Exploratory analysis of the safety climate and safety behavior relationship
Introduction
Safety climate (Zohar, 1980) is a term used to describe shared employee perceptions of how safety management is being operationalized in the workplace, at a particular moment in time (Byrom & Corbridge, 1997). These descriptive perceptions provide an indication of the “(true) priority of safety” (Zohar, 2000) in an organization with regard to other priorities such as production or quality. Safety climate is considered to be a sub-component of the “safety culture” construct (International Atomic Energy Agency [IAEA], 1988) by some (Cooper, 2000, Glendon & Stanton, 2000, Neal et al., 2000, Silva et al., 2004, Zohar, 2000) or a reflection of actual safety culture by others (Arboleda et al., 2003, Cabrera & Isla, 1998, Cox & Flin, 1998, Fuller & Vassie, 2001, Guldenmund, 2000, Lee & Harrison, 2000, O'Toole, 2002, Vredenburgh, 2002, Williamson et al., 1997).
Over the last 25 years, safety climate research has taken four directions: (a) Designing psychometric measurement instruments and ascertaining their underlying factor structures (e.g., Brown & Holmes, 1986, Coyle et al., 1995, Dedobbeleer & Beland, 1991, Garavan & OBrien, 2001, Zohar, 1980); (b) Developing and testing theoretical models of safety climate to ascertain determinants of safety behavior and accidents (e.g., Cheyne et al., 1998, Neal et al., 2000, Prussia et al., 2003, Thompson et al., 1998); (c) Examining the relationship between safety climate perceptions and actual safety performance (Glendon & Litherland, 2000, Zohar, 2000); and (d) exploring the links between safety climate and organizational climate (Neal et al., 2000, Silva et al., 2004).
Factor analysis is a statistical technique used to identify a relatively small number of non-observable, underlying factors that characterize underlying “constructs” (e.g., management attitudes to safety). In safety climate research, these “factors” are used to represent relationships among many sets of inter-related perceptual questions about safety. The identified “factors” simplify interpretation of these relationships by reducing the observed correlations into as few “constructs” as possible. As yet there is no universal consensus about a key set of underlying factors for the concept of safety climate, or even if one exists (Coyle et al., 1995). It has been suggested that analogous to the personality literature (Barrick & Mount, 1991) there is a Big 5 safety climate structure (Flin, Mearns, O'Connor, & Bryden, 2000). An excellent review (Guldenmund, 2000) indicates the complexity of safety climate as a psychological construct and exposes the Big 5 concept as somewhat premature. Many researchers point to the differences between the factor structure in Zohar's (1980) study and those of Brown and Holmes (1986) and Dedobbeleer and Beland (1991) that used the same instrument (or variants) on different populations in different industries and countries to justify why their factor structure differs from other instruments. However, reported differences in the key underlying factor structures may simply reflect methodological differences in question generation (e.g., focus group exercises, literature reviews), sample populations (within or between companies) across industries, and the labeling of constructs according to the theoretical model driving the research (Guldenmund, 2000). In most instances, the purpose of measuring safety climate is to provide opportunities for enquiry or change (Carroll, 1998) so as to improve safety performance in the measured organization. This means that industrial organizations are the major stakeholders of safety climate research. As such, it is very important that a safety climate factor should only be viewed as key if it predicts actual, or ongoing, safety performance in organizations.
Three studies have attempted to validate their factor structures by replication (Rummel, 1970) at two points in time with different (Coyle et al., 1995) or the same (Glendon & Stanton, 2000, Thompson et al., 1998) sample populations. Coyle et al. (1995) obtained a different factor structure, whereas both Glendon and Stanton (2000) and Thompson et al. (1998) obtained similar structures. Such results imply that obtained safety climate factor structures are specific to particular industries and/or sample populations (McDonald & Ryan, 1992) or that different instruments measure distinctly different safety climate concepts (Glendon & Litherland, 2000).
