You are on page 1of 13

Safety Science Vol. 25. No. 1-3. pp.

15-27, 1997
Pergamon 0 1997 Elsevier Science Ltd. All rights reserved
Printed in the Netherlands
092%7535/97 $17.00 + 0.00

PII: SO9257535(97)00020-9

THE DEVELOPMENT OF A MEASURE OF


SAFETY CLIMATE: THE ROLE OF SAFETY
PERCEPTIONS AND ATTITUDES ’
Ann M. Williamson a**, Anne-Marie Feyer b,
David Cairns ‘, Deborah Biancotti d
a University of New South Wales, School of Psychology, Sydney, 2052, Australia
b New Zealand Occupational and Environmental Health Research Centre, PO Box 913,
Dunedin, New Zealand
’ Macquarie University, School of Behavioural Sciences, Sydney, 2109, Australia
d National Institute of Occupational Health and Safety (Worksafe Australia) PO Box 58,
Sydney, 200 1, Australia

Abstract-In understanding the safety climate or culture of a workplace, the perceptions and
attitudes of the workforce are important factors in assessing safety needs. Safety solutions may
fail if they do not take into account these prevailing attitudes and perceptions. Also, changes in
attitudes and perceptions about safety are often likely outcomes of safety interventions. The aim
of this study was to develop a measure of perception and attitudes about safety as an indicator
of safety culture for use with working populations. After reviewing the structure and content of
a number of older, related measures of safety attitudes, a 67-item questionnaire was developed.
The questionnaire was distributed in a self-administered form to 1560 workers in a wide variety
of types of jobs and 660 questionnaires were returned from 7 workplaces with a response rate of
42%. Items were evaluated by Cronbach alpha which showed that the item set shared
considerable common variance. Scale development was approached in two ways. First items
which were highly skewed were removed, resulting in removal of around half of the items, with
very little loss in alpha. The remaining items were then subjected to factor analysis which
revealed five factors; personal motivation for safe behaviour, positive safety practice, risk
justification, fatalism and optimism. This factor structure showed acceptable psychometric
properties. Second, in an effort to produce a unidimensional scale with a small, but representa-
tive set of items, additional items were removed which had high correlation with other items.
This resulted in loss of around half of the remaining items. The short scale contained 17 items
with again overall acceptable psychometric properties. The most striking finding in the
development of this questionnaire was that there was little variation between respondents on a
very large proportion of the questions originally selected. Clearly there are well-known beliefs
about safety in the working community which need to be understood in order to progress the
concept of safety culture. 0 1997 Elsevier Science Ltd.

* Corresponding author.
’ The views expressed in this paper are those of the authors and do not necessarily reflect the
views of their Institutions.

15
16 A.M. Williamson et al.

Safety climate is argued to be one of the contributors to the climate in organisation,


