Abstract
Due to multicollinearity and autocorrelation time series data are quite often the least satisfactory method of obtaining empirical results. One solution is offered via pooling of time series and cross-section data. Such a technique was employed in the context of total energy demand for eight EEC countries. Out of the existing literature on pooling techniques the traditional covariance analysis approach was chosen to be the most suitable specification when our sample consists of a number of countries. A basic static demand model with income, price and temperature as explanatory variables was estimated, and the variability of the derived elasticities was observed over time. It was inferred that this relationship is inherently unstable but good predictions can be obtained if the period of estimation is sufficiently small. Indeed the predictions made for the crisis period 1971-74 showed no change of the basic structural relationship. As total demand for energy is subdivided into the demand by the domestic and the industrial sectors, predictions for 1974 become slightly worse. This might be taken to suggest, that at a lower level of disaggregation the expectations generated during the crisis period "show up" more, in which case the model as specified in a static sense must have somehow been affected. The dynamic behaviour of oil demand in the domestic and transportation sectors was investigated through a model that allows for the stock of energy using appliances and expectations to vary. Upon estimation, it was found that the basic elasticities and especially price become "weaker" as we approach 1970, while predictions in the crisis period were badly in error. A partial adjustment dynamic model used for the industrial sector indicated that the dynamic specifications have changed at the firm level earlier than at the household level. It was concluded, that a dynamic approach for the time period under investigation and the distributed lags used, is not able to capture the great changes in expectations that occurred in the oil market. Thus, although a dynamic analysis is more "pragmatic" in describing our economic relationship it is less valid in the crisis period. There are indications that for this limited period a simple static approach, inadequate as it is, could yield better results.