Safe Speed Forums

..

"The survey also reveals that an overwhelming majority – more than eight out of ten drivers (84%) – think speed limits should be varied according to weather and traffic conditions. "
That's interesting.
What would you put the relative percentage of good/bad weather accidents to be?

Good/Bad
50/50
40/60
60/40
30/70
70/30
20/80
80/20

The assumption of the IAM is that the accident ratio is high on the bad weather side is this true?

..

I think you'll find that that's what the survey indicates that the driving public thinks - not necessarily the IAM.

But the driving public doesn't really know, does it?

Why is comprehensive accident data so thin on the ground and hard to pin down?

Are the powers-that-be hiding this data from the public, and if so, why?
Or don't they know themselves?

A bit like, "we don't know what the disease is, but we have a cure for it"

Cheers
Peter

_________________
Only when ideology, prejudice and dogma are set aside does the truth emerge - Kepler

Before conducting any analysis you'd need to understand the differences in traffic behaviour in good and bad conditions.

for example the stats show that it is much safer to ride a motorcycle in january than august, but we know the risk of having an accident is higher. There are just fewer riders.

To have a proper comparison we'd need to understand how a sample group reacts to good or bad conditions and how that affected the accidents.

It's too simple to say were there more or less accidents in bad conditions. A bit like were there more or less crashes after installing a scamera

Unfortunately this survey failed to ask all the right questions.

Well, the brainwashed numpties who, for some reason seem to think that so long as they don't exceed the limit they'll be fine (wonder which organisations, agencies, partnerships etc. give them that idea...) just see the limit signs, forget about the effects of weather, and plough on regardless down L2 of the nearest motorway doing 70 in torrential rain, fog, snow etc. If having VSL would get them to slow down a bit, even if it's for the wrong reasons (we should be encouraging drivers to think for themselves and modify their speeds based on their own observation of road conditions, not just because they've seen a set of digits ringed in red), then it'd be something.


I find it interesting that you'd appear to be against the idea of reducing limits to try and reduce accidents, but then I realised that the "more than eight out of ten drivers think speed limits should be varied according to weather and traffic conditions" comment doesn't say anything there about just lowering the limits, merely that they should vary in accordance with the conditions - I suspect many of those motorists surveyed answered that question with the thought of having the limit raised when conditions were good, and of course you couldn't possibly be seen to support free-minded thinking like that, could you

This isn't the correct question to ask to establish whether it would be useful to vary speed limits according to weather conditions. Suppose the ratio were 50/50, say. If this were the case, it would most likely be because drivers are automatically equalising their risk in different conditions, taking more care in more dangerous conditions, perhaps including driving slower. So you would need to know not only how accident rates vary with weather, but how people's driving varies with weather, too, so the two things can be linked together in the analysis.

A reasonable point Stevei but why is there repeated calls for variable limits? Is it because there is increased risk in bad conditions or because there is actually more accidents in bad conditions.
Lets say that there are more accidents in poor conditions, that would indicate that drivers are not modifying their behaviour and displaying a serious disregard to the conditions. If the rate was 50/50 then that would indicate that there is a neutral or slight modification of behaviour as there is no differenc in the accident rate in conditions that would suggest a higher accident rate, i.e. bad weather.
If the rate is 30% of accidents in bad conditions and 70% of accidents occuring in good conditions then this suggests that there may be an over compensation for the bad conditions with a tendancy towards an increase in risky behaviour in good conditions.
Would you agree?
If the last case is true I would suggest that drievrs are displaying self regulation in poor conditions and there is no need for an enforced lower limit in poor weather conditions.
The situation would have to be that 70% of accidents happen in bad weather and 30% in good weather to suggest an enforced and variable limit was required.
Now back to my question, what is the actual situation regarding the percentage of accidents in good/bad conditions?

But before any of this makes sense you first need to find out what percentage of vehicle mileage is driven in bad conditions.

For example, we know that there tends to be a sharp rise in motorcycle fatalities when we have a good summer. Using your logic we'd conclude that riding motorcycles is more dangerous in good weather!

_________________
CSCP Latin for beginners...
Ticketo ergo sum : I scam therefore I am!

Maybe its because South Wales police helpfully put 'dangerous weather conditions' on their next matrix signs when its raining.

