Eric Morris writes on the FREAKONOMICS blog that improving a gas-guzzling car to moderate efficiency is better than improving an already efficient car to even better efficiency. That is, the 10-mpg improvement from 15 to 25 mpg is better than the 10-mpg improvement from 25 to 35 mpg. This isn’t contingent on the distribution of efficiencies; it’s just better, period.
I didn’t find his description super-perspicuous, so I’m going to make up my own with fake rather than real examples. Let’s think about car brands A, B, and C with 33, 50, and 100 mpg. The difference between brands B and C (50 mpg) is almost three times the size of the difference between brands A and B (17 mpg). But instead of thinking about how many miles you can extract from a gallon of gas, think about how many gallons you need to go, say, 100 miles. Car A requires 3 gallons of gas to go 100 miles (100 mi / 33 mpg); Car B requires 2 (100 mi / 50 mpg). But that isn’t to say that the 17-mpg increment buys you one gallon of gas per hundred miles, because Car C, with three times the efficiency increment as expressed in mpg, yields only an additional gallon of savings for that distance (100 mi / 100 mpg == 1). The next 100-mpg increment buys you only half a gallon per hundred miles, and so on.
Morris cites a SCIENCE article showing that only 1% of college students understand this (without being told). I would not have been one of those 1%.
Anyway, the next time I see an ad for a 20-mpg SUV, I won’t chuckle so much.