reduce the significance level in a relatively unordered search for a suitable specific model. The wide range of estimated elasticities, however, indicates that such estimation results depend heavily on the choice of the model specification.