Causal and Corrective Organisational Culture: A Systematic Review of Case Studies of Institutional Failure

There are pockets of consensus regarding how to define organisational culture. It is generally accepted that culture provides the rather stable and shared system of values, beliefs, and assumptions which provides approved modes of thought and behaviour, is resistant to change, and maintained through social interaction (Schall 1983; Schein 1984; Schneider et al. 2013). Schein (1984) further suggests culture is stratified by different levels of meaning, where the deepest level comprises the underlying and pervasive assumptions which organisational members tacitly accept, the intermediary level comprises what they espouse to believe, and the highest level consists of visible or audible patterns of behaviour and artefacts which are a manifestation of the other levels. Studies of culture divide according to whether they orient ethnographically to organisations “as cultures,” or measure culture through surveys and questionnaires as a variable or “something an organisation has” (Smircich 1983, p. 347, original emphasis).

Various attributes and types of culture have been associated with financial performance (e.g. Denison 1984; O’Reilly et al. 2014). Barney (1986) suggests culture can afford a sustained competitive advantage if characterised by uncommon qualities which cannot be imitated by other organisations. Similarity in survey responses, as an indicator of cultural strength, has also been linked to performance (e.g. Denison 1990; Gordon and DiTomaso 1992). However, as Reason (1998) highlights, the very same processes of internal integration and external adaptation which maintain group cohesion and thus comprise the core function of a culture (Schein 2010) can threaten an organisation’s survival when applied to goals which undermine good practise. Namely, through normalising maladaptive behaviour, “cultures create problems as well as solving them” (Kroeber and Kluckhohn 1952, p. 57). This duality corresponds to the sub-field of sociology which draws on Merton (1936, 1940, 1968) and Durkheim (1895/1966) to investigate how the same processes which produce positive organisational outcomes, are also responsible for the ‘dark side’ that generates mistakes, misconduct, and disaster (Vaughan 1999).

Institutional failure is “a physical, cultural, and emotional event incurring social loss, often possessing a dramatic quality that damages the fabric of social life” (Vaughan 1999, p. 292). Concretely, it is typically used to refer to large-scale avoidable failures, for example accidents (e.g. Chernobyl, Deepwater Horizon) or scandals (e.g. Enron, Barings Bank), that have consequences for those within an organisation (e.g. employees), stakeholders (e.g. passengers, investors, the public), the environment (e.g. pollution), and integrity of an institution itself (e.g. collapse, or huge reputational damage). Within the diverse conceptual models that are used to explain failure, for example by Turner (1978; Turner and Pidgeon 1997), Reason (1990, 2016), Perrow (1984, 1999), and Rasmussen (1997), organisational culture is often a key element.

Turner (1978) was first to describe failure as a socio-technical phenomenon, rather than an event which is divine, coincidental, or purely technical (Turner and Pidgeon 1997). He conducted a systematic qualitative analysis of 84 British accident and disaster reports published between 1965 and 1975, developing a six-stage developmental sequence model of failure (‘man-made disaster’) as being preceded by several preconditions which develop during a ‘disaster incubation period.’ The incubation period involves the slow ‘accumulation’ of events which deviate from the culture’s beliefs and norms regarding hazards. This accumulation can continue for many years and is enabled by people’s incorrect assumptions about hazards, problems in information-handling, rigidities of perception, and inappropriate or outdated formal procedures. The incubation period ends when a ‘precipitating incident’ such as an explosion, fire, or plunge in share prices, exposes the actual state of affairs. Culture is at the centre of Turner’s model which equates failure sociologically to a cultural collapse (Pidgeon and O’Leary 2000). Yet, culture operates at a meta-level in the divergence between what people believe is an accurate perception of affairs, and what is actually true. This highlights how disaster occurs despite people believing they are taking the necessary precautions against failure, but does not identify common ways in which it is precipitated by specific cultural problems.

Reason’s (1990, 2016) model of accident causation also considers the role of organisational culture. Reason posits an organisation’s layers of defence are somewhat akin to layers of Swiss cheese: each has gaps representing weaknesses, through which an accident ‘trajectory’ can pass if gaps momentarily align. Defence weaknesses are constantly moving, making their alignment—and thus failure—a rare occurrence (Reason 1998). Reason makes the useful distinction between the errors or violations at the ‘sharp-end’ of operations which ‘trigger’ failure (‘active failures’)—the final slice in the Swiss cheese model—and the latent error- and violation-producing conditions. Latent conditions include for example an emphasis on cost-cutting (e.g. understaffing), aspects of organisational structure and how business is conducted, inadequate hardware in terms of tools and equipment, poor system design, and procedures which are unclear or not applicable. They persist undetected and are a product of culture, as well as management decisions, and organisational processes. However, only culture is ubiquitous enough to influence all aspects of defence. Culture “can not only open gaps and weaknesses but also—and most importantly—it can allow them to remain uncorrected” (Reason 1998, p. 297).

