Road accident stat trends - an alternative theory



Status
Not open for further replies.
S

Simon Proven

Guest
(alternative to Paul's Smith's "it's the focus on speed" theory)

I read an article in New Scientist this week about the dangers of SUVs (in particular, they're more
likely to roll than ordinary cars, and more likely to have the roof crushed in a roll). They also
mentioned in the article that occupants of normal cars are 3x more likely to be killed in an
accident with an SUV than in an accident with a normal car. Heaven knows what they do to peds,
cyclists, etc.

Now, SUVs (vehicles of this type, anyway - the term isn't normally used here) are far less common
over here, but still account for 4% of the vehicle market.

Given their acknowledged safety issues, I was wondering how the numbers of this type of vehicle have
changed over the past 10 years and if an increase in collisions involving such behemoths could
account for any change in trend. Note that for the purposes of this discussion I'm assuming that
there might be a change in trend, though if you look at PS's graphs there have been other occasions
where there has been a brief plateau before continuing the downward trend.

Simon
 
On Wed, 12 Mar 2003 01:11:34 +0000, Simon Proven <[email protected]> wrote:

>(alternative to Paul's Smith's "it's the focus on speed" theory)

>I read an article in New Scientist this week about the dangers of SUVs (in particular, they're more
>likely to roll than ordinary cars, and more likely to have the roof crushed in a roll). They also
>mentioned in the article that occupants of normal cars are 3x more likely to be killed in an
>accident with an SUV than in an accident with a normal car. Heaven knows what they do to peds,
>cyclists, etc.

>Now, SUVs (vehicles of this type, anyway - the term isn't normally used here) are far less common
>over here, but still account for 4% of the vehicle market.

>Given their acknowledged safety issues, I was wondering how the numbers of this type of vehicle
>have changed over the past 10 years and if an increase in collisions involving such behemoths could
>account for any change in trend. Note that for the purposes of this discussion I'm assuming that
>there might be a change in trend, though if you look at PS's graphs there have been other occasions
>where there has been a brief plateau before continuing the downward trend.

I'll be delighted to look into the idea, but I don't think there are anywhere near enough SUVs to
account for the changes. Then their occupant safety is good, which would tend to balance out some of
the other factors. The cigarette packet approach suggests 4% * 2x danger = 8% of fatalities (don't
forget to subtract 4% of fatalities from the previous total, because the SUVs are switched from
other vehicles) so we get:

3500* .96 = 3360 3360* 1.08 = 3629 3629-3500 = 129.

Or something like 10% of the figure we're looking for. I can't see anything that could be anywhere
near a factor ten error in the quick calculation.

No previous plateau comes anywhere close to what we now have. Most previous plateaus are noise in
the figures - i.e. there's a balancing falling trend nearby. Note also that later figures are less
"noisy" because traffic increases (=bigger populations) help smooth the numbers.

btw, I hate the trend towards ridiculous SUVs.
--
Paul Smith Scotland, UK http://www.safespeed.org.uk please remove "XYZ" to reply by email speed
cameras cost lives
 
"Paul Smith" <[email protected]> wrote in message
news:[email protected]...
> On Wed, 12 Mar 2003 01:11:34 +0000, Simon Proven <[email protected]> wrote:
>
> >(alternative to Paul's Smith's "it's the focus on speed" theory)
>
> >I read an article in New Scientist this week about the dangers of SUVs (in particular, they're
> >more likely to roll than ordinary cars, and more likely to have the roof crushed in a roll). They
> >also mentioned in the article that occupants of normal cars are 3x more likely to be killed in an
> >accident with an SUV than in an accident with a normal car. Heaven knows what they do to peds,
> >cyclists, etc.
>
> >Now, SUVs (vehicles of this type, anyway - the term isn't normally used here) are far less common
> >over here, but still account for 4% of the vehicle market.
>
> >Given their acknowledged safety issues, I was wondering how the numbers of this type of vehicle
> >have changed over the past 10 years and if an increase in collisions involving such behemoths
> >could account for any change in trend. Note that for the purposes of this discussion I'm assuming
> >that there might be a change in trend, though if you look at PS's graphs there have been other
> >occasions where there has been a brief plateau before continuing the downward trend.
>
> I'll be delighted to look into the idea, but I don't think there are anywhere near enough SUVs to
> account for the changes. Then their occupant safety is good, which would tend to balance out some
> of the other factors. The cigarette packet approach suggests 4% * 2x danger = 8% of fatalities
> (don't forget to subtract 4% of fatalities from the previous total, because the SUVs are switched
> from other vehicles) so we get:
>
> 3500* .96 = 3360 3360* 1.08 = 3629 3629-3500 = 129.
>
> Or something like 10% of the figure we're looking for. I can't see anything that could be anywhere
> near a factor ten error in the quick calculation.
>