I agree with all of this with two exceptions:

1. I suspect people in some cases want an increased speed limit in good conditions rather than a reduced speed limit in bad conditions. Perhaps they see variable speed limits as a way to get limits that they see as more realistic in good conditions without encouraging lunatic driving in bad conditions. It also can only be a good thing to encourage people to see the need for a link between speed and conditions.

2. As JT has said, we need to correct for mileage in good and bad conditions, so it is the relative accident rate per vehicle mile that should be compared. Once that adjustment is made, though, I totally agree with the inferences drawn from whether that rate is higher or lower in good conditions in terms of over / under compensating.

A few more thoughts on this - you would actually need to do more than just correct for mileage, what would be needed would be to sample traffic at various locations in good and bad conditions, as well as sampling environmental factors at those same locations. For example:

- Different sectors of the motoring population may make different choices as to whether to drive in bad conditions, e.g. a mother taking her children to school may be more likely to drive them in bad weather, a motorcyclist may be less likely to ride in bad weather. We would need to establish the makeup of the driving population in good and bad conditions, and evaluate the accident rate per mile for each specific group in good and bad conditions.

- The environment might change, e.g. in bad weather, pedestrians and cyclists may choose not to go out, and this might actually improve the environment in bad conditions from an accident risk perspective.

- When people talk about variable speed limits in good and bad conditions, they don't just mean weather, they mean other factors as well, e.g. have lower speed limits near schools only when there is school-related activity. This would clearly be preferable to having humps etc to reduce speeds at all times.

of course we already had a variable limit some time ago - it was called the national speed limit. On many of the roads, this limit was simply unachievable. hence drivers chose an appropriate speed. The police could still prosecute dangerous driving and the mostorist actually worked out the appropriate speed rather than relying on the speedo.

sadly the NSL is now mostly history and in its place we see 30, 40 and 50mph speed limits set according to environmental, political and economical needs.

The net result is widespread confusion and frustration. In some cases at some times during the day a 30 limit may be appropriate. For example the road by chessington world of advantures has a 30 limit. when people are queuing to get in or out the safe speed would probably be nearer 20. At all other times 50mph feels slow. An ideal candidate for a variable limit.

I don't think that relative mileage is significant in these considerations at all.

Motorcyclists chose not to venture out in wet or freezing conditions for several reasons. One may be that they don't like to get wet and cold, I know I don't, but the main reason is that it is more dangerous to ride on 2 wheels in wet and freezing conditios so motorcyclists "don't choose to commit suicide", quoted directly from an IAM motorcylist.

To raise the speed limit in good (apparent) conditions would be to raise the risk of collision and injury unacceptably. Why is it OK to raise the speed limit by 17% say but not to accept driving at a blood alcohol level at the drink driving prosecution threshold when the change is risk is identical? Source WHO.

i can't see why there is a major call for speed cameras aoutside of schools, are there loads of accidents happening there or is it that the potential victims are somehow more precious and emotive?

..

Of course relative mileage is significant. Motorcyclists do far less mileage in bad conditions, for the reasons quoted above.
But that tells us nothing about the ratio of crashes/mile travelled in bad weather compared to crashes/mile travelled in good weather.



Any chance of a link to that source? That 17% figure makes it sound ominously like a piece of Aussie 'research' I've seen - in which case it's unmitigated nonsense.



Because they're so brainwashed into the 'speed kills' mentality, and they want their little darlings to be safe, don't they?
What they should be doing is educating said little darlings in how to cross the road safely, rather than baying for speed cameras.

Cheers
Peter

_________________
Only when ideology, prejudice and dogma are set aside does the truth emerge - Kepler

So why is 20mph acceptable? This is double 10mph, so why are all 20mph limits not reduced to 10mph? But 10mph is double 5mph, so why not reduce them further, etc. Each of these changes, from 5 to 10, 10 to 20, by your argument must be increasing risk by the same amount as being very drunk indeed.


It certainly seems to be perceived as a major problem - does anyone know what percentage of fatal accidents and serious injuries involve schoolchildren travelling to or from school?


Right, onto the major issue, time for a crash course in statistical analysis. Let me consider something much simpler than speed to illustrate the techniques, which incidentally are universally accepted, and standard practice in clinical trials of drugs, for example. I will consider how colour of car affects accident risk.

Suppose I tell you that 80% of accidents involve red cars. What would you conclude from this? Some might think that driving a red car gives you a high risk of having an accident, but the correct answer is that nothing at all can be concluded from this.

