Statistically is the UK railway network now less safe?

[Annetts key]

UK railways certainly have, and if you do have money to spend it could be spent far more effectively on safety in other areas of life. IDK about Boeing, but I worked in an industry with potentially serious danger to the public, and we had a price on life for the purpose of evaluating the cost/benefit of safety measures. I attended meetings with the Health and Safety Executive where these things were discussed. I am not going to state the value (which was not based on payouts) because any value will upset some people and it will have changed by now anyway.

But that does not mean we should not investigate accidents thoroughly and look for (and evaluate) safety improvements.

Two things, the value of a safety improvement often depends on the chances of an accident occurring if said safety improvement was not implemented compared to the chances if it was, combined with the cost of the work to implement the safety improvement. So these very much depends on the figures used, and how the statistics were derived. I say that just using data / numbers from fatalities, injuries or accidents that have resulted in a publication of an investigation is not sufficient to capture the real world risk.

We don't use the number of invasions by enemies in the past fifty years to decide what size our military is or how much we spend on it. The government uses other information (some of which is not disclosed) to make that decision.

With the railway, a big factor is where would any money come from to make changes, and often for anything expensive, that's a decision made by the government. But government dislikes spending money on the railway, especially for something that may not happen while they are in power.

QUOTE="Lucan, post: 7843944, member: 60310"]
But just to show the potential absurdity (a reductio ad absurdum argument), you could increase railway safety (even though very slightly) by having two signal posts where there are currently one, in case one fell over. Or why not three in case two fell over? Or a hundred in case 99 fell over? You have to draw the line at throwing money at some point.
[/QUOTE]
The safer option is not to have signal posts (instead having in cab signalling like ETCS/ERTMS).

---

As I tried to say earlier, do you use principles to make a decision or do you use statistics? IMHO statistics only tell you about the past and the past is not always reliable at predicting the future.

We have recently had a train crash whereby two passenger trains were involved resulting in one train driver being killed and a large number of injuries including some that were serious.

Did past statistics predict this? Would statistics based on the data that is not publicly available showed something different (that is including all the incidents that the various railway companies know about but which are not published in the public domain)? What about the mistakes and misjudgments that were not reported or picked up?

The point is, as things stand at the moment, on our "modern" railway , there still exists the very real possibility of a train crash that could be much worse than what happened at Bedford.

Technology has existed since the 1990s that not only would substantially reduce the risk of trains passing signals at red / danger, but would also substantially reduce the risk of trains speeding including over points and junctions and on curves and at other speed restrictions. The modern version of this technology is ETCS/ERTMS.

Let me put it another way, would our attitude still be the same if later this year or next year we had another train crash where another train went past a signal at red (at a signal that was not fitted with TPWS, ATP or ETCS) and slammed into a stationary train, with more people killed and injured? Would it still be the same if instead of just one person being killed, a larger number of people were killed?

 

[WillC]

Well the railway definitely has a number they use, called the Value of preventing a fatality (VPF) the number in 2025 was £2,856,000 (happy to be corrected).

I feel quite chuffed to be worth that much
But agree, there has to be a line, and we have such lines everywhere in life - who gets benefits and who doesn't, which treatment does the NHS fund and does not etc. There's always a - usually well-publicised case - whereby a war veteran with missing limbs fails to get a disability payment because they managed to crawl to the benefits office, or a dreadful case of a child with cancer which is deemed not worth treating (in pure national financial terms). It's a grim fact of life we just have to deal with.
I seem to remember a large debate after one of the 90s crashes (Southall?) about the value of installing improved train warning systems and if money might be better spent elsewhere.
Who'd be a politician.

 

[Surreytraveller]

Well the railway definitely has a number they use, called the Value of preventing a fatality (VPF) the number in 2025 was £2,856,000 (happy to be corrected).

I seem to recall after Clapham, when they were trying to put a cost on a life, one life was worth the same as ten serious injuries.
I'm not sure how a serious injury is defined, though

 

[anothertyke]

deleted-- failed to master multi-quote function!

