Why people take risks at work

An evidence review of organisational risk appetite, individual risk propensity, and the behavioural science of workplace risk-taking, with implications for work health and safety practitioners

Abstract

Almost every organisation promises to keep its people safe at work. The same organisations run on schedules, budgets and output targets that reward getting the job done, and much of what safety practitioners label risk-taking arises from the tension between the two. The evidence indicates that this risk-taking is, for the most part, normal behaviour by ordinary people rather than a defect of character. This review draws that evidence together to explain why people take risks at work, and what the explanation means for those whose duty is to keep them safe.

The review keeps two ideas distinct: how much risk an organisation is willing to accept, and how much risk an individual tends to take, including how that individual perceives and judges risk in the first place. It weighs the evidence at three levels: the principal theories, the strongest large-scale studies that test them, and current Australian safety law and injury data.

A central observation follows in that organisational leaders state how much risk they will accept, but the risk an organisation actually rewards and tolerates is often different, and workers respond to the second. Work therefore tends to drift toward the edge of what is safe, and what the organisation rewards sets the strength of that drift. An individual’s judgment and personality then shape the path taken within those limits. The situation generally matters more than the person, though both contribute, and a place remains for the genuinely reckless individual choice.

Australian safety law already favours controls that do not depend on how people behave. For practitioners, the implication is to direct safety effort at the organisational conditions that create risks in the business: choosing controls that remove the hazard matters far more than relying on people to follow instructions consistently.

1. What this review covers

1.1 What the organisation tolerates and what the worker brings

Almost every organisation says it wants no one harmed at work, yet the same organisations run on schedules, budgets and output targets that quietly reward getting the job done. Workers, supervisors and managers live inside that tension on every shift, and much of what safety practitioners call risk-taking is the visible residue of decisions made within it. The historical reflex, including among some safety practitioners, is to instantly read an unsafe act as a defect in the person who performed it: a lapse of attention, a poor attitude, a failure to follow the rule. The research reviewed here points elsewhere. Risk-taking at work is, for the most part, normal behaviour, produced by ordinary people responding sensibly, often on their own terms, to the situation they are placed in. To understand why people take risks, and what that means for those whose duty is to keep them safe, two ideas must be held in view at once and kept distinct.

The first idea is organisational risk appetite which comes from enterprise risk management and corporate governance, and it describes how much risk an organisation is willing to accept, in pursuit of its objectives, before it acts to treat that risk. The second is individual risk propensity, together with how individuals perceive risk and decide under uncertainty. That concept draws on personality psychology, behavioural economics, cognitive and affective neuroscience, and evolutionary biology. Everyday safety conversation tends to run the two together, as though an organisation that tolerates a hazardous shortcut and a worker who takes that shortcut were expressing the same thing. They are not, and the distinction matters for both diagnosis and control. An organisation expresses appetite through its rules, its resourcing, its incentives and, above all, through what it actually tolerates day to day. An individual expresses propensity through dispositions and states that are partly stable and partly situational. The two meet at the point where work is done, and that meeting point is where this review argues practitioners should look.

The argument advanced here is that organisations communicate a revealed appetite for risk through production pressure and through the shortcuts they reward or ignore, and that this revealed appetite shapes, and sometimes compels, the risk-taking of individuals. The appetite stated in policy and the appetite revealed in the workflow are often different things, and workers respond to the second.

1.2 How the evidence was chosen and weighed

This is a narrative review rather than a systematic one, written primarily for work health and safety practitioners and others concerned with building safer workplaces. The regulatory framing and the Australian injury data serve as the anchor to which the behavioural science is referred. Because a narrative review selects and arranges its sources, the basis of selection is stated by the author rather than left implicit.

The review makes no claim to the exhaustive, protocol-driven coverage of a systematic review. The literature was identified by the author’s judgment rather than by a single reproducible database search: foundational works were taken from the standard handbooks and authoritative reviews in each field, and the empirical sources were the most highly cited and methodologically strongest meta-analyses and large primary studies known in the safety, decision-science and organisational literatures, supplemented by targeted searching for more recent work and for the contested-construct debates. The Australian legal and statistical sources were drawn from the current primary materials. Coverage therefore spans foundational theory from the 1970s onward, the integrative meta-analyses of roughly 2005 to 2015, and current Australian law and data to 2025. The selection is consequently susceptible to the biases of expert selection, and the review is best read against a recognised standard for narrative-review quality such as the SANRA criteria, which call for explicit statements of aim, search, referencing and level of evidence (Baethge, Goldbeck-Wood and Mertens 2019); the limitations this method imposes are stated in section 9.

The literature was assembled in three layers. The first is the set of foundational theoretical traditions that any serious examination of risk-taking must address; these are drawn from the standard reviews and handbooks in each field and represented by the defining statement of each tradition. The second is the strongest available quantitative evidence on how those traditions actually perform: the meta-analyses and large primary studies that test the theories against work safety outcomes. The third is the current Australian regulatory text, verified against the legislative source, together with the most recent national injury statistics. Where the layers conflict, quantitative and convergent evidence is given more weight than single studies, and contested constructs are flagged as contested rather than presented as settled.

Three cautions follow from this method and apply throughout. First, much of the safety-climate, leadership and motivation evidence is correlational, cross-sectional and based on self-report, so it is open to reverse causation and common-method variance, and the causal language the field often uses is not always earned. Second, several of the foundational constructs have been seriously challenged in the decade since the integrative meta-analyses were published, and a review written in 2026 should engage those challenges rather than rest on the original statements. Third, the empirical backbone of the safety literature is now ten to sixteen years old, so its conclusions should be read as well established in their time and checked against newer work where currency matters. These cautions do not weaken the overall thrust of the evidence; they bear on how confidently any single claim can be made, and the review tries to mark that confidence honestly.

This review moves from the organisational construct of appetite in section 2, through the individual in section 3, to the behavioural and organisational mechanisms that turn inclination into action in sections 4 and 5, and then to the Australian regulatory and statistical framework in section 6. Section 7 is an integrative synthesis built on a single systems model, which seeks to adjudicate between the traditions rather than list them. Section 8 is the purpose of the exercise: it sets out in one place what the evidence means for practice. Section 9 states the review’s limitations, and section 10 concludes.

2. What an organisation will tolerate and what the law requires

2.1 What ‘appetite’, ‘tolerance’ and ‘capacity’ each mean

The vocabulary of risk appetite originates in the enterprise risk management and governance literature. The international risk management standard, ISO 31000:2018, frames risk as the effect of uncertainty on objectives and directs organisations to set risk criteria against which the significance of risk is evaluated (International Organization for Standardization 2018). The companion vocabulary standard, ISO 31073:2022, supplies the agreed definitions and replaced the earlier ISO Guide 73:2009 (International Organization for Standardization 2022). The widely used enterprise risk management framework of the Committee of Sponsoring Organizations of the Treadway Commission treats risk appetite as the types and amount of risk an organisation is willing to accept in pursuit of value (Committee of Sponsoring Organizations of the Treadway Commission 2017).

These terms are worth separating. Risk appetite is the amount and type of risk an organisation is willing to seek or accept. Risk tolerance is the acceptable variation around that level for a particular objective or activity, the boundary of what will be endured before action is required. Risk capacity is the maximum risk the organisation could bear before its viability is threatened, whether or not it would choose to bear it. Hillson and Murray-Webster (2007) add a useful distinction between an organisation’s risk attitude, its chosen response to a given situation of uncertainty, and the deeper traits and culture that produce that attitude. A stated position and an enacted one are different things.

2.2 When risk management becomes box-ticking

Not all of the literature treats enterprise risk management as an unalloyed good, and the critique bears directly on the gap between stated and revealed appetite. Power (2004, 2007) argues that risk management has grown into a generalised apparatus of organised uncertainty, in which producing auditable records, registers and assurances can become an end in itself and displace attention from managing the hazard. On this account, risk management can manufacture comfort. Renn (2008), writing from a risk governance perspective, makes a related point: the handling of risk in real institutions is shaped by perception, framing and social process at least as much as by technical assessment, and good governance must integrate the analytic and the social rather than privilege one. These are not arguments against managing risk. They are warnings that a documented appetite and a controlled hazard are different achievements, and that the first is easily mistaken for the second.

2.3 How WHS law frames the same problem

Australian work health and safety law does not use the language of risk appetite. The primary duty of care requires a person conducting a business or undertaking to ensure, so far as is reasonably practicable, the health and safety of workers and of others affected by the conduct of the business (Work Health and Safety Act 2011 (Qld) s 19). The standard is not a chosen appetite but an externally fixed obligation, qualified only by what is reasonably practicable. The Act defines that qualifier, and the definition is essentially a structured weighing exercise. What is reasonably practicable is that which is reasonably able to be done to ensure health and safety, weighing the likelihood of the hazard or risk occurring, the degree of harm that might result, what the person knows or ought reasonably to know about the hazard and the ways of eliminating or minimising it, the availability and suitability of those ways and, only after the extent of the risk and the available controls have been assessed, the cost, including whether that cost is grossly disproportionate to the risk (Work Health and Safety Act 2011 (Qld) s 18). Cost enters the frame last and in a constrained sense: it cannot be weighed until the risk has been assessed, and it may defeat a control measure only where it is grossly disproportionate. The regulator’s interpretive guidance confirms this ordered, objective construction of the test, and stresses that the standard is fixed objectively rather than by the duty holder’s capacity to pay (Safe Work Australia 2013).

The regulatory means of giving effect to this duty is the hierarchy of control. Where eliminating a risk is not reasonably practicable, the duty holder must minimise it so far as is reasonably practicable: first by substitution, isolation or engineering controls, then, for any remaining risk, by administrative controls, and then by personal protective equipment, with combinations permitted where a single control is insufficient (Work Health and Safety Regulation 2011 (Qld) r 36). The architecture embeds a strong preference for controls that do not depend on individual behaviour, and it places the obligation on the duty holder, not the worker. This is the single most important point of contact between the law and the behavioural science that follows. The science explains why controls that rely on people behaving as instructed are fragile; the law, read carefully, already prefers the controls that do not.

A note on jurisdiction belongs here, because this review speaks to practice across Australia while citing a single statute. The provisions relied on throughout are those of the model Work Health and Safety Act and its Regulation, as enacted in Queensland. The same model law, with minor local variation, is in force in the Commonwealth, New South Wales, the Australian Capital Territory, the Northern Territory, South Australia, Tasmania and, since 2022, Western Australia, so the Queensland sections cited here stand in for those jurisdictions. Victoria is the exception: it retains the Occupational Health and Safety Act 2004 (Vic), which places a comparable duty on the employer, qualified by what is reasonably practicable (s 21), but uses different drafting and distributes the control hierarchy across hazard-specific control-of-risk regulations in the Occupational Health and Safety Regulations 2017 (Vic), for example r 325 for construction work, rather than stating it in one general provision.

