A cross-validation of safety climate scale using confirmatory factor analytic approach

https://doi.org/10.1016/j.jsr.2004.04.006Get rights and content

Abstract

Problem: Given the lack of a consistent factor structure of safety climate, this study tested the stability of a factor structure of a safety climate scale developed through an extensive literature review using confirmatory factor analytic approach and cross-validation. Methods:A cross-sectional sample of 722 U.S. grain industry workers participated in the questionnaire survey. Results: The safety climate scale developed through the generation of an item pool based on a table of specifications, subsequent scientific item reduction procedures, reviews from experts, and pilot test yielded adequate reliabilities for each dimension. Each item showed proper discriminative power based on both internal and external criteria. Criterion validity was manifested by the significant positive correlation of the scale with five criteria. Evidence of construct validity was provided by both exploratory and confirmatory factor analyses. Both calibration and validation samples supported a consistent factor structure. Management commitment and supervisor support were found to influence other dimensions of safety climate. Discussion:This study provides an insight into the primary reason why previous attempts have failed to find a consistent factor structure of safety climate: No specification of the influence of management commitment and supervisor support on other dimensions of safety in their models. Impact on industry: The findings of this study provide a framework upon which accident prevention efforts can be effectively organized and underscore the importance of management commitment and supervisor support as they affect employee safety perceptions.

Introduction

Since the 1980s industry as well as researchers have paid a great deal of attention to the role of organizational and cultural factors as antecedents to accident occurrences (Brown et al., 2000, Flin et al., 2000, Glendon & Litherland, 2001, Hale et al., 1998, Hofmann & Stetzer, 1996, Lee & Harrison, 2000, Mearns et al., 1998, Mearns et al., 2003, Niskanen, 1994, Oliver et al., 2002, O'Toole, 2002, Rundmo et al., 1998, Tomas et al., 1999, Turner & Pidgeon, 1997, Vuuren, 1999, Wagenaar & Groeneweg, 1987, Weick et al., 1999, Wright, 1986, Zohar, 1980). All of these studies, either quantitative or qualitative, revealed that organizational and cultural factors are underlying causal factors of accidents.

Wright (1986) investigated the causes of fatal accidents in the UK offshore oil industry by attending and observing court hearings in 1983–1984 and was able to identify the influence of organizational factors on accidents. Among the several factors Wright revealed was a strong pressure throughout the entire organization to complete the work as quickly as possible even at the expense of safety. The production pressure was mainly due to high interdependency of offshore production processes where delay in one area can cause costly delays in other areas. Workers who perceive a high degree of production pressure will focus their attention more on completing the work fast, readily taking short cuts, than on their safety.

Another similar finding was made by Wagenaar and Groeneweg (1987), who attempted to classify the types of human errors involved in accidents by reviewing 100 sea accidents that were investigated by the Dutch Shipping Council between the years 1982–1985. Wagenaar and Groeneweg found that workers made more frequent errors under high situational stress than would be expected by chance. Wagenaar and Groeneweg also found that social pressure was more powerful for job performance than formalized rules and procedures.

In addition to the organizational factors, cultural factors have drawn attention as antecedents to accidents since the mid 1980s. The 1986 Chernobyl disaster triggered the fusion of the two concepts, safety and culture (Cooper, 2000, Pidgeon & O'Leary, 2000). The Chernobyl disaster was attributed in part to a poor safety culture within the former Soviet nuclear industry as well as at the Chernobyl plant (International Atomic Energy Agency [IAEA], 1988). Since then, interest in the concept of safety culture has grown as safety researchers and practitioners have attempted to define and operationalize the concept. Also, increasingly, there has been confusion between the terms, safety climate and safety culture (Glendon & Stanton, 2000, Hale, 2000).

