5.-In Safety We Trust

“In Safety We Trust”
Safety, Risk and Trust in the Offshore

Petroleum Industry
by
Jorunn-Elise Tharaldsen
Thesis submitted in fulfillment of

the requirements for the degree of
PHILOSOPHIAE DOCTOR
(PhD)
Faculty of Social Sciences

University of Stavanger
2011
University of Stavanger
N-4036 Stavanger
NORWAY
© 2011 Jorunn Elise Tharaldsen
ISBN: 978-82-7644-440-7
ISSN: 1890-1387
Acknowledgements
This work is based on four articles having exploration and understanding of
safety and trust within the offshore oil and gas industry as its point of
departure. My way into this research area started with funding from the
Norwegian Research Council back in 2002 (Petromaks) under the umbrella
“HSE-culture”, led by Knut Haukelid at the University of Oslo. The real
work with my PhD thesis, however, did not start before 2004. Looking back, I
feel privileged to have had the opportunity to perform research on this
literally, exciting industrial sector. As such, it has consisted of a range of both
opportunities and challenges – with regards to learning, much hard work and a
constant struggle to find time and place for the thesis between other, ongoing
research projects, and not to mention the balancing between family and work.
At times this has been the most challenging part, and as a side-effect I have
probably become a very good juggler. During this PhD period, I have also got
the opportunity to accomplish two research stays, one at the University of
Aberdeen and the other at the University of Berkeley. These stays and this
thesis would not have been possible to accomplish without funding and
facilitation from Norwegian Research Council, Seawell AS, IRIS and the
University of Stavanger – which I am highly grateful of.
Along the way I have had many good helpers, and first of all I want to thank
my supervisors Knut Haukelid, Karina Aase and Knud Knudsen for their
support and contributions during this PhD-process. Knut Haukelid has in his
inspiring manner shared with me his anthropological insights from his own
research and longtime experience within the oil and gas industry. Karina Aase
has been a continuously inspiring mentor, a thorough reader and very good at
pulling me back when I drifted away. And to Knud Knudsen, thank you for
inspiring discussions and reflections on quantitative statistics, factor analysis
and logistic regression models. In addition, I will also direct a special thanks
to Kathryn Mearns both for your hospitality when including me so generously
in your research group in Aberdeen, and for your reflections and comments on
my work. Later on, this research visit resulted in co-writing on the fourth
article and also collaboration in other research projects. Thank you all for
your willingness to share your time, experience and knowledge with me, and
for your reading and comments on this thesis.
I also want to thank all co-authors for their contributions. Espen, we started
both yours and my PhD work with the first article and working with you is
always inspiring. You also introduced me to Torbjørn Rundmo, who with his
strong engagement for safety research guided us through our early factor
iii
analyses in the longitudinal study, which prepared me for further work with
this method later on. During the first years of the HSE-culture research
program Knut Haukelid initiated several national workshops and seminars
which gave safety researchers, me included, a good overview of the research
field. In this network I also met Dordi Høivik, which later on resulted in
collaboration and co-writing with her and her fellow researchers at the
University of Bergen – Valborg Baste and Bente Elisabeth Moen. From the
HSE-culture network there are also a handful of other researchers I would like
to thank for inspiring discussions and collaboration: Gunnar Lamvik, Ragnar
Rosness, Jakob Kringen and Tor Olav Nævestad. I also need to mention
members of a Nordic network of safety climate researchers in which we
worked on the development of a safety climate tool: Thank you Pete, Jorma,
Kim, Anders and Marianne for engaged discussions on factor analysis and
long-lasting meetings at safety workshops and conferences. And not to forget;
my fellow co-workers at IRIS! Several of you are or have been in the same
situation as me and we all know something about time priorities. Thank you
all for being understanding and great supporters! A special thanks to Kari
Anne for reading some of my work and for your company at concerts – maybe
now is the time to start the music club we’ve talked about?
In addition to providing financial funding, I would also like to thank Seawell

AS for opening their doors and spending valuable time and information with
me and our research team. In this respect I will especially mention Alf Ragnar
Løvdal, Christian Cappelen-Smith, Harald Bratthammar, Kenny Dey and
Gunn Ekenberg Søndervik for helping me out. I also would like to thank
Øyvind Lauridsen and Inger Danielsen from the Petroleum Safety Authorities
and their helpful attitude, providing us with all the data and statistics we
needed in the first study. Without the PSA’s funding and establishment of an
industry level measurement of safety climate – we wouldn’t have had this
quantitative database available for research.
And finally, to the greatest supporters of all, my family! Thank you, Atle, for
your patience, and for keeping up with this long-lasting finishing stage. Jakob
and Frida – thank you for just being such lovely children and for your
amazing capacity to bring me back to basic! I am really looking forward to
spending more time with you all!
Stavanger, August 2010
Jorunn Elise Tharaldsen
iv
Scientific Environments – Co-authors
 Jorunn Elise Tharaldsen, University of Stavanger, Faculty of

Social Sciences, Risk Management and Societal Safety, Norway
and International Research Institute Stavanger, Norway.
 Dordi Høivik, Bente E. Moen, Valborg Baste: University of

Bergen, Department of Public Health and Primary Health Care,
research group for Occupational and Environmental Medicine,
Norway.
 Espen Olsen, University of Stavanger, Faculty of Social Sciences,

Risk Management and Societal Safety, Norway.
 Kathryn Mearns, University of Aberdeen, School of Psychology,

Industrial Psychology Research Centre, Scotland, United
Kingdom.
 Knud Knudsen, University of Stavanger, Faculty of Social

Sciences, Change Management, Norway.
 Knut Haukelid, University of Oslo, Centre for Technology,

innovation and Culture.
 Torbjørn Rundmo, Norwegian University of Science and

Technology (NTNU), Faculty of Social Sciences and Technology
Management, Department of psychology, Trondheim, Norway.
v
Summary
This thesis draws the attention towards offshore work environments and
their safety challenges. It contributes both to explanations of safety
climate dynamics and to an understanding and framing of this specific
industrial culture. Five research problems are examined: I) How are valid
and reliable safety climate dimensions established? II) What are important
factors explaining the distribution of safety climate perceptions? III)
Where do cultural or behavior based approaches to safety lead us? IV)
What are the links between trust, distrust and safety, and V) Do safety,
trust and cultural aspects differ between offshore employees on the
Norwegian and UK Continental shelves? These problems are further
specified through their own research questions and examined in four
journal articles. Mainly, the research questions may be divided in two
types: those searching for explanations, and those promoting
understanding. As such, the thesis involves a combined methods approach
with the use of both quantitative and qualitative methods. Research
questions searching for explanations deal with the establishment of valid
safety climate concepts, safety performance indicators, how safety climate
and performance are linked and distributed on the Norwegian and UK
shelves, and examination of how safety and other organizational aspects
may be linked to each other statistically. Research questions aimed at
understanding look at distinctions between cultural and behavioral
approaches to safety, the functions and dysfunctions of safety specific
trust, and the overall framing of cultural differences and characteristics of
the industrial offshore culture.
The thesis is divided into two main parts. Part I includes the overall
framework embracing all studies and has eight chapters, while Part II lists
the four journal articles. In Part I, chapter one introduces the main research
problems, aims and research questions for the thesis, while the second
chapter gives a description of the industrial context in which the studies
are situated. Thereafter theoretical framing and clarifications are given in
chapter three, followed by the methodological approach used involving
research designs and the introduction of a conceptual work model in the
fourth chapter. The main results are briefly summarized in chapter five and
vi
further elaborated in chapter six. Chapter seven discusses important
limitations and implications, while conclusions are given in chapter eight.
The thesis draws upon theory from different disciplines – mainly

sociology, social anthropology and psychology. In this thesis culture is
treated as the superior concept and safety as an aspect of it. A “safe”
culture will reflect common symbols, sense-making and meaning systems
distinguished by robust and heedful practices – enabling groups,
organizations or societies to protect themselves, the environment and
others from harm. As such, I make a clear distinction between culture and
climate, reserving the latter for aggregated perceptions or attitudes of a
group of people, and the first to deal with sense-making, construction of
meaning, and tacit, taken for granted beliefs about our “world”.
The first and second articles involve testing and development of the
NORSCI questionnaire tool, resulting finally in six safety climate
dimensions. Sound fit measures and reliabilities during test-retests of our
models indicate that the content validity of the instrument is good. Our
factors behave consistently when tested for both in the first and second
study and belonging items are well anchored both within the safety
research field and within the industrial sector.
The first and fourth papers deal with three different kinds of safety
performance indicators: accident rates, risk perception and incident
involvement. All in all they show a reasonable relationship with our safety
climate dimension. However, some of the aggregated level correlations in
the first study are weak, and Safety prioritization shows an unexpected
positive relation with both risk perception and accident rates. Qualitative
data also suggest that public statistics may be associated with bias’
problems both within the industry (strong incentives of maintaining good
safety statistics), and across shelves due to differing payment regulations
when having to stay off from work because of an injury. As a consequence
of the results from these studies and experiences from other studies,
recommendations are made in the direction of a self-report measure either
as a substitute or as a supplement to public accident statistics in self-
completion questionnaires. An anonymous self-report of involvement in
accidents or near misses may be experienced as less threatening than other
company reporting systems, which more often become “public” and may
contribute to “destruction” of company statistics.
vii
Installation membership has been shown to be one of the most important
characteristics framing employees’ safety perceptions and offshore work
life. Our qualitative findings suggest that employees develop identity and
affective based trust towards the installations and their colleagues, where
identification and belonging are wrapped in family metaphors. Other
explanatory factors are related to shelf differences, type of company, work
area, type of installation, and work factors related to flexibility, stability
and rotation patterns. Hence, the platforms may function as important
culture creating “places”. A possible explanation and reason for such
identification may be related to common work characteristics: work is
accomplished on remote worksites, management commitment to safety is
exercised locally, and employees share approximately the same goals,
common dangers and high risk scenarios. Such creation of local identities
and subcultures may enable groups to control their work situation, to
reduce uncertainty, maximize autonomy and make fertile ground for
functional and identity based trust relations offshore.
One of the main aims in paper four is to examine the safety exposure of
contractor employees working with drilling and wireline services across
the UK and Norwegian Continental Shelves. These contractors are known
to have a more nomadic and unpredictable work life and one of my main
anticipations was that such unstable work conditions would give higher
accident exposure. A surprising finding, though, is that nomads and
employees with unpredictable shift rotations show less exposure to
incident involvement compared to more stable personnel. A possible
explanation for this lower exposure may be that such shifting work
demands may lead to workers becoming a combination of more flexible
and careful navigators. Qualitative findings also suggest that these
contractor workers seem to experience a high degree of “soft” clan control
and institutionalized control from client companies. Hence, a nomadic,
contractor position may enhance both personal and organizational
alertness.
A model envisaging dynamics between functional and dysfunctional trust

and distrust with regards to safety is introduced in the final chapter. In the
comparative study I find that a division of trust and distrust into functional
and dysfunctional aspects makes intuitive sense, and that trust and distrust
with regards to safety are understood in similar ways across the shelves.
Trust with an element of control is spoken of as good for safety, while too
high trust is considered risky. Functional distrust is further anticipated to
viii
be crucial for safety and related to a well-developed “creative imagination
– a kind of “what if” ability, while dysfunctional distrust is always
considered negative. Both trust and distrust seem to be used in parallel to
minimize uncertainty, reduce risk, to promote (social) order and
predictability with regards to safety.
Both the level and form of trust are found to differ across shelves. The UK
sample shows, as expected, higher Assertiveness and Power Distance
compared to the Norwegian sample. Taken together with higher Safety
compliance and higher Trust in management commitment to safety than
Trust in colleagues, the UK culture and climate are considered to be
distinguished by a more rule based trust, whilst being more heavily based
on equality and a common identity among Norwegian workers and their
managers. None of these characteristics are found to have a clear effect on
safety performance.
The epistemological roots of cultural and behavioral perspectives are

different, as is their methodology. A cultural approach often has a holistic
and inter-subjective point of departure and makes mainly use of qualitative
methods. Behavior based safety (BBS) approaches are rooted in behavioral
psychology and makes use of stimuli – response models and principles of
operant conditioning and reinforcement. A BBS approach will look for
problematic behavior and will mainly be concerned with observable
behavioral outputs. A two-by-two table focusing on culture-behavior and
tacit-explicit aspects pinpoints the implications of differing safety
intervention approaches. I argue for a balanced and multi-method
approach taking into consideration how formal (canonical) practice is
related to the informal (non-canonical) sides, and how underlying, not-so-
easily observable or external factor may influence workers’ safety.
The industrial culture seems to be geared towards unity, shared norms,

rules and procedures as the most valued organizational properties, even if
reality seems to be characterized by subcultures and safety as differently
distributed. Hence, both safety practitioners and safety research ought to
develop better knowledge of how work related safety is distributed and
what consequences such distribution may have. If not, safety management
systems may conflict with informal work practices in unintended ways,
and well intended ambitions may lead to unforeseen paradoxes.
ix
Contents
Acknowledgements .............................................................................. iii
Scientific Environments – Co-authors................................................ v
Summary .............................................................................................. vi
Contents................................................................................................ xi
Part I .................................................................................................... 15
1 Introduction ............................................................................ 15

1.1 The role of trust with regards to safety..................................... 16
1.2 Main research problems and research questions ...................... 16
1.3 Aims of the articles................................................................... 17
2 Industrial context ................................................................... 21

2.1 Regulatory regime and industrial relations .............................. 24
2.2 The Trends in Risk Level Project ............................................. 25
3 Theoretical framing and conceptual clarifications ............. 29

3.1 The concept of culture .............................................................. 29
3.2 Culture and behavioral approaches to safety............................ 30
3.3 Safety culture and safety climate ............................................. 32
3.4 Safety climate and indicators of safety performance ............... 35
3.5 Trust and distrust with regards to safety .................................. 37
3.6 Cross-cultural comparison........................................................ 40
3.7 Summary – theoretical framework ........................................... 42
4 Methodology ........................................................................... 45

4.1 Research design ........................................................................ 45
4.2 Samples and data collection ..................................................... 46
4.3 Statistical methods, tools and techniques ................................. 49
4.4 Tools and measures .................................................................. 53
xi
4.5 Ethical issues ............................................................................ 55
4.6 Philosophy of science considerations ....................................... 56
4.7 Conceptual work model ........................................................... 59
5 Results ..................................................................................... 61

5.1 Summary of article I ................................................................. 61
5.2 Summary of article II ............................................................... 62
5.3 Summary of article III .............................................................. 62
5.4 Summary of article IV .............................................................. 63
6 Discussion ................................................................................ 67

6.1 Instrument qualities - psychometrics ........................................ 67
6.2 Safety climate and safety performance .................................... 69
6.3 Differentiating patterns............................................................. 72
6.4 Influence of culture, trust and distrust on safety ...................... 73
6.5 Industrial culture....................................................................... 77
7 Limitations and implications ................................................. 81

7.1 Common method variance and potential bias’ ......................... 81
7.2 Sample sizes ............................................................................. 83
7.3 Cross-cultural comparison........................................................ 85
7.4 Safety performance indicators .................................................. 87
7.5 Dynamics between safety, trust and distrust ............................ 88
7.6 Implications for safety practitioners ......................................... 90
8 Conclusions ............................................................................. 93
References ........................................................................................... 95
xii
Part II ................................................................................................ 111
List of Articles................................................................................... 111
Article I.............................................................................................. 113
Article II ............................................................................................ 115
Article III .......................................................................................... 117
Article IV ........................................................................................... 119
xiii
Introduction
Part I
1 Introduction
Sectors which are exposed to high risk scenarios and accidents constitute a
matter of societal, industrial, company and human concern. An often
outspoken goal in such industries is to build strong “safety cultures” in
order to protect the employees, the environment or technology from harm.
Safety culture is a frequently used lay concept and as a research concept it
has gained attention since its first appearance in the International Nuclear
Safety Advisory Group’s (INSAG) report on the Chernobyl accident
(INSAG, 1986). It has been defined as:
That assembly of characteristics and attitudes in organizations and

individuals which establishes that, as an overriding priority, nuclear
installation safety issues receives the attention warranted by their
significance (INSAG 4, 1991, p.1).
Since then the hunt for an all exhaustive definition or instrument able to
define or measure “safety culture” or “safety climate” started, and is still
running. Many different safety climate instruments exist, often referred to
as safety climate tools. Which one should be used or how should they be
developed? Are we looking for a once and for all establishment of “The
Big Five” (Barrick and Mount, 1991), “The Big Six” (Flin, Mearns,
O’Conor and Bryden, 2000) or are safety measurement instruments
dependent on contextual frames? The goodness of a tool will be dependent
on its construct and discriminant validity and the Dov Zohar's (Zohar,
1980, 2003, 2008) instruments have until now shown to be the best
predictors of safety performance. So why should we bother making new
ones? Also; is safety to be considered a universal value (Helmreich &
Merrit, 1998) or dependent on culture (Douglas & Wildavsky, 1983)? Are
climate and culture stable matters or changeable? Finally, how do we deal
with cultural approaches to safety when others talk of Behavior Based
Safety (BBS) as the answer? What are main differences or similarities
between them and where do behavioral approaches to safety leave us?
These questions will be addressed through my main aims and research
questions in four articles.
15
Introduction
1.1 The role of trust with regards to safety

Within safety research trust has been regarded as an important element of
a sound safety culture (Reason, 1997). When developing an understanding
of phenomena such as safety, risk and trust, one discovers that they are
interrelated – often implying a sense of social predictability, social order
or lack thereof. Recently within safety research the downsides to trust and
the benefits of distrust have received increasing attention (see Risk
Analysis, 2006, Vol. 26, No. 5). Trusting someone or something may be
inherently risky, trusting blindly is naïve and dangerous, and diverse,
complex systems are supposed to need aspects of both trust and distrust in
order to stay functional (Luhmann, 1979; Weick, 1987). The question
whether the form and level of trust differs across cultures has also been
raised (Zaheer & Zaheer, 2006), leaving us with new challenges when
performing cross-cultural research.
1.2 Main research problems and research

questions
The thesis operates with four main research problems each of them
expressed in specific research questions:
 Establishing valid and reliable safety climate dimensions:
 What theoretical concepts best explain safety performance and

how are they developed?
 How do these concepts relate to each other and to accident

exposure?
 What makes a sound indicator of safety performance?
 The distributed character of Safety climate:
 What underlying dynamics contribute to explanations of

safety climate perceptions and to the understanding of
(safety) culture?
 How are safety climate, risk perception and accident exposure

differentiated?
16
Introduction
 What factors could explain such a distribution and what

consequences may they have?
 Culture and behavioral perspectives on safety:
 What are main roots of cultural and behavioral perspectives on

safety?
 How do these approaches differ and/or how may they be said

to be similar?
 Where do differing approaches lead us and what kind of

consequences do various perspectives have on how we
understand our safety problems and how to solve them?
 The links between trust, distrust and safety:
 What are the functions or dysfunctions of trust and distrust

with regards to safety?
 What kind of impact may trust at different levels have on

accident involvement?
 Cross cultural comparison of offshore safety:
 Do safety climate perceptions differ across the UK and

Norwegian Continental shelves?
 How may cultural differences influence trust relations,

leadership practices and safety aspects?
1.3 Aims of the articles

The thesis involves four separate studies with their own main aims. Below
all aims are listed for each study.
1. A longitudinal study of safety climate on the Norwegian Continental

Shelf:
17
Introduction
The first study uses survey data covering all offshore personnel at two
moments in time; 2001 and 2003. The aims of the study are to:
 test and retest the psychometric qualities of a questionnaire

survey according to a model developed by Cooper (2000)
 examine longitudinal development of the safety climate

concepts measured in the instrument (from 2001 to 2003)
 investigate the hypothesized association patterns between the

safety climate concepts and perception of risk and accident
rates
 examine the differentiating capacities of our safety climate

concepts
2. What is most important for safety climate: The company belonging or

the local working environment? – A study from the Norwegian offshore
industry:
This second article makes use of survey data from the same questionnaire
tool as the first study – collected two years later (2005). The aims of this
study are to:
 retest the safety climate model developed in the former study
 examine the relative influence of offshore installation and

company membership on employees’ perceptions of the safety
climate concepts
3. Culture and behavioral perspectives on safety – towards a balanced

approach:
The third study aims at examining differences and similarities between

culture and behavior based perspectives on safety. In order to do so, an
examination of the main historical roots of these approaches, main
concepts used, their methods and ‘techniques’ with regards to safety
improvements, are made. A two-by-two table is developed to highlight
main differences between a cultural and behavioral approach to safety
along two dimensions: Culture-Behavior and Tacit-Explicit.
18
Introduction
4. Perspectives on safety: the impact of group membership, work factors

and trust on safety performance in UK and Norwegian drilling
employees:
The fourth paper involves both quantitative survey data and qualitative
interviews gathered on the UK and Norwegian shelves. An overall aim is
to examine work and safety conditions for contractor employees having a
more nomadic and unpredictable work life than other offshore workers. It
operates with four main hypotheses:
 H1 - involves an anticipated impact of installation and Shelf

membership on Safety Performance (incident involvement). However,
the effect of Shelf should mainly work indirectly, or being moderated
by installation membership, group identity (Team/work area), trust and
self-reported safety behavior.
 H2 – Employees who tend to move between installations (“nomads”)

or who work more irregular rotations will to a greater extent have to
adjust themselves to new surroundings, safety systems, working
cultures and sleeping patterns. We expect such continuous adjustment
to increase exposure to accidents.
 H3 – Trust in workmates and management commitment to safety will

have a significant positive impact on safety performance.
 H4 – Safety Compliance and Safety Participation will have a

significant positive impact on safety performance.
19
Introduction
Below an overview of all articles are given.
I Tharaldsen, J., Olsen, E., Rundmo, T., 2008: A longitudinal study of safety
climate on the Norwegian Continental Shelf. Safety Science, vol. 46, no. 3,
427-439.
II Høivik, D., Tharaldsen, J.E., Baste, V., Moen, B.E., 2009. What is most
important for safety climate: The company belonging or the local
environment? A study from the Norwegian offshore industry. Safety
Science,Vol. 47, no. 10, 1324-1331.
III Tharaldsen, J.E., Haukelid, K., 2009. Cultural and behavioural approaches to
safety – towards a balanced approach. Journal of Risk Research, vol. 12, No.
3-4, April-June 2009, 1-14.
IV Tharaldsen, J.E., Mearns, K.J., Knudsen, K., 2010. Perspectives on safety: the
impact of group membership, work factors and trust on safety performance in
UK and Norwegian drilling company employees. Safety Science, 48, 1062-
1072.
Table 1 Overview all articles
20
Industrial context
2 Industrial context
Organizations situated in the offshore oil and gas industry face huge
challenges with regards to safety. As such they share characteristics with
other organizations facing the risk of major accidents named High
Reliability Organizations (HRO) (Weick, 1987; LaPorte & Consolini,
1991) or characterized as systems in which interactions are complex and
couplings tight (Perrow, 1999). Such organizations operate technologies
that are highly beneficial, yet costly and hazardous (LaPorte & Consolini,
1991:19). Despite high risk exposure and organizational complexity, the
capacity of error free performance in these domains is nevertheless
impressive.
The oil and gas industry plays an important role in the Norwegian
economy. In 2007 the sector provided approximately 31% of the
Norwegian state cash flow (KonKraft-report 7, 2009). Due to the
Norwegian tax regime 90% of the industry profits fall into the public purse
– providing Norwegian citizens with hospitals, roads etc. As such, the
industry constitutes an important role in the building and maintenance of
the Norwegian welfare state. However, Norway reached its “peak oil”
status in 2004, and since then production is pointing downwards. The total
state cash flow is estimated to have decreased from 31% in 2007 down to
1.5% in 2060 (White paper no 9, 2008-2009). In addition to tax income,
the spin-off effects are huge; the supply chain includes a workforce of
100 000 employees and just below 150 000 employees in total are directly
involved in the oil and gas industry in Norway (KonKraft-report 7, 2009).
Approximately 30000 of these people work offshore.
It is stated from the Norwegian government that the oil and gas industry
shall be world-leading on health and safety (White paper no. 7, 2001-
2002). In its approximately 50 year long history the Norwegian oil and gas
industry has undergone major technical, organizational, economic and
social changes. Accordingly it has been divided into the following phases:
Texas (from 1966), The Great Change (80s), The Great Systems (90s), and
The Cultural Solution (Haukelid, 2008). The present period is
characterized by a strong emphasis on safety culture as a key to success. In
the early phase of oil drilling on the Norwegian Continental Shelf (NCS)
the industry was dominated by American culture, technology (rigs and
drilling equipment were American), expertise and language (Smith-
21
Industrial context
Solbakken, 1997). Accident rates in this phase were high and the workers
identified with a macho working culture accepting accidents and injuries
as a part of their daily work life. This phase culminated with the capsizing
of the accommodation platform Alexander Kielland in March 1980. The
accident killed 123 workers and led the NCS into the “Great Change”. The
Piper Alpha fire and explosion marked in similar ways the UKCS situation
in July 1988, killing 167 men. Public claims of safety improvements arised
and changes were made – resulting in higher focus on management
commitment to safety, new directives with requirements for education,
training, safety courses and other safety measures being issued on both
shelves. From the end of the 80s the big oil companies also introduced
comprehensive safety management systems like the International Safety
Rating System (ISRS), Tripod, accident reporting systems and new
technology – involving less manual work, higher focus on safety reporting
and decreased personnel risk. The consequences were manifested in a
lowering of the number of accidents in the industry, as illustrated below:
Period 1967-1979 1980's 1990's 2000-2009 Total

Construction and marine syst 6 134 0 1 141
Helicopter 34 0 15 0 49
Falls 25 12 6 1 44
Diving 10 6 0 0 16
Lifting 3 0 2 3 8
Work accidents vessels 1 1 3 3 7
Fire and explosion 5 1 0 0 6
Drilling operations 0 1 3 0 4
Poisoning 3 1 0 0 4
Other 1 1 1 1 4
Total fatal accidents 92 156 30 9 286
Table 2 Types and numbers of industrial accidents in the oil and gas industry on
the Norwegian Continental Shelf in the period 1956-20091
For the total period (1967-2009) there are 286 deaths, and approximately
half of the deaths occurred in the Aleksander Kielland accident. If we
1
The table shows total number of accidents and is not corrected for levels of
employment or activity.
22
Industrial context
subtract the Aleksander Kielland deaths from the 1980s, we find that there
have been 214 fatal accidents from the start until 1980 (75%) and 72
deaths from 1981 until today (25%), which illustrates the above mentioned
“Great Change”.
Daily work tasks on oil and gas facilities and their coordination are highly
regulated by procedural systems. The industry is also characterized by
complex client-contractor chains, organized as strategic non-equity
alliances (Das & Teng, 1998), where the largest portion of employees
work in contractor companies within frameworks heavily defined by their
customers. Contractors often find themselves at the sharp end or closer to
the sources of risk with regards to several aspects; e.g. having more
unpredictable working hours, high production demands, physical strain
and a more “nomadic” work life (moving between platforms, working for
different customers). Such work conditions imply constant adjustment to
various clients and local platform cultures (Tharaldsen & Haukelid, 2007).
Working conditions offshore are generally known to involve challenges

with regards to physical work conditions, psychosocial stress and health
risks (Lauridsen & Tønnesen, 1990; Eide, Brandsdal & Bjorseth, 1992;
Rundmo, 1992; Parkes, 1998; Mearns, Whitaker & Flin, 2001; Morken,
Bratveit & Moen, 2005), but there is a lack of knowledge on if and what
kind of exposure structural work factors may give for offshore employees’
safety performance. Work environments are characterized by harsh climate
and weather conditions, complex technology, challenging shift rotations,
chemical exposure, noise, risks of injuries and major accidents. The work
depends on cooperation and communication both within and across
organizational borders and between actors on- and offshore. Hence,
interdependence between the actors is anticipated to be crucial, and the
organizations face a constant vulnerability to human error. If we add a
long chain of client (operating companies), contractor and subcontractor
companies to this picture, we catch a glimpse of a work life characterized
by complex client-supplier relations and multisided organizational
interfaces. One of the main anticipations in this thesis is that risk and
safety are distributed along such chains and interfaces. Thus, the challenge
will be to trace fruitful paths in order to examine such dynamics in
meaningful ways.
23
Industrial context
2.1 Regulatory regime and industrial relations