Culture is less focal in Perrow’s (1984, 1999) theory of normal accidents which suggests failure is a normal and unpreventable part of organisational systems working with high-risk technology (e.g. nuclear power plants, air traffic control) because they are characterised by ‘tight coupling’ (e.g. little scope for slack, delays, and alternative procedures) and high ‘interactive complexity’ of system components. Perrow gives external forces of capitalism a larger role in failure than culture. He notes,

If culture plays a role, as many argue it does (…), it is not the most important one, and while efforts to change the culture to one that favors high reliability operations are certainly of high priority, restricting the catastrophic potential of our enterprises is of higher priority. (…) Rather than look to national cultures, or even to cultures of companies and the workplace, we might look at plain old free-market capitalism (Perrow 1999, p. 416, original emphasis).

Perrow defines culture in positive terms as a source of reliability, and thus does not associate culture with the negative potential of adverse outcomes. Instead, this potential is attributed to external production pressures of the outside economic system. However, the ways in which an organisation manages the two opposing goals of efficiency and safety can shed light on what is valued and tolerated within an organisation (i.e. culture), and would explain why not all those operating in the same economic conditions experience failure.

Rasmussen (1997) includes culture as a possible preventative mechanism. According to Rasmussen (1997), organisations are constantly under pressure to maintain an acceptable workload, safe performance, and avoid economic failure. These are depicted as boundaries, where the organisation is at the centre, managing the tensions between them. The organisation moves toward the ‘functionally acceptable' (i.e. safe) performance boundary when it reduces employees’ workload or increases productivity. Errors and accidents occur if the organisation moves so far that it crosses the boundary of functionally acceptable performance. Rasmussen proposes that informational campaigns for safety culture can counter the pressures of efficiency and thus improve control of performance. Safety culture is defined in terms of people’s knowledge of the boundary of functionally acceptable performance. Like Perrow, this highlights the positive role of culture in fostering an awareness of risk and danger, but does not highlight the negative contributing aspects of culture.

In conclusion, different and seminal models of organisational failure view failure as something which emerges gradually and sequentially over several contributing factors (Reason 1990; Turner 1978). Organisational culture permeates these models through providing an explanatory framework for understanding the drivers of behaviour within an organisation (e.g. the values that underlie, and are expressed, through cost-cutting, incentivisation, system design, procedures), and explaining how norms and values towards risk (e.g. normalisation and tolerance) determine how managers and employees identify and respond to hazards. Subsequent research on the role of organisational culture in institutional failures has tended to focus on two domains: safety culture and ethical culture.

Safety culture and ethical culture are the main cultural dimensions applied to investigate the relationship between culture and institutional failures (Cooper 2000; Guldenmund 2000; Kaptein 2008; Treviño and Weaver 2003). Although they focus on different sets of values (e.g. the importance of safety, adhering to ethical standards), behaviours (e.g. risk-taking, dishonesty), and outcomes (e.g. accidents, scandals), safety culture and ethical culture have many parallels in how they are used to explain organisational failures. For instance, both stress the importance of senior leadership in setting standards, supervisors in guiding behaviour, organisational and group norms in determining what practices are acceptable, giving employees the knowledge and skills to behave effectively, and ensuring employees can speak-up (and are listened to) when they raise concerns (Ardichvili and Jondle 2009; Guldenmund 2000; Kaptein 2011; Neal and Griffin 2002; Reader and O’Connor 2014; Zohar 2010). However, both models diverge in their origins, and the variables they use to explain organisational failures.