The behavioural argument that follows does not turn on the specific jurisdiction, and the legal point, that the framework already prefers controls which do not depend on individual behaviour, holds in every Australian jurisdiction, including Victoria.

2.4 What an organisation says versus what it actually rewards

The organising idea of this review is that an organisation has both a stated risk appetite and a revealed one, and that the second is what actually reaches the worker. Because the idea does real work in the argument, it should be defined precisely and defended rather than asserted, and its relationship to existing theory made explicit.

The revealed risk appetite of an organisation is defined here as the level and type of risk it actually accepts in practice, inferred from its resourcing decisions, its incentive structures and the deviations from its own standards that it tolerates without correction. It is distinct from the stated appetite expressed in policy: in the safety management system, the risk management system and in public commitments. This is the safety analogue of a distinction familiar in economics: that between stated and revealed preference. What a party says it values and what its choices show it values can diverge, and the choices are the better evidence.

The concept is not a new theory, and it is not offered as one. It reframes, in the vocabulary of governance, two well-established findings in the safety literature. Hopkins (2000, 2005, 2008, 2012) showed across a sequence of disaster studies that production pressure, the practical prioritising of output over safety, is a recurring organisational cause of catastrophic failure. Rasmussen (1997) modelled the same force as a gradient that pushes activity steadily toward the boundary of acceptable safety. The revealed appetite is the organisation-level decision field that sets the strength and direction of that gradient. Its value is not that it adds a mechanism but that it connects two audiences: boards and executives reason in the language of risk appetite, while practitioners and the safety literature reason in the language of production pressure and drift. Naming the revealed appetite makes the production-pressure mechanism legible to the people who set it, which is a precondition for changing it.

Two objections must be met for the idea to be meaningful. The first is circularity: if the revealed appetite is inferred from the risk-taking, it explains nothing. The answer is that the antecedents of revealed appetite are observable independently of the behaviour they produce. An organisation’s budget for controls, its staffing relative to workload, its bonus and schedule structures, and the specific deviations it has left uncorrected can all be read directly, before any incident, and they predict where the gradient points. The concept is therefore falsifiable in practice: assess an organisation’s revealed appetite from its decisions, then test whether sharp-end behaviour follows. The second objection is redundancy with production pressure. The reframing earns its place because production pressure, stated as such, does not travel into the boardroom, whereas risk appetite does. The contribution is translational, and for a practitioner audience that translation is the point. With those qualifications, the stated-versus-revealed distinction is retained as the core of the argument, and the rest of the review can be read as an account of how the revealed risk appetite enters the perception, the decisions and the body of the individual at the coalface.

2.5 Reading revealed appetite in practice: indicators and a worked case

For the construct to be usable rather than merely suggestive, it has to be observable. Revealed appetite can be read from five classes of indicator, each available before any incident and each independent of the front-line behaviour it is invoked to explain. They are set out in Table 1.

Table 1. Observable indicators of revealed risk appetite

A documented case shows the indicators carrying explanatory weight. In Hopkins’s (2000) analysis of the 1998 Esso Longford gas plant explosion in Victoria, which killed two workers, injured eight and cut gas supply across Victoria for an extended period, the organisation’s stated commitment to safety was not in doubt. Its revealed appetite, read through the indicators above, pointed the other way. Esso had relocated its engineers from the Longford site to Melbourne, stripping the plant of on-site engineering expertise (a resourcing decision); a hazard study of the relevant process had not been carried out (a hazard-study failure); and recurring abnormal plant conditions had been observed and worked around without correction over a long period (tolerated deviation). None of this required hindsight or the explosion to be visible: each was a standing feature of how the organisation allocated resources and tolerated departure from its own standards. The point is not that the operators who reintroduced warm oil to cold, embrittled metal were the cause; Hopkins’s argument is precisely that they were positioned by an organisation whose revealed appetite had already moved the safety boundary. The indicators would have registered that appetite in advance, which is what makes the construct diagnostic rather than circular.

3. The person: who takes risks and why they take them

3.1 Personality: thrill-seeking and acting on impulse

The longest-standing line of research treats risk-taking as partly a stable individual difference. The most developed single hypothesis is sensation seeking, defined by Zuckerman (1994) as the seeking of varied, novel, complex and intense sensations and experiences, and the willingness to take physical, social, legal and financial risks for the sake of such experience. Zuckerman’s Sensation Seeking Scale, refined across decades, separates facets including thrill and adventure seeking, experience seeking, disinhibition and boredom susceptibility, and brief measures such as the Brief Sensation Seeking Scale have since been validated for applied use (Hoyle et al. 2002). Impulsivity is a second concept, overlapping with sensation seeking but distinct from it. The Barratt Impulsiveness Scale operationalises it across attentional, motor and non-planning components (Patton, Stanford and Barratt 1995), and the tendency to discount delayed outcomes steeply in favour of immediate ones (one engine of impulsive choice) was given an early and durable behavioural account by Ainslie (1975). Whiteside and Lynam (2001) brought order to a crowded field by mapping impulsivity onto the five-factor model and identifying four facets (urgency, lack of premeditation, lack of perseverance and sensation seeking) in what became the UPPS model. What gets called impulsiveness is therefore several different things, and only some of them involve a taste for risk.

Beyond these focused theories, several instruments try to measure risk-taking propensity more directly. The Risk-Taking Index distinguishes risk-taking across six life domains and finds only moderate consistency across them (Nicholson et al. 2005), and short self-report measures of general risk propensity have been developed and validated for organisational settings (Meertens and Lion 2008). The most important finding to carry forward is that risk propensity is not one trait but many, and that they do not travel together. The Domain-Specific Risk-Taking scale shows that an individual’s willingness to take risks in the financial, health and safety, recreational, ethical and social domains is only weakly correlated across them, so a person who is bold in one is not reliably bold in another (Weber, Blais and Betz 2002; Blais and Weber 2006). That work also separates risk-taking attitude from risk perception. Much of the apparent difference in how much risk people take turns out to reflect not a differing taste for risk but a differing perception of how risky the activity is: people who engage in an activity often do so because they see it as less risky, not because they like risk more. This has a consequence for the word ‘propensity’. There is no single, transferable trait of riskiness, so the term is used here as shorthand for a family of weakly related, domain-bound dispositions, not for a unified characteristic that could be measured once and relied upon.

Personality’s link to actual workplace incidents is likely real but modest, and meta-analysis reads it far better than single studies do. A meta-analytic review of personality and occupational accident involvement found that low conscientiousness and low agreeableness were the most consistent dispositional predictors, with validities of a similar order, around the mid-twenties in correlation terms (Clarke and Robertson 2005). A later and more theoretically organised meta-analysis confirmed that broad personality traits relate to safety behaviour and outcomes in interpretable ways, while making clear that these are general tendencies of modest size, not strong individual predictors that could responsibly be used to sort people (Beus, Dhanani and McCord 2015).

Two further dispositions round out the picture. Locus of control, the degree to which a person believes outcomes are within their own control rather than down to chance or others, comes from Rotter (1966) and has been linked to safety-relevant behaviour. Self-efficacy, the belief in one’s capacity to carry out the actions a situation requires, comes from Bandura (1977, 1997) and cuts both ways, since confidence supports safe performance while overconfidence can support exposure. Taken together, stable individual differences do contribute to workplace risk-taking, but the contribution is small, domain-specific and heavily mediated by perception. That is a poor basis for selecting risk out of a workforce, and a strong reason to design work that tolerates the ordinary distribution of human behaviour.

3.2 How people judge risk when they are unsure

If personality answers part of who, decision science answers much of how. Human decision-making departs systematically from the rational ideal that classical economics assumes. Simon (1955) introduced bounded rationality: the recognition that real decision-makers have limited information, limited time and limited computational capacity, and so satisfice, choosing an option that is good enough rather than optimising. Tversky and Kahneman (1974) showed that under uncertainty people lean on a few judgmental heuristics, such as availability and representativeness, which are usually serviceable but produce predictable biases, including the systematic misjudgement of probabilities. Prospect theory (Kahneman and Tversky 1979) changed how choice under risk is understood. People evaluate outcomes as gains and losses against a reference point rather than as final wealth; they are loss averse, weighting losses more heavily than equivalent gains; and they are risk averse for gains but risk seeking for losses, so the same decision can flip depending on how it is framed. Cumulative prospect theory extended this to a wider range of choices and added the systematic overweighting of small probabilities and underweighting of moderate to large ones (Tversky and Kahneman 1992).

One element of this account warrants caution, because it has been seriously contested. Loss aversion, the claim that losses loom larger than equivalent gains, is among the most cited ideas in the behavioural sciences, and its status as a general law has been challenged. Gal and Rucker (2018) reviewed the evidence and argued that, on balance, current data do not support the proposition that losses are systematically more impactful than gains, and that the belief persists partly through its appeal. Their critique drew a direct rebuttal from Mrkva and colleagues (2020), who accepted that loss aversion has moderators and boundary conditions but concluded that reports of its death were greatly exaggerated. For a work safety audience the defensible reading is narrower than the headline: reference dependence and framing are well supported, and risk-seeking in the domain of losses is well supported, but the specific quantitative claim of a fixed loss-aversion multiplier should be treated as contested rather than settled. The framing implication survives the debate. A worker who frames a task as protecting something already counted on, such as finishing on time or avoiding a reprimand, is reasoning in the domain of losses and will tend to accept more risk. Framing is not a matter of presentation; it changes the choice.

A second strand of decision science concerns the architecture of thought itself. Dual-process accounts distinguish a fast, automatic, intuitive mode from a slow, effortful, deliberative one (Stanovich and West 2000; Evans 2008; Kahneman 2011). Much routine work runs on the fast mode, which is efficient and usually adequate to the outcome, but it is also where habit, rule of thumb and feeling dominate, and where a familiar hazard can be processed without a conscious risk assessment at all. Affect plays a related role. The affect heuristic describes how people read both risk and benefit off a quick overall feeling about an activity, so that activities which feel good are judged both more beneficial and less risky, even though risk and benefit are not actually inversely related in the world (Finucane et al. 2000). That finding has since been tested under the higher evidentiary standards now expected in psychology, and it held up: a well-powered pre-registered replication reproduced the inverse relationship between perceived risk and perceived benefit across several technologies, while qualifying it as somewhat asymmetric and sensitive to incidental mood (Efendić et al. 2022). Loewenstein and colleagues (2001) developed the risk-as-feelings hypothesis, showing that emotional reactions to a risky situation often diverge from cognitive assessments of it and, when they do, frequently drive behaviour. Slovic and colleagues (2004) drew this together into the distinction between risk as analysis and risk as feelings, arguing that the experiential, affective system is the older and the more powerful, and that deliberative assessment competes with it rather than replacing it. Gigerenzer (2014) enters a useful dissent: simple heuristics are not merely error-prone shortcuts but can be ecologically rational, performing well in the uncertain, information-poor settings where they evolved, and the deficit framing of human judgment can be overstated. The implication for practice is plain. Intuitive, feeling-led judgment is the human default at the sharp end; it is often good enough; and it turns dangerous exactly where a hazard is severe but does not feel dangerous. That is the situation engineering controls and isolation exist to address.

