Abstract
The usefulness of the Environmental Protection Agency's (EPA) passenger car and light truck fuel economy estimates has been the subject of debate for the past three decades. For the labels on new vehicles and the fuel economy information given to the public, the EPA adjusts dynamometer test results downward by 10% for the city cycle and 22% for the highway cycle to better reflect real world driving conditions. These adjustment factors were developed in 1984 and their continued validity has repeatedly been questioned. In March of 2005 the U.S. Department of Energy (DOE) and EPA's fuel economy information website, , began allowing users to voluntarily share fuel economy estimates. This paper presents an initial statistical analysis of more than 3,000 estimates submitted by website users. The analysis suggests two potentially important results: (1) adjusted, combined EPA fuel economy estimates appear to be approximately unbiased estimators of the average fuel economy consumers will experience in actual driving, and (2) the EPA estimates are highly imprecise predictors of any given individual's in-use fuel economy, an approximate 95% confidence interval being +/-7 MPG. These results imply that what is needed is not less biased adjustment factors for the EPA estimates but rather more precise methods of predicting the fuel economy individual consumers will achieve in their own driving.