I think the authorities have excellent statistics of this. One has to look at the kind of accidents. 60-70% of fatal accidents are either controlled flight into terrain and unintended VFR flight into IMC. Looking at the total accidents, 80-90 % are some runway related accident (runway excursion, hard contact, loss of ground control), but they are seldom fatal.
While UIMC and CFIT are random events (independent of hours flown), the runway related ones are highly dependent on proficiency and the runway itself.
Besides, human factors and limitations, what causes risk and how to cope, is part of the theoretical training for PPL. Only UIMC can be prevented by risk management, simply by flying when the weather is nice, but that also counts for 30-40 % of the fatal accidents. One can also manage ground related accidents by flying in no wind at large runways for instance, but that is seldom a real option regarding runway. Flying in nice weather (wind) helps of course. That would however reduce the number of accidents drastically, but hardly make an effect on the number of fatal accidents.
Either way, one can argue that up to 30-40% of fatal accidents and maybe 50-60% of total accidents can be prevented simply by flying in nice weather (sun is shining and wind is calm). This is true whether you are a low time or high time pilot. The main difference here is that a high time pilot is a better judge of his own limitations and the (weather) factors affecting them, and this requires pushing the limit from time to time. But the thing is, for most low time pilots, pushing the limit is not even remotely interesting. What they want is a nice flight in nice weather looking at the world and meeting other people, they want to have a nice and pleasant time.
Another thing which will mess up the stats, in Europe for sure and probably all other places, is that most people give up almost right away, another chunk gives up at the initial license/medical renewal (some info allegedly from the UK CAA suggests 90% don’t revalidate at that point), so the ones left over to show up in the accident stats are really a small group which is probably highly self selected by various character traits, and by favourable combinations of having both time and money (or opportunity to fly with somebody else paying).
And I don’t think it is the most careful/cautious people who make it through these barriers. For example, there will be a % who are very driven individuals with a dubious attitute to risk – ask any instructor 
But in any case I would expect an accident blip as per The Killing Zone (which I have not read), but it doesn’t have any meaning because the cause is largely an artefact of the numbers versus the timescale spent in GA.
and this requires pushing the limit from time to time.
The problem about pushing the limit is that you do not know where it is until you have crossed it. Children are not risk averse because they do not know the consequences of their actions. Once they’ve burned or hurt themselves they learn the limit and will be less likely to do the same thing again. Just look at kids in the ski slopes. They have no fear.
I have heard several numbers for the killing zone. Some claim that it is cyclical like a sinus curve where the amplitude diminishes with time. It oscillates around a horizontal line. Above the line is the killing zone. 90% of the area enclosed by the curve and the line lies below the killing zone, and the remaining 10% above. The abcissa represents the experience (total hours) and the ordinate the risk.
Could someone give a short resume of what this Killing Zone is all about?
It is about cycles of confidence and reaching a point where over-confidence induces decisions which are beyond the pilot’s capacities and cause accidents.
The book suggests that after a safe period learning to fly, there is a “Killing Zone” from 50-350H where there is an elevated risk of fatal accident. After 350h that the risk reduces again.
Paul Craig dug out 17 years of US accident statistics broken down by the experience category of the pilot, and published them, and that alone was worth buying the book.
Just because the statistical analysis is lacking, doesn’t mean the conclusion is wrong!
Personally, I think his figures do make the case that some period after 100H is more dangerous than the 0-100H experience.
You can see it mostly simply by thinking about drop out rates, which have to be positive, i.e. you can’t get to 100-149H without having been in 0-49H and 50-99H first.
If accidents were equally likely in each experience category, then the accident statistics, broken down by category, reflect the experience cross-section of the pilot community.
For example, if the accident frequency over a long period were 100 for 50-100H, and 70 for 100-150H, this would imply that 30% of the 50-100H pilots drop out.
But the statistics do not look like that:
the accident frequency goes up to a maximum at 100-149H before dropping, which implies either a negative drop-out rate or some kind of Killing Zone, i.e. risk rises initially with experience.
It gets harder to argue the safety at 350+ from the numbers, because there are only 17 years of statistics, and some people will take many years to get that many hours, and the experience cross-section at the start might have been significantly different from the cross-section at the end.
By contrast, for the lower experience bands, I think the faster throughput makes 17 years enough to go on. Fewer ab-initio starters each year will flatter the lower hour pilot statistics, but not by enough to make a difference.
This analysis is not watertight, but IMHO any sensible drop out profile and ‘before’ and ‘after’ cross section of experience force you to accept some kind of “Killing Zone”, although I wouldn’t have called it that.
Interesting, but only if there exist some facts showing that over confidence actually is a factor in accidents. It might as well be that on average a pilot moves on to a different, a more complex, higher performance aircraft at 100-150 h, which can be shown for a fact to produce higher risk.
The killing zone theory is one of many that generalize rather then take into account specific factors. It may have a grain of truth as all these things have but that is about it.
