Friday, February 03, 2006

Example of Subtle Pushing of a Prejudice?

Consider:

By Ken Thomas ASSOCIATED PRESS
1:33 p.m. January 30, 2006

WASHINGTON – People love compact cars as gas prices soar, but there's a sobering cost: The government says they have the highest fatality rate.


That is okay. But the headline given in a San Diego paper was
"Study: Compact cars, SUVs have highest fatality rates by vehicle type"
Which is almost true but a horrible distortion which, with others of a similar nature, has lead people I know personally into wrong conclusions about SUVs. The distortion comes from a wrongly chosen meaning of "fatality rate."

The article presents these numbers:
Compacts had a fatality rate of 17.76 per 100,000 vehicles in 2004, followed by compact pickup trucks with 16.87 and subcompact vehicles with 16.85, according to a report Monday by the National Highway Traffic Safety Administration. Large vans had the lowest rate, 9.34, while pickup trucks and sport utility vehicles had rates of about 15 deaths.


It seems time-costly to confirm the actual calculations (I have found no simple version on the Web), so I will presume the numbers mean what they seem to be said to indicate.
The flaw involves ignoring miles traveled and the number of occupants in a crashed vehicle. Consider the low number for "large vans." I do not know what a large van is but I suspect that the number of occupants of some classes of large vans is usually one. The number of occupants of a compact car, on average, is lower than that for an SUV. There was a wreck of an SUV of teenagers here in Baton Rouge which killed several (10?) of them. The more occupants on average in a class of vehicle the higher, other things being equal, will be the percentage of fatalities per vehicle; though, I do not know by what factor. Also, the that percentage will be higher the greater the average number of miles traveled, doubling one probably doubles the other.
The conclusion that some people will make that SUVs are less safe than, say, a medium-sized sedan, is likely to be false.
The fraction F = Number of Fatalities / Number of Vehicles is flawed.
A better measure of the inverse of safety would be
F*(fudge factor)*(Average Number of Occupants)*(Average Number of Miles Traveled).
Not perfect, but much more honest than F for deciding what to purchase for the safety of your family.

Why might an author use the inferior calculation?
1) sheer ignorance of the author,
2) its math is easier than that of a superior truth,
3) it supports, if only weakly, the author's prejudice against SUVs?
Or, have I missed something?

1 comment:

Rick Ballard said...

Hi CP,

Here is the data sheet that the innumerate journo used. And here are the data sets available at the very same site from which alternative conclusions might be drawn.

The guys that set insurance rates offer the best proof of how 'dangerous' one type of vehicle is in comparison to another. I'm pretty sure that passenger miles traveled is a significant factor in their calulations. Take a look and see which data set seems most useful in determing vehicle safety. I'll check in the morning - it's a bit late for me to do any thinking. 'Course in the morning, it'll be a bit early.