[Paper Review] Investigating Severity of Motorcycle-Involved Crashes in a Developing Country

Sina Asgharpour, Mohammadjavad Javadinasr+3 more|arXiv (Cornell University)|Sep 19, 2021

Traffic and Road Safety33 references4 citations

TL;DR

This study investigates motorcycle crash severity in Iran using ordered logistic regression on crash data from March 2018 to March 2019, revealing that motorcycle-pedestrian collisions significantly increase the likelihood of injury and fatal outcomes. Key findings show that rider characteristics, road conditions, and temporal factors are major contributors, leading to recommendations for stricter enforcement, education, and targeted road design.

ABSTRACT

Despite paying special attention to the motorcycle-involved crashes in the safety research, little is known about their pattern and impacts in developing countries. The widespread adoption of motorcycles in such regions in tandem with the vulnerability of motorcyclists exacerbates the likelihood of severe crashes. The main objective of this paper is to investigate the underlying factors contributing to the severity of motorcycle-involved crashes through employing crash data from March 2018 to March 2019 from Iran. Considering the ordinal nature of three injury classes of property-damage-only (PDO), injury, and fatal crashes in our data, an ordered logistic regression model is employed to address the problem. The data statistics suggest that motorcycle is responsible for 38% of injury and 15% of all fatal crashes in the dataset. The results indicate that significant factors contributing to more severe crashes include collision, road, temporal, and motorcycle rider characteristics. Among all attributes, our model is most sensitive to the motorcycle-pedestrian accident, which increases the probability of belonging a crash into injury and fatal crashes by 0.289 and 0.019, respectively. Moreover, we discovered a significant degree of correlation between young riders and riders without a license. Finally, upon the insights obtained from the results, we propose safety countermeasures, including 1) strict traffic rule enforcement upon riders and pedestrians, 2) educational programs, and 3) road-specific adjustment policies.

Motivation & Objective

• To understand the patterns and contributing factors behind the severity of motorcycle-involved crashes in a developing country context.
• To address the gap in knowledge regarding motorcycle crash severity in developing nations, where motorcycle use is high and rider vulnerability is pronounced.
• To identify significant predictors of crash severity—ranging from rider behavior to environmental and temporal factors—using empirical crash data.
• To inform evidence-based safety policies by analyzing the ordinal injury outcomes: property-damage-only (PDO), injury, and fatal crashes.

Proposed method

• An ordered logistic regression model is applied to analyze the ordinal nature of crash severity outcomes: PDO, injury, and fatal crashes.
• Data from motorcycle-involved crashes in Iran between March 2018 and March 2019 are used, covering a range of crash attributes.
• The model evaluates the impact of variables including collision type, road conditions, time of day, and rider characteristics such as age and license status.
• Sensitivity analysis identifies the most influential predictors, particularly motorcycle-pedestrian collisions.
• Statistical significance and odds ratios are computed to quantify the effect of each factor on crash severity levels.
• Model validation ensures reliability in estimating the probability of a crash belonging to a higher severity category.

Experimental results

Research questions

• RQ1 What are the key factors that contribute to increased severity in motorcycle-involved crashes in Iran?
• RQ2 How do rider characteristics such as age and licensing status influence crash severity?
• RQ3 To what extent do road conditions and temporal factors (e.g., time of day) affect the likelihood of fatal or injury crashes?
• RQ4 How does a motorcycle-pedestrian collision compare to other collision types in terms of severity impact?
• RQ5 What policy interventions can be derived from the identified risk factors to reduce crash severity?

Key findings

• Motorcycle-pedestrian collisions increase the probability of a crash being classified as injury by 0.289 units on the ordered logit scale.
• These same collisions increase the probability of a fatal crash by 0.019 units, indicating a statistically significant but smaller effect on fatal outcomes.
• Motorcycles were responsible for 38% of injury crashes and 15% of all fatal crashes in the dataset, highlighting their disproportionate role in severe outcomes.
• A strong correlation was found between young riders and those without a driver’s license, suggesting a behavioral or regulatory gap.
• Road conditions, time of day, and collision type were identified as significant contributors to higher crash severity.
• The model demonstrated high sensitivity to rider-related variables, particularly age and licensing status, underscoring their importance in severity prediction.