3.3 Why some hazards feel riskier than they are

How risky something feels, as distinct from how hazardous it is, is itself a structured and predictable thing, and two research traditions explain its shape. Both, it should be said at the outset, explain less of it than their prominence suggests. The psychometric paradigm grew from the foundational question of how safe is safe enough (Fischhoff et al. 1978) and was consolidated by Slovic (1987), who showed that lay perceptions of risk are organised along a small number of dimensions. The most important are dread (which captures perceived lack of control, catastrophic potential, dread of consequences and inequitable distribution) and the unknown, which captures how observable, how immediate and how well understood the risk is. Hazards that are voluntary, familiar, controllable and immediate in their effects are systematically perceived as less risky than hazards that are involuntary, unfamiliar, uncontrollable and delayed, whatever the actuarial numbers say. This explains a recurring frustration in safety practice. Many serious occupational hazards, such as repetitive manual tasks, noise, dusts and routine work at height, feel voluntary, familiar and controllable, and so are read as low risk by the people exposed to them even when the statistics say otherwise.

The cultural tradition widens the frame from the individual to the group. Douglas and Wildavsky (1982) argued that what a society or a group picks out as risky, and how much it worries, reflects its way of life and its values rather than any neutral reading of the hazards, so that risk perception is in part a social and moral act. The social amplification of risk framework joined the psychological and the social, describing how signals about a hazard are amplified or dampened as they pass through individuals, the media, institutions and cultural groups, producing social and economic effects out of proportion, in either direction, to the physical hazard (Kasperson et al. 1988).

Both traditions, though, are weak at predicting how risky a particular individual will judge a particular hazard. In a direct quantitative assessment, the psychometric model accounted for only about a fifth of the variance in perceived risk, and cultural theory for no more than 5 to 10 per cent (Sjöberg 2000). These are not the numbers of a strong predictive theory, and for the practitioner the perception literature is useful conceptually rather than predictively. It reliably identifies the direction of the bias, that familiar, voluntary and controllable hazards feel safer than they are, and it establishes that perception is partly a collective property of the work group, which can attenuate a real hazard into invisibility. It does not support confident prediction of who will misjudge which hazard by how much, and should not be cited as though it did. The cultural framing is a reminder that the group shares and shapes a sense of which hazards matter; it is not a measurement model.

3.4 Why risk-taking peaks in young people

Risk-taking is not spread evenly across the lifespan; it peaks in adolescence and young adulthood. The usual explanation is the dual-systems, or maturational-imbalance, model, though that explanation is more contested than its popularity suggests. The model runs as follows. A reward-processing system centred on the ventral striatum becomes highly responsive around puberty, while the cognitive control system, centred on the prefrontal cortex, matures more slowly and is still settling into the mid-twenties (Steinberg 2008; Casey, Getz and Galván 2008). For a few years the result is a mismatch: reward sensitivity is high while the capacity to restrain reward-driven impulses is still catching up, and behaviour tilts toward risk.

The model has been challenged, and the challenge bears on how the evidence is used. Romer, Reyna and Satterthwaite (2017) argued that the imbalance account blames too much adolescent harm on a structural deficit, and that the data support a more discriminating picture. They drew a line between two kinds of risk-taking. One kind, sensation seeking and impulsive action, does peak in adolescence and reflects exploration where the odds are unclear. The other, impulsive choice under known risk, declines steadily from childhood as executive function matures. On their reading, much adolescent risk-taking looks more like exploration than lost control, and the harmful form sits with a minority of young people who have weak self-regulation. Proponents of the dual-systems model have defended and refined it (Shulman et al. 2016), so this remains a live debate. For the workplace, the outcome changes little: both sides agree that younger workers show heightened sensation seeking and reward sensitivity, sharpest among peers, and that this is normal development, not a flaw of character.

The peer effect is well evidenced. In a controlled experiment, adolescents and young adults took more risks, and rated the benefits of risky options more highly, when they believed peers were watching, while adults were unmoved by the same manipulation, and risk-taking fell with age across the study (Gardner and Steinberg 2005). Hormonal state matters in adults too, though the evidence is thinner and warrants care. On a single trading floor, morning testosterone predicted a trader’s profit for the day, and cortisol rose with both the variance of returns and the level of uncertainty, suggesting that physiology shifts the appetite for risk at work and is in turn shaped by recent wins and losses (Coates and Herbert 2008). That finding came from a small, all-male sample on one floor, so it points to an interaction between physiology and risk-taking rather than proving a general one. Two things follow for practice. The greater appetite for risk in younger workers is developmental and sharpest in front of peers, so it should be designed for, not disciplined. And, as section 6 shows, this is not the group that Australian injury data put most at risk, which limits how far the point can be carried.

3.5 Why a taste for risk exists at all

The deepest layer of explanation asks why a disposition to take risks should exist at all. The evolutionary literature offers a coherent answer, and it is also the most speculative material in this review, treated accordingly. Costly signalling theory, originating with Zahavi (1975), holds that an organism can advertise underlying quality precisely by undertaking displays that are costly or dangerous, because only a high-quality individual can afford the cost. Applied to humans, conspicuous risk-taking can work as a signal of competence, confidence or status to rivals and to potential mates. This connects to a documented demographic pattern. Wilson and Daly (1985) described a young male syndrome of elevated risk-taking and competitiveness, which they linked to the intensity of male reproductive competition, and their broader work on lethal violence develops the argument that the same competitive psychology underlies a range of dangerous male behaviour (Daly and Wilson 1988).

Two cautions are needed. First, evolutionary-psychological explanations of specific modern behaviour are hard to test and easy to over-apply, and the claim that workplace risk-taking is a fitness display is plausible rather than demonstrated. Second, the empirical pattern it is invoked to explain is more modest than the theory implies. A meta-analysis of gender differences in risk-taking, across many studies and many tasks, found that men take more risks than women on average, that the difference is real but modest, and that its magnitude varies considerably by domain and by context and has shifted across age groups over time (Byrnes, Miller and Schafer 1999). So the evolutionary account explains, parsimoniously, why a capacity for risk-taking exists and why there is an average sex difference. It cannot carry the weight of explaining occupational death, and section 6 shows why: the Australian fatality pattern is driven far more by who is exposed to dangerous work than by any disposition. The material is included to complete the explanatory picture of the individual, not because it points to a lever for practice.

4. How traits and pressures turn into unsafe acts

Disposition, perception and decision-making do not act in a vacuum. A set of well-developed mechanisms describes how they become the specific behaviours practitioners observe, and how those behaviours drift over time. This section covers the mechanisms that operate largely at the level of the individual and the work group; section 5 turns to the organisational conditions that drive them.

4.1 Why safety gains get partly used up

One of the more provocative ideas in the field is that people carry a target level of risk and adjust their behaviour to keep perceived risk near it, so that a safety improvement which lowers perceived risk is partly offset by riskier behaviour elsewhere. Wilde (1982) proposed this as risk homeostasis theory. The strong version, that compensation is complete and safety measures therefore futile, is not supported by the evidence; the underlying phenomenon of risk compensation is real and well documented. The best known instance is the Peltzman effect, named for the economist who observed that mandated vehicle safety devices were associated with offsetting changes in driver behaviour (Peltzman 1975). A careful review concluded that behavioural adaptation to safety interventions does occur but is usually partial, and that how much occurs depends on four conditions: whether the change is visible to the person, whether it affects their perceived risk, whether they are motivated to use any margin gained, and whether they have the control to do so (Hedlund 2000). The lesson for practice is not that controls fail, but that a control which only changes perceived risk, while leaving the person visible margin, motive and the means to use it, will be partly consumed by behaviour. A control that removes the hazard, or removes the person’s ability to take the margin, is not subject to that leakage.

4.2 Mistakes versus rule-breaking

A foundational distinction for any practitioner is between an error and a violation, because the two have different psychology and call for different responses. Reason (1990) drew the line clearly. Errors are unintended: the person meant to do the right thing and failed, whether through a slip or lapse in executing a good plan or through a mistake in the plan itself. Violations are intended deviations from a known rule or procedure, where the action is deliberate even if the harm is not. Reason’s generic error-modelling system organised the unintended failures, while later work built a taxonomy of the deliberate ones (Reason, Manstead et al. 1990; Reason, Parker and Lawton 1998; Lawton 1998). Violations are usually divided three ways. Routine violations are habitual corner-cutting that has become the normal way of working, often because the rule is unworkable or the shortcut is simply more efficient. Situational violations are forced by the conditions of the task: a procedure that cannot be followed with the time, tools or staffing provided. Optimising or thrill violations are undertaken for their own sake. The practical force of this is large. Errors call for better design, better tools and less demand on memory and attention. Violations call for an inquiry into why the rule is being broken, because routine and situational violations are usually evidence that the rule is wrong, the resourcing is inadequate or the deviation is being quietly rewarded, while the smaller number of optimising violations reflect a genuine choice to take risk for its own sake and may warrant a different response. Discipline the individual for a situational violation and the cause stays in place; ignore a genuinely reckless optimising violation and a real hazard stays in place.

4.3 Behaviour-based safety: what it can and cannot do

The recognition that behaviour matters gave rise to behaviour-based safety, an approach with a respectable research pedigree and a serious critique. Early experimental work showed that defining safe behaviours, observing them and giving feedback could measurably increase safe performance (Komaki, Barwick and Scott 1978), and the approach was built out into structured programs (Geller 2001). Well-implemented behavioural approaches can change the targeted behaviours in the short term. The critique is substantial, and it matters for how the approach is used. By putting the spotlight on front-line workers, behaviour-based safety can imply that workers are the primary problem, can divert attention from the upstream conditions that produce unsafe behaviour, and can encourage under-reporting where observation is tied to consequences. A more balanced position holds that behavioural and situational factors operate together, and that interventions should address the organisational and design causes of behaviour rather than the behaviour alone (DeJoy 2005). Used well, observation becomes a diagnostic of those upstream causes; used as a tool for correcting individuals, it suppresses the reporting safety depends on.

