Pooled models for car ownership

This paper which describes pooling techniques and their application at a regional level (using British data), utilizes recent developments in econometric theory which enables estimates of regression coefficients to be made using both cross-section and time-series data simultaneously. Comments are made on aspects such as simple pooling, dummy variables, restricted least squares, and the error term model. The results of empirical work are discussed and it is noted that the major weakness is the persistently high level on a regional basis, of over and underprediction. This probably reflects the omission of one or more spatially related explanatory variables. This may reflect a lack of available data at the required level of aggregation. The real disposable income parameter has frequently proved both positive and significant. The values attached to this parameter in the equations suggest an income elasticity of about 0.5. The parameter relating to the percentage of each regions population living in connurbations or large cities has also been consistently significant.