I think you'll find you've missed out all the dark SUV's.
 
On Wed, 12 Mar 2003 03:00:29 +0000, Paul Smith <[email protected]> wrote: <snip>
>>also mentioned in the article that occupants of normal cars are 3x more likely to be killed in an
>>accident with an SUV than in

<snip>

> 3500* .96 = 3360 3360* 1.08 = 3629 3629-3500 = 129.
>
> Or something like 10% of the figure we're looking for. I can't see anything that could be anywhere
> near a factor ten error in the quick calculation.
>

SUV's have replaced 4% of the market so assume that, had those drivers been in normal cars they
would have accounted for 4% of the fatalities:

3500*.04 = 140.

SUVs are three times more likely to kill so new fatalities = 140*3 = 420.

Change = 280.

(In my experience SUV drivers are far more likely to be the aggressive types on the road - I
don't know whether the car makes the driver aggressive or the aggressive driver selects the car
but it could well be that the SUV drivers would have accounted for 8% of the fatalities making
the change 560)

Of course, we should be starting from where we should have been. However, I don't agree with your
estimates but even so here are my plucked out of the air figures:

SUVs
2500x.08 = 200 200x3 = 600 change = 400.

Mobile phones 4% risk increase 5x
2500y.04 = 100 100x5 = 500 change = 400

25% increase in motorway traffic which has lead to congestion on most motorways at certain times.
(In my first year - 2002 in this job I recall the M1 being shut once sufficently locally to affect
people who work here. In the last 6 months I know of two occasions including in both directions on
one occasion)

So lets say motorways have got 4x more dangerous and account for 2% of fatals (I don't know the 2%
but I'm sure you have it at your fingertips so I can't be bothered to look it up)

2500x0.02 = 50 50x4 = 200 change = 150

Total change 400+400+150 = 950.

Expected fatals = 2500+950 = 3450.

Well within what I would anticipate was a reasonable estimate for noise.

Now I think your extrapolation is wildly optimistic but lets assume it is only a factor of two too
optimistic:

2500z.04x3 = 360 (ignore my 2x for aggressive drivers)
2500z.04x5 = 600
2500z.02x4 = 240

Total = 1200

3000 + 1200 = 4200

So speed cameras are saving 700 lives a year. And I've accepted a handicap of 360 in my
calculation as well.

Of course I'm not putting forward this as a rigourous and robust defence of speed cameras. However,
it is at least as logical and consistent as your arguments against them. Picking holes in my
arguments doesn't make your argument right, until and unless you answer the criticisms addressed at
your pet theory. You need to show a: that your theory is robust against all criticism and b: that
there is no other theory that can adequately explain what your theory explains. i.e. in this case
you need to show that there really are 1000 lives a year being unnecessarily lost (you haven't) and
b: that speed cameras are the only possible cause for this (you haven't) (Personally I feel the both
the "you haven't" should be "one can't" but there are a huge number of people much more clever than
me and maybe there are techniques that can be brought to bear than I haven't even heard of, let
alone understand, and which other knowledgeable people will agree are sound.)

> No previous plateau comes anywhere close to what we now have. Most previous plateaus are noise in
> the figures - i.e. there's a balancing falling trend nearby. Note also that later figures are less
> "noisy" because traffic increases (=bigger populations) help smooth the numbers.
>
Fewer statistics = more noise not less. One major coach/bus[1] crash could add 2% to the K figures
for one year now. In 1950 it was less than 1%

Regards,

Tim.

