An Exploration of Factors Affecting Drivers’ Daily Fuel Consumption Efficiencies Considering Multi-Level Random Effects

This paper investigates the factors affecting drivers’ vehicle fuel consumption efficiency, which was defined as the daily average fuel consumption for a unit of driving mileage. Based on the long-term Controller Area Network (CAN) data collected from private cars during 10 months in Toyota City, Japan, we explored the relationships between drivers’ fuel consumption efficiencies, and factors including drivers’ characteristics, car attributes, date-specific environmental attributes, and travel behavior. Furthermore, a multi-level model was applied to explicitly incorporate the effects of individual-specific, date-specific, and observation-specific unobserved factors. According to the estimation results, it was found that, on working days, model fit was significantly enhanced by incorporating all three error terms. Several findings regarding the relationships between observed factors and drivers’ fuel consumption efficiencies were also obtained