On 31 Dec 2004 11:38:32 -0800, "Jim Rosinski" wrote:

Ron Wanttaja wrote:

We'll probably never really know the true answers, until somebody
makes a long term comprehensive study of hours flown and types of
flying versus accidents across the whole fleet.

Ahem. KITPLANES magazine, October 2004.

How about a brief summary for those of us who don't subscribe? One
wouldn't expect a magazine of that title to be completely unbiased
about such a subject, but even biased information can be useful.

A brief summary is tough, when we're talking a 3000+ word article with a bunch
of graphs (I was the author of the article).

The biggest problem with a summary is that it's tough to include all the
cautions and caveats about the input data used. I can't guarantee the results
the article show are accurate, except within the framework of the data and
assumptions used. I've had several email exchanges with folks who didn't agree
with some of the assumptions I used. That's fine...*that's* why I explained my
processes in the article.

But on a summary... it's tough to make it clear where errors could have been
introduced. Given a couple of days, I can probably get the entire article
online. But let me give what summary I can, now.

The study was based on homebuilt aircraft accidents from 1998 to 2000,
inclusive. I downloaded the full NTSB accident summaries for each of those
years. For each accident involving a homebuilt, I studied the narrative and
made my *own* assessment of the cause of the accident. I did the same for
Cessna 172 and 210 accidents (to provide a baseline of comparison of causes). I
referred to these as "Accident Initiators," leaving the phase "Probable Cause"
to the NTSB.

In addition, I already possessed the FAA Registration databases for July 1997
and January 2001. I determined the average "fleet size" for homebuilts and for
the total US registered aircraft for the 1998-2000 time period.

This leads to one problem with the input data. Each registry entry includes a
field for an Airworthiness Classification code. This code will be "1" for a
Standard Category aircraft, "2" for Limited category, "3" for restricted, "4"
for Experimental, and so forth.

Supposedly, this code is assigned when the airplane receives an airworthiness
certificate. Unfortunately, this doesn't always happen with homebuilts...I've
found a number of operational aircraft that have a blank in this field. I've
also found a number of aircraft still under construction that *do* have an
entry.

A while back, I did a step-by-step analysis of the FAA registration database,
and found about 4000 aircraft with "homebuilt-like" names, that have
airworthiness column blank. How many of these airplanes are currently flying?
No one knows. But the FAA and EAA *only* count aircraft that are positively
indicated as Experimental, and have the appropriate code in another column that
indicates that they are Experimental Amateur-Built aircraft. These are also the
only planes *I* counted in my analysis...basically because there was no reliable
way to tally the unmarked aircraft.

With that said: My analysis showed an average annual fleet accident rate of
1.05% for homebuilt aircraft, and 0.68% for all US-registered aircraft. The
Cessna-alone rate was probably more indicative of the GA rate, that was 0.72%.

If homebuilt aircraft during their first 40 hours of flight are eliminated from
the homebuilt accidents, the overall homebuilt rate drops to 0.85%.

Of more interest was determining the accident rate on a per-hour basis. There
are no real figures available. I attempted to approximate this, using the NTSB
accident reports. They include the model year of the accident aircraft, the
date of the accident, and the total time at the time of the accident. I used
these figures to determine the average hourly rate for various types of
aircraft.

Of importance was not, so much, the actual magnitude of the figures, but the
*relative* magnitude, between the two types of aircraft.

But this method had problems as well. It's easy to figure what "1972" in the
model year column means for a Cessna 150...but what does it mean for a
homebuilt? Was it the first year it was registered (which might be ten years
before the first flight), or the predicted completion date, or the actual date
the airplane made its first flight?

But I ran the figures. I came up with an average annual utilization rate for
homebuilts of around 55 hours. For single-engine, fixed-wing, non-agricultural
aircraft, the rate was about 155 hours per year. It results in homebuilts having
a accident rate per 100,000 flight hours about five times higher than the GA
average.

I'm personally skeptical of this figure. I think if one could extract the
*equivalent* operations from the production-aircraft accidents...e.g, only
aircraft that were personally owned and operated (homebuilt can't be rented,
etc.), I think the comparative figures would be closer to the fleet rate.

I'm skeptical of the five times higher rate... but that's the way my numbers
came out, that's what I put in the article, and that's what KITPLANES magazine
printed.

I'll probably do some work and get the whole article online in a bit.

Ron Wanttaja