Statistically is the UK railway network now less safe?

[CapabilityB]

Or to simply answer the question... No.

 

[thenorthern]

One thing to remember with rail safety is we learn from our mistakes. I was born in the 1990s and growing up as a train fan I reckoned from 1997 until 2004 we would have 1 major train crash per year. There is no doube train crashes do happen but they are much rarer than they used to be. Some of the mistakes we have learned from:

• Ladbroke Gove 1999, in my opinion, led to by far the biggest changes in rail safety. it led to the creation of the Rail Accident Investigation branch. Also it led to the Train Protection and Warning System which has no doubt made our railways much safer. Also it meant that driver training and other safety standards were massively improved. Given driver Michael Hodder had only qualified 2 weeks before he should not have been driving out of Paddington so early on given Paddington is one of the most complex stations in the world and the Cullen Report lead to massive changes in railway operations.
• Southall 1997, after this if a train had a defective AWS it meant the train was taken out of service. While it causes inconvenience and delays, you can't compromise on safety.
• Hatfield 2000, lead to the end of Railtrack PLC, which let's face it was a badly run company from the start.
• Potters Bar 2002, I remember my local line being shut for a while after this as Railtrack PLC checked all the tracks in the country to ensure they met safety standards.

While accidents do unfortunately happen, as I say we very much learn from our mistakes with rail accidents in this country. The creation of the RAIB has in my mind made our railways much safer.

 

[Djgr]

Very interesting reading, thanks for posting

== Doublepost prevention - post automatically merged: 28 Jun 2026 ==


This is exactly what prompted my original question, I remember we did 'statistical significance' at school but that was a while ago and the methodology escapes me

A useful approach is to purchase a "for dummies" type book and you can certainly cover the basics in a couple of weekends.

 

[Swedenorer]

The figures here are too tiny to reach any conclusion. I think Mikeb42 above is right. I've never been convinced of any of the tests of statistical significance beyond looking at the data for trends since the quality of the data being worked on is seldom assessed - it's just assumed to be reliable. In the case of the OP chart the only thing the events had in common was fatality. You have to really look carefully at the screening of the data being looked at.

Mind you looking at the table in the OP the trends on the left they mirrored what I felt about things generally about the railway at the time but compared to the number of passenger journeys or trains run there is no statistical significance really. I certainly did not stop travelling since I accepted that once I left the house my risk index was higher anyway.

It's far more important to interrogate individual incidents as there is where the significance lies. Southall, Ladbrook Grove and Potters Bar revealed a railway straggling. Had they not happened the factors underlying them would still have been there. Which is a management issue not a statistical one

You can of course look at things the other way round. Take the thread on Barns Green. Considering the number of AHBs in the UK and the fact that in a normal days working on West Coastway between Chichester and Havant each one is going to operate at least 200 times - call it 64000 times a year allowing for closures etc - Barns Green on the less intensely used Arun Valley isn't even a statistical whisper overall and is insignificant for that location.

But the railway fitted them all with monitoring cameras. There was, I would argue, no statistical reason for doing so - or painting them yellow. But I would have done the same probably.

 

[Eyersey468]

Ah good spot

== Doublepost prevention - post automatically merged: 28 Jun 2026 ==


Very interesting thanks

To be fair to the railway the crashes at Great Heck and Ufton Nervert were caused by vehicles on the line,the railway wasn't directly at fault in either crash

== Doublepost prevention - post automatically merged: 29 Jun 2026 ==

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

 

[Harpo]

Less safe? No. The railway in the UK is undoubtedly a much safer operation than it was 50 years ago.

On a statistical slant, part of the safety improvement has come from a workforce that’s been highly trained to reduce individual variation, but there are times when I sense that it is almost causing individuals to work by rote and lack the ability to deal with varied circumstances.

 

[edwin_m]

There was clearly a big shift around 2002, around the end of Railtrack.

Also about the time TPWS came in, and other safety measures about the same time. Obviously difficult to disentangle the effects of different factors, but you'd expect introducing a new mitigating factor to show up in the statistics.

 

[anothertyke]

Given the small number of events, a longer time series is desirable. I'd suggest reading Prof Andrew Evans's various papers on this subject which examine trends in fatal accidents and fatalities over a long period eg 1967-2016. It would not be difficult to add the last decade on to that series.

