Seat belts and helmets

[StuartG]

I would have thought a significant reduction is deaths/mile must have generated by the huge motorway construction from 1959 until the eighties. Long distance traffic was largely shifted from relatively dangerous fast A roads to the much safer motorways. And away from cyclists and pedestrians. At the same time it would probably have increased crash speeds making seatbelts more effective.

At the same time crumple zones and safer interiors were making it harder to die. A lot was going on which can make it difficult to pin down what was responsible for what. Indeed I think its probably the greater congestion and hence slower speeds which are keeping the death rate down these days. Its all very complicated and perceived risk is not the same as real risk which confounds Adams' risk compensation theory.

 

[srw]

Do you think you can discern a reduced risk attributable to the step change in seat belt use in 1983, as opposed to the general decline in risk with time?

My response was in the context of this bit:

Thus, seat belts, ABS*, better road surfaces etc have all been seamlessly absorbed, presumably leading to faster driving or taking less care, but not significantly impacting on risk; and logically, spikes in steering wheels would be expected to lead to slower journeys but no great lasting change to risks for drivers.

My point is that those things have not been seamlessly absorbed but that they have all, along with what Stuart identifies, contributed to a general reduction in risk for drivers.

I'm all for testing hypotheses (I'm currently sitting in an EU scientific meeting in Athens where the current speaker must have mentioned "hypothesis" twenty times so far). But the test is not "has there been an increase or decrease in cycling risks over time?", the test is surely "do the statistics on cycling/pedestrian risk show any change associated with the step change in seat belt use in 1983?"

Like Stuart, I don't think there was a step change in seat belt use in 1983, and I don't think you can separate out any particular safety feature. I'm on a crappy tablet so I'm not going to go data-hunting, but my recollection is that there has been a significant drop in cycling casualties over the last few decades roughly in line with the significant drop in motorised vehicle casualties you started the thread with. If so, then the conclusion that making things safer for drivers makes them worse for non-drivers is erroneous.

 

[snorri]

g, but my recollection is that there has been a significant drop in cycling casualties over the last few decades roughly in line with the significant drop in motorised vehicle casualties you started the thread with. If so, then the conclusion that making things safer for drivers makes them worse for non-drivers is erroneous.

I would have thought the fall in casualty rates for cyclists and pedestrians was due to the reduction in cycle and pedestrian traffic on our rural roads which is where most of the fatal crashes occur?

 

[srw]

I would have thought the fall in casualty rates for cyclists and pedestrians was due to the reduction in cycle and pedestrian traffic on our rural roads which is where most of the fatal crashes occur?

What fall? You're making a heck of an assumption there. The data shows that cycling rates have gone up a lot in the last few years particularly - and, guess what, a lot of that will be on rural roads.

 

[buggi]

But your graph shows that there has been a significant change to risk for drivers...

I'm not going to go searching now, but my memory is there has also been a commensurate reduction in deaths of cyclists and pedestrians.

Yours is an easy hypothesis to test, and the data is easily available. If making driving safer for drivers has made it more dangerous for others, then when you divide pedestrian and cyclist deaths by billion vehicle kilometres you ought to see an upward trend. I don't think you will.

Not according to this.. http://www.roadsafetyobservatory.com/Summary/riders/pedal-cyclists

 

[snorri]

What fall? You're making a heck of an assumption there. The data shows that cycling rates have gone up a lot in the last few years particularly - and, guess what, a lot of that will be on rural roads.

It was you who mentioned a "significant drop in cycling casualties", perhaps I have misinterpreted.
Please point me to the data regarding numbers cycling on rural roads, I can't find separate data for rural and urban.

 

[david k]

I appreciate many drive to the limits of the car, therefore more safety advancements and increased speed cancels them out, but how many does this relate to? My Mum drives just as slow now as she always has done!

 

[swansonj]

I don't have an agenda either way on whether seat belts are effective. I came across John Adams' arguments and statistics a few years ago and thought it would be interesting to bring the critical resources of cyclechat to bear.