4.4 How small shortcuts quietly become normal

Perhaps the most important body of work for understanding serious failures is the literature on how organisations slide, gradually and without alarm, into dangerous states. Vaughan (1996), in her study of the Challenger disaster, named the normalisation of deviance: a deviation from the expected standard, once it has occurred without harm, is reinterpreted as acceptable, so the boundary of acceptable risk is redrawn outward in small, locally reasonable steps until catastrophe arrives. Her later work generalised this to the dark side of organisations, the routine production of mistake, misconduct and disaster by ordinary organisational processes (Vaughan 1999). Rasmussen (1997) supplied the systems account, modelling the workplace as a space bounded by economic failure, unacceptable workload and acceptable safety, and arguing that natural gradients, the pressure toward efficiency and the pressure toward least effort, push activity steadily toward the safety boundary, where an accident becomes possible. Snook (2000), studying a friendly-fire shootdown, described practical drift, the slow uncoupling of actual practice from written procedure as local adaptations accumulate. Dekker (2011) drew these into the idea of drift into failure: accidents in complex systems often emerge not from a single broken component but from the system’s normal, adaptive behaviour gradually eroding its own margins. Amalberti (2001) added the paradox that ultra-safe systems face distinctive pressures, and that migration toward boundaries is a property of the system rather than of careless individuals. The decisive point is consistent across all of it. The most dangerous risk-taking is rarely a dramatic individual act. It is a slow, collective, unremarkable normalisation of small deviations, each of which seemed sensible at the time, and it is a property of systems under pressure that interventions aimed at individual attitude cannot reach.

The contemporary extension of this systems view is the body of work known as Safety-II and resilience engineering, which shifts attention from why things occasionally go wrong to how work usually goes right, and treats the everyday variability and adaptation of normal work as the source of both success and failure rather than as deviance to be stamped out (Hollnagel 2014). Building on it, Provan and colleagues (2020) recast the role of the safety professional around creating foresight about the changing shape of risk and enabling people to adapt safely, rather than enforcing compliance after the event. This ‘new view’ bears directly on the person-versus-situation question at the centre of this review: it reinforces the finding that front-line behaviour is shaped by system conditions, while cautioning, as the drift literature does, that several of its propositions remain conceptual and are not yet matched by strong predictive validation.

4.5 Trading thoroughness for speed

A unifying way to read much of the above is Hollnagel’s efficiency–thoroughness trade-off principle (Hollnagel 2009). People and organisations constantly trade thoroughness – doing everything needed to be sure of a good outcome – against efficiency – doing enough to get the result with acceptable effort and cost. The trade-off is necessary, and usually benign, since no one can be maximally thorough about everything. But it is the mechanism through which production pressure becomes a shortcut, and through which the gap between work as imagined in the procedure and work as actually done opens up. The question is rarely whether people will make the trade-off. It is where the organisation has set the balance, and whether it has made thoroughness affordable.

5. The workplace conditions that drive risk-taking

The individual mechanisms of section 4 are driven by conditions the organisation creates. The evidence in this section shows, repeatedly, that the situation explains more about workplace risk-taking than the person does, though it carries the correlational cautions noted in the method.

5.1 Pressure to get the job done

Across the study of major accidents, the most consistent finding is that production pressure, the implicit or explicit prioritising of output over safety, is a recurring contributor. Hopkins, across a sequence of detailed case studies of Australian and international disasters, showed how schedule and cost pressure, combined with structures that diffused responsibility and learning, produced catastrophic failures later, wrongly, attributed to front-line error (Hopkins 2000, 2005, 2008, 2012). Rasmussen’s (1997) gradients give the mechanism: the pressure toward efficiency is constant, it is rewarded and it is exerted on everyone, so without a countervailing force the system migrates toward the safety boundary. In the terms of this review, production pressure is the principal channel through which an organisation’s revealed appetite for risk reaches the individual. A worker who takes a shortcut under time pressure is enacting an appetite the organisation has set, whether or not any policy admits it.

5.2 How seriously a workplace really takes safety

Safety climate, the shared perception among workers of how safety is actually prioritised and managed in their workplace, is the most heavily researched organisational construct in the field. Zohar (1980) introduced it, and later sharpened the point that climate is fundamentally about the alignment between what management says about safety and what it actually rewards and tolerates (Zohar 2000, 2010). Safety culture, the deeper and more durable set of shared values and assumptions, was reviewed and clarified by Guldenmund (2000), who showed how loosely the term had been used and located it within established models of organisational culture. Westrum (2004) offered an influential typology separating pathological organisations, which suppress safety information, from bureaucratic ones, which tolerate it, and generative ones, which actively seek it. Parker, Lawrie and Hudson (2006) developed a five-level maturity ladder running from a pathological culture, through reactive, calculative and proactive stages, to a generative one, which has become a common diagnostic frame. The Westrum typology and the maturity ladder are conceptual, practitioner-facing frameworks. They are useful for structuring a conversation and tracking direction, but they have limited predictive validation against injury outcomes, and they should be offered to clients as organising heuristics, not as measured predictors.

What the climate construct delivers empirically is qualified, and must be read with the correlational cautions in mind, since most of this evidence is cross-sectional and self-report. A meta-analytic review found that safety climate was reliably associated with safety behaviour, both compliance and participation, but that the onward link from climate to actual accidents and injuries was weak, which suggests climate works largely by shaping behaviour rather than translating directly into injury rates (Clarke 2006). The most comprehensive integrative meta-analysis of the antecedents of safety outcomes found that situational and contextual factors – including safety climate and the safety-related leadership and design of the work – were generally stronger predictors of safety performance and outcomes than individual-difference factors (Christian et al. 2009). This anchors the review, but it should be stated as an interactionist conclusion, not as a victory of situation over person. The situation predicts more because it sets the field within which everyone acts; it does not follow that disposition is irrelevant, only that it operates within, and is generally dominated by, that field. The claim is not that the person does not matter. It is that the lever with the most leverage, and the lever the law assigns to the duty holder, is the situation.

5.3 What makes people follow safety rules and push further

Researchers have usefully modelled the behaviour through which climate and leadership act. Griffin and Neal (2000) separated safety compliance – meeting the required standards and following procedures – from safety participation – the discretionary behaviours such as helping colleagues, raising concerns and contributing to a safe environment that are not strictly required. Both, they showed, are driven by safety knowledge and safety motivation, which are in turn shaped by safety climate, and the model was extended in later work (Neal and Griffin 2006). A job demands–resources meta-analysis broadened the account: job resources support safety motivation and engagement, while job demands erode them through strain and burnout, so the design of the job itself – its demands and its resources – feeds into safety behaviour (Nahrgang, Morgeson and Hofmann 2011). Compliance can be specified and required. Participation cannot. It includes the reporting every safety system depends on, and it has to be earned through the conditions of work and the conduct of management.

5.4 The kind of leadership that improves safety

Leadership is one of the stronger situational levers, and the evidence is reasonably specific about which kind matters, though it too is largely correlational. A field study found that transformational leadership – leadership that sets a vision, models commitment and attends to individuals – was associated with better safety climate and fewer injuries (Barling, Loughlin and Kelloway 2002). A meta-analytic review confirmed that transformational leadership, and active forms of transactional leadership that monitor and respond to safety performance, were associated with safety behaviours, while passive leadership, which intervenes only after problems escalate, was associated with worse outcomes (Clarke 2013). These designs do not establish causation, so the safest statement is the modest one: visible, active, safety-specific leadership is consistently associated with better safety behaviour, and is therefore a well-supported candidate for intervention rather than a proven cause. It is among the better-evidenced situational factors a practitioner has to work with.

5.5 How the crew shapes risk-taking

Risk-taking is also shaped by the group, and the social-psychological evidence is long-standing. Group discussion does not simply average individual positions; it tends to push the group further in the direction it was already leaning, first observed as the risky shift (Stoner 1961) and later understood more generally as group polarisation (Moscovici and Zavalloni 1969). Put that together with the developmental evidence on peer effects in younger workers (Gardner and Steinberg 2005), and a crew can collectively endorse a level of risk that few of its members would choose alone. The work group is a unit of risk-taking in its own right, not merely a collection of individuals.

6. Who is harmed in Australia, and what the law requires

6.1 Who is being hurt at work now

The behavioural science has to be read against the actual pattern of harm in Australian workplaces, and the national data carry several lessons that bear directly on how the preceding evidence should be applied. On the most recent national statistics, there were 188 worker fatalities in Australia in 2024, down from 200 the year before, against a five-year average of around 191 a year (Safe Work Australia 2025, p. 3). The fatality rate for 2024 was around 1.3 per 100,000 workers, which sits below the five-year average of 1.4 and represents a fall of about a quarter over the preceding decade, down from 1.7 in 2014 (Safe Work Australia 2025, p. 3). Serious workers’ compensation claims tell a different story. They numbered 146,700 in the most recent reporting period – more than 400 a day – and have risen substantially over the decade (Safe Work Australia 2025, p. 8). Of those serious claims, around two-thirds were for injuries and one-third for diseases and conditions, and the disease share has grown over the decade (Safe Work Australia 2025, p. 14). Fatalities were concentrated in a few industries, with 80 per cent occurring in six, and those same six accounted for 61 per cent of serious claims (Safe Work Australia 2025, p. 2). The leading mechanisms of traumatic fatality were vehicle incidents, at around 42 per cent, falls from a height, and being hit by moving objects (Safe Work Australia 2025, p. 5). Conditions involving mental health now account for 12 per cent of serious claims, a share that has grown sharply, and these claims carry a median time off work several times longer than other serious claims (Safe Work Australia 2025, pp. 14–15). Two features of this distribution complicate the behavioural picture, and both need careful handling. Table 2 summarises the key figures.

Table 2. Key Australian work health and safety statistics (Safe Work Australia 2025)

6.2 Reading the sex and age figures correctly

The first feature is sex. The overwhelming majority of worker fatalities, around 96 per cent (180 of 188 in 2024), are men, and the male fatality rate is many times the female rate (Safe Work Australia 2025, p. 4). Serious claims are made more often by men too, who accounted for 58 per cent of them, though the female share has been rising with female workforce participation, and the male serious-claims frequency rate is only modestly higher than the female rate once hours worked are taken into account (Safe Work Australia 2025, p. 10). This pattern fits the evolutionary and developmental evidence on male risk-taking reviewed above, and the easy conclusion is that male risk propensity drives occupational harm. That conclusion is incomplete, and for a practitioner it is misleading. Men are heavily over-represented in the highest-exposure occupations and industries. The six industries that account for most fatalities are male-dominated, and construction alone, one of those six, is around 86 per cent male, with women making up only about 14 per cent of its workforce (Master Builders Australia 2024, p. 7; Australian Bureau of Statistics 2024, Table EQ06). The fatality disparity therefore reflects exposure to hazardous work at least as much as any difference in disposition, and the meta-analytic evidence is in any case that the average sex difference in risk-taking is modest and varies by context (Byrnes, Miller and Schafer 1999). The gendered pattern of harm is not uniform either: women carry a markedly higher share of mental health claims – around 17 per cent of their claims against 8 per cent for men – a pattern more plausibly explained by the different work the sexes do than by any difference in character, though the data establish the pattern and not its cause (Safe Work Australia 2025, p. 10). Attributing the death toll to male recklessness both overstates the dispositional evidence and points away from the controllable cause: the hazardous design and conduct of the work.