During the last 25 years, the regulatory regime of the offshore oil and gas
industry has changed its approach from a “counterparty” to a “co-player”
culture. In principle there are two broad approaches to industry regulation,
(1) a compliance based regime and (2) a self-regulatory regime. The latter
is characteristic for the Norwegian petroleum sector and implies a goal-
setting regime in which legislation is less detailed and more general –
often in the form of functional requirements through legal standards. The
goal has been to move away from a “compliance mentality” – with a
passive duty-holder role – to an active and flexible approach in which the
duty-holder’s responsibility becomes more explicit (Logstein, 2009). The
concept is coined enforced self-regulation (internal control in the
Norwegian context) and means that the regulatory process is delegated to
stakeholders (license holders), but under conditions given by the
regulating authority (Kringen, 200; Lindøe & Olsen, 2009). This means a
regulator role which has to balance delicately between a reactive,
controlling inspector and a dialogue oriented facilitator towards the
industry and the unions (Bruhn, 2006; from Braut & Lindøe, 2009). Such
use of functional requirements and legal standards require a high level of
trust between regulators and the actors being regulated. Presently, the
overall picture on the NCS is changing from the involvement of 14-15
large, highly competent oil companies to a mosaic diversity of 70
companies, many of the newcomers being smaller and anticipated to be
less competent and experienced with the Norwegian oil and gas industry
(ibid.).
Another principle characterizing the Norwegian work life in general, and

the oil and gas industry in particular, is the tripartite cooperation between
regulator, enterprises and unions – often referred to as the “Nordic model”
(Lindøe & Karlsen, 2006). The occupational health and safety
organization is soundly anchored within the Norwegian Working
Environment Act (1977, revised 2006) which instructs all companies of a
certain size to elect safety representatives and establish a safety committee
where employer and employees are equally represented. Ensuring rights of
worker participation is stated in national industrial health and safety
legislation and in the EU Framework Directive 89/391 (ibid.). Trade union
influence has a long history and is anticipated to play an important role
within the Nordic model, and might be characterized as especially strong
within the petroleum sector. There are several examples of revolts and
24
Industrial context
strikes among Norwegian unions in the early phases of the industry,

attacking an attitude and practice of “anti-unionism” from global oil
operating companies. One finds for instance several examples and
instances where Anglo-American management styles clashed with
Norwegian workers’ expectations of involvement and co-determination
(Kringen, 2009).
During the nineties extensive processes of change took place in the

industry, probably with an insufficient focus on the consequences for
health, safety and environment. There were huge disagreements between
the industrial actors; union representatives argued that the safety level was
declining because of downsizing and intensive cost reductions, while
company representatives claimed that it had never been better. In the wake
of heated arguments, a new cooperative climate emerges and several
tripartite collaborative arenas are established, like the Safety Forum and
Working Together for Safety. Together with a revival of the tripartite
collaboration, a new legal framework was developed introducing the term
HSE-culture, together with a White Paper (no 7, 2001-2002). The
development of a sound HSE-culture was in the White Paper being closely
linked to the role of management and asserts that what management gives
systematic attention and priority to become the prevailing HSE culture
(Tharaldsen & Lindeberg, 2004). The introduction of the HSE-culture
concept led to a lot of activities within the industry, and large “HSE-
culture programs” were being implemented by the companies involving
both clients and contractors (Olsen, 2009). Simultaneously, a research
program was started and carried out by the Norwegian Research Council
with the concept of HSE-culture as one of the main areas of interests. As
part of this revived dialogue and move towards consent, the Norwegian
Petroleum Directorate initiates the Trends in Risk Level Project (2000) in
order to monitor the overall level of risk and safety within the Norwegian
oil and gas sector. Since an organizational restructuring in 2004 the TRL
has been pursued by the Norwegian Petroleum Safety Authorities, and like
the other above mentioned arenas or initiatives, this project has also been
anchored within the framework of the Nordic model.
2.2 The Trends in Risk Level Project

The TRL project aims at (1) Measuring the effect of HSE work in the
sector, (2) Contributing to identifying areas which are critical for HSE,
and (3) Increasing insights into potential causes of accidents and
25
Industrial context
undesirable conditions and their significance for the risk scenario (see
www.ptil.no/rnns). The project is unique and involves the possibility of
measuring the effect of the collective HSE work in the industry. It keeps
under surveillance the overall risk level and occupational health and safety
both offshore and at petroleum land facilities.2 TRL stresses
methodological triangulation and collects a variety of data; e.g. leading
and lagging incident indicators, qualitative and quantitative data etc., and
results are published in annual reports. Of special interest for this thesis is
the major survey being accomplished every second year since 2000
involving all offshore personnel. Both survey data and other indicators are
in general being handled by researchers and HSE experts. The database
established in TRL represents a treasure trove for the industry and its
actors, but also for various research environments, both nationally and
internationally.
The most important risk indicators are the 21 so called DFUs (major
hazards, accidents and near misses, “Definerte Fare- og
Ulykkessituasjoner” in Norwegian); e.g. ignited and non-ignited
hydrocarbon leaks, loss of well control, fire/explosions, drifting objects,
vessels on collision course, leaks or damage to subsea equipment,
helicopter crash, man overboard, diving accidents, H2S emissions,
occupational illness, personal injuries etc. These indicators are either
collected by the industry itself, the PSA or in cooperation between the
industry and the authorities. Especially targeted areas in TRL have been
towards the reduction of hydrocarbon leaks and increased well control,
both projects showing a positive development.
Below the development of serious injuries3 to personnel on production

installations and mobile units are shown for the period 1998-2008 (TRL
Report 2008). As can be seen from the figures below, serious injuries
show a peak in 2000/2001 on both installation types with a decreasing
trend until 2008. Exposure on mobile units seems in general to be higher
2
Onshore facilities are included from 2006.
3
Serious injury includes death, absence continuing into the next shift and medical
treatment.
26
Industrial context
than on production installations with 1.4 against 0.7 serious injuries per
million man-hours, respectively.
Figure 1 Serious injuries to personnel per million manhours, production installations

(TRL report, 2008)
Figure 2 Serious injuries to personnel per million man-hours, mobile units (TRL
report, 2009)
The latest TRL report (2010) shows that serious injuries per million man
hours for the total shelf is further decreasing from the year 2008 to 2009.
However, production and mobile installations show opposite tendencies
with production platforms having a small increase in accidents per mill
man-hours from 0.7 in 2008 up to 0.9 in 2009. Mobile units, on the other
hand, show a marked improvement from 1.42 in 2008 down to 0.55 in
2009. When comparing serious personal injuries per million man-hours
27
Industrial context
reported to the authorities on the UK Continental Shelf (UKCS) and the

NCS in the period 2001 to 2009, the rates are approximately the same with
0.95 in Norway and 1.01 in the UK. However, when looking at the fatal
accidents the picture becomes somewhat different, showing a significantly
higher exposure on the UKCS compared to the NCS with 2.7 fatalities per
million man-hours in UK against 1.5 per million man-hours in Norway.
28
Theoretical framing and conceptual clarifications
3 Theoretical framing and conceptual

clarifications
Safety research constitutes a cross-disciplinary field and as such this thesis
draws upon theories and concepts from different disciplines; mainly
sociology, anthropology and psychology. What do we mean by culture?
How is the culture concept related to organizational culture, safety culture
and safety climate? The thesis also deals with exploration and validation
of safety climate concepts. How may safety climate be defined and what
are the dynamics between safety climate, other organizational concepts,
trust and safety performance? What role may trust and distrust play with
regards to safety? And how do we deal with cross-cultural comparisons
with regards to safety?
3.1 The concept of culture

Before we head for the safety culture and safety climate concept
comparison, we will have to take a detour and have a look at the concept
of culture. The culture concept has its roots in sociology and anthropology
and there exist numerous definitions of it. It is a concept difficult to grasp
and “a tricky concept as it is easily used to cover everything and nothing”
(Alvesson, 2002:3). There is no fixed or agreed meaning or definition of
culture, even in anthropology (ibid.). Most of them are related to some
kind of sharedness with regards to ways of thinking, feeling and reacting,
norms and rules, meaning construction or common practices (Geertz,
1973; Alvesson, 2002).
According to Geertz (1973) culture can be viewed as an ordered system of

meaning and symbols in which social interaction takes place, while the
social system can be seen as the pattern of interaction itself. Culture then
becomes the fabric of meaning by which human beings interpret their
experience and which guides their actions (ibid.). In this thesis the concept
of culture will be related to humans’ common, shared sense-making, story-
telling, construction of meaning, collective identity shaping and
interpretation processes:
Talking about organizational culture seems to mean talking about the

importance for people of symbolism – of rituals, myths, stories and legends –
and about the interpretation of events, ideas, and experiences that are
29
influenced and shaped by the groups within which they live (Frost et al.,
1985:17, from Alvesson, 2002:3).
Culture is then closely connected to interpretation processes, systems of

common symbols and the meaning people make of such symbols. If we
move such an understanding towards the field of safety, then a “safe”
culture would reflect common symbols, sense-making and meaning
systems which are distinguished by robust and heedful practices –
enabling groups, organizations or societies to protect themselves, the
environment and others from harm. However, different groups within a
society or an organization may to a varying degree be able to protect
themselves from harm. As such we may argue that safety constitute a
resource that is socially and structurally distributed. In similar ways,
Keesing (1987) discusses culture as phenomena involving processes of
power:
I will suggest that views of cultures as collective phenomena, of symbols and

meanings as public and shared, need to be qualified by a view of knowledge
as distributed and controlled. […] Even in classless societies, cultural
ideologies empower some, subordinate others, extract labor of some for the
benefit of those whose interests the ideologies serve and legitimate. Cultures
are webs of mystification as well as signification (Keesing, 1987:161-162)
So when talking about culture one should also take into consideration
processes of power; i.e. what values, norms, ideas and knowledge that
prevail in a group, who(m) they serve and for what reasons, is not to be
regarded as neutral matter: “Few do the actual spinning while the …
majority is simply caught” (Scholte, 1984:140, from Keesing, 1987:162).
And with regards to safety an important question to ask is who gets
exposed to what?
3.2 Culture and behavioral approaches to safety

In the field of safety research there is an on-going debate about the
difference between behavioral and cultural approaches to safety and how
to perform effective interventions in order to improve safety related
aspects in an organization. This debate is also echoed in the Norwegian
petroleum industry and tends to result in trench-war-like positioning of the
approaches as opposites in theory and as excluding each other. However in
practice, we often tend to use insights from both perspectives, but an
30
exploration of their distinctions is needed and might inform both theory

and practice.
The main principles of behavior based safety can be traced back to early
behavioral theory starting with controlled laboratory experiments on
animals. It is associated with Pavlov in Russia (classical conditioning) and
Thorndike, Watson and Skinner in USA (operant conditioning). It makes
use of stimuli – response models and principles of operant conditioning
and reinforcement. Behavior based safety has also much in common with
safety management perspectives which often make use of goal theory
(Krause, Seymour & Sloat, 1999; Depasquale & Geller, 1999; Geller,
2001; Johnson, 2003; Cox, Jones & Rycraft, 2004; Seo, 2005). A
distinction is often made between behavior modification and applied
behavior analysis, reserving the last term for applications in natural
settings, such as work places (DeJoy, 2005).
Unlike behavioral approaches, cultural analysts see mind and culture as a

necessity for human nature and, according to Geertz (1973):
There is no such thing as a human nature independent of culture (...) They

would be unworkable monstrosities with very few useful instincts, fewer
recognizable sentiments, and no intellect: mental basket cases (Geertz,
1973:49).
In short, we need culture to create order in the world, and the way we
think and the way we act are to be interpreted as culturally mediated. The
fourth paper addresses four themes which distinguish cultural from
behavioral perspectives: 1) culture as construction of meaning, 2) sub-
cultures and power, 3) the importance of tacit and embodied knowledge,
and 4) culture as a process – not as a noun.
A cultural approach to safety often takes a holistic and inter-subjective

starting point taking into consideration employees’ construction of
meaning towards work related risks. In contrast to behavior based safety,
which mainly is concerned with observable behavioral outputs, it tries to
grasp the tacit or embodied sides of culture, to balance the formal and
informal sides and may consider how external and internal conditions
influence employees’ possibility of behaving in a safe or unsafe manner.
Methods used are often qualitative in character and the epistemological
roots are hermeneutic.
31
Behavior based safety approaches make use of stimuli – response models

and principles of operant conditioning and reinforcement. They examine
the specific context of the employees, diagnose and treat the critical
behavior, and work with the outcomes of culture (DeJoy, 2005). In this
respect, a behavioral and a cultural perspective will have different starting
points, even if their goal may be the same – that is to develop safe
practices and behaviors. The strengths of the BBS approach are related to
its specific “technology”, the systematic collection and analysis of
observed data, its focus on frontline personnel, a participatory approach
and its positive focus on desired safety behaviors (ibid.).
I argue that the characterization of BBS approaches as bottom-up and

cultural approaches as top-down (DeJoy, 2005) may be somewhat
misleading, and actually depending more on how the actual intervention is
carried out – rather than as striking characteristics of both. Challenges
related to BBS interventions may be: lack of trust between employees and
management, disagreement of what constitutes a just organizational
culture, employee rejection of performance goals, efforts may not give
expected improvement, it may be hard to hold focus after the intervention,
lack of risk awareness, and a too high focus on individual behavior –
losing track of external conditions which influence safety behavior
significantly (Reason, 1997; Lingard & Rowlinson, 1997; Depasquale &
Geller, 1999; Cox et al., 2004).
3.3 Safety culture and safety climate

The sense or nonsense of the concept “safety culture” has been discussed
for a couple of decades, and we now see a move towards treating
(organizational) culture as the main concept and safety as an aspect of it.
This latter clarification I find both necessary and fruitful, and it will be
made the basic foundation of this thesis. As such, it rests on a clear
distinction between culture and climate, reserving the latter for aggregated
perceptions or attitudes of a group of people, while the first deals with
sense-making, meaning construction, tacit and taken for granted beliefs
about our “world”.
Frost, Moore, Louis, Lundberg and Martin (1991) recap the research on
organizational culture into three main perspectives: Integration,
Differentiation and Fragmentation. The integration perspective treats
culture as distinguished by unity and strength. Agreement then becomes
32
the measure by which one decides whether the culture is strong/weak,

present or absent. Organizational culture may then be reduced to a
management tool by which one may shape, strengthen or mould by one’s
own will and efforts. Disagreement and differing subcultures then become
a threat and a symptom of the lack of one common and shared
organizational culture. The differentiation perspective is more open
towards the existence of sub-cultures and countercultures, often living
harmoniously side by side within the same organization. The third
approach, the fragmentation perspective, takes into account that the same
events or acts can be interpreted differently by different people or groups
of people, and that conflicting interpretation frames constitute an
important part of a company’s real life – dividing the organization into
various sub-cultures.
Since the INSAG definition of safety culture, several alternatives have
been made and most of them derive from the organizational culture
literature. Pidgeon (1991) argues that safety culture should be understood
as the constructed systems of meanings through which a given worker, or
group of workers, understands the hazards of their world. As such, it is
deemed to have relative stability and not to change on an hourly, daily or
weekly basis (Cox & Cox, 1996). Gherardi & Nicolini (2000) bring the
safety culture concept towards the “communities of practice” and define it
as “the collective ability to produce organizational and inter-organizational
work practices that both protect individual welfare and the environment.”
Safety is then considered to be a competence realized in practice (Gherardi
& Nicolini, 2000:8). This (safety) competence, knowledge or expertise is
then viewed as circulating in webs of practices.
Safety climate, on the other hand, is often used to describe employees’
perceptions of how safety is dealt with at the specific workplace. These
perceptions are often measured by questionnaires and provide us with a
“snap shot” of the current state of safety (Mearns & Flin, 1999). Safety
climate dimensions are mainly treated as leading indicators and predictors
of safety performance (Mearns, 2009), and it is fairly well documented
that a favorable safety climate is essential for safe operations (Mearns,
Whitaker & Flin, 2003). Zohar (2008) argues that safety climate refers to
shared perceptions regarding safety policies, procedures and practices and
that safety climate tools should, as far as possible, reflect policies-in-use
(enacted policies) rather than their formal counterparts. In addition, safety
climate tools should reveal real management priorities of safety when
safety is under pressure – named “acid-test-indicators” of “true” priorities
33
of safety, e.g. competing operational demands, safety versus production,

time pressure etc. Measurement sensitivity could also be enhanced by
developing industry or sector specific items (Zohar, 2008).
According to Cooper and Phillips (2004), the last 25 years of safety
climate research can be split into the following four directions: (1) The
design of psychometric measurement instruments and their underlying
factor structure, (2) development and testing of theoretical safety climate
models to ascertain determinants of safety behavior and accidents, (3) the
examination of relationships between safety climate perceptions and actual
safety performance; and (4) exploration of the links between safety
climate and organizational climate. The field is dominated by the search
for the right inventory or dimensions able to grasp the “true priority of
safety” (Dedobbeleer & Beland, 1998; Anderson, McGovern, Kochevar,
Vesley & Gershon, 2000; Cox & Cheyne, 2000; Flin, Mearns, O’Connor
& Bryden, 2000; Griffin & Neal, 2000; Guldenmund, 2000; Zohar, 2003;
Cooper & Phillips, 2004). The most frequently used statistical method to
determine the dimensional structure of safety climate is factor analysis and
yet there exists no agreement on what dimensions safety climate consists
of. However, safety management and colleague involvement dimensions
have been found in several studies (Zohar, 1980, 2003; Flin et al., 2000;
Guldenmund, 2000; Rundmo, 2003), and have also been found to play a
crucial role in the implementation of behavioral safety interventions
(Lingard & Rowlinson, 1997; Depasquale & Geller, 1999; Krause et al.,
1999; Geller, 2001; Johnson, 2003; Lund & Aarø, 2004; Cox et al., 2004;
Seo, 2005). Other common dimensions are related to safety systems,
procedural compliance, work pressure, competence and risk (Flin et al.,
2000).
An assumption within safety climate research has been that differences in
key underlying structures may reflect methodological differences in
question generation, sample populations across industries, labeling of
constructs according to the theoretical model driving research or that
different instruments measure distinctly different or overlapping safety
climate concepts (Cooper & Phillips, 2004). Safety climate measures
seem to be useful, however, for ascertaining employee’s perceptions,
despite differences in how safety climate is conceptualized (Carroll, 1998;
Cooper & Phillips, 2004).
34
3.4 Safety climate and indicators of safety

performance
The past seems incredible, the future implausible.
(Woods, 2009:498).
In safety research, it is generally assumed that there is a relationship
between safety climate and safety performance; i.e. that the employees’
safe or unsafe practices or behaviors are a function of the underlying
organizational safety culture and the reflected or measured safety climate.
However, it has been difficult to provide empirical evidence for causal
links between them (Zohar, 2003; Cooper & Phillips, 2004). Some have
suggested that the relationship might be mutual (Clarke, 2006) or that “the
climate-behavior-accident path is not as clear cut as commonly assumed”
(Cooper & Phillips, 2004:497). A mutual relation implies that safe
behavior may lead to a safer culture and less accidents or reverse, that
accidents may urge the organization towards a safer culture and better
scores on safety climate measures. One might also finds situations where
an improved safety culture involves better incident reporting and hence an
apparent worsening of safety performance, and that employees with high
safety awareness might expect more and, hence, be more critical than
those who expect less. Safety climate, culture and reporting practices may
depend on both local level engagement and trusting mechanisms between
workers and management (Reason, 1997; Clarke, 2006).
Safety performance on the other hand can relate to different aspects and
levels, and the development of sound safety indicators in a company, an
industry or a sector depends on careful design. According to Hopkins
(2009) some of the most used concepts like “lead” and “lag” indicators are
not fully understood, used inconsistently and are problematic. A similar
problem exists with regards to the tendency of conflating personal and
process safety indicators. Mainly, mixing up personal and process safety,
he argues, stems from a misinterpretation of the Heinrich triangle (1931),
implying that minor accidents are precursors to major accidents (Hale,
2001; Hopkins, 2009). While process safety indicators refer to hazards
arising from the processing activity in which a plant may be engaged,
personal safety is related to hazards that may affect individuals and have
little or nothing to do with process safety. Process safety accidents may,
typically, damage or threaten to damage a plant and cause multiple
fatalities. Hence, personal safety indicators will not tell you how well you
are managing process safety and vice versa.
35
Hopkins sorts the lead/lag and personal/process safety in a two-by-two

table, which shows that both personal and process safety indicators can be
lead or lag (ibid.:461). It is worthwhile to point out also, as Hopkins does,
that indicators are indicators of something; a surrogate for more latent
phenomena, and in the case of lead/lag they are indicators of something
like “the state of the safety management system” (ibid.:509). Safety
management systems are put in place in order to protect people and
property from harm, and, intuitively then, most of us will agree with the
idea that lag indicators are direct measures of harm, while lead indicators
are in some ways precursors of harm (ibid.:461). From the Leading
Performance Indicator guide for Health and Safety in the UK oil and gas
industry leading indicators are defined as “something that provides
information that helps the user respond to changing circumstances and the
actions to achieve desired outcomes or avoid unwanted outcomes”
(Mearns 2009:491). Leading indicators, then, provide us with information
regarding the current situation which can affect future performance. Lag
indicators of personal safety are mostly unproblematic and usually refer to
injuries or fatality rates. Another concern Hopkins delineates, is the fact
that once there are incentives or consequences attached to any kind of key
performance indicators (leading or lagging), there is an incentive to
manage the indicator itself instead of the phenomenon it is an indicator of.
Among safety researchers it is commonly agreed that the correlations
between personal and process accidents is limited, if there is any at all
(Hopkins, 2009). Hudson (2009), however, argues that there actually
exists a level of commonality, but that this relation is to be found at an
abstract level; at the level of organizational culture where prioritization of
safety may affect both personal and process safety. The BP organizational
culture at the refinery in Texas City resulted on the 23rd of March 2005 in
one the most serious workplace disasters in the US with 15 deaths and
more than 170 injuries (Baker, 2007). The accident investigation clearly
shows an organization with a non-compliance culture at many levels
undermining both its personal and process safety (Hudson, 2009). The
problem was, however, neither lacking personal safety information nor
process safety indicators, but the organization’s willingness or ability to
treat them as important and real.
As the interesting debate among 19 researchers/practitioners in the special
issue of Safety Science show (Safety Science, April 2009, vol. 47, Issue 4),
there is still disagreement regarding the lead/lag distinction and how to
conceptualize it. One of the main issues at stake relates to the challenge of
36
establishing suitable indicators which leads us into a common problem

among safety researchers: What actually characterizes a valid and reliable
safety performance indicator – at what level? Kjellén (2009) goes back to
Rockwell’s (1959) definition of the concept. He puts forward the
following characteristics as necessary: They should be (a) quantifiable and
permitting statistical inferential procedures, (b) valid or representative of
what is to be measured, (c) provide minimum variability when measuring
the same conditions, (d) sensitive to change in environmental or
behavioral conditions, (e) costs of obtaining and using measures is
consistent with the benefits and (f) comprehended by those in charge with
the responsibility of using them. Fatalities happen for instance infrequently
and, hence, do not constitute a suitable indicator at a company level, but
may constitute a good industry or sector level indicator.
In this and preceding chapters I have made a distinction between the
culture and climate concepts, pointed at organizational aspects or
dimensions which have predictive qualities towards safety performance,
and discussed the relation between safety climate concepts and indicators
of safety performance. In the following, I want to focus on trust, another
very common and often taken for granted concept, presumed to play an
important role in establishing a sound safety culture. Within differing
research fields, trust is most commonly treated as something that in some
way creates positive outcomes. Within safety research lately, and in this
thesis, I want to make this picture a bit more nuanced and reflect upon
both the potential advantages and disadvantages of trust and distrust.
3.5 Trust and distrust with regards to safety

The role of trust has been argued to play an important role in modern,
global, complex, and ambivalent risk societies (Barber, 1983; Beck, 1991;
Giddens, 1991) and to be a cornerstone in the creation of social order
(Luhmann, 1979). The trust concept is also regarded to be a complex,
multidisciplinary, multifaceted, multilevel and multiplex phenomena
(Deutsch, 1958; Luhmann, 1979; Doney, Cannon & Mullen, 1998;
Lewicky, Mcallister & Bies, 1998; Rousseau, Sitkin, Burt & Camerer,
1998; Lewicki, Tomlinson & Gillespie, 2006). Despite differing
conceptualizations, there seems to be some agreement on the meaning of
trust, the conditions that must exist for trust to arise and its functions. It
implies positive expectations about others’ intentions and behavior, it
reduces complexity and conflict, involves vulnerability and risk, and it
37
involves interdependence between different types of actors. Research on

trust has commonly been related to positive organizational outcomes, like
increased competitiveness (Seppänen, Blomqvist & Sundqvist, 2007),
higher organizational performance (Child & Möllering, 2003), reduced
transaction costs (Dyer & Chu, 2003), increased communication and
knowledge exchange (Andrews & Delahaye, 2000), enhanced mutual
learning (Nonaka & Takeuchi, 1995; Bakker, Leenders, Gabbay, Kratzer,
& Van Engelen, 2006; Gubbins & MacCurtain, 2008), learning from
accidents (Reason, 1997), and is seen to have a positive impact on safety
culture and safety performance (Reason, 1997; Hale, 2000; Burns, Mearns
& McGeorge, 2006; Conchie & Donald, 2006, 2008; Conchie, Donald &
Taylor, 2006; Cox, Jones & Collinson, 2006; Jeffcot, Pidgeon, Weyman &
Walls, 2006).
Most of the research on trust and distrust relies on the assumption that
trust on different levels is beneficial and that lack of trust is bad. Lately
this view has been challenged, and there now seem to be new perspectives
emerging that place emphasis on the limits of trust and the potential
benefits of distrust (Reason, 1997; Hale, 2000; Adler, 2001; Dirks &
Ferrin, 2001; Poortinga & Pidgeon, 2003; Lewicky et al., 2006; Burns et
al., 2006; Conchie et al., 2006; Conchie & Donald, 2006; Conchie &
Burns, 2008). One might have too much trust ending in naivety or
blindness, while too high a level of distrust, for instance in the form of too
much monitoring and control might result in a non-sharing environment,
inefficiency or other non-intended side-effects. Also a view of trust-
distrust as a bipolar phenomenon – indicating that trust and distrust exist at
opposite ends of a single trust-distrust continuum – has also been
challenged. In the bipolar approach low trust expectations become
indicative of high distrust and distrust is treated as a psychological,
organizational or societal disorder that should be corrected (Lewicky et al.,
1998). To replace such assumptions Lewicky et al. (1998) argue that
relations should be treated as more dynamic, complex and multifaceted.
Parties may hold different views of each other, views that may be accurate,
but inconsistent, and balance and consistency are most likely temporary
and transitional states. Lewicky et al. (ibid) also argue for taking into
account the bandwidth and richness of ongoing relationships; the broader
the experience across multiple contexts, the broader the bandwidth. Hence,
trust (and distrust) may vary in scope as well as in degree. Trust research
38
with a non-normative perspective is limited and more research is needed in

order to understand these dynamics.
My understanding of trust relies on an assumption that both trust and

distrust may be functional as well as dysfunctional – the overall aim with
both is to create predictability. This is also in accordance with a
Luhmanian perspective on trust, which claims that both a trusting and
distrusting strategy involves reduction of complexity, i.e. trust reduces
social complexity by allowing specific undesirable conduct to be removed
from consideration, while distrust functions to reduce complexity by
allowing undesirable conduct to be seen as likely – even certain
(Luhmann, 1979, from Lewicky et al., 1998:444). This perspective on trust
implies that trust and distrust may have both a functional and
dysfunctional impact on safety and effective risk regulation. Dysfunctional
trust and dysfunctional distrust are hypothesized to give poor safety
performance, while functional trust and functional distrust should lead to
sound safety performance. Further, we may argue that functional distrust
and functional trust may be treated as a necessary part of a sound safety
culture and, conversely, dysfunctional trust would be detrimental for
safety (Hale, 2000; Conchie & Donald, 2006; Conchie et al., 2006). In this
picture both trust and distrust depend on risks being kept under control,
which is of crucial importance in high-reliability systems and
organizations. Highly complex systems are supposed to need both more
trust and distrust, often in the case of institutionalized distrust or control
(Luhmann, 1979). With increased diversity, trust is anticipated to become
more difficult to sustain. A solution to this could be to treat both
confidence and doubt as important in the production of reliability and to
keep an eye on the limits to faith (Weick, 1987) and to the benefits of
doubt or distrust.
Trust building factors has been linked to individual characteristics like,

ability, integrity, benevolence, being truthful, openness, showing respect
etc. (White & Eiser, 2006; Seppänen et al., 2007). Some authors identify a
single trustee characteristics, while others delineate as many as 10. Mayer,
Davis and Schoorman (1995) compress these factors into three attributes:
Ability, Benevolence and Integrity. Ability refers to the trustee’s skills and
competencies to do the job and should be regarded as dynamic.
Benevolence is the extent to which a trustee is believed to want to do good
to the trustor, aside from an egocentric motive. The relationship between
39
integrity and trust involves the trustor’s perception that the trustee adheres
to a set of principles that the trustor finds acceptable (Mayer et al.,
1995:719). Integrity has also been linked to concepts like moral integrity,
consistency of the party’s actions, credible communications, a strong
belief in the trustee’s sense of justice and congruence between his or her
words and actions. Mayer et al. (1995) treat the three traits as interrelated
and as varying independently of each other.
With regards to the industrial frames formerly described, both trust

building and control mechanisms are in use in order to secure reliable and
effective work performance. Findings, in general, suggest that permanent
workers (client staff) invest less trust in temporary (contractor) workers
(Moreland & Levine, 2002), that contractors are being treated unfairly and
as inferior B-team members, and that such aspects might create undesired
side-effects (Collinson, 1999). If we consider trust with regards to client-
contractor relations, Conchie & Donald (2006) show that low trust
between client and contractors seem to be most common in the initial
phases of a relationship. In the course of time, positive trust relations
between the customer company and its supplier might be built, depending
on demonstration of ability and good workmanship (ibid.).
3.6 Cross-cultural comparison