But I would say this. Like, I suspect, many cyclists, I came to the cycle helmet debate with an assumption that helmets reduce risk (it's only common sense, isn't it); it was only when I started looking at the data with an open mind that I realised they don't. And now, like I suspect many cyclists, I get annoyed by other people who are so determined that helmets must be effective (it's only common sense, isn't it) that they won't address the data. I think we have an obligation to bring the same open, data-driven approach to seat belts.

My response was in the context of this bit:

OK, sorry, I misunderstood.

AFAIR there was no step change. Seatbelts were introduced and then made mandatory on new cars. So it took around 10/15 years for most of the nation's car fleet to have drivers who could belt up. The usage of belts grew voluntarily prior to mandation which, in any case only applied to cars with seatbelts. Arguably Jimmy (cough) Savile did more with the clunk click campaign a decade earlier. Later auto adjustable seatbelts would become more effective than the earlier ones. We didn't suddenly go from a non-seatbelt to a seatbelt world. Or from two seats to all seats. It was much smoother than that.

I agree - it is obvious - that if each safety measure came in progressively, the benefit of each would overlap and merge in a time sequence, and the progressive decrease in risk could be attributed to the progressive and overlapping introduction of all the various measures. But I think, for seat belts at least, there was a step change in 1983:

Would you not think that a step change from 30-40% to >90% should, if seat belts are nearly as effective as claimed, have produced a visible step in the risk trend?

I don't think the data (or at least the data presented by Adams) show any such step change.

My point is that those things have not been seamlessly absorbed but that they have all, along with what Stuart identifies, contributed to a general reduction in risk for drivers.

I think that what we are trying to explore is whether that is true or not, or perhaps more precisely, whether that is a conclusion that follows from the data or not..

 

[srw]

I think that what we are trying to explore is whether that is true or not, or perhaps more precisely, whether that is a conclusion that follows from the data or not..

The graph you opened the thread with demonstrates there has been a significant observed reduction in risk for drivers over the last 50 years. This has coincided with a lot of individual factors all of which have been intended to reduce risk for drivers. Applying Occam's Razor, the simplest explanation by far is that collectively they've worked.

Similarly (and I can't immediately find the data) I believe the statistics show a significant observed reduction in risk for cyclists over the same sort of time period. The Occam's Razor explanation is that what's made the roads safer for drivers has also worked for cylists.

 

[swansonj]

The graph you opened the thread with demonstrates there has been a significant observed reduction in risk for drivers over the last 50 years. This has coincided with a lot of individual factors all of which have been intended to reduce risk for drivers. Applying Occam's Razor, the simplest explanation by far is that collectively they've worked.

Similarly (and I can't immediately find the data) I believe the statistics show a significant observed reduction in risk for cyclists over the same sort of time period. The Occam's Razor explanation is that what's made the roads safer for drivers has also worked for cylists.

I'm not sure we're in the realm of Occam's razor, though. Occam's razor applies when you are comparing a simple explanation to a complicated explanation. Here, we have a set of data showing a reduction in risk for drivers. But we have two almost equally simple hypotheses to explain it. One is that it's a result of the safety measures that have been introduced over that period. The other is that it is a consequence of the reducing appetite for risk following the increasing prosperity over the same period. I don't know which (or what balance of combinations of them both) is the actual explanation, but I don't think it is at all obvious that it's the first not the second. Several factors seem to me to challenge the first hypothesis: why is there no discernible effect of an extra 50% of drivers starting to wear seatbelts almost overnight in 1983? Why was there (as I understand it) an increase in fatalities/injuries to cyclists, pedestrians, and back-seat passengers when the seatbelt law was introduced? Why have (according to Adams) fatality rates gone down no faster in countries which introduced seatbelt laws than those that didn't? Why (as I understand it) do developing countries, who nowadays drive cars with almost as many safety features built in as in developed countries, have massively higher accident rates? There may be explanations for all those issues, but I think they challenge a simple conclusion that the reduction in risk is because of safety measures.

(It's only seat belts, I think, that were introduced with enough of a step change to expect to see a step change in risk. As Stuart and you have pointed out, other safety measures, introduced gradually, would not produce any effect on the risk statistics that could be distinguished from a general trend. But if - if - we conclude that risk compensation has largely eaten up any benefit of seat belts, we might justifiably wonder if the same phenomenon applies to other safety measures.)