The second feature is age, and it cuts hard against the most intuitive application of the developmental evidence. The neuroscience establishes that younger workers show elevated sensation seeking and reward sensitivity, especially among peers, so it would be natural to expect a youth-dominated injury profile. The Australian data do not show one. Once hours worked are taken into account, the serious-claims frequency rate is highest among the oldest workers – at around 9.5 claims per million hours worked for those aged 55 to 64 and around 10.0 for those aged 65 and over – and lowest among workers aged 25 to 44, at around 5.8; workers under 25 sit a little below the all-claims average, at around 6.4 claims per million hours worked, against an all-claims average of 6.8 (Safe Work Australia 2025, pp. 8 and 11). Frequency, time lost and cost of serious claims all generally rise with age, and the share of claims borne by older cohorts is rising as the workforce ages (Safe Work Australia 2025, p. 11). Fatalities skew older as well, with workers aged 45 and over accounting for around half of the total (Safe Work Australia 2025, p. 4). The reasons are several: the concentration of older workers in some high-hazard roles, the cumulative and latent nature of some fatal exposures, and physiological vulnerability and longer recovery in older workers. The practical consequence is clear. The developmental risk-taking evidence is sound as an account of behaviour and of specific high-energy, peer-influenced contexts, and it should be applied there. It is not an account of who is actually injured or killed in Australian workplaces, where older workers carry the higher injury frequency and the heavier fatalities. A safety program that treats the young and the bold as the central problem is aiming at the wrong population for the outcomes that matter most. Once again, the data redirect attention to the hazardous work rather than the worker’s age.

6.3 What the law’s logic tells practitioners to do

The law, as set out in section 2, puts the duty on the person conducting the business or undertaking, not on the worker, and it requires the elimination or minimisation of risk so far as is reasonably practicable through a hierarchy that prefers controls independent of individual behaviour (Work Health and Safety Act 2011 (Qld) ss 18, 19; Work Health and Safety Regulation 2011 (Qld) r 36). Lay that legal logic over the behavioural science and the two point the same way. The science says risk-taking is normal, motivated, perception-dependent and driven more by situation than by disposition, and that controls relying on people behaving as instructed are partly defeated by risk compensation and by the slow drift of practice. The law already prefers controls that remove the hazard or the exposure over controls that depend on instruction, training and supervision. A duty holder who reaches for behaviour change and toolbox talks while a higher-order control is reasonably practicable is on weaker ground scientifically and, in substance, on weaker ground legally, because the hierarchy required the higher control to be considered first (Work Health and Safety Regulation 2011 (Qld) r 36; Safe Work Australia 2013).

6.4 Paperwork versus what happens on the job

A final Australian and applied theme is the relationship between written rules and actual work, which the behavioural evidence on violations and drift makes unavoidable. Hale and Borys (2013a, 2013b) reviewed the management of workplace safety rules and contrasted two paradigms. In the first – top-down and rationalist – rules are fixed and comprehensive limits on the operator’s freedom, deviations are violations to be suppressed, and more rules are the answer to failure. In the second – bottom-up and constructivist – rules are local, situated and necessarily incomplete, expert operators must adapt them to the diversity of real conditions, and competence consists partly in knowing how to adapt safely. Neither paradigm alone is adequate, they argue, and good rule management draws on both; they also draw the general principle that explicit written rules should not be the first means of control proposed for a hazard, with design and layout taking precedence. Borys (2012), studying safe work method statements in Australian construction, found that these documents often functioned as compliance artefacts produced for audit rather than as live tools that shaped work – the gap between the paper system and the practised system that the drift literature predicts. Quinlan, Mayhew and Bohle (2001) added the structural dimension, showing through an extensive review that precarious and contracted employment, work disorganisation and the fracturing of responsibility are associated with worse occupational health and safety outcomes, partly because they degrade the conditions, training and continuity that safe work depends on. Three things follow. Paperwork is not practice. A rule that cannot be followed as written will be adapted or ignored. And the structure of the workforce itself, including how much of it is contracted and precarious, is a determinant of risk-taking that sits well above the level of the individual.

7. Putting it together: how work drifts toward danger

The traditions surveyed above are usually presented in parallel, as this review has so far presented them. They can instead be integrated within a single model, and doing so both adjudicates between them and shows the practitioner where the leverage lies. The model is Rasmussen’s (1997), and the other literatures sit inside it. The model earns that role by its integrative reach rather than by independent predictive validation; like the safety-culture maturity ladders discussed in section 5.2, it is best treated as a conceptual framework that organises the evidence, not as a measured predictor of harm, and it is held to that standard here.

Rasmussen pictured any work system as occupying a space bounded on three sides: a boundary of economic failure, beyond which the enterprise is not viable; a boundary of unacceptable workload, beyond which people cannot or will not go; and a boundary of acceptable safety performance, beyond which an accident becomes likely. Activity does not sit still inside that space. Two gradients act on it constantly. Management pressure for efficiency pushes activity away from the economic boundary, and the human tendency toward least effort pushes it away from the workload boundary. Both push the same way – toward the boundary of acceptable safety. Unless a counter-gradient is deliberately created, the system drifts toward that safety boundary, and the drift is made of small, individually reasonable steps. Figure 1 represents the model schematically.

Figure 1. The migration of work toward the boundary of acceptable safety (adapted from Rasmussen 1997; conceptual, not to scale)

Both the efficiency gradient (management pressure for output) and the least-effort gradient (the human tendency to economise effort) push the operating point in the same direction: toward the boundary of acceptable safety. The organisation’s revealed appetite sets the strength of the efficiency gradient, and, as the drift literature shows, the safety boundary itself can migrate outward as deviations are normalised.

Each tradition in this review describes one part of that picture. The organisation’s revealed appetite, defended in section 2.4, sets the strength and direction of the efficiency gradient: the budget for controls, the staffing relative to workload, the incentive structure and the deviations tolerated together determine how hard the system is pushed toward the safety boundary, independently of any individual. The decision-science and perception literatures of sections 3.2 and 3.3 describe how an individual positioned near that boundary reads the situation: framing places the choice in the domain of losses, affect and the fast intuitive system process familiar hazards without alarm, and the dread and unknown dimensions make exactly the wrong hazards feel safe. The dispositional literature of section 3.1 shifts the distribution of where individuals sit, modestly and by domain, without fixing any individual’s position. The developmental, endocrine and group literatures of sections 3.4 and 5.5 amplify or dampen the movement, with peers and group polarisation pushing crews further toward the boundary than individuals would go alone. And the drift literature of section 4.4 describes the boundary itself moving: as deviations are normalised, the line that counts as acceptable migrates outward, so a system can be travelling toward catastrophe while every local actor believes they are operating safely.

The model resolves the person-versus-situation question that the parallel presentation leaves hanging. The situation – expressed as the revealed appetite and the gradients it sets – determines the field of action and the location of the boundary, which is why situational factors predict safety outcomes more strongly than dispositional ones (Christian et al. 2009). The person – expressed as perception, framing, affect and disposition – determines an individual’s trajectory within that field. Neither is sufficient alone, and the relationship is interactionist rather than competitive. But the two are not symmetrical in what they offer the duty holder. The field can be re-engineered by the organisation; the trajectory of any one individual within it cannot be reliably controlled. That asymmetry – not a denial that the person matters – is why the evidence and the law both direct effort at the situation.

Read through this model, the findings converge on a set of propositions, stated here as the basis for the practical guidance that follows. First, stated and revealed risk appetite differ, and workers respond to the revealed one, because it sets the gradient they actually feel. Second, risk-taking is normal, motivated and usually locally rational; it is rarely deviance, and is not, in the main, a defect of character. Third, there is no single trait of riskiness: propensity is domain-specific, small in effect and heavily mediated by perception. Fourth, felt risk and actuarial risk diverge in patterned ways, with familiar, voluntary and controllable hazards feeling safe even when they are not. Fifth, safety improvements that only change perceived risk are partly consumed by behaviour, so controls that remove the hazard or the exposure are systematically more reliable. Sixth, serious failure is usually a slow, collective drift in which the boundary itself moves, not a single reckless act. Seventh, the situation predicts workplace risk-taking better than the person does, because the situation sets the field, though the relationship is interactionist and the person is not irrelevant. Eighth, the Australian demographic patterns are real but must be read with care: the developmental evidence explains youthful risk-taking behaviour, yet older workers carry the higher injury frequency and the heavier fatalities, and the male predominance in fatalities reflects exposure to dangerous work more than disposition – so the controllable variable is the hazardous work, not the worker’s age or sex.

These conclusions sit alongside genuine, unresolved tensions that the model does not dissolve. The strong form of risk homeostasis is unsupported, even though risk compensation is real. Loss aversion as a fixed quantitative law is contested, even though reference dependence and loss-domain risk-seeking are well established. The dual-systems account of adolescent risk is the subject of a live debate. The predictive validity of the perception models and the safety-culture maturity ladders is modest, so they are better used as conceptual aids than as measurement tools. And the causal status of the climate, leadership and motivation evidence is limited by its correlational design. None of these reverses the central direction of the evidence. Each one bounds the confidence with which a particular claim can be made, and a practitioner who carries the tensions as well as the conclusions will give better advice.

8. What this means in practice

This section draws the review into practice. It is organised around the decisions a practitioner actually makes, and each recommendation rests on the evidence set out above. It deliberately keeps, rather than discards, the legitimate role of individual capability and accountability, because an honest reading of the evidence places those within the systemic account, not against it. Table 3 maps each body of evidence to the synthesis proposition it supports and the recommendation that follows.

Table 3. From theory to synthesis to practice

8.1 Treat most unsafe acts as a symptom not individual character flaws

The most consequential shift is in diagnosis. Because risk-taking is normal, motivated and situationally driven, an unsafe act should usually be treated as the visible end of a causal chain that runs back into the design of the work, the resourcing of the task and the conditions set by management, rather than as a defect to be corrected in the person (Reason 1990; Christian et al. 2009; DeJoy 2005). In practice, when a practitioner observes or investigates risk-taking, the first questions should be about the situation. What was the time and production pressure? What tools and staffing were available? Could the procedure be followed as written? What behaviour was actually being rewarded? The error and violation distinction should be used explicitly, because routine and situational violations are nearly always evidence that the rule is wrong or the resourcing inadequate (Reason, Parker and Lawton 1998; Lawton 1998). Even so, this is not an argument against individual accountability in every case. A small class of optimising or reckless violations reflects a genuine, informed choice to take risk for its own sake. A just-culture response separates those from the error and the forced violation, holding individuals fairly accountable for the former while redesigning the conditions that produce the latter. The error is to stop the inquiry at the individual by default, not to recognise individual responsibility where it genuinely sits.