Interest in safety culture came from a shift in focus from models of causation to how crisis and risk management might be improved to provide institutional resilience (Pidgeon and O’Leary 2000). Safety culture relates to the norms and practises surrounding health and safety within an organisation (Cooper 2000; Guldenmund 2000), and is highly related to safety climate (perceptions on the priority of safety) (Zohar 2010). Pidgeon and O’Leary (2017) suggest a ‘good’ safety culture is characterised by senior management’s commitment to safety, a shared concern for hazards and how they impact people, realistic norms and procedures for managing risk, and continual processes of reflection and organisational learning. Safety culture gained traction in the 1980s to account for large-scale failures such as the Chernobyl nuclear disaster (Pidgeon 1998) and Piper Alpha oil rig explosion (1993) where shared patterns of belief and behaviour were found to have played a significant role in the disasters. In both cases, a prioritisation of other concerns (e.g. productivity) by senior managers led to operational decisions that weakened safety (e.g. on the use of resources, reduced safety inspections, pushing safety capabilities), the normalisation of unsafe practices (e.g. unsupervised staff undertaking maintenance routines), and a lack of preparedness for managing safety emergencies. By focussing accident investigations on the system and cultural context of organisations, safety culture theory departed from earlier research which had attributed accidents to fallible mental processes (e.g. forgetfulness, negligence) in individuals working directly with the system (Reason 2000). Rather, safety is understood as part of an organisation’s value-system, with the consideration of safety in everyday practices being a product of, and revealing, these values (Guldenmund 2000).

Interest in the domain of ethical culture has come about due to organisational failures that are considered a consequence of unethical conduct (e.g. scandals of Enron, LIBOR, and Odebrecht). Similar to safety culture, the concept of ethical culture emerges from the rationale that unethical acts within an organisation are likely to reflect values within an organisation for ethical conduct, rather than individual failings. This diverges from the perspective that unethical behaviour is determined by individual factors such as a person’s sensitivity to moral issues (Rest 1979), level of moral judgement (Kohlberg 1969), and guilt proneness (e.g. Cohen et al. 2012). Within the ethical culture framework, ethical behaviour is conceptualised as determined by immediate job pressures, institutional values and norms on the importance of ethics (e.g. for indicating the appropriateness of behaviours), and the embedding of these values into formal systems (e.g. rules and polices) (Treviño 1986; Treviño et al. 2014). As with safety culture, ethical culture is conceptualised as a subset of organisational culture, with specific domains of activity—for instance on transparency or the sanctioning of unethical behaviour—constituting an ethical culture (Kaptein 2008; Kish-Gephart et al. 2010). Furthermore, ethical culture is influenced by the values and behaviours of leaders (see Ardichvili and Jondle 2009), and associated through various dimensions with reported (un)ethical behaviour and intentions (e.g. Kaptein 2011; Sweeney et al. 2010; Zaal et al. 2019). Several contributing factors to ethical culture have been identified, including an organisational commitment to employees (Fernández and Camacho 2016), the presence of ethics programmes such as a dedicated ethics department or committee (Martineau et al. 2017), the role-modelling of ethical practises by management (Kaptein 2011), and reward systems that reinforce ethical behaviour (Treviño et al. 1998). Ethical climate, understood as the perceptions of organisational values for ethics (Victor and Cullen 1988), has been distinguished from ethical culture, which relates more to the systems of control that shape ethical behaviour (Kaptein 2011; Treviño et al. 1998). As the dimensions of ethical culture and ethical climate are highly related (Treviño et al. 1998), we, like others (e.g. Ardichvili and Jondle 2009), regard ethical climate as a sub-category of ethical culture.

The fields of both safety culture and ethical culture have arisen to explain organisational failures, and to investigate this, researchers in both fields have relied on psychometrically validated surveys (Guldenmund 2007; Kaptein 2008). These are used to measure employee perceptions of the culture (e.g. management commitment to safety, reporting safety incidents, clarity of expected conduct, reporting ethical concerns), with responses being associated at an individual, unit, or organisational level with adverse outcomes. For instance, research shows safety culture to be associated with safety behaviours, reporting, and lost-time injuries (Beus et al. 2016; Christian et al. 2009; Petitta et al. 2017), and ethical culture to be associated with ethical choices and reports of unethical behaviours (e.g. intentions to report misconduct) (Schaubroeck et al. 2012; Kish-Gephart et al. 2010; Kaptein 2011). Associations between safety culture and ethical culture and larger-scale organisational failures (e.g. corruption, process safety failures) are absent due to their rarity (e.g. in comparison to individual reports on behaviour), unpredictability (e.g. in identifying and accessing a failing organisation), and the challenges of expecting employees to recognise and report on sensitive topics like safety and ethics (e.g. Antonsen 2009a; Arnold and Feldman 1981; Fischer and Fick 1993). Furthermore, whilst survey methods have provided valuable insight on ‘what’ cultural dimensions are associated with adverse outcomes, they have not necessarily shown ‘how’ various aspects of culture interact to create the conditions for failure. This is important because failure is defined by the nature of its sequential development through several events over time (Reason 1990; Turner 1978).