Validity is a quality standard for evaluating safety climate and other psychometric measures that refers to their accuracy and appropriateness for predicting or drawing inferences about certain criteria (e.g., safety performance). Relying solely on discriminant validity (e.g., statistical differences in scale scores by demographic variables such as age) most studies have made no attempt to assess the concurrent or predictive validity of their instrument, or identified factors, against independent variables such as accident rate or other measures of safety performance (Brown & Holmes, 1986, Carroll, 1998, Cheyne et al., 1998, Cox & Cheyne, 2000, Cox & Cox, 1991, Dedobbeleer & Beland, 1991, Donald & Canter, 1994, Fuller & Vassie, 2001, Hayes et al., 1998, Mearns et al., 1998, Neal et al., 2000, Prussia et al., 2003, Rundmo, 1992, Rundmo, 1994, Williamson et al., 1997). The universality of discriminant validity reported across all published studies suggests that sub-group differences within the same organization are a given. This accords with the purpose of safety climate measurement: to identify and explore such differences so as to implement the appropriate remedial interventions (Budworth, 1997). Psychometric [safety climate] instruments are deliberately designed to discriminate between people on various demographic dimensions (Cook, Hepworth, Wall, & Warr, 1993). Thus, any between sub-group differences merely inform about the degree to which the measure has reached its initial design goals. They do not inform about the ability of the measure to assess or predict actual ongoing safety performance. Moreover, correlating demographic data collected at the same time as responses to safety climate questions are collected is not concurrent validation (Bausell, 1986, p216).
Some researchers have attempted to assess concurrent validity (i.e., safety performance at the time of distribution) or predictive validity (i.e., forecast future safety performance) by correlating the scale or factor scores against actual accident rates (e.g., Lee & Harrison, 2000, Mearns et al., 2003, Niskanen, 1994, O'Toole, 2002, Silva et al., 2004, Varonen & Mattila, 2000, Vredenburgh, 2002, Zohar, 2000), expert ratings (Arboleda et al., 2003, Diaz & Cabrera, 1997, Zohar, 1980), human error analysis (Glendon & Stanton, 2000), ratings of behavioral compliance (Garavan & OBrien, 2001), and actual safety behavior (Glendon & Litherland, 2000). With the exception of Zohar (2000) who found a predictive relationship with “micro-accidents” five months after distribution, no safety climate instrument has yet been found to predict actual safety behavior or ongoing levels of safety performance.
Attempts to delineate the underlying safety climate constructs and their relationships with self-report indices of safety activity have been undertaken using a priori Structural Equation Modelling (SEM). Safety activities include subjective appraisals of the physical work environment and workplace hazards (Brown et al., 2000, Cheyne et al., 1998), managerial assessments of employee's safety compliance (Prussia et al., 2003), safety hazards and self-reported compliance (Neal et al., 2000, Thompson et al., 1998), and safety participation (Neal et al.). In the same way that differences are reported in factor structures, vast differences are found in theoretical models derived from this process. Importantly, in all of these studies the path correlations between safety climate and the self-report safety activities show the degree of association between constructs to be moderate at best. Given that correlations between two perceptual constructs tend to be inflated (Miller & Monge, 1986) these modelling results may even be over-estimates of actual relationships. Nonetheless, some (e.g., Glendon & Litherland, 2000) argue that the utility of SEM resides in the revelation that safety climate exerts an indirect effect on safety behavior, which is mediated by further variables such as transformational leadership (Barling, Loughlin, & Kelloway, 2002), the work context (Hofmann & Stetzer, 1996), and production pressures (Brown et al., 2000). The validity of the various SEM models is difficult to ascertain as all are based on self-report instruments and none have used independent variables to verify any relationships obtained. However, all the SEM studies report that the relationship between safety climate and safety activity (i.e., behavior) is mediated by other variables. Overwhelmingly, this body of evidence suggests that there is no direct link between perceptual safety climate constructs and actual safety behavior.
Measurement of safety performance is notoriously problematic as measures such as accident rates and compensation costs tend to be reactive (after the event) and relatively infrequent. This focus on safety results (Cohen, 2002) often means that the success of safety is measured by lower levels of system failure. Many modern approaches (Strickoff, 2000) advocate the use of proactive measures (e.g., safety climate, hazard identification and /or observed percent safe behavior) that focus on current safety activities to ascertain system success rather than system failure. In combination both approaches can help organizations to ascertain the effects of their safety programs.
Derived from behavioral safety, the observed percent safe score is thought to be one of the most useful indicators of current safety performance (Reber, Wallin, & Duhon, 1989). Based on randomized behavioral sampling, employee observers record the number of safe and unsafe behaviors performed by their peers, against predetermined checklists of safety related behaviors derived from accident/incident reports. The observation results are used to compute a percent safe score, which is used in many ways (e.g., set improvement targets) but is primarily intended to provide ongoing feedback (Cooper, Phillips, Sutherland, & Makin, 1994) so that people can adjust their performance accordingly. Reviews of behavioral safety studies have demonstrated dramatic improvements in safety performance (Grindle et al., 2000, McAfee & Winn, 1989, Sulzer-Azaroff et al., 1994) in terms of reductions in accidents, workers compensation costs, and insurance premiums. To date, no published study has established a clear direct link between measures of safety climate and actual safety behavior. The reported relationships between safety climate and safety behavior have largely been inferred from structural equation models based on a variety of self-report instruments. The notable exception to this trend is Glendon and Litherland's (2000) attempt to measure both safety climate perceptions and actual safety behaviors in road construction. Contrary to expectations, but in accordance with structural equation models, this study failed to establish a direct relationship between the two. The authors speculated that the information obtained from the two forms of measurement is so independent that safety climate and safety behavior exist independently under a super-ordinate safety construct (Culture?). However, the authors also postulated that the number of observations conducted over the course of one day in five-minute periods violated recommendations (Tarrants, 1980) for this type of measurement. This suggests two competing explanations: (a) there is no direct link between safety climate and safety behavior or (b) that a relationship between safety climate scale scores will be found if behavioral measurements are taken over longer periods of time.