8.2 Read what the organisation rewards and not just what it says

Practitioners should treat the gap between stated and revealed risk appetite as a primary object of assessment, and section 2.4 shows that the revealed appetite can be read from observable decisions before any incident. The stated appetite is in the policy. The revealed appetite is in the schedule, the bonus structure, the staffing levels and the shortcuts that go unremarked (Power 2004; Rasmussen 1997; Zohar 2000). A practical diagnostic is to ask what the organisation actually does when safety and production conflict at the front line, because that answer – not the policy statement – is the appetite the workforce is responding to. Where the two diverge, the most useful thing a practitioner can do is make the divergence visible to the people who set the production pressure, since the revealed appetite is set above the worker and can only be changed there. Auditable documents and registers should be watched for the failure mode the critical literature identifies, in which producing assurance is mistaken for controlling the hazard (Power 2007).

8.3 Prefer controls that do not rely on people behaving

The hierarchy of control is not only a legal requirement but the most direct application of the behavioural evidence, and practitioners should defend it on both grounds (Work Health and Safety Regulation 2011 (Qld) r 36). Intuitive judgment misreads familiar hazards. Risk perception is an unreliable guard for exactly the hazards that matter. Any margin left to the worker is partly consumed by risk compensation. So controls that eliminate the hazard, substitute a lesser one, isolate it or engineer it out tend to be more reliable in practice than administrative controls, training and personal protective equipment, because they do not depend on perception or behaviour holding up over time (Slovic et al. 2004; Hedlund 2000; Finucane et al. 2000). Where a higher-order control is reasonably practicable, reaching first for behaviour change is weaker practice and, given the structure of the duty, weaker compliance. Where reliance on administrative controls or protective equipment is unavoidable, practitioners should expect behavioural adaptation and design against it – by removing the visible margin, the motive or the means that lets the gain be taken back (Hedlund 2000).

8.4 Use training for what it can do and build real skill where skill is the control

Training and behaviour-based safety have a place, and the evidence defines that place. Behavioural methods can change targeted behaviours in the short term, but they do not by themselves address the upstream conditions that generate unsafe behaviour, and tying observation to consequences suppresses the reporting safety depends on (Komaki, Barwick and Scott 1978; DeJoy 2005). Practitioners should therefore use behavioural observation mainly as a diagnostic that surfaces situational causes, not as a corrective aimed at individuals, and should be sceptical of programs that locate the problem in worker attitude while leaving production pressure, design and resourcing untouched. The systemic account does not make competence irrelevant. For some hazards – especially those that cannot be fully engineered out – genuine skill, knowledge and supervision are themselves part of the control, and the dispositional evidence that self-efficacy supports safe performance (Bandura 1997) is a reason to build real competence rather than nominal compliance. Training is a weak control for a hazard that could be eliminated, and it should never be the highest control offered where elimination or engineering is reasonably practicable. Well-designed competence and supervision, though, are a legitimate and sometimes necessary layer, not a failure of nerve.

8.5 Design for real people including the young and the old

Because risk propensity is domain-specific, modest in its effect on outcomes and impossible to select out reliably, the workplace should be designed to tolerate the ordinary range of human dispositions instead of depending on recruiting a uniformly cautious workforce (Clarke and Robertson 2005; Beus, Dhanani and McCord 2015). For younger workers in particular, the developmental and peer-effect evidence supports specific, non-punitive measures: structuring early work so that high-energy tasks are not performed unsupervised among peers, providing genuine mentoring rather than nominal sign-off, and treating the elevated sensation seeking of the young as a designable feature – strongest in the presence of peers – rather than a character problem (Gardner and Steinberg 2005; Steinberg 2008; Romer, Reyna and Satterthwaite 2017). The group itself should be treated as a unit of risk, since crews can collectively endorse exposures that individuals would refuse, which is an argument for supervision and control design that does not assume the group will self-moderate (Stoner 1961; Moscovici and Zavalloni 1969).

8.6 Target dangerous work

The national data should be used to direct effort toward hazardous work, not toward demographic groups, and section 6.2 shows why the intuitive reading is wrong. Fatalities are overwhelmingly male and skew older, and serious-claim frequency is highest among the oldest workers, all of which reflects who is exposed to the most dangerous tasks at least as much as any difference in disposition (Safe Work Australia 2025; Master Builders Australia 2024; Byrnes, Miller and Schafer 1999). A program that frames young or male workers as the problem will misallocate effort, particularly for the most serious outcomes, and will tend toward blame rather than control. The young-worker behavioural evidence should be applied where it holds – to specific high-energy and peer-influenced contexts. The resources for preventing death and serious injury should follow the exposure, which means attending to older workers too, whose injury frequency and recovery burden are the heaviest.

8.7 Use climate and culture surveys but truly know their limits

Safety climate is worth measuring, because it predicts safety behaviour and because it is a usable leading indicator of how safety is actually prioritised, and the maturity models offer a defensible way to structure and track progress (Zohar 2000; Parker, Lawrie and Hudson 2006). Two limits should be stated honestly. The link from climate to behaviour is solid, but the onward link from climate to injury is weak, so a good climate score must not be presented as proof of a safe site (Clarke 2006). And the maturity ladders and culture typologies have limited predictive validation, so they should be offered as organising heuristics, not as measured predictors of harm. The most defensible use of climate measurement is to read the alignment between what management says and what it rewards and tolerates – because that alignment is the substance of climate – and to pair it with hard evidence about the controls actually in place.

8.8 Invest in visible and hands-on safety leadership

Because leadership is one of the better-evidenced situational factors, practitioners should treat the active, visible engagement of managers in safety as a high-value intervention, while describing it accurately. Transformational and active transactional leadership are consistently associated with better behaviour and outcomes, and passive leadership that responds only after escalation is associated with worse ones, though the evidence is correlational rather than causal (Barling, Loughlin and Kelloway 2002; Clarke 2013). Securing genuine, monitored, safety-specific leader behaviour – and removing the passive default in which managers attend to safety only after a problem has grown – is among the higher-value moves available, and one practitioners should press for at the level where production pressure is also set.

8.9 Make it safe to report and to speak up

Every safety management system depends on discretionary behaviour: the reporting of incidents and near misses, the raising of concerns, the willingness to stop unsafe work. That behaviour is participation, not compliance. It cannot be mandated, and it is earned through the conditions of work and the conduct of management (Griffin and Neal 2000; Nahrgang, Morgeson and Hofmann 2011). Practitioners should therefore protect the conditions that make participation safe for the individual: separating learning processes from blame, ensuring that raising a concern carries no penalty, and designing jobs whose demands and resources do not erode the motivation to engage. This connects directly to the reframing in section 8.1. An organisation that punishes the visible symptom destroys the reporting it needs to find the cause, which is also why the place kept for individual accountability must be confined to genuine reckless choice and applied through a just-culture process, not extended to honest error.

8.10 Keep procedures usable and constantly watch for shortcuts creeping in

Finally, because practice drifts from procedure, and because rules that cannot be followed as written will be adapted or ignored, procedures should be designed as usable tools rather than compliance artefacts, and the gap between work as imagined and work as done should be monitored as a live indicator (Hale and Borys 2013a, 2013b; Borys 2012; Hollnagel 2009). Practical measures include involving the people who do the work in writing the rules that govern it, testing whether a procedure can actually be followed with the time, tools and staffing provided, and looking specifically for normalised deviations – the small shortcuts that have quietly become the standard way of working – since those are the early signs of the drift that precedes serious failure (Vaughan 1996; Snook 2000; Dekker 2011). The structural conditions of the workforce, including the extent of contracting and precarious employment, should be recognised as determinants of this drift, not treated as outside the practitioner’s scope (Quinlan, Mayhew and Bohle 2001).

9. Limitations

The caveats stated through the review are gathered here so that its claims are read with the right confidence. First, this is fundamentally a narrative review: the literature was selected by the author’s judgment rather than by a reproducible protocol search (section 1.2), so it is open to selection bias and makes no claim to exhaustive coverage. Second, the empirical backbone ie the integrative meta-analyses of safety climate, leadership and motivation and dates are largely from 2005 to 2015, and although this review engages more recent person-versus-system work (Hollnagel 2014; Provan et al. 2020), the core quantitative estimates have not been refreshed against a comparable body of post-2015 meta-analysis. Third, that backbone is predominantly correlational, cross-sectional and self-report, so the causal language the field uses, and which this review sometimes adopts in its recommendations, is not fully earned by the designs. Fourth, two of the review’s organising devices such as the revealed-appetite construct (section 2.4) and Rasmussen’s migration model (section 7)are conceptual frameworks valued for their integrative reach, not measurement instruments with independent predictive validation; the same caution applies to the perception models and the safety-culture maturity ladders. Fifth, the legal analysis rests on a single jurisdiction, the model law as enacted in Queensland, with Victoria noted but treated more briefly. Finally, the evolutionary material (section 3.5) is the most speculative content in the review and is included to complete the explanatory picture, not as a basis for practice. None of these failings points the other way on the central question; each simply marks the limits within which a given claim should be read.

10. Conclusion

The evidence in this review points to one practical change in how safety is handled at work places. Most risk-taking at work is not the work of careless or reckless people; it is inherently normal behaviour for most. Personality definitely plays a small part, and so does the way people see and judge risk. But the largest influence by far is the conditions the organisation creates such as the time and cost pressure, the example managers set, the way the job is designed and the equipment it uses, and the gap between the risk an organisation says it will accept and the risk it actually allows.

These pieces fit neatly together in that work tends to drift toward the edge of what is safe, because the pressure to be quick and to save effort keeps pushing it that way. What an organisation rewards, and what it lets slide, decides how strong that push is. An individual’s own judgment and personality then shape the path taken within those limits. A group can carry a crew further than any single member would go alone. And over time, the line for what counts as normal quietly moves outward. The most powerful place to act, therefore, is on the situation rather than the individual.

This of course does not let the individual off the hook as skill, supervision and holding someone fairly responsible for a genuinely reckless choice all still matter at any workplace. Those characteristics simply sit inside this larger picture rather than against it.

Two factors reinforce the conclusions here in that Australian work health and safety law already leans toward controls that do not depend on how a person behaves and the national figures point to dangerous work, not to the age or sex of the worker.

For practitioners, the message is to focus safety effort on the conditions that create risk, not on the worker at the end of the chain. That means definitively reading what an organisation really rewards and allows, and closing the gap between that and what its policy says. It ultimately means choosing controls that remove the hazard ahead of controls that rely on people following instructions. It means using training and behaviour programs for the real but limited good they do, while building genuine skill where skill is itself the control. It means designing work for the normal range of people, not for an ideal worker who, in essence, never exists. It means aiming work safety effort at hazardous and dangerous work rather than at the kind of person doing it.

Done well, this aligns everyday practice with both the science of why people take risks and the law that sets the duty to keep people safe.

