This analysis exploits new data from the Vehicle Ownership and Alternatives Survey, which elicits beliefs over the financial benefits of owning higher fuel economy vehicles. The data are used to test for underestimation and to document evidence of "MPG Illusion": consumers think as if fuel costs scale linearly in miles per gallon instead of gallons per mile. Counterfactuals suggest that the MPG Illusion reduces welfare by less than $4 per new vehicle. Furthermore, even the most severe plausible underestimation of the financial benefits of fuel economy cannot account for the consumer welfare gains attributed to fuel economy standards. (JEL D12, D83, L62)C onsumers' choices depend both on preferences over outcomes and on beliefs about how choices map into these outcomes. Typical demand models assume that consumers have unbiased beliefs about how a product attribute maps into utility, implying that aggregate choice patterns map directly into underlying preferences. However, there is empirical evidence from different contexts that beliefs are sometimes systematically biased. For example, Starbucks customers tend to overestimate the calories in drinks and underestimate the calories in food (Bollinger, Leslie, and Sorensen 2011). People signing up for gyms are overconfident about their future attendance and about their likelihood of cancelling automatically renewed memberships (DellaVigna and Malmendier 2006). More than 70 percent of seniors choosing between Medicare plans underestimate potential cost savings from switching (Kling et al. 2012).This paper focuses on the choice of automobiles, one of the most expensive and complicated decisions that consumers make. In particular, I focus on consumers' beliefs about the differences in fuel costs between vehicles with different fuel economy ratings. This is motivated by recent work documenting that consumers are