Schneider (1975) defined organizational climate as “molar perceptions people have of their work settings” (p. 473). People are bound to develop the molar perceptions, added Schneider, because individuals attempt to apprehend order in their environment to be used as a framework for adaptive behavior. While Schneider did not distinguish organizational climate from culture, he argued that distinctions should be made between “perceptions of organizational practices and procedures” and “reactions to those same practices and procedures” (p. 464), the former being the climate. Jones and James (1979) affirmed this, describing climate as “a set of perceptually based, psychological attributes” (p. 205) that are descriptive and cognitive rather than affective and evaluative in nature. Jones and James also stated that the climate is “multidimensional, with a central core of dimensions that apply across a variety of situations” (p. 205).

In management studies, the term organizational culture, defined by Guldenmund (2000) as “a relatively stable, multidimensional, holistic construct shared by (groups of) organizational members that supplies a frame of reference and which gives meaning to and/or is typically revealed in certain practices” (p. 225), replaced the term organizational climate gradually during the 1980s (Hale, 2000). However, many researchers, especially in the safety field, have distinguished the two terms, most of them advocating the retention of both terms (Cox & Flin, 1998, Glendon & Stanton, 2000, Glick, 1985, Guldenmund, 2000, Hale, 2000, Moran & Volkwein, 1992, Schein, 1992). Reichers and Schneider (1990) who reviewed the evolution of the two concepts concluded that “culture exists at a higher level of abstraction than climate, and climate is a manifestation of culture” (p. 29). This distinction was widely advocated by many other researchers (Cox & Flin, 1998, Glendon & Stanton, 2000, Guldenmund, 2000, Schein, 1992). Culture has signified the broad multi-faceted concept that incorporates shared attitudes, values, beliefs, assumptions, and practices whereas climate is conceived of as a limited set of dimensions indicating an organization's culture. Thus, climate is commonly associated with terms such as “superficial” (Glendon & Stanton, 2000, p. 198), “snapshot” (Flin et al., 2000, p. 178), “quantitative” (Guldenmund, 2000, p. 220), and “state” (Cheyne, Cox, Oliver, & Tomas, 1998, p. 256), whereas culture with “deep” (Hale, 2000, p. 5), “stable” (Schein, 1992, p. 5), “qualitative” (Guldenmund, 2000, p. 220), and “trait” (Cheyne et al., 1998, p. 256).

Safety culture is a part of organizational culture. Accordingly, safety culture must take on the characteristics of organizational culture with the scope limited to or related to safety. There are many different definitions of safety culture adopted by researchers (Pidgeon, 1998) because of the extensive dimensionality of safety culture. Among various definitions, which as suggested by Cox and Cox (1991) seemed to be the most succinct and common for many different definitions, “the attitudes, beliefs, perceptions, and values that employees share in relation to safety” (p. 93), whereas the definition proposed by the Advisory Committee on the Safety of Nuclear Installations (Health and Safety Commission [HSC], 1993) has been more widely used (especially in Europe) than any other definitions. This definition taking on a social psychological perspective says, “The safety culture of an organization is the product of individual and group values, attitudes, perceptions, competencies, and patterns of behavior that determine the commitment to, and the style and proficiency of, an organization's health and safety management” (p. 23).

Some researchers, including Lee and Harrison (2000), assessed safety culture in three nuclear power stations in the UK and attempted to develop a measuring instrument of safety culture. However, Cox and Flin (1998) conducted an extensive literature review on safety culture and concluded that safety climate, rather than safety culture, is the preferred metric because safety climate studies provide a limited set of variables that can be operationalized and measured. Glendon and Stanton (2000) supported Cox and Flin, reporting “Questionnaires or similar measures will be inadequate to measure all aspects of organizational culture. Validated questionnaires are acceptable as climate measures” (p. 196). Hale (2000) and Guldenmund (2000) also affirmed this, arguing that it is extremely difficult, if not impossible, to measure safety culture.

Thus, a great deal of attention has been paid to safety climate, especially to the definition of the concept as well as to the development of a reliable and valid instrument to measure safety climate. Although many researchers (Brown & Holmes, 1986, Coyle et al., 1995, Dedobbeleer & Beland, 1991, Niskanen, 1994, Williamson et al., 1997, Zohar, 1980) have proposed different definitions of safety climate, none of them seem to be quite divergent from that proposed by Zohar who defined safety climate as “a summary of molar perceptions that employees share about their work environments” (p. 96). The fact that almost all of the definitions of safety climate include the words shared and perceptions implies that a general agreement has been reached on the concept of safety climate.