 

[mrcheek]

Ultimately you can prove almost everything you want to with statistics, especially when the actual figures are very low (as they thankfully are in the case of UK railway accidents). Statistically, the railway was considerably less safe in 2026 five minutes after Bedford than it was 5 minutes before. In reality, there was no change (or if anything an infinitesimal improvement).

very well said.

Statistics can be used in many different ways.

But the fact is, that anyone who has any connection to rail, as an employee, traveller, or just a fan, knows that the railways are safer than they have ever been

 

[TUC]

Less safe? No. The railway in the UK is undoubtedly a much safer operation than it was 50 years ago.

On a statistical slant, part of the safety improvement has come from a workforce that’s been highly trained to reduce individual variation, but there are times when I sense that it is almost causing individuals to work by rote and lack the ability to deal with varied circumstances.

I agree, and the wider rules-based culture in the railway adds to that. To be clear, I'm not arguing against clear processes, but there is a big difference between rules which should be followed come what may vs processes which should almost always be followed but where good judgement is needed in individual exceptional circumstances as to what actions would best achieve the outcomes which the processes are aimed at achieving.

 

[Gerald Fiennes]

Former nuclear safety analyst here. Read up on Poisson distributions, which often used to be taught through case studies of the number of Prussian soldiers injured by sudden horse kicks and Bayesian statistics which really attempts to answer the question you are asking. Very rare events sometimes bunch together as one would expect completely random events to do from time to time but how bunched do they be to flag up a problem?

Statistics is the means of looking of data but won't provide definitive answers, one needs to pick a measure of statistical relevance that you think is appropriate and see where that takes you. 95% relevance is often used -- but this is a value judgement, not a physical law.

The further factor one must bear in mind when looking at trends over such timescales is the increasing collision resistance of the vehicles themselves. which certainly to me seems the main headline one should take from Bedford. Trains with better impact resistance mean (fortunately) fewer fatalities. The standards are much higher so one needs to look at deviations not from the mean but from a mean in which there is continuous improvement.

That said, I think one should indeed look at the most recent fatal collisions to see if there are common factors involved, which is really the thought process behind using this kind of stats. Actually, train adhesion seems to have been a factor and so one hopes that the industry is revisiting this issue, including eg checking that sand hoppers are in working order. In other words, it is an earlier set of incidents that could be 'outlier', not the current one.

 

[Basher]

A very difficult question to answer, Things that would have to be considered are:
miles of track
number ojournies made
total hours stock runs
was it a railway or roadway cause the accident
it is endless the list

 

[anothertyke]

The further factor one must bear in mind when looking at trends over such timescales is the increasing collision resistance of the vehicles themselves. which certainly to me seems the main headline one should take from Bedford. Trains with better impact resistance mean (fortunately) fewer fatalities. The standards are much higher so one needs to look at deviations not from the mean but from a mean in which there is continuous improvement.

As a very crude comparison look at the difference in outcome between Clapham Jct (speed at impact 40mph, 35 fatalities) and at Elstow (49mph, one fatality).

I agree also that if, for example, the incidence of heavy weather conditions changes, creating increased infrastructure related risks, that needs modelling. No amount of TPWS will prevent Talerrdig or Carmont.

 

[positron]

One other thing again depending on how you define safety, if you mean passenger safety but still purely look at deaths and not serious injury, then discounting driver deaths would be something to look at too. (Though they're obviously an awful thing to happen too).

Though I'd also say that a fatal crash with 1 death is very different from a fatal crash with 60 injured so again it's not super useful data.

One other thing is the cut off of 1990 is pretty arbitrary too. If you included the 1980s I think your chart would look quite different.

 

[King Lazy]

I know I’ll get shot down for this. But I think we have a hyper focus on scientific, statistical explanations for everything these days.

I posted a lengthy reply in another thread but removed it as it wasn’t strictly related to that thread but was answering a very similar question to this one.

All I’ll say is there is a lot that to me looks worse at the moment on the UK railway. I’m not talking major newsworthy incidents only but little things I see and hear. But I’m only going from what I’ve seen in my career. Some of the things have statistics to back it up. Some don’t.

