Mobility tools in pre- and post-COVID Toronto: ownership patterns and model transferability

The reliability of travel behaviour forecasts depends heavily on the temporal stability of their sub-models, such as those for mobility tools ownership. The COVID-19 pandemic introduced a major structural shock to urban travel behaviour, raising important questions about the temporal stability and transferability of mobility tools ownership models. This study examines changes in household mobility tool ownership and model transferability between the pre- and post-COVID eras using data from the 2016 and 2022 Transportation Tomorrow Surveys in the Greater Golden Horseshoe region of Ontario, Canada. A bivariate ordered probit model is estimated to jointly model household car ownership and public transit pass ownership, accounting for socio-demographic characteristics, built environment factors, accessibility measures, daily activity patterns, and survey mode effects. Temporal transferability is assessed using multiple disaggregate and aggregate metrics, including likelihood-based tests, prediction error measures, and mean absolute error statistics. Possession of a driver’s licence, availability of free parking at one’s work location, one’s home municipality, daily activity patterns and accessibility are found to have a significant influence on car and transit pass ownership. Results indicate that car ownership model parameters are comparatively stable over time, while transit pass ownership exhibits substantially greater parameter instability and a larger unexplained component in the post-COVID period. Accessibility-related variables generally decline in importance after the pandemic, while socio-economic factors, particularly income, play a stronger role. Although neither model transfers without significant performance loss, the pre-COVID model demonstrates superior transferability to the post-COVID context than the reverse. These findings suggest that the pandemic amplified pre-existing behavioural instability, particularly for transit-related decisions, with important implications for long-term travel demand forecasting and model updating practices.