Private pilot dropout hours - US

[abgd]

I'm trying to make sense of chart 1.2 in Paul Craig's book 'The Killing Zone' which seems to show that pilots are most at risk of a fatal accident between 100 and 350 hours.

If I understand correctly, the numbers on the chart are simply the raw numbers of fatal accidents, binned by the number of pilot hours at the time of the accident. This approach will only be valid, if almost all the 200 hour pilots (who don't die) carry on flying until they have at least 1000 hours (the maximum on the chart).

For example, if there are more than 4 times the number of 300 hour private pilots compared with 1000 hour pilots, then we should actually conclude that 1000 hour pilots are more likely to have a fatal accident after making the necessary correction. Now this seems unlikely to me, but I do wonder whether the 'killing zone' is less significant than Paul Craig appears to argue.

Is there anywhere I can find statistics about the average PPL's flying experience in the US, or whether the concept of the 'killing zone' been corroborated by other investigators? The book mentions a 1974 NTSB study but doesn't reference it and I can't find it online.

[Genghis the Engineer]

I think you're doomed to find very little information.

The USA seems to regard central collation of statistics as some massive infringement of personal liberties, so hours on either pilots or airframes are not really available.

You might get something from either US-AOPA or the US Air Taxi Survey, but I'd not hold out much hope.


For the UK, there are statistics sourced from the CAA that are regularly published in Flight Training News. I doubt many other countries are so helpful to the safety researcher.

G

[SDB73]

abgd,

You've alluded to one of the key problems with "The Killing Zone", which is that the statistics have been mis-represented in terms of conclusions.

The book is an incredibly valuable resource for a new PPL, in that the majority of the advice in it, and the case studies are filled with valuable tips and learnings, but the conclusion that there is a "killing zone" in terms of hours flown has been based on a lack of understanding of the statistics being analysed, as well as of the subject of statistical analysis.

As well as your own observation about pilots who don't die needing to fly for X hours longer, there are also countless other aspects which haven't been normalised properly, but one of the biggest issues is that there is no way to draw those conclusions from the data unless those who died in "the killing zone" were allowed to carry on flying after they were dead to see whether they killed themselves again every 100 hours or so.. making them (for possibly some other reason) predisposed to dieing every 100 hours or so - for example. This would no longer show a "killing zone" at all.

The fact that the people who die in the "killing zone" are then removed from the future statistics means that you are unable to draw those conclusions on them.

Ignore the stats, they're fairly meaningless. But take GREAT care to look at each subject (which really do seem to be the key reasons for pilots dieing) and use your best endeavours to avoid the pitfalls / mitigate against the biggest risks. In that, alone, the book is a really valuable resource.

[abgd]

unless those who died in "the killing zone" were allowed to carry on flying after they were dead to see whether they killed themselves again every 100 hours or so.. making them (for possibly some other reason) predisposed to dieing every 100 hours or so - for example. This would no longer show a "killing zone" at all.You mean, are there a few people who are just so likely to kill themselves, that they almost inevitably do so within a few hundred hours of getting their licences? That could fit too. I can't help but think of that poor chap who landed in the river Derwent before recently flying into a mountain.

I found quite a few bits in the medical chapter that just struck me as not being so much 'wrong' as... confused ("Alcohol is absorbed into the blood's haemoglobin faster than oxygen") so I confess I was rather losing faith in the book. For the price, and from an academic publisher, I was expecting something rather more rigorous.

There are bits that seem well written so I shall do as you suggest and plow through it before coming to any further judgement.

[peterh337]

You mean, are there a few people who are just so likely to kill themselves, that they almost inevitably do so within a few hundred hours of getting their licences?That is virtually guaranteed to be true, since only a tiny % of pilots get past a few hundred hours :)

It's like saying that virtually nobody succeeds in stealing the crown jewels before they reach 120 years of age :)

If you read American pilot forums (as I have been since Usenet used to be the main "forum" for aviation years ago) you find that the USA has the same problem as the UK in that nearly everybody chucks it in very quickly, and only the most determined hang on, for a bit longer, and then only the most obscessive types hang on after that :)

Over there they like to moan that the average age at Oshkosh gets 1 year higher every year and is now around 60 :)

That said, Americans tend to hang in there a good bit longer than the Brits, probably simply because GA has a lot of utility value out there (H24 airports everywhere, good "going places" pilot training, etc, etc).