In summary, researchers have provided and validated conceptual models for measuring safety culture and ethical culture, and these models are the most widely used to understand the cultural conditions under which institutional failures occur. However, for reasons of methodology and data availability, both concepts are limited in the extent to which their underlying components are demonstrated and understood (e.g. in terms of sequence within an event) to have a role in explaining large-scale and avoidable institutional failures (e.g. accidents, scandals). Indeed, it is not clear that safety culture and ethical culture are entirely distinct in explaining failure: for example, investigations of hospital failures in the UK (e.g. Mid Staffordshire hospital, Shrewsbury and Telford hospital) have revealed a combination of poor safety culture (e.g. staff not reporting on sub-standard care) and poor ethical culture (e.g. dismissing patient concerns, concealing poor care) to have led to unnecessary patient deaths (Francis 2013; Lintern 2019). There may be aspects of organisational culture important for explaining institutional failures that are generalisable and not unique to either safety culture or ethical culture (e.g. reporting on concerns), or factors that are not considered within either model. To better explain the role of organisational culture in institutional failures, and specifically the contribution of safety culture and ethical culture, we undertake a systematic review of articles investigating the relationship between culture and institutional failures.

In this study, we undertake an inductive content analysis of case studies investigating the role of organisational culture in institutional failures. Case study analyses draw on a multitude of secondary sources to understand causes of an institutional failure, including investigative reports, organisational documents, as well as first-person accounts, which only become available after a failure because of a need to establish what occurred (Cox and Flin 1998; Feagin et al. 1991). This mitigates the previously mentioned limitations with survey-based methods, and supports an approach which can consider the perspectives of all people and sub-cultures involved, enabling analysis of ‘how’ culture led to failure (Glendon and Stanton 2000; Tellis 1997). Given these methodological affordances, case studies may provide insight on the most commonly identified aspects of safety culture and ethical culture that underlie organisational failures, alongside identifying cultural factors not established within survey-based models of safety culture and ethical culture. Moreover, their inductive and retrospective nature offers the opportunity to determine whether failures of safety and ethics differ by the cultural factors they involve, or if the conceptual boundaries of safety culture and ethical culture become less distinct when considering failure. In this study, we systematically identify and analyse case studies of institutional failure in order to address four research questions. We define institutional failure as an event with multiple causes which had developed over time (Turner 1978; Turner and Pidgeon 1997).

First, we identify and extract the aspects of organisational culture reported as contributing to institutional failures. Our aim is to determine whether a common set of cultural factors can be established—from across the multiple and independent case studies—as leading to institutional failure. We think that establishing what cultural factors have been used to explain failure provides a logical starting point for our analysis because these cultural factors have not been systematically catalogued before and it will enable comparisons with existing models of safety culture, ethical culture, and failure.

Second, we examine how different cultural factors are used to explain institutional failure. We explore whether, as specified in various models of institutional failure (e.g. Reason 1990; Turner 1978), culture is used to both account for the organisational conditions that underlie failure (e.g. norms, values), and problems in responding to threats that endanger an organisation (e.g. a developing accident). This exploration of how culture is used to explain failure will shed light on the mechanisms by which culture contributes to failure. These mechanisms are beyond the scope of survey-based methods which measure cultural elements to the degree that they are present or absent (see Reason 2000).

Third, we consider the cultural factors identified in case studies of institutional failure in relation to the safety culture and ethical culture models. We investigate the extent to which the cultural factors identified by analyses of institutional failure map onto existing models of safety culture and ethical culture. We identify cultural factors not typically included with models of safety culture and ethical culture, and consider whether they indicate other aspects of organisational culture which may be important for explaining institutional failure. As a third step, this establishes the extent to which retrospective studies diverge from survey-based studies of safety culture and ethical culture.

Fourth, we examine whether the cultural factors identified as contributing to institutional failures vary according to failure-type (safety or ethics) and failure-outcome (e.g. loss of life, environmental damage). Here, we are interested in the extent to which models of safety culture and ethical culture are exclusive and explanatory of institutional failures, or whether a more generalisable model of institutional failure might be derived.