The present study is an extension of behavioral safety (Cooper et al., 1994) and safety climate work, utilizing a modified version of Zohar's (1980) safety climate instrument (Cooper & Phillips, 1994) carried out in a manufacturing facility. Within the same organization, at the same time that a behavioral safety intervention was conducted, a safety climate survey was completed. The intervention was designed and implemented as a continuous process that would eventually involve all plant personnel in safety observations for a period of three to five months each, in order to improve levels of safe behavior. Twelve months after the first safety climate survey a second survey was conducted with the original survey instrument. Archival behavioral safety data made available to the first author by the study organization presented an opportunity to test various hypotheses pertaining to the current status of safety climate research. Based on the current safety climate literature, the following hypotheses were formulated.
- 1.
When used to survey the same sample population, at different points in time, a similar factor structure will be obtained from the same safety climate instrument.
- 2.
Differences in perceptions will be demonstrated across a variety of demographic variables such as self-reported accident involvement, age, job experience, and functional departments, for both pre and post distributions of the safety climate measure.
- 3.
No direct relationship will be obtained between safety climate scale scores and safety behavior.
- 4.
Measured changes in safety behavior will not be reflected in similar changes in pre and post test safety climate scores.
Section snippets
Sample
The study population was the plant personnel (n=540) of a packaging production plant. The questionnaire, along with return envelopes addressed to the researchers, was initially distributed through the internal mail system to all members of the organization. At the same time a behavioral safety initiative was being implemented across the organization. The response rate for this distribution was 69% (n=374). The average age of all respondents was 45.3 years (SD=10.22) with a range from 18–63
Safety climate
Utilizing the Statistical package for the Social Sciences (SPSS), the data were analyzed to evaluate the safety climate instruments factorial structure for each distribution. The method of factor extraction was principal components, rotated according to a varimax solution when two or more factors emerged. In addition, a “second order” factor analysis of the scales was conducted. The overall reliability of the measure for both distributions was also assessed using Cronbach's Alpha.
Discussion
In accordance with the multiple directions taken by safety climate researchers, this study examined the underlying factor structure of an adapted safety climate measure originally developed by Zohar (1980), and attempted to ascertain the instruments discriminant, concurrent, and predictive validity. This study also explored the relationship between safety climate and safety behavior. Based on previous evidence, four hypotheses were tested.
Impact on industry
The finding that safety climate perceptions will not necessarily match actual levels of safety performance strongly suggests that industry should focus its primary safety improvement effort on changing unsafe situations and conditions as well as people's safety behavior at all organizational levels, rather than concentrating on improving people's attitudes, beliefs, and perceptions about safety. It is reductions in the frequency of unsafe behaviors and their antecedents (i.e., unsafe conditions
Summary
This study has established an empirical link between a limited set of safety climate perceptions and actual safety behavior. It has also demonstrated how complex the overall relationship is: changes in climate perceptions do not necessarily reflect changes in levels of behavioral safety performance. Equally, changes in safety behavior are not necessarily reflected in safety climate perceptions. Such results challenge many of the assumptions that have typified previous research. This degree of
A Registered Safety Practicioner (RSP), A Fellow of the Institute of Occupational Safety & Health (IOSH) and a Chartered Psychologist, Dominic Cooper is an associate professor of safety education and a visiting professor in industrial / organizational psychology at Indiana University, Bloomington. Dr. Cooper holds a BSc from the University of East London, an MSc from the University of Hull and a PhD from the University of Manchester Institute of Science & Technology. His research focuses on
References (80)
- et al.
Management practices as antecedents of safety culture within the trucking industry: similarities and differences by hierarchical level
Journal of Safety Research
(2003) - et al.
Predicting safe employee behavior in the steel industry: Development and test of a sociotechnical model
Journal of Operations Management
(2000) - et al.