conceptualised by Zohar (1980) as a summary of the beliefs and perceptions of employees
about safety in the workplace. In its original conception, it was assumed that the safety climate
acts as a frame of reference which guides behaviour, such that employees develop “coherent
sets of perceptions and expectations regarding behaviour-outcome contingencies and behave
accordingly” (Zohar, 1980, p. 96). In short, safety climate is a summary concept describing
the safety ethic in an organisation or workplace which is reflected in employees’ beliefs about
safety and is thought to predict the way employees behave with respect to safety in that
workplace. The need for the concept has arisen from various areas, including the need to
describe the factors underpinning safe behaviour in the workplace, the need to define the
‘flavour’ of safety in an organisation, and as one of the factors which will moderate change in
workplaces.
Safety climate has been researched for approximately 25 years and there have been many
attempts to develop methods which can be used to measure the concept. Zohar (1980)
developed the first measure, after reviewing the literature, reporting characteristics which
differentiated high and low accident-rate companies. After factor analysis, his final model
based on an Israeli sample included dimensions covering workers perceptions of: the
importance of safety training, management attitudes towards safety, effects of safe conduct on
promotion, level of risk at workplace, effects of work pace on safety, status of safety officer,
effects of safe conduct on social status and status of safety committee. An attempt to replicate
this structure using confirmatory factor analysis did not support these dimensions in a sample
of workers from the USA (Brown and Holmes, 1986). Further analysis of their data set
revealed a three factor structure which was shown to have practical importance as it showed
differences in safety perception between employees who had experienced an accident and
those who had not. The three factors revealed by this analysis were employee perceptions of
management concern about their well-being, management activity in responding to problems
with their well-being and their own physical risk. Dedobbeleer and Beland (1991) tested this
3-factor model in construction workers and found that it was supported by their data, but that a
two factor solution was superior. The two factors were interpreted to be management
commitment to safety and workers’ involvement in safety.
This work has been followed by further attempts at modelling the concept of safety climate.
For example, Seppala’s work produced a three factor model of safety climate; organisational
responsibility for safety, workers’ concern about safety and workers’ indifference towards
safety (Seppala, 1992), as did work by Donald et al. (199 1) which revealed three facets of
safety attitude; people or the organisational roles which make up the safety climate (e.g.,
supervisor, manager, workmates), attitude behaviour or aspects of an individual’s safety
behaviour (e.g., knowledge, satisfaction, actual behaviour) and safety activity or type of safety
behaviour (passive, e.g., wearing safety clothing; active, e.g., attending safety meetings). A
study by Cox and Cox (1991) of employee attitudes in an organisation which manufactured
industrial gases produced five factors; personal skepticism, individual responsibility, the
safeness of the work environment, the effectiveness of arrangements for safety and personal
immunity. A recent study by Niskanen (1994) studying road workers found two separate four
factor solutions for workers and for supervisors. Both included attitudes to safety in the
organisation, changes in work demands and safety as part of productive work. The structure
for each group also included a similar factor describing perceptions of the work which were
labelled workers appreciation of the work and for supervisors, value of the work.
There is clearly little real agreement amongst these previous studies on the dimensions
The development of a measure of safer?, climate 17

which should be incorporated into a safety climate model. Across all studies, two areas seem
to be reflected consistently, views about management attitudes to safety and about workers’
involvement or attitudes to safety. The interpretation of these two factors is however very
broad and has been generated from quite different types of questions. For example, Zohar’s
( 1980) original study and consequently that of Brown and Holmes (1986) used questions about
workers’ perceptions about the situation in their workplace with respect to safety. Dedobbeleer
and Beland (1991) also mainly used questions which called for workers to report how they
saw their workplace, for example, “How important do you think the workers’ safety practices
are to the management of your company?“. In contrast, studies by Niskanen (1994) and Cox
and Cox (199 1) included questions which were much more about attitudes to safety in general,
for example, “Accidents occur by chance” and “If I worried about safety, I would not get my
job done”. In addition, there has been relatively little cross matching of data from previous
studies in helping to develop the concept of safety climate and also very little involvement of
theory (Niskanen, 1994). It is not surprising therefore, that there has been so little consensus
about the safety climate concept and its dimensions. Clearly, further work is needed to gain
better understanding of the concept of safety climate and to develop an appropriate measure of
the concept.
In this study the aim was to develop a measure of attitudes, perceptions and awareness of
safety that are considered to be related to safety climate in workplaces. The approach used was
to develop an initial pool of items and then to use factor analysis and an item analysis
approach to develop scales to measure workplace perceptions and attitudes about safety
(DeVellis, 199 1).

1. Method

1.1. Development of an item pool

Based on an extensive review of the literature, eight aspects were identified by one or more
studies of safety climate. These aspects are shown in Table 1 along with the number of items

Table 1
Factors included in the original questionnaire design, the total number of items included for each one and the number
of items from existing questionnaires (in brackets)
Factor Description Number of items
1. Safety awareness Attitudes to hazards and risks and possibility personal injury 2 (2)
in the workplace
2. Safety responsibility Attitudes about whose role is ensuring safety in the workplace 7 (3)
3. Safety priority Beliefs about the importance of safety in the workplace 5 (2)
4. Management safety commitment Perceptions of management commitment to safety issues 6 (1)
5. Safety control Attitudes to the controllability of accidents 9 (4)
6. Safety motivation Attitudes and perception relating to the influences motivating 14 ( 1)
safe or unsafe behaviour
7. Safety activity Perceptions of the individual’s own safe behaviour 10 (1)
8. Safety evaluation Perceptions of safety in the individual’s own workplace. 9 (2)

Total 62 (16)
18 A.M. Williamson et al.