But I know if I raised concerns (I have told managers what I’m seeing) the statistics could dismiss them. I also know I could be told I have confirmation bias and am seeing things because I now believe I’m aware of them and am looking for extra evidence. Maybe I am.

To me, as someone who has worked on the railway for over 20 years there is a lot that makes me think things are declining. But possibly not enough to flag up a problem. Or because different types of incidents are studied in isolation they aren’t being looked at as a whole.

And yes, I am aware many cleverer, more qualified people than me are looking at things as well. That is why I’m often reluctant to make posts like this. I have nothing to offer statistically nor do I have the training to voice my concerns eloquently against statistics. But I sense all is not great.

 

[Djgr]

very well said.

Statistics can be used in many different ways.

But the fact is, that anyone who has any connection to rail, as an employee, traveller, or just a fan, knows that the railways are safer than they have ever been

I still think that investing a few hours with a introductory statistical textbook is better than the "my mates down the pub" approach to this.

 

[gazthomas]

As above, this is not a statistical approach. There needs to be much greater granularity and proper measures over than "major accidents". Number of deaths per passenger mile as an example

 

[edwin_m]

Ladbroke Gove 1999, in my opinion, led to by far the biggest changes in rail safety. it led to the creation of the Rail Accident Investigation branch. Also it led to the Train Protection and Warning System which has no doubt made our railways much safer.

TPWS was devised before Ladbroke Grove, arising from a recognition in 1994 that ATP was not going to be widespread any time soon and other measures were needed to address SPADs and their consequences. By 1999 the system had been developed but was awaiting funding for network-wide fitment.

Potters Bar 2002, I remember my local line being shut for a while after this as Railtrack PLC checked all the tracks in the country to ensure they met safety standards.

This was Hatfield, not Potters Bar, with gauge corner cracking thought to be widespread due to Railtrack not having the engineering knowledge to understand the problem. Severe speed restrictions were imposed on all sections of track where they thought it might be present.

Potters Bar was probably the nail in the coffin of contracted-out maintenance, being due to a defect in a set of points that were maintained by a contractor and nobody really got to the bottom of who was responsible. There may have been some checking of similar sets of points but it would have been much less significant than Hatfield.

 

[bramling]

very well said.

Statistics can be used in many different ways.

But the fact is, that anyone who has any connection to rail, as an employee, traveller, or just a fan, knows that the railways are safer than they have ever been

The latter point is one for debate. Clearly the railway is safer than it was in the 80s or 90s. But is it safer than it was in the 2010s? One could argue that it isn’t simply because there’s been several accidents since 2020 that could reasonably be defined as serious, whereas there weren’t for a near 15-year period prior to that, although including Croydon alters that.

From a perception point of view, I suspect a lot will depend on what arises from the report into the Bedford collision. If it transpires it was a relatively inexplicable error from a driver with an unblemished record then it will likely just be a case of looking at whether the TPWS risk criteria could benefit from being altered. If however it turns up a pile of mess like we saw at Southall or Ladbroke Grove, especially in relation to driver training/supervision/assurance, it’s reasonable to suggest that things could have started to slip.

 

[mrcheek]

The point about statistics makes me think of the fascinating stat about Concorde.

For many years, it was statistically the safest plane in the world.

The it suddenly went to statistically being the least safe........

 

[edwin_m]

Being a UK monarch is also either the (joint) safest occupation or one of the most dangerous, depending how far back you go for the sample.

 

[Harpo]

In the past, SPADs were the railways biggest risk factor but TPWS mitigations pushed level crossings to the top of the list.

That the two most recent fatal accidents have both been a SPAD/movement authority exceedance and that one of then happened on a full L2 ETCS railway is very uncomfortable.

 

[The exile]

I know I’ll get shot down for this. But I think we have a hyper focus on scientific, statistical explanations for everything these days.

I posted a lengthy reply in another thread but removed it as it wasn’t strictly related to that thread but was answering a very similar question to this one.

All I’ll say is there is a lot that to me looks worse at the moment on the UK railway. I’m not talking major newsworthy incidents only but little things I see and hear. But I’m only going from what I’ve seen in my career. Some of the things have statistics to back it up. Some don’t.

