Forecasting Car Ownership in the Sydney Area
F Tsang, A Daly, RAND Europe, UK
A car ownership model developed for Sydney, Australia, in 1999-2000 has recently been updated and reestimated on data from a continuous household survey for 1997-2008. This paper describes the methods used and discusses the results of the two models.
Car ownership can reasonably be considered the most important single variable in describing mobility, but in any case it is clear that an accurate forecast of car ownership is fundamental to forecasting traffic in any area. The paper describes the methods that have been used to develop car ownership forecasts for the Sydney area in New South Wales. A model was developed in 1999-2000 and that model has recently been updated and re-estimated on data from a continuous household survey for 1997-2008.
The model operates at household level, allowing the full detail of the socio-economic characteristics of the household to be taken into account. While it is possible for households to own more cars than they can drive at one time, this is rare and constraining the number of cars to the number of licensed drivers in the household gives a good account of the impact of household size on car ownership. In particular, this estimate of car ownership indicates the car availability of the household.
The use of licensed drivers, rather than adults, improves the model because there remains a significant fraction of the population that does not have a licence. In particular, these tend to be women migrants to Australia, indicating that the differential between men?s and women?s licence-holding is likely to persist as long as migration continues at approximately the same level, even though this difference has declined almost to zero among those born in Australia. A detailed study was made of this issue and this is reported in the paper. Licence holding is projected using cohort analysis methods, applied separately for men and for women and for migrants and those born in Australia, a method that has proved successful in several other countries.
The car ownership model itself represents cars held through companies separately from those privately owned, because the mechanisms influencing these types of ownership are somewhat different. The alternatives of owning none, one and two or more company cars are represented and, conditional on these, the choice is modelled of owning none, one, two and three or more cars in total.
The car ownership alternatives are modelled conditional on the household characteristics and in particular on income. Additionally, parking conditions in the zone of residence of the household have been found to be significant; these are represented by the average price per hour payable for on or off-street parking. These parking costs must be seen as a proxy to a large extent, as residents will not usually have top pay, particularly overnight, to park cars near their homes. However, it has been found that this variable gives a significant improvement to the model, both for company and for private cars, so it is concluded that it gives a good measure of parking difficulty in the zone.
A further variable that has been found to be significant in the case of private car ownership is the accessibility improvement given by car ownership at different levels. Each increase in the number of cars owned by a household increases the car availability and thus improves accessibility for all members of the household (including those who do not have licences). This variable allows the impact on car ownership of car costs per kilometre and of the accessibility and cost of other modes, such as public transport, to be forecast.
The paper presents the models that have been estimated, comparing those of 2000 with the latest ones, and gives some elasticity estimates derived from the models.
Association for European Transport