There is no "killing zone" once you adjust for hours flown etc. and that data doesn't publicly exist - in the USA or in the UK.

The UK CAA does have quite a lot of data which it for some reason doesn't publish (e.g. how long pilots retain their medical after the initial issue) - I believe because it would show GA in poor light but like the FAA they will have this kind of data so one could estimate the dropout rate fairly well.

What there is no decent data on however is annual flying hours. Informal surveys suggest that the UK average is somewhere in the 10-20 hours/year, but there will be a big standard deviation on that.

[abgd]

That is virtually guaranteed to be true, since only a tiny % of pilots get past a few hundred hours http://images.ibsrv.net/ibsrv/res/src:www.pprune.org/get/images/smilies/smile.gif

Ah... I wasn't very clear. I meant, you sometimes meet people for whom having an accident seems almost a certainty - whether in aviation or in other spheres. Not just that it's a certainty that any flying accidents they have, will occur whilst they're still flying...

There is no "killing zone" once you adjust for hours flown etc. and that data doesn't publicly exist - in the USA or in the UK.


Thanks for confirming my suspicions.

[CharlieDeltaUK]

Are the stats really so misleading? Since one can only have one fatal accident, isn't it relevant that most of those relate to low hours pilots? The dropout rate along the way to high hours doesn't change that fact does it? Whether or not I go on to acquire higher hours doesn't affect the chances of my having a accident tomorrow. And, if I do get those higher hours, my chances of a fatal accident on the same day 10 Years hence are lower I think. But, I've not studied stats so happy for someone to tell me why my analysis is flawed.

[abgd]

I don't know how misleading the statistics are... It's possible that they're a good approximation, but from what Peter says this seems unlikely.

Try this imaginary scenario for size... Due to an administrative error at EASA it becomes impossible to legally train any new private pilots as of 2012... By April 2020 almost all the pilots still flying will have over 200 hours, though a few stragglers may have just 93. If you plotted a graph of fatal accidents by pilot airtime, you would now have very low figures for low airtime pilots, because there simply wouldn't be that many around. Your graph would seem to imply that higher airtime pilots are more likely to have accidents, but this would be wrong too.

I follow your reasoning that a more experienced pilot will be a safer pilot. On the other hand - it may not be true. More experienced pilots may be doing more challenging flying, or may be getting sloppy - counteracting the benefits of their experience. But what I would say is that if you can't put together decent evidence to support an idea, it's better not to give any evidence at all.

[Big Pistons Forever]

Statistics are only as good as the underlying data and require a very clear understanding of what the question was.

Personally I think the advancing the idea that any one persons likelihood of having an accident will vary by the amount of flight time they have is pointless.

What is worth discussing IMO is the fact that the risks of any particular flight should be well understood and ameliorated as much as practicable.

For example the accident rate as expressed by accidents per 100,000 hrs is much worse for PPL's flying VFR at night as compared to those flying during the day. As well a much higher percentage of night accidents result in fatalities as opposed to daytime accidents. I see a lot of PPL's that treat day and night flights pretty much the same. By not acknowledging and dealing with the extra risks they greatly increasing their exposure to an accident and that exposure is pretty much independent of total flight time.

[peterh337]

To draw conclusions from accident stats one needs to have an idea of the population size, and also of the kind of flying they do.

For example let's say you have two groups of pilots, 100 each, flying same hardware and having had the same training and total hours. Both are instrument rated.

If you get one group to do day VFR only, and the other group to do IFR night flights, there will be a big difference in the accident numbers.

The biggest piece of stats which is missing from GA is what kind of flying different people do. To get that, somebody would have to do serious work and understandably nobody seems keen to do that :)

The regulators (CAA or FAA) could do postal surveys but few pilots are going to give frank answers in those to questions like how much ice they picked up ;)

[SDB73]

Peter, even that would be misleading if it were used to jump to conclusions.

BPF is on the money. Stats are just stats. The conclusions that have been drawn from them are the problem.