But I know if I raised concerns (I have told managers what I’m seeing) the statistics could dismiss them. I also know I could be told I have confirmation bias and am seeing things because I now believe I’m aware of them and am looking for extra evidence. Maybe I am.

To me, as someone who has worked on the railway for over 20 years there is a lot that makes me think things are declining. But possibly not enough to flag up a problem. Or because different types of incidents are studied in isolation they aren’t being looked at as a whole.

And yes, I am aware many cleverer, more qualified people than me are looking at things as well. That is why I’m often reluctant to make posts like this. I have nothing to offer statistically nor do I have the training to voice my concerns eloquently against statistics. But I sense all is not great.

Indeed - you can measure the number of “near misses” as an indicator of “accidents waiting to happen” and extrapolate from them. But how do you measure “near misses waiting to happen”? (Ie the previous stage)

 

[Bald Rick]

I haven’t read all of the thread, apologies.

Railway safety is about far, far more than fatalities caused through train accidents.

There is a whole system safety model, regularly updated, which shows how safe the railway is. It has been on an improving trend for a long time.

Being a UK monarch is also either the (joint) safest occupation or one of the most dangerous, depending how far back you go for the sample.

*like button*

 

[positron]

In the past, SPADs were the railways biggest risk factor but TPWS mitigations pushed level crossings to the top of the list.

That the two most recent fatal accidents have both been a SPAD/movement authority exceedance and that one of then happened on a full L2 ETCS railway is very uncomfortable.

I think Talerddig is unnerving because it shows that just focusing on signalling and driver assistance is not enough to prevent fatalities. Hopefully one silver lining is that network rail might now specify the additional feature in future ETCS rollouts that is missing from Cambrian and may have helped the situation. We also need to be investing far more in vegetation management too as ultimately aside from level crossings I think weather events are going to be the biggest issue going forwards.

 

[The exile]

I haven’t read all of the thread, apologies.

Railway safety is about far, far more than fatalities caused through train accidents.

There is a whole system safety model, regularly updated, which shows how safe the railway is. It has been on an improving trend for a long time.




*like button*

Certainly compared to the risks of being a King of an independent Scotland and called James!

 

[bramling]

I think Talerddig is unnerving because it shows that just focusing on signalling and driver assistance is not enough to prevent fatalities. Hopefully one silver lining is that network rail might now specify the additional feature in future ETCS rollouts that is missing from Cambrian and may have helped the situation. We also need to be investing far more in vegetation management too as ultimately aside from level crossings I think weather events are going to be the biggest issue going forwards.

One could certainly argue that three of the last four incidents appear to have been loosely related to weather, albeit both Salisbury and Telerddig were simply typical British autumn weather rather than anything special. However the other side of this coin is that one can very reasonably argue that the weather simply exposed a flaw in the railways safety systems that would at some point have come to light.

 

[TheSel]

Another facet to bring into the equation when discussing fatalities is the response of the emergency services, and the NHS' (more widely) ability to treat people quickly and effectively. After any given event, the likelihood of a person's survival is not solely determined by the event itself, but also by the speed and effectiveness of the after care - particularly the immediate after care.

Thus the same event occurring in a remote location 50 years ago was far more likely to be fatal than that exact same event happening close to a major trauma centre today.

 

[WillC]

I suspect the long term trends will continue towards safer railways, following the usual path of upward jumps after incidents and the post-analysis improvements. But I wonder if we start to reach a point whether the law of diminishing returns kicks in, and the tough question of how much it costs and is it worth it kicks in? Weren't Boeing the first company to put a price on human life, based on pay outs to bereaved families? We face lots of risks in life and thankfully railways are nowhere near the top.

 

[FormerBritInMe]

The point about statistics makes me think of the fascinating stat about Concorde.

For many years, it was statistically the safest plane in the world.

The it suddenly went to statistically being the least safe........

Although there you have the problem of a fairly small population of people that ever flew Concorde, so that one crash with the loss of all passengers and crew really distorts the statistics. As opposed to say a Boeing 737 where millions have flown on it. Especially since the crash wasn't initially caused by a defect in the plane itself, but by foreign objects on the runway (although the plane's design may have contributed to the loss). So it may fall more into the category of "crashes due to FOD" rather than "Concorde crashes".