Cirrus Safety: The Real Numbers
This interesting story on the ditching of a Cirrus SR22 last weekend probably made the news for two reasons: It was a slow news day and the sinking airplane and inflated parachute made good visuals. Otherwise, it was an unremarkable event. A well-prepared pilot made good decisions, had the survival gear aboard and executed his plan. This sort of thing happens a handful of times a year in the Caribbean and elsewhere.
The main distinguisher was that it represented the 32nd time a Cirrus BRS/CAPS had been deployed intentionally or unintentionally and again focused our attention on the overall safety record of the Cirrus line. In a nutshell, despite the parachute system and stall/spin resistant wing, the Cirrus airplanes have, over a dozen years in the market, produced a middle-of-the-pack safety record compared to other GA models, but a fatal accident rate that's a little worse than average. Its percentage of accidents that are fatal is nearly 50 percent and possibly the highest in the industry. Clearly, if the BRS has helped reduce the fatal rate, it hasn't made it better than average. Why this is so is subject to speculation, but I addressed my theory on it in a previous blog.
The exact numbers sort out like this: First, the GA overall accident rate is 6.3/100,000 and the fatal rate is 1.2. The Cirrus combined overall rate is 3.2 and the fatal rate is 1.6. This data alone gives lie to the claim of some Cirrus bashers that the line has a terrible accident rate. It doesn't. Perhaps the perception that Cirrus airplanes crash more frequently results from that news story I mentioned in the lead. Cirrus accidents seem to get news coverage while Cessna crashes don't. The reality is that Cirrus airplanes are involved in accidents a little less than some models and a little more than others. But the likelihood of a Cirrus crash being fatal is higher than it is for other models, according to the accident data. (The actual numbers: 48 percent of SR22 accidents have involved fatalities; 56 percent of SR20s have.)
Comparisons with other models are useful, especially the Columbia/Corvalis. It has an overall accident rate higher than the Cirrus, at 3.9, but a fatal rate—at 1.0—that's lower. Why is this? Again, the reasons are not obvious, because the airplane compares favorably with the Cirrus models. The small-number paradox has some effect. And that's not that the Columbia/Corvalis has about 700 airframes while the Cirrus has over 5000. Rates are rates and exposure is exposure, so the measure is fair and equal. Small numbers mean a few accidents either way can swing the results noticeably. Cirrus has about 5 million fleet hours and in the scheme of things, that's not a big number.
People who insist that composite construction is the best way to build airplanes and that composites are much safer in a crash have to confront some mixed data. Controlled lab tests have shown that composites should do well in protecting occupants in crashes. The theoretical data NASA developed showed this. The field experience is quite different. Both the Jurassic metal 172 and 182 have lower fatal rates than the Cirrus or Columbia and both have much lower percentages of accidents that are fatal than do the Columbia, Cirrus or Diamond, all of which are composite. On the other hand, the Diamond line—the DA40 and DA42—have both the lowest overall and lowest fatal rates by a wide margin. The DA40 fatal rate is .42, less than a quarter of the SR20's fatal rate. Also, the DA40 has no incidence of post-crash fire, while about 16 percent of SR22 airplanes burned after crashing.
Pilot experience and skill obviously play a role here, but total hours of experience might not. Experience data for pilots involved in fatal accidents is sparse, but what data I could develop showed that Cirrus accident pilots had an average of 1454 hours total time, with 240 in type, while Columbia/Corvalis fatal accident pilots had 1631 total, with 237 in type. To my eye, there's no meaningful difference between the two.
So what accounts for this disparity in the accident rates? Could it be just a statistical anomaly that will wash out over more years of experience? That's quite possible, because between the Cirrus and Columbia, the rates aren't all that that different. But compared with the Diamonds, they're getting to the outer edge of that plausibility envelope. Yes, the DA40 is used a little differently than the SR20—more training, less long distance trips where weather is a factor. But there were plenty of SR20 training accidents, too. So I don't buy that it's all due to the way the airplanes are used.
Then what? In a previous blog, I mentioned that I thought risk homeostasis played some role here, although I don't know how much. I don't accept that it's no factor. The reason I think this is that after reading these hundreds of accident reports, I saw many examples of Cirrus pilots making what I would call edgy judgment calls. (The Norden crash I cited in a previous blogs was one of these.) I saw that in other types, too, such as the Mooney pilot who took off downwind 250 pounds over weight. It's just my impression that Cirrus accident pilots, as a group, seemed to have more than their share of what-what-was-he-thinking moments.
And this brings me back to the five- to-one sales ratio between Cirrus and the Columbia/Corvalis. These are like models, with similar performance and selling for similar prices. Virtually everyone in the industry I speak to attributes the Cirrus sales success to its marketing of BRS as an additional safety feature, the oft-mentioned "spouse factor." Five years ago, we surveyed Cirrus buyers and only a third said the parachute swayed them to buy. A third said it had no influence or they weren't sure. But I think the BRS has a much more powerful and subtle influence than that. I also think that BRS's potential was lumped into a general sentiment of Cirrus airplanes being safer, without much deep thought given to exactly what that meant. In other words, I think many buyers did not—and still don't—understand that the BRS's potential is not passive. It's not like seatbelts or airbags. It requires active intervention and possibly no small degree of skill to reach the decision to employ.
Consider this. There have been 83 fatal Cirrus accidents. If just half of those—about 42—had been BRS saves instead, the Cirrus fatal rate would be .82/100,000, substantially lower than the GA average. Acknowledging that if pigs had wings they could fly, it seems to me after all this analysis that Cirrus airplanes, for all their potential, aren't much different than other airplanes. They're only as safe as their owners make them. And thus far, with regard to fatal accidents, that's worse than average.