CharlieDeltaUK. A good illustration of how the *meaning* of the stats are changed by the people who died :

Imagine you're on a gameshow, and there are three boxes. One of them contains a million pounds, the other two contain nothing.

You are asked to select a box, and you select box number 2 (for instance).

The host then explains he'll remove one of the other two boxes, but in doing so he will not remove the million pound box. (eg. If both unselected boxes are empty, he'll just take one at random, but if one of the two boxes is the million pound box, he will remove the empty one).

There are now two boxes. One of them has the million pounds in it. You are given one final chance to stick with the box you had already selected, or to swap.

Statistically speaking - what should you do? Stick or swap?

Most people would say "it makes no difference, it's 50/50" but it isn't.

(Spoiler : Answer / explanation if you're thinking about it : The million pounds is TWICE as likely to be in the OTHER box. So you will double your chances by swapping. The first selection you made had a 2 in 3 chance of being empty. and a 1 in 3 chance of being a million pounds)

[abgd]

Thankfully I don't think enough of us are going to die flying, that it becomes worth worrying about the Monty Hall problem! I don't even believe it's applicable to the question - the difference is that there is a rule that in one of the boxes there will be a prize. If you take 100 pilots, there is no rule that a given number of them will die - just a statistical probability.

But I think it would be wrong to dismiss the 'killing zone' as meaningless, if it were valid. For example, it would give you some guidance as to how many hours you should have before you can get a CPL and take paying passengers.

I think it alarmed me when I heard about it, because I thought 'My... - I have to fly 300 hours before I will be able to think of taking friends and family for a joy ride?!'

[Torque Tonight]

Thread drift.

Why is the prize twice as likely to be in the other box? After an empty box has been removed you are left with two boxes and the prize is in one of them. If there was an equal (1/3) probability of the prize being in each of the 3 original boxes and then an empty boxes is removed, the two remaining boxes had an equal (to each other) probability of the prize having been placed in either of them. Having completely removed one box from the situation and leaving two, with equal probabilty, then surely the probability can only be 50% and there is no benifit to changing box.:bored:

I think your logic is flawed. At the start you think that there is a 1/3 chance of choosing the winning box and therefore a 2/3 chance the winning box not being the one you chose. When one box is removed leaving only one other box, you are thinking that the 2/3 chance of the prize being in a box you did not choose, versus 1/3 chance of it being in your choice remains valid. This is not the case.

Think of it this way. At the beginning one box 'contains the prize'. Another box is 'empty and will be left'. The last is 'empty and will be removed'. You are attaching a 1/3 probability to each box, but if you think about it, you cannot possibly select 'empty and will be removed'. So the real probabilities are:

'contains the prize' 50%
'empty and will be left' 50%
'empty and will be removed' 0%

[abgd]

The Monty Hall problem is all about information. When the gameshow host gives you information that the prize is not in one of the boxes, you have to adjust your estimates of the probabilities of it being in the other boxes. If you type 'monty hall' into Wikipedia you'll probably find a better explanation.

The only way I've ever found to understand the problem intuitively, is to imagine that there are a lot more boxes.

For example, imagine that there are 100 boxes rather than 3. Your first guess is that the prize is in box 100. If the host now opens all the boxes except box 100 and box 57, the chance is now 99% that the prize is in box 57, and only 1% that it is in box 100.

Or even imagine that you never picked a box to start off with - the host opens all the boxes except no. 57 - the chance of the prize being in that box is now 100%

I remember once discussing it with a professional statistician who agreed that the result was correct, but who at the time (either the problem had just been formulated, or it had just got popular) admitted that she just didn't get it.

[SDB73]

abgd,

With respect, you're missing the point, which is :

Statistics are altered by previous events : eg. You selected a box, which affects which box the hosts takes, which affects your odds of then selecting a winner. AND .. by dieing at 100 hours, you affect whether or not you can die at 200 hours, or 1000 hours.

There are so many ways to recreate the "killing zone" stats without it proving that the "number of hours flown so far" is the factor which causes it.

The simplest (but really there are unlimited ways) is if you imagined the following to be true :

1) 90% of pilots only fly for up to 150 hours before giving it up.
2) Most pilots are in training / have a safety pilot with them for the majority of their first 100 hours.