Both the UK and Norwegian Continental Shelves constitute mature oil
producing regions. In addition, major petroleum companies operate
globally and often have an outspoken goal of international growth. This
implies extensive cross-cultural contact, and that safety priorities have to
be communicated consistently to employees with different cultural
backgrounds and sometimes societies characterized by completely
differing realities and risk scenarios. Operating across national borders
both within and outside Europe requires a delicate sensitivity towards
cultural aspects and an awareness of what the potential impact may be for
the management of safety. Employees within the industry are also
expected to be able to work wherever the job takes them, anticipating a
flexible attitude towards work, work hours, work place, and that
employees’ competencies’ are transferable all over the world.
Globalization, cultural differences and how these processes may influence

occupational safety have recently been reviewed and studied within the oil
and gas industry (Mearns & Yule, 2008). Due to the fact that various
40
elements in institutional environments (such as culture and regulations)

vary between nations – organizational and management practices are
anticipated to vary across countries (Gooderham, Nordhaug & Ringdal
1999; Kostova & Roth, 2002; House, Hanges, Javidan, Dorfman & Gupta,
2004). With regards to safety, it is assumed that organizational safety
culture may be embedded in deeper rooted values (as cultural values), but
also that safety may constitute a “universal value” (Helmreich & Merrit,
1998). Mearns & Yule’s (2008) study used the Hofstede cultural value
dimensions (Hofstede, 1984) and their results suggested that more
proximal influences such as perceived management commitment to safety
and the efficacy of safety measures exerted a higher impact on workforce
behavior and accident rates than fundamental cultural values (Mearns &
Yule, 2008). Previously, only one comparative analysis of offshore safety
climate across the UK (UKCS) and Norwegian Continental shelves (NCS)
has been performed (Mearns, Rundmo, Flin, Gordon & Fleming, 2004).
The results showed clear differences in how UK and Norwegian workers
evaluate various social and organizational aspects that might have an
impact upon safety. For most of the scales, a greater percentage of the
variance was explained by installation rather than sector (Mearns et al.,
2004). Another study found differences in safety performance between the
Nordic countries within the building and construction sector
(Spangenberg, Baarts, Dyreborg, Jensen, Kines & Mikkelsen, 2003).
With regards to this thesis, the participants in the comparative study

belong to the same company, and they perform comparable work tasks
(drilling and well services) in approximately similar work environments
(production platforms) though working on different shelves. At some
places offshore these production platforms are so close to one another that
one might salute each another from platform to platform, only that the
invisible border separates them into a UK and a Norwegian zone and
workers with different national backgrounds. This context makes a unique
opportunity of examining the influence of national and organizational
culture on a range of issues; how is organizational culture expressed across
shelves? What matters most; national, organizational or installation
culture? And, do trust relations and leadership styles have different
dynamics or expressions, and what kind of role may such potential
differences have for safety?
For our purpose, I decided to take a further look into the GLOBE study
and the dimensions developed there (House et al., 2004). The GLOBE
41
study divides the world (62 participating nations) into 10 clusters

according to their perceived values and practices across 9 cultural value or
practice dimensions. Organizational cultures are anticipated to reflect the
societies in which they are embedded and psychometrically tested
dimensions reflect either practices (the way things are) or values (the way
things should be done). In this thesis, scales reflecting practices are used.
According to cluster scores in the GLOBE study, the Anglo and Nordic
cluster should differ on dimensions like Assertiveness (aggressiveness and
dominance) and Power Distance, indicating that lower aggressiveness and
low Power distance are more accepted or prominent characteristics in the
Nordic than in the Anglo cluster. Even if differences may be large both
between and within countries, the Anglo scores are anticipated to reflect
our UK sample employees and the Nordic scores to mirror our Norwegian
sample. Another important assumption we made was that differences
found in the House cluster scores, would reveal themselves within one
single company, implying that that differing national cultural backgrounds
would overwhelm organizational culture.
Former research on the relationship between cultural clusters and trust

building processes have also shown differences both with regards to the
level and form of trust across international borders (Doney, Cannon &
Mullen, 1998; Madhok, 2006; Sullivan, Peterson, Kameda, Shimada,
1981; Zaheer & Zaheer, 2006). Zaheer & Zaheer (2006) link these results
to the fact that the nature and the institutional and cultural support for trust
seem to vary across nations. With regards to the former mentioned
differences in Assertiveness and Power distance, Doney et al. (1998) have
for instance suggested that more Masculine and higher Power distance
cultures would be linked to a more calculative based trust building
compared to their counterparts in Feminine and low Power Distance
cultures. As such, to examine whether such dynamics are found in our
comparative study and what kind of role cultural differences may have for
safety, make interesting issues.
3.7 Summary – theoretical framework

This chapter outlines main theoretical framework and concepts used in the
thesis. Culture makes a concept difficult to grasp, but most definitions
refer to some kind of sharedness; like common ways of thinking, feeling,
reacting etc. In this thesis culture will be closely connected to
interpretation processes, common sense making, and taken for granted
42
values and norms, systems of common symbols, and the meaning people
make of such symbols.
Culture expresses itself through our behaviors and practices – which may
be safe or risky. I have pointed at differences in behavioral and cultural
approaches to safety. However, their aims are often similar – to make safer
workplaces or to change risky work practices. Differences relate to
scientific roots, methods used and what is treated as important; outcomes
of culture or underlying forces shaping it.
Safety culture will in this thesis be defined as an aspect of the

organizational culture, reflecting common symbols, sense-making,
meaning systems and practices which enables or disables group,
organizations or societies to protect themselves, the environment and its
members from harm. I have also drawn a quite clear distinction between
safety culture and safety climate, reserving the latter for aggregated
perceptions or attitudes of a group of people, measured by questionnaires
and statistical techniques (Flin et al., 2000). Safety climate scales are
treated as leading indicators of safety performance and ought to
demonstrate predictive qualities towards actual safety practices or
behaviors (safety performance). Safety climate dimensions should,
according to Zohar (2008), reveal real management priorities of safety
(when safety is under pressure), and measurement sensitivity may be
enhanced by developing sector specific items (ibid.).
Different groups within a society or in an organization may to a varying

degree be able to protect themselves from harm. Contractors, for instance,
often find themselves closer to the “sharp end” compared to their clients,
and, because of this, suppliers may experience higher exposure to
accidents or other kinds of risks compared to client personnel, like for
instance, personal injuries and health complaints. As such, we may argue
that safety constitutes a resource that is socially and structurally
distributed, and that such distribution involves processes of power – often
embedded in culture and taken for granted within different cultural
contexts.
Trust has both within safety and general organizational research been
found to have positive outcomes, while distrust, on the other hand, mostly
has been related to negative consequences. I lean on a Luhmanian view of
trust and distrust treating them as parallel forces targeted towards social
43
order and predictability. With such an approach, both trust and distrust
may be used in functional and dysfunctional ways with regards to safety;
there may be too much trust having bad outcomes and functional distrust,
may have a positive impact on safety.
All in all, the thesis touches upon a theoretical field which is distinguished
by being multidisciplinary, using insights from mainly sociology,
anthropology and psychology. This cross-disciplinary approach also
applies to the following chapter, illustrating overall methodology and
different methods used for the four studies.
44
Methodology
4 Methodology
Research design refers to the planning and execution of the different
studies; how I decided to go about in order to examine specific research
problems and considerations regarding the strength and limitations of the
methods used (Blaikie, 2000). The thesis consists of four papers with their
own aims and research problems and differing research strategies have
been chosen to enlighten them in a reasonable manner. This chapter
provides an overview of the research design and the strategies used for the
different studies; qualitative and quantitative methods, samples and data
collection, accomplishment of focus groups (Magnaghten & Myers, 2004;
Morgan, 1997); statistical methods, tools and techniques that have been
used, followed by ethical considerations and philosophy of science
positioning. Finally, a conceptual work model is set up, illustrating
concepts which are examined in the thesis and the anticipated links
between them.
4.1 Research design

Most of the studies within safety climate research use questionnaire
surveys and psychometric measurement techniques to establish sound
climate constructs enabling analysis of safety dynamics within
organizations. This is also the case with three of the articles in this thesis
(first, second and fourth). The fourth also includes qualitative data in the
form of focus group interviews.
The point of departure in article III is not an empirical, but a theoretical

one. We urge for a sharper distinction between and a better balanced
understanding of cultural and behavioral approaches to safety. As such, the
third paper mainly involves a review of literature discussing the roots of
various theoretical safety perspectives, differences and similarities, and
potential consequences for the understanding of safety and safety
interventions.
The three quantitative studies embrace large and unique cross-sectional

samples of petroleum employees gathered at four moments in time: 2001
(N=3310), 2003 (N=8567), 2005 (N=4479) and 2007 (N=791). The first
three involve Norwegian offshore workers, while the last includes
45
Methodology
Norwegian and UK drilling employees working on production platforms

on both shelves.
Several authors have urged for the use of combined methods and touched
upon the differences and similarities between qualitative and quantitative
approaches (Denzin, 1970; Brannen, 1992, 2004; Hammersley, 1992;
Creswell, 1994; Blaikie, 2000). Article IV brings different datasets into
play, quantitative and qualitative, in different phases (or sequences) in
order to complement each other. The motivation for combining methods
was both to cross-validate quantitative findings, and to seek the actor’s
point of view on the results (Blumer, 1956).
Figure 4 below shows time of measurement, sample sizes, response rates

and targeted populations for all four questionnaire surveys.
Paper 1: Paper 2: Paper 4: Paper 4:

2001: N=3310 N=4479 N=791 2007:
Response: 55% Response: Response: N=2x15
2003: N=8567 50% 67% Focus
Response: 50% groups
Population:
Population: NCS & NCS &
Populations: NCS UKCS UKCS
NCS
2001 2003 2005 2007 2008
Figure 3 Measurement time, samples, response rates and population
4.2 Samples and data collection

The first study involves a cross-sectional design, using two quantitative
datasets (2001 and 2003) from the TRL project. The two datasets
constitute main point of departure to perform testing and retesting of the
psychometric qualities of the survey instrument. All employees working
46
Methodology
on offshore installations (production platforms, mobile units and vessels)

on the NCS were asked to fill in a self completion questionnaire. Each
year the questionnaire was distributed by the medical staff offshore. In
addition, a network of distribution contacts in different companies was
established, encouraging offshore workers to participate in the survey. The
questionnaires were filled inn during the employees’ shift period and
returned in closed envelopes to the medical staff at the end of the shift.
Thereafter completed forms were sent back to the research institutes (IRIS
in 2001 and DNV in 2003)4 and integrated into a database for statistical
analysis (SPSS). Due to an extension of the survey from two weeks in
2001 to six weeks in 2003, the samples vary considerably in size, but are
considered to be representative at both moments in time. The samples have
been checked according to demographic background information provided
by the companies to the PSA.
The second paper uses survey data from the same database (TRL), but
from the questionnaire study in 2005. The core database involves N=9820
respondents with a response rate at 50%. For theoretical reasons we
selected i) respondents who had their ordinary work offshore (spending
more than 50% of their work time offshore), ii) who worked for a client
company operating more than two offshore oil and gas installations, and
iii) employees who reported that they were working all the time or mostly
on the same installation. Our sample size was now reduced to N=4479
respondents. The accomplishment period was the same as the survey in
2003 (six weeks), as was the target group. However, the distribution
procedure was slightly changed. This time the respondents received the
questionnaire either through their own company routines, a web version or
at the heliports where employees are shuttled to their installation. All
workers were encouraged to participate and return the completed
questionnaire in a closed envelope either to the medical staff offshore or to
send it directly to the responsible research institution (IRIS).
Questionnaires were treated anonymously and respondents were promised
confidentiality.
4
This questionnaire survey is accomplished and announced in a public tender process
by the PSA. Hence, which research institute gets to carry out the survey may vary.
Nevertheless; IRIS has been main responsible actor for the carrying out of these
questionnaire surveys, except from the one in 2003.
47
Methodology
Article IV uses a combined method design integrating both quantitative

and qualitative data. The first phase involves a quantitative survey
involving all company offshore employees working on the UK and
Norwegian Continental Shelves. The survey is carried out
Summer/Autumn 2007 and is based on self completion questionnaire data.
In the UK all questionnaires were distributed by post, with an information
letter from IRIS and a request from their Human Resource manager to fill
in the questionnaire. In Norway a mixed distribution process was used.
Some of the questionnaires were sent and returned by post, while others
were distributed by company contacts at the heliports. Differences in
distribution procedures were related to pragmatic reasons. Some of the UK
workers would have been hard to reach, due to little work in the period of
the survey accomplishment. Hence, they were easier got to by post. In
both countries the questionnaires are pre-stamped and returned in closed
envelopes to the research team in Stavanger. Participation was done
voluntarily and anonymously.
Qualitative data were gathered in the second phase, consisting of two

focus group interviews accomplished separately in each country (Spring
2008). Each focus group held 15 participants comprising: offshore
workers, rig managers, human resource personnel, safety delegates and
union representatives. The focus groups were organized as structured
discussions between researchers and informants focusing on three main
themes measured in the questionnaire study: 1. Self reported safety
behavior and Safety performance, 2. Functions of safety specific trust and
distrust with regards to safety, and 3. Power distance and Assertiveness
(House et al., 2004). The Trust/Distrust reflections in the focus groups
were taped and transcribed, while the other sections were summarized,
distributed among the research team members for validation and sent back
to the company. The transcriptions and summaries have been content
analyzed along a cultural comparison axis. Between the two phases, short
summaries of results were presented and commented on by company
representatives. The Seadrill office in Aberdeen was not a very old one –
established in February 2008. One thing that became obvious for us during
the focus group interviews was that especially UK workers seemed to have
a very positive opinion of their employer. One of the main reasons for this
may be related to the fact that they, despite of hard financial times (periods
with little work) kept their employment. The company is also back in
Norway known to be loyal towards their employees (“safe” employer”),
48
Methodology
which is anticipated to yield long employment relations and a strong

identification towards the company.
4.3 Statistical methods, tools and techniques

Factor analysis methods: In the quantitative articles principles from
psychometric theory, factor analysis, and the most common scientific
criteria employed within scale development have been used and two
statistical tools; SPSS and Lisrel.
Factor analysis represents a method that commonly seeks to explore the

underlying structure of a particular phenomenon. It is based on complex
structure-analyzing procedures, identifying the interrelationship among a
large set of observed variables. As such a factor represents a linear
combination or cluster of related variables that represents a specific
underlying dimension of interest. The underlying phenomenon a scale
intends to reflect is often called a latent variable. This notion implies that
it is not a directly observable or a manifest phenomenon and that the
construct is variable rather than constant (Devillis, 2003). Often several
factors are included, and a much used criteria used is that factors should be
as distinct from one another as possible (Pett, Lackey & Sullivan, 2003).
Different wordings are often used for the same phenomena; e.g. factor,
scale, dimension, facet, construct etc., referring to a latent (organizational)
aspect which we want to explore the importance of.
The most common methods used are Exploratory Factor Analysis (EFA)
and Confirmatory Factor Analysis (CFA). EFA is in general used when the
researcher does not know how many factors are necessary to explain the
interrelationship among a set of indicators or items. Hence, it is an
effective way of ascertaining the minimum number of hypothetical factors
that can account for the observed co-variation, and to explore the data for
potential data reduction (Kim & Mueller, 1978). In contrast, CFA is used
to assess the extent to which the hypothesized organization of a set of
identified factors fits the data. It is used when the researcher (or the
research field as a whole) has some knowledge about the underlying
structure of the construct under investigation. One could say, somewhat
simplified however, that EFA represents a data driven and CFA a theory
based approach. When using CFA the purpose could for instance be to test
a specific hypothesis of five underlying dimensions and anticipate that
49
Methodology
certain variables belong to one dimension, while other variables belong to

other specified dimensions.
Reliability and validity: Studies involving factor analysis should include

reporting of reliability measures and validity concerns (Pett et al., 2003).
Scale reliability refers to the proportion of variance that is attributable to
the true score of a latent variable or to the internal consistency of our
factor. This measure is typically equated with Cronbach’s coefficient
alpha, α, and ranges from 0.0 to 1.0 (Cronbach, 1951; Devillis, 2003). A
common criterion for the alpha value is that it should generally not be
below 0.7. However, alpha scores are shown to be sensitive to the number
of items in the dimension (Schmidt, 1996) and should also be interpreted
in relation to other validity concerns (Devillis, 2003). Validity is typically
divided into Content validity, Criterion-Related Validity and Construct
Validity (ibid.). Content validity refers to the extent a specific set of items
reflects a certain domain. It has been defined as the degree to which
“elements of an assessment instrument are relevant to and representative
of the targeted construct for a particular assessment purpose” (Haynes,
Richard & Kubany, 1995:238, from Flin et al., 2006:110). Content validity
is easier to evaluate when a domain is well defined and refers to whether
our constructs really measure what they intend to do. It may be determined
by a number of sources like, relevant theory, empirical literature, expert
judgement etc. (ibid.:110). Criterion validity is a second type of validity,
and typically a pragmatic, objective criterion is defined with which to
evaluate our measure and is also often referred to as predictive validity. In
criterion-related validity the key issue is whether the instrument functions
as a useful predictor of subsequent behaviors, experiences or other
conditions. However, criterion-related validity does not necessarily
involve a causal relationship among variables (Devillis, 2003). With
regards to criterion related validity of a safety climate measure, it is
common to correlate climate scores with some kind of safety outcome
indicator or safety performance data. Such indicators should preferably be
collected through other methods than the questionnaire instrument in use.
Examples of safety outcome indicators used in safety climate studies are
individual or organizational level accident rates or ratings of unsafe
behaviors (Flin, Burns, Mearns, Yule & Robertson, 2006). Some sources
of common method variance biases are anticipated to constitute a problem
in studies where predictor or criterion variables are obtained from the
same source or by the same method or measurement instrument
(Podsakoff, MacKenzie, Lee & Podsakoff, 2003) – for instance in the case
50
Methodology
of self reports of unsafe behaviors or involvement in accidents (Flin et al.,

2006). Safety performance indicators used in this thesis will be described
under 4.4. The construct validity of an inventory is concerned with the
theoretical relationship a variable has to other variables, or to what extent
the construct “behaves” the way it is expected to do, often seen in relation
to measures of other constructs (ibid.).
Fit measures in CFA: The purpose of CFA fit measures is to assess to

which degree our model as a whole is consistent with our empirical data.
There exist a wide range of goodness-of-fit indices, unfortunately none of
them showing to be superior to the rest (Diamantopoulous & Siguaw,
2005). Hence, all measures have to be interpreted in relation to sample
size, estimation procedures, model complexity, underlying assumptions of
multivariate normality etc. The chi-square is the traditional measure used
to evaluate overall fit, and assumes that our model fits the population
perfectly. Since a null hypothesis of perfect fit is mostly broken, it has
been suggested that large χ2 values correspond to bad fit and small χ2 to
good fit. Because of the often broken assumption of a perfect fit, other
measures have been developed which seem to function well when
estimating model fit, and which have been used in my papers: Root mean
square error of approximation (RMSEA), Comparative Fit Index (CFI),
Goodness of Fit Index (GFI), Adjusted Goodness of Fit Index (AGFI) and
Critical N. The RMSEA is proposed by Steiger (1990) and it has been
suggested that an RMSEA value of less than 0.05 indicates a close fit,
values between 0.05 and below 0.08 reflect reasonable fit, between 0.08
and 0.10 mediocre fit and that values >1.0 reflect poor fit (Browne &
Cudeck, 1993; Diamantopoulus & Siguaw, 2005). The GFI and AGFI of
Jöreskog & Sörbom (1989) do not depend on sample size explicitly and
measure how much better the model fits as compared to no model at all.
These indices are supposed to lie between 0 and 1, but values outside this
interval can occur, and, since the independence model almost always has a
huge chi-square, one often obtains values very close to 1.
To justify aggregated analysis of climate perceptions (e.g. installation

level), the respondents’ climate perceptions should exceed a threshold of
homogeneity to index consensus. An rwg value > .70 was used in the first
article as an indicator of within-group agreement (Lawrence et al., 1993;
Klein et al., 2000; from Zohar, 2005).
51
Methodology
Analysis Of Variance: Multivariate analysis of variance (MANOVA)

deals with situations where there is more than one dependent variable and
one or more independents to examine the main and interaction effects of
categorical variables on multiple dependent interval variables. Unlike
ANOVA (Analysis of variance), where there is only one dependent
variable, MANOVA uses more than one dependent variable, and tests the
differences in the centroid (vector) of means of the multiple interval
dependents for various categories of the dependents. The F-test is the first
of the two-step MANOVA process of analysis and tests the zero
hypothesis of no difference in the means of the dependent variables for the
different groups formed by categories of the independent variables. Wilks’
lambda is a measure of the difference between groups of the centroid
(vector) of means on the independent variables; the smaller the lambda,
the greater the differences. Eta2 is used as a measure of effect size.
Logistic regression: To use the Ordinary Least Squares (OLS), which is

the basis of linear regression, a number of assumptions should be met
(Menard, 2002; Eikemo & Høyvarde Clausen, 2007). When the dependent
variable is dichotomous, as our performance measure is in article IV, a
series of assumptions are broken for linear regression. These problems are
solved by using logistic regression which calculates the regression
coefficients on the basis of Maximum Likelihood instead of linear
reasoning. Mainly, a logistic regression analysis predicts the probability of
a case to fall into the higher of two categories on a dependent variable; for
instance in the case of employees’ accident exposure; one has either been
exposed to an accident or not. Logistic regression analysis involves
interpretations of odds ratios (Exp(B)), corresponding confidence intervals
(CI=.95), significance levels and R-square. An odds ratio of 1 indicates a
50% probability of “success” or “failure”. If the probability of being
involved in an incident is higher than 50% the odds ratio will be larger
than 1 and, vice versa, if the probability is less than 50% the odds ratio
will be smaller than 1. Hence, when the CI includes 1.0, the results will
tend to be non-significant. It should be noted that the R-square reported in
logistic regression represents a pseudo-explained variance and hence
should be interpreted differently than in linear regression. The R-square in
logistic regression presupposes a two-divided dependent variable (in my
case: involved in incidents or not), while linear regression presupposes
linear dependency.
52
Methodology
4.4 Tools and measures

Article I develops a safety climate tool named the Norwegian Offshore
Risk & Safety Climate Inventory (NORSCI) involving five safety climate
scales (i) Safety prioritization, (ii) Safety management, (iii) Safety versus
production, (iv) Individual motivation and (v) System comprehension. In
paper II this tool is expanded with a sixth dimension, Safety competence.
Scoring followed the Likert scale from 1 to 5 expressing agreement or
disagreement.
In article IV a selection of scales from the Offshore Safety Questionnaire

was used (Mearns et al., 2001): two self reported safety behavior scales
(Safety Compliance and Safety Participation) and a reporting climate scale
(reporting of hazards, near misses and incidents). The questionnaire also
included two scales measuring Trust in colleagues and Trust in
management commitment to safety (Conchie & Donald, 2006). The Likert
scale (1 to 5) expressing degree of agreement or disagreement was also
used in this survey. The respondents were also given the opportunity to
choose a sixth category (“Uncertain”). During analysis “uncertain” were
coded as missing and hence held out of the analysis.
In order to measure cultural values in the organization two GLOBE

organizational culture scales were integrated (i) Power distance (two items
measuring either social distance between superior and subordinates or
tendency to obey your boss even when in disagreement) and (ii)
Assertiveness (including two items; one reflecting degree of
Aggressiveness and the second measuring degree of Dominance) (House
et al., 2004). The Power distance indicators were kept separate. Scales
ranged from one to seven expressing opposites, e.g. “1=aggressive to
7=non-aggressive”. The respondents were guided to express what norms,
values and practices that are dominant in the organization in which they work
(“The way things are in your organization”). In other words, we were
interested in “the way things are” – not “the way things should be done” in
the respondents’ company.
53
Methodology
Outcome measures and indicators of safety performance
As mentioned above, common safety performance or safety outcome

measures are individual or organizational level accident rates, ratings of
unsafe behaviors and different kinds of self reports; e.g. unsafe behaviors
or involvement in accidents or near misses. In article I two types of safety
performance measures were used: i) self reports of perceived risk (Marek
et al., 1985; Rundmo, 1989, 1992) and installation level accidents. The
risk perception scale included nine industry specific risk scenarios where
respondents were asked to judge the probability of specific risk scenarios
to occur on their platform, e.g. helicopter accident, gas leakages, fire,
blow-out etc. Scoring followed a scale ranging from one to six, where
1=Very little risk and 6=Very high risk. Installation or field level accidents
reported from the companies to the petroleum safety authorities
constituted the second safety performance indicator. The accidents
mentioned below were included, summed up in one score and weighted
per million work hours produced on installation level: Death or severe
personal injury, work disability with absence and medical treatment given
by a medical doctor or under supervision of a doctor. Accidents which
occurred in the subsequent quarter as the survey was carried out were used
for both years (2001 and 2003).
Article four included three safety performance measures: i) Have you been
involved in an incident on this installation the last year, ii) Have you been
involved in a near miss on this installation the last year, and iii) Have you
suffered any injury on this installation the last year and did not seek
medical attention? The first item was used in our logistic regression model
as dependent variable and, hence, as a criterion validity measure. Further,
if the respondents had experienced and injury they were asked to classify it
according to the following categories: a) examination only, b) first aid, c)
medical treatment, d) alternative work, e) absence due to injury, f) a
serious personal injury, or as g) not relevant.
Overview of statistical methods: In table 2 below follows an overview of

all statistical methods used throughout the papers.
54
Methodology
Statistical methods used Art. 1 Art. 2 Art. 4

Exploratory factor analysis x
Confirmatory factor analysis x x x
Crohnbach's alpha x x x
Correlations (Pearsons r) x x
Analysis of Variance (ANOVA) x x
Multivariate Analysis of Variance (MANOVA) x
Logistic regression analysis x
T ‐tests x x x
r wg x
Qualitative interviews x
Table 3 Statistical methods used in all papers
4.5 Ethical issues

Ethical issues are often not that obvious in the social sciences as in other
fields of research, such as within medicine or the natural sciences. No
physical, individual or environmental experiments are performed, and
there is normally no risk of personal or environmental harm. Although
ethical problems are not that obvious, they must be dealt with. The most
evident aspects are related to confidentiality, clarification of expectations
regarding the research project and establishment of a contract between the
researcher and the participants (Pettigrew, 1999). Informants or
respondents should know the main aims of the study, how data are being
treated and reported, give their (informed) consent to participate and be
promised confidentiality (Ryen, 2004). All in all, participants should be
given relevant information of the purpose and relevance of the research
project, and how results are going to be used.
Paper 1 and Paper 2: The Petroleum Safety Authorities are the “owner”
of the TRL project. In order to use datasets from the TRL, an application
has to be sent and approved by the PSA comprising a declaration of
confidential treatment of the data for all researchers involved in the study.
IRIS was responsible for the accomplishments of the questionnaire

surveys in 2001 and 2005, while DNV had main responsibility in 2003.
When the survey was accomplished, participants were given relevant
information regarding the aim of the study, data handling procedures and
55
Methodology
use of the results – both in an introduction session in the questionnaire and

via mail to main contacts. Respondents were promised confidentiality and
participation was done voluntarily. Name or birth date were not requested
and filled in forms are returned in closed envelopes or dropped in a sealed
box offshore. Health and safety statistics constitute an important corporate
identity and reputation factor in the petroleum industry, and is made
relevant in contract negotiations. Hence, as a rule, no company or
installation names were revealed.
Paper 4: Participation in the questionnaire study and the focus groups

were done voluntarily and the respondents were assured confidentiality.
The participants were informed of the main goals of the study, and how
data was going to be used and reported on. For the part of the focus groups
that were recorded, the participants signed and gave their formal consent
to be a part of the study. Total anonymity in focus groups is hard or
impossible to keep. When using citations, there is always a possibility that
participants recognize their own or others’ statements. However, if
citations were going to be used, we promised to do this in a manner that
people external to the project should not be able to identify the informant.
The methodological approach of the project was approved by the
Norwegian Social Science Data Services (NSD).
4.6 Philosophy of science considerations