Incidentally, if you want an example of where Occam's razor - or simply assuming that the most obvious explanation is the most likely - is dangerous, think cycle helmets, and Thompson, Thompson and Riverra, the infamous paper that showed helmeted kids having dramatically less severe injuries. The simple explanation, which Occam's razor encourages us to adopt, is that helmets reduce injury severity. But of course we now understand that the real explanation is the rather more complex one, that there was a confounding factor about the type of kids wearing helmets, the type of cycling they do, and their propensity to attend hospital.

 

[srw]

View attachment 40836

I've gone right back to post one, because frankly your second explanation - that increasing prosperity has led to reduced risk appetite and directly to a 10-fold reduction in risk over 60 years - is so unlikely as to be preposterous. If risk-taking were connected with prosperity then you'd see something cyclical, going up in recessions - and that woud be obvious in the graph. In fact (it's a link I found a couple of days ago and didn't keep, but it's out there if you search), the number of accidents goes down in recessions in line with the reduction in traffic*. The rate stays roughly the same.

I'm frankly not that interested in the impact of single interventions. What's been important is a series of interventions (publicity, numerous safety features, motorways etc etc etc), each one of which has contributed.

Absolutely agreed about Occam's Razor and stats - but I don't think T, T, R is a good example.

 

[swansonj]

Absolutely agreed about Occam's Razor and stats - but I don't think T, T, R is a good example.

Of course TT&R is not the best example of how Occam’s razor can be misleading. But, of course, I chose that particular example to emphasise the parallel between cycle helmets and seat belts. I strongly suspect that most of us (it certainly applied to me) initially approached both issues with the same assumption – that the safety intervention is effective. But in the case of cycle helmets, most of us who look at the data with a reasonably open mind then conclude that our assumption was wrong; the advocacy of cycle helmets, and perhaps even the mere wearing of them, actually increases risk. And we now get irritated with people who, in respect of cycle helmets, insist on maintaining their initial assumption and won’t address the data. Whereas they, no doubt, can’t understand why we are pushing convoluted and counter-intuitive (they might even say “preposterous”) interpretations of some of the studies, and why we keep bleating on about risk compensation.

The parallel with seat belts is not perfect. For starters, there’s no dispute that if you have an accident, a seat belt will help, whereas there certainly is dispute as to whether, if you have an accident, a cycle helmet will help. But there are parallels. Knowing what we know about risk compensation as it applies in cycling – knowing what we know about how the biggest effect of helmet wearing is in knock-on effects on people’s behaviour rather than the immediate effect of the helmet – would we not be surprised if those effects didn’t operate in motoring as well? I don’t know to what extent these various things apply (Adams claims 100% risk compensation, and I doubt that is true, but I also doubt 0% is true either), but I sure as heck want to interrogate any relevant data to try to find out. And there's a parallel about our willingness to let our assumptions be challenged by the data.

I'm frankly not that interested in the impact of single interventions.

I agree that in general it’s not meaningful to try to separate the effect of individual overlapping and progressive safety interventions. The reason why seat belts are an exception is because of that step change in usage in 1983. To all of us interested in modelling and data, a step change of a factor of 2 in one of the explanatory variables is a godsend – we’d better see some impact on the output, or be prepared to ask ourselves some searching questions about our model.

Does making a driver feel safer cause them to drive less safely, and does a driver driving less safely lead to bigger risks for cyclists and pedestrians? I’d be absolutely amazed if this wasn’t true to some extent. Here is the results from a broadly pro-seat belt analysis:

They were arguing that the increase in cyclist and pedestrian (and back-seat passenger) injuries did not outweigh the reduction in front-seat injuries, so seat belts are a Good Thing. Maybe they’re right or maybe they’re not about whether the overall balance is positive (there’s debate as to how much of the reduction was attributable to changes in drink-driving law in the same year). But it looks to me like further reinforcement that risk compensation eats up at least some of the benefits of road safety measures .... and as often as not, it’s we cyclists who bear the consequences.