 

[styles]

Well the railway definitely has a number they use, called the Value of preventing a fatality (VPF) the number in 2025 was £2,856,000 (happy to be corrected).

Transport Scotland has a very similar figure for road fatalities - £2,718,861 in 2023 prices (roughly £2.9m in May 2026 prices).

They even publish a breakdown of this figure: https://www.transport.gov.scot/publ...r-casualty-by-severity-for-gb-at-2022-prices/

Police and damage to property costs for GB:
Casualty related costs for GB:
Fatal	Serious	Slight	Damage only
	Collision Severity
(b) Costs per collision by element of cost and severity for Great Britain (£) at 2023 prices
(a) Cost per casualty by severity: average costs for Great Britain (£) at 2023 prices
	Killed	Seriously Injured	Slightly Injured	Average all casualties
Average cost per casualty for Great Britain	2,411,659	271,003	20,892	99,048
Lost output	897,022	35,672	4,186
Medical/ambulance	8,840	21,441	1,776
Pain, grief, suffering	1,770,228	243,408	19,946
Police/administration	26,253	3,051	789	52
Insurance	442	275	167	79
Damage to property	Total	16,076	7,250	4,267	2,749
	- Motorways	24,878	21,228	10,740	3,745
	- Non built-up roads	19,558	8,916	5,910	3,897
	- Built-up roads	11,532	6,180	3,646	2,607
Total costs per collision for GB	2,718,861	311,098	31,132	2,880

 

[WillC]

I wonder how costs are attributed to ‘grief and suffering’?

 

[styles]

I wonder how costs are attributed to ‘grief and suffering’?

No idea. I'm sure it's either in the public domain or at least FoIable though for anyone keenly interested.

Edit...

apparently these sources:

Hopkin, J.M. and Simpson, H. (1995) Valuation of Road Accidents, TRL Research Report 163, Crowthorne, Transport Research Laboratory

O’Reilly, D. M. (1993) Costing road traffic accidents: the value of lost output, TRL Working Paper WP/SRC/09/93, unpublished

Simpson, H.F. (1993) Damage costs in road traffic accidents: a survey of insurance claims, TRL Project Report PR/SRC/36/93, unpublished

Simpson, H.F. and O’Reilly, D.M. (1994) Revaluation of the accident related costs of road accidents, TRL Project Report 56, Crowthorne, Transport Research Laboratory

Presumably updated to today's prices!

 

[edwin_m]

Well the railway definitely has a number they use, called the Value of preventing a fatality (VPF) the number in 2025 was £2,856,000 (happy to be corrected).

The guidance including the number (which hasn't changed) is at https://www.rssb.co.uk/-/media/Proj...ty-related-cost-benefit-analysis-guidance.pdf.

This document gives guidance on the use of cost-benefit analysis (CBA) when determining whether a measure is necessary to ensure safety so far as is reasonably practicable (SFAIRP). Ultimately, the decision maker needs to apply their professional judgement to decide whether to implement the measure. CBA can inform this but should not be the sole driver of the decision. Safety-related CBA should always be undertaken as part of or alongside a suitable and sufficient risk assessment.

I remember reading a while ago (which does of course not mean that it's true even if I have remembered correctly) that the value was set higher for rail than road on the grounds that society views rail deaths as less acceptable than those on roads.

Of course on the roads there are a very large number of people making safety decisions every day and they are not generally doing so in a formal fashion.

As far as I recall it was set originally based on research into how much people would be prepared to spend to save the life of a random person, that is not themself or someone known to them. I believe it was originally more widely different between contexts than now suggested by #94 above.

Same at Talerddig.

According to the report, impact of the Talerddig victim's chest with the table was probably contributory to their death.

 

[WillC]

No idea. I'm sure it's either in the public domain or at least FoIable though for anyone keenly interested.

Edit...

apparently these sources:









Presumably updated to today's prices!

I get that some parameters might have specific costs that are measurable, like last work days etc. but grief feels very subjective, unless translated into depression, lost days at work, medication etc. I do remember getting £5000 from a claim against an errant car driver who failed to notice me (through not looking one presumes) and sent me flying leading to a fractured arm. They had a list for body parts I think.