What research strategies one chooses to solve a research problem involve
a positioning of the researcher according to both some specific ontological
and epistemological assumptions. Ontological assumptions deal with the
nature of social reality; what exists, what does it look like, what units is it
made of and how do these units interact with each other, while
epistemological assumptions refer to the nature and scope of human
knowledge (Blaikie, 2007). This thesis encompasses theory and concepts
from various social scientific disciplines. In order to remain consistent,
such a strategy requires a consciousness on how perspectives are
combined and what consequences such blending will have.
Blaikie (2007) distinguishes between four research strategies, or logics of

enquiry: the inductive, deductive, retroductive and abductive. Inductive
and deductive strategies are based on linear reasoning, while the other two
rest on cyclic or spiral reasoning processes. Inductive strategies are related
to the epistemological principle of empiricism which assumes that the
56
Methodology
world is “out there”. It can be described and explained rather non

problematically as long as the scientist adopts an objective standpoint and
objective procedures. Deductive strategies, on the other hand, are
proposed as alternatives to inductive ones and are often referred to as
hypothetico-deductive methods, with Karl Popper as its most outspoken
advocate and it is related to the epistemology of falsificationism, also
referred to as critical realism (Popper, 1959; Gilje & Grimen, 1993). The
third research strategy, the retroductive, aims at the revelation of
underlying forces that may explain observed regularities, and implies
moving back and forth from observations to explanations. It follows the
“logic of realist explanation” which involves the identification of causal
powers and illustrations of how structure and agency combine to produce
regularities. Finally, the fourth research strategy, the abductive, is often
being referred to as the appropriate method of theory construction in the
interpretive social sciences and is traditionally based on an idealist
ontology and the epistemology of constructionism. It is concerned with
how actors view and understand their social world, how they give meaning
to, make sense of and interpret their everyday life and how these meanings
guide behavior. This research strategy implies an “insider” point of view
trying to uncover the largely taken for granted and common sense
knowledge which guides peoples’ actions (Blaikie, 2007).
In this thesis, and most commonly within general research practice,

strategies are used in combinations. Article I, II and IV are all concerned
with the discovery of regularities, identification of underlying causes or
dynamics which may or may not have an impact on organizational safety.
As such, it follows the “logics of realist explanation”. Also, discussions of
what safety climate might “look” like, what kind of dimensions it may
consist of etc. are done. However, the intention is not to grasp the “Big
Five” once and for all. My view is that safety climate tools may be
dependent on culture as well as branch, and as such implies an
epistemology of constructionism belonging to the abductive approach.
Article III promotes more of a purely abductive approach by its emphasis
on the importance of taking into consideration people’s construction of
meaning, sense making, tacit and taken for granted knowledge in order to
understand and improve safety. The fourth article also involves an
abductive approach involving combined methods taking into consideration
the actors’ point of view on dynamics between culture, trust and safety.
However, and as mentioned above, it also aims at the discovery of
regularities and causal explanations in the quantitative section.
57
Methodology
The field of safety research is characterized by being multidisciplinary

encompassing differing social scientific perspectives, sometimes hardly
referring to or looking at each other’s work (Turner & Gray, 2009). At the
opposite outer edges one could place an empirical positivist approach and
on the other side those treating safety as socially constructed. The first
approach risks the danger of ending in naïve objectivism and the other in
total relativism (Tharaldsen, 2004). A special issue of the journal Human
Relations (Vol. 69(9)) presents safety research which promotes an
understanding of risk and safety as socially constructed phenomena in
contrast to those treating safety according to a “classically individualistic
approach” (Turner & Gray, 2009). The latter approach is criticized as
treating safety as a “disembodied, tangible, and easily quantifiable
phenomenon”, developing or using tools that predict safety outcomes in
organizations, leading to top-down de-contextualized prescriptions on how
to control safety in the workplace (ibid.). By placing differing paradigms
as opposites to each other, one may promote a better distinction between
them, but one risks undermining the fact that the field is characterized by
being cross-disciplinary, and that there may be some gain in combining
approaches. Also, I would disagree that quantifiable analyses necessarily
leads to top-down and de-contextualized prescriptions on how to deal with
safety at the workplace or that a socially constructive perspective solves
all problems – that would depend on the quality of the analysis. Hence, I
would rather argue for a multi-method approach where both qualitative
and quantitative data are used to inform each other in various ways or in
sequences. Surveys do have the benefit of quite effectively reaching a
large number of respondents, but often needs to be complemented with
qualitative data in order to be interpreted in sound ways. Also, I agree with
the point made by Zohar 2008) that the validity of safety climate surveys
may be enhanced by including industry or sector near items which may
serve to reflect sector specific aspects and safety dilemmas.
All science and research will in some way or another serve a purpose; by
its focus, intentions, goals and results. Safety science may be said to serve
moral or ideal goals more explicitly than other research fields – by its
search for – not only understanding and new insights, but also techniques
and methods for improvements; to make safer societies, protect humans
and environments from harm, invent safer technology, models or tools for
how to mitigate risk etc. Hence, the field of safety science may be said to
serve two main goals: 1) development of valid and reliable instruments,
tools, methods etc. which are able to measure (explain) or contribute to the
58
Methodology
understanding of risk and safety, and 2) development of techniques or

approaches which are able to improve safety or decrease risk. As such, this
thesis includes contributions to both.
4.7 Conceptual work model

Both research on safety culture and climate argue for the importance of
getting as close to the natural social unit as possible (Gherardi & Nicolini,
2000), or to find the right level of analysis or natural social units of people
working together (Zohar, 2003). As such, this makes one main strategy
both in the quantitative and the qualitative study.
Safety climate is conceived of as a “leading” indicator of safety
performance (Mearns, 2009), a multidimensional (Flin, Mearns, O’Connor
& Bryden, 2000) and a multilevel construct (Zohar, 2005, 2008). This
means that safety climate dimensions ought to reflect or – in a statistical
sense – contribute to a sound prediction of safety performance at specific
levels. Even though there exists no agreement within the research field on
what dimensions safety climate actually consists of, some dimensions have
shown to be more important than others (e.g. management commitment to
safety), and the methods used to explore, define or test them are more or
less the same. To move the research field forward, there is a need to focus
on the antecedents and mediators of a positive safety climate (Mearns,
2009), and how safety climate may be linked to other organizational
concepts, as suggested by Zohar (2008). This thesis falls within this scope.
I would also like to add that the examination of how safety climate
perceptions are distributed and how such distribution may be explained
makes a key focus of this thesis. The thesis also constitutes one out of few
which combines methods and perspectives in an attempt to both explain
and understand how climate and cultural aspects are intertwined. Below a
working model for the thesis has been set up. The model embraces
concepts which will be subject to examination and may act as an
illustration of anticipated dynamics between them.
59
Methodology
TIME Safety
Safety climate and
safety culture performance:
Group membership and 1. Accidents
characteristics 2. Risk perception
Trust and distrust
3. Self‐reported
involvement
Cultural aspects: Power
distance & Safety behavior
Assertiveness
Figure 4 A conceptual working model
Article one and article two explore relations between time, criteria
variables, safety climate, safety culture and two of our safety performance
indicators: Accident rates and risk perception (article one). The third paper
deals with theoretical clarifications and distinctions between cultural and
behavioral approaches to safety. The fourth study looks into dynamics
between specific criteria variables, cultural aspects, trust/distrust, safety
behavior and safety performance.
60
Results
5 Results
5.1 Summary of article I
Tharaldsen, J., Olsen, E., Rundmo, T., 2008: A longitudinal study of safety
427-439.
The main aims of the study were i) to test and retest the psychometric
qualities of a questionnaire survey according to a model developed by
Cooper (2000), ii) to examine longitudinal development of our safety
climate concepts (from 2001 to 2003), iii) to investigate the hypothesized
association patterns between our safety climate concepts with our safety
performance measures; risk perception (RP) and accident rates both on an
individual and aggregated level, and ix) to examine the differentiating
capacities of our safety climate concepts.
A combination of EFA and CFA showed a five dimensional safety climate

solution consisting of: (i) Safety prioritization, (ii) Safety management and
involvement, (iii) Safety versus production, (iv) Individual motivation, and
(v) System comprehension. A three factor solution, which would have
been in prefect accordance with the Cooper model, was not found.
However, the five factor model may be reallocated according to Cooper’s
three main components. The model has satisfactory fit indices, and the
factors contain reasonable internal consistency and conceptual clarity.
Four out of five factors demonstrate a significant improvement from 2001

and 2003, with a corresponding decline in Accident Rates (AR).
Associations between all safety climate scales and risk perception on an
individual level indicate that the higher perceived risk on the installations,
the lower the safety climate scores. The aggregated level correlations
between AR, RP and our safety climate dimensions also show negative,
but weak relations – with the exception of Safety prioritization, which
shows a positive correlation with both our performance measures. With
regards to differentiating capacity; platform membership show the
strongest impact on employees’ safety attitudes both years, followed by
work area, company (client or contractor) and installation type (production
of mobile facilities).
61
Results
5.2 Summary of article II

Høivik, D., Tharaldsen, J.E., Baste, V., Moen, B.E., 2009. What is most
Science,Vol. 47, no. 10, 1324-1331.
The dimensional structure developed in the first paper was tested using
CFA. The analysis showed satisfactory results for a six factor solution.
The sixth factor may be explained by an adding of two items regarding
safety competence in the 2005 questionnaire. Also, a second order analysis
showed that our hypothesized model fitted the data. Internal consistency
was good, ranging from 0.65 to 0.84.
Five out of six dimensions showed significant differences across operating

and contractor companies, except for Safety management and
involvement, which seems to be high across all companies. Differences
between installations were significant for both operator and contractor
employees. The greatest variance on installation mean scores was found
for Safety versus production and System comprehension, and the smallest
variation on HSE behavior and Competence. All in all, the results of the
ANOVA analyses illustrated that installation membership was of greater
importance for employees’ perception of safety climate than the company
you work for.
5.3 Summary of article III

Tharaldsen, J.E., Haukelid, K., 2009. Cultural and behavioural approaches
to safety – towards a balanced approach. Journal of Risk Research, vol.
12, No. 3-4, April-June 2009, 1-14.
The point of departure in this paper is a debate echoed both in the

Norwegian oil and gas industry and within the field of safety research –
what are the pros and cons with behavioral versus cultural approaches to
safety? What are their main roots? Where do cultural or behavioral
approaches lead us?
By establishing a two-by-two table along a (i) cultural-behavioral and (ii)

tacit- explicit dimension, we were able to pinpoint implications of
differing safety intervention approaches. By taking into account all aspects
62
Results
in the table, one may get a fuller understanding of safety and risk at the
workplace – and it represents a more balanced approach. Explicit and
formal sides of organizational safety practices attract our attention, are
easier to track, count and measure, maybe at the expense of
comprehending more slowly developing, not so easily observable risks. In
a balanced approach one would try to understand how formal (canonical)
practice is related to informal (non-canonical) practice, and to perceive
how underlying or external factors influence employees’ safety behaviors;
why do for instance well intended (formal) safety systems create
unforeseen side-effects when they face (informal) practice? This would
call for a multi-method approach, sensitivity towards the tacit sides of
organizational culture, and an attitude of slow change, slow learning and
slow fix.
5.4 Summary of article IV

Tharaldsen, J.E., Mearns, K.J., Knudsen, K., 2010. Perspectives on safety:
the impact of group membership, work factors and trust on safety
performance in UK and Norwegian drilling company employees. Safety
Science, 48, 1062-1072.
An overall aim of the fourth study was to examine the influence of work
and safety conditions for drilling and wire-line contractor employees. It
operates with four main hypotheses. The first relates to an anticipated
influence of cultural and group related factors on our output measure,
involvement in incidents. The second deals with the potential influence of
stability versus flexibility oriented variables on our dependent measure. The
third tests an expected positive impact of two trust dimensions on safety and
the fourth deals with the potential effects of Safety compliance and Safety
participation on our safety performance indicator. Qualitative interview data
were used to give a better understanding of cultural differences across shelves
and to compliment and interpret statistical results.
The CFA were performed at an earlier stage and demonstrated that our
suggested models fitted our data well (Tharaldsen, Mearns & Knudsen,
2008). All scales showed satisfactory internal consistency, except for
Safety Participation with an alpha below the recommended level
(α=0.584). Descriptive and correlation analysis were performed, and a
five-step logistic model was set up using involvement in incidents (‘yes’
or ‘no’) as the dependent variable and the following independents: Shelf
63
Results
membership, installation, work area, nomadic or stable status on the

installation, shift rotations, Trust in workmates, Trust in management,
Safety compliance and Safety participation.
Significant differences were found across shelves on six out of seven

scales; Trust in management commitment, Safety compliance, Safety
participation and even larger differences on Power distance and
Assertiveness. UK workers reported higher scores on the dimensions Trust
in management commitment, Safety compliance, both Power distance
indicators and Assertiveness compared to their Norwegian co-workers, but
were found to have lower scores on Safety participation. Norwegian
workers reported higher incident involvement, even though they were
found to have higher scores on Reporting climate.
Results from the correlation analysis suggest that among UK employees

the organizational climate may be judged as quite Assertive, but still
Safety compliant. Although among UK respondents Assertiveness and
Power distance is judged as higher than among Norwegian respondents, a
negative association between Assertiveness and Power Distance is found
in the UK sample – indicating that high Assertiveness may be associated
with low Power Distance among UK respondents. Trust in colleagues,
management and Safety compliance showed positive associations with all
scales on both shelves, but among the UK workers Safety participation
indicated a non-significant relation with Safety compliance.
The five-step logistic regression model explained in total 14.9% of

employees’ exposure to involvement in incidents. H1: The effect of shelf
membership on safety performance worked as expected through
installation membership, group characteristics, trust and self-reported
safety behavior. With the introduction of Safety compliance and Safety
participation in the last step, shelf membership lost its significance. H2:
Flexibility/stability related variables showed a significant impact, but in
the opposite direction than expected; employees with varying shift
rotations and a more nomadic work life than others showed lower risk of
being involved in incidents. H3: Hypothesis three was partly accepted
with trust in workmates having an anticipated positive effect on safety, but
the influence of management commitment on safety performance was non-
significant. H4: Our fourth hypothesis also had to be partially rejected.
Safety compliance was associated with a decrease in employees’ exposure
64
Results
to incident involvement, but Safety participation indicated a not-expected

positive relation.
Findings from the focus groups illustrated interesting reflections regarding

differences and similarities across the shelves related to our three main
topics. Trust with an element of control was judged as functional and
positive for safety on both shelves. Too high trust was seen as dangerous
and as constituting potential pitfalls with regards to safety. Beneficial
aspects of distrust like being able to stop colleagues or managers when
engaging in risky behavior were expressed as crucial. However, stopping
or confronting others was experienced as hard; one does not want to
intimidate others. Trusting authority too much and having too much faith
managers, experienced workers or in procedural systems were considered
as involving danger and being risky.
Culturally rooted differences might explain differences in forms of trust in

managers and differences in manager-subordinate trusting relations.
Management styles were also experienced as varying across installations.
Higher Assertiveness taken together with higher Safety compliance and
Power distance among UK workers were linked to a more rule-based trust,
while being connected to equality and democracy among Norwegian
employees. Hence, both forms and levels of trust differed across shelves.
However, differing trust relations were not unambiguously linked to a
better or worse safety performance.
65
Discussion
6 Discussion
This chapter includes a discussion of overall results for all papers seen in
relation to my main research problems given in the introduction: I) The
establishment of sound safety climate dimensions, II) The distributed
character of safety climate, III) Culture and behavior based safety
perspectives, and IV) The links between trust, distrust and safety, and v)
Cross-cultural comparison of offshore safety. For each research problem
succeeding research questions were formulated, specific research aims
were set up, and different research designs and strategies have been chosen
to investigate them. The discussion of the results is addressed in five
sections: (i) Instrument qualities – psychometrics, (ii) The relation
between safety climate concepts, other organizational dimensions and
safety performance indicators, (iii) Differentiating patterns, (iv) Influence
of culture on trust and distrust, and (v) Industrial culture.
6.1 Instrument qualities - psychometrics

The questionnaire survey of the TRL project represents an important
database covering an entire industrial sector. It involves a large number of
respondents and its accomplishment every second year makes a unique
possibility of monitoring longitudinal development. Thus, validation of
this instrument was an important goal – both in order to know whether
data gathered represent valid indicators of the sector’s overall safety level
and for future research. The first and second articles both deal with the
establishment of valid safety climate dimensions in this survey (the
NORSCI), with three tests and re-tests of hypothesized models on
comparable samples in 2001, 2003 and 2005. Our statistical procedures
involved both EFA and CFA, with EFA on our first sample (2001) and all
three factorial solutions were thereafter tested with CFA. The first article
resulted in a five factor solution, and was extended in article two with a
sixth dimension. The factor solution in the first study solution was not
perfectly in accordance with the Cooper model claiming that Safety
practice, Individual safety skill and Situational aspects constituted the
three most important aspects to consider when establishing a picture of an
organization’s safety climate. However, the dimensional meaning content
was possible to reallocate according to these three components. Our first
dimension Safety prioritization comprised items reflecting safety practice,
dimension five included aspects of individual safety skills and the second,
67
Discussion
third and fourth dimension touched upon situational aspects relevant for
safety behavior. Content in dimension two and three may be treated as
influencing safety practice and dimension four could be seen as a
consequence of how the workforce perceive safety management and the
role of safety in relation to production. The dimensions were also in
accordance with the most common factors found in other safety climate
measures (Flin, et al., 2000).
In the first study, we found that offshore employees’ perceptions of risk

and the number of reported accidents decreased significantly during the
period from 2001 to 2003. Taken together with more positive perceptions
of safety climate, this result seems reasonable. Only System
comprehension indicated a small but not significant decline, which might
be related to the introduction of a new procedural system for work permits
and “safe job analysis” during the year before the second survey
accomplishment on the Norwegian Continental Shelf. Shelf scores in the
first paper were not compared specifically with safety climate results for
the 2005 sample, however, the main TRL report shows that safety scores
were slightly improved from 2003 to 2005 (www.ptil.no/rnns). As such,
the industrial sector seems to have experienced a notable improvement in
their safety climate level during the 2001 to 2005 period. A new regulatory
framework on HSE-culture was introduced in the same period, which
promoted an increased safety focus, both from the petroleum safety
authorities’ side and the companies within the industry. As a
consequence, operating companies, contractors and the petroleum safety
authorities introduced their own culture programs focusing on safety
management and leadership, safety compliance, safe behavior programs,
behavioral tracking systems etc. Hence, this progress might be explained
by a combination of an overall improved industrial focus, stronger
surveillance by the PSA, and local company and management efforts.
The fourth paper differs from the others in both its comparative aim and
by its examination of how safety aspects may be linked to other
organizational aspects. Unlike article I and II it also includes qualitative
data. Formerly performed tests of our hypothesized factor model showed
sound CFA fit measures and internal consistency of the concepts involved
were considered (Tharaldsen, Mearns & Knudsen, 2008). Safety
participation held, however, a lower consistency than recommended and
also seemed to behave somewhat irrationally in the logistic regression
model. The logistic regression model was set up in five steps, following
68
Discussion
the logic of our four hypotheses: The most general and basic contextual
characteristics were placed in the first step, thereafter group specific
characteristics related to stability or flexibility, followed by two trust
scales and, finally, the safety behavior scales. Two interesting and
unexpected observations are worth noticing: i) nomadic workers showed
less exposure to incident involvement, and ii) Trust in management
commitment to safety did not exert a significant impact on involvement in
incidents. And, as will be discussed later, these findings seemed to make
better sense when complemented by qualitative findings from our focus
group interviews. Also, Safety compliance exerted a significant positive
impact on safety performance and had a contributing effect on shelf
membership losing its importance when entered in the final step. As such,
the results highlight interesting theoretical dynamics between group level
characteristics, structural work factors, organizational safety aspects and
trust dimensions on safety performance. And as an overall evaluation, the
model seems to predict safety performance in a reasonable manner.
Zohar (2008) argues that climate perceptions shall reveal policies in use;
i.e. enacted practices and not their formal counterparts. He also claims that
indicators of enacted safety practices may touch upon the more tacit sides
of our perceptions. There exists disagreement, however, whether
questionnaires – even if the indicators are directed towards enacted safety
practices – are able to capture non-canonical practices and deeper rooted
cultural aspects. Some research disciplines within cultural sociology and
social anthropology would give a definite “no” to this question, while
others would claim that it is perfectly possible. As an overall evaluation,
the dimensions developed within the NORSCI questionnaire tool involve
items reflecting perceptions of safety policies, procedures and practices. A
strength of this tool, I would argue, is that its indicators are closely related
to enacted practices, that it reveals “acid-indicators” of safety when it is
under pressure (Safety versus production), that it reflects management and
company priorities, and that it encompasses’ a sensitivity towards industry
specific items, as proposed by Zohar (2008). Deeper rooted cultural
aspects, however, I argue, ought to be examined through qualitative
methods.
6.2 Safety climate and safety performance

Article I and article IV make use of three different kinds of output
measures or safety performance indicators: Risk perception and accident
69
Discussion
statistics reported to the PSA in the first study, and a self report measure of
accident involvement in the fourth study.
With regards to criterion related validity, the first paper includes a test-
retest of our anticipated safety climate-safety performance relation. On an
individual level, our findings show that all safety climate dimensions are
significantly and negatively associated with risk perception. Since accident
rates were reported at the installation level, an aggregated analysis had to
be performed. The aggregated installation level results showed, as
anticipated, negative associations between our safety climate dimensions,
risk perception and accident rates for both the 2001 and the 2003 measure,
except for the Safety prioritization construct which in 2001 indicated a
positive association with both perceived risk and accident rates. In general,
higher perceived risk on platforms is associated with lower scores on
safety climate, and the less ideal perceptions of safety climate, the more
accidents are reported. The positive relation to Safety prioritization both
with regards to risk perception and accident rates in 2001, though, may be
related to the lower homogeneity score (rwg) estimated for this dimension,
indicating higher variance than recommended for aggregated analysis, or
that the relation between Safety prioritization and accidents is not clear
cut. A similar low homogeneity score is also found for our concept Safety
versus production, which on the other hand behaves as anticipated. Hence,
this explanation does not give the total picture, suggesting that we either
have to look elsewhere for underlying contributing factors or that our
performance measures are problematic. However, the aggregated analysis
demonstrates that ten out of twelve dimensions are negatively related to
our output indicators and that the individual level analysis yielded
consistent results.
Our measure on risk perception reflects employees’ subjective judgments

of the probability of major accidents to occur on their installation, like gas
leakages, blow-outs, fires etc., but also scenarios which may threaten
personal safety like serious work accidents. Correlations between risk
perception and all other dimensions were performed both on an individual
and aggregated (installation) level and show more consistent negative and
stronger relations with our safety climate dimensions. These correlations
may be interpreted in the light of Hudson’s argument, that there might be a
relation between our safety climate scales and both personal and process
safety indicators. Hence, such dynamics would be interesting to
investigate more thoroughly and should be developed further. Our analysis
70
Discussion
does not estimate causal relations between our safety scales, risk
perception and accident rates. Thus, theoretically, the relation goes both
ways, with for instance high risk perception or a high number of accident
rates being related to low safety climate scores. However, when using
accident rates happening after the accomplishment of the survey, the
safety climate scales are at least theoretically treated as leading indicators
of safety performance.
In the fourth study a causal model was set up with involvement in

incidents as our dependent safety performance measure. The results from
the regression analysis showed that this output measure differentiated
significantly both with regards to our criteria variables, our scales and in
the overall model. In the study, three safety performance measures were
originally included and involvement in incidents represented the least
serious category of these three. Hence, the numbers reported for
involvement in incidents were generally higher than for the two other
categories; being involved in a near miss or having suffered an injury on
this installation the last year without seeking medical attention. The UK
workers generally reported higher Safety compliance and lower
involvement in incidents also for the two other measures. Hence, the
accident involvement pattern holds the same tendency when controlling
for other types of accident exposure and injury involvement. The lower
involvement in incidents among UK workers, though, represented a
surprising finding, especially when seen in the light of a generally higher
fatal accident rate on the UKCS for approximately the last 10 years. This
may indicate that company level accident rates and industrial level rates
may vary independently, but also, as suggested in our focus groups,
according to local level efforts and over time.
During the interviews it became apparent a potential bias on the reporting

of serious injuries, especially among UK workers. While Norwegian
employees keep full payment while being sick or recovering from an
accident, this is not the case for employees on the UKCS whom lose parts
of their payment. As such, different national rules regarding regulation of
absence from work may influence employees’ willingness to report
involvement in serious personal injuries. Also, offshore schedules
(involving between two and four weeks off between shift periods) may
make it possible to recover between work shifts. Hence, long off-duty
periods might make it possible to “hide” personal injuries, because
workers may have recovered until their next work period. Such contextual
71
Discussion
based bias may threaten the reliability of safety statistics making it

problematic to compare statistics across borders and industrial sectors.
Another issue which we at least know is present on the NCS concerns
contractors’ experience of pressure towards keeping the vision of “zero
accidents”. This may indicate that there exist a general industrial or sector
specific bias on the reporting of accidents – even if the reporting focus is
known to be high.
6.3 Differentiating patterns

The results from our MANOVA in study one suggest that installation
membership is among the strongest safety climate differentiators along
with work area, company and platform type. The approximate same
characteristics are found in the second study indicating that the installation
you work on matters more for your perceptions than company
membership. Hence, installation membership seems to constitute an
important aspect “framing” offshore work environments and creating
employee identification towards the place they work, live, and make
friends. It also suggests that local management efforts or local installation
“cultures” might constitute a strong safety climate promoting “force”.
In the fourth study, the causal influence of installation membership was

tested, though not showing the same pattern as in article I and II. However,
it has to be taken into consideration that sampling procedures and sample
sizes are different. Article four involves data from one company, having
employees on both continental shelves (NCS and UKCS) spread on a
number of production platforms. Also, lot fewer installations are included
from the UKCS and a low number of respondents on some of our sample
installations may have made it difficult to test our hypothesis properly.
The main finding in paper four is that both cultural and structural factors
are found to play an important role in safety performance. However, not
always as expected. Shelf membership loses its influence when Safety
compliance and Safety participation are introduced, and our group and
structural factors exercise a certain influence, even though in other
directions than expected, e.g. with nomadic employees showing less
exposure to incident involvement than more stable personnel. When
checking for exposure time, nomadic workers in the UK spent just as
much time offshore “in the last year” as other employees, while on the
NCS the nomads had a slightly lower time exposure than more stable
72
Discussion
personnel. However, also 85% of the Norwegian nomads spent between

