Developing an optimal sampling design to monitor the vehicle fuel consumption gap

Monitoring the fuel consumption gap between official and real-world measurements is of great interest to policy makers and researchers. This study explores how sampling methods (simple random, stratified and quota sampling) can be used to supplement and validate the monitoring. Three user datasets were utilised to simulate the fuel consumption gap of the 11.6-15.5 million vehicles registered annually in the European Union (2018-2020). Results suggest that a simple random sample of 16,240 vehicles is sufficient to estimate accurately the fleets' average fuel consumption gap. Stratified sampling can reduce the sample size to less than 4,500 vehicles. To estimate accurately the fuel consumption gap of each manufacturer, the sample size increases to approximately 17,200 vehicles. The increase in sales of plug-in hybrid vehicles in 2020 led to an increase of the average fuel consumption gap by 8% and its standard deviation (variability) by 20%. This higher variability resulted in a more than double sample size, compared to previous years. It was also found that the introduction of the Worldwide Harmonized Light vehicles Test Procedure (WLTP) reduced the average gap by 20-24%. This study highlights the viability of a sampling scheme to estimate the fuel consumption gap by monitoring less than 0.05% of the fleet. Moreover the study draws attention to the need for further analysis and understanding of the real-world use and fuel consumption of plug-in hybrid vehicles.