Return to Alycidon Rail.

Return to Archive -by date - by topic - 2000 Archive.



Risk, perception and the cold numbers

There is much talk of restoring confidence in railway safety – but is confidence a rational emotion?


Like it or not, sooner or later, we live and die by the cold numbers. Even if you pass away peacefully in your sleep at a ripe old age, some actuary will note your demise and pension annuity rates will shift imperceptibly.

And, of course, we don't like it. If, that is, we think about it, because risk is one of those things that you don't really want to think about.

Instead we have perceptions of risk and in the modern world, perception is reality – perhaps it always was. Thus, we are told that the risk of being assaulted is far less than you might think from viewing Crimewatch, but late night trains and empty stations are scary even though the statistics may say otherwise.

All of which is why ‘Public Perception of Risk' is usually worth a paper at safety conferences. Indeed, by the time you read this, the LGRI will have held a seminar on just that subject as part of preparation for Part 2 of Lord Cullen's investigation.


Perception matters

Should we worry about public perception of risk when we, that is we in the railway, understand and appreciate the cold numbers? Yes, for two reasons.

Above all we have a Government which follows an agenda set by the media and focus groups. If the press or TV says that railways are dangerous, then they are dangerous and ‘something must be done'. Decisions may then be taken which could harm our industry.

And not just Government. The previous Chairman of the HSE took the view that the public will not tolerate accidents.

Second, while we can calculate risk precisely, safety cases are analogue not digital. There is no hard and fast line between ‘tolerable' and ‘intolerable' risk. If risk is tolerable, you can't sit back and relax. Under a safety case it is management's duty to make tolerable risk as low as reasonably practicable and what is practicable depends in part on perception.

Let me give you an example. After Mk 2 and Mk3 InterCity stock was introduced, the interior door handles had to be removed to stop people falling out with the train on the move.

With the handles on the outside, people still kept falling out. The railway view was that this was due to drunkenness or fooling around – call it ‘inappropriate behaviour' if you want to gain PC points, and not BR's responsibility.

So for several years, around 20 people a year were dying in falls from InterCity trains. And while Her Majesty's Railway Inspectorate muttered in its annual reports, nothing was done, partly because, like road accidents, it was a steady drip of personal tragedies.

Then, perceptions changed, one or two papers began to press the cause of people who were patently not drunk or confused and had still disappeared though an open door. The HMRI commissioned a study of door lock reliability. And under this pressure, driven by changing perceptions, central door locking came in and the death toll virtually ceased.

When you did a cost per fatality averted for central locking it was amazingly good value, or very reasonable in terms of practicability as we might say today.


Commercial impact

There is another, grimmer, aspect of public perception of risk. Put bluntly, what does it take to stop people travelling by train?

When British Airways cancelled its Concorde flights briefly after the Paris accident and before the aircraft had its Certificate of Airworthiness withdrawn, some people had cancelled their flights anyway. Others rolled up at Heathrow and complained bitterly when they could not fly supersonic. Clearly the advantage of getting to New York in three hours outweighed the perceived risk of another accident.

Many survivors of accidents still travel on the same service. Taking comfort in the cold statistics is as much a form of courage as discounting them when they are not in you favour.


Hard facts

So what are the cold numbers. Well in the 1990s the number of annual passenger accidental fatalities on the railways ranged from 9 in 1993/94 to 40 in 1999/00, the year of Ladbroke Grove. The second worst year was 1997/98 when the seven fatalities at Southall boosted the total to 29.

But the third highest number of fatalities was recorded in the year between the two disasters when no one was killed in an accident. Table 1 shows the statistics for this year when as far as perception was concerned the railways were ‘safe'.


Table 1

Passenger accidental fatalities 1998/99

Door related accident in running 3

Boarding alighting 2

Platform falls/too close to edge 9

Stair/escalator falls at stations 2

Surfing on trains 1

Passenger trespass: crossing/on track 9


Total 26


Fatalities per billion passenger -

km Journeys Hours
Air 0.05 Bus/Coach 4.3 Bus/Coach 11.1
Bus/Coach 0.4 Rail 20 Rail 30
Rail 0.7 Van 20 Air 30.8
Van 1.2 Car 40 Water 50
Water 2.6 Foot 40 Van 60
Car 3.1 Water 90 Car 130
Pedal cycle 44.6 Air 117 Foot 220
Foot 54.2 Pedal cycle 170 Pedal cycle 550
Motorcycle 108.9 Motorcycle 1,640 Motorcycle 4,840

Source DETR study


So already, the statistics are diverging from perception. From the media viewpoint 1998/99 was a non event as far as rail safety was concerned and yet 26 people died. If you look across the decade, in four successive years the death toll of passengers crossing the tracks equalled or exceeded that in the Southall accident. In three years (1997/98–1998/99 inclusive) the number of people being killed by passing trains as a result of being too close to the platform edge or falling onto the track averaged seven a year.