100 and 75 percent of their working time offshore the last year. Hence,
less exposure time for the nomads is not anticipated to play a major role in
the explanation of less involvement in incidents. Some of the nomads may
even experience higher exposure time because they sometimes make
highly skilled personnel and to hold desired expertise.
Taken together, the papers show various differentiation patterns with

regards to organizational safety, trust and behavioral aspects. Generally we
find that risk, safety and trust are distributed across shelves, installations,
groups and along several structural work factors. Sometimes as
anticipated, sometimes not. Even though surprising, quantitative results
were interpreted as making sense when supported by qualitative findings,
which undermines the importance of combining qualitative and
quantitative methods.
6.4 Influence of culture, trust and distrust on

safety
Article 4 involved a cross-cultural comparison of various safety aspects in
one company having operations, crews and employees on two shelves. A
much used assumption is that organizational culture influences its
members with regards to values and practices; how things are done, how
reality is understood, what is valued, seen as right or wrong etc. In short, it
provides patterns of meaning and systems of symbols which guide our
actions. Such patterns of meaning are not always shared, however, and as
a quantitative phenomenon differentiating explanation patterns are
possible to explore, and with qualitative methods it is possible to make
sense of such patterns. With regards to quantitative cross-cultural
comparison, differences in respondents’ values and perceptions might be
linked to various sources. National culture may constitute one of these
along with organizational and crew or installation culture. As such, the
fourth paper represents a revisited comparison of the UK and Norwegian
offshore safety climates (Mearns et al., 2004), but it also implies an
extended horizon with quantitative measures being complemented with
qualitative methods seeking the actors’ interpretations of statistical results.
With regards to the functions and dysfunctions of trust and distrust in

relation to safety, the results from our focus groups show that our
informants tend to reflect similarly around the dynamics between trust,
73
Discussion
distrust and safety; independently of national culture. Functional distrust

was expressed as being important for safety and seemed to constitute a
crucial risk if not being exerted properly:
Too much trust can lead to disaster, in worst cases death (Norwegian
informant). The downside to trust is if you admire someone. Then you can
lean back too much, you are not alert (Norwegian informant).
If there is too much trust, people are not doing what they are supposed to do.
Then they are taking a chance, a risk. It is abdication of responsibility (UK
informant).
Trust, with an element of control is anticipated to be good for safety, while

too much trust is considered risky. Trusting colleagues, management or
systems too much, are given as examples of a dysfunctional trusting
attitude.
I trusted someone too much once, and it caused an accident because I didn’t
check. I was taking over for someone, and all I had to do was to press that
button. I didn’t check. I trusted him. It caused basically a spill. I always did
check myself, but for some reason I just trusted him on that occasion (UK
informant).
As such, dynamics between safety and trust seem to be related to common

“universal” trust dynamics. A trusting attitude is confirmed to involve risk
– either it is directed towards your co-workers and managers or towards
friends and family, because it may always be “lost” or exerted in a blind or
naïve manner, which may lead to unforeseen consequences and undesired
side-effects. Hence, an understanding of trust as only positive and distrust
as negative gives overly simple answers.
As expected, both Power distance indicators (“tendency of obeying your

boss even when in disagreement” and “social distance between superior
and subordinate”) and the Assertiveness dimension (measuring
aggressiveness and dominance) were found to be higher in our Anglo
compared to our Norwegian sample, which is in line with the findings in
the GLOBE study (House et al., 2004). The correlation analysis showed,
however, interesting differences across shelves between our safety
behavior scales and the GLOBE measures. For the Norwegian respondents
Assertiveness and both Power distance indicators were positively
associated with each other, and, as expected, negatively related to all other
scales. The UK results indicated, however, that Assertiveness was (non-
74
Discussion
significantly) negatively associated with one of the Power distance items

(social distance between superior and subordinates) indicating that even if
the climate may be Assertive (aggressive and dominant), social distance
between managers and employees may be small or vice versa. Also,
among UK respondents a non-significant positive correlation was found
between Assertiveness and Safety compliance, which also represents an
opposite pattern compared to the Norwegian sample. This result suggests
that among UK employees, high Safety compliance may happen despite of
an assertive climate. Since these associations were non-significant, they
should, however, be treated with care.
A surprising finding, though, is the non significant effect of Trust in

management commitment on safety performance. However, a more
nomadic and unpredictable work-life may explain why Trust in colleagues
may constitute a better predictor of safety performance than Trust in
management commitment to safety. Being situated on remote work sites,
exposed to flexible rotation patterns, a nomadic work life, combined with
temporarily isolation from family or common friendly relations onshore –
may create fertile conditions for affective (McAllister, 1995) and identity
based trust relations towards near colleagues (Coleman, 1990).
Interpersonal trust relations take time to build and for some of these
employees, their managers change continuously as well as client
management relations. Thus, a near colleague relationship may increase in
importance, maybe at the cost of Trust in management commitment to
safety. Affective and identity based trust is distinguished by reliability and
interdependence in long-term interactions, reciprocated care and concern,
and is anticipated to give rise to positive expectations about the trustee’s
intentions. Hence, the non-significant influence of Trust in management
may be explained by short-term interactions, less interdependence and less
experience with one another.
How trust and distrust elements are made important in client-contractor

relations may also play a role in this complex. Our qualitative findings
suggest that contractor employees sense a strong element of client control
when entering new installations, expressed as an almost tiring need to
prove themselves and their competence. Such institutionalized distrust and
“clan control” towards contractors may be especially important in the
beginning of a relationship, but is anticipated to loosen up as the trust
relation grows in depth and bandwidth (Lewicki, 1998). Some of these
nomadic workers are on the move most of the time. Hence, such loosening
75
Discussion
up might hardly ever happen. For nomadic employees, though, this type of
distrust may be of special importance. It may create both functional
personal and organizational alertness, which may actually enhance
organizational reliability. Near colleague relations combined with a careful
navigator attitude and a functional distrusting attitude from clients, may as
such lead to less accident exposure – even though it is experienced as
tiring.
Among UK employees Trust in management commitment to safety is

reported as higher than trust towards colleagues, while Norwegian workers
report trust towards colleagues as higher. The qualitative interviews
suggest that Trust in management patterns seen in relation to the trends of
higher Safety compliance and Power distance among UK respondents may
be related to differing management cultures across shelves and
dissimilarities with regards to rule compliance. Norwegian leaders are
understood on both shelves to be more open, easier going and emphasizing
equality in the manager-worker relation, while the UK worker-leader
relation may be more heavily based on obedience, assertiveness and
upward respect. Although some informants claimed that UK management
culture is changing, it still seems to be characterized by “shouting” on
some installations. Also, the higher UK rule compliance, Assertiveness
and Power distance are among our respondents anticipated to yield a more
rule-based trust, while lower scores on these dimensions are anticipated by
the researchers to give an identity and equality based trust in the
Norwegian employee-leader relation. Both management styles are not
associated unambiguously with positive or negative influences on safety.
Too high a degree of involvement and equality between superior and
subordinates were related to hierarchical inefficiency and indecision
among Norwegian respondents, and for the UK workers a too high degree
of rule-based trust or too high respect towards leaders/authorities were
associated with specific safety challenges. Both compliance with safety
rules and ability to critically question managers, colleagues or
management systems reliability are considered as crucial for safe
organizational performance. Hence, both a rule-based and an equality
based trust may have their upsides and downsides and need to be balanced.
Seen in relation to the hypotheses suggested by Doney et al. (1998) that

more Masculine (Assertive) and higher Power distance clusters should be
linked to a more calculative based trust building compared to their
counterparts Feminine (non assertive) and low Power distance cultures,
76
Discussion
our results could be said to agree to a large extent with this reasoning.
Rule-based trust may correspond well with or be closely related to a form
of trust based on a calculative way of reasoning, which is a more
prominent and outspoken characteristic among our UK informants.
Obedience may, for instance, be used and appreciated both as effective
and as a strategy to keep in with the boss among UK employees, while
Norwegian managers may be more willing to accept critical comments
regarding their work in correspondence with a non-assertive and low
Power distance culture. Values among Norwegian workers related to
equality, identity and democracy may also be understood in the light of
values and principles highly appreciated in the “Nordic model” of tripartite
cooperation (Lindøe & Karlsen, 2006). Differing union cooperation
traditions within the UK and Norwegian context are also mentioned in the
Mearns et al. study (2004) indicating lower trade union membership in the
UK compared to the Norwegian sector. Such a difference seems to be
apparent also today with stronger emphasis on the role of the labor unions
and higher union membership rates on the NCS compared to the UKCS.
Taken together this may indicate that the level and form of trust may differ
across shelves and that these patterns may overwhelm organizational
membership.
6.5 Industrial culture

Some see accidents as inevitable and claim that the combination of our
desire to extract, produce and refine energy, the technology we use to do
so, and the complex organizational interfaces that have been generated,
yield new systemic risks and accident pictures. Human perception is
limited and humans make errors. Accidents then become hard to predict,
organizational and technological systems become (too) complex to
comprehend, leaving them difficult to control (Perrow, 1999). Others
claim that accidents are avoidable, that a combination of thorough safety
management systems, a strong emphasis on managers as key role models
combined with individual and organizational mindfulness will create
highly reliable organizations (Weick & Sutcliffe, 2001). The latter view
corresponds better with the vision of error free performance and zero
accidents than the first, and as such with the oil and gas industry – without
concluding that one is right and the other is wrong. The zero vision is
often used in this industrial sector as a leading star; hardly being
questioned neither as a global nor an organizational goal.
77
Discussion
The point of departure in article III stems from a debate being raised
within the Norwegian oil and gas industry regarding behavioral
approaches being accused of leading into blame games, but also as a
debate continuing in the field of safety research. As my elaboration has
shown, there exist differing scientific routes to safety improvements and
interventions. A concern raised from employee unions in several
collaboration arenas in Norway has been that companies’ safety efforts
have been too one-sided and with a too heavy focus on individual safety
behavior. The claim has been that company management often chooses the
cheaper solution of urging employees to engage in safe behaviors instead
of investing in safer technology or work environment improvements. Or
when companies have spent huge amounts of money on safe
culture/behavior programs, it has been claimed that resources could have
been better spent elsewhere. The establishment of our two-by-two table
focusing culture-behavior and tacit-explicit aspects enables us to pinpoint
the implications of differing safety intervention approaches. I argue for a
balanced and multi-method approach when trying to understand and
improve safety. Formal (canonical) practice may not always correspond
with informal (non-canonical) practice, and in order to define how to solve
a problem one ought to get hold of a rich understanding of it.
Safety is highly prioritized within the oil industry. On the NCS this
tendency appears in overall high safety climate scores, our longitudinal
results show decreased variance in perceptions and attitudes, and a
considerably large drop in accident rates. However, oil prices have been
high, providing the companies with an anticipated strong capacity to
afford the valuation of safety. But how do we keep commitment high if
safety should come under pressure? Oil prices fell drastically during the
financial crisis, and we still cannot say whether this is a temporary state or
if it has come to stay. We also know from accident investigations what
happens when safety is devalued (Baker, 2007), and, as such, global
market dynamics may affect safety more than cultural differences in
perceptions or individual safety attitudes.
For future business on the NCS there are challenges to be dealt with: The
shelf is mature, installations are growing old and rusty (gives greater need
of maintenance), integrated operations are increasingly used involving
higher dependency between off- and onshore units and exploration and
drilling is moving into deeper and vulnerable waters and environments. In
a situation with lower or decreasing oil prices, there will probably also
78
Discussion
emerge a need to run operations more efficiently, but hopefully with an

eye for potential safety outcomes. As I see it, economy makes perhaps the
most important external framework condition for companies’ abilities,
capacities and ambitions to act safely.
A predominant characteristic of our modern world is that we seem to be

geared towards an extensive development and use of repertoires of rules
and routines that create order and predictability (Rip, 1991). The oil and
gas industry makes no exception; having to control, reduce and remove by
various means organizational risk and safety. What we sometimes seem to
be blind to, though, is that rules, procedures and systems always meet and
have to face a reality where dilemmas live and paradoxes develop, or as
one of our informants put it: It would have been nice with a small hole in
the wall, where I could place my finger, get it cleaned, treated and nothing
reported. (Haukelid, Tharaldsen, Lamvik, Bye & Dahle, 2006). As all
other companies and facilities operating offshore, this installation had an
excellent medical service onboard, but the workers seemed hesitant to use
it, most likely because of fear of destroying company or client statistics. In
public statistics, injuries are categorized according to seriousness.
Together with the challenge of underreporting there also seem to be a
tendency of negotiating the categorization of the injury towards the least
serious type, transforming the management system to actually fit with
reality but away from its original purpose – organizational learning from
errors in order to protect workers from harm (ibid.). As such, it represents
a sociological phenomenon illustrating an industry with an extremely high
prioritization of safety, while at the same time living with paradoxical
consequences of well intended ambitions. Unity, shared norms and values
are highly valued organizational properties, but real life safety practice
show that risks are socially and structurally distributed and that everyday
work practice tend to be filled with dilemmas and unforeseen paradoxes.
Offshore facilities and fields are very different; i.e. varying from huge
complex interconnected production fields with a high number of
employees – to smaller installations, floating production units etc. framing
more intimate and perspicuous work environments. What they
nevertheless have in common is that work is accomplished on remote
worksites (“in the middle of nowhere”), management commitment to
safety is exercised locally, and that employees often share approximately
the same goals, common dangers and high risk scenarios. Installation and
crew membership is in that respect wrapped in family metaphors and trust
79
Discussion
dynamics towards near colleagues and first line supervisors in the fourth
study. Turner (1981) has earlier shown that isolation and sharing of
dangers may create occupational subcultures, and that such subcultures
have specific functions:
The subculture provides shared perspectives which enhance the group’s

control over their work situation, which maximise autonomy and minimize
dependence upon outsiders. […] group norms are very important both in
controlling and testing new members and in encouraging them to behave
predictably in the face of danger (Turner 1981:67, from Rip, 1991, pp. 354-
355).
Such forces may also contribute to the understanding of our findings and
to my claim that offshore installations should be treated as important
culture creating “places”, enabling groups to continue with their activities.
Miners continue to go underground and offshore workers continue to go
offshore, even if we as outsider may perceive such environments as ever-
present with danger – but that’s why we, according to Rip, are outsiders
(ibid.:355).
80
Limitations and implications
7 Limitations and implications

This chapter draws up main limitations and implications for the thesis with
regards to methods and measurements, cross-cultural comparisons, safety
performance indicators and dynamics between safety, trust and distrust.
Finally, implications for safety practitioners will be outlined.
7.1 Common method variance and potential bias’

According to Podsakoff et al. (2003) bias may be produced by the
measurement items themselves, the context of the items within the
measurement instrument, and/or the context in which the measures are
obtained (ibid., p. 881). With regards to predictor or criterion variables, the
most common sources of method bias are anticipated to result from the
fact that they are obtained from the same source. Nevertheless, various
kinds of self reports are often used within safety climate surveys –
frequently due to lack of suitable data or to some other concern; e.g. costs,
that collection may be too time consuming or that other measures may be
inflated with other kinds of biases. Spector (2006), on the other hand,
challenges the widely accepted view that variables collected with the same
method are inflated due to common method variance (CMV), often
referred to as mono-method bias. He argues that although several authors
have noted that the CMV problem is overrated (e.g., Spector, 1987, 1994;
Crampton & Wagner, 1994; Lindell & Whitney, 2001; from Spector,
2006), it seems that this automatic criticism of cross-sectional self report
has become invoked so broadly and automatically that it has achieved the
status of a methodological urban legend (p. 222). Spector suggests that the
term CMV should be abandoned in favor of focus on measurement bias
that is the product of the interplay of constructs and methods by which
they are assessed (p.221). This leads to a view that, regardless of source,
self report variables (for instance our self report surveys) or performance
indicators may be associated with different kinds of method biases, an
approach of which I find meaningful.
In this thesis, the installation level accident rates constitute an

independently collected criterion variable (of safety performance), while
all others stem from self report surveys. Leaning on Spector’s suggestion
then both self report variables and independently collected performance
indicators may be associated with different kinds of method biases. With
81
regards to the data collection methods used in this thesis, I will point out
two main sources of bias and how they are dealt with: Social desirability
and contextual factors.
Social desirability is known as one of the most common sources of bias

that may affect the validity of survey results. As a phenomenon it reflects
the tendency of subjects to deny socially undesirable traits, to claim social
desirable ones and to say things which place the person in a favorable light
(Nederhof, 1985), referred to by Goffman (1992) as image or impression
management. As such, measuring organizational safety dimensions within
an industry which prioritizes safety highly provides us with a challenge. It
may both constitute an individual problem and a company level challenge
regarding image and impression management. Neutral items are difficult
to formulate, because the respondent will know what he/she is supposed to
answer or what might represent the most favorable answer. A
recommended method used to avoid such bias, is the use of self-
administered questionnaires. However, it has not been shown to remove
the problem completely (Nederhof, 1985).
Research design in all of the quantitative studies consists of self-

administered or self-completion questionnaires. No personal information is
asked for, which might have revealed the identity of our respondents, even
if some demographic background data has had to be treated with care, e.g.
gender, age, discipline, installation, company name etc. In order to
reassure respondents and participants, main aims of the research
projects/surveys were given in information letters or in the questionnaire.
It might also have been perceived credible that the accomplishment of the
self report surveys, analysis of the results and the reporting of them was
done by an independent actor (IRIS) and not by the companies or the
Petroleum Safety Authorities themselves. Questionnaires were also sent
back in closed, pre-stamped envelopes, and respondents were promised
anonymity.
With regards to the gathering of qualitative data in study four, the research
team put up a list of desired participants for the focus groups, but the final
selection was done by the company’s own human resource personnel and
company management. This procedure might have involved a sense of
obligation to participate. Also during focus groups, it may be experienced
as difficult for subordinates to speak freely about their “true” safety
priorities and problems when superiors are present. We also risked that
82
company representatives chose agreeable employees, in order not to

compromise the company. However, the research team’s own evaluation
was that participants spoke freely during interviews. Management may
however be more comfortable in taking the lead in discussions, but we
also found that this was not totally dependent on whether informants were
managers or not, but dependent on personal style and personality.
The last decades the industry show enormously improved injury statistics,
which despite of underreporting issues must be understood as a general
safety improvement. Combined with high safety climate scores and less
statistical variance, we may statistically talk of an industrial “ceiling”
effect, which of course is positive for safety, but leaves it more
challenging with regards to statistical analysis and further industrial
improvement. Even though the level is high, this thesis shows that
differentiation patterns exist. A future challenge will be to gain a better
understanding and methods on how to work with exposed groups and still
challenging safety aspects. And finally, rather than emphasizing the
excellent sector safety performance, it may be more important to focus on
maintenance of an already high level of safety and maybe functional trust
and distrust would be useful tools in such an approach.
In section 6.2 and 6.5 I also pointed out contextual issues that might lead
to reporting bias with regards to our safety performance indicators. These
problems were both related to the industrial sector as such (paradoxical
consequences of the zero vision) and national differences across shelves
regarding payment when being off from work (loss time accident) because
having experienced a personal workplace injury. Implications of these
challenges will be discussed below (7.4).
7.2 Sample sizes

The four samples involved in the thesis are of different sizes and the
targeted populations differ. In study 1 and 2 an entire industrial sector is
covered at three points in time, while study 4 includes surveying a group
of employees from one company having common work tasks across
shelves (contractor employees working within drilling and well-service) at
one moment in time. The large samples in the first and second study
provide us with a unique gateway to the understanding of occupational
health and safety issues for an entire industrial sector, which is different
from the more common way of investigating safety climate, namely with
83
company/organization or installations/group level units as point of

departure. Even though the response rates are quite low (ca 50%),
background demographics follow the approximately same pattern across
all three measurements (2001, 2003 and 2005), which indicate that the low
response rates should not be considered a large threat. It should be
commented though that the 2001 sample have a shorter survey
accomplishment period (two weeks), while from 2003 onwards the entire
population of offshore employees are actually included. The response rate
in study 4 is considered to be fairly good (67%).
Differences in sample sizes leave us with different challenges with regards

to the probability of making type I or type II errors. With large samples,
the chance of making type I errors increases. Hence, differences between
groups or measurements may come out as “significant” or systematic –
even though they may be very small. Hence, one may more easily
conclude with for instance an improvement or worsening in safety climate
scores from one measurement to another even though the null hypothesis
should have been kept. One may control such challenges by introducing
stricter confidence intervals or just to pay less attention to tests of
significance. In short, large samples make it easier to confirm “significant”
results or differentiation patterns than in smaller samples. However, the
larger samples, the better our chances are for generalization.
In article IV the opposite problem emerges. Some of our hypotheses

involve examination of group level data, for instance our regression model
and test of differences across shelves and installations. Due to a lower
number of installations participating from the UK shelf and a
corresponding lower number of UK respondents, a proper examination of
our hypothesis of installation membership was shown to be statistically
difficult. Hence, our hypothesis had to be rejected, but the result may just
as much be a result of sample size than our anticipated installation
membership influence not existing, and might have shown other results
with a larger sample.
The fact that our sample in article IV represents data from only one
organization, our possibility of drawing representative conclusions
remains limited. However, taking into consideration that we deal with a
contractor company and employees working with drilling and well-
services, we have no indication that this organization is a very unique one,
or that the results would have been largely different with a sample with
84
employees coming from a comparable contractor company or from a

collection of contractors. However, it would be worth investigating,
though, whether another sampling technique would give other types of
results. I will also argue that our research design, where the intention of
examining work and safety exposure for a special group of employees
have proven useful, and that the combination of qualitative and
quantitative methods was fruitful. Some of the findings related to our
GLOBE organizational culture scales also behave as anticipated compared
to larger representative, cluster samples.
As an overall methodological issue, it becomes obvious when performing

factor analyses (both EFA and CFA) that the researcher has to exercise her
own judgment when deciding which items are to be included or excluded
from a dimension. Different models can always be tested and other
factorial solutions may have given other results or solutions, and
hypotheses could have been differently angled and formulated. Hence,
one’s perspective, anticipated dynamics between certain concepts and
hypothetical models, put certain opportunities and restrictions in place
both with regards to what you look for and what you find. An interesting
finding, though, is related to the fact that our safety climate measures in
the NORSCI tool were replicated at all three measurements. As such, we
may argue that safety climate may be meaningful investigating at a higher
level than the organization, as in our case at a sector or industrial level,
and that our concepts seem to behave relatively stable through time.
7.3 Cross-cultural comparison

Cross-cultural comparison is a challenging exercise and should be done
with care. The work of Hostede, which the GLOBE dimensions are a
further development of, has both its followers and opponents. When
reading both Hofstede and the work accomplished in the GLOBE study,
there seems to be an awareness of strengths, weaknesses and potential
pitfalls regarding cross-cultural comparison, and dividing the world (in the
GLOBE study 62 participating nations) into 10 clusters according to their
perceived values and practices across 9 cultural value and practice
dimensions, necessarily involves simplification. Differences within
clusters are being undermined, cultural processes hard to grasp empirically
are not seen as important, and the link between individual, organizational
and societal level data characteristics may become blurred. Several critical
voices have given sound examples of how cultural scores on dimensions
85
like Collectivism, Confucian dynamism etc. give too simple answers and
undermine contrasts and paradoxes within cultures (Tayeb, 2001; Fang,
2003). Cluster analysts, their predecessors and contemporaries have also
basically been accused of belonging to a structuralist, cross cultural
functionalism accomplishing ethnocentric social science from a western
worldview – treating living, thinking, learning adaptive systems in a
mechanistic way (Lowe, 2001). So, the critics have both been directed to
the researcher’s worldview, the theoretical validity of the instruments and
the empirical consequences of using them – with regards to potential
fallacies.
Taking such critics into consideration, I have tried to cultivate an

awareness of the problems associated with them, not to overestimate
differences and to make explicit what kind of level the results reflect;
national, organizational or individual level. Also, when gathering and
analyzing qualitative data, a more nuanced picture emerges. Our findings
with regards to our two GLOBE scales suggest, interestingly, that
differences between our UK and Norwegian samples on Power Distance
and Assertiveness are in accordance with general findings from the
GLOBE study, if we use the Nordic and the Anglo clusters as reference
groups. When corroborated with focus groups interviews implications of
these finding were in specific ways linked to safety results and differing
leadership styles across shelves. The GLOBE scales also came up with a
reasonable factorial solution (Tharaldsen et al., 2008), which indicated that
the dimensions “fitted” across national membership. This may, however,
have been more of a challenging exercise if participating nations differed
substantially. Whether factorial solutions, theoretical concepts and models
are transmissible across cultures constitutes a research area that needs
further investigation. It has for instance been claimed recently that
Western-European safety and risk models are challenging to apply in
substantially differing cultural settings (Lund & Rundmo, 2009), while
others argue for the transferability and universality of safety climate
instruments across sectors (Flin, Burns, Mearns, Yule & Robertson, 2006,
Flin, 2007; Olsen, 2009).
Another issue may relate to the understanding of the indicators used. The
petroleum industry is global in its nature. As such, the indicators used
seem to be well integrated into company management systems, and they
are apparently the same all over the world. Nevertheless, we do not know
for sure whether our respondents understand our concepts, wording of
86
items and indicators exactly the same way. Accident statistics have been
difficult to compare, both on a national and international level, and to be
associated with bias both with regards to underreporting and differing
cultural frameworks (Lamvik & Ravn, 2006; Bye & Lamvik, 2007).
However, participants in the fourth study all belonged to the same
company, their safety management systems are the same for all
employees, and there seemed to be extensive cross-cultural contact
between the UK and Norwegian offices. Such contact and common
management communication may increase common understandings across
shelves and in the long-run facilitate important safety learning across
cultures.
7.4 Safety performance indicators

Safety climate dimensions have in my papers either been shown to be
negatively associated with or to hold predictive qualities towards three
different kinds of output measures: accident rates, risk perception and
incident involvement. Within the Norwegian and UK industrial sector
reporting of work related injuries is regulated by law. However, public
statistics are generally supposed and shown to be marked with
underreporting problems and dependent on the incentives of reporting
(Hurst et al., 1996). Whether underreporting is an equally large issue
within the oil and gas industry as a whole, remains an unanswered
question, but the general reporting practice is anticipated to be well
integrated into companies’ management systems and practices.
With accident rates playing such an important role with regards to

company identity and reputation, “bad statistics” means intensified
attention towards local level management of safety, often in the form of
individual training or compliance programs combined with various kinds
of incentives in order to improve safety performance. Such local level
efforts and safety management are expressed as yielding positive results,
but well intended incentives may not always have unambiguous effects.
The most important challenge may be related to the maintenance of high
efforts when scores are good and to avoid falling into self complacency.
Generally, contractors are dependent on having good safety scores, if they

are to be considered as qualified partners in contract negotiations. Hence,
there may be stronger underreporting incentives among contractors than
among client companies, which by and large defines contractor frames and
87
decides what is important. Such mechanisms combined with the above

mentioned differing national frameworks regarding losing or maintaining
full payment when having a lost time accident, may result in reporting bias
in public statistics. With regards to defining a meaningful output measure,
there seems to be fewer problems associated with the reporting of less
severe incidents than the more serious ones – except for the reporting of
fatalities which should be hard to question. Rather than only depending on
public statistics, I suggest that a less severe self–report-performance
measure may be used and that such self-reports may be associated with
smaller bias’ problems. Such self-report indicators have also been
recommended in other studies (Hurst, Young, Donald, Gibson, Muyselaar,
1996). Also, if we remind ourselves of the characteristics of a valid and
reliable safety performance indicator, a self-report indicator is quantifiable
and permits statistical inferential procedures, and it seems to be a valid
indicator of individual safety performance providing good variability.
Further, it also seems to be sensitive to context and behavioral conditions,
the costs of obtaining and using such measures are small and consistent
with the benefits, and it seems to make an indicator well comprehended by
the respondents who are going to use it (Rockwell, 1959).
7.5 Dynamics between safety, trust and distrust

With regards to safety, the results show that our focus group participants
tend to reflect upon the functions and dysfunctions of trust and distrust
similarly – surpassing national background and organizational
membership. As such, trust and distrust make a research area which is of
high relevance for safety. Both trust and distrust seem to be used in
parallel to minimize uncertainty, reduce risk, to promote (social) order and
predictability with regards to safety, which constitute dynamics well in
accordance with the luhmanian way of conceptualizing trust and distrust.
Similar ways of talking about and reflecting upon the functions and
dysfunctions of trust and distrust may also be related to the fact that these
groups of workers share much the same dangers and challenges with
regards to safety. Then direct and near challenges with regards to their
proximate work environments, like being situated on oil and gas
installations, dealing with similar types of accident exposure, comparable
work tasks, technologies etc., may constitute an important source of
similarities – rather than company membership or national culture.
88
On the other hand, I have also shown that trusting and distrusting make an
important issue when your near environment becomes more unpredictable or
characterized by higher degree of instability. In line with theoretical
reasoning, increased uncertainty will most likely affect trusting relations –
both with regards to personal, organizational and institutionalized trust. Near
colleagues may become more important, shifting work rotations yield higher
management and client control, and nomadic workers seem to become
comparable with careful drivers in new work environments. Such
precautionary behavior and alertness give higher predictability (and trust)
which in general is regarded as crucial for safety and reliability.
In figure 3 I have developed a model where I try to illustrate and summarize

my findings with regards to the functions and dysfunctions of trust for safety.
Functional trust seem to yield a trusting, but verifying attitude, while
dysfunctional trust is associated with naivety, blindness and a too high degree
of agreeability. Risk blindness happens with complacency, and a too
agreeable atmosphere may yield colleagues hesitating to correct each other or
management – even if it should be necessary. Functional distrust, on the other
hand, would give a precautionary strategy, while dysfunctional trust comes
with too high degree of monitoring and control. Dysfunctional distrust may
lead to a desire for sabotage and revenge or to low worker commitment and
involvement.
Figure 5 Functional and dysfunctional aspects of trust and distrust5
5
The model represents an adjustment and simplification of work done by Pidgeon,
Walls, Weyman and Horlick-Jones’ (2003) and Conchie & Donald (2006) on the
functions of safety specific trust and distrust.
89
As mentioned earlier, the positive impacts of trust and negative influence

of distrust is well documented in former research, while it remains to give
higher attention especially towards the potential dysfunctions of trust and
the positive functions of distrust. When communicating dynamics
illustrated in the model above, the benefits of trust are more or less self-
explanatory. The negative impact of dysfunctional distrust is also quite
obvious. Too much trust and blindness is also a recognizable phenomenon.
Functional distrust, however, may constitute a more challenging
phenomenon to communicate the meaning of, much because distrust in
itself gives negative associations. As such, the model illustrates the
potential positive influence that distrust may have for safety. Hence, trust
as ‘good’ and distrust as ‘bad’ give a too simplistic picture of reality and
rather than a normative view on trust, I will suggest taking into
consideration how trust, distrust and control elements are being balanced
and how they are being restrained or promoted by contextual factors at
various levels.
7.6 Implications for safety practitioners