in the initial questionnaire relating to each one. Both perceptual and attitudinal items were
included in roughly equal proportions (attitudinal/philosophical based = 42%,
perceptual/reality based = 58%). Attitudinal questions were defined as focussing on beliefs
about safety (e.g., “Everyone has an equal chance of having an accident” and “Workmates
should encourage each other to work safely”) whereas perceptual/reality questions focussed
on the individual views about their safety situation (e.g., “All the safety rules and procedures
in my workplace really work”; “When I have worked unsafely it has been because I was not
trained properly”). Sixteen of the items were adapted from two other questionnaires,
specifically those of Cox and Cox (1991, 12 items) and Dedobbeleer and Beland (199 1, 4
items). Sixty-two items covered all of these eight factors. In addition, a number of items were
included as possible validation items. Three questions were included regarding the existence
of dangers in the workplace and the likelihood of accidents in the workplace, two questions
about the meaning of safety and nine demographic-type questions were included at the end of
the questionnaire. Specifically the demographic questions asked about age, gender, language
background, education, experience in the current job, employment status and type, accident
experience and severity of accident.

1.2. Determining the format for measuring

The initial format for responses to the items in the questionnaire was a 100 millimetre
visual analogue scale with descriptors (True/False or Always/Never) as anchors at either end
of the scale. This version of the questionnaire was administered to a sample of 250 workers in
order to assess their reaction to the format and content of the items. Responses were collected
from 69 workers (27% of sample). Based on comments and responses to the questionnaire, it
was clear that respondents had difficulty with the visual analogue scale format and with some
of the descriptors used as anchor points.
The questionnaire was modified to use likert-format questions with five categories of
response for all 62 items. For most items the categories ranged from strongly agree to strongly
disagree but for eight items focussing on worker perception of safety activities in their
workplace, the categories ranged from always to never. This second version also contained
four questions from the safety evaluation section which were in a Yes/No response or
open-ended question format.

1.3. Administering the questionnaire to a development sample

The second version of the questionnaire was distributed to 1570 workers from seven
workplaces. These workplaces covered heavy and light manufacturing industry and outdoor
workers. The workplaces were selected on the basis that they represented the types of
industries where the final version of the scale will be used. All workgroups were invited
personally to participate by a member of the research team who assured anonymity. No names
were required on the questionnaires. A sealable envelope was handed out in which all
questionnaires were returned. In some workplaces returns were put into a locked return box
which was only opened by a member of the research team and for others the questionnaires
were returned independently by each respondent by mail.
The development of a measure of safety climate 19

2. Results

Responses were obtained from 660 workers which represents a 42% overall response rate.

2.1. Evaluation of the items

The 62 likert-type questions were subjected to traditional item evaluation. Cronbach alpha
was used as the first step to estimate the internal consistency of the items or the proportion of
the total variance across all items that is attributable to a common source. With all 62 items,
the obtained Cronbach alpha was 0.81 (n = 543) which was judged to be good. In addition,
the mean and distribution of scores were examined for each item. Thirty items were found to
have highly skewed distributions with more than 70% of respondents in one of the extreme
categories. Skewed items tended to come from only some of the factors used in the
construction of the item pool. The factors with a majority of highly skewed items were (see
Table 2):

(i> Safety responsibility (86% of items) For example, almost all respondents agreed with
the statement “Safety is the responsibility of both management and the worker
together”.
(ii) Management safety commitment (83% of items), For example, almost all respondents
agreed with the statement “Management should be as concerned with people’s safety
as it is with profits”.
(iii) Safety priority (60% of items). For example, almost all respondents agreed with the
statement, “My own safety is a high priority for me in my job.”
(iv) Safety activity (50% of items) For example, almost all respondents agreed with the
statement that “I wear safety equipment when required”.

Almost half of the questions from other questionnaires included inthis questionnaire were
skewed (7 out of 16 questions). The distribution across companies was also very consistent
and a significant number of questions showed no response range or a very low range of
responses. For seven questions there was virtually no response range as 90 to 100% of
respondents in all were in the two extreme categories, for 17 questions there was very low
response range with three-quarters of respondents in all companies being skewed and for 28
questions more than 60% of respondents in all companies were skewed. In addition, only 23%
of the attitudinal items were not skewed, compared to 69% of perceptual/reality based items,
showing that consensus occurred mostly on attitudinal items.

