The concept of organizational culture is hard to define, hard to analyze and measure, and hard to manage'', Schein states in the preface of his 1992 book (p. xi). Yet culture has become a panacea for our “ills” in safety and it has spurned a prolific increase in safety culture models, surveys and “culture change experts.”

Here’s the most basic and important question (Guldenmund, 2000): Is what we define as culture something the organization is, or something it does? 

A popular notion is that culture is embedded in the organization (something it is) as a set of values, beliefs and attitudes and it directly impacts on behaviors (something it does) as “the way we do things around here.”

Impossible task

These simplistic statements are sometimes little more than hollow buzzwords and generalizations. Values, beliefs and attitudes cannot actually be measured in valid or meaningful ways and to really capture the “way we do things” is utterly complex. It will require extensive and impossible levels of behavioural observation and the “way we do things” will change from team to team, from circumstance to circumstance and be dependent on when, who and if someone is watching… or not watching. So the definition can be expanded to: “the way we do things around here when no one is watching.” But someone is almost always perceived to be watching!

Which brings us to perceptions. Some believe an organization’s culture ‘”lives” in the perceptions of the workforce, and is more correctly measured by perception surveys. Perceptions about what? It has to be perceptions about safety — this is what we are interested in.

Defining safety performance

The problem starts with the definition of “safety performance.” It is actually an extremely complex and contentious parameter to use because high performance is mostly defined as a condition of not having accidents — a low accident rate. And so here’s the problem: measuring safety through accident rates.

An organization without high rates of accidents may well be safe; it also may be simply fortunate — still having a lot of unreported near miss events. Safety performance may also be the result of suppressing the willingness of people to report accidents. In some classic cases, organizations were rated as superior in safety performance until they experienced dramatic disasters, such as the BP Texas City Refineries (Baker report). Another caveat: the extensive use of “modified work” strategies where injured employees are placed in “office jobs” or are assigned to administrative tasks so that they are not counted as off-work. They are simply fudged out of the injury statistics.

Dubious data

The reliability of these data sets is highly dubious. Their validity can be questioned, too. Consider everyday, in a typical organization, where millions of actions and decisions are taken that lead to safe outcomes, most of the time. Most or many of those actions could have been very risky in multiple ways, almost resulting in a near miss, an accident, a serious accident, a fatal accident, or even multiple fatalities, but only, fortuitously, did not happen. In other circumstances, most actions could have been highly safe, but one of the few risky actions resulted, unfortunately, in a serious accident. Both of these examples show how invalid incident indicators are of the level of overall safety in organizations, because we are fooled by randomness and tricked into this kind of thinking by the traditional triangles (ratios) of Heinrich and Bird.  

Counting meetings

Most safety culture models are “proven” by looking at the difference between high and poor performing companies, using these invalid and unreliable incident rates. It’s claimed that employees in high-performing companies give more positive ratings to a number of “structural” factors in a safety culture model.

For example, a certain popular safety culture model claims that frequency of safety meetings or belief that injuries can be prevented are related to a “good safety culture” as shown by companies with lower accident rates. Yet each of these (and other) factors is contestable. How can it be scientifically demonstrated that frequency and/or quality of safety meetings (a subjective assessment) indicate a good culture and relate to excellent safety performance?

Many readers have experienced a typical safety meeting where boring safety instructions are read out, rules and reminders are drummed into people or managers condescendingly point fingers to “workers not paying attention,” etc. 

Readers have also heard that a “belief in the preventability of injuries” in an organization is consistently and causally related to improved safety performance, or that employees with that belief are also more cautious, safer in their work,

Is the better performance of those organizations a result of less accidents and incidents occurring, or is it a result of less accidents being reported due to fear of retribution? The so-called Bradley curve may well be a deceptive curve, attained by falsification, fudging and fear to report.

What makes a strong culture?

Here’s an important second question: is the culture is good or bad, positive or negative, or strong or weak?

James Reason (1997) was the first to suggest which set of perceptual patterns, or cultural characteristics can readily point towards a safety culture. These can be described as an organization in which people are willing to report risks, wrong-doing, mistakes etc, from which they can continuously learn from and improve, and therefore maintain a high level of flexibility and transparency.

This can only happen if:

  • Employees feel free and unchallenged to enthusiastically make these reports and the changes that they deem necessary, and
  • Feel that they will be treated justly if mistakes occur (Dekker, 2012).
  • This can be sustained if there is a high level of trust between the various organizational layers;
  • If there is a reasonable balance between production objectives and safety requirements; and,
  • If the organization has a high level of readiness and capability to respond to risk.

It may therefore be more correct to talk about a risk culture, as against a safety culture. The direction of these ratings by employees points towards a positive or “good” safety culture. 

Widespread agreement among employees that the organization is indeed flexible, just, trusting, transparent, etc, is an indication of the strength of the culture.

Just like an onion

Culture is just like an onion, said Trompenaars (1993). There is no core “thing” in the middle of it. The layers of the onion are the onion, and if peeled away, nothing is left in the center. This is the same with safety culture. All the layers, or perceptual factors, point toward culture as the level of allegiance that exists among employees in the organization and culture “marshals” the practices that will give the best chance for excellence in results. Still, the term “excellence” does not mean that no, or few, or fewer accidents occur…

Safety is best defined as: “the organization is as poised as it could be to minimize the risk for adverse events to occur.” “Poised” refers to a multitude of competencies, organizational structures, capabilities, systems and perceptual patterns.

But ultimately, there is no such thing as a safety culture!


Baker, J.A. (2007). The Report of the BP US Refineries Independent Safety Review Panel. Washington, DC: Baker Panel.

Dekker, S., (2012). Just Culture. Ashgate, Aldershot, Hampshire, UK.

Guldenmund, F.W. (2000).  The Nature of Safety Culture: A Review of Theory and Research. Safety Science 34 215-257, NL-2628 EB Delft, The Netherlands.

Reason, J.T. (1997). Managing the Risks of Organizational Accidents. Ashgate, Aldershot. , Hampshire, UK.

Schein, E.H. (1992). Organizational Culture and Leadership, 2nd Edition. Jossey-Bass, San Francisco, USA.

Trompenaars, F. (1993). Riding The Waves of Culture. Understanding Cultural Diversity in Business. Nicholas Brealy Publishing. London.