The articles focus on aspects which may be of importance for safety
practitioners, and, as I see it, there are no simple answers or quick fixes,
but some important issues to reflect upon. What kind of organizational
factors are important for safety and how might we improve them? How
could cross-cultural differences be dealt with? How are safety and risk
distributed, and what kind of implications may such differences have?
How do we know what safety level an organization is at? How do we
maintain a high level when safety climate scores and safety performance is
good? What kinds of dangers are associated with complacency? What is a
sound indicator of our organizational safety performance?
Organizational decision makers operating in global environments have to

deal with differences both in cultural values and differing contextual risk
levels. “Safety” is not the same in Norway, Nigeria, Brazil, Canada or in
the UK. As such, this thesis has touched upon the potential consequences
of cultural differences between the NCS and UKCS and how such
differences might influence organizational safety aspects, trust building
relations and safety performance. As a main result safety compliance
seems to constitute an important organizational aspect that should be
properly dealt with, also owing to the fact that this dimension seems to
moderate the effect of differences across shelves in the fourth study.
90
Safety compliance will be dependent on how safety is being prioritized in

daily work practice, and especially how “truly” it is being practiced by, for
instance, local installation management and work crews. Further, it will
probably also be dependent on work pressure and employees’ possibility
of actually giving safety priority, for instance when employees and
managers are under time pressure.
Article IV also shows that culture may have an influence on both the form
and level of trust, but that these differences do not have an unambiguous
positive or negative effect on safety performance. This indicates that what
is considered “the way to do it” in the UK would probably not be the best
way of doing it on the NCS; high assertive leader styles or an introduction
of high Power distance would probably neither be appreciated nor
accepted among Norwegian workers or vice versa. This indicates that
cultural differences are hardly possible or beneficial to directly translate or
transfer. However, differences should be recognized and cultural
awareness should be appreciated and developed.
Another important issue is the delicate relationship between trust, distrust

and other institutional control elements: When does functional trust
become dysfunctional, and how may control elements (functional distrust)
meant to protect become destructive (dysfunctional distrust)? Complex
supply chains constitute the industrial modus operandi. When safety is
constructed both in horizontal and vertical interfaces, organizational
borders become too limited. This makes relations between clients and
contractors crucial. The positive sides of trust are fairly well documented
and proven (see chapter 3.5). However, the downsides to trust and the
beneficial sides of distrust need further examination. Hence, safety
practitioners would probably benefit from reflections on how safety is
taken care of in organizational interfaces and what kind of role trust
relations play, for instance between clients and contractors.
Three of the four papers show that safety is distributed across various
structural factors, such as, installation membership, company types
(contractor or client), disciplines, rotation types etc. For safety
practitioners it becomes important to know and understand how different
groups are exposed. This makes it important to not only acquire local
knowledge, but also to understand how systemic and formal elements are
being anchored in actual (informal) work practice. Hence, safety
improvements ought to be developed as a combination of systemic and
91
“tailored” approaches. Safety officers who rely on accident statistics or

quantitative data alone would have poor grounds for decision making. And
as for researchers, a multi-method approach would be beneficial.
92
Conclusions
8 Conclusions
The first study provided a five concept solution for two measurements, and
was extended with a sixth dimension in the second study. The safety
climate dimensions are: (i) Safety prioritization, (ii) Safety management
and involvement, (iii) Safety versus production, (iv) Individual motivation,
(v) System comprehension, and (vi) Safety competence. Criteria related
reliability and validity are satisfactorily met in all three measurements.
The dimensions behave consistently, differentiate in a reasonable manner,
and they are shown to be negatively associated with two safety
performance measures: accident rates and risk perception.
Potential reporting bias problems regarding public and company accident

statistics have been shown to be important considerations. My
recommendation goes in direction of a self-report measure to be included
as a substitute or as a supplement to public accident statistics in self-
completion questionnaires; a self report of involvement in incidents or
accidents.
Differentiating qualities have been examined in various ways in three

articles; the first, second and fourth. Both quantitative and qualitative
results show that installation membership is one of the most important
characteristics “framing” employees’ perceptions and offshore work life in
important ways. Other structural factors like type of company, work area,
type of installation and work factors related to flexibility or stability have
also been shown to have a significant impact on safety attitudes and safety
performance. Hence, the distributed character of safety climate and its
impact on safety performance is quite well documented in the thesis.
The epistemological roots of cultural and behavioral perspectives are very

different, so is their point of departure with regards to methodology. A
cultural approach often has a holistic and inter-subjective starting point,
while BBS approaches are mainly concerned with observable behavioral
outputs. Our two-by-two table focusing on culture-behavior and tacit-
explicit aspects points at the implications of differing safety intervention
approaches. Hence, I would argue for a balanced and multi-method
approach when trying to explain, understand or improve workplace safety.
93
Conclusions
Nomads and employees with more unpredictable shift rotations were

shown in our study to be less exposed to incident involvement than other
employees. A consequence of such shifting work demands, I argue, may
yield employees that become a combination of more flexible and “careful
navigators”. These workers also seem to experience a higher degree of
“soft” or institutionalized control from client companies. Hence, a
nomadic position may enhance both personal and organizational alertness.
Cultural differences are found across shelves with regards to both the level
and form of trust. Organizational culture among the UK respondents is
distinguished by higher Assertiveness and Power distance compared to the
Norwegian sample climate. Taken together with higher scores on Safety
compliance among UK respondents, the quantitative and qualitative results
indicate a more rule-based trust with regards to safety and between UK
workers and managers, while being characterized by equality and a
common identity among our Norwegian workers and their superiors. None
of these characteristics are found to have a clear effect on safety
performance.
Reflections on our model of functional and dysfunctional trust/distrust

show similar results on both shelves. Trust with an element of control, and
functional distrust are expressed by our informants as being good for
safety. The challenge of following rules or procedures blindly, respecting
authority too much, being too agreeable, not being alert enough or trusting
the system too much are all examples given by our informants on the
potential downsides to trust, expected to produce poor safety performance.
94
References
References
Adler, P.S., 2001. Market, Hierarchy and Trust: The Knowledge Economy
and the Future of Capitalism. Organization Science 12(2), 215-234.
Alvesson, M., 2002. Understanding Organizational Culture, Sage

Publications.
Anderson, E., McGovern, P.M., Kochevar, L., Vesley, D., Gershon, R.,. 2000.
Testing the Reliability and Validity of a Measure of Safety Climate. Journal
of Healthcare Qualit, 22(2), 19-24.
Andrews, K.M., Delahaye, B.L., 2000. Influences of knowledge process in

organizational learning: The psychological filter. Journal of Management
Studies, 37(6), 797-811.
Baker Report, 2007. The Report of the BP US Refineries Independent Safety

Review Panel.
Bakker, M., Leenders, R.T.A.J., Gabbay, S.M., Van Engelen, J.M.L., 2006. Is
trust really social capital? Knowledge sharing in product development
projects. Learning Organization, 13 (6), 594-605.
Barber, B., 1983. The Logics and Limits of Trust. Rutgers University Press.
Barrick, M. R., Mount, M. K., 1991. The big five personality dimensions and
job performance: a meta-analysis. Personnel Psychology Inc., 1-26.
Beck, U., 1991. Risk Society. Towards a New Modernity. New Dehli: Sage.
Translated from the German Risikogesellschaft published in 1986.
Blaikie, N., 2000. Designing Social Research. The Logic of Anticipation.

Polity Press, in association with Blackwell Publishers.
Blaikie, N., 2007. Approaches to social enquiry. Polity Press. Cambridge-UK,

Malden-USA. 2nd edition 2008.
Blumer, H., 1956. Sociological Analysis and the “Variable”. American

Sociological Review 21(6), 683-690.
95
References
Brannen, J., (ed) 1992. Mixing Methods: Qualitative and Quantitative

Research. Avebury: Aldershot.
Brannen, J., 2004. Working qualitatively and quantitatively. Chapter 14 in

Seale, C., Gobo, G., Gubrium, J.F., & Silverman, D., (eds), Qualitative
Research Practice. Sage Publications, pp. 312-326.
Braut, G.S., Lindøe, P., 2008. Risk Regulation in the North Sea: A common
law perspective on Norwegian legislation. WorkingOnSafety, September 30
October 4, 2008, Crete, Greece.
Browne, M.W., Cudeck, R., 1993. Alternative ways of assessing model fit. In
K.A. Bollen & J.S. Long (Eds.): Testing Structural Equation Models. Sage
Publications.
Bruhn, A., 2006. The inspector’s dilemma under regulated self-regulation.

Policy and Practice in Health and Safety, 4(2), 3-23.
Burns, C., Mearns, K., McGeorge, P., 2006. Explicit and Implicit Trust
Within Safety Culture. Risk Analysis, 26(5), 1139-1150.
Bye, R., Lamvik, G.M., 2007. Professional culture and risk perception:
Coping with danger onboard fishing boats and offshore service vessels.
Reliability Engineering and System Safety, 92, 1756-1763.
Carroll, J., 1998. Safety culture as an on-going process: culture surveys as

opportunities for enquiry and change. Work & Stress, 12 (3), 272-284.
Child, J., Möllering, G., 2003. Confidence and Active Trust Development in
the Chinese Business Environment. Organization Science, 14(1), 69-80.
Clarke, S., 2006. The Relationship between Safety Climate and Safety
Performance: A Meta-Analytic Review. Journal of Occupational Health
Psychology 11(4), 315-327.
Collinson, D.L., 1999. Surviving the Rigs: Safety and Surveillance on North
Sea Oil Installations. Organization Studies, 20(4), 579-600.
Conchie, S.M., Donald, I.J., 2006. The role of distrust in offshore safety
performance. Risk Analysis, 26(5), 1151-1160.
96
References
Conchie, S.M., Donald, I.J., Taylor, P.J., 2006. Trust: Missing Piece(s) in the
Safety Puzzle. Risk Analysis, 26(5), 1097-1104.
Conchie, S.M., Burns, C., 2008. Trust and Risk Communication in High-Risk
Organizations: A Test of Principles from Social Risk Research. Risk Analysis,
28(1), 141-149.
Conchie, S. M., Donald, I.J., 2008. The functions and development of safety-
specific trust and distrust. Safety Science, 46(1), 92-103.
Cox, S., Cox, T., 1996. Safety, Systems and People. Butterworth-Heinemann.
Cox, S., Flin, R., 1998. Safety culture: Philosopher’s or man of straw? Work
& Stress, 12(3), 189-201.
Cox, S.J., Cheyne, A.J.T., 2000. Assessing safety culture in offshore

environments, Safety Science, 34 (1-3), 111-129.
Cox, S., Jones, B., Rycraft, H., 2004. Behavioural approaches to safety
management within UK reactor plants. Safety Science, 42 (9), 825-839.
Cox, S., Jones, B., Collinson, D., 2006. Trust Relations in High-Reliability
Organizations. Risk Analysis, 26(5), 1123-1138.
Cooper, M. D., 2000. Towards a model of safety culture. Safety Science,

36(2), 111-136.
Cooper, M. D., Phillips, R. A., 2004. Exploratory analysis of the safety

climate and safety behaviour relationship. Journal of Safety Research 35(5),
497-512.
Crampton, S.M., Wagner, J.A., 1994. Percept-percept inflation in micro-

organizational research: An investigation of prevalence and effect. Journal of
Applied Psychology, 79, 67-76.
Creswell, J.W., 1994. Research design, Qualitative and quantitative

approaches. Thousand Oaks, London, New Delhi: Sage Publications , Inc
Cronbach, L. J., 1951. Coefficient alpha and the internal structure of tests.
Psychometrika, 16 (3), 297-334.
97
References
Das, T.K., Teng, B.S., 1998. Between Trust and Control: Developing
Confidence in Partner Cooperation in Alliances. Academy of Management
Review, 23(3), 491-512.
Dedobbeleer, N., Beland, F., 1998. Is risk perception one of the dimensions of
safety climate? In A. Feyer, & A. Williamson (Eds.), Occupational Injury:
Risk, Prevention and Intervention. Taylor & Francis, London.
DeJoy, D.M., 2005. Behaviour change versus culture change. Divergent

approaches to managing workplace safety. Safety Science, 43(2), 105-129.
Denzin, N., 1970. The Research Act in Sociology. London: Butterworths.
Depasquale, J. P., Geller, E. S., 1999. Critical Success Factors for Behavior-
Based Safety: A Study of Twenty Industry-wide Applications. Journal of
Safety Research, 30(4), 237-249.
Deutsch, M., 1958. Trust and suspicion. Journal of Conflict Resolution 2,

265-279.
Devillis, R.F., 2003. Scale Development. Theory and Applications. Second

edition. Applied Social Research Methods Series. Volume 26. Sage
Publications. Thousand Oaks – London – New Dehli.
Diamantopoulos, A., Siguaw, J.A., 2005. Introducing Lisrel. A guide for the
Uninitiated. Sage Publications. London, Thousand Oaks, New Dehli. First
published in 2000.
Dirks, K.T., Ferrin, D.L., 2002. Trust in Leadership: Meta-analytic findings

and implications for research and practice, Journal of Applied Pyschology,
87(6), 611-628.
Doney, P.M., Cannon, J.P., Mullen, M.R., 1998. Understanding the Influence
of National Culture on the Development of Trust. Academy of Management
review, 23(3), 601-620.
Douglas, M., Wildavsky, A., 1983. Risk and culture. An Essay on the
Selection of Technical and Environmental Dangers, First paperback printing.
University of California Press Berkeley and Los Angeles, California.
98
References
Dyer, J.H., Chu, W., 2003. The Role of Trustworthiness in Reducing

Transaction Costs and Improving Performance: Empirical Evidence from the
United States, Japan and Korea. Organization Science, 14(1), 57-68.
Eide, I., Brandsdal, E., Bjorseth, O., 1992. Subjective and objective
evaluation of the indoor environment in an enclosed 24-hour society on an
offshore installation. Environment International, 18(4), 371-379.
Eikemo, T.A., Høyvarde Clausen, T., (eds), 2007. Kvantitativ analyse med
SPSS. En praktisk innføring i kvantitative analyseteknikker. (Quantitative
analysis with SPSS. A practical introduction to quantitative analysis
techniques). Tapir akademisk forlag.
Fang, T., 2003. A Critique of Hofstede’s Fifth National Culture Dimension.

International Journal of Cross Cultural Management, 3(3), 347-368.
Flin, R., Mearns., K., O’Connor, P., Bryden, Rl., 2000. Measuring safety
climate: identifying the common features. Safety Science, 34(1-3), 177-192.
Flin, R., Burns, C., Mearns, K., Yule, S., Robertson, E.M., 2006. Measuring
safety climate in health care. Qual. Saf. Health care, 15(2), 109-115.
Flin, R., 2007. Measuring safety culture in healthcare: A case for accurate
diagnosis. Safety Science, 45(6), 653-667.
Frost, P.J., Moore, L.F., Louis, M.R., Lundberg, C.C., Martin, J. (eds) 1991.
Reframing organizational culture. London and New Delhi: Sage.
Geertz, C., 1973. The Interpretation of Culture. London: Fontana Press.
Geller, S., 2001. Behavior-based safety in industry: Realizing the large-scale

potential of psychology to promote human welfare. Applied & Preventive
Psychology, 10(2), 87-105.
Gherardi, S., Nicolini, D., 2000. The organizational learning of safety in

communities of practice, Journal of Management Inquiry, 9(1) 7-18.
Giddens, A., 1991. Modernity and Self-Identity: Self and Society in the late
Modern Age. Cambridge: Polity Press.
Gilje, N., Grimen, H., 1993. Samfunnsvitenskapenes forutsetninger.

Universitetsforlaget. [Premises of the social sciences].
99
References
Glendon, A. I., Stanton, N. A., 2000. Perspectives on safety culture. Safety

Science, 34(1-3), 193-214.
Goffman, E., 1992. Vårt rollespill til daglig. Ny utgave Pax Forlag A/S, Oslo.
Original: The Presentation of Self in Everyday Life. Copyright © 1959 Erving
Goffman.
Gooderham, P., Nordhaug, O., Ringdal, K., 1999. Insitutional and rational
determinants of organizational practices: human resource management in
European firms. Administrative Science Quarterly, 44(3), 507-531.
Griffin, M. A., Neal, A., 2000. Perceptions of Safety at Work: A Framework

for Linking Safety Climate to Safety Performance, Knowledge and
Motivation. Journal of Occupational Health Psychology, 5(3), 347-358.
Gubbins, C., MacCurtain, S., 2008. Understanding the Dynamics of

Collective Learning: The Role of Trust and Social Capital. Advances in
Developing Human Resources, 10(4), 578-599.
Guldenmund, F. W., 2000. The nature of safety culture: a review of theory

and research. Safety Science, 34(1-3), 215-257.
Hale, A., 2000. Editorial: Cultures confusion. Safety Science, 34(1-3), 1-14.
Hale, A., 2001. Conditions of occurrence of major and minor accidents.

Institution of Occupational Safety and Health Journal, 5(1), 7-21.
Hale, A., Guldenmund, F., 2005. Much ado about safety culture. Keynote
presentation at Nordic Safety Conference, Nofs, Gilleleje, Denmark, June 8.-
10th.
Hammersley, M., 1992. Deconstructing the qualitative-quantitative divide. In

Brannen, J., (ed). Mixing Methods: Qualitative and Quantitative Research.
London: Gower, 39-57.
Haukelid, K., Tharaldsen, J.E., Lamvik, G.M., Bye, R., Dahle, I.B., 2006.
What works – Lessons learned from the Norwegian Petroleum Industry. Paper
proceedings at The 3rd International Conference Working-On-Safety (WOS)
2006, Netherlands.
Haukelid, K., 2008. Theories of (safety) culture revisited – An

anthropological approach. Safety Science, 46(3) 413-426.
100
References
Heinrich, H.W., 1931. Industrial Accident Prevention. McGraw Hill, New

York.
Helmreich, R.L., Merritt, A.C., 1998. Culture at work: National,

organizational and professional influences. Aldershot, United Kongdom:
Ashgate.
Hofstede, G., 1984. Culture’s consequences: International differences in work

related values. Beverly Hills, CA: Sage.
Hopkins, A., 2009. Thinking about process safety indicators. Safety Science
47(4), 460-465.
House, R.J., Hanges, P.J. Javidan, M., Dorfman, P.W., Gupta, V. (eds.)
(2004). Culture, Leadership, and Organizations. The GLOBE Study of 62
Societies. SAGE Publications.
Hudson, P., 2009. Process indicators: Managing safety by the numbers. Safety
Science 47(4), 483-485.
Hurst, N.W., Young, S., Donald, I., Gibson, H., Muyselaar, A., 1996.
Measures of safety management performance and attitudes to safety at major
hazard sites. Journal of Loss Prevention. Process Industry, 9, 161-172.
INSAG, 1986. International Nuclear Safety Advisory Group. Summary

Report on the Post-Accident Review on the Meeting on the Chernobyl
Accident. Safety Series No. 75 – INSAG 1, International Atomic Energy
Agency, Vienna.
INSAG, 1991. International Nuclear Safety Advisory Group. Safety culture.

Series No. 75 – INSAG 4, International Atomic Energy Agency, Vienna.
Jeffcott, S., Pidgeon, N., Walls, J., 2006. Risk, Trust and Safety Culture in the
U.K. Train Operating Companies. Risk Analysis, 26(5), 1205-1121.
Johnson, S. E., 2003. Behavioral Safety Theory. Understanding the theoretical

foundation. Professional Safety, 48(10), 39-44.
Jöreskog, K.G., Sörbom, D., 1989. LISREL 7 - A guide to the program and
applications. Chicago: SPSS Publications.
101
References
Karlsen, J.E., Lindøe, P., 2006. The Nordic OHS model at a turning point?
Policy and Practice in Health and Safety, 04.1, 17-30.
Keesing, R.M., 1987. Anthropology as an Interpretive Quest. Current

Anthropology, 28, 161-176.
Kim, J.O., Mueller, C.W., 1978. Introduction to factor analysis. Sage

publications. Beverly Hills / London.
Klein, K.J., Bliese, P., Kozlowski, S., Danserau, F., Gavin, M., Griffin, M.,
2000. Multilevel analytical techniques: Commonalities, differences and
continuing questions. In Klein, K.J., Koslowski, S.W. (eds). Multilevel theory,
research and methods in organizations, pp. 512-553. San Francisco: Jossey-
Bass.
KonKraft report 7, 2009. Economic consequences of petroleum activities.
Kostova, T., Roth, K., 2002. Adoption of an organizational practice by

subsidiaries of multinational corporations: institutional and relations effects.
Academy of Management Journal, 45(1), 215-233.
Kjellén, U., 2009.The safety measurement problem revisited. Safety Science

47(4), 486-489.
Krause, T. R, Seymour, K.J., Sloat, K.C.M., 1999. Long-term evaluation of a

behaviour-based method for improving safety performance: a meta-analysis of
73 interrupted time-series replications. Safety Science, 32(1), 1-18.
Kringen, J., 2009. Culture and control. Regulation of risk in the Norwegian
petroleum industry. Doctoral thesis, University of Oslo, Faculty of social
sciences, Centre for technology, innovation and culture.
Lamvik, G.M., Ravn, J.E., 2006. National culture, work practice and safety.
Offshore drilling in South East Asia vs. the North Sea. SINTEF report; STF50
S06027.
LaPorte, T.R., Consolini, P.M., 1991. Working in Practice But Not in Theory:
Theoretical Challenges in “High-Reliability-Organizations”. Journal of Public
Administration Research and Theory, 1, 19-47.
102
References
Lauridsen, Ø., Tønnesen, T., 1990. Injuries Related to the Aspects of Shift
Working – A Comparison of Different Offshore Shift Arrangements. Journal
of Occupational Accidents, 12, 167-176.
Lawrence, R. J., et al., 1993. rwg: An Assessment of Within-Group Interrater

Agreement. Journal of Applied Psychology, 78(2), 306-309.
Lewicky, R.J., Mcallister, D.J., Bies, R.J., 1998. Trust and distrust: New
relationships and realities. Academy of Mangement review, 23(3), 438-458.
Lewicki, R.J., Tomlinson, E.C:, Gillespie, N., 2006. Models of Interpersonal

Trust: Theoretical Approaches, Empirical Evidence, and Future Directions.
Journal of Management, 32(6), 991-1022.
Lindell, M.K., Whitney, D.J., 2001. Accounting for common method variance
in cross-sectional research designs. Journal of Applied Psychology, 86, 114-
121.
Lindøe, P., Olsen, O.E., 2009. Conflicting goals and mixed roles in risk
regulation: a case study of the Norwegian Petroleum Directorate. Journal of
Risk Research, 12(3-4), 427-441.
Lingard, H., Rowlinson, S., 1997. Behavior-based safety management in

Hong Kong's construction industry. Journal of Safety Research, 28(4), 243-
256.
Logstein, H.S., 2009. Myndighetenes forvaltning av sikkerhetsregelverket i

petroleumsvirksomheten. [Authorities’ management of safety regulations in
the petroleum sector].
Lowe, S., 2001. In the Kingdom of the Blind, the One-eyed Man is King.
International Journal of Cross Cultural Management, 1(3), 313-332.
Luhmann, N., 1979. Trust and power. Chicester, England; Wiley.
Lund, J., Aarø, L. E., 2004. Accident prevention. Presentation of a model

placing emphasis on human, structural and cultural factors. Safety Science,
42(4), 271-324.
Lund, I.O., Rundmo, T., 2009. Cross-cultural comparisons of traffic safety,

risk perception, attitudes and behavior. Safety Science, 42(4), 547-553.
103
References
Madhok, A., 2006. How much does ownership really matter? Equity and trust
relations in joint venture relationships. Journal of International Business
Studies, 37, 4-11.
Magnaghten, P., Myers, G., 2004. Focus groups. Chapter 4 in Seale, C.,
Gobo, G., Gubrium, J.F., & Silverman, D., (eds). Qualitative Research
Practice. Sage Publications, pp. 65-79.
Mearns, K. J., Flin, R., 1999. Assessing the state of organizational safety —
Culture or climate? Current Psychology, 18(1), 5–17.
Mearns, K., Whitaker, S.M., Flin, R., 2001. Benchmarking Safety Climate in
Hazardeous Environments. A Longitudinal, Interorganizational Approach.
Risk Analysis, 21(4), 771-786.
Mearns, K., Whitaker, S., & Flin, R., 2003. Safety climate, safety
management practice and safety performance in offshore environments. Safety
Science, 41(, 641–680.
Mearns, K., Rundmo, T., Flin, R., Gordon, R., Fleming, M., 2004. Evaluation
of psychosocial and organizational factors in offshore safety: a comparative
study. Journal of Risk Research, 7(5), 545-561.
Mearns, K., Yule, S., 2008. The role of national culture in determining safety
performance: Challenges for the global oil and gas industry. Safety Science,
47(6), 777-785.
Mearns, K., 2009. From reactive to proactive – Can LPI’s deliver? Safety
Science 47(4), 491-492.
Menard, S., 2002. Applied Logistic Regression Analysis. SAGE University

Papers. Series/Number 07/106.
Moreland, R.L., Levine, J.M., 2002. Socialization and Trust in Workgroups.

Group Processes & Intergroup Relations, 5(3), 185-201.
Morgan, D. L., 1997. Focus groups as qualitative research. Thousand oaks,

California: Sage Publications, Inc.
Nederhof A.J., 1985. Methods of coping with social desirability bias: a

review. European Journal of Social Psychology, 15, 263-280.
104
References
Nonaka, I., Takeuchi, H., 1995. The knowledge creating company. New York:
Oxford University Press.
Olsen, E., 2009. Safety climate and safety culture in health care and
petroleum: Empirical studies on longitudinal change, structural models and
psychometric quality. Doctoral thesis submitted July 2009. University of
Stavanger, Faculty of Social Sciences, Societal Risk and Safety.
Parkes, K.R., 1998. Psychosocial aspects of stress, health and safety on North
Sea installations. Review. Scandinavian Journal of Work, Environment and
Health, 24(5), 321-333.
Perrow, C., 1999. Normal accidents. Living with High-Risk Technologies.

Princeton University Press. First published by Basic Books 1984.
Pett, M. A., Lackey, N. R., Sullivan, J. J. 2003. Making Sense of Factor

Analysis. The Use of Factor Analysis for Instrument Development in Health
Care Research. Sage Publications.
Pettigrew, A.M., 1999. Longitudinal field research on change: Theory and

Practice, Organization Science, 1(3), 267-292.
Pidgeon, N.F., 1991. Safety culture and risk management in organizations.