A reliable and valid measurement of safety climate offers immense advantages. First of all, measurement of safety climate can overcome the limitations of traditional safety measures such as lost time accident rates and accident investigation reports. Some of the serious limitations of traditional safety measures are that: (a) they are not sensitive enough to provide useful information about safety problems of a specific work site, mainly because accidents are such rare events compared to widespread work hazards and much more frequent near-misses; (b) they do not provide a means to evaluate risk exposure of employees; and (c) they are invariably retrospective (Glendon & McKenna, 1995). Safety climate measurement, however, can offer information about safety problems before they develop into accidents and injuries (Lutness, 1987).

Second, a safety climate survey can focus on safety efforts to improve problematic areas (Cox & Cheyne, 2000), which may also improve other functions of a company (including productivity).

Third, a safety climate survey offers a valuable tool for identifying trends in an organization's safety performance as well as establishing external benchmarks (Cox & Cheyne, 2000, Coyle et al., 1995).

Fourth, compared to other proactive means of accident prevention efforts such as a safety audit, a safety climate survey costs a company much less money and time (although a safety climate survey cannot replace other diagnostic tools and safety activities).

Finally, but most importantly, research has revealed growing evidence that safety climate is associated with safety practices (Zohar, 1980), accidents (Mearns et al., 1998, Mearns et al., 2003), and unsafe behavior (Brown et al., 2000, Cabrera & Isla, 1998, Hofmann & Stetzer, 1996, Tomas et al., 1999). Safety climate has become a “leading indicator” (Flin et al., 2000) of safety performance.

There have been at least 16 different data-based studies published in refereed journals to identify the appropriate constructs of safety climate. However, after an extensive literature review of safety culture and safety climate studies, Guldenmund (2000) pointed out the lack of a reliable and valid instrument to measure safety climate as follows:

While the importance of the concept of safety climate or culture is stressed by most authors, very few have attempted to support their claim by reporting an indication of its construct validity or predictive validity. Most efforts have not progressed beyond the stage of face validity. Basically, this means that the concept still has not advanced beyond its first developmental stages (p. 216).

Besides poor evidence of validity, few instruments have been reused by other researchers, contributing to the little agreement on the number of safety climate dimensions. One possible explanation for the divergence of factor structures would be the use of different populations in different industries or cultures. Another probable explanation arises from the fact that factor labeling is left to the discretion of each researcher. Thus, it is very possible that, due to arbitrary labeling, re-examination of diverse factors of the same construct among different studies may lead to emergence of a more common factor structure. Accordingly, it is imperative to thoroughly examine all the factors of different safety climate scales found in previous studies, which may lead to some insight into ubiquitous dimensions of safety climate. The selection criteria of the safety climate scale development studies were that the study should be: (a) data-based, (b) published in a refereed journal, (c) presented in English; and (d) based on a sample of greater than 150 employees (Cliff, 1987, Gorsuch, 1983).

One of the first safety climate scales was developed by Zohar (1980). Zohar initially developed a questionnaire from seven sets of items that were found to discriminate high accident-rate companies from low accident-rate companies. Next, data were collected from 20 factories in Israel. One of the strengths of this study lies in the use of stratified random sampling procedure. Zohar randomly selected five factories from each of four production industries: metal fabrication, food processing, chemical industry, and textile industry. A stratified random sample of 20 workers was finally selected from each factory (N = 400). The factor analysis using principal component analysis with Varimax rotation method resulted in the retention of eight factors with 40 items. The retained factors were labeled as: (a) perceived importance of safety training programs, (b) perceived management attitudes toward safety, (c) perceived effects of safe conduct on promotion, (d) perceived level of risk at workplace, (e) perceived effects of required work pace on safety, (f) perceived status of safety officer, (g) perceived effects of safe conduct on social status, and (h) perceived status of safety committee.

