Ups and Downs: What Accident Statistics can tell us

This piece is loosely based on an article I wrote in preparation of the Congres Veilig Werken on 11 April 2018 in Ahoy, Rotterdam. You can find the original text (in Dutch) here.

I do not read all the news about safety, but internet and social media often alert me to the main headlines. Like press releases that the Dutch health and safety authority (Inspectie SZW) is concerned about the number of accidents that apparently has been on the rise for the past few years - while they at the same time tell us that the likelihood of having an occupational accident in the Netherlands is very low. Other organisations come with similar messages. Maybe it is a good moment to reflect on what these metrics and news items can tell us.

Do mind, I do not have any concrete answers. I do have many questions, however, and thinking about those may help us to deal with statistics in a healthy way.

Primal Reaction

Let’s start by wondering how this affects us. Accidents are horrible, in particular when people lose their lives or are permanently injured, physically or mentally. Events like that demand our attention and appeal to our emotions. Those primal reactions quickly lead to social, political and moral pressure to do “something”. Suddenly there is this sense of urgency: this horrible thing shall never happen again!

But... is that primary reaction always the best? Often the reaction on accidents and statistics is little but a knee-jerk. Unthinkingly one jumps to the first ‘cause’ at hand (“human error”, “bad safety culture”) and comes up with a quick fix (“more regulatory oversight”, “stricter rules”, “a campaign”). Often, these are ineffective and draw resources away from other, structural safety initiatives. The smart thing to do would therefore be to think things over, before acting.

A Measure for Safety?

Accidents are horrible, but what do they actually tell us about safety? Instinctively we think that “many accidents = unsafe”. The number of accidents surely gives some kind of indication about (un)safety, but this is rather limited. And this works both ways.

We perform countless “unsafe acts” without this ever resulting in an accident. Say, someone gets into his car after an evening of heavy boozing, and arrives home safely without so much as a scratch. No one will think of calling that scenario particularly safe, the term to describe it would rather be ‘lucky’. No accident is not necessarily safe.

What about the opposite? Does an accident mean that the situation or act was unsafe? Intuitively we may be inclined to answer, “Of course, else there had not been an accident”. But that is speakig with hindsight. Before the accident happened, one probably assumed that things were safe. That assumption may have been correct, and maybe false.

The accident may be the result of an unforeseen combination of circumstances, but also of an accepted risk. After all, we take many risks willingly. When you build your house in a flood plain, you should not be surprised if you get wet feet every once in a while. We often regard a one-in-a-million chance as safe, but that does not mean that no accident can happen. Then there are the risks we cannot control fully, despite our best efforts. The prime reason for having airbags, seat belts and energy absorption zones.

Accidents are an even worse measure for safety when it comes to injuries. Randomness plays an even larger role. One and the same accident, for example a collision of a car on a railway level crossing, can have limited consequences (the driver is physically unharmed), serious consequences (the only person in the car, the driver, is killed) or extremely serious (a family of five dies in the accident). The accident was the same; the consequences are very diverse.

A Sign for Urgent Action?

What does a peak or an increasing or declining line actually tell us? Our brains are wired to recognize patterns. This ability helps us to function effectively and efficiently. Sometimes we perceive patterns that are not there, however, and we attribute these perceived patterns some significance that they do not have. Three accidents on three successive days does not necessarily mean that suddenly a wave of unsafety has struck us.

If you see a peak, or some decrease of increase, put it into context. Check whether where this happens (e.g. circumstances, sector, or location) really is the ‘main’ problem, or if there are maybe matters that deserve more attention. For example, more people die as the consequence of exposure to various chemicals than from occupational accidents.

Numbers can vary heavily over the years. Do look into the reasons for these fluctuations. Sometimes this can be randomness. Sometimes you can find a factor that correlates. If activity in the construction business doubles, it should be not surprising that the number of accidents goes up.

Interests?

Look always how the numbers are presented. Lying, or telling half-truths, with help of statistics is rather easy. Statistics are easy to manipulate without even faking the numbers. By adjusting scales, cutting them, by limiting to a certain period or by extending it, by presenting things in percentages, ratios or in absolute numbers you can amplify your message.

Three fatalities per year on level crossings sounds serious; 0,006 fatalities per level crossing with barriers is considerably less serious and 0,0009 fatalities per level crossing won’t worry many people. Or you can make an increase in the number of fatalities look dramatic by only looking at the previous year, and ignoring the period that came before (Cobouw did this in the Netherlands, declaring 2016 a “disaster year”, while looking at a five year period did not show any pattern).

Ask yourself what message the way of presenting is meant to support and how the numbers would look (and what they would say) if they were presented in another way. It is always wise to stop up for a moment when dramatic numbers are presented and wonder what interests this message serves. This is not always to the benefit of safety!

A Better Approach

Do not regard a number, a trend or a peak as some kind of an answer; something that leads to apparently “decisive” panic actions. Rather regard them as something that tickles your curiosity. Something that sparks the question “Hey, I wonder why this is how it is”. Something to be regarded in an historical perspective and its context.

Accident metrics? My advice: do not give them more attention than they deserve from a professional safety perspective. Accidents send us a signal that we most likely have one or more opportunities for improvement. They give information about unknown or not recognized risks. About unexpected side effects and competing objectives. Approach that with sober and critical thinking, and not with a knee-jerk.

---

Carsten Busch is a self-declared Safety Mythologist and author of the well-received book Safety Myth 101. This book collects 123 (and then some) of Safety Myths. Crisp and compact discussions address weaknesses of conventional safety ‘wisdom’ and give suggestions for alternative approaches and improvement. You can find many of the subjects discussed above in the book, including measuring safety, application of numbers, and the usefulness of outcomes.

https://www.mindtherisk.com/the-book