Journal of Cross-Cultural Psychology, 22(1), 129-140.
Pidgeon, N., Walls, J., Weyman, A., Horlick-Jones, T., 2003. Perceptions of
and Trust in the Health and Safety Executive as a Risk Regulator. Research
Report 2003/100. Norwich: HSE Books.
Podsakoff, P.M., MacKenzie, S.B., Lee, J., 2003. Common Method Bias in
Behavioral Research: A Critical Review of the Literature and Recommended
Remedies. Journal of Applied Psychology, 88(5), 879-903.
Poortinga, W., Pidgeon, N.F., 2003. Exploring the Dimensionality of Trust in

Risk Regulation. Risk Analysis, 23(5), 961-972.
Popper, K., 1959. The logic of scientific discovery. First English edition
published by Hutchinson & co. First published by Routledge 1992.
Reason, J., 1997. Managing the Risk of Organizational Accidents. Ashgate,

USA.
105
References
Rip, A., 1991. The danger culture of industrial society. Chapter 15 in

Kasperson, R.E., Stallen, P.J.N. (eds.). Communicating Risks to the Public.
Kluwer Academic Publishers, Printed in the Netherlands.
Rockwell, T.H., 1959. Safety performance measurement. Journal of

Industrial Engineering 10, 12-16.
Rousseau, D.M., Sitkin, S.B., Burt, R.S., Camerer, C., 1998. Not so different
after all: A cross-discipline view of trust. Academy of Management Review,
23(3), 393-404.
Rundmo, T., 1992. Risk perception and safety on Offshore Petroleum

Platforms. Part II: Perceived risk, job stress and accidents. Safety Science,
15(1), 53-68.
Rundmo, T., Hale, A., 2003. Managers’ attitudes towards safety and accident
prevention. Safety Science, 41(7), 557-574.
Ryen, A., 2004. Ethical issues. Chapter 15 in Seale, C., Gobo, G., Gubrium,
J.F., Silverman, D., (eds). Qualitative research practice. London: Sage, pp.
230-247.
Schmidt, N., 1996. Uses and abuses of coefficient alpha. Psychological

Assessment, 8(4), 350-353.
Seo, D. C., 2005. An explicative model of unsafe work behaviour. Safety

Science, 43(3), 187-211.
Seppänen, R., Blomqvist, K., Sundqvist, S., 2007. Measuring inter-

organizational trust – a critical review of the empirical research in 1990-2003.
Industrial Marketing Management, 36, 249-265.
Smith-Solbakken, M., 1997. Oljearbeiderkulturen. Historien om cowboyer og

rebeller. [Oilworker culture. The story of cowboys and rebels.] Doctoral
thesis, NTNU Norwegian University of Science and Technology, Faculty of
Humanities. Department of history and classical studies.
Spangenberg, S., Baarts, C., Dyreborg, J., Jensen, L., Kines, P., Mikkelsen,
K.L., 2003. Factors contributing to the differences in work related injury rates
between Danish and Swedish construction workers. Safety Science, 41(6),
517-530.
106
References
Spector, P.E., 1987. Method variance as an artifact in self-reported affect and

perceptions at work: Myth or significant problem? Journal of Applied
Psychology, 72, 438-443.
Spector, P.E., 1994. Using self-report questionnaires in OB research: A

comment on the use of controversial method. Journal of Organizational
Behavior, 15, 383-392.
Spector, P.E., 2006. Method Variance in Organizational Research. Truth or

Urban Legend? Organizational Research Methods, 9(2), 221-232.
Steiger, J.H., 1990. Structural model evaluation and modification: An interval

estimation approach. Multivariate Behavioral Research, 25, 173-180.
Sullivan, J.H., Peterson, R.B., Kameda, N., Shimada, J., 1981. The
Relationship Between Conflict Resolution Approaches and Trust – A Cross
Cultural Study. Academy of Management Journal, 24(4), 803-815.
Tayeb, M., 2001. Conducting Research Across Cultures. Overcoming

Drawbacks and Obstacles. International Journal of Cross Cultural
Management, 1(1), 91-108.
Tharaldsen, J.E., Lindeberg, R., 2004. Regulatory requirement for a sound

health, environment and safety culture in the Norwegian Petroleum Acitvities.
Paper proceedings at the International SPE conference on HSE (Society of
Petroleum Engineers), Calgary 2004.
Tharaldsen, J., 2004. Teorier og metoder innen risikostyring og

samfunnssikkerhet. [Theories and methods within risk management and
societal safety]. Exam phd-course, May 2004, University of Stavanger.
Tharaldsen, J.E., Haukelid, K., 2007. The underdogs on the Norwegian

Continental Shelf. High risk, high exposure and high satisfaction! Abstract
presented at Nofs, Nordic conference on safety, 13th to 15th of June, 2007,
Tampere, Finland.
Tharaldsen, J., Høydal, R., Junge, A., Westnes, P., 2007. Ny vår på feltet.
Vurdering av den nye organisasjonen på XX. [A new spring on the field.
Evaluation of the new organization at XX]. IRIS report 2007/052.
Confidential.
107
References
Tharaldsen, J.E, Mearns, K.J., Knudsen, K., 2008. A comparative study of

safety climate on the Norwegian and Continental Shelves. Proceedings SPE
international conference, HSE, Nice (France), April 2008.
Tharaldsen, J.E., 2008. “Safety Science” – Focusing what in the service of

whom? Paper accepted for a PhD-course in “Philosophy of Science with
Ethics”, University of Stavanger.
Turner, N., Gray, G.C., 2009. Socially constructing safety. Human Relations
62(9), 1259-1266.
Weick, K.E., 1987. Organizational Culture as a Source of High Reliability.

California Management Review, 24(2), 112-127.
White, M. P., & Eiser, R. 2006. Marginal trust in risk managers, building and
losing trust following decisions under uncertainty. Risk Analysis, 26(5), 1187-
1203.
White paper no. 7, 2001-2002. ”Health, environment and safety in the

Petroleum activities.”
White paper no. 9, 2008-2009. “Long-Term Perspectives for the Norwegian

Economy”.
Woods, D.D., 2009. Escaping failures of foresight. Safety Science 47(4), 498-
501.
Zaheer, S., Zaheer, A., 2006. Trust across border. Journal of International
Business Studies, 37, 21-29.
Zohar, D., 1980. Safety Climate in Industrial Organizations: Theoretical and

Applied Implications, Journal of Applied Psychology, 65(1), 96-102.
Zohar, D., 2003. Safety Climate: Conceptual and Measurement Issues. In:
Quick, J. C., Tetrick, L. E., (Eds). Handbook of Occupational Health
Psychology. Washington, DC: American Psychological Association, 96-123.
Zohar, D., Luria, G., 2005. A Multilevel Model of Safety Climate: Cross-level
Relationships Between Organization and Group-Level Climates, Journal of
Applied Psychology, 90(4), 616-628.
108
References
Zohar, D., 2008. Safety climate and beyond: A multi-level multi-climate

framework. Safety Science 46(3), 376-387.
109
Part II
List of Articles
I Tharaldsen, J., Olsen, E., Rundmo, T., 2008: A longitudinal study of safety
427-439.
II Høivik, D., Tharaldsen, J.E., Baste, V., Moen, B.E., 2009. What is most
Science,Vol. 47, no. 10, 1324-1331.
III Tharaldsen, J.E., Haukelid, K., 2009. Cultural and behavioural approaches to
safety – towards a balanced approach. Journal of Risk Research, vol. 12, No.
3-4, April-June 2009, 1-14.
IV Tharaldsen, J.E., Mearns, K.J., Knudsen, K., 2009. Perspectives on safety: the
impact of group membership, work factors and trust on safety performance in
UK and Norwegian drilling company employees. Safety Science, 48, 1062-
1072.
Contribution to appended papers

The table below illustrates my contributions to the articles included in the
thesis:
Type of contribution Art I Art II Art III Art IV

Idea, formulation and planning 3 1 3 3
Data acquisition 3 2 N/R 3
Data Analysis 3 2 N/R 3
Writing of manuscript 3 1 3 3
3=Main responsibility 2=Contributed to a high extent 1=Contributed

0=Did not contribute N/R=Not relevant
111
Article I
Available online at www.sciencedirect.com
Safety Science 46 (2008) 427–439

www.elsevier.com/locate/ssci
A longitudinal study of safety climate on the

Norwegian continental shelf
J.E. Tharaldsen *, E. Olsen, T. Rundmo
IRIS – International Research Institute of Stavanger, Postbox 8046, N-4068 Stavanger, Rogaland, Norway
Received 23 March 2006; received in revised form 24 April 2007; accepted 4 May 2007
Abstract
The objective of the study was to examine the psychometric qualities of a questionnaire (Norwegian offshore risk and
safety climate inventory) and whether employee perceptions of safety climate changed over time. The aim of the question-
naire was to measure safety climate and risk on offshore oil platforms on the Norwegian continental shelf (NCS). The
results were based on two surveys carried out on all offshore oil personnel in 2001 and 2003. The response rate was
55% (N = 3310) in 2001 and 50% (N = 8567) in 2003. A combination of exploratory factor analysis and confirmatory fac-
tor analysis resulted in a safety climate structure of five dimensions: safety prioritisation, safety management and involve-
ment, safety versus production, individual motivation and system comprehension. Structural equation modelling indicated
that the suggested factor model fitted the data in 2001 and in 2003. Safety climate was significantly improved from 2001 to
2003 on four dimensions. Platform, work area, type of company and type of platform constituted important differentiating
variables.
Ó 2007 Elsevier Ltd. All rights reserved.
Keywords: Safety climate; Safety culture; Norwegian continental shelf
1. Introduction
During the nineties, there were diverging opinions among the partners in the petroleum industry whether
safety on offshore oil platforms located on the Norwegian continental shelf (NCS), had decreased or stag-
nated. This must be understood in relation to extensive processes of technological, operational and organisa-
tional changes – perhaps without sufficient or corresponding focus on the consequences for health, safety and
environment (HSE). In order to meet this challenge, a new legal framework and White Paper No. 7, 2001–
2002 was introduced, and the project ‘‘Trends in risk levels – Norwegian shelf” (TRL) was initiated by the
Petroleum Safety Authorities (PSA). The objective of TRL was to: (1) measure the effect of HSE work in
the industry, (2) contribute to identifying areas that are critical to HSE, and (3) increase insight into potential
causes of accidents and undesirable conditions and their significance for the risk scenario (see www.ptil.no).
*
Corresponding author. Tel.: +47 51875158; fax: +47 51875200.
E-mail address: jet@irisresearch.no (J.E. Tharaldsen).
0925-7535/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ssci.2007.05.006
428 J.E. Tharaldsen et al. / Safety Science 46 (2008) 427–439
In order to throw light on the actual risk level on the NCS, the TRL stressed the importance of methodolog-
ical triangulation. Triangulation refers to the combination of methodologies in the study of the same phenom-
enon. Based on the TRL-initiative, an expert committee with representatives from various research
environments and the industry developed a questionnaire aimed at measuring safety climate and risk during
the year 2000.
The core objective of this study was to examine the psychometric qualities of the questionnaire called the
Norwegian offshore risk and safety climate inventory (NORSCI) and to compare employee perceptions of the
safety climate in the years 2001 and 2003. Based on the current literature on safety climate (see section below),
the aims of the study were as follows: (1) to test the dimensional structure of NORSCI using a combination of
exploratory (EFA) and confirmatory factor analysis (CFA) on data from 2001 to 2003, (2) to examine whether
or not employee perceptions of safety climate changed from 2001 to 2003, (3) to investigate differences in per-
ceptions across various demographic variables, (4) to explore the associations between safety climate measures
and risk perception and, (5) to check whether or not safety climate scores aggregated at the platform level were
correlated with risk perception and actual safety performance (accidents rates) on the platforms in 2001 and
2003.
2. Previous studies
In the Norwegian oil industry safety culture was in focus at the end of the eighties (Haukelid et al., 1991;
Haukelid, 1999). The industry has undergone major technical, organizational, economic and social changes
and has been divided accordingly into the following phases: ‘‘Texas” (from 1966), ‘‘the great change”
(1980s), ‘‘the great systems” (1990s) and ‘‘the cultural solution” (present). The prevailing period is character-
ized by a strong emphasis on safety culture as a key to success (Haukelid, 1998; Haukelid, 2001). The term
safety culture first appeared in the International Atomic Energy Agency’s report on the Chernobyl accident
(IAEA, 1991). There are many definitions of the concept and most of them are derived from the organizational
culture literature. A much used definition of organizational culture is: ‘‘the shared values (what is important)
and beliefs (how things work) that interact with an organization’s structure and control system to produce
behavioral norms (the way we do things around here)” (Reason, 1997, p. 192). The term safety culture is often
understood as a subcomponent or indicator of organizational and societal culture (Mearns and Flin, 1999;
Glendon and Stanton, 2000; Guldenmund, 2000). However, we will, in accordance with Gherardi and Nicc-
olini (2000) move towards the working communities as the places were safety practice is enacted and evolving,
and argue for a more dynamic understanding of safety culture. Safety culture is then to be considered as the
collective ability to produce organizational and interorganizational work practices that both protect individ-
ual welfare and the environment (ibid.).
Safety climate, on the other hand, is often used to describe employees’ perceptions, attitudes and beliefs
about risk and safety. These perceptions are often measured by questionnaires and provide us with a ‘‘snap
shot” of the current state of safety. Safety culture, however, is considered to be of a more complex and endur-
ing phenomenon than safety climate, reflecting fundamental values, norms, assumptions and expectations,
which, to some extent, reside in societal culture. Hence it follows that safety culture should be examined
and complemented with qualitative methods and other measures, e.g. interviews, observations/audits and
fieldwork. Further, safety climate is either defined as a sub-component of safety culture (Glendon and Stan-
ton, 2000; Zohar, 2003; Cooper and Phillips, 2004) or as a reflection of the actual safety culture by others (Cox
and Flin, 1998; Mearns and Flin, 1999; Guldenmund, 2000). In this study, we will lean towards the last view;
i.e. safety climate as a reflection of an underlying safety culture of a work group, plant or organisation. How
near this reflection is to the actual safety culture, will depend on the quality of the instrument trying to mea-
sure it.
According to Cooper and Phillips (2004), the last 25 years of safety climate research can be split into the
following four directions: (1) designing of psychometric measurement instruments and ascertaining their
underlying factor structure, (2) developing and testing theoretical models of safety climate to ascertain deter-
minants of safety behaviour and accidents, (3) examining the relationship between safety climate perceptions
and actual safety performance; and (4) exploring the links between safety climate and organisational climate
(Cooper and Phillips, 2004). The field is dominated by the search for the right inventory or dimensions able to
J.E. Tharaldsen et al. / Safety Science 46 (2008) 427–439 429
grasp the ‘true priority of safety’ (Mearns and Flin, 1995; Williamson et al., 1997; Dedobbeleer and Beland,
1998; Anderson et al., 2000; Cox and Cheyne, 2000; Flin et al., 2000; Griffin and Neal, 2000; Guldenmund,
2000; Rundmo, 2000; Zohar, 2003; Cooper and Phillips, 2004). The most frequently used statistical method
to determine the dimensional structure of safety climate is factor analysis and yet there exists no agreement
on which dimensions safety climate consists of. However, safety management and colleague involvement
dimensions have been found in several studies (Zohar, 1980; Zohar, 2003; Flin et al., 2000; Guldenmund,
2000; Rundmo and Hale, 2003). Safety management and involvement have also been found to play a crucial
role in the implementation of behavioural safety interventions (Lingard and Rowlinson, 1997; Depasquale and
Geller, 1999; Krause et al., 1999; Geller, 2001; Johnson, 2003; Lund and Aarø, 2004; Cox et al., 2004; Seo,
2005).
In safety research, it is hypothesized that there is a link between safety culture, climate and performance; i.e.
that the employees safe or unsafe practices or behaviour is a function of the organisational safety culture and
the safety climate. However, it has been difficult to establish causal links between them (Felknor et al., 2000;
Zohar, 2003; Cooper and Phillips, 2004). Some have suggested that the relationship may be mutual or that
‘‘the climate-behaviour-accident path is not as clear cut as commonly assumed” (Cooper and Phillips,
2004, p. 497). A mutual relation implies that safe behaviour may lead to a safer culture or reverse. Accidents
may also urge the organization towards a safer culture and better scores on safety climate measures. A better
safety culture may also paradoxically lead to better incident reporting and hence an apparent worsening of
safety performance. In order to get an overall picture of the organisation’s priorities of safety, Cooper and
Phillips (2004) argue for a combined focus on both (low level) system failures (accident rates) and proactive
measures, e.g. safety climate, hazard identification and/or observed percent of safe behaviour.
Differences in key underlying structures may reflect methodological differences in question generations,
sample populations across industries, labelling of constructs according to the theoretical model driving
research or that different instruments measure distinctly different safety climate concepts. Safety climate mea-
sures seem to be useful for ascertaining employee’s perceptions of the way in which safety is being operationa-
lised, despite differences in how safety climate is conceptualized (Carroll, 1998; Cooper and Phillips, 2004).
The aims of the study therefore became an important task, both in order to explore the dimensional structure
of the instrument and the stability of safety climate over time, but also to further investigate the relationship
between safety climate and the actual safety performance on the platforms on the NCS.
3. Methods
3.1. Samples
The results of the study were based on self completion questionnaire data from two samples of personnel on
offshore oil platforms on the NCS. The number of participating platforms/fields was 52. Data collection were
carried out in the years 2001 (N = 3310) and 2003 (N = 8567) with response rates of 55% in 2001 and 50% in
2003 (information about how the questionnaire was administrated is more thoroughly described on the PSA
website). Due to an extension of the survey from two weeks in 2001 to six weeks in 2003, the samples varied
considerably in size, but were considered to be representative at both moments in time. Each year, the ques-
tionnaire was distributed by the medical staff on the platforms. A network of distribution contacts (HSE rep-
resentatives) from the operator companies was established. They requested their employees to cooperate with
the study and kindly asked them to fill in the questionnaire. Participation was done voluntarily and anony-
mously. According to PSA, the distribution of responses for different categories of personnel corresponded
closely to information from other sources on demographic factors and working hours produced in the period.
The demographic characteristics of both samples also showed a steady distribution on different indicators for
both years.
3.2. Measures used in NORSCI
The development of NORSCI can be described as a five-stage process: (1) examination of previous
research on safety climate/culture, (2) design of a preliminary version, (3) test of the preliminary version,
(4) incorporation of comments and suggested improvements, and (5) development of a final version. Relevant
literature was reviewed and different dimensions, scales and questionnaires were examined (Bandura, 1986;
Reason, 1990; Mearns and Flin, 1999; Cooper, 2000; Glendon and Stanton, 2000; Guldenmund, 2000).
A questionnaire was then developed by safety researchers. A group consisting of specialists on occupational
health and safety and representatives from the employee unions were asked to review, test and examine the
questionnaire and to come up with improvements. In order to validate and secure the integration of the most
relevant items on risk and safety offshore, interviews were also conducted with representatives from the indus-
try and summarised in a pilot report. Based on the recommendations done by the expert group as well as the
pilot testing of the questionnaire, safety climate was defined in the inventory according to the model developed
by Cooper (2000).
Safety climate
Person
Conse-
quences
Situation Practice
Model 1 Safety Climate (Copper, 2000, from Bandura 1977, 1986)
Cooper (2000) stresses the importance of empirical investigation of the links between personal, behavioural
(practice) and situational aspects and that these links are interactive or reciprocal. People are neither determin-
istically controlled by their environments nor entirely self-determining; they influence and are influenced by
their environments in a dynamic interplay. Dependent on how these reciprocal aspects are linked to each
other, they will influence on peoples perception of safety climate and on their ability to perform safely or unsa-
fely. These reciprocal aspects were operationalised into items included in the questionnaire, focusing on safety
climate aspects and employees’ perception of risk and safety. The superior TRL project also stresses the
importance of methodological triangulation in order to get a picture of the safety culture on the NCS.
The final version of the instrument consisted of (1) demographic data, (2) safety practice, (3) individual
safety skills and experience, (4) situational aspects relevant for safety behaviour and (5) risk perception. Part
one consisted of several demographic variables, i.e. gender, age, platform name, platform type, work area, etc.
Part two, three and four comprised a total of 49 items. Safety practice was related to matters of daily safety
prioritisation and risk communication. Individual safety skills and experience were connected to role clarity,
safety training and competence. Situational aspects were divided into aspects influencing behaviour, e.g. safety
management and managerial safety prioritisation, and consequences, e.g. individual motivation and the fol-
lowing up of systems and procedures. The risk perception items originate from a study made by Marek
et al. (1985) and has further been used by Rundmo (1989, 1992). Risk perception were related to nine industry
specific risk scenarios and the respondents were asked to judge the probability of the risk scenarios to occur on
their platform, e.g. helicopter accident, gas leakages, fire, blow-out, etc. Risk perception was treated as
depending on the employees’ perception of safety climate; high perceptions of safety climate ought to be cor-
related negatively with the employees’ judgement of the probability of different risk scenarios to occur. The
accidents rates used in the study, included death or severe personal injury, work disability with absence
and medical treatment given by a medical doctor or under supervision of a doctor. The accident rates were
registered on platform level and accident rates (AR) reported in the same quarter as the survey was carried
out were used. Accident rates were standardised to correspond with million man hours worked at the plat-
form/field level.
Scoring on safety climate followed the Likert scale from 1 to 5 expressing agreement or disagreement. Risk
perception was given a rating scale from 1 = very little risk to 6 = very high risk. Certain safety climate items
were recoded so that all the responses could be compared in the same direction of 1 being lowest or least desir-
able response and 5 being the highest or most desirable response. For Risk perception and AR, we kept the
original coding; with low values expressing low perceived risk or few accidents and vice versa.
3.3. Statistical procedures
Principal component analysis with iteration and varimax rotation was used to explore the dimensional
structure of safety climate in the 2001 sample. Thereafter, the model fit of the structure which emerged
through the exploratory analyses, was tested by applying SEM confirmatory factor analyses and tested on
both samples (2001 and 2003), separately. The fit measures applied were the following: comparative fit index
(CFI), root mean square error of approximation (RMSEA), goodness of fit index (GFI), adjusted goodness of
fit index (AGFI) and critical N (CN). Internal consistency was tested by Cronbach’s a separately for the two
years. Discriminant and criterion validity was analysed by Pearson’s r. The associations between safety climate
and risk perception were tested on individual level (total shelf) separately for the two years. In order to keep
the data on the same level of analysis, the correlations between accident rates, risk perception and safety cli-
mate dimensions were aggregated at the platform level and performed separately for the two years.
Multiple analysis of variance (MANOVA) was used to test whether or not there was an overall change in
employee perceptions of safety climate in 2001 and 2003 and to test which of the dimensions showed signif-
icant differences on various demographic variables. The MANOVA including the demographic variables was
performed separately for the two years.
4. Results
4.1. Demographics
The sample was, as expected, male dominated (91% of the respondents both years). The majority of the
respondents were employed by contractor companies, with 57% of the respondents in 2001 and 59% in
2003. Most of the respondents worked on production platforms (77% in 2001 and 80% in 2003), within con-
struction/maintenance (28% in 2001 and 29% in 2003), drilling (23% in 2001 and 19% in 2003), process (16% in
2001 and 14% in 2003) and well service (6% in 2001 and 7% in 2003). Catering personnel constituted approxi-
mately one tenth of the total for both years, as did the ‘‘other” group.
4.2. Dimensional structure of NORSCI
According to the underlying theories driving the research, we expected to explore and confirm dimensions
related to the Cooper model mentioned above. This should be dimensions related to safety practice, individual
safety skills and Situational aspects relevant for safety behaviour. First EFA was performed on the 2001 sam-
ple with the 49 items on safety climate. Items which were found to be unrelated to any factor or had high load-
ings on several factors were removed. This resulted in a safety climate structure of 6 dimensions comprising 42
items. The next step was to examine this dimensional structure of safety climate using CFA. Analysis was car-
ried out separately for 2001 and 2003. Based on theoretical considerations, modification indices and loadings,
the structure was then reduced to five dimensions comprising a total of 32 items. The items that made up the
dimensions, as well as the factor loadings, are shown in Appendix 1. Nineteen of the 32 loadings showed a
drop from 2001 to 2003. The rest (13) were either equal to or higher than the loadings in 2001. The largest
drops were found in the first dimension, which indicates that the unexplained variance for this dimension is
bigger in 2001 than in 2003. The most consistent pattern was found for Safety management and involvement,
but also safety versus production, individual motivation and system comprehension kept most of the loadings
at the same level. The SEM fit statistics also indicated that the suggested model fitted to the data in 2001 as
well as in 2003. The fit measures were: RMSEA = 0.039 (2001) and 0.041 (2003), CFI = 0.97 (2001) and 0.98
(2003), GFI = 0.95 (2001) and 0.95 (2003), AGFI = 0.94 (2001) and 0.94 (2003) and CN = 661.35 (2001) and
702.55 (2003).
The first dimension was entitled Safety prioritisation and comprised a total of eight items reflecting safety
prioritisation and involvement, e.g. following procedures, violations and daily safety prioritisation. The next
factor was Safety management and involvement which consisted of 11 items. This dimension reflected the
management’s priorities towards safety, the role of the safety deputies and the employees feeling of involve-
ment and influence in safety matters, e.g. ‘‘The company I work for takes HSE seriously” and ‘‘I can influence
the HSE conditions at my workplace”. The third dimension was named Safety versus production. It reflected
external framework conditions affecting safety or safety prioritisation and the relative significance of safety
versus production. Safety versus production was made up of four items, e.g. ‘‘lack of maintenance has resulted
in reduced safety” or ‘‘in practice the concern for production precedes the concern for HSE”. Factor number
four was individual motivation and had reference to individual safety motivation, prioritisation of safety and
use of personal protective equipment. The dimension consisted of five items all dealing with what ‘‘I” do; e.g.
‘‘I report dangerous situations when I see them”, or ‘‘I stop working when I think it’s dangerous for me or
others to continue” etc. Factor number five, system comprehension, intended to measure conditions concern-
ing understanding and functionality of safety related procedures and management systems and comprised four
items; for instance, ‘‘I think it’s easy to find the right steering document” or ‘‘I always know which person
within the organisation to report to”.
Cronbach’s a ranged from 0.672 to 0.857. Factor 3–5 showed lower alphas than recommended in 2001
(Cronbach, 1951), but reached a satisfactory level in 2003 for dimension 3. Alpha scores are shown to be sen-
sitive to the number of items in the dimensions and this may be an explanation of the alpha’s being less than
0.70 (Schmidt, 1996). However, the alpha scores and the SEM fit measures were on the same level for both
measures and the dimensions seem to reflect distinctly different safety aspects.
A three dimensions solution, which would be in perfect accordance with the Cooper model, was not found.
However, the five dimensions could be re-allocated according to the Cooper model. The first dimension safety
prioritisation was related to safety practice, dimension five included aspects of individual safety skills and
dimension two, three and four touched upon situational aspects relevant for safety behaviour. Dimension
two and three could be treated as influencing safety practice and dimension four could be seen as a conse-
quence of how the workforce perceive safety management and the role of safety in relation to production.
The dimensions also seem to be in accordance with the most common dimensions found in other safety climate
measures (Flin et al., 2000).
4.3. Safety climate scores from 2001 to 2003
The second aim in the study was to examine whether or not employee judgement on safety climate changed
from 2001 to 2003. Descriptive statistics for the two measurements on all the safety climate dimensions are
shown in Table 1.
The results showed an improvement on four of the five safety climate dimensions, when data from 2001 and
2003 were compared. The most marked improvements, as perceived by the employees, were related to safety
prioritisation and safety versus production. But safety management and involvement and Individual motiva-
tion also showed better scores in 2003. As can be seen in the table, system comprehension showed a small
decline.
Table 1
Employee judgement of safety climate in 2001 and 2003
Safety climate dimensions Sample in 2001 Sample in 2003
Mean Standard deviation Mean Standard deviation
Safety prioritisation 3.18 1.02 3.90 0.69
Safety management and involvement 4.11 0.58 4.23 0.55
Safety versus production 3.02 0.96 3.20 0.97
Individual motivation 4.63 0.44 4.69 0.42
System comprehension 3.89 0.74 3.86 0.74
Further, we used MANOVA to test the model of difference between the two measures for all five dimen-
sions, with year as an independent variable and the safety climate dimensions as dependents. The results
showed a statistically significant difference in the safety climate dimensions when data from year 2001 was
compared to data in 2003, with an overall Wilks’ Lambda of 0.827 (df = 5), p < 0.001, effect size = 0.173
(g2). ANOVA analyses indicated significant differences (p < 0.001) in employee judgements for all except
the last dimension, system comprehension.
MANOVA was then performed to examine differences in perceptions of safety climate dimensions across
four demographic variables: platform (platform name), platform type (production/mobile), company type,
(customer/contractor), and work area (maintenance/construction, drilling, well service, process, catering
and others). The demographic variables were treated as independents and the safety climate dimensions as
dependents. The analysis was done separately for the two years. Table 2 shows the results from the
MANOVA.
The smallest Wilks’ Lambda (indicating greatest differences) was found on platform belonging then work
area, followed by company type and platform type. This pattern was found in 2001 and 2003, however, with
some differences between the two years on each variable, e.g. differences between the platforms were bigger in
2001 than in 2003, while differences between work areas and company types indicated bigger differences in
2003. Differences between platform types showed the same results both years.
ANOVA showed that platform differentiated significantly (p < 0.05) on all safety climate dimensions both
years. Work area differentiated significantly (p < 0.05) on all safety climate dimensions, except for Safety ver-
sus production in 2001, but on all dimensions in 2003 (p < 0.001). Drilling, often together with the Other
group, were found to have better scores on the safety climate dimensions compared to other work areas
for both years. Well service, often together with catering personnel, came out with the lowest scores on all
dimensions both years.
For company type, whether you work in a customer or contractor company, ANOVA indicated that three
of the safety climate dimensions differentiated significantly (p < 0.01) on company type, except for leadership
and involvement and Individual motivation both years. The customer companies had significantly (p < 0.01)
higher scores on safety prioritisation and safety versus production both years. Employees from contractor
companies reported better scores on system comprehension in 2001 and 2003 (p < 0.001).
Platform type, whether you work on a production platform or a mobile rig, differentiated significantly
(p < 0.05) on individual motivation and system comprehension in 2001. In 2003, the results showed a signif-
icant (p < 0.001) differentiating pattern for Safety versus production and system comprehension. In 2001, the
production platforms were found to perceive their Individual motivation as higher than employees on mobile
rigs, while the employees working on mobile rigs perceived their system comprehension to be better than
employees on the production platforms. In 2003, the mobile rigs perceived the safety versus production
and system comprehension to be better than employees on production platforms.
All in all, the strongest and most consistent differentiating patterns were found for platform then work area,
followed by company type and platform type. This indicates that safety climate should be further investigated
at a group level, suggesting that the differences between the grouping variables were greater than within. The
differentiation pattern also gives support to the idea that safety climate dimensions seem to be related to actual
and natural working units, indicating that there exist different safety climates on different platforms and in
different working areas.
Table 2
Differentiating patterns for four demographic variables (MANOVA) in 2001 and 2003
Independents Sample in 2001 Sample in 2003
Wilks’ F df Significance Wilks’ F df Significance
Platform 0.702 4.32 265 0.000 0.880 4.20 255 0.000
Work area 0.948 5.84 30 0.000 0.926 20.50 30 0.000
Company type 0.973 18.44 5 0.000 0.968 56.44 5 0.000
Platform type 0.988 7.68 5 0.000 0.988 20.58 5 0.000
5. Associations between safety climate dimensions and risk perception
The next step was to examine the associations between the safety climate dimensions and risk perception.
The correlation analysis was performed separately for the two years and for the total shelf (individual level).
Risk perception was computed as an index (mean) of the nine items concerning employee judgement of the
probability of different hazardous situations, resulting in a Cronbach’s a equal to 0.878. The employee’s per-
ception of risk was significantly (p < 0.001) improved from 2001 (M = 2, 52) to 2003 (M = 2, 33). Table 3
shows the correlations between safety climate dimensions and risk perception for 2001 above the diagonal
and for 2003 beneath the diagonal.
All correlates in Table 3 were significant at the 0.01 level (two-tailed) and ranged from 0.31 to 0.68. As
expected, all safety climate dimensions were significantly correlated, although not very highly in 2001. This
indicated conceptual differences and that they measure different aspects of safety climate. In 2001 Safety pri-
oritisation was moderately correlated with safety versus production (0.39). Safety management and involve-
ment was strongly associated with Individual motivation (0.56) and system comprehension (0.64).
Individual motivation was associated with system comprehension (0.43). Risk perception was negatively cor-
related with all safety climate dimensions, strongest with leadership ( 0.27) and system comprehension
( 0.22).
The most striking result of the correlation analysis is the big increase in the correlations from 2001 and
2003, both between all safety climate dimensions and the negative correlations between risk perception and
the safety climate dimensions. This may be related to the improved scores on safety climate from 2001 to
2003 and that the standard deviations are either on the same level or smaller in 2003 (see Table 1). Thus, it
seems to be a tendency of becoming more united in their safety efforts throughout the total shelf.
6. Associations between safety climate, risk perception and accident rates
The fifth main aim of the study was to examine whether or not safety climate dimensions were negatively
correlated with safety accidents rates and risk perception on the platforms in 2001 and 2003. Each platform
was given their specific mean score on all safety climate dimensions, risk perception and their correspond-
ing, standardised accident rate. These aggregated mean scores on the safety climate dimensions and risk per-
ception result in smaller variance in the dataset and hence another basis for the correlation analysis. This
also explains why the correlations in Tables 3 and 5, differ. The number of platforms in the analysis was 52
both in 2001 and 2003. Some of the large fields were given one accident rate, even if they consisted of sev-
eral platforms, due to the way data was reported to the PSA. Table 4 shows the range of accident rates in
2001 and 2003.
The AR on the NCS showed a general significant (p < 0.05) decrease from 2001 to 2003. The same ten-
dency we find for all the other statistics: Median, modus, standard deviations and max scores are decreas-
ing from 2001 to 2003. This made sense considering the generally improved scores on the safety climate
dimensions and ought to give support for a hypothesis of safety climate dimensions to be negatively cor-
related with AR and risk perception. To justify aggregation to the relevant level of aggregation (platform
level), the members’ climate perceptions should exceed a threshold of homogeneity to index consensus,
Table 3
Association between safety climate dimensions and risk perception in 2001 (above diagonal) and 2003 (beneath diagonal)
Concepts 1 2 3 4 5 6
Safety prioritisation – 0.09 0.39 0.07 0.06 0.05
Safety management and involvement 0.66 – 0.14 0.56 0.64 0.27
Safety versus production 0.68 0.58 – 0.08 0.11 0.08
Individual motivation 0.44 0.55 0.28 – 0.43 0.13
System comprehension 0.52 0.63 0.44 0.42 – 0.22
Risk perception 0.31 0.29 0.34 0.13 0.25 –
Table 4
Range of accident rates on the platforms in 2001 and 2003
Year Mean Md Mo Standard deviation Minimum Maximum
2001 23.97 18.82 11.82 18.65 0 82.34
2003 12.6 10.02 0 11.24 0 70.71
using rwg > .70 as indication of within-group agreement (from Zohar and Luria, 2005; Lawrence et al.,
1993). In 2001, the homogeneity test for safety prioritisation (mean rwg = 0.59) and safety versus produc-
tion (mean rwg = 0.54) were weaker than recommended. The other three ranged from mean rwg = 0.74 to
0.90. In 2003, safety versus production also showed lower homogeneity than the rest (mean rwg = 0.55).
The other safety climate dimensions ranged from mean rwg = 0.74 to 0.92 in 2003. The low homogeneity
within safety prioritisation in 2001 and safety versus production in 2001 and 2003 indicated larger differ-
ences in perceptions than recommended. This may affect the results of the group-level analysis. Table 5
shows associations between the safety climate scales, risk perception and AR aggregated on platform level
for each year.
The associations between safety climate scales and risk perception generally showed a significant (p < 0.01),
negative relation for both measures, indicating higher perceived risk on platforms with lower scores on safety
climate. The correlations ranged from 0.07 to 0.67. However, risk perception and safety prioritisation
showed a weak but positive correlation in 2001 (0.07). In 2001, we found strongest correlations between risk
perception and system comprehension ( 0.55) and safety management and involvement ( 0.46). Individual
motivation ( 0.27) and safety versus production ( 0.15) were not highly, but still negatively correlated with
risk perception. In 2003, the overall tendency was stronger correlations between risk perception and all five
safety climate dimensions, all significant and negative.
For both years all safety climate dimensions, except the first dimension safety prioritisation, were negatively
associated with AR per million work hours; i.e. the less ideal the safety climate was perceived, the more acci-
dents were reported. The correlations ranged from 0.17 to 0.18 and were all significant. Safety prioritisation
was positively correlated with both AR (0.17) and risk perception (0.16) in 2001. The weak rwg score in 2001,
which indicated less within group homogeneity on this dimension, may have contributed to this, but could also
reflect that despite of high safety prioritisation on the platforms – accidents may occur, or that perceptions of
risk may be high – despite of high safety prioritisation.
7. Discussion and conclusions
The present stage in the Norwegian petroleum industry has been named ‘‘the cultural solution”, character-
ized by a strong emphasis on safety culture (Haukelid, 1998; Haukelid, 2001). It is tempting to relate an
increased focus on safety culture, both in the new petroleum regulations and by the industry itself, to the
improved safety climate scores and the decreasing accident rates. Still, we have no proof of such a relation.
Table 5
Correlations (Pearson’s r) aggregated on platform level between safety climate scales, risk perception and AR
Dimensions 2001 2003
RP AR RP AR
Safety prioritisation 0.16** 0.17** 0.17** 0.10**
Safety management and involvement 0.42** 0.07** 0.50** 0.06**
Safety versus production 0.07** 0.13** 0.67** 0.07**
Individual motivation 0.25** 0.11** 0.30** 0.14**
System comprehension 0.52** 0.12** 0.53** 0.13**
**
p < 0.01.
The non-significant decline in the fifth dimension system comprehension can probably be related to a thor-
ough implementation of new systems for Work permits and Safe Job Analysis, which was accomplished
throughout the NCS in the same period as the measurement in 2003. The work permit- and safe job analy-
sis-systems are fundamental in the daily working practice and an introduction of new systems would probably
bring along a period of ‘‘confusion”.
This study has thrown light on the psychometric qualities of NORSCI as a tool in the TRL project. Results
indicate that NORSCI measures important aspects regarding safety climate and risk at work, differentiated on
various demographic variables and seem to be well suited for the monitoring of the level of risk and safety on
the NCS. In this way NORSCI contributes in identifying areas that are critical to safety and follow up studies
using NORSCI will enhance the understanding of potential causes and development of undesirable conditions
regarding safety and risk on the NCS.
Correlations between safety climate dimensions ranged from 0.07 to 0.64, the correlations were stronger in
2003 than in 2001, and the dimensions measured different aspects of safety climate. The different dimensions
can be reallocated according to the reciprocal perspective suggested by Cooper (2000). The first dimension
safety prioritisation included aspects related to safety practice, dimension five could be related to Individual
safety skills and dimension two, three and four touched upon Situational aspects relevant for safety behaviour.
Dimension two and three could be treated as influencing safety practice and dimension four could be seen as a
consequence of how the workforce perceive safety management and the role of safety in relation to produc-
tion. The dimensions also seem to be in accordance with the most common dimensions found in other safety
climate measures (Flin et al., 2000).
The overall picture showed that safety climate dimensions were negatively associated with risk percep-
tion and AR. However, risk perception and AR were positively correlated with safety prioritisation in
2001. The weak rwg score in 2001, which indicated less within group homogeneity on this dimension,
may have contributed to this, but could also reflect that despite of high safety prioritisation – accidents
may occur or that perceptions of risk may be high despite of high employee perception of safety priori-
tisation. Zohar (2003) has also suggested that injuries may influence supervisory practices and group cli-
mate. It may also be that AR do not substitute a reliable indicator of safety performance. Hence,
safety climate scales should be further investigated across other outcome and performance measures on
the NCS.
Safety has high priority in the oil industry and the level must be characterised as good. For the time
being, oil prices are high and the industry is distinguished by a positive drive and energy, which also influ-
ences in the prioritisation of safety. The working communities situated on platforms on the NCS have much
in common with high reliability organisations, signified by complex working operations, coordination chal-
lenges, intensive technology, high professional knowledge and being vulnerable to human and organisa-
tional error. The organisations are governed by rules, procedures and invest heavily in the training of
their personnel (Weick et al., 1999; Weick and Roberts, 1993). However, there exist future challenges to
comply with, with relevance for safety. The NCS must be characterised as a mature shelf with ‘‘aging” plat-
forms. The need for maintenance and modifications will increase, the wells and reservoirs are complex, dril-
ling is accomplished in deep waters and integrated operations become more usual. Integrated operations
result in fewer people offshore or that several operations are monitored by people onshore. The risk scenar-
ios change and may influence employees’ attitudes to and perceptions of risk and safety. At the moment,
exploitation of the northern continental shelf also constitutes an environmental (harsh climate, vulnerable
ecological environment) and a political challenge, meeting strong resistance from environmental protection
organisations.
In this picture, NORSCI seem to be a useful tool in ascertaining employee perceptions of the level of risk
and safety on the platforms on the NCS. The dimensions, the test–retest results, reliabilities and the differen-
tiating patterns were reasonable. However, the instrument has no ambition of defining the ‘‘Big Five” (Barrick
and Mount, 1991; Flin et al., 2000), even if the factor analysis resulted in five dimensions. It is important to
bear in mind that these kinds of instruments always could have been defined differently. However, for future
research, we would recommend to further investigate the safety climate dimensions and their relation to other
measures of safety performance, to examine the eventual causal relation between them and how they relate to
the different safety climates on a (hierarchical) group level.
Appendix 1. Dimensional structure of NORSCI, Cronbach’s a and loadings in 2001 and 2003
Dimensions and items Loadings