A couple of points are noteworthy in Zohar's (1980) study. First, he used .49 as a cut-off point for factor loading, which is uncommon. While the decision on the cut-off point is arbitrary, the factor structures resulted from use of different cut-off points are likely to be different. Considering that the majority of safety climate researchers use .40 as a cut-off point for factor loading (Cox & Cox, 1991, Coyle et al., 1995, Niskanen, 1994, Williamson et al., 1997), which is supported by Velicer, Peacock, and Jackson (1982), simple comparison of a factor structure found by other researchers with that of Zohar could result in misleading interpretation. The other notable point is that among the eight factors, only two factors—perceived management attitudes toward safety and perceived importance of safety training programs—accounted for 60% of the variance. Also, all the question items except for those asking perceived risk and work pace were very closely related to management commitment to safety. These imply that management commitment to safety is one of the most important factors in safety climate.

Brown and Holmes (1986) attempted to replicate the factor structure Zohar (1980) found using confirmatory factor analysis of the Linear Structural Relationship (LISREL) program, version 6. They administered Zohar's 40-item questionnaire to an American sample of 425 production workers who were conveniently selected from companies in Wisconsin and Illinois. Brown and Holmes, however, failed to support Zohar's factor structure. Instead, they found a three-factor structure using an exploratory factor analytic approach of LISREL program, which indicated a good fit to the data. The three factors retained were: (a) employee perception of how concerned management was with their well-being, (b) employee perception of how active management was in responding to this concern, and (c) employee physical risk perception. Brown and Holmes employed a sound and rigorous methodology to extract factors from the data and succeeded in demonstrating factorial validity of their three-factor structure. However, they did not validate their safety climate scale by showing its association with an external criterion such as unsafe work behavior, which limits the validity of their scale.

Dedobbeleer and Beland (1991) tested Brown and Holmes' (1986) three-factor structure of safety climate scale on an American sample of 272 construction workers. The sample was conveniently selected from nine nonresidential construction sites located in the Baltimore metropolitan area. Question items were similar but not identical to those included in Brown and Holmes' questionnaire. When the maximum likelihood method in LISREL program chosen by Brown and Holmes was used, Brown and Holmes' three-factor structure was retained. However, when the weighted least squares method in LISREL program was used, which was more appropriate for Dedobbeleer and Beland's data because their data were ordinal level, the best fit model was a two-factor model. The two factors were management commitment to safety and workers' involvement in safety. A slightly modified version of Dedobbeleer and Beland's questionnaire was used by Gillen, Baltz, Gassed, Kirsch and Vaccaro (2002) to 255 construction workers in California in which showed a significant difference in perceived safety climate between union workers and nonunion workers. Gillen et al. however, did not investigate the factor structure of the scale.

Unlike the previous three safety climate scales that were focused on shared perceptions regarding safety, Cox and Cox (1991) incorporated more individual constructs (such as personal attitudes to safety) into their safety climate questionnaire and administered their survey to employees in industrial gas distribution depots of a European company. The data were collected from a large number of locations in five European countries: The Netherlands, the United Kingdom, France, Germany, and Belgium. A total of 630 cases was factor analyzed and resulted in the retention of five factors: (a) personal skepticism, (b) individual responsibility, (c) safeness of work environment, (d) effectiveness of arrangements for safety, and (e) personal immunity.

Although Cox and Cox (1991) conducted data analyses appropriately, a few weaknesses of their study should be discussed. First of all, Cox and Cox did not show evidence of criterion validity and content validity of their instrument. No association was established between the measured safety attitudes and a safety performance measure. Also, the development of questionnaire items was primarily based on a discussion with the company safety manager, rather than on accumulated findings of previous research. Second, Cox and Cox surveyed all the company's personnel including managers, supervisors, and production workers and aggregated the data regardless of hierarchical level. This is not warranted because employees of different hierarchical level can have different perceptions and attitudes (Gonzalez-Roma, Peiro, Lloret, & Zornoza, 1999). Gonzalez-Roma et al. tested the validity of the collective climate concept and found that collective climate membership was significantly related to hierarchical level. Lastly, as Cox and Cox acknowledged, their question items were more related to personal attitudes than shared perceptions. Thus, their instrument is not parallel to other safety climate instruments when safety climate is defined as shared perceptions among employees regarding safety. Despite these weaknesses, Cox and Cox's study provided useful perspectives on possible individual contributory factors to unsafe work behavior – personal skepticism and personal immunity from accidents.