References

Ainslie, G 1975, ‘Specious reward: a behavioral theory of impulsiveness and impulse control’, Psychological Bulletin, vol. 82, no. 4, pp. 463–496, https://doi.org/10.1037/h0076860

Amalberti, R 2001, ‘The paradoxes of almost totally safe transportation systems’, Safety Science, vol. 37, no. 2–3, pp. 109–126, https://doi.org/10.1016/S0925-7535(00)00045-X

Australian Bureau of Statistics 2024, Labour Force, Australia, Detailed, ABS, Canberra, https://www.abs.gov.au/statistics/labour/employment-and-unemployment/labour-force-australia-detailed

Baethge, C, Goldbeck-Wood, S & Mertens, S 2019, ‘SANRA—a scale for the quality assessment of narrative review articles’, Research Integrity and Peer Review, vol. 4, article 5, https://doi.org/10.1186/s41073-019-0064-8

Bandura, A 1977, ‘Self-efficacy: toward a unifying theory of behavioral change’, Psychological Review, vol. 84, no. 2, pp. 191–215, https://doi.org/10.1037/0033-295X.84.2.191

Bandura, A 1997, Self-efficacy: the exercise of control, W.H. Freeman, New York.

Barling, J, Loughlin, C & Kelloway, EK 2002, ‘Development and test of a model linking safety-specific transformational leadership and occupational safety’, Journal of Applied Psychology, vol. 87, no. 3, pp. 488–496, https://doi.org/10.1037/0021-9010.87.3.488

Beus, JM, Dhanani, LY & McCord, MA 2015, ‘A meta-analysis of personality and workplace safety: addressing unanswered questions’, Journal of Applied Psychology, vol. 100, no. 2, pp. 481–498, https://doi.org/10.1037/a0037916

Blais, A-R & Weber, EU 2006, ‘A domain-specific risk-taking (DOSPERT) scale for adult populations’, Judgment and Decision Making, vol. 1, no. 1, pp. 33–47, https://doi.org/10.1017/S1930297500000334

Borys, D 2012, ‘The role of safe work method statements in the Australian construction industry’, Safety Science, vol. 50, no. 2, pp. 210–220, https://doi.org/10.1016/j.ssci.2011.08.010

Byrnes, JP, Miller, DC & Schafer, WD 1999, ‘Gender differences in risk taking: a meta-analysis’, Psychological Bulletin, vol. 125, no. 3, pp. 367–383, https://doi.org/10.1037/0033-2909.125.3.367

Casey, BJ, Getz, S & Galván, A 2008, ‘The adolescent brain’, Developmental Review, vol. 28, no. 1, pp. 62–77, https://doi.org/10.1016/j.dr.2007.08.003

Christian, MS, Bradley, JC, Wallace, JC & Burke, MJ 2009, ‘Workplace safety: a meta-analysis of the roles of person and situation factors’, Journal of Applied Psychology, vol. 94, no. 5, pp. 1103–1127, https://ingaa.org/wp-content/uploads/2018/11/35441.pdf

Clarke, S 2006, ‘The relationship between safety climate and safety performance: a meta-analytic review’, Journal of Occupational Health Psychology, vol. 11, no. 4, pp. 315–327, https://doi.org/10.1037/1076-8998.11.4.315

Clarke, S 2013, ‘Safety leadership: a meta-analytic review of transformational and transactional leadership styles as antecedents of safety behaviours’, Journal of Occupational and Organizational Psychology, vol. 86, no. 1, pp. 22–49, https://doi.org/10.1111/j.2044-8325.2012.02064.x

Clarke, S & Robertson, IT 2005, ‘A meta-analytic review of the Big Five personality factors and accident involvement in occupational and non-occupational settings’, Journal of Occupational and Organizational Psychology, vol. 78, no. 3, pp. 355–376, https://doi.org/10.1348/096317905X26183

Coates, JM & Herbert, J 2008, ‘Endogenous steroids and financial risk taking on a London trading floor’, Proceedings of the National Academy of Sciences, vol. 105, no. 16, pp. 6167–6172, https://pmc.ncbi.nlm.nih.gov/articles/PMC2396566/

Committee of Sponsoring Organizations of the Treadway Commission 2017, Enterprise risk management: integrating with strategy and performance, COSO, https://www.coso.org/guidance-erm

Daly, M & Wilson, M 1988, Homicide, Aldine de Gruyter, New York.

DeJoy, DM 2005, ‘Behavior change versus culture change: divergent approaches to managing workplace safety’, Safety Science, vol. 43, no. 2, pp. 105–129, https://doi.org/10.1016/j.ssci.2005.02.001

Dekker, S 2011, Drift into failure: from hunting broken components to understanding complex systems, Ashgate, Farnham.

Douglas, M & Wildavsky, A 1982, Risk and culture: an essay on the selection of technological and environmental dangers, University of California Press, Berkeley.

Efendić, E, Chandrashekar, SP, Lee, CS, Yeung, LY, Kim, MJ, Lee, CY & Feldman, G 2022, ‘Risky therefore not beneficial: replication and extension of Finucane et al.’s (2000) affect heuristic experiment’, Social Psychological and Personality Science, vol. 13, no. 7, pp. 1173–1184, https://doi.org/10.1177/19485506211056761

Evans, JStBT 2008, ‘Dual-processing accounts of reasoning, judgment, and social cognition’, Annual Review of Psychology, vol. 59, pp. 255–278, https://doi.org/10.1146/annurev.psych.59.103006.093625

Finucane, ML, Alhakami, A, Slovic, P & Johnson, SM 2000, ‘The affect heuristic in judgments of risks and benefits’, Journal of Behavioral Decision Making, vol. 13, no. 1, pp. 1–17, http://stanford.edu/~knutson/jdm/finucane00.pdf

Fischhoff, B, Slovic, P, Lichtenstein, S, Read, S & Combs, B 1978, ‘How safe is safe enough? A psychometric study of attitudes towards technological risks and benefits’, Policy Sciences, vol. 9, no. 2, pp. 127–152, https://doi.org/10.1007/BF00143739

Gal, D & Rucker, DD 2018, ‘The loss of loss aversion: will it loom larger than its gain?’, Journal of Consumer Psychology, vol. 28, no. 3, pp. 497–516, https://doi.org/10.1002/jcpy.1047

Gardner, M & Steinberg, L 2005, ‘Peer influence on risk taking, risk preference, and risky decision making in adolescence and adulthood: an experimental study’, Developmental Psychology, vol. 41, no. 4, pp. 625–635, https://doi.org/10.1037/0012-1649.41.4.625

Geller, ES 2001, The psychology of safety handbook, Lewis Publishers, Boca Raton.

Gigerenzer, G 2014, Risk savvy: how to make good decisions, Viking, New York.

Griffin, MA & Neal, A 2000, ‘Perceptions of safety at work: a framework for linking safety climate to safety performance, knowledge, and motivation’, Journal of Occupational Health Psychology, vol. 5, no. 3, pp. 347–358, https://doi.org/10.1037/1076-8998.5.3.347

Guldenmund, FW 2000, ‘The nature of safety culture: a review of theory and research’, Safety Science, vol. 34, no. 1–3, pp. 215–257, https://doi.org/10.1016/S0925-7535(00)00014-X

Hale, A & Borys, D 2013a, ‘Working to rule, or working safely? Part 1: a state of the art review’, Safety Science, vol. 55, pp. 207–221, https://doi.org/10.1016/j.ssci.2012.05.011

Hale, A & Borys, D 2013b, ‘Working to rule, or working safely? Part 2: the management of safety rules and procedures’, Safety Science, vol. 55, pp. 222–231, https://doi.org/10.1016/j.ssci.2012.05.013

Hedlund, J 2000, ‘Risky business: safety regulations, risk compensation, and individual behavior’, Injury Prevention, vol. 6, no. 2, pp. 82–90, https://doi.org/10.1136/ip.6.2.82

Hillson, D & Murray-Webster, R 2007, Understanding and managing risk attitude, 2nd edn, Gower, Aldershot.

Hollnagel, E 2009, The ETTO principle: efficiency-thoroughness trade-off: why things that go right sometimes go wrong, Ashgate, Farnham.

Hollnagel, E 2014, Safety-I and Safety-II: the past and future of safety management, Ashgate, Farnham.

Hopkins, A 2000, Lessons from Longford: the Esso gas plant explosion, CCH Australia, Sydney.

Hopkins, A 2005, Safety, culture and risk: the organisational causes of disasters, CCH Australia, Sydney.

Hopkins, A 2008, Failure to learn: the BP Texas City refinery disaster, CCH Australia, Sydney.

Hopkins, A 2012, Disastrous decisions: the human and organisational causes of the Gulf of Mexico blowout, CCH Australia, Sydney.

Hoyle, RH, Stephenson, MT, Palmgreen, P, Lorch, EP & Donohew, RL 2002, ‘Reliability and validity of a brief measure of sensation seeking’, Personality and Individual Differences, vol. 32, no. 3, pp. 401–414, https://doi.org/10.1016/S0191-8869(01)00032-0

International Organization for Standardization 2018, ISO 31000:2018 Risk management: guidelines, ISO, Geneva, https://www.iso.org/standard/65694.html

International Organization for Standardization 2022, ISO 31073:2022 Risk management: vocabulary, ISO, Geneva, https://www.iso.org/standard/79637.html

Kahneman, D 2011, Thinking, fast and slow, Farrar, Straus and Giroux, New York.

Kahneman, D & Tversky, A 1979, ‘Prospect theory: an analysis of decision under risk’, Econometrica, vol. 47, no. 2, pp. 263–291, https://doi.org/10.2307/1914185

Kasperson, RE, Renn, O, Slovic, P, Brown, HS, Emel, J, Goble, R, Kasperson, JX & Ratick, S 1988, ‘The social amplification of risk: a conceptual framework’, Risk Analysis, vol. 8, no. 2, pp. 177–187, https://doi.org/10.1111/j.1539-6924.1988.tb01168.x

Komaki, J, Barwick, KD & Scott, LR 1978, ‘A behavioral approach to occupational safety: pinpointing and reinforcing safe performance in a food manufacturing plant’, Journal of Applied Psychology, vol. 63, no. 4, pp. 434–445, https://doi.org/10.1037/0021-9010.63.4.434

Lawton, R 1998, ‘Not working to rule: understanding procedural violations at work’, Safety Science, vol. 28, no. 2, pp. 77–95, https://doi.org/10.1016/S0925-7535(97)00073-8

Loewenstein, GF, Weber, EU, Hsee, CK & Welch, N 2001, ‘Risk as feelings’, Psychological Bulletin, vol. 127, no. 2, pp. 267–286, https://doi.org/10.1037/0033-2909.127.2.267

Master Builders Australia 2024, The building and construction industry workforce, Master Builders Australia, Canberra, https://www.mba.org.au/wp-content/uploads/2024/07/2024_July_State-of-building-and-construction-industry-workforce.pdf

Meertens, RM & Lion, R 2008, ‘Measuring an individual’s tendency to take risks: the risk propensity scale’, Journal of Applied Social Psychology, vol. 38, no. 6, pp. 1506–1520, https://doi.org/10.1111/j.1559-1816.2008.00357.x

Moscovici, S & Zavalloni, M 1969, ‘The group as a polarizer of attitudes’, Journal of Personality and Social Psychology, vol. 12, no. 2, pp. 125–135, https://doi.org/10.1037/h0027568

Mrkva, K, Johnson, EJ, Gächter, S & Herrmann, A 2020, ‘Moderating loss aversion: loss aversion has moderators, but reports of its death are greatly exaggerated’, Journal of Consumer Psychology, vol. 30, no. 3, pp. 407–428, https://nottingham-repository.worktribe.com/output/3681587

Nahrgang, JD, Morgeson, FP & Hofmann, DA 2011, ‘Safety at work: a meta-analytic investigation of the link between job demands, job resources, burnout, engagement, and safety outcomes’, Journal of Applied Psychology, vol. 96, no. 1, pp. 71–94, http://www.morgeson.com/downloads/nahrgang_morgeson_hofmann_2011.pdf

Neal, A & Griffin, MA 2006, ‘A study of the lagged relationships among safety climate, safety motivation, safety behavior, and accidents at the individual and group levels’, Journal of Applied Psychology, vol. 91, no. 4, pp. 946–953, https://doi.org/10.1037/0021-9010.91.4.946

Nicholson, N, Soane, E, Fenton-O’Creevy, M & Willman, P 2005, ‘Personality and domain-specific risk taking’, Journal of Risk Research, vol. 8, no. 2, pp. 157–176, https://doi.org/10.1080/1366987032000123856

Occupational Health and Safety Act 2004 (Vic), s 21, viewed via the Victorian legislation register, https://www.legislation.vic.gov.au/in-force/acts/occupational-health-and-safety-act-2004.