The full-text of included articles was retrieved and extraction was carried out according to the four research questions. To establish what cultural factors are cited as contributing to institutional failure, a content analysis of included articles was carried out at the sentence-level. “Content analysis is a research technique for making replicable and valid inferences from texts (or other meaningful matter) to the contexts of their use” (Krippendorff 2013, p. 24). This content analysis was inductive because no pre-existing model of culture or climate was used to identify the cultural factors cited by case studies. Avoiding pre-conceived cultural factors was considered important for methodological integrity given the untraditional nature of this review. Cultural factors were collapsed based on similarity over three rounds of consolidation into 23 categories.

To establish how case studies use cultural factors to explain failure, each factor was additionally coded for the presence of preceding cultural causes and subsequent cultural outcomes (i.e. which it elicited or supported). Cultural factors were not double-coded unless they indexed distinct values or behaviours, and where it was necessary to capture the unconnected causes or outcomes of a single cultural factor.

To establish whether the cultural factors identified in case studies correspond to the items and dimensions of models of safety culture and ethical culture, we examined two reviews of safety culture and climate scales (Flin et al. 2000; Guldenmund 2000), two widely-cited models of ethical culture (Kaptein 2008; Treviño et al. 1998), and conducted additional literature searches where a cultural factor was not captured by these resources (e.g. Hessels and Wurmser 2020; Singhapakdi et al. 1996).

Finally, each failure analysed was coded according to type (i.e. safety or ethics) and outcome (e.g. environmental damage). The cultural factors cited in relation to each failure type and outcome were examined for patterns to see if any cultural factors are exclusively involved in safety failures or ethical failures and whether some aspects of culture more frequently precede specific failure outcomes.

To establish whether case studies identify a common set of cultural factors, we systematically coded and collapsed all cultural factors invoked by case studies. Twenty-three cultural factors were identified as contributing to failure, indicating that failure arises from a generally common set of values and practises (see Table 1).

To establish whether usage of the concept of culture was congruent with the dynamic failure models of Reason (1998) and Turner (1978), we counted the number of factors cited and coded for possible sequential relationships between them. With a mean of 7.28 cultural factors cited per case study, usage of culture was in line with these failure models. Cultural factors were also frequently linked sequentially (87.84%, n = 65) and case studies typically cited more than one sequence of cultural factors (59.46%, n = 44). This reflects popular models of failure development which suggest the complex and dynamic nature of failure development (Reason 1990; Turner 1978) and is a divergence from survey-based studies of safety culture and ethical culture which commonly measure culture by its absence (see Reason 2000).

Sequences of cultural factors tended to consist of two cultural factors (77.03%, n = 57). Fewer case studies cited sequences of three (22.97%, n = 17), four (2.7%, n = 2), five (1.35%, n = 1), and six cultural factors (1.35%, n = 1). Importantly, sequential cultural factors were not necessarily consecutive. For example, in a case study of the Piper Alpha oil rig explosion, it was the combination of (a) inadequate regulation, (b) the primacy of production, and (c) disbelief that failure was possible, which together led to (d) design decisions that created (e) an unsafe physical work environment ( Paté-Cornell 1993).

Culture was also used to explain why an organisation failed to correct a problem before it was able to develop into failure (72.97% of case studies, n = 54). Issues of corrective culture manifested as ‘problems dealing with problems.’ This encompassed eight of the 23 cultural factors, including: bullying, listening, learning, problem acceptance, problem response, rhetoric, speaking-up, and speaking-up system. These may be organised into the two categories of voicing and hearing. Voicing factors refer to the failure of employees to voice concerns about institutional problems to people in authority. Hearing factors refer to the failure by management to act on information received about institutional problems. This distinction maps onto the two critical phases of an organisation’s ‘adaptive capacity’: first, disruptive events need to be identified and relayed appropriately, and second, information must be heeded with the proper deployment of resources (Burnard and Bhamra 2011). Problems of corrective culture are distinct from other cultural factors because they do not advance the development of a failure. Instead, they represent missed opportunities to address a problem and potentially avert failure.

Case studies cite an average of 4.99 causal cultural factors and 2.3 corrective cultural factors. As depicted in Fig. 3, the maximum number of causal and corrective factors cited by any case study is 17 and nine, respectively. Twenty case studies (27.03%) cite only causal factors, whilst five case studies (6.76%) describe failure arising from corrective factors alone. Indeed, 66.22 percent of case studies describe a combination of causal and corrective cultural factors. This indicates that failure commonly involves more aspects of causal culture, but that failure is typically also preceded by at least one missed opportunity to avert failure.

