“Down in North Carolina, Duke management professors Jack Soll and Richard Larrick passed their morning commutes pondering the vagaries of fuel efficiency, and came up with the following thought experiment: What saves more fuel: (A) Replacing a 10mpg Escalade with a 20mpg crossover? Or (B) replacing a 25mpg Accord with a 50mpg Prius?
Counterintuitively, the correct answer is (A). Why? Because over the course of 10,000 miles, a 10mpg SUV will use 1,000 gallons while the 20mpg crossover will use 500 gallons — resulting in a 500-gallon savings. On the other hand, an Accord will burn 400 gallons and the Prius 200, for a net conservation of only 200 gallons.
If you guessed wrong, you’re in good company: After all, mpg is how automotive fuel efficiency has always been presented to us, and converting efficiency into consumption (as we have above) is a convoluted process. But Soll and Larrick found that when they expressed the same question in tems of gallons per miles, their subjects immediately realized that replacing the 10-gallon-per-hundred-mile megamonster with a 5-gallon-per-hundred-mile mini-monster brought the bigger gain. According to their study, the EPA should replace mpg ratings with a gallons-per-mile metric — such as X gallons per 10,000 miles — which would give consumers a better sense of their annual consumption. Definitely an idea worth discussing during your next car pool.” – - VSL
http://news.duke.edu/2008/06/gpmfuqua.html