 

[The exile]

I get that some parameters might have specific costs that are measurable, like last work days etc. but grief feels very subjective, unless translated into depression, lost days at work, medication etc. I do remember getting £5000 from a claim against an errant car driver who failed to notice me (through not looking one presumes) and sent me flying leading to a fractured arm. They had a list for body parts I think.

The irony is that, unless you have dependents, the actual “cost” of being killed is a lot less than that of being seriously injured. Logically, it makes sense but flies in the face of our view of “gravity”.

 

[AdamWW]

As I tried to say earlier, do you use principles to make a decision or do you use statistics? IMHO statistics only tell you about the past and the past is not always reliable at predicting the future.

Statistics on past behaviour can be a valid way of making very good predictions of the likelihood of events in the future.
Otherwise annuities and insurance policies would look very different, if they even existed.

But you have to have enough data to be statistically significant, and of course you have to have grounds to make the assumption that there won't be some change in circumstances that invalidate the conclusions.
Current life expectancy tables will turn out to be horribly wrong if a large asteroid hits the Earth next year but. (That probably wouldn't be a problem for companies offering life insurance though).

Let me put it another way, would our attitude still be the same if later this year or next year we had another train crash where another train went past a signal at red (at a signal that was not fitted with TPWS, ATP or ETCS) and slammed into a stationary train, with more people killed and injured? Would it still be the same if instead of just one person being killed, a larger number of people were killed?

I suspect that it would change things, but from a purely statistical point of view it probably shouldn't.

 

[Horizon22]

There's also the crash on the level crossing near Preston this year missing as well where the car driver was killed if we are including incidents of vehicles on the line

There have been a few of those - there was an Amazon driver at a level crossing in Kent too who also died IIRC.

 

[Bald Rick]

There have been a few of those - there was an Amazon driver at a level crossing in Kent too who also died IIRC.

Indeed. There‘s several such incidents each year, and many more with pedestrians. These are all still fatalities on the railway system.

 

[Annetts key]

Statistics on past behaviour can be a valid way of making very good predictions of the likelihood of events in the future.
Otherwise annuities and insurance policies would look very different, if they even existed.

But you have to have enough data to be statistically significant, and of course you have to have grounds to make the assumption that there won't be some change in circumstances that invalidate the conclusions.

I think you misunderstand what I was saying. There are good reasons to use statistics. As you say, you do however need to have plenty of data and the data must be relevant without missing out any significant information.

But even with plenty of relevant data and without missing out any significant information, they still only tell you about the past. Yes, in many cases, routine events in the future will follow the pattern of the past. But that does not apply to everything.

Does anyone win often by using statistics on the national lottery? What about horse racing? What about other sports?

The risk of a train collision, or death or serious injury due to a train collision are, IMHO, one of the things where the current method of using statistics are not suitable due to the reasons I pointed out previously.

On insurance, if an individual takes out insurance but fails to declare some significant and relevant information, the insurance will declare the policy invalid and will refuse cover and any payouts.

 

[AdamWW]

I think you misunderstand what I was saying. There are good reasons to use statistics. As you say, you do however need to have plenty of data and the data must be relevant without missing out any significant information.

But even with plenty of relevant data and without missing out any significant information, they still only tell you about the past. Yes, in many cases, routine events in the future will follow the pattern of the past. But that does not apply to everything.

Does anyone win often by using statistics on the national lottery? What about horse racing? What about other sports?

The risk of a train collision, or death or serious injury due to a train collision are, IMHO, one of the things where the current method of using statistics are not suitable due to the reasons I pointed out previously.

On insurance, if an individual takes out insurance but fails to declare some significant and relevant information, the insurance will declare the policy invalid and will refuse cover and any payouts.

I generally agree and going back to your previous comment: "the past is not always reliable at predicting the future." perhaps I did indeed misunderstand.