[2501] Petrol tanker hits a coach and explodes sort of thing.

--
God said, "div D = rho, div B = 0, curl E = - @B/@t, curl H = J + @D/@t," and there was light.

http://tjw.hn.org/ http://www.locofungus.btinternet.co.uk/
 
"Paul Smith" <[email protected]> wrote in message
news:[email protected]...

> anywhere near enough SUVs to account for the changes. Then their occupant safety is good, which
> would tend to balance out some of the other factors.

But be ****** away with risk compensation, which will affect anyone in sensible cars as well as
bikes and on foot. (or on trains for that matter!)
 
Paul Smith <[email protected]> wrote in message
news:<[email protected]>...
> On Wed, 12 Mar 2003 01:11:34 +0000, Simon Proven <[email protected]> wrote:
>
> >(alternative to Paul's Smith's "it's the focus on speed" theory)
>
> >I read an article in New Scientist this week about the dangers of SUVs (in particular, they're
> >more likely to roll than ordinary cars, and more likely to have the roof crushed in a roll). They
> >also mentioned in the article that occupants of normal cars are 3x more likely to be killed in an
> >accident with an SUV than in an accident with a normal car. Heaven knows what they do to peds,
> >cyclists, etc.
>
> >Now, SUVs (vehicles of this type, anyway - the term isn't normally used here) are far less common
> >over here, but still account for 4% of the vehicle market.
>
> >Given their acknowledged safety issues, I was wondering how the numbers of this type of vehicle
> >have changed over the past 10 years and if an increase in collisions involving such behemoths
> >could account for any change in trend. Note that for the purposes of this discussion I'm assuming
> >that there might be a change in trend, though if you look at PS's graphs there have been other
> >occasions where there has been a brief plateau before continuing the downward trend.
>
> I'll be delighted to look into the idea, but I don't think there are anywhere near enough SUVs to
> account for the changes. Then their occupant safety is good, which would tend to balance out some
> of the other factors.

"And in terms of fatalities from all types of crash, SUVs are slightly more unsafe than passenger
cars. In 1999, for every 100,000 registered vehicles in the US, there were 16.4 fatalities in
passenger cars, and
17.8 in SUVs. "

Though these figures are for the US, we can't assume that SUVs are safer.

> The cigarette packet approach suggests 4% * 2x danger = 8% of fatalities (don't forget to
> subtract 4% of fatalities from the previous total, because the SUVs are switched from other
> vehicles) so we get:

More sloppy maths. You've subtracted 4% from the total then failed to add them back in again at the
end, creating a large error. Also, you've made an erronous assumption that SUVs are safe for their
occupants.

> 3500* .96 = 3360 3360* 1.08 = 3629 3629-3500 = 129.

What you could have done is:

3500 * 0.96 = 3360 (non-SUV risk) 3500 * 0.04 = 140 (SUV risk)

3360 * 1.12 = 3763 (increased risk to non-SUV occupants) 140 * 1.10 = 154 (increased risk to people
switching to SUVs)

3917 - 3500 = 417

Which is a substantial part of your observed deviation from trend. Note: there are potential
problems with my maths, too. Can you work out what they are?

The true figure is likely to be inbetween the two. Certainly cause for questioning your speed
hypothesis.

Simon
 
[email protected] (Simon Proven) wrote in message
news:<[email protected]>...

> 3500 * 0.96 = 3360 (non-SUV risk) 3500 * 0.04 = 140 (SUV risk)
>
> 3360 * 1.12 = 3763 (increased risk to non-SUV occupants)

Erk. Should be 1.08. Giving 3628.

> 140 * 1.10 = 154 (increased risk to people switching to SUVs)

> 3917 - 3500 = 417

For a total of 268.

> Which is a substantial part of your observed deviation from trend. Note: there are potential
> problems with my maths, too. Can you work out what they are?

(Apart from the 1.12 vs 1.08 mistake :)

The overall point I'm making is that you've basically discounted all other plausible explanations
and have concentrated on the "speed kills" campaign as the sole reason. Other factors to include are
mobile phones, the increase in road haulage, and perhaps a slowing in the rate of medical advances
turning what would have been a K in the past into an SI.