Suppose I further tell you that 90% of cars are red. What can you now conclude? The temptation is to say that driving a red car reduces your risk of accident, however the correct answer is still that nothing at all can be concluded.

Suppose I further tell you that 95% of red cars are never driven by their owners, and remain in their garage. What can you conclude now? The temptation is to say that only 0.05x90% of cars on the road are red, and that driving a red car therefore increases your risk of accident, however the correct answer is that you can still conclude nothing at all.

Suppose I further tell you that 99% of all cars that aren't red are never driven by their owners, and remain in their garage. What can we conclude now? This would mean that 98% of the cars on the road are red, so we might now reverse our judgement again, and believe that driving a red car gives you a lower risk of having an accident, but the correct answer is still that you can conclude nothing at all.

Suppose I now tell you that there are some roads that are incredibly dangerous, and some roads are incredibly safe. I'll stop giving numbers at this point, but the issue is that people who drive red cars might be massively more likely to drive on one or the other type of road, and I could concoct suitable numbers to still swing the result one way or the other.

The solution to all this confusion is to adopt the standard practice of a control group. The idea of a control group is that it differs from the test group in only one respect. In this case we would need to have two groups where the only difference is the colour of car, everything else about them must be identical. For clinical trials of drugs, this is achieved by having, say, three groups that are statistically identical, one will receive the new drug, one an existing drug, and one a placebo.

To gather similar data for cars is incredibly difficult. We certainly need to sample from cars on the road. We even need to sample from accident locations at the time of accidents. This implies having permanent data gathering in place, so that when an accident does occur, we have data for the period around when the accident occurred. Even this data won't be perfect, for example we won't know the occupations of people driving the vehicles unless we are constantly stopping people and asking them a set of questions to profile them, and insurance companies seem to think your occupation affects your accident risk, as does the age of the driver, for example.

Without all this data, you can't conclude anything at all for certain. You can guess that some of these factors aren't varying much, but your guess might be wrong. I have seen studies done where this sort of depth of analysis was undertaken, but they seem to be exceedingly rare when speed is being studied. Some analyses seem to fail at the first hurdle, for example the statement that "one third of accidents are caused by excessive speed for the conditions" might not be incorrect in itself, but if the inference is then drawn from that, that speed is a problem, then that is like concluding red cars are dangerous given only the information that 80% of accidents involve red cars.

And, at the end of the day, after you've sorted out your control groups and eliminated all sources of error, you establish a correlation between red cars and accidents. But that still tells us nothing, because correlation is not causation. In order to establish causation, there has to be a plausible mechanism And I reckon you'd search forever for any mechanism by which red cars cause accidents.

Cheers
Peter

_________________
Only when ideology, prejudice and dogma are set aside does the truth emerge - Kepler

Yes, this is an interesting point, and I was thinking of going on to cover correlation not being causation. I'm going to be cautious and say that I'm not 100% certain on this point, but I think that if correlation is not causation, then it means that you have failed to identify and correct for another difference.

Here's an example - suppose we find that men who drive fast cars are more likely to be bald. Fast cars correlated with baldness. Can you avoid becoming bald by not driving a fast car? No, because the two factors probably share a common cause - higher than average testosterone levels. So we would only get the misleading result if we failed to analyse the testosterone levels of our two groups - bald men, and men who drive fast cars. The two effects would still be correlated, but not if you study only people with a fixed testosterone level, you would then find no correlation within that group. If you did find a correlation within that group, there is either a causative link (e.g. the adrenaline rush of driving a fast car causes hair loss), or you have missed another factor that is an underlying cause of both things. Of course there can be so many underlying reasons for all sorts of things that it is very hard indeed to be certain of when you have established a causative link vs a mere coincidental correlation.

Yes, I see your point - although very often if you establish a correlation between x and y, having eliminated all other influences, you're still not sure whether x causes y or y causes x - unless you have a plausible mechanism.
And the lack of mechanism is very evident in most speed vs accidents studies, and very weak in the studies which purport to show one. It's simply taken as a 'given' and no attempt is made to show a quantitative or qualitative relationship between the mechanism and the results.
Putting it another way, if the mechanism shows a risk ratio of 2:1 and the stats show a risk ratio of 4:1 then there's something wrong with either the mechanism, the stats or both.

If the above is about as clear as mud, it's probably because I've had a few jars.

Cheers
Peter

_________________
Only when ideology, prejudice and dogma are set aside does the truth emerge - Kepler