Decoding household multi-car transactions: A Bayesian belief network approach

Abstract

Cars exert a substantial influence on people’s daily lives, shaping travel behaviors while playing a pivotal role in the sustainability of our living environment. As economies advance and more women join the workforce, households owning multiple cars have become increasingly common. It is believed that the recent advances in mobility tools such as shared cars and bikes as well as Mobility-as-a-Service have the potential to diminish the need for having a second car in the households in which more than one member possesses a driving license. Despite this, the real-world data suggest that buying a second car in households with more than one driving license is still an attractive solution. Understanding the determinants of buying an additional car is vital in devising policies aiming to lessen such interest and the accompanying externalities. This study develops a Bayesian belief network (BBN) model that encompasses households’ socio-demographics and their life events as well as built environment information to capture their interdependences in unraveling the influential factors of household multi-car transaction decisions. The population microdata from the Netherlands CBS (Census Bureau of Statistics) are utilized to train and test the BBN model. This study uncovers concurrent and lagged effects of life events on car transactions as well as the interaction effects between the household head and the partner’s car transaction decisions. The simulation results indicate that the proposed BBN model achieves a high prediction accuracy of over 83.6% for all transaction decisions.