This alone would create a "killing zone" even if all pilots had EXACTLY the same chance of dieing per hour flown.

Then let's say that there are just some types of people who are more likely to die. Those people remove themselves from the statistics pool once they die.

I'm not saying "THERE IS NO KILLING ZONE".. I'm saying "There is no evidence in that book which even begins to suggest there is one".

My gut instinct is that the longer you fly, the safer you are. It seems fairly intuitive to believe that, and your suggestion "i need to fly 300 hours before I'm safe" might not be wrong. But the thing you'd be very brave to assume is "I'm safe to take family and friends when I'm at 75 hours, but then once I get to about 120 hours, I need to stop flying them for a while as I'm less safe".

[Torque Tonight]

Stats can be a real mind@#&? and it's a long time since I was a studied it but I'm pretty certain that this can't be right

Imagine that there are 100 boxes rather than 3. Your first guess is that the prize is in box 100. If the host now opens all the boxes except box 100 and box 57, the chance is now 99% that the prize is in box 57, and only 1% that it is in box 100.

All the boxes other than 57 and 100 are in the 'empty and will be removed' category with 0% chance of being winners. Box 57 and 100 (or more generically the box you chose originally and the final alternative) will have carried 50% probability from the start.

Any learned fellows able to comment on the logic?

[SDB73]

Torque Tonight

When you first select. You DO (unquestionably) have twice as much chance of selecting an empty box.

When you get the chance to switch, nothing has changed about your original odds. It is still true that it is twice as likely that you had selected an empty box.

[SDB73]

Stats can be a real mind@#&?

You're dead right. Which is why it's so common to see poor conclusions drawn from them, and how easily we all (I'm not pretending to be better here, me too!) accept them willingly.

[Fuji Abound]

Based entirely on the pilots i have flown with there is a degree of complacency that sets in after the first 100 hours running up to the first 600 hours. Beyond 600 hours it would seem they have usually had a fright or two, which is inevitable; to still be flying after 600 hours also means they probably take their flying more seriously.

I have flown with some pilots in the former category that i doubt would pass their ppl.

Were there a properly controlled study which adjusts for the factors raised earlier i would not be surprised if there were some truth in the premise.

[Torque Tonight]

SDB73, There has been a fundamental change to the odds as you know with certainty that one empty box has been removed. Where is the flaw in my logic? :confused:

[abgd]

sdb...

I agree with your reasoning that if you die at 100 hours you can't kill yourself at 300 hours. However, only about 2% of GA pilots will die in flying accidents, whereas a much higher (and currently unknown to us) proportion will give up early in their flying careers.

The question is, I guess, simply whether a high proportion of the deaths were 'accidents waiting to happen', and what proportion happened to generally competent, appropriately cautious people who just got unlucky.

If the former is predominantly true, then I guess it may be important to take the effect into account. If the latter is predominantly true, then I doubt it will have much influence on the statistics.

I'm still not convinced it's exactly analogous to the monty hall problem though.

Interactivate: Simple Monty Hall (http://www.shodor.org/interactivate/activities/SimpleMontyHall/)

for anyone in need of convincing.

[Whopity]

In 1995 the CAA published CAP 667 -Review of General Aviation Fatal Accidents 1985-1994. This appeared to be a very authoritative document so based upon my own observations, I asked them what proportion of owner operators were involved in in these accidents. They had no idea. I believe on further investigation it was shown to be a substantial proportion indicating that they had omitted to consider a major factor in their analysis.

[abgd]

I wonder whether 'freedom of information' requests would be a good way of getting at some of the data - alas currently too busy to really formulate questions and mow through vast amounts of data.

[peterh337]

I asked them what proportion of owner operators were involved in in these accidents. They had no idea. I believe on further investigation it was shown to be a substantial proportion indicating that they had omitted to consider a major factor in their analysis.

I wonder what that means.

Probably just the fairly self evident fact that nearly every pilot who has clocked up more than a few hundred hours is an aircraft owner.

It gets awfully expensive to do a lot of hours when renting.