Table 2
Percentage of items dropped from each of the factors covered in the original item pool because of skewedness
Aspects covered in original item pool % items dropped due to skew

1. Safety awareness 0
2. Safety responsibility 86
3. Safety priority 60
4. Management safety commitment 83
5. Safety control 33
6. Safety motivation 43
7. Safety activity 50
8. Safety evaluation 22
20 A.M. Williamson et al.

2.2. Optimising scale length

Although the Cronbach alpha was good, at 62 items, the questionnaire was too long to be
acceptable for use in workplaces. It was not possible to eliminate items using the criterion of
deleting those which did not have much effect on alpha because the sample size was large and
alpha did not change much with deletion of any item. Items were eliminated, therefore, based
on their skewed distributions, resulting in 30 items being dropped. The remaining 32 items
had a Cronbach alpha of 0.75 (n = 563) which was still very acceptable. Almost halving the
scale length produced only a small drop in alpha.
At this point in the scale development it was decided to investigate two approaches. The
first approach was to attempt to explore the underlying factor structure of the 32 items. This
would have the advantage of revealing further information about the roots of the safety climate
concept and allow understanding of the contribution of specific aspects. The tradeoff for this
first approach was that for workplace applications, there were still too many items. The second
approach therefore, was to maximise the coverage of the scale, but minimise redundancy by
eliminating one member of intercorrelated pairs, taking the ‘best’ (most well-written and
easily understood) item of each pair, thus keeping the items in one scale. The advantage of the
second approach was that it emphasises the practical applicability of the scale for workplace
use. The tradeoff for the second approach was that with little correlation between items, the
final product would be unidimensional with no underlying factor structure. Details of the
development of these two approaches is given in the next two sections.

2.2.1. Approach I: Long version scale with factor structure


The 32 items were subjected to a factor analysis with principal component extraction and
varimax rotation. This yielded an interpretable solution of five factors using the scree test and
accounted for 47.5% of variance. The final solution included 27 items which loaded at above
0.4 on one of the five factors while not loading similarly on another factor (see Table 3). Five
items failed to load to this level on any factor. Table 3 shows the percentage of variance
accounted for by each factor, the Cronbach alpha for each factor, the name given to each
factor and the component items of each factor as well as their factor loadings. The first factor
extracted was interpreted as Personal motiuation for safety as it comprises 8 items on aspects
which would promote safer behaviour in respondents. The second factor was interpreted as
Positive safety practice because it contains items which reflect safety activity in the work-
place. The interpretation of the third factor was as Risk justification since it is a collection of
items about the instances when and the reasons why the individual worked unsafely or took
risks. The fourth factor was interpreted as Fatalism because it contains items which focus on
lack of control over being safe. Lastly, the fifth factor was interpreted as Optimism as the
items reflect a favourable view of personal accident risk. All factors contained at least four
items and the internal consistency across items in each factor (alpha) was high for all factors
except factor 5, Optimism. On closer inspection of the data, the relatively low alpha for factor
5 was found to be due to one negative correlation between two items, v23 (People who work
to safety procedure will always be safe) and v29 (Not all accidents are preventable, some
people are just unlucky). The deletion of one of these items (~29) did not improve alpha very
much at all. It was decided to keep factor 5 and explore its characteristics further in the
validation phase of the development of this long scale.
The factors in the long scale were found to be largely independent from one another as the
intercorrelations of the factor scores were very low (see Table 4). The best correlation was for
The development of a measure of safety climate 21

Table 3
Results of the rotated factor analysis showing name of each factor, percentage of variance accounted for by each
factor, the internal consistency between items for each factor (alpha) and the factor loadings for each item
Factor 1: Personal motivation for safe behaviour (l&O%, alpha = 0.86)

v41 It would help me to work more safely if my supervisor praised me on safe behaviour 0.62
v42 It would help me to work more safely if safety procedures were more realistic 0.7 1
v43 It would help me to work more safely if management listened to my recommendations 0.77
v44 It would help me to work more safely if we were given safety training more often 0.70
v45 It would help me to work more safely if the proper equipment was provided more often 0.79
v46 It would help me to work more safely if management carried out more workplace safety checks 0.75
v47 It would help me to work more safely if my workmates supported safe behaviour 0.73
v48 It would help me to work more safely if I was rewarded (paid more) for safe behaviour 0.51