Occupational Health and Safety Regulations 2017 (Vic), r 325, viewed via the Victorian legislation register, https://www.legislation.vic.gov.au/in-force/statutory-rules/occupational-health-and-safety-regulations-2017/022.

Parker, D, Lawrie, M & Hudson, P 2006, ‘A framework for understanding the development of organisational safety culture’, Safety Science, vol. 44, no. 6, pp. 551–562, https://doi.org/10.1016/j.ssci.2005.10.004

Patton, JH, Stanford, MS & Barratt, ES 1995, ‘Factor structure of the Barratt Impulsiveness Scale’, Journal of Clinical Psychology, vol. 51, no. 6, pp. 768–774, https://doi.org/10.1002/1097-4679(199511)51:6%3C768::AID-JCLP2270510607%3E3.0.CO;2-1

Peltzman, S 1975, ‘The effects of automobile safety regulation’, Journal of Political Economy, vol. 83, no. 4, pp. 677–725, https://doi.org/10.1086/260352

Power, M 2004, The risk management of everything: rethinking the politics of uncertainty, Demos, London.

Power, M 2007, Organized uncertainty: designing a world of risk management, Oxford University Press, Oxford.

Provan, DJ, Woods, DD, Dekker, SWA & Rae, AJ 2020, ‘Safety II professionals: how resilience engineering can transform safety practice’, Reliability Engineering & System Safety, vol. 195, https://www.sciencedirect.com/science/article/pii/S0951832018309864

Quinlan, M, Mayhew, C & Bohle, P 2001, ‘The global expansion of precarious employment, work disorganization, and consequences for occupational health: a review of recent research’, International Journal of Health Services, vol. 31, no. 2, pp. 335–414, https://doi.org/10.2190/607H-TTV0-QCN6-YLT4

Rasmussen, J 1997, ‘Risk management in a dynamic society: a modelling problem’, Safety Science, vol. 27, no. 2–3, pp. 183–213, https://backend.orbit.dtu.dk/ws/files/158016663/SAFESCI.pdf

Reason, J 1990, Human error, Cambridge University Press, Cambridge.

Reason, J, Manstead, A, Stradling, S, Baxter, J & Campbell, K 1990, ‘Errors and violations on the roads: a real distinction?’, Ergonomics, vol. 33, no. 10–11, pp. 1315–1332, https://doi.org/10.1080/00140139008925335

Reason, J, Parker, D & Lawton, R 1998, ‘Organizational controls and safety: the varieties of rule-related behaviour’, Journal of Occupational and Organizational Psychology, vol. 71, no. 4, pp. 289–304, https://doi.org/10.1111/j.2044-8325.1998.tb00678.x

Renn, O 2008, Risk governance: coping with uncertainty in a complex world, Earthscan, London.

Romer, D, Reyna, VF & Satterthwaite, TD 2017, ‘Beyond stereotypes of adolescent risk taking: placing the adolescent brain in developmental context’, Developmental Cognitive Neuroscience, vol. 27, pp. 19–34, https://repository.upenn.edu/asc_papers/518

Rotter, JB 1966, ‘Generalized expectancies for internal versus external control of reinforcement’, Psychological Monographs: General and Applied, vol. 80, no. 1, pp. 1–28, https://doi.org/10.1037/h0092976

Safe Work Australia 2013, How to determine what is reasonably practicable to meet a health and safety duty, Safe Work Australia, Canberra, https://www.safeworkaustralia.gov.au/doc/how-determine-what-reasonably-practicable-meet-health-and-safety-duty

Safe Work Australia 2025, Key work health and safety statistics, Australia 2025, Safe Work Australia, Canberra, https://data.safeworkaustralia.gov.au/sites/default/files/2025-10/Key_Work_Health_and_Safety_Statistics_Australia_2025.pdf

Shulman, EP, Smith, AR, Silva, K, Icenogle, G, Duell, N, Chein, J & Steinberg, L 2016, ‘The dual systems model: review, reappraisal, and reaffirmation’, Developmental Cognitive Neuroscience, vol. 17, pp. 103–117, https://www.sciencedirect.com/science/article/pii/S1878929315001292

Simon, HA 1955, ‘A behavioral model of rational choice’, Quarterly Journal of Economics, vol. 69, no. 1, pp. 99–118, https://doi.org/10.2307/1884852

Sjöberg, L 2000, ‘Factors in risk perception’, Risk Analysis, vol. 20, no. 1, pp. 1–11, https://doi.org/10.1111/0272-4332.00001

Slovic, P 1987, ‘Perception of risk’, Science, vol. 236, no. 4799, pp. 280–285, https://doi.org/10.1126/science.3563507

Slovic, P, Finucane, ML, Peters, E & MacGregor, DG 2004, ‘Risk as analysis and risk as feelings: some thoughts about affect, reason, risk, and rationality’, Risk Analysis, vol. 24, no. 2, pp. 311–322, https://doi.org/10.1111/j.0272-4332.2004.00433.x

Snook, SA 2000, Friendly fire: the accidental shootdown of U.S. Black Hawks over northern Iraq, Princeton University Press, Princeton.

Stanovich, KE & West, RF 2000, ‘Individual differences in reasoning: implications for the rationality debate?’, Behavioral and Brain Sciences, vol. 23, no. 5, pp. 645–665, https://doi.org/10.1017/S0140525X00003435

Steinberg, L 2008, ‘A social neuroscience perspective on adolescent risk-taking’, Developmental Review, vol. 28, no. 1, pp. 78–106, https://doi.org/10.1016/j.dr.2007.08.002

Stoner, JAF 1961, A comparison of individual and group decisions involving risk, unpublished master’s thesis, Massachusetts Institute of Technology, Cambridge, MA.

Tversky, A & Kahneman, D 1974, ‘Judgment under uncertainty: heuristics and biases’, Science, vol. 185, no. 4157, pp. 1124–1131, https://doi.org/10.1126/science.185.4157.1124

Tversky, A & Kahneman, D 1992, ‘Advances in prospect theory: cumulative representation of uncertainty’, Journal of Risk and Uncertainty, vol. 5, no. 4, pp. 297–323, https://www.psych.fullerton.edu/mbirnbaum/psych466/articles/Tversky_Kahneman_JRU_92.pdf

Vaughan, D 1996, The Challenger launch decision: risky technology, culture, and deviance at NASA, University of Chicago Press, Chicago.

Vaughan, D 1999, ‘The dark side of organizations: mistake, misconduct, and disaster’, Annual Review of Sociology, vol. 25, pp. 271–305, https://doi.org/10.1146/annurev.soc.25.1.271

Weber, EU, Blais, A-R & Betz, NE 2002, ‘A domain-specific risk-attitude scale: measuring risk perceptions and risk behaviors’, Journal of Behavioral Decision Making, vol. 15, no. 4, pp. 263–290, https://doi.org/10.1002/bdm.414

Westrum, R 2004, ‘A typology of organisational cultures’, Quality and Safety in Health Care, vol. 13, suppl. 2, pp. ii22–ii27, https://doi.org/10.1136/qshc.2003.009522

Whiteside, SP & Lynam, DR 2001, ‘The five factor model and impulsivity: using a structural model of personality to understand impulsivity’, Personality and Individual Differences, vol. 30, no. 4, pp. 669–689, https://doi.org/10.1016/S0191-8869(00)00064-7

Wilde, GJS 1982, ‘The theory of risk homeostasis: implications for safety and health’, Risk Analysis, vol. 2, no. 4, pp. 209–225, https://doi.org/10.1111/j.1539-6924.1982.tb01384.x

Wilson, M & Daly, M 1985, ‘Competitiveness, risk taking, and violence: the young male syndrome’, Ethology and Sociobiology, vol. 6, no. 1, pp. 59–73, https://doi.org/10.1016/0162-3095(85)90041-X

Work Health and Safety Act 2011 (Qld), ss 18, 19, 20, viewed via the Queensland legislation register, https://www.legislation.qld.gov.au/view/xml/inforce/current/act-2011-018.

Work Health and Safety Regulation 2011 (Qld), r 36, viewed via the Queensland legislation register, https://www.legislation.qld.gov.au/view/xml/inforce/current/sl-2011-0240.

Zahavi, A 1975, ‘Mate selection: a selection for a handicap’, Journal of Theoretical Biology, vol. 53, no. 1, pp. 205–214, https://doi.org/10.1016/0022-5193(75)90111-3

Zohar, D 1980, ‘Safety climate in industrial organizations: theoretical and applied implications’, Journal of Applied Psychology, vol. 65, no. 1, pp. 96–102, https://doi.org/10.1037/0021-9010.65.1.96

Zohar, D 2000, ‘A group-level model of safety climate: testing the effect of group climate on microaccidents in manufacturing jobs’, Journal of Applied Psychology, vol. 85, no. 4, pp. 587–596, https://doi.org/10.1037/0021-9010.85.4.587

Zohar, D 2010, ‘Thirty years of safety climate research: reflections and future directions’, Accident Analysis & Prevention, vol. 42, no. 5, pp. 1517–1522, https://doi.org/10.1016/j.aap.2009.12.019

Zuckerman, M 1994, Behavioral expressions and biosocial bases of sensation seeking, Cambridge University Press, New York.

You might like to read other articles by this author. What is Safe? examines the use of the word in organisational context