So while one would expect owner pilots to have the most prangs because they are the highest hour group, one would also expect them to have the highest currency :) So what does this tell us? It could be that currency doesn't matter much because low currency owners simply avoid nontrivial flights. I think that is probably true because most low hour pilots I know (owners or renters) don't go anywhere past the local burger run on a nice day.

But all this could have been surveyed, by a body which gets access to the CAA database and which can anonymise the data so the CAA cannot see individual identities.

[Whopity]

So what does this tell us?No ongoing supervision! Pilots who fly in clubs are subject to currency and annual checks whereas the owner operator has no oversight allowing some to develop bad practices.

[abgd]

Well, you have to have 1 hour every 2 years with an instructor, in the UK. Whether that's sufficient to make a difference is another matter.

[SDB73]

Probably just the fairly self evident fact that nearly every pilot who has clocked up more than a few hundred hours is an aircraft owner.

Exactly. Again, it may be an irrelevant factor or a massive contributor. Who knows without more data and proper analysis.

abgd,

Sorry, you're absolutely right. I didn't mean that the killing zone is an exact parallel to the game show thing (I didn't realise it way Monte Hall, thank you), merely that it was an example of how stats are altered by each changing event. I think we're agreeing! :)

SDB73, There has been a fundamental change to the odds as you know with certainty that one empty box has been removed. Where is the flaw in my logic?

Box 57 and 100 (or more generically the box you chose originally and the final alternative) will have carried 50% probability from the start.

That's the bit that's flawed.

At the start, each box has a 1% chance of being selected by you. Which means that you have 1% chance of selecting box 57. Agreed?

If so, then no matter what happens after that, there is still only a 1% chance that the box you selected is a winning box. Because nothing can change the past. In the past you selected a box, and there is only a 1% chance that the box you selected was a winner. Agreed?

If so, when 98 empty boxes are then remove, there is STILL a 1% chance that the box you originally selected is a winner. If you disagree at this point, go back to the previous paragraph.

So, if you know for sure that there is only a 1% chance that the box you are currently selecting is a winner, and that there is only one other box remaining, AND (critically) that one of the two boxes is a winner, it is a concrete mathmatical / statistical fact that there is a 99% chance that the other box is a winner.

I hope this is clear now. and trust me, I sound clever now, but I argued 50/50 originally also, until it clicked.

[peterh337]

Well, you have to have 1 hour every 2 years with an instructor, in the UK. Whether that's sufficient to make a difference is another matter.

That 2-yearly flight seems to a formality in most cases. It e.g. allows a pilot who cannot read tafs and metars and who never gets notams to just carry on. ISTM that you get the logbook signed if the instructor has survived the flight :)

It is a big missed opportunity.

[abgd]

I think we're agreeing!

Yes - and it was a good point. I've enjoyed thinking about the question rather deeper than I otherwise would have.

[soaringhigh650]

Despite the shortcomings in the data used for The Killing Zone, I think most people would be inclined to agree that inexperienced pilots by hours on type are more likely to be involved in an incident than not.

Insurance companies employ actuaries to compute premiums. The best is to get yourself a few insurance quotes and play around with the different parameters to see how it affects that quote.

[SDB73]

Way off topic, but I had to post this. I watched a repeat of QI yesterday where Stephen Fry gave the following good example of extremely simple statistics which can be very easily completely misinterpreted.

It is true that Iceland has more Nobel Prize winners per capita than any other place on earth.

It is also true that Iceland has only one Nobel Prize winner!

[Genghis the Engineer]

In 1995 the CAA published CAP 667 -Review of General Aviation Fatal Accidents 1985-1994. This appeared to be a very authoritative document so based upon my own observations, I asked them what proportion of owner operators were involved in in these accidents. They had no idea. I believe on further investigation it was shown to be a substantial proportion indicating that they had omitted to consider a major factor in their analysis.

CAA publish one of those aviation safety reviews every 3-4 years and they are very thorough and useful. Each normally covers the last 10 years.

I *think* that the most recent was the 2008 CAP 780, downloadable here (http://www.caa.co.uk/docs/33/cap780.pdf) and covering 1998-2007.

You probably could analyse UK statistics to see if the alleged Killing Zone exists or not, but I doubt any other country provides a well enough documented population to find out. It would be interesting and worthwhile to do so, but would take a while - although all you'd actually need to do is look up all the individual accident reports from the AAIB website and note the captain's hours. Easy enough.