2001 2003
Safety prioritisation, Cronbach’s a: 2001 = 0.857, 2003 = 0.715
1. Occasionally I’m required to work in a manner that jeopardizes safety 0.79 0.60
2. When it comes to one’s carrier it is a disadvantage to be too concerned with HSE 0.74 0.61
3. I sometimes violate safety rules to get the job done 0.74 0.57
4. My lack of knowledge of new technology can sometimes lead to an increased risk of accidents 0.66 0.44
5. I do not participate actively at safety meetings 0.58 0.33
6. My work site is often untidy 0.61 0.38
7. I find it uncomfortable to call attention to violations of safety rules 0.63 0.52
8. The regulatory requirements on HSE are not good enough 0.51 0.47
Safety management and involvement, Cronbach’s a: 2001 = 0.849, 2003 = 0.849
1. The company I work for take HSE seriously 0.65 0.65
2. My supervisor is committed to working with HSE on the installation 0.72 0.67
3. The safety deputies’ suggestions are taken seriously by the management 0.61 0.64
4. Risky work operations are always carefully examined before they are commenced 0.49 0.54
5. My colleagues are very preoccupied with HSE 0.64 0.63
6. I can influence the HSE-conditions at my workplace 0.48 0.56
7. Information about undesirable incidents are effectively used to prevent them recurring 0.55 0.54
8. The emergency preparedness is good 0.56 0.53
9. The safety deputies are doing a good job 0.50 0.50
10. It is easy to tell the nurse/company health service about worries and sickness related to the 0.52 0.48
work situation
11. The work permit system is always lived up to 0.47 0.47
Safety versus production, Cronbach’s a: 2001 = 0.699, 2003 = 0.731
1. Lack of maintenance has resulted in reduced safety 0.40 0.54
2. In practice the concern for production precede the concern for HSE 0.64 0.74
3. Reports on accidents or dangerous situations are often ‘‘smartened up” 0.72 0.69
4. There are often parallel work operations proceeding that leads to dangerous situations 0.64 0.58
Individual motivation, Cronbach’s a: 2001 = 0.672, 2003 = 0.666
1. I report dangerous situations when I see them 0.69 0.71
2. Safety has top priority when I do my job 0.67 0.64
3. I ask my colleagues to stop work when I think the job is being done in a risky manner 0.63 0.62
4. I stop working if I think it can be dangerous for me or other to continue 0.44 0.39
5. I use personal protective equipment 0.33 0.33
System comprehension, Cronbach’s a: 2001 = 0.666, 2003 = 0,673
1. I think it’s easy to find the right steering document 0.50 0.56
2. I always know which person within the organisation to report to 0.52 0.59
3. The HSE procedures are suitable for my work tasks 0.70 0.71
4. I have received sufficient safety training 0.69 0.55
References
Anderson, E. et al., 2000. Testing the reliability and validity of a measure of safety climate. Journal of Healthcare Quality 22 (2), 19–24.
Bandura, A., 1986. Social Foundations for Thought and Action: A Social Cognitive Theory. Prentice Hall, Englewood Cliffs, NJ.
Barrick, M.R., Mount, M.K., 1991. The Big Five Personality Dimensions and Job Performance: a Meta-analysis. Personnel Psychology,
Inc., pp. 1–26.
Carroll, J., 1998. Safety culture as an on-going process: culture surveys as opportunities for enquiry and change. Work & Stress 12 (3),
272–284.
Cox, S., Cheyne, A.J.T., 2000. Assessing safety culture in offshore environments. Safety Science 34 (1-3), 111–129.
Cox, S., Flin, R., 1998. Safety culture: philosopher’s stone or man of straw? Work and Stress 12 (3), 189–201.
Cox, S. et al., 2004. Behavioural approaches to safety management within UK reactor plants. Safety Science 42 (9), 825–839.
Cooper, M.D., 2000. Towards a model of safety culture. Safety Science 36 (2), 111–136.
Cooper, M.D., Phillips, R.A., 2004. Exploratory analysis of the safety climate and safety behaviour relationship. Journal of Safety
Research 35 (5), 497–512.
Cronbach, L.J., 1951. Coefficient alpha and the internal structure of tests. Psychometrika 16 (3), 297–334.
Dedobbeleer, N., Beland, F., 1998. Is risk perception one of the dimensions of safety climate? In: Feyer, A., Williamson, A. (Eds.),
Occupational Injury: Risk, Prevention and Intervention. Taylor & Francis, London.
Depasquale, J.P., Geller, E.S., 1999. Critical success factors for behavior-based safety: a study of twenty industry-wide applications.
Journal of Safety Research 30 (4), 237–249.
Felknor, S.A. et al., 2000. Safety climate and its association with injuries and safety practices in public hospitals in costa rica. International
Journal of Occupational Environmental Health 6 (1), 18–25.
Flin, R. et al., 2000. Measuring safety climate: identifying the common features. Safety Science 34 (1–3), 177–192.
Geller, S., 2001. Behavior-based safety in industry: realizing the large-scale potential of psychology to promote human welfare. Applied &
Preventive Psychology 10 (2), 87–105.
Gherardi, S., Niccolini, D., 2000. The organizational learning of safety in communities of practice. Journal of Management Inquiry 9 (1),
7–18, March.
Glendon, A.I., Stanton, N.A., 2000. Perspectives on safety culture. Safety Science 34 (1–3), 193–214.
Griffin, M.A., Neal, A., 2000. Perceptions of safety at work: a framework for linking safety climate to safety performance, knowledge and
motivation. Journal of Occupational Health Psychology 5 (3), 347–358.
Guldenmund, F.W., 2000. The nature of safety culture: a review of theory and research. Safety Science 34 (1–3), 215–257.
Haukelid, K., 1998. En historie om risiko: antropologiske betraktninger om sikkerhet, bedriftskultur og ledelse i norsk oljevirksomhet [A
history of risk: anthropological reflections on safety, organisational culture and management in the Norwegian oil industry]. Centre for
Technology and Human Values, University of Oslo. TMV Report number 32.
Haukelid, K., 1999. Risiko og sikkerhet: forståelser og styring [Risk and safety: comprehension and regulation]. Universitetsforlaget.
Haukelid, K., 2001. Oljekultur og sikkerhetskultur del 1 (Oil culture and safety culture part 1), Working paper 10. Centre for Technology,
Innovation and Culture, University of Oslo.
Haukelid, K., et al., 1991, first ed. 1990. Retningslinjer for sikkerhetsmotiverende tiltak på flyttbare rigger [Guidelines for motivational
safety measures on mobile rigs]. Rogaland Research-Report 12/91.
IAEA, 1991. Safety Culture (Safety Series No 75- INSAG-4). International Nuclear Safety Advisory Group, International Atomic Energy
Authority, Vienna.
Johnson, S.E., 2003. Behavioral safety theory. Understanding the theoretical foundation. Professional Safety, 39–44.
Krause, T.R. et al., 1999. Long-term evaluation of a behaviour-based method for improving safety performance: a meta-analysis of 73
interrupted time-series replications. Safety Science 32 (1), 1–18.
Lawrence, R.J. et al., 1993. rwg: an assessment of within-group interrater agreement. Journal of Applied Psychology 78 (2), 306–309.
Marek, J. et al., 1985. Experience of risk and safety on a north sea oil platform. In: Hellesøy, O.H. (Ed.), Work Environment: Statfjord
Field. Work Environment, Health and Safety. University Press, pp. 142–174.
Lingard, H., Rowlinson, S., 1997. Behavior-based safety management in Hong Kong’s construction industry. Journal of Safety Research
28 (4), 243–256.
Lund, J., Aarø, L.E., 2004. Accident prevention. Presentation of a model placing emphasis on human, structural and cultural factors.
Safety Science 42 (4), 271–324.
Mearns, K.J., Flin, R., 1995. Risk perception and attitudes to safety by personnel in the offshore oil and gas industry: a review. Journal of
Loss Prevention in the Process Industries 8 (5), 299–305.
Mearns, K.J., Flin, R., 1999. Assessing the state of organizational safety – culture or climate? Current Psychology 18 (1), 5–17.
Reason, J., 1990. Human Error. Cambridge University Press, Cambridge.
Reason, J., 1997. Managing the Risk of Organizational Accidents. Ashgate, USA.
Rundmo, T., Hale, A., 2003. Managers’ attitudes towards safety and accident prevention. Safety Science 41 (7), 557–574.
Rundmo, T., 1989. Risk perception and safety. Norwegian Petroleum Directorate, Trondheim/Stavanger, Research Report No. STF75.
Rundmo, T., 1992. Risk perception and safety on Offshore Petroleum Platforms. 3. Perceptions of risk. Safety Science 23 (15), 39–68.
Rundmo, T., 2000. Safety climate, attitudes and risk perception in Norsk Hydro. Safety Science 34 (1–3), 47–59.
Schmidt, N., 1996. Uses and abuses of coefficient alpha. Psychological-Assessment 8 (4), 350–353.
Seo, D.C., 2005. An explicative model of unsafe work behaviour. Safety Science 43 (3), 187–211.
Weick, K.E., Roberts, K.H., 1993. Collective mind in organizations. Heedful interrelating on flight decks. Administrative Science
Quarterly 38, 330–358.
Weick, K.E., Sutcliffe, K.M., Obstfeld, D., 1999. Organizing for high reliability: processes of collective mindfulness. Research in
Organizational Behavior 21, 81–123.
Williamson, A.M., Feyer, A., Cairns, D., Biancotti, D., 1997. The development of a measure of safety climate: the role of perceptions and
attitudes. Safety Science 25 (1–3), 15–27.
Zohar, D., 1980. Safety climate in industrial organizations: theoretical and applied implications. Journal of Applied Psychology 65 (1), 96–
102.
Zohar, D., 2003. Safety climate: conceptual and measurement issues. In: Quick, J.C., Tetrick, L.E. (Eds.), Handbook of Occupational
Health Psychology. American Psychological Association, Washington, DC, pp. 96–123.
Zohar, D., Luria, G., 2005. A multilevel model of safety climate: cross-level relationships between organization and group-level climates.
Journal of Applied Psychology 90 (4), 616–628.
Article II
Safety Science 47 (2009) 1324–1331
Contents lists available at ScienceDirect
Safety Science
journal homepage: www.elsevier.com/locate/ssci
What is most important for safety climate: The company belonging or the
local working environment? – A study from the Norwegian offshore industry
Dordi Høivik a,b,*, Jorunn E. Tharaldsen c, Valborg Baste a, Bente E. Moen a
a
Occupational and Environmental Medicine, Department of Public Health and Primary Health Care, University of Bergen, Kalfarveien 31, N-5018 Bergen, Norway
b
StatoilHydro ASA, Occupational Health and Working Environment, Forusbeen 50, N-4035 Stavanger, Norway
c
IRIS – International Research Institute of Stavanger, P.O. 8046, N-4068 Stavanger, Norway
a r t i c l e i n f o a b s t r a c t
Article history: Obtaining knowledge about factors affecting health, safety and environment (HSE) is of major interest to
Received 18 November 2008 the petroleum industry, but there is currently a severe shortage of relevant studies. The aim of this study
Received in revised form 2 April 2009 was to examine the relative influence of offshore installation (local working environment) and company
Accepted 2 April 2009
belonging on employees’ opinions concerning occupational health and safety. We analyzed data from a
safety climate survey answered by 4479 Norwegian offshore petroleum employees in 2005 on the dimen-
sions ‘‘Safety prioritisation”, ‘‘Safety management and involvement”, ‘‘Safety versus production”, ‘‘Indi-
Keywords:
vidual motivation”, ‘‘System comprehension” and ‘‘Competence” using one way analysis of variance
Safety climate
HSE
(ANOVA), effect size and mixed model. The companies differed significantly for ‘‘Safety prioritisation”,
Offshore ‘‘Safety versus production”, ‘‘Individual motivation”, ‘‘System comprehension” and ‘‘Competence”. The
Petroleum industry local offshore installation explained more of the safety climate than the company they were employed
Working environment in or worked for did.
1. Introduction production involves several flammable substances. The location

makes such events even more dangerous than onshore, because
1.1. Health and safety in the petroleum industry of the long distances to medical aid and hospitals. Health, safety
and environment were at this time considered separate issues.
Health, safety and environment (HSE) has been an important is- While safety issues seemed to be most widely emphasized, envi-
sue in the Norwegian petroleum industry since the offshore oil and ronmental factors became more important for the industry parallel
gas drilling and production started in Norway in 1969, as it has with the growing public opinion regarding sustainable develop-
been in other countries with this kind of production (Gardner, ment, and in the face of several industry-related pollution disas-
2003). Today more than 50 oil and gas fields are being operated ters. In the 1980s and 1990s systematic work on issues related to
on the Norwegian continental shelf, and petroleum-based activi- health and working environment in the petroleum industry be-
ties employed approximately 80,000 people in 2005 (Ministry of came strengthened, partly due to the Working Environment Act
Petroleum and Energy, 2005). In 2004 approximately 20,000 peo- in Norway coming into force on the Norwegian continental shelf
ple had their workplace offshore (Mhlum and Kjuus, 2006). The from 1979, regulating both the physical and the psychosocial
oil installations related to these operations on the Norwegian con- working environment. The three concepts ‘‘health”, ‘‘safety” and
tinental shelf came to be located at between 40 and 185 miles from ‘‘environment” became more integrated in the petroleum industry
the coast, creating workplaces in new and potentially dangerous during the 1990s.
working environments. The offshore working environment has been described as
In the seventies and eighties a major focus was put on the safety stressful, with psychosocial stressors such as difficult working
part of the HSE work at the offshore installations. Work at offshore and living conditions (Parkes, 1998; Gardner, 2003; Mearns et al.,
installations involves risks related to fires and explosions, as oil 2003), long working days and shift work including night work
(Lauridsen and Tonnesen, 1990; Parkes, 1999, 2003) as well as
physical stressors like noise (Morken et al., 2005), ergonomics
* Corresponding author. Address: Occupational and Environmental Medicine, (Chen et al., 2005) and chemical hazards (Steinsvag et al., 2007).
Department of Public Health and Primary Health Care, University of Bergen, All these factors may affect health, environment and safety in a
Kalfarveien 31, N-5018 Bergen, Norway. Tel.: +47 91693599.
negative manner.
E-mail addresses: doh@statoilhydro.com (D. Høivik), Jorunn.Tharaldsen@iris.no
(J.E. Tharaldsen), Valborg.Baste@isf.uib.no (V. Baste), Bente.Moen@isf.uib.no (B.E. The risk of major offshore accidents increased in the late 1990s
Moen). (Ministry of Government Administration and Reform, 2002).
doi:10.1016/j.ssci.2009.04.001
D. Høivik et al. / Safety Science 47 (2009) 1324–1331 1325
Several actions were undertaken in the industry in order to reverse important predictor variables for employee satisfaction with safety
this development, and HSE-related work was assigned a higher pri- measures (Rundmo, 1994). Two studies suggest that the local
ority. In this situation the need for evaluating the ongoing HSE working environment at the offshore installation is important for
work arose among the petroleum companies. The Petroleum Safety the safety climate but neither considered whether this is related
Authorities Norway (PSA) initiated the project ‘‘Trends in Risk Lev- to the company the workers are employed by. One of these studies
els on the Norwegian Shelf” in 1999/2000 to measure the impact of has compared results from the workers’ perception of social and
HSE-related work in the industry, and subsequently to contribute working environment factors in Norway and the United Kingdom
to identifying areas that are critical for working environment and (Mearns et al., 2004). The other study (Tharaldsen et al., 2008)
safety and to increase insight into potential causes of accidents examined the importance of the localization on the platform com-
and undesirable conditions offshore (Petroleum Safety Authority pared to work areas such as drilling or catering, company type like
Norway, 2006). Data from this project are used in the present operating and contractor companies and installation type, such as
study. drilling or production. Offshore petroleum employees in Norway
are employed in an operating company or in a contractor firm. In
1.2. Organizational belonging and health and safety climate 2005 contractors carried out 63.6% of the reported work hours on
the Norwegian Continental shelf (Petroleum Safety Authority Nor-
The