The impact of mandating a driving lesson for elderly drivers in Japan using count data models: Case study of Toyota City

Abstract

As one crucial mobility problem for elderly drivers, vehicle crashes due to elderly drivers account for an increased ratio of total vehicle crashes in recent years in Japan. Mandating a driving lesson for elderly drivers was implemented and revised continuously to reduce the number of vehicle crashes. To perform a practical driving lesson for elderly drivers, it is essential to evaluate whether it can reduce vehicle crashes significantly or not. Most previous studies only investigated its effects on fatal or severe rates using rather simple methods, without consideration for the increasing number of elderly drivers and the variability of vehicle crashes in different months. To bridge these research gaps, this study examines the impact of mandating a driving lesson for elderly drivers by some advanced statistical methods based on a monthly level. Vehicle crash records from April 2005 to December 2019 collected in Toyota City, Japan are used for empirical analysis. Three types of count data models, i.e., the Poisson regression model, the Negative Binomial regression model, and the Poisson Integer-Valued Autoregressive (INAR) (1) model are proposed in this study. A comparison of three proposed models was implemented to indicate the similarity and distinction of estimation results. The significant findings of this study suggest that: 1) three proposed models have the same prediction accuracy referring to indexes of Root Mean Squared Error, Mean Absolute Error, and Mean Squared Error; 2) all of them indicate that the revision of license renewal legislation for elderly drivers in March 2017 is a significant factor negatively affecting the number of vehicle crashes at a 10 % significance level; 3) all of them indicate that the number of drivers aged 65 years or older and month-variability are significant factors affecting the number of vehicle crashes at least a 10 % significance level; 4) the time-series nature of vehicle crashes due to elderly drivers was not existing indicated by the result of the Poisson INAR (1) model. Statistical methods proposed in this study can be referred by researchers and engineers to evaluate the effects of traffic safety measures in the research field of traffic and transportation engineering.

Keywords: Driving lesson; Elderly drivers; Negative Binominal regression model; Poisson INAR (1) model; Poisson regression model.