Factor 2: Positive safety practice (11.9%, alpha = 0.84)

v59 Our management supplies enough safety equipment 0.78


v60 Our management checks equipment to make sure it is free of faults 0.76
v61 There is adequate safety training in my workplace 0.74
v62 Management in my workplace is as concerned with people’s safety as it is with profits 0.73
v63 Everybody works safely in my workplace 0.60
V&l All the safety rules and procedures in my workplace really work 0.69

Factor 3: Risk justification (7.2%, alpha = 0.79)

v51 When I have worked unsafely it has been because I was not trained properly 0.79
v52 When I have worked unsafely it has been because I didn’t know what I was doing wrong 0.83
at the time
v53 When I have worked unsafely it has been because I needed to complete the task quickly 0.65
v54 When I have worked unsafely it has been because the right equipment was not provided 0.65
or wasn’t working

Factor 4: Fatalism (5.9%, alpha = 0.65)

v12 Safety works until we are busy then other things take priority 0.66
v16 If I worried about safety all the time I would not get my job done 0.65
v24 I cannot avoid taking risks in my job 0.63
v26 Accidents will happen no matter what 1 do 0.57
v25 I can’t do anything to improve safety in my workplace 0.48

Factor 5: Optimism (4.2%) alpha = 0.39)

v27 It is not likely that I will have an accident because I am a careful person 0.62
v29 Not all accidents are preventable, some people are just unlucky 0.46
v23 People who work to safety procedure will always be safe 0.43
v4 In the normal course of my job, I do not encounter any dangerous situations 0.4 1

Not included:

v3 Everyone has an equal chance of having an accident


v9 People who do not take the necessary precautions are responsible for what happens to them
v22 Management claims it is concerned with safety, but I do not believe this
v50 I make sure I know about new equipment and procedures
v58 Our management only notices when we perform a job unsafely, not when we work safely.

Factors 1 and 3 (Personal motivation for safe behaviour and Risk justification) but the shared
variance for these two factors was only 15%. Most of the other intercorrelations revealed less
than 6% of variance.
22 A.M. Williamson et al.

Table 4
Intercorrelations between factors identified for the long scale
Factor 1: Factor 2: Factor 3: Factor 4: Factor 5:
Personal Positive Risk Fatalism Optimism
motivation safety justification
for safety practice

Factor 1: Personal motivation for safety 1.00 -0.011 0.39 0.12 0.08
Factor 2: Positive safety practice 1.00 - 0.20 -0.18 0.23
Factor 3: Risk justification 1.00 0.13 0.78
Factor 4: Fatalism 1.00 0.16
Factor 5: Optimism 1.00

Validation of the long scale: The validity of the factor structure of the scale was examined
using responses to two of the additional questions in the original item pool, “Do you think
there are any dangers in your workplace?” and “Have you ever had an accident while
working?“. Factor scores were created and compared for respondents answering yes or no to
each of these questions using one-way ANOVA. The results of this analysis can be seen in
Table 5. For the question about perceived dangers in the workplace, Factors 1, 2, 3, and 5
showed significant differences between the groups of respondents. Respondents who perceived
that there were hazards in their workplace showed significantly higher need for Personal
motivators to be safe (Factor l), reported poorer safety practice in their workplace (Factor 2),
a greater tendency to justify unsafe working (Factor 3) and to be optimistic about personal risk
in the workplace (Factor 5). For the second validity question regarding accidents while
working, two factors discriminated significantly between respondents who had experienced

Table 5
Results for the validation of the long version of the scale, showing mean summed scores for each factor and ANOVA
results for each validation item, “Have you had an accident while working. 7” and “Do you think there are any
dangers in your workplace?”
Response Accident while working

Factor 1: Factor 2: Factor 3: Factor 4: Factor 5:


Motivation Positive Risk Fatalism Optimism
for safety safety justification
practice

Yes 19.6 17.3 11.7 16.8 13.4


No 20.4 16.1 12.4 17.1 12.4

F&624, = 1.94 q1.630) = 1o.5 5, ,623) = 1.62 qI.632) = o.4 ‘--(I ,633) = 19.6
n.s. p < 0.001 n.s. n.s. p < 0.001

