AN EVALUATION OF CAR OWNERSHIP FORECASTING TECHNIQUES

In this paper we see how different techniques have been developed for deriving car ownership forecasts which are required for differing purposes. The simplest thing to do is to extrapolate past time series. This is largely the basis of Tanner's work at the T.R.R.L. which is used by the D.O.E. in its forecasting. Since it was thought unreasonable that car ownership would continue to expand without check for ever it is general practice to fit S-shaped curves. The most popular of these, at the moment, is the logistic. The background to the theory of extrapolation techniques is given in section 2, and the practice of, and objections to, this method are discussed in section 3. It emerges that it is very difficult to predict the saturation level of car ownership. Burrell argues that the saturation level will be ever increasing. Adams says that Tanner's 0.45 is the "figure he first thought of" rather than the result of a particular calculation. Analysis of Adams' arguments upheld Adams' contention that the national saturation level should not be estimated by a single growth (g) on level (c) regression using cross-sectional data for counties all heading towards different saturation levels, but did not support Adam's objection to the use of aggregate time-series data in this way. In section 4 the background to the econometric models of car-ownership is explored. Elementary utility theory is invoked to illustrate why people should wish to own a car. It is found that proper explanatory models of car ownership are not suitable for estimation. Instead simple linear, loglinear or logit formulations are used in multiple regression models. As an alternative the method of category analysis has been used. Results from these methods are discussed in section 5. /Author/

Media Info

Subject/Index Terms

Filing Info

  • Accession Number: 00177149
  • Record Type: Publication
  • Files: TRIS
  • Created Date: Aug 19 1978 12:00AM