Simon
 
On 12 Mar 2003 08:10:10 -0800, [email protected] (Simon Proven) wrote:

>For a total of 268.

>> Which is a substantial part of your observed deviation from trend. Note: there are potential
>> problems with my maths, too. Can you work out what they are?

>(Apart from the 1.12 vs 1.08 mistake :)

>The overall point I'm making is that you've basically discounted all other plausible explanations
>and have concentrated on the "speed kills" campaign as the sole reason. Other factors to include
>are mobile phones, the increase in road haulage, and perhaps a slowing in the rate of medical
>advances turning what would have been a K in the past into an SI.

But previous trends are continuing in other western countries. (e.g. Germany, France and Italy.) I'm
quite sure they have mobile phones, SUVs and "other factors" too.

I've been able to eliminate some of the other potential factors, but I confess to not being all that
motivated to prepare and present irrelevant information.

Mobile phones: RoSPA trawled 12 years of Coroners reports and found 19 cases. Deaths are well
investigated and to suggest that 4,500+ deaths were missed is really stretching credibility too far.

SUVs: I'm watching with interest and so far have made a few efforts t gather figures.

Medical advances: constant input: constant improvement I reckon.

But with any of these factors we'd expect to see a worldwide loss of trend. So far I've only found
it in Norway, Sweden, Belgium, Australia and the UK. And these are the "worst" countries for
"speed kills".
--
Paul Smith Scotland, UK http://www.safespeed.org.uk please remove "XYZ" to reply by email speed
cameras cost lives
 
Paul Smith wrote:
> On 12 Mar 2003 08:10:10 -0800, [email protected] (Simon Proven) wrote:
>
>
>>For a total of 268.
>
>
>>>Which is a substantial part of your observed deviation from trend. Note: there are potential
>>>problems with my maths, too. Can you work out what they are?
>
>
>>(Apart from the 1.12 vs 1.08 mistake :)
>
>
>>The overall point I'm making is that you've basically discounted all other plausible explanations
>>and have concentrated on the "speed kills" campaign as the sole reason. Other factors to include
>>are mobile phones, the increase in road haulage, and perhaps a slowing in the rate of medical
>>advances turning what would have been a K in the past into an SI.
>
>
> But previous trends are continuing in other western countries. (e.g. Germany, France and Italy.)
> I'm quite sure they have mobile phones, SUVs and "other factors" too.

The case of germany has already been addressed - they're suffering from death rates similar to what
we had 20 years ago. It's hardly surprising that they are achieving larger improvements through road
safety policy.

> I've been able to eliminate some of the other potential factors, but I confess to not being all
> that motivated to prepare and present irrelevant information.

If you want to be convince people of its irrelevance, then you have to *show* that it is irrelevant.

> Mobile phones: RoSPA trawled 12 years of Coroners reports and found 19 cases. Deaths are well
> investigated and to suggest that 4,500+ deaths were missed is really stretching credibility
> too far.

Strawman.

> Medical advances: constant input: constant improvement I reckon.

You reckon. Well that proves it then. It's *not* a fundamental law that accident rates must drop
over time. The same goes for medical advances. Saving the first n lives is much easier than saving
the next n. Notice how with massive increases in spending on health, worldwide, cancer survival
rates are only improving very slowly.

> But with any of these factors we'd expect to see a worldwide loss of trend.

Why?

In any case, the statistical evidence for your loss of trend in this country is very weak. It's also
interesting that you consider a lower number of fatalities to result in less noise. Oops!