G

[abgd]

Insurance companies employ actuaries to compute premiums. The best is to get yourself a few insurance quotes and play around with the different parameters to see how it affects that quote. I'd always thought that insurance companies were the last bastion of objective political incorrectness. Kind of like Jeremy Clarkson, except with some kind of link to reality.

Car insurance? Under 21? Male? That'll be £3500 please sir.

A pension Madam? I'm afraid you'll have get much less than your brother - you're just likely to live longer, and get used to the idea of a little poverty. That's the fact of it, I'm afraid.

But when I applied for life insurance recently, I had a chat with the broker about the exact question, and he was quite dismissive of the models insurance companies used to calculate premiums for aviation. Specifically, he felt that they were poor at taking into account different mission types, and different aircraft types, which every pilot knows have very different risk profiles.

It was a source of a lot of contention between hang glider pilots and paragliders back in the day - travel insurance costs rocketed after paragliders were invented, because their pilots were considerably more likely to have accidents. Yet insurance companies at least initially lumped them into the same category.

Before paragliders were invented, hang-gliding travel insurance was cheaper than insurance for skiing. The thought occurred to me that this might not be because they were 'safer' but because it cost less to repatriate a dead body than a person with tetraplegia.

[abgd]

You probably could analyse UK statistics to see if the alleged Killing Zone exists or not, but I doubt any other country provides a well enough documented population to find out.

Where would I look for population statistics? (i.e. about the distribution of experienced / neophyte pilots who have not had accidents)

[Genghis the Engineer]

Where would I look for population statistics? (i.e. about the distribution of experienced / neophyte pilots who have not had accidents)

Try Flight Training News - they get a fair bit of info from the CAA and publish a summary every few months.

If that doesn't do the job, and you have a genuine safety research purpose, then I'd suggest contacting the CAA and asking nicely. They will certainly have this information because it's in everybody's periodic medical forms.

G

[peterh337]

However we are back to square one, because nobody has the data on flying patterns.

Sure one could analyse the AAIB reports and plot them according to commander TT, but let's say there is a peak at 500hrs. It tells us nothing unless we know the distribution of flying hours etc.

One would need to do a fair size survey to get that.

The FAA (and possibly the CAA also) have annual hours data, from the medical applications where you state your current TT. But that TT is not broken down according to aircraft type...

I think the info FTN publish is just what the CAA have on their stats website.

[Genghis the Engineer]

At risk of repeating myself - every medical revalidation I have to fill in a box stating my total hours. That goes to the CAA.

So, with the exception of the DVLA/NPPL medical, that information should all exist on a central database at Gatwick. I agree that it's not on the CAA stats website (http://www.caa.co.uk/default.aspx?catid=286&pagetype=90&pageid=11777), or at-least I can't find it. I've certainly found in the past that Gatwick are pretty helpful to safety researchers who ask questions with a clear objective.

G

[airpolice]

Well, you have to have 1 hour every 2 years with an instructor

Not if you are world champion rally driver, you just keep on flying anyway, without supervison or mentoring.

I'd be interested to see some stats on deceased/serious injury commander's admin issues.

I seem to read about more people with expired licence/medical/permit type of issues than seems reasonable. Some folk will say that having an expired medical or licence is not a contributing factor, but it may be symptomatic of an unsafe approach to aviation.

[abgd]

Sure one could analyse the AAIB reports and plot them according to commander TT, but let's say there is a peak at 500hrs. It tells us nothing unless we know the distribution of flying hours etc.

It's a very good point, but even with this information I doubt there would be sufficient UK data in order to determine anything with confidence. There were only 140 fatal GA aeroplane accidents between 1985 and 1994, for example. This would be too few for it to be valid to include many categories using ANCOVA. And I can think of loads of categories - mission (aerobatics, x-country, ifr), twins, singles, turboprop, currency, gender, age... that might be significant.

Perhaps it also reflects your situation, which is that you fly interesting and demanding x-countries in sometimes challenging conditions in a complex aircraft (yes, I spent an interesting few hours on your website). I'm interested in low airtime pilots in simple aircraft because that's me, and I'd be delighted if I could analyse only such pilots, as I agree it would make the analysis more valid. But both areas are interesting.

