NRTF Working Paper 1: Car Ownership: Modelling and Forecasting executive summary

Executive Summary

This Working Paper describes the methods used to produce the forecasts of car ownership presented in the National Road Traffic Forecasts (NRTF) 1997. The car traffic forecasts comprise two components; a forecast of car ownership (defined as the total number of cars), and a forecast of car use (defined as the average annual distance travelled per car). Details of the car use forecasts are given in Working Paper No. 2. Car traffic accounted for 83% of total traffic in 1996, so the forecasts of car traffic are an important element of NRTF 1997.

1. After the publication of the NRTF 1989, an extensive review of the econometric methods and models used to forecast car traffic was instigated. In the course of this review, the Department commissioned MVA Ltd. to report into improved car ownership models 1 . The car ownership model used for the purpose of the NRTF 1997 follows the specification and structure suggested by MVA Ltd., and this Working Paper draws heavily on the research findings 2 .

2. The main points of the car ownership forecast are as follows;

  • car ownership is forecast to increase by 46% between 1996 and 2031, from 23 million cars 3 in 1996 to around 33.5 million cars in 2031,
  • the forecast growth in car ownership differs significantly across households groups. For example, households with one adult (not retired) are forecast to increase car ownership from 2.2 million in 1996 to 4.1 million in 2031, an increase of 85%. Households with three-plus adults, no children are forecast to own 0.3 million in 2031, compared to 2.9 million cars in 1996, a predicted fall of 77%,
  • the number of cars owned (in millions) is forecast to grow most rapidly in rural and metropolitan areas. However, since most of the growth in rural car ownership can be explained by an increase in the number of households living in rural areas, in terms of cars per head / per household, car ownership in rural areas is forecast to increase at a lesser rate than other area-types.

3. In summary, the NRTF 1997 car ownership model comprises two main elements. Firstly, a household structure sub-model is derived in order to forecast the levels and composition of households over the forecast period. Secondly, probabilities of car ownership by household type, car-type 4 and area-type are developed using (pooled) cross-sectional household based models. An unresolved difficulty of the car ownership model is the omission of either car purchase prices or fuel prices as explanatory variables, and the Department continues to review its approach to modelling and forecasting car ownership.

4. The remainder of this paper proceeds as follows. Section 2 provides an overview of the determinants of car ownership. The historical experience of car ownership in Great Britain is the subject of Section 3. Section 4 details the preferred car ownership modelling and forecasting method for NRTF 1997. The section includes explanations of the determining variables, input assumptions, model forms, estimation results, the forecasting process and finally the forecast results. Annex 1 outlines alternative approaches to modelling car ownership that were considered in the course of preparing the NRTF 1997. Annex 2 summarises some international comparisons of car ownership and licence holding. Annexes 3 and 4 describes the procedures used to forecast the number of households, and the number of car driving licences using two sub-models developed for the purposes of the NRTF. Annex 5 presents detail of the intermediate estimation results derived in the modelling process, and Annex 6 contains tables of selected forecast year elasticities of car ownership with respect to income.

1 “Improved Car Ownership Models”, research report by MVA Ltd for the Department of Transport, March 1996.

2 A dual approach of national and local car ownership models (the former for the NRTF and the latter for the National Trip End Model (NTEM) software system) has been developed. To maintain consistency, the local models are constrained to the NRTF forecasts. Further details of the local car ownership model used in NTEM are contained in the Design Manual for Roads and Bridges, Volume 12, Section 2, Part 3 - the National Trip End Model.

3 The NRTF 1997 car ownership model uses the Census definition of cars, which includes all cars and vans available to the household for private use, a total of 23.02 million vehicles in 1996. This differs from the Driver and Vehicle Licensing Agency (DVLA) definition of ‘body-type’ cars (which excludes vans) and unregistered vehicles, amounting to 22.24 million vehicles in 1996. Forecast growth rates in the numbers of vehicles in the car stock presented in the Working Paper are based on the Census definition, but can also be applied to the ‘body-type’ cars definition.

4 Car type is used in these Working Papers to represent the distinction between ‘single’ car ownership and ‘multi-car’ ownership.