I thought you were making the point that circumstances change making conclusions from past behaviour invalid but perhaps you were just saying that - as with a fair lottery - understanding past behaviour doesn't always provide any useful guide to the future even if everything does still act in the same way.

I'd say that in a fair lottery events in the future exactly follow the pattern of the past so probably we're defining "pattern" differently here.

I think rail accidents are rare enough to make it hard to use them to predict the likelihood of future accidents, which is now making me ponder on how an insurance company would set a premium for such things.

 

[paninaro]

Does anyone win often by using statistics on the national lottery? What about horse racing? What about other sports?

I would say that's exactly what bookmakers do to ensure they consistently "win"; using statistics from past events to ensure they are offering odds that give punters less than they stake.

 

[Andrew1395]

As an ex rail employee the railway I would say is indeed safer now. If you want a better indicator than crash fatalities to measure “safeness”, I would look at the trends on staff injuries and deaths. You could compare those events in the UK to say comparable sorts of railways in Europe, both individual systems and the European average.

I am sure that data is studied. I have no idea if say Japanese or South Korean railway systems are safer, but I suspect the GB railway (including LUL and other regional metro rail systems) are safer than the European average.

 

[Annetts key]

I would say that's exactly what bookmakers do to ensure they consistently "win"; using statistics from past events to ensure they are offering odds that give punters less than they stake.

This is going off topic, but bookmakers rig the system in favour of themselves by the odds that they offer. Just like any gambling system is rigged. Yes, they do use statistics to do this.

My point was that the punter if they use statistics cannot reliably pick a winner. Even if they had access to the exact same data that the bookmakers have.

== Doublepost prevention - post automatically merged: 1 Jul 2026 ==

As an ex rail employee the railway I would say is indeed safer now. If you want a better indicator than crash fatalities to measure “safeness”, I would look at the trends on staff injuries and deaths. You could compare those events in the UK to say comparable sorts of railways in Europe, both individual systems and the European average.

I am sure that data is studied. I have no idea if say Japanese or South Korean railway systems are safer, but I suspect the GB railway (including LUL and other regional metro rail systems) are safer than the European average.

Regarding injuries and deaths on the railways, I've not looked for details of other countries. The details on U.K. railways are available from the ORR (link), the Key results for April 2024 to March 2025 are:

Key results​

• There were 14 non-workforce fatalities (passenger or public) in the year April 2024 to March 2025: 11 fatalities (eight RIDDOR and three non-RIDDOR reportable) occurred on the mainline network and three passenger fatalities (two RIDDOR and one non-RIDDOR reportable) on the London Underground.
• There were two workforce fatalities across all the networks in the year April 2024 to March 2025, one on the mainline and the other on the non-mainline network.
• There were 22 people who died in accidents while trespassing in the year April 2024 to March 2025. They involved 17 trespass fatalities on the mainline, four on the London Underground and one on the non-mainline network. There were five fatalities at level crossings which involved pedestrians on the mainline.
• There were 368 suicide or suspected suicide attempts on the mainline, of which 293 were fatalities, the highest number recorded since that start of the time series in April 2002.

This year (2026), two railway employees have died that I am aware of: a train driver died in the collision of two passenger trains near Bedford and a Network Rail track worker was killed near Radlett (Rail magazine reported on this).

 

[positron]

Same at Talerddig.

I know that it was reported by some media as a heart attack but I remember there being some controversy around that at the time and I don't know if the cause of death was ever published. The RAIB certainly never mentioned it.

 

[mikeb42]

My point was that the punter if they use statistics cannot reliably pick a winner. Even if they had access to the exact same data that the bookmakers have.

They can't, for the very good reason that it's impossible to reliably pick a winner. As the thing they're betting on isn't really random since one horse or football team or whatever really is better than another, if they're competent enough they will likely be able to predict winners on average though. However, the bookmakers are at least as competent at doing this so on average punters always lose. This is no argument against statistics - it proves they work at predicting the future in ensemble rather than for individual events.