It does not take into account people who e.g. have resumed flying after a break. With that taken I would regard myself as a 200 hr pilot now rather than what my logbook sais. After a certain periode of inactivity, the counter goes down pretty much to zero again.
I agree with Peter that decision making taught by many flight schools are way too focused on play it supersafe every time which leads to a cancellation minded rather than go minded focus. I myself am still working to overcome some of that. How often have I cancelled flights because the feeling was ambiguos rather than based on hard facts. Sobyes,go no go should be done on technical facts much more than on hunches.
The other problem today is that our society is moving towards a risk phobia in all aspects of life. That is partly why overregulation is such an issue. Nobody wants to take risks or worse be held responsible for anyone else taking risks. So the logical consequence is to ban and restrict.
Risk management is today a very sore subject. In many cases bad risk management leads to paralysis and stagnation.
I think part of it all is that we as humans are just very bad at estimating risk when it comes to low probability events. For example (far) more people have died in car accidents because they were afraid of flying commercially, or wanted to avoid the hassle of increased security measures, after the 911 attacks than in the attacks itself. [1,2]
So I ended up asking myself this question: How dangerous is GA flying? Well there is a measure for that which is kinda useful. A micromort: Its just a 1 in 1 million probability of dying, or 1 death in 1 million people.
Of course the whole calculation and argument which follows assumes equal risk for everybody, which is probably not really true, but lets just roll with it for now.
For the US I found some good accident rate information, it comes down to 1.05 fatal accidents per 100 000 hours of flying. 3 This is equal to 10.5 micromorts per flight hour.
Now lets compare that to some other activities to gain some perspective. First we want to establish a baseline, what are the chances of dying without any additional activities. This is strongly dependent on your age, the US CDC has some nice tables for that 4, for my age the chance of dying is 2.8 micromort per day. (I am rather young ;) ) If you are 55 this would put you at about 17 micromorts per day.
In my personal case that means that flying 100 hours a year (which I am not doing, cause I cant afford it) would double my risk of dying that year. But in total it is still a 2*0.1% probability. Not exactly an extreme sport not exactly risk free either.
Lets see how some other activities measure up: 5
Scuba diving gives you between 5-10 micromorts per dive.
Skydiving is about 9 micromorts per jump.
Running a marathon puts you at a risk of 7 micromorts.
Hang gliding is 8 micromorts per trip.
But maybe a more interesting way to look at it is to compare it to other travel methods, this time well see how far 10 micromorts takes you: 5
97 km on a motorbike
270 km on foot
3700 km by car
16 000 km on a commercial jet airliner
190 000 km on a commercial jet and dying due to terrorism
The take away from that is that flying is safer than taking your motorbike but the bad news is that driving would be associated with about a factor 10 less risk.
Looking a bit further into this, if we take an average speed of 400 km/h for the airliner, this gives us 0.25 micromorts per hour. For an airline pilot who flies 1000 hours per year it would be 250 micromorts per year, equal to 25 hours of GA flying, which puts an airline pilot on about the same risk level as probably most GA pilots.
Now comes the statistically sketchy but “fun” part, micromorts are not very good for measuring chronic health risks but lets see what one hour of flying equals: 5
5 liters of wine
14 cigarettes
400 tablespoons of peanut butter
Concerning the whole killing zone thing. I also thought about buying the book at some point, read the Amazon extract and noticed that the methodology is severely lacking within the first couple of pages. And also ended up reading the papers cited by @kwlf. I think to conclude an initial increase in accident rates with total flight hours is scientifically sound. Concerning the decrease with higher flight hours and the fitting of the gamma function, I remain highly skeptical. When you take a look at the data laid out in the paper you notice that the population size decreases like a exp(-b x), when you take that into account the estimated errors on the “tail” of the data set increase exponentially. Looking at the quality of the fit and taking this knowledge into account, I would claim that pretty much any fit is highly unreliable for the tail proportion of this distribution as the sampling error just becomes too large. But I am a physicist and therefore by nature quite skeptical when it comes to very noisy data sets and such fits.
1: Michael Sivak and Michael J. Flannagan. 2004. “Consequences for road traffic fatalities of the reduction in flying following September 11, 2001.” Transportation Research Part F: Traffic Psychology and Behavior 7 (4).
2: https://www.schneier.com/blog/archives/2013/09/excess_automobi.html
3: http://www.flyingmag.com/technique/accidents/fatal-general-aviation-crashes-drop-all-time-low
4: United States Life Tables, 2010. NVSR Volume 63, Number 7. 16 pp. (PHS) 2014 -1120
5: https://en.wikipedia.org/wiki/Micromort
Mooney_Driver wrote:
After a certain periode of inactivity, the counter goes down pretty much to zero again.
I don’t believe that’s true. At least the counter goes up again very much faster than the actual hours flown.
