Sunday, October 19, 2014

(F)lying With Statistics -2

This will make more sense if you start back at the October 12 post...

"Passenger-miles" is the main sleight-of-hand trick the airlines use to work their statistical magic, but it isn't the only gimmick in use; this from an air travel industry Internet page:
"The number of U.S. highway deaths in a typical six-month period -- about 21,000 -- roughly equals all commercial jet air travel fatalities worldwide since the dawn of jet aviation decades ago. In fact, fewer people have died in commercial air travel accidents in America over the past 60 years than are killed in auto accidents in a typical three-month period."
Here we momentarily drop the "passenger-miles" device in favor of what seem like a couple of straight numbers-to- numbers comparisons; but on examination these, too, turn out to be impressive-sounding yet uninformative factoids. Maybe more people travel on highways every six months today than traveled by air in all those decades ("since the dawn..."). Maybe there are more automobile trips every three months now than there have been commercial air travel flights over the past 60 or 70 years. We don't get real numbers. (Note also the language: people die in air travel accidents but they're killed in their cars.)

But it's the passenger-miles gambit that provides the basis for the most competitive comparisons the air travel industry makes. Here's another quote from that Internet page:
"In the United States, it's 22 times safer flying in commercial air travel than driving in a car, according to 1993-95 study by the U.S. National Safety Council comparing accident fatalities per million passenger- miles traveled."
Why "passenger-miles"? Seen any 200-seat cars on the road? No, the automobile figures are compiled at -- what? -- maybe two or three passengers per mile? as compared to the airlines' hundreds? Think how many more individual trips highway travelers took -- the vastly greater number of opportunities for accidents they exposed themselves to -- to rack up the same millions of miles. How many air travel passengers would have died (or been killed) in that number of trips?

But even if you accept passenger-miles as a concept, there's still the question of scorekeeping. One hundred passengers fly 1,000 miles, at which point the plane crashes. Sixty passengers die. Does the airline get 40,000 passenger-miles credit for the survivors?

No, "passenger-miles" may help the airlines figure their ROI or decide on what type of equipment to buy next, but as backup for safety claims -- that ain't getting me out of my car. Let me know, better, when they develop a jetliner that can coast over to the side of that highway in the sky if its engines cut out, so you can ring up the aero club at the emergency callbox. You can improve your chances of surviving an auto accident by wearing a seat belt. What will your seat belt do for you if the view out the front window of your plane becomes cornfield?

                          #                #            #
I had referred to airline industry web pages as the source, but I’m reminded now that it was primarily these newspaper articles, although some websites figured into it later. 

I can also now cite the names of the experts who published the statistics I refute so decisively. They were H.W. Lewis, a Physics professor, and Richard H. Wood, a professor of Safety Science.    I believe it was a coincidence that both articles appeared the same day, as there doesn’t appear to have been a special section on safety in that issue of The Times.  I had written a letter to the editor in 1980 that went nowhere, but the impetus to actually pursue the subject came by way of the late Times columnist, Jack Smith, who touched on it in a column five years later.

Maybe the subject’s been debated someplace, but I’ve never seen it done – a surprise to me considering the prevalence of bright minds much better at mathematics and statistics than mine. That raises the possibility that I could be talking through my hat, something I’ve been known to do – but I don’t think so this time.