Response Workplace hazards exist


Factor 1: Factor 2: Factor 3: Factor 4: Factor 5:
Motivation Positive Risk Fatalism Optimism
for safety safety justification
practice
Yes 19.8 17.6 11.7 16.8 13.4
No 20.9 14.8 12.4 17.1 12.4

%599) = 3.73 %59,) = 6.49 51,607) = 1.06 51.608) = 19-ti


p = 0.05 p = 0.011 n.s. p < 0.0001
The deuelopment of a measure of safety climate 23

accidents and those who had not. Factor 2, Positive safety practice, showed that respondents
who had experienced accidents reported poorer safety practice in their workplaces and Factor
5, Optimism, also showed that respondents who had experienced accidents showed less
rationalisation of the risks in their workplace.
The relationship between the two validation questions was not strong. While study
participants who reported no dangers in their workplace were less likely to have experienced
an accident (29.5% accidents compared to 70.5% no accidents), the group who reported
dangers in their workplace were equally likely to have experienced an accident (48.0%
accidents compared to 52.0% no accidents).

2.2.2. Approach 2: Short unidimensional version of scale


In the interests of keeping the scale as brief, but as comprehensive as possible, the 32 item
scale was shortened by dropping items which were highly correlated (correlations > 0.4). In
deciding which of two correlated items was to be kept the better-written or clearest question
was retained. In this way the redundancy of items was reduced. The final questionnaire
contained 17 questions with an acceptable alpha of 0.61 (n = 5881, indicating that some alpha
was sacrificed for the sake of ensuring a workable length of questionnaire but maintaining
coverage of the factors included in the original item pool. The development of the short scale
was done independently of the development of the long scale. Nevertheless, the items retained
in the short scale contained representatives of all of the five factors in the long form,
specifically, 2 from 8 in factor 1, 2 from 6 in factor 2, 3 from 4 in factor 3, 4 from 5 in factor
4 and 4 from 4 in factor 5. Two items from the short scale were not included in the factor
structure of the long version scale.

Validation of the short scale: The final unidimensional scale was validated in two ways.
First, the summated scores from the final 17 item scale were tested against responses to the
item, “Do you think there are any dangers in your workplace?” Using one-way ANOVA, the
summated scores of respondents who answered ‘yes’ were compared to those who answered
‘no’ (see Table 6). The analysis showed that the summated scale score discriminated
differences between these groups of respondents at a 0.06 level of significance (F(,,,,,, = 3.47,
p = 0.06). Second, summated scale scores were compared on responses to the question,
“Have you ever had an accident while working?“. One-way ANOVA showed, again, that
summated scale scores were different between respondents who said ‘Yes’ compared to those
who said ‘No’ to this question at the 0.06 significance level (FC,,633j = 3.44, p = 0.06, see
Table 6).
Subsequent to the development of the scales, the short scale was used in a study to be
reported elsewhere, of safety attitudes and behaviour of employees in a Supermarket. It is

Table 6
Results for the validation of the short version of the scale, showing mean summed scores across all items and
ANOVA results for each validation item, “Have you had an accident while working?” and “Do you think there are
any dangers in your workplace?”
Response Accident while working (mean summed score) Workplace hazards exist (mean summed score)

Yes 51.7 51.8


No 50.6 50.7

%,,a,, = 3.47 F(1.633,= 3.44


p = 0.06 D = 0.06
24 A.M. Williamson et al.

noteworthy that the Cronbach alpha calculated for a sample of the employees in a Supermarket
(n = 71) was 0.60. This replicates the alpha coefficient found for the initial development
sample.

