Why mpg lies to you
Going from 10 to 15 mpg and going from 40 to 45 mpg both look like five mpg. Over the same 12,000 miles, the first saves 400 gallons a year and the second saves 33.
That is a factor of twelve between two changes that appear identical. It is the clearest example of a measure being the wrong way round, and it is why most of the world does not use miles per gallon at all.
Fuel used is the reciprocal of mpg
Miles per gallon tells you distance per unit of fuel. What you actually pay for is fuel per unit of distance — the other way up. Because the relationship is a reciprocal rather than a straight line, equal steps in mpg are not equal steps in consumption.
At 10 mpg, 12,000 miles takes 1,200 gallons. At 15 mpg it takes 800 — a saving of 400. At 40 mpg the same distance takes 300 gallons and at 45 it takes 267, a saving of 33. Same five mpg, twelve times the difference.
The practical consequence is that improving the worst vehicle in a household is worth far more than improving the best one, and it is worth more by a margin most people would never guess. Replacing the thirsty second car usually beats upgrading the efficient first one.
12,000 miles a year at $3.40 a gallon
10 -> 15 mpg saves 400 gallons, $1,360 a year
40 -> 45 mpg saves 33 gallons, $113 a year
Same 5 mpg, twelve times the saving
Work it out: Fuel Savings Calculator →
You cannot average mpg
This is the second consequence and it catches out fleets constantly. Drive 100 miles at 20 mpg and 100 miles at 40 mpg and your average is not 30 mpg. You used 5 gallons and then 2.5, so 200 miles on 7.5 gallons, which is 26.7 mpg.
Litres per 100km does not have this problem, because it is linear in the thing you buy. Two equal distances at 10 and 20 L/100km really do average 15. It can be summed, averaged and compared, and mpg cannot.
That is the actual reason most countries switched, and it is a better reason than metrication. A measure you can average is a measure you can manage.
Work it out: Fuel Economy Converter →
And there are two different gallons
A US gallon is 3.785 litres. An imperial gallon is 4.546. The same car reads about twenty per cent higher in imperial mpg than in US mpg, purely because the gallon is bigger.
32 US mpg is 38.4 imperial mpg, and both are 7.35 L/100km. Any comparison drawn from a car magazine, forum post or manufacturer figure needs the gallon identified before it means anything, and it very often is not.
This is a real source of confusion in online comparisons between American and British figures, where the same vehicle appears to be meaningfully more efficient on one side of the Atlantic.
What to do with the number once you have it
The honest way to measure your own economy is brim to brim: fill the tank, drive, fill it again, and divide the distance by what went in. The trip computer is an estimate, and it is usually optimistic by a few per cent.
One full tank is a noisy sample, because how much a pump puts in before cutting off varies. Two or three tanks averaged — as litres per 100km, not as mpg — gives a figure worth trusting.
Range is worth working out at the same time, and worth being pessimistic about. A tank that gives 448 miles on paper should not be planned as 448 miles: the reserve exists because the last of a tank is not reliably usable, and economy on the day is rarely the economy on the sticker.
Questions
Why is 10 to 15 mpg worth more than 40 to 45?
Because mpg is a reciprocal. Over 12,000 miles, 10 to 15 saves 400 gallons while 40 to 45 saves 33 — twelve times as much for the same apparent gain.
Can I average two mpg figures?
No. 100 miles at 20 mpg and 100 at 40 mpg averages 26.7, not 30. Litres per 100km can be averaged, which is one of the main reasons it is preferred.
Why is UK mpg higher than US mpg for the same car?
Because an imperial gallon is 4.546 litres against 3.785 for a US one. The same car reads about twenty per cent higher in imperial figures.
How do I measure my real fuel economy?
Brim to brim: fill up, note the odometer, drive, fill up again, and divide the distance by the fuel added. Average two or three tanks, since one is a noisy sample.
Is the trip computer accurate?
Close, and usually optimistic by a few per cent. It is fine for watching trends and worth checking against a brim-to-brim measurement occasionally.
Which car should I replace to save fuel?
The thirstiest one, almost always. The saving from improving a bad figure dwarfs the saving from improving an already good one over the same distance.