In summary, where studies of safety culture and ethical culture have prospectively measured culture through the perceptions of organisational members, case studies have retrospectively analysed how cultural factors interact in a sequential way to produce failure. Case studies thus reveal that, rather than exerting a top-down influence on behaviour, cultural factors divide according to whether they represent causes of failure or missed opportunities to avert failure. Indeed, a majority of case studies highlight that failures could have been prevented.

Twenty of the 23 cultural factors which contribute to failure are typically part of models of safety culture (Flin et al. 2000; Guldenmund 2000; Hessels and Wurmser 2020) or ethical culture (Kaptein 2008; Singhapakdi et al. 1996; Treviño et al. 1998) (see Table 2). This indicates that case study analyses and survey-based studies largely overlap in the cultural factors they identify. However, three cultural factors—listening, bullying, and homogeneity—were novel, and are important to identify.

Listening is not typically measured by survey-based studies of safety culture and ethical culture. This is noteworthy given that a failure to listen to signs or information about problems is mentioned by a third of case studies. A diverse literature outside of culture recognises the importance of listening in organisations (e.g. Gillespie and Reader 2016). Some draw on the Foucauldian concept of ‘parrhesia’ (true speech) to analyse whistleblowing and listening (e.g. Catlaw et al. 2014; Vandekerckhove and Langenberg 2012; Weiskopf and Tobias-Miersch 2016). For instance, Vandekerckhove and Langenberg (2012) suggest that parrhesia in the workplace may not be heard because information typically travels up a hierarchy, requiring that a series of people have the courage to speak-up to the next person above them. Others use the term ‘deaf effect’ to understand when organisations persist with projects despite reports of trouble (Cueller et al. 2006; Keil and Robey 1999). Jones and Kelly (2014) suggest that when whistle-blowers are listened to, it can create a mutually-reinforcing cycle that benefits organisational learning, whilst ‘organisational disregard’ can create a norm of silence, preventing people from coming forward about other issues (see also Mannion and Davies 2018). Burris (2012) focusses on the information communicated and finds that managers are more likely to agree with information that supports rather than challenges the organisation’s goals, and that this relationship is mediated by the perceived loyalty and threat of voicing employees. Perceptions of management listening have also been linked to psychological safety and creativity (Castro et al. 2018).

The case studies indicate that not listening (i.e. not heeding information) was the outcome of culture in two ways (see Table 3 for examples). First, not listening occurred because the information received (i.e. X is unsafe or unethical) conflicted with the receiver’s taken-for-granted assumptions about the world as a cultural member (i.e. X is safe, X is ethical, or failure cannot happen), resulting in a breakdown of intersubjectivity (e.g. Grice 1975). Choo (2008) refers to this misalignment between information and available cognitive frame as an ‘epistemic blindspot.’ In terms of excluding employees from decision-making, this misalignment could cause employee involvement to be regarded as extraneous. Second, not listening occurred because the information, whilst decoded ‘correctly’ in terms of intent and content, conflicted with competing cultural values and demands. These include a pressure to maintain performance levels (e.g. Broadribb 2015; Mason 2004) and having to manage with limited resources (e.g. MacLean et al. 2004). Competing values and demands may also account for those cases where managers excluded employees from decision-making. Involving employees may conflict with organisational goals if employees communicate inconvenient information or deplete valued time.

Bullying is the second cultural factor not typically a part of models of safety culture and ethical culture. Bullying deterred employees from speaking-up about organisational problems (e.g. Dimeo 2014; Patrick et al. 2018). Bullying had pragmatic qualities somewhat similar to gossip: it was a form of social control directed at someone perceived to have transgressed the rules, which seemed to reinforce the group’s normative boundaries of right and wrong (Gluckman 1963; see also Waddington 2016). Bullying is naturally more hostile than gossip, and was indelibly dysfunctional because it enabled unsafe or unethical practises to persist and remain acceptable to (most) group members. As such bullying can have adverse outcomes for organisations, as well as individuals (e.g. self-esteem, Randle 2003; job satisfaction, Quine 1999).