3. Discussion

This study demonstrated the development of two versions of a scale to measure workplace
perceptions and attitudes about safety, where both versions appear to have acceptable internal
consistency. They also appear to have reasonable validity in measuring what they are intended
to measure since the employees who had experienced accidents differed from those who had
not and workers who perceived workplace hazards differed from those with no hazards in their
workplaces. The two versions are aimed at two different needs. The long version provides a
factor structure which describes employee perceptions and attitudes which are likely to
contribute to safety climate and which will allow interpretation of their relative contributions.
The short scale contains a representative group of items from the long version but is of a
suitable length for application in workplaces where lengthy questionnaires are likely to affect
response rates. The extent that the scales are alternatives for each other remains to be seen
through further research.
The five factors generated in the final long solution are a combination of general safety
attitude and perceptions of real workplace conditions. The strongest factor was Personal
motivation for safe behaviour, reflecting the perceived deficiencies in the workplace which
prevent respondents from acting safely. Related to this factor was the risk justification factor
which revealed workplace deficiencies from an alternative viewpoint, that is, the circum-
stances in which unsafe behaviour actually occurred. The Positive safety practice factor is
most like the factors generated from other scales as it reflects perceptions of the workplace
conditions, including the role and commitment of management to safety. The last two factors
reflected aspects of respondents attitudes and beliefs about safety. Fatalism reflects views of
the importance and controllability of safety whereas Optimism reflects the extent that the
individual believes that their level of personal risk is favourable. Unlike the other factors, the
Fatalism factor appears to be resistant to change due to experience as it did not differ between
groups with accident experience or between those with different perceptions of their work-
place. It is possible that this is a more enduring personal characteristic which contributes to
safety climate, but which may not be very amenable to change.
There are some similarities between the five factors identified in this scale and those from
previous studies. The closest factor structure is that of Cox and Cox (1991) with the
effectiveness of arrangements for safety, personal skepticism and personal immunity factors
from their study being very close in flavour to the factors generated in this study identified as
positive safety practice, fatalism and optimism respectively. This provides further support for
the factor structure generated in the current study.
These workplace perception and attitude scales were examined on the extent to which they
detect differences between individual workers in accident experiences and hazard perceptions.
The extent that these workplace perception and attitude scales reflect differences in safety
climate in different work groups, companies or different workplaces with varying exposure to
hazards, however, is not really clear. In the validation of a safety climate scale, the first step
should be to measure against individual experiences and perceptions. As the safety climate
concept is seen as the sum of the perceptions and attitudes of individuals, it is essential that
The development of a measure of safety climate 25

any measure of the components of safety climate can demonstrate a relationship at the
individual level between fairly potent safety experiences such as having an accident, with
safety attitudes. If a relationship cannot be demonstrated, then the measure will be unlikely to
reflect differences between company cultures which may be much more subtle and certainly
more affected by ‘noise’ from other variables. Another aspect of the validation of a company
safety climate scale would then be to demonstrate that it can detect safety attitude and
perception differences between companies where climate differences should exist, for example
high and low hazard workplaces, companies with high and low lost time injury rates or
companies with good and poor overall safety practices. In the development of earlier safety
climate scales, measures were validated against individual accident experiences (Brown and
Holmes, 1986; Donald et al., 1991) and against different industries with different hazards
and/or accident rates (Zohar, 1980; Niskanen, 1994; Donald et al., 1991). Unfortunately in
this study it was not possible to collect the sort of information that could possibly be used to
detect differences in climate, such as accident rates. While this step in the validation process
remains to be done, the scales have demonstrated strengths for assessing individual safety
attitudes and perceptions.
One of the most striking aspects of the development of these safety climate scales is that so
many of the items in the original item pool reflected a high level of consensus among
respondents no matter what company they came from. For around half the item pool,
responses were severely skewed, in all cases in the direction of ‘good’ safety. Most
respondents in most companies had a common answer to a broad range of items, although
some areas of safety seem to be more well-known than others. Specifically the well-known
areas corresponded to the original factors relating to who is responsible for safety, manage-
ment’s role and commitment to safety, the priority of safety in the workplace and the
respondents perceptions about their activity in the workplace. It is notable that most of the
skewed items were attitudinal, while skewed perceptual or reality-based questions were in the
minority.
These findings are intriguing as there are a number of possible explanations. It is possible
that the skewedness reflects general beliefs or stereotypes about safety which are shared by
most of the workforce. The extent that these reflect the study participants’ true beliefs or
attitudes or socially acceptable responses cannot be determined from this study. It is clear,
though, that differences in accident history and perceived exposure to hazards in workplace do
not affect these skewed questions as ratings were the same for the vast majority of all study
participants, no matter which company they came from. An alternative possibility is that the
questions were simply not phrased in a way that would emphasise individual differences.
Further refinement of these questions may produce variation in responses. It should be
remembered, however, that the skewed items were not independent of item type. This suggests
that the skewedness is due to more than the phrasing of the questions. It should also be
remembered that a significant number of skewed items were borrowed from existing question-
naires. Another possibility is that the skewed items reflect shared positive safety attitudes
which have been generated by consistently good conditions in the companies. This possibility
cannot be dismissed entirely in this study as no accident or risk exposure information was
collected systematically. However, the fact that there was such a high level of consistency
across all companies on the clear majority of skewed items, makes it unlikely. It is improbable
that such a strong effect should be due to sampling bias. The companies which participated in
this study came from very different industry sectors and involved exposure to different hazards
and different levels of the same hazard.
26 A.M. Williamson et al.