I would expect, though, that you wouldn't find any sharp peaks amongst private pilots. Without the tyranny of a schedule, some people will get complex or aerobatic training right away, whilst others may take much longer. In my view you'd be more likely to see gradual trends, perhaps with an initial peak +- a dip. You might find an exception to this at around 200 hours, due to the 'dropout' as modular students converted their ppls to cpls.

What you might still be able to do usefully, would be to exclude the concept of a substantial 'killing zone', or at least put an upper limit on how large the effect is. And you could potentially draw on other sources to support your interpretations.

I appreciated Big Piston's point that, at the end of the day, the important thing is to know what the most common types of accidents are, and avoid the accident du jour.

[peterh337]

As Genghis says, the CAA do have TT data (at least TT between medicals) but they don't have data on the type of flying people do.

Pilot A could do 50hrs/year flying between Goodwood and Beachy Head on sunny Sundays (as indeed great many do). Pilot B could do 50hrs on business-type trips, doing DIY approaches into Welshpool :)

Whether A or B is more likely to get killed depends on

- how careful they are (B is IMHO likely to be more careful, perhaps?)
- how well maintained (A is likely to be a renter and thus flying a less well maintained plane, but since others fly that plane most of its hours, they are more likely to have an engine failure)
- weather (B will be more at risk)
- equipment (B is likely to have better kit, e.g. GPWS)

Yes I do long trips and 100-150hrs/year so I should be more likely to get killed (especially with this (http://i101.photobucket.com/albums/m74/peterh337/alps-2010.jpg) kind of stuff, although to be fair my time actually crossing the Alps is only 1-2% of my TT) but I am quite picky about avoiding hazardous high altitude weather, and I fly IFR so don't scud run; also maintenance is money-no-object.

So one would need to do a pretty detailed flying pattern survey to make sense of what really makes people crash.

[abgd]

In a sense, we already have a fair qualitative idea of what kills pilots - flying into IMC; carb icing; engine failure at night; the impossible turn etc... One could also argue that it doesn't really make sense to worry too much about the proportions of most of those accidents, because whichever one we neglect is the one that will catch us out. One has to think about all of them.

The 'killing zone' asks a much more general question - what the time course of developing a reasonable general level of competence and judgement is. The problem it tries to address isn't so much people doing demanding and risky things, as people doing risky or stupid things through ignorance rather than intent.

[Crankshaft]

I think there's too much obsession about the fatal accident.
As said before there are simply not enough fatal accidents to make a reliable statistics on. For an accident to be fatal it has to be a rather spectacular one such as the ones agbd mentions above. Most accidents don't actually have a fatal outcome unless it's a high speed accident such as loss of control or CFIT, or there might be the odd case of very unlycky circumstances making the survival prospects very small. An engine failure daytime, even if caused by gross negligence, is normally a very survivable accident. And the most frequent accidents are very rarely fatal. Such as the mishandled crosswind landing, or the continued landing in to a wet short grass strip with too much speed, etc.

A few years ago (in the 90's) there was a survey regarding accident rates with gliders but it did not focus on fatalities, but all accidents. Compared to the large variety of different private flying operations, gliding is somewhat more consistent. Most of the flights takes place in similar weather conditions, similar type of flying, not too much variety of aircraft complexitiy and performance. (Mountain flying was excluded). The conclusion was quite similar to the "Killing Zone", that there was an increased rate of accidents at a certain level of experience. But once again, this was statistics which could be questioned just as we have done here.

[SDB73]

Crankshaft makes an excellent point.

IF there are many more non-fatal accidents than there are fatal ones, then those accidents could include horrendous injuries. Often, when considering safety we focus too much on whether we'll be killed or not.

A really irritating example of this are the muppets who take their seatbelt off too early / stand up on commercial flights. You can see them thinking "i'm hardly going to be killed if it stops abruptly", well that's true, but you could also lose an eye or have horrific facial scarring, etc, etc. Or worse still, these utter morons could cause the same to another innocent passenger.

Aside from that rant! :) ... You'd probably find that if you were able to look at all GA accidents, then once an accident becomes and accident it's only a matter of luck whether it ends up a fatal one or not. So ...