The final cultural factor not typically a part of models of safety culture and ethical culture is workforce homogeneity (i.e. the absence of diversity). In the case studies, homogeneity was predominantly fostered through recruitment and promotion practises which rewarded unethical behaviour with a sense of belonging (Crawford et al. 2017) or (continued) employment (Fallon and Cooper 2015; Sims and Brinkmann 2002, 2003). Homogeneity has been studied as a source of cultural strength, associated with more reliable performance in stable environments (Sørensen 2002), job satisfaction, organisational commitment, and less role stress (Barnes et al. 2006). Conversely, diversity research shows that demographic diversity amongst members of the board and management has a positive effect on financial performance (e.g. Erhardt et al. 2003), promotes cognitive diversity (see Horwitz 2005), and mitigates groupthink (Janis 1972). Indeed, strongly shared values and norms can be maladaptive (Syed 2019), but there is need to further conceptualise homogeneity as a cultural phenomenon that is conducive to failure.

In summary, whilst the cultural factors identified by case studies largely map onto models of safety culture and ethical culture, listening, bullying, and homogeneity as cultural phenomena need further research as they relate to adverse organisational events. However, the large overlap suggests failure and less severe events arise from a common set of cultural factors.

This review analysed 58 publications which use case study methods to investigate how organisational culture contributes to institutional failures. The findings advance existing models of failure in three ways. First, we found that case studies identify 23 contributing factors of culture. Several of these cultural factors have been established by the failure literature. For example, the most common cultural factor involved organisations prioritising productivity over ethics and safety. Such goal conflict is central to the theories of Perrow (1999), Reason (1990), and Rasmussen (1997). Where an underlying cause was provided, this goal conflict was owing, as Perrow (1999) suggests, to factors of capitalism, including privatisation (Dien et al. 2004) and permissive legislation (e.g. Behling et al. 2019). Yet, in a majority of cases it was not apparent why this goal conflict emerged. Perhaps, as Rasmussen (1997) suggests, organisations systematically move towards greater efficiency. Second, failure was preventable in a majority of case studies, but problems of corrective culture prevented the timely and effective resolution of issues. The distinction identified here, between those aspects of culture which cause failure and those which impede its aversion, adds to Rasmussen’s (1997) theory that organisations continually manage the boundaries of safety, economic failure, and workload. Causal culture pushes an organisation toward failure, creating disequilibrium. Corrective culture pulls an organisation back from failure, maintaining equilibrium. Third, most case studies described failure as arising from a sequential set of factors as described by the failure models of Reason (1990) and Turner (1978). Namely, cultural factors were not normally treated as being present or absent to varying degrees (Reason 2000), but as interacting with other cultural factors in dynamic ways. For example, inadequate regulation could lead to inadequate training, and inadequate training could lead to employees who feel unqualified to speak-up about problems and so remain silent. Dimensions of culture do not exist in isolation from one another: problems in one dimension can produce problems in other dimensions. Where Reason (1990) suggests ‘safety culture’ both opens and closes institutional weaknesses, we suggest the ‘opening’ of weaknesses is achieved by causal factors of culture, whilst the ‘closing’ of weaknesses is achieved by corrective factors of culture. Only a minority of case studies did not report both causal and corrective factors of culture, and thus we suggest that failure is characteristically preceded by both types of cultural factor. Namely, failure must involve at least one causal factor of culture which enables a problem to arise, and usually at least one corrective factor of culture which enables the problem to persist and develop into failure. This model needs to be substantiated with further research which explicitly identifies the causal and corrective cultural factors involved in failures.

Large-scale failures are difficult to track prospectively using traditional methods. As such we sought to explore the proposed role of safety culture and ethical culture in large-scale failures through a review of case study analyses. We found that the cultural factors identified in survey-based studies of safety culture and ethical culture overlap on 20 of the 23 cultural factors implicated in failures. This establishes the role of safety culture and ethical culture in causing failure, which has not been substantially established. Yet, it is important to recognise that safety culture and ethical culture differ in terms of their motivations and consequences. Safety culture generally relates to the management of physical risk towards employees (e.g. occupational injuries) and organisational processes (e.g. an oil rig). A combination of regulatory factors, and the insight that safety failures are always damaging to an organisation (e.g. in aviation), underlie the drive to have a culture that prioritises safety. Ethical culture is arguably more complex, because it relates to the upholding of both legal codes and what is regarded as morally acceptable within a society. Ethical failure is defined by subjective and shifting beliefs about right and wrong (both within and outside an organisation), and can be less apparent. From this perspective, the outcomes of safety culture are more stable and objective than ethical culture, although both concepts attempt to capture the conflict that can arise in organisations through having goals that simultaneously emphasise performance (e.g. productivity, profit) and avoiding risk-taking.

