No.46 In 2024
Title
Analysis of primary-party traffic accident rates per driver in Japan from 1995 to 2015:
Do older drivers cause more accidents?
Recipient
Kyoungmin Kim，Keisuke Matsuhashi，Masahiro Ishikawa
Reason for award
Although the number of traffic accidents in Japan is on a downward trend, society still demands traffic safety initiatives that respond to contemporary needs--especially appropriate measures for an increasingly aging population.
In particular, addressing the issue of older drivers has become essential, and the topic is being debated across many fields. Against this backdrop, the present study not only examines accident rates (accidents per license holder) using Japanese traffic accident data, but also analyzes age specific time series accident data to clarify factors--such as traffic safety education and policy measures of the time--and the accompanying period effects.
The analysis adopts Bayesian age-period-cohort (BAPC) methodology, widely used in epidemiology and medicine. By focusing on
1. age effects, which change as individuals grow older;
2. period effects, which change over calendar time; and
3. cohort effects, which reflect characteristics rooted in year of birth,
the study offers a fresh analytical perspective on traffic accidents.
Regarding accident rates, the authors point out that traditional analyses based on numbers of license holders include people who rarely drive. Consequently, age specific accident rates derived this way fail to capture true age characteristics. Using the number of active drivers instead, they found that older drivers do not have markedly more multi vehicle collisions per driver, whereas single vehicle accidents are significantly more frequent among the elderly.
The age effect analysis shows that for men accident rates peak at 20-29 years and 80 years or older, and reach their lowest between 35-69 years. For women, rates peak at 20-24 years and 75 years or older, and are lowest between 30-64 years. In both sexes, accident rates rise sharply with advanced age (≥80 for men, ≥70 for women), indicating a higher likelihood of single vehicle crashes caused by diminished physical abilities or judgment errors that do not involve other parties.
Period effect analysis reveals a decline after a peak around 2000, with rates significantly lower by 2015. Contributing factors include widespread basic traffic safety programs, the diffusion of anti lock braking systems (ABS), and the spread of advanced electronic vehicle control systems.
In sum, this study is highly valued because it:
1. applies Bayesian methods to disentangle age, period, and cohort effects--techniques borrowed from
epidemiology and medicine;
2. demonstrates the need to exclude non driving license holders when analyzing traffic accident data; and
3. provides detailed insight into elderly age groups for whom single vehicle accidents per driver are significantly more common. These contributions offer a novel viewpoint for future analyses of traffic accident data.