Simon
 
"Simon Proven" <[email protected]> wrote in message
news:[email protected]...
>
> (alternative to Paul's Smith's "it's the focus on speed" theory)
>
> I read an article in New Scientist this week about the dangers of SUVs (in particular, they're
> more likely to roll than ordinary cars, and more likely to have the roof crushed in a roll). They
> also mentioned in the article that occupants of normal cars are 3x more likely to be killed in an
> accident with an SUV than in an accident with a normal car. Heaven knows what they do to peds,
> cyclists, etc.

http://www-nrd.nhtsa.dot.gov/departments/nrd-11/aggressivity/980908/980908.h tml // Light trucks and
vans (LTVs) currently account for over one-third of registered U.S. passenger vehicles. Yet,
collisions between cars and LTVs account for over one half of all fatalities in light
vehicle-to-vehicle crashes. //

http://www.bts.gov/btsprod/nts/AppA_web/Auto99R.htm // Average miles traveled per vehicle Passenger
car and motorcycle 14,404 miles Other 2-axle 4-tire vehicle 12,173 miles //

We know they cause more casualties in a crash. The mileage proportion makes things worse. Any number
that can be conjured up for additional casualties in actual crashes does not take into account that
additional crashes may also have happened.

David Roberts
 
DR wrote:

[interesting quotes snipped]

Thanks for those.

> We know they cause more casualties in a crash. The mileage proportion makes things worse. Any
> number that can be conjured up for additional casualties in actual crashes does not take into
> account that additional crashes may also have happened.

Indeed. What I was trying to illustrate was that the assumption that "speed kills" was the sole, or
major cause of a slowing in the decline of fatals is highly suspect.

Job done, I think.
 
On Wed, 12 Mar 2003 21:05:24 +0000, Simon Proven <[email protected]> wrote:

>In any case, the statistical evidence for your loss of trend in this country is very weak. It's
>also interesting that you consider a lower number of fatalities to result in less noise. Oops!

You didn't read what I wrote:

"Note also that later figures are less "noisy" because traffic increases (=bigger populations) help
smooth the numbers."
--
Paul Smith Scotland, UK http://www.safespeed.org.uk please remove "XYZ" to reply by email speed
cameras cost lives
 
"Paul Smith" <[email protected]> wrote in message
news:[email protected]...

> But previous trends are continuing in other western countries. (e.g. Germany, France and Italy.)
> I'm quite sure they have mobile phones, SUVs and "other factors" too.

They have higher death rates though.

Prepare yourself for a big shock Mr. Smith. I heard on the radio yesterday (french), that they have
witnessed a 35% decrease in road accidents in the last two months.
 
On Thu, 13 Mar 2003 10:09:02 -0000, "W K" <[email protected]> wrote:

>> But previous trends are continuing in other western countries. (e.g. Germany, France and Italy.)
>> I'm quite sure they have mobile phones, SUVs and "other factors" too.

>They have higher death rates though.

>Prepare yourself for a big shock Mr. Smith. I heard on the radio yesterday (french), that they have
>witnessed a 35% decrease in road accidents in the last two months.

That's an absurd claim, with nothing similar available anywhere in the history of motoring. They
must have been talking about something narrow and specific, or they just misreported the facts.
--
Paul Smith Scotland, UK http://www.safespeed.org.uk please remove "XYZ" to reply by email speed
cameras cost lives
 
> Mobile phones: RoSPA trawled 12 years of Coroners reports and found 19 cases. Deaths are well
> investigated and to suggest that 4,500+ deaths were missed is really stretching credibility
> too far.

I heard on the radio just the other day that road accidents where mobile phones were in use at the
time were very much under reported.

It wasn't a requirement that use or not was investigated, it was only
reported if it was obvious at the time that the phone was being used.

So it may be that phones are responsible for the lack of improvement in the death rate but
they're just not being counted because the police just don't have the time/inclination/resources
to follow it up.

BigRAb
 
W K <[email protected]> wrote:
>
> Prepare yourself for a big shock Mr. Smith. I heard on the radio yesterday (french), that they
> have witnessed a 35% decrease in road accidents in the last two months.

That quite clearly correlates with the French decision to oppose war in Iraq. If we can get Mr
Blair to take a similar attitude to the French perhaps we will see the same reduction in road
accidents ;-).

Tony

PS. http://www.coxar.pwp.blueyonder.co.uk/ is well worth a visit.
 
On 13 Mar 2003 03:24:28 -0800, [email protected] (Robert McDonald) wrote:

>> Mobile phones: RoSPA trawled 12 years of Coroners reports and found 19 cases. Deaths are well
>> investigated and to suggest that 4,500+ deaths were missed is really stretching credibility
>> too far.

