Modelling car ownership dynamics based on irregularly spaced panel data

Abstract

This paper investigates car ownership dynamics based on irregularly spaced panel data of Denmark. The data originates from a national travel survey, where a fraction of the respondents are recurrent due to the random sampling scheme of the survey. While this creates a rich sample collected over a long time period and with desired variation between the panel observations, it introduces an estimation challenge due to the irregular spacing of the panel observations. This challenge is addressed by estimating changes in car ownership based on a generalised ordered logit model in which the irregular nature of the spacing between panel observations is controlled by including panel-specific weights in the log-likelihood function. The estimated model, which is formulated as a first-difference approach, includes several variables explaining the change of car ownership. Specifically, it is found that accessibility improvements, measured as the number of people that can be reached by public transport within a certain time interval, significantly reduce the likelihood of acquiring additional cars in the observation period. In line with the literature, it is also confirmed that changes in income, number of adults and driver’s licenses within the household have a significant impact on household car ownership changes.