Re: Air Safety (was Re: A3xx vs B747-600)

Date:         08 Dec 96 04:12:31 
From:         kls@ohare.Chicago.COM (Karl Swartz)
Organization: Chicago Software Works, Menlo Park, California
References:   1
Followups:    1 2
Next article
View raw article
  or MIME structure

>I visited the Air Safety Home Page ...

I assume you're referring to http://airsafe.com/.

>and I saw that the B747 was more "dangerous" than the A320 and the
>DC9 "safer" than the B737. All this according to a "fatal event rate
>per million flights".

It's counting flights in which there were *passenger* fatalities.

>Are these misleading statistics?

Unfortunately, the best answer is "it depends."  What exactly are you
trying to measure?  If you're interested in measuring the safety of a
given aircraft design, there are probably better metrics.  I would
count flights which resulted in loss of life and/or the aircraft, at
least as a better starting place.

The 17 fatal accidents listed for the 747 include KAL 007 (shot down
by a Soviet fighter) and no less than five hijackings and bombings.
Similarly, of the five fatal accidents listed for the A300, one was
shot down by the US Navy and two were hijackings.  The 757 gets
dinged for having been stopped in the wrong spot when a hijacked
737 hit it.

If the goal is to study the safety of various aircraft, are these
sorts of accidents really of relevance?  One could perhaps argue
that the 747, being the largest Western airliner in service, is an
attractive target for the world's loonies, and thus risker a riskier
plane on which to fly.

How, though, does one justify accident #14 for the 747, in which a
passenger aboard an Aerolineas Argentinas died of food poisoning?
I guess it might make sense *if* there's something about the design
of the 747 and its galleys that led to the fatality.

More troubling, from an aircraft safety stanpoint, is counting only
*passenger* fatalities.  This leads to the omission from the Fatal
Event Rates table of both the Aloha "convertible" (the only fatality
was a flight attendant) and the US Air Force T-43A crash in Dubrovnik
(because it was not an airline flight, and thus, apparently, the
passengers were not "passengers").  These two accidents are included
in the detailed list, however, with notes explaining why they were
included even though they don't meet the stated criterion.

The inclusion of the A320 crash at Habsheim but not the A330 crash at
Toulouse seems odd, too.  The Habsheim flight was a demonstration
flight, not an airline flight, though the aircraft had been delivered
to Air France three days prior to the crash.  While the A330 flight
was officially a test flight, it was also serving as a demonstration
flight -- only three of the seven people aboard were involved in the
flight testing, so at least four "passengers" were killed.  I find it
difficult to conjure up a defensible argument for only counting one
of these two flights.

I think Todd Curtis, the author of the Air Safety page, has done an
excellent job of research and documentation, but IMO, the choice of
metrics for the statistical analysis of aircraft safety was poor.

--
Karl Swartz	|Home	kls@chicago.com
		|Work	kls@netapp.com
		|WWW	http://www.chicago.com/~kls/
Moderator of sci.aeronautics.airliners -- Unix/network work pays the bills