Likewise, the question "but how can an insurance company predict the future likelihood of a car/train/whatever crash?" is answered by the observation that their actuaries clearly can on the whole. Otherwise they'd all go bust, or have a huge gulf between premia and payouts, neither of which is generally true. The very best guide to the future is the past (maybe plus some intuitive twiddling).

Thus maybe the best way to answer the founding question of the thread is - ask an insurance company. Though doesn't the railway self-insure to a large extent? Is this also true of the ROSCOs?

 

[The exile]

They can't, for the very good reason that it's impossible to reliably pick a winner. As the thing they're betting on isn't really random since one horse or football team or whatever really is better than another, if they're competent enough they will likely be able to predict winners on average though. However, the bookmakers are at least as competent at doing this so on average punters always lose. This is no argument against statistics - it proves they work at predicting the future in ensemble rather than for individual events.

Likewise, the question "but how can an insurance company predict the future likelihood of a car/train/whatever crash?" is answered by the observation that their actuaries clearly can on the whole. Otherwise they'd all go bust, or have a huge gulf between premia and payouts, neither of which is generally true. The very best guide to the future is the past (maybe plus some intuitive twiddling).

Thus maybe the best way to answer the founding question of the thread is - ask an insurance company. Though doesn't the railway self-insure to a large extent? Is this also true of the ROSCOs?

Although once the company has built up sizeable reserves they can afford to risk getting it a bit wrong - especially as in most cases cover is limited. The experience of Lloyds (of London) “names” (whose liability was then unlimited) at the end of the 1980s shows that “they” didn’t always get it right and that very rare disasters are very hard to predict precisely because they are so rare.

 

[Bald Rick]

Though doesn't the railway self-insure to a large extent?

For some things, but not others. The Network Rail ‘book’ is very keenly competed for in the London Insurance market.

 

[AdamWW]

For some things, but not others. The Network Rail ‘book’ is very keenly competed for in the London Insurance market.

Fascinating!

 

[Bald Rick]

Fascinating!

To be fair everything is keenly competed for on the London Insurance market at present. After a decent run over a good few years (for the insurers / reinsurers) there is more supply than demand and prices have taken quite a hit. Whether this feeds through into the domestic insurance market is hard to say. Nevertheless, the NR book is more keenly competed for than most.

 

[APT618S]

Well the railway definitely has a number they use, called the Value of preventing a fatality (VPF) the number in 2025 was £2,856,000 (happy to be corrected).

Interestingly that figure is significantly higher than what NICE (The National Institute for Health and Care Excellence) use to approve NHS treatment which is around the £25k-£30k per year or QALY (Quality Adjusted Life Year):

Changes to NICE’s cost-effectiveness thresholds confirmed

Changes to NICE’s cost-effectiveness thresholds confirmed

www.nice.org.uk

Given the average rail traveller probably has around 40 years ahead of them that would equate to just £1m - £1.2m.

 

[dastocks]

I'm not sure how a serious injury is defined, though

When I processed road accident data for a police authority it was any casualty that required an overnight stay in hospital.

 

[blueberry11]

I could say Iceland has the safest railways in the world always having zero deaths but there isn't a single railway track in the country. Even if it did, its low population (400k) would mean there would be zero railway deaths for many years, possibly decades and even a single death would skew the data.

 

[zwk500]

I could say Iceland has the safest railways in the world always having zero deaths but there isn't a single railway track in the country.

Conceptually, is it possible for something that does not exist to be safe? You could certainly make the argument that Iceland is in joint 1st ranking for the place least likely to experience a rail accident, or perhaps semantically the joint-safest place for an individual to avoid rail accidents.

 

[edwin_m]

Iceland had a couple of industrial railways in the past, but I don't know if they had any accidents. I think one or two of the locomotives are still on display - if someone stubbed their toe on one then would that count?

 

[anothertyke]

I could say Iceland has the safest railways in the world always having zero deaths but there isn't a single railway track in the country. Even if it did, its low population (400k) would mean there would be zero railway deaths for many years, possibly decades and even a single death would skew the data.

That's why Ireland was top of the European league table for years. Passenger fatalities zero, passenger kms xxx. I don't know whether that is still true.