I presume car manufacturer design and build vehicles to perform with some specified degree of variability. If I had 100 cars of the same year and model and were somehow able to drive them all at the same time, on the same road, at the same temperature, at the same speed, with the same load, etc., by how many miles per gallon or percent would the fuel consumption of all cars vary from a median value?
Under ideal conditions, nil. BTW, your list (under etc,) is a long one.
The variation would have been much higher in the old days before statistical process control.
However, today’s production lines very carefully control variation through every step of the process such that it’s extremely small, control the distribution curve such that it stays centered around the ideal dimensions, and the data is also used to analyze and monitor process capabilities to ensure that the process itself is well capable of producing parts or assemblies that stay well within these parameters. In addition, processes for doing things like sheetmetal forming, casting, machining, stamping, coating, and assembling have been developed that produce far more consistancy.
Cars coming off the end of production lines today are far more consistantly the same than they used to be. I’d suspect the mileage to be extremely close.
Agree, first of all the CAFE fuel efficiency trials (in the lab) have to have a certain amount of repeatability before the government accepts them. Secondly, as Mountainbike pointed out, the statistical process control is much better these days. A few years ago, Ford was fined heavily by the EPA for cheating on their CAFE economy tests. The process is much more transparent now, but it is still possible to have the odd car that is way below the average, due to a defect in assembly or manufacturing. It almost never happens that a car is way above the figure, for obvious reasons. The warranty program should cover the defective ones.