I did a little research on the accident history of the Cirrus SR20/22 versus the "Corvalis". These two aircraft are remarkably similar except for the chute on the Cirrus, so perhaps there is some insight as to the benefit or not of the chute by comparing them. It doesn't make sense to compare the Cirrus against, say, a Cessna 182, not the same kind of airplane.

Assumptions:

In this context, "Corvalis" means any Lancair, Columbia, Cessna, or Textron airplane that was the 300, 350, 400 model (LC40, LC41, LC42 type). It takes a fair bit of effort to dig out all the variants here and find all the accidents due to the name changes.

Cirrus is SR20 and SR22s of all variations. Finding all Cirrus was easier due to stability of company name.

Delivery information came from GAMA and represents worldwide shipments. For 2014, I had only 3 quarters of data, so I extrapolated deliveries for the missing 4th quarter. Total airplanes is 5,885 Cirrus and 852 Corvalis (which was somewhat larger than I expected, a meaningful number).

Accident data is from NSTB web site. If a foreign accident was not mentioned on the NTSB web site, then it wasn't included. Data through 2014 was used. I did no qualification of the accidents, if the NTSB called it fatal, I counted it. Only fatal accidents were counted.

Not included were any fatal accidents involving experimental versions before they were delivered to customers (curiously, both Lancair and Cirrus each had one such occurrence in 1999). I decided that those didn't count, only customer delivered certified units in play.

I assumed every airplane in the fleet did 125 hours a year. This number was achieved by taking the 6 million fleet hours Cirrus claimed this year and dividing it by the total aircraft years in the fleet when they made that announcement. That came right in around 125 hours per aircraft per year. There is good evidence that new aircraft fly more than old ones and I don't take that into account. I doubt it would swing the numbers very much, particular in the last few years where most of the interest is.

I then assumed that every Corvalis flew as much as every Cirrus. I got nothing to prove that is the case other than the argument the two kinds of airplanes are very similar in performance, cost, capability, etc. If we had some better means of measuring that, I'm all for it.

I also assumed there were no hull losses in each fleet. That is, every airplane delivered is still flying. This is obviously incorrect technically, but if you assume similar hull loss rates, the conclusions won't be affected by making this assessment more precise. The planes are all relatively new, surely 90% or more of them are still flying, so this would be a small adjustment in any case.

I also counted CAPS events that did NOT result in a fatal accident. These are the supposed "saves", or fatal accidents averted by the chute. Whether one believes this is true or not, it will be interesting to graph that as well. Note that I was careful not to double count, a CAPS event that did have a fatality is not in this count.

Analysis:

Okay, so we put all this data into the spreadsheet meat grinder and what do we get. See the graphs attached.

The first graph, Yearly in Isolation, is the single year accident rate for Corvalis, Cirrus, and Cirrus w/CAPS events added. As expected, looking at only one year produces some wild swings in the data. Since I assume no hull losses, the fleet size is slightly overestimated as you move forward in time, but let's ignore that for now.

For the Corvalis, there are very wild swings which is characteristic of having fewer airplanes in the fleet so more sensitivity to each event.

For the Cirrus, the yearly rates are actually reasonably steady even so.

Note the up tick in 2011 for Cirrus. That was a bad year by number of events, but the fleet size was large, so it wasn't particularly out of character, bouncing up and down around 2.

Note the downward trend in 2012-2014. Note interestingly that not only was fatals way down (red line), but the fatals+CAPS (green line) also decreased. One could start to say that not only are Cirrus pilots not having as many fatal accidents, but they are also getting into somewhat fewer "fatal situations" when viewed as a rate not an absolute number.

2014 was a very good year for Cirrus, was a very bad year for Corvalis, both types had 3 fatals.

The second chart is Cummulative which includes the entire history of the type from the year indicated into the past.

As expected, this smooths things out as the years go by. In fact, from the period 2004 to 2011, the graph is remarkably smooth for the Cirrus hovering about 2 for actual fatals and 2.7 for fatals+CAPS.

The Corvalis has no fatals until 2005, then tracks about 2 as well. A period of no fatals gradually lowers the rate down to 1.2 or so for the years 2010 onward. The bad 2014 causes an up tick but due to the total history, the effect is muted.

When Cirrus had the bad 2011 year, and people made comparisons to the Corvalis, it was easy to see why they could point to the Corvalis having a better record. I think it was about this time that Bertorelli wrote the article that quoted the 1.0 Corvalis, 1.6 Cirrus numbers. My charts give a higher number, but none the less consistent with his when scaled. It was from this time I concluded Corvalis was having fewer accidents than Cirrus despite the chute.

The third chart is a rolling 3 year average. This is a compromise between the noisy year by year numbers and the smooth full history chart which fails to show recent trends.

The Corvalis shows a definite bump 2005-2007, somewhat worse than Cirrus, then a quiet period, then a gradual uptick through 2014. In the 2011 time period, the Corvalis 3 year average was much better than the Cirrus.