The three cultural factors involved in failure which are not normally included in existing models of safety culture and ethical culture were listening, homogeneity, and bullying. Of particular importance is listening which was disproportionately linked to a loss of human life. Problems of listening occurred in more than a third of case studies and were nearly as prevalent as well-established preconditions to failure (e.g. procedural deviations, inadequate training). The problems of listening are disconcerting, particularly considering that speaking-up about problems often comes from a passion for the organisation (Kenny et al. 2020). Whilst Turner (1978, Turner and Pidgeon 1997) describes how communication problems could prevent individuals from seeing an imminent failure, these relate to variables of the message (e.g. clarity) and individual cognition. Problems of listening are different because the message is received, often repeatedly, but not acted upon because of a culturally-prescribed worldview or competing cultural factors. Further research into problems of listening is needed to identify predictive variables and possible measures to counteract them. For instance, drawing on Foucault, Catlaw et al. (2014) suggest the ability to listen and enact change in response to ‘parrhesia’ (i.e. information about problems) requires “a certain kind of relationship with ourselves that is grounded in an attentive examination of how we live our lives” (p. 199; original emphasis). In contrast to listening and bullying, homogeneity as a factor in failure has been highlighted before (Syed 2019), but further research is needed to understand homogeneity as a cultural phenomenon. In the case studies analysed here, bullying maintained unsafe or unethical norms by deterring employees from speaking-up about problems, and punishing those who did. As such bullying is particularly detrimental to corrective culture because it fosters not only employee silence, but also fear. More research is needed to understand how bullying facilitates unsafe and unethical behaviour.

Problems of corrective culture can be conceived in terms of what Argyris (1976) called ‘single loop learning’ in contrast to ‘double loop learning’ that enables large-scale corrections. Organisations which engage in single loop learning favour control as their preferred ‘behavioural strategy': they are characteristically defensive, and inhibit free choice and information. Problems which conflict with the organisation’s accepted views of itself are not normally reacted to, irrespective of whether they are detected. It is clear how every factor of corrective culture that we identified evinces this behavioural strategy in response to problems.

Corrective culture is also complementary to Westrum’s (2004, 2014) notion of information flow. Westrum (2004) suggests different culture-types exhibit different patterns of response to warning signals, ranging from the suppression and isolation of the message-sender (pathological culture) to the launch of a public inquiry to identify the root causes of a problem (generative culture). This review adds to Westrum’s (2004, 2014) model by cataloguing the values, beliefs, and practises which stifle information flow and obstruct measures to self-correct and avert failure.

The findings of this review indicate that interventions to prevent failure, and evaluations of an organisation’s ability to prevent failure, should focus on corrective culture. An organisation will inevitably encounter problems of causal culture from time to time, but these will not develop into failure so long as the organisation is able to swiftly identify and appropriately deal with them.

Organisations can improve corrective culture through simulations, management, and introducing relevant cultural artefacts. Role-play simulations of bullying (e.g. amongst nurses, Ulrich et al. 2017), whistleblowing, and listening could be used to develop employees’ understanding of these corrective aspects of culture and readiness to deal with them. Managers could promote the value of corrective action by displaying an active interest in corrective culture, promoting the occurrence of corrections and rewarding employees who have detected and raised problems (e.g. with social recognition). Corrective culture could also be improved by addressing relevant cultural artefacts. For example, an organisation could evaluate the effectiveness of its whistleblowing hotline.

Organisations might also assess their corrective culture through cultural measurement and testing whether the corrective culture is working. Corrective culture could be measured by survey, with questions to assess employees’ confidence in raising issues, whether employees are listened to when they do raise concerns, and the presence of recurring problems in an organisation. Unobtrusive indicators of culture (‘UICs’, Reader et al. 2020) could be used to assess the extent to which corrective culture manifests in artefacts and behaviours. For example, one could assess whether incident reports are recorded in the meeting minutes of the executive team or board, and if the organisation has a telephone number for customer service. Online customer reviews could be analysed for content related to factors of corrective culture (e.g. listening). Responding only to selected people is not a good indication of corrective culture, and so handling of complaints from outsiders (e.g. customers), employees, and management could be compared to establish selective listening. The volume of complaints received by an organisation could also be assessed: as shown in the incident reporting literature, a lack of complaints may reflect a poor corrective culture rather than a lack of problems.