Two important considerations have become apparent from this study about the concept of
safety climate. The first consideration is that it is essential not to confuse separate approaches
to the concept. Approaches differ on what are assumed to be the components of safety climate.
One approach is based on the assumption that safety climate is generated by the actual
characteristics of the workplace and that these can be discovered by asking workers about their
perceptions of the current status of these characteristics. This approach produces models such
as that of Zohar (19801, and Dedobbeleer and Beland (1991) and is similar to the Performance
Shaping Factors scale being developed by Glendon et al. (1994).
Another approach assumes that safety climate is generated by worker attitudes about safety
in general as well as worker perceptions of the characteristics of the workplace. The item pool
used in this study contained questions about attitudes to safety as well as questions about
perceptions of safety in the respondents workplace. The use of both types of items has the
advantage of providing insight about the respondents orientation towards safety from two
vantage points. The attitudinal questions reveal aspects of respondents beliefs about safety
which are likely to have been developed through experiences inside and outside the work-
place. The perceptual questions also reveal aspects of safety beliefs but which are directed
towards the respondents’ perception of reality in their workplace. In this way the scale covers
safety beliefs in general as well as perceived workplace problems. The factors identified are
therefore broader than those identified in a number of studies (e.g., Zohar, 1980; Brown and
Holmes, 1986; Dedobbeleer and Beland, 1991). As a consequence, they are likely to be much
more informative in comparing individuals, work groups and companies and in looking at any
changes in safety climate. The present study has clearly demonstrated the important discrimi-
natory role of some worker attitudes and beliefs.
The second important consideration about the concept of safety climate, is that it cannot be
defined as a simple summary of the perceptions of workers about their work environments. It
seems that there are safety stereotypes which are shared by most workers which contribute
little to discriminating the particular safety culture in an organisation. Instead it seems that as
far as worker attitudes influence, or are influenced by, the prevailing culture or climate in an
organisation, only some aspects of safety attitude and perception vary across individuals and
organisations. The existence of these apparently consistent safety attitudes deserves further
research attention.

Acknowledgements

We are grateful to the employees of the seven companies for their participation. We are
grateful to Zahra Haji Habibzadeh for her role in data collection.

References

Brown, R.L. and Holmes, H. (1986) The use of a factor-analytic procedure for assessing the validity of an employee
safety climate model. Accidents Analysis and Prevention 18(6), 455-470.
Cox, S. and Cox, T. (1991) The structure of employee attitudes to safety: a European example. Work and Stress 5(2),
93-106.
Donald, I., Cantaer, D. and Chalk, J. (1991) Measuring safety culture and attitudes, In: Proceedings of the Firsf
International Conference on Health, Safety and Enuironment in Oil and Gas Exploration and Production. Society
of Petroleum Engineers, The Hague, 1 l- 14 Novermber.
The development of a measure of safety climate 27

Dedobbeleer, N. and Beland, F. (1991) A safety climate measure for construction sites. Journal @Safety Science 22,
97-103.
DeVellis, R.F. (1991) Scale Development. Theory and Applications. Sage, Newbury Park.
Glendon, AL, Stanton, N.A. and Harrison, D. (1994) Factor analysing a performance shaping concept questionnaire.
In Contemporary Ergonomics 1994: Ergonomics for All, ed. S.A. Robertson, Taylor and Francis, London, pp.
340-345.
Niskanen, T. (1994) Safety climate in the road administration. Safety Science 17, 237-255.
Seppala, A. (1992) Evaluation of safety measures, their improvement and connections to occupational accidents. Cited
in Niskanen, T., 1994. Safety climate in the road administration. Safety Science 17, 237-255.
Zohar, D. (1980) Safety climate in industrial organizations: Theoretical and applied implications. Journal qf Applied
f.y%iology 65(l), 96-102.

You might also like