The Cirrus is again quite stable. The 3 year moving average is just starting to be sensitive to the downward trend. This is an indication that the change is real, not noise, since it survives a 3 year averaging. At the end of this graph, the actual fatal accident rate of the Cirrus is more than the Corvalis, but Cirrus pilots are finding themselves in more fatal situations when CAPS events are included.

If one assumes that half of all CAPS events are actual fatals averted, then the Cirrus and Corvalis would have basically identical "fatal situation" rates on a 3 year moving average including 2012-2014. That is, the blue line is halfway between the red and green.

I think it is safe to say there has been a real change in Cirrus fatal accident rates. Given the large number of recent chute pulls, I was a bit surprised to find the total "fatal situations" (fatals+CAPS) was actually going down as well. I'm a firm believer in accident avoidance so the "pull early, pull often" teachings are having some avoidance benefits as well. Again, teaching skills engenders judgment, not at all an obvious connection.

If you go back to the yearly chart, you can see the Cirrus fatals (red) and fatals+CAPS (green) are not far apart for most years. They start to separate more distinctively at the end which is the real sea change.

What will the next 3 years bring? Don't know. If Cirrus pilots manage to keep the trend going, they will have demonstrated a remarkable change in accident history by changing only the mindset of the pilots. The plane is identical. I've been through such a project myself, but that was with regulatory force. The Cirrus change is entirely community driven.

As for Corvalis, I don't think there has been any fundamental change in their fleet, so I hope 2014 is proven to be an aberration. Time will tell.

There are an unbelievably large set of valid criticisms for the data presented here. Such is the nature of doing this kind of work.

I've included the spreadsheet (in ODS format, OpenOffice, Libre Office) so anyone can see the data and if I made any mistakes.

Mike C.

Why include the SR20? It's not really comparable to the Corvallis the way the SR22 is and there's plenty of data without it. No clue whether leaving out the SR20 would make the numbers better or worse although I doubt it would change the trend, which seems pretty robust.

You can separate it out if you like.

I suspect it won't make much difference.

The proportion of accidents and the proportion of SR20s fleet years seem about the same.

Mike C.

Excellent, excellent post.

Hi Mike,

Great information. Thanks for putting this together.

That is interesting, but all built on an assumption that Corvalis fly the same hours as Cirri.

Anecdotally, I rarely see a Corvalis on the ramp. I always see Cirrus.

At this moment in time there are 29 SR22s an 2 SR20s on Flightaware.

There are 2 Cessna 400s.

Very nice work Mike. I'm amazed you have the kind of time to devote to little projects like these. I appreciate your diligence.

I had the same reaction as Jesse. I rarely ever see a Corvalis flying and I fly to different airports every week. Since the #of Corvalis' registered is much higher than I expected (5,885 Cirrus and 852 Corvalis per you info above), I would think you'd see many more on the ramp or on Flightaware.

If you're really looking to get solid info here, I believe Fligthaware offers services for custom reporting. You could have them compile a comparison of total # of flights and/or hours of SR22 v. Corvalis in their system to really get a fair comparison. My feeling is that Corvalis hours / aircraft are likely less than SR22.

Can you articulate an underlying mechanism for that?

That is, why would a Corvalis owner fly their airplane less?

Mike C.

Can you articulate an underlying mechanism for that?

That is, why would a Corvalis owner fly their airplane less?

Mike C.

First, they won't give it up without $$$$. I asked in the past about this sort of thing.

Second, it doesn't go all the way back to 2000 when these planes first appeared.

Third, it covers only IFR flying, and only in the US.

I would find it inconsistent to say both these things:

"Corvalis pilots fly less despite their being no obvious reason why"

and

"We can extrapolate US IFR hours to VFR and worldwide hours"

Seems to me that if the US IFR hours are different, then whatever causes that could easily make the VFR and worldwide hours flip the other way. If you don't understand the underlying mechanism for the usage difference, then you can't make reasonable assessments as to how to extrapolate that information.

Mike C.

It may not cover all flying, but it is an appropriate direct comparison.

Sample it for a week, and there is your data.

Many times, it's all about the underlying assumptions.

Do 300/400 planes fly less (or more) IFR than Cirrus planes? Would an extrapolation of Flightaware data therefore be wrong? Why?

I covertly sampled some Flightaware data about 8 years ago. At that point, I got the following data on mid-winter IFR flying between Cirrus and Columbia. Each dot represents average IFR mileage flown/registered airframe in the sampling period. X axis is year of production.

The above is old data. Not sure how widely applicable it is to estimating total fleet hours. I sampled many other types and the new=flies more association was really robust. (This was before registrations expired, which explains some of it)

As for do Cirri fly more than Corvali ... I think so, but only have anecdotal thoughts. The typical Corvalis purchaser seems to be older and purchasing a new airplane because they can and they enjoy it. Cirrus seems to have more young businessmen who are going to put 1000 hours on the plane in 2-3 years then get a Mustang. I have no data to back this up.

You can separate it out if you like.

I suspect it won't make much difference.

The proportion of accidents and the proportion of SR20s fleet years seem about the same.

Mike C.