Safety Climate
Journal of Safety Research
(1995) - et al.
A safety climate measure for construction sites
Journal of Safety Research
(1991) - et al.
Employee attitudes and safety in the chemical industry
Journal of Loss Prevention in the Process Industries
(1994) - et al.
Perspectives on safety culture
Safety Science
(2000) The nature of safety culture: a review of theory and research
Safety Science
(2000)- et al.
Measuring Perceptions of Workplace Safety: Development and Validation of the Work Safety Scale
Journal of Safety Research
(1998) - et al.
Assessing safety culture in nuclear power stations
Safety Science
(2000) - et al.
Accident prevention. Presentation of a model placing emphasis on human, structural and cultural factors
Safety Science
(2004)
The use of incentives/feedback to enhance workplace safety: A Critique of the literature
Journal of Safety Research
Evaluation of peer intervention training for high school alcohol safety education
Accident Analysis and Prevention
Safety climate, safety management practice and safety performance in offshore environments
Safety Science
The impact of organizational climate on safety climate and individual behavior
Safety Science
Understanding the organizational structure-job attitude relationship through perceptions of the work environment
Organizational Behavior and Human Performance
Assessing the safety environment in work organization of road maintenance jobs
Accident Analysis & Prevention
The relationship between employees' perceptions of safety and organizational culture
Journal of Safety Research
Mental models of safety: do managers and employees see eye to eye?
Journal of Safety Research
Risk perception and safety on offshore petroleum platforms—part II: Perceived risk, job stress and accidents
Safety Science
OSCI: an organisational and safety climate inventory
Safety Science
Where the safety rubber meets the shopfloor: A confirmatory model of management influence on workplace safety
Journal of Safety Research
Cultural influences on risks and risk management: Six case studies
Safety Science
The safety climate and its relationship to safety practices, safety of the work environment and occupational accidents in eight wood-processing companies
Accident Analysis and Prevention
Organizational safety: Which management practices are most effective in reducing employee injury rates?
Journal of Safety Research
The Development of a measure of safety climate: The role of safety perceptions and attitudes
Safety Science
The use of supervisory practices as leverage to improve safety behavior: A cross-level intervention model
Journal of Safety Research
Development and Test of a Model Linking Safety-Specific Transformational Leadership and Occupational Safety
Journal of Applied Psychology
The big 5 personality dimensions and job performance: A meta analysis
Personnel Psychology
Practical Guide to Conducting Empirical Research
The use of a factor-analytic procedure for assessing the validity of an employee safety climate model
Accident Analysis and Prevention
The development and evaluation of a safety climate measure as a diagnostic tool in safety management
IOSH Journal
A tool to assess aspects of an organisations health & safety climate
The role of safety climate in a safety management system
Safety culture as an ongoing process: Culture surveys as opportunities for enquiry and change
Work & Stress
Modelling safety climate in the prediction of levels of safety activity
Work & Stress
Statistical Power Analyses for the Behavioral Sciences
Statistical Power Analyses for the Behavioral Sciences
Measuring safety performance in construction
Occupational Hazards
The Experience of Work: A Compendium and Review Of 249 Measures and Their Use
Towards a Model of Safety Culture
Safety Science
Cited by (547)
Modelling yielding dilemma zone for drivers at non-signalised crosswalks
2024, Transportation Research Part F: Traffic Psychology and BehaviourSafety leadership: A bibliometric literature review and future research directions
2024, Journal of Business ResearchThe role of ethical climates in employee information security policy violations
2024, Decision Support Systems
A Registered Safety Practicioner (RSP), A Fellow of the Institute of Occupational Safety & Health (IOSH) and a Chartered Psychologist, Dominic Cooper is an associate professor of safety education and a visiting professor in industrial / organizational psychology at Indiana University, Bloomington. Dr. Cooper holds a BSc from the University of East London, an MSc from the University of Hull and a PhD from the University of Manchester Institute of Science & Technology. His research focuses on safety culture, safety management systems, behavioral safety and safety climate. He has published in Safety Science, Professional Safety, Journal of Occupational and Organizational Psychology, Leadership and Organisational Development Journal, European Review of Applied Psychology and others.
Robin Phillips is a Behavioral Safety consultant with Marsh UK Ltd. A Member of the Institute of Occupational Safety & Health (IOSH), he holds a Masters in Construction Management from the University of Manchester Institute of Science & Technology. His research interests lie in behavioral safety, Safety climate and Human Error. He has published in the Journal of Occupational and Organizational Psychology, Leadership and Organisational Development Journal, European Review of Applied Psychology and others.