Niskanen (1994) examined the dimensions of safety climate of workers (N = 1,890) and supervisors (N = 562) working for the Finnish National Road Administration in road maintenance, road and bridge construction, and central repair shops. Niskanen aggregated data separately for workers and supervisors and conducted an exploratory factor analysis for the two groups of data. Four factors were retained for the safety climate of workers: (a) attitude toward safety in the organization, (b) changes in work demands, (c) appreciation of the work, and (d) safety as a part of productive work. Little weight, however, should be put on Niskanen's findings because Niskanen did not test any reliability and validity of the safety climate instrument. The same serious limitation is found in Coyle et al.'s (1995) study in which they examined the dimensions of safety climate of two Australian clerical and health care organizations.

Diaz and Cabrera (1997) measured the dimensions of safety climate of three Spanish airport ground handling companies using 166 subjects from those companies. The principal component analysis resulted in six factors: (a) company policies toward safety, (b) emphasis on productivity versus safety, (c) group attitudes toward safety, (d) specific strategies of prevention, (e) safety level perceived in the airport, and (f) safety level perceived on the job. Examination of factor contents was not possible because question items were not reported in the article. Of the 60.8% of the total variance explained, 38.9% was accounted for by company policies toward safety, which reflects management commitment to safety. The strength of this study was that Diaz and Cabrera showed criterion validity of their safety climate scale. The ratings of general safety level appraised by 29 experts in airport ground handling activities aligned with safety climate rankings. However, this study did not use an appropriate sample size required to yield reliable results. One of the widely used criteria to determine the number of cases in factor analysis is subjects-to-variables (STV) ratio, where the ratio is recommended to be five or greater (Bryant & Yarnold, 1995). Since Diaz and Cabrera used 40 variables, the sample size should be at least 200 according to the STV rule. Another limitation was that Diaz and Cabrera did not assess the reliability of each dimension of their safety climate scale.

Subsequent studies for safety climate scale development had relatively few methodological problems in terms of appropriate use of statistical techniques and sample size, although Cox and Cheyne (2000) and Glendon and Litherland (2000) barely maintained the STV ratio at five. Among the nine subsequent studies (Table 1), three studies (Cox & Cheyne, 2000, Mearns et al., 1998, Mearns et al., 2003) were based on the sample of offshore oil and gas installations in the UK; Lee and Harrison (2000) on three nuclear power stations in the UK; Glendon and Litherland (2001) and Williamson et al. (1997) on Australian workplaces; Cheyne et al. (1998) on four plants of a manufacturing company in the UK and France; and Brown et al. (2000) and O'Toole (2002) on U.S. workplaces.

A couple of studies among the nine (Table 1) are notable. Williamson et al. (1997) experimented with an interesting approach in developing their safety climate scale. Williamson et al. included attitudinal as well as perceptual items in their safety climate item pool and found, during the item analysis, that most of the highly skewed items were attitudinal. This implies that perception-based items are more likely to have discriminative power than attitude-based items. Also, the finding that all the skewed responses were without exception in the favorable direction indicates that attitudinal items are more susceptible than perceptual items to “social desirability” response bias (Singleton & Straits, 1999, p. 308).