But abgd's last post says it all in my opinion :
In a sense, we already have a fair qualitative idea of what kills pilots - flying into IMC; carb icing; engine failure at night; the impossible turn etc... One could also argue that it doesn't really make sense to worry too much about the proportions of most of those accidents, because whichever one we neglect is the one that will catch us out. One has to think about all of them.

It's pointless worry about the number of hours you've had as P1 as a safety factor, as you literally cannot affect that (except not to fly into the "killing zone" at all).. instead look at what causes people to crash, and work as hard as you can to avoid the causes.

Otherwise all you're doing is saying "hold on, I'm in the killing zone, so I better be a bit more careful for 100 hours or so until I'm out of it, and can then just fly around carefree".

[abgd]

A few years back, my obsession was cycling accidents rather than flying accidents, and there was pretty much the same problem. There are just over 100 deaths a year, versus 14 fatal accidents for GA.

At least for cyclists, it's almost impossible to infer anything from 'seriously injured' data, because this includes so much, from people rendered paraplegic to things that might spoil your day but which are not life changing. Rather a lot of 'seriously injured' people don't feel that they are 'seriously injured' enough to bother going to hospital, as an example.

You end up finding that your average cycling injury happens to a younger person (child, teenager) who is doing something dumb, but who probably gets away with a trip to A&E and no lasting harm. However, your average cycling fatality is an older person, possibly cycling to work after dark and often hit from behind - an almost vanishingly rare type of accident that is also exceedingly likely to prove fatal.

In the end I realised (as had some academic researchers) that fatal accidents were a much better proxy for 'life-changing' accidents, than 'serious' accidents were. I guess that's my bias coming to aviation statistics. But now you mention it, I can see it may have less justification in aviation.

On the other hand, the example you gave of a forced landing is one that may or may not affect both inexperienced and experienced pilots equally (somebody posted the other day that 80% were due to engine mismanagement) but where I would argue that experience would tend to swing things very much in your favour, all else being equal.

I wonder what proportion of forced landings do so little damage to the plane, that it's not notifiable as an accident? If it were a low proportion, one could search the NTSB database by type (e.g. C172, PA28) and chart the proportion of engine failures that were survived, by pilot hours. This might give some indication of how pilot experience affects manoeuvring skills.

[24Carrot]

Going back to abgd's original post, there is another way of judging the significance of Fig 1.2 in Paul Craig's "The Killing Zone", and I have always meant to do it, but never got round to it till now.

In simple terms:
Dropout Rate + Accident Stats -> Conclusion

Sadly we don't know the drop-out rates, but why don't we try it backwards?
Assumptions + Accident Stats -> Dropout Rate

With suitable assumptions we can start with the accident numbers and work out the drop-out rates they imply. Then we can judge how reasonable the drop-out rates are.

I had a stab at this, and there certainly seems to be a "safety zone" during pilot training, and then there might well be some kind of a "risky zone" for the next 250 hours or so.

Here is a link to what I found (you have to scroll past the flying pig):

TwoDodecaCarrot (http://www.sites.google.com/site/twododecacarrot/)


PS For anybody who liked the Monty Hall problem, try googling: "boy born on Tuesday"

[abgd]

The first graph shows two drop-out profiles: The dashed "Raw Data" line shows the dropout rate if pilot safety is completely unrelated to pilot experience. It is obviously wrong. The initial negative drop-out rates show only 40% of pilots starting with no experience, another 40% starting at 50H experience without ever having been in the sub-50H category and a final 20% start in the 100-150H category! Thanks for doing that - it' an interesting way to look at it.

One thought is that 50 hours is a minimum for getting the ppl, so many people may still be working towards their licence at 50-80 hours, presumably remaining under instructor supervision for some time further after that.

In the USA you need 250 hours for a CPL - it would be interesting to take that into account too. I saw some training figures (on paper, not the internet) showing that in the UK, the annual number of fixed wing CPLs is now not very much less than the number of PPLs awarded each year, so despite the existence of straight-to-cpl programs, I think this is likely to be a big 'timed' effect.

I'll see if I can find the data (or ideally, USA data) again.