This raises the issue of railway culpability in such events. Some of the platform fatalities were clearly due to inappropriate behaviour and the question then is whether the railway should let young men (as is often the case) lark about on platforms? How many deaths would be avoided if stations were manned with staff able to intervene effectively?

Remedial measures are being taken and last year saw passenger deaths due to trespass cut from nine to three and platform deaths from nine to two compared to the previous year. Before Ladbroke Grove 1999/00 was going to be the second safest year of the decade.


Measuring risk

At which point, we need to digress into how we measure risk in transport. There are three possible bases: how far we travel, how long we spend travelling and how many journeys we make. The choice makes a considerable difference to the cold numbers, unless you ride a motor bike which is always the most dangerous mode.

If you calculate the risk on the basis of distance travelled, air travel wins by an order of magnitude. The reason is obvious: aircraft travel thousands of kilometres with everything working nicely and the only hairy moments are on take off and landing.

On the other hand, an EU study compared modes for journeys of less than 1,500km, that is continental rather than international. In this scenario there are a lot more take offs and landings per billion kilometres and rail turns out to be twice as safe as air.

If you use passenger hours as the base, you get another result. This favours slow forms of transport which cover long distances and haven't the capacity to kill a lot of people if catastrophic risk strikes.

Stand up bus and coach travel which is three times as safe as rail and air, which now have pretty well identical fatality rates.


Journey risk

Finally there is my preferred parameter, or rather the one I find most meaningful as I climb on a train. Fatalities per 100million or billion journeys, or its inverse, one fatality per so many journeys

This is, I think, (and readers may disagree) how risk is perceived at the moment of travel. Am I likely to survive this train journey, flight, drive and so on?

Railtrack certainly disagrees and argues that the passenger thinks in terms of making the same journey, that is travelling the same distance, in different ways. Hmm, but statistics show, for example, that a majority or car accidents happen within three miles of home, for example.

So, I still prefer journeys as the parameter. And on this basis, the DETR figures show buses and coaches at 4.3 per billion journeys, rail at 20, car at 40, air at 117 and motor bikes and mopeds at 1,640.

But I think (though readers may disagree), that the risk of death expressed one in so many journeys is more easily grasped.

Now, statistics such as those in the Table are riddled with inconsistencies. For example, as shown above, the rail figures cover falls at stations and passenger trespass. Coach and air don't include people standing too close to the runway, wing walking on 747s or falling down the escalator in the bus station.


Who counts?

You can argue, and I would, that fatality risks should include only occupants of the vehicle, so rail statistics should include only those who are killed in accidents, fall from moving trains or are killed getting on and off the train. If you make this adjustment, the Rail figure in the DETR statistics falls to 9.5 per billion passenger journeys.

In practice, Railtrack Safety & Standards Directorate uses 100 million journeys which is easier to relate to real life, as we shall see,

If we take 1999/2000 in isolation, the overall risk was 4.4 fatalities per 100 million journeys. Correcting for occupancy gives 3.52 fatalities per 100million journeys or 1 fatality in 28million journeys. This compares with the Railway Group Safety Plan target of 1 per 133 million journeys (also excluding trespass) by 2008/09.

What does one fatality in 28 million journeys mean? Well 28million is just above the total number of journeys on LTS Rail in 1999-00. What about the 2008/09 target of 1 in 133 million journeys? South West Trains last year had 132 million passenger journeys.

In fact, cruel though it may seem, we should not take one year, in isolation, good or bad. To prevent over reaction to tragedy or complacency in most years, the cold numbers have to be impersonal, which is why ten year rolling averages are preferred.

And as you can see from the table, during the 1990s, once the secondary door locking had come into effect (14 deaths in 1991/2 were falls from trains) the fatality rate per 100 million journeys has been broadly in the range 0.5 – 1.5. The risk had been steadily reducing over the decade with the cumulative average rising from 1 in 44 million journeys to 1 in 95 million by the end of 1998/99



What all this tells us is that the railway is very safe, but that this routinely high level of safety means that catastrophic risk now dominates the statistics and has a much greater impact on perception than in the past. My reference to the railway being very safe will, of course get short shrift from survivors of major accidents and the bereaved.

Which brings us back to perception. One passenger fatality a year on SWT would go unremarked, but, the trouble is that rail fatalities are not shared around pro-rata across the year like road accidents. Or rather, they are, until catastrophic risk strikes and the cold numbers which predict so many deaths per 100 million journeys on average are fulfilled in a few terrible seconds.

What's the answer? Only total commitment to running the railway by the book, all day, every day. It's called professionalism and it's very hard work


Table 1 Tables from DETR on fax





















per 100m journeys










Adjusted fatalities










per 100m journeys










Adjusted journeys per fatality(m)











Continues.........Return to Alycidon Rail.