Lee and Harrison's study (2000) was quite different from the other 15 scale development studies in terms of scope of measurement. Lee and Harrison attempted to measure various aspects of safety culture, mostly safety-related attitudes, using a 120-item questionnaire that covers eight different domains. Among them were 23 questions regarding job satisfaction and 20 questions regarding contractors. Lee and Harrison conducted principal component factor analysis for each domain, which resulted in the retention of 28 different factors. This study has several limitations. First of all, the selection of the eight domains upon which development of question items was based was somewhat arbitrary. Lee and Harrison did not provide rationale for the selection of the eight domains. Also, they did not present an operational definition of safety culture that the questionnaire was purported to measure. These converge on the probable lack of content validity as well as of construct validity of the instrument. Secondly, factor analysis that was conducted for each domain rather than for the entire variables was flawed. Because many items in different domains were inter-correlated as Lee and Harrison acknowledged, the factor analysis should have been performed for the entire data set. They precluded the possibility of correlated variables in different domains being loaded on a common factor. For example, management's concern for safety and management's concern for health, two among the 28 factors, might have loaded on one single factor if factor analysis had been performed for the entire variables. The same might be true for the other two factors: Quality of training induction and general quality of training. Thirdly, as noted in findings of Williamson et al. (1997), the attitudes-based questionnaire might have yielded relatively unreliable estimates of safety-related parameters. Lee and Harrison's failure to find any relationship between personal risk taking and accident probability measured by nine different criteria of accident history might be due to the inherent limitation of the attitudinal items.

As can be seen in Table 1, leadership support that includes management commitment to safety and supervisor safety support was the most common dimension out of the various, differently labeled dimensions of the 16 safety climate scales, appearing in two-thirds of the questionnaires. The next common dimension was employee participation, appearing in seven questionnaires, followed by work pressure, competence level, and hazard level in the work environment that appeared in a third of the questionnaires. Others were perceived risk, coworkers' safety support, and perceived barriers to safety. Among these nine emergent themes in safety climate, the following five constructs appear to constitute the core of generic safety climate concept, shared perceptions regarding safety that is less affected by site specificities: (a) management commitment to safety, (b) supervisor safety support, (c) coworker safety support (hereafter referred to as coworker support), (d) employee participation in safety-related decision making and activities (hereafter referred to as employee participation), and (e) competence level of employees with regard to safety (hereafter referred to as competence level).

Section snippets

Overview

This study used a purposive sample of 722 U.S. workers throughout the nation who worked for grain elevator facilities of a multi-national grain company based in Illinois (Standard Industrial Classification code: 5153). The target population of this study was all the floor workers in the U.S. working at grain elevator facilities. The selection of the target population was based on potential applicability of research findings to other industries in terms of injury incidence rate and common work

Sample characteristics

Seven-hundred-thirty-six employees in 102 grain elevator locations throughout the nation were invited to participate in the study and 722 (98%) completed the survey. Of the 722 returned surveys, 620 were determined to be usable based upon the criteria regarding percentage of missing data and presence of systematic response patterns. The sample of this study (N = 620) represented 15 different states and was composed of 530 non-clerical workers and 90 clerical workers. As shown in Table 2, the

Discussion

Provision of a valid and reliable safety climate scale may bring enormous benefits to the industry. It can provide proactive, rather than reactive, information about safety problems before they develop into accidents and injuries (Lutness, 1987). It allows focusing of safety efforts for improvement of problematic areas (Cox & Cheyne, 2000), which is cost-effective with limited safety resources. In addition, it enables a company with multiple plants to make valid comparisons of safety climate

Dong-Chul Seo, Ph.D. MS, CHES, is a full-time lecturer at Indiana University. A professional with more than 12 years experience at Korean OSHA and expertise in measurement and evalution, he teaches statistics and research methodology and conducts research in the field of safety climate, risky health and safety behavior, and policy analysis.

References (81)

  • A.R. Hale

    Culture's confusions

    Safety Science

    (2000)
  • A.P. Jones et al.

    Psychological climate: Dimensions and relationships of individual and aggregated work environment perceptions

    Organizational Behavior and Human Performance

    (1979)
  • T. Lee et al.

    Assessing safety culture in nuclear power stations

    Safety Science

    (2000)
  • K. Mearns et al.