>I heard on the radio just the other day that road accidents where mobile phones were in use at the
>time were very much under reported.

> It wasn't a requirement that use or not was investigated, it was only
> reported if it was obvious at the time that the phone was being used.

>So it may be that phones are responsible for the lack of improvement in the death rate but
>they're just not being counted because the police just don't have the time/inclination/resources
>to follow it up.

In fatals? Really? They found mobile phones in just 0.4% of cases where the mobile phone was a
factor? (19/4,500)

And why hasn't the same happened in France, Germany and Italy?
--
Paul Smith Scotland, UK http://www.safespeed.org.uk please remove "XYZ" to reply by email speed
cameras cost lives
 
"Paul Smith" <[email protected]> wrote in message
news:[email protected]...
> On Thu, 13 Mar 2003 10:09:02 -0000, "W K" <[email protected]> wrote:
>
> >> But previous trends are continuing in other western countries. (e.g. Germany, France and
> >> Italy.) I'm quite sure they have mobile phones, SUVs and "other factors" too.
>
> >They have higher death rates though.

Ah, no dispute on that one.

> >Prepare yourself for a big shock Mr. Smith. I heard on the radio
yesterday
> >(french), that they have witnessed a 35% decrease in road accidents in
the
> >last two months.
>
> That's an absurd claim, with nothing similar available anywhere in the history of motoring. They
> must have been talking about something narrow and specific, or they just misreported the facts.

Its not absurd, and can be well within "noise" factors. sampling etc. One or two weather incidents,
or good weather in Jan/Feb could easily make it 35% better than usual.

I'm not holding out for the above figures to be a real long term effect, but I'm sure you'd quote
something like that if it went in your favour.
 
Paul Smith wrote:
> On Wed, 12 Mar 2003 21:05:24 +0000, Simon Proven <[email protected]> wrote:
>
>
>>In any case, the statistical evidence for your loss of trend in this country is very weak. It's
>>also interesting that you consider a lower number of fatalities to result in less noise. Oops!
>

> You didn't read what I wrote:

Considering only one possible explanation for observed events again?

> "Note also that later figures are less "noisy" because traffic increases (=bigger populations)
> help smooth the numbers."

The other explanation might be that I removed the Smith Spin and restated the situation.

Simon
 
Paul Smith wrote:
> On 13 Mar 2003 03:24:28 -0800, [email protected] (Robert McDonald) wrote:
>
>
>>>Mobile phones: RoSPA trawled 12 years of Coroners reports and found 19 cases. Deaths are well
>>>investigated and to suggest that 4,500+ deaths were missed is really stretching credibility
>>>too far.
>
>
>>I heard on the radio just the other day that road accidents where mobile phones were in use at the
>>time were very much under reported.
>
>
>>It wasn't a requirement that use or not was investigated, it was only reported if it was obvious
>>at the time that the phone was being used.
>
>
>>So it may be that phones are responsible for the lack of improvement in the death rate but
>>they're just not being counted because the police just don't have the time/inclination/resources
>>to follow it up.

> In fatals? Really? They found mobile phones in just 0.4% of cases where the mobile phone was a
> factor? (19/4,500)

Yawn. That's the same strawman you used yesterday.

> And why hasn't the same happened in France, Germany and Italy?

Perhaps because of all the other reasons that you've failed to consider as possibilities:

Their worse starting point, meaning that any upwards effect from things such as SUVs, mobile phones,
your pet theory, etc, might well be swamped by a larger downtrend.

Different culture; for instance the fact that driving is less safe in France than here, but cycling
is much safer here than in France, illustrating that a simplistic comparison of death rates hides
what's really going on.

Different levels of mobile phone penetration into the market. Different levels of SUV penetration
into the market. Etc, etc, ad infinitum.

Simon
 
Status
Not open for further replies.

Similar threads

P
Replies
3
Views
379
Mountain Bikes
Crescentius Vespasianus
C
B
Replies
2
Views
632
UK and Europe
Wafflydirtycatl
W