    Safety climate, safety management practice and safety performance in offshore environments

    Safety Science

    (2003)
  • T. Niskanen

    Safety climate in the road administration

    Safety Science

    (1994)
  • M. O'Toole

    The relationship between employees' perceptions of safety and organizational culture

    Journal of Safety Research

    (2002)
  • N. Pidgeon et al.

    Man-made disasters: Why technology and organizations (sometimes) fail

    Safety Science

    (2000)
  • T. Rundmo et al.

    Organisational factors, safety attitudes and workload among offshore oil personnel

    Safety Science

    (1998)
  • W.A. Wagenaar et al.

    Accidents at sea: Multiple causes and impossible consequences

    International Journal of Man-Machine Studies

    (1987)
  • A.M. Williamson et al.

    The development of a measure of safety climate: The role of safety perceptions and attitudes

    Safety Science

    (1997)
  • P.M. Bentler

    On the fit of models to covariances and methodology to the Bulletin

    Psychological Bulletin

    (1992)
  • P.M. Bentler et al.

    Significance tests and goodness-of-fit in the analysis of covariance structures

    Psychological Bulletin

    (1980)
  • K.A. Bollen

    A new incremental fit index for general structural models

    Sociological Methods and Research

    (1989)
  • K.A. Bollen

    Structural equations with latent variables

    (1989)
  • M.W. Browne et al.

    Alternative ways of assessing model fit

  • F.B. Bryant et al.

    Principal-components analysis and exploratory and confirmatory factor analysis

  • Bureau of Labor Statistics [BLS] (n.d.). 2001 Incidence rates of nonfatal occupational injuries and illnesses by...
  • B.M. Byrne

    Structural equation modeling with LISREL, PRELIS, and SIMPLIS: Basic concepts, applications, and programming

    (1998)
  • D.D. Cabrera et al.

    The role of safety climate in a safety management system

  • A. Cheyne et al.

    Modelling safety climate in the prediction of levels of safety activity

    Work and Stress

    (1998)
  • C.P. Chou et al.
  • N. Cliff

    Analyzing multivariate data

    (1987)
  • G. Coenders et al.

    Alternative approaches to structural equation modeling of ordinal data: A Monte Carlo study

    Structural Equation Modeling

    (1997)
  • M.D. Cooper

    Towards a model of safety culture

    Safety Science

    (2000)
  • S. Cox et al.

    The structure of employee attitudes to safety: A European example

    Work and Stress

    (1991)
  • S. Cox et al.

    Safety culture: Philosopher's stone or man of straw?

    Work and Stress

    (1998)
  • R.F. DeVellis

    Scale development: Theory and applications

    (2003)
  • R.I. Diaz et al.

    Safety climate and attitude as evaluation measures of organizational safety

    Accident Analysis and Prevention

    (1997)
  • J.C. Flanagan

    The critical incident technique

    Psychological Bulletin

    (1954)
  • D.W. Gerbing et al.

    Validity of exploratory factor analysis as a precursor to confirmatory factor analysis

    Structural Equation Modeling

    (1996)
  • Cited by (0)

    Dong-Chul Seo, Ph.D. MS, CHES, is a full-time lecturer at Indiana University. A professional with more than 12 years experience at Korean OSHA and expertise in measurement and evalution, he teaches statistics and research methodology and conducts research in the field of safety climate, risky health and safety behavior, and policy analysis.

    Mohammad R. Torabi, Ph.D., Chancellor's Professor at Indiana University is Chairperson of the Department of Applied Health Science. His research focus has been in the area of measurement and evaluation of school and public health education programs and factors associated with health behavior.

    Earl Blair, Ed.D., is an Associate Professor of Safety Management at Indiana University. He conducts research for improving safety performance through leadership, safety culture, and management. He is an award winning teacher and author and a frequent keynote speaker at safety conferences.

    Nancy Ellis, H.S.D., is an Associate Professor of Public Health in the Department of Applied Health Science at Indiana University. Her research focus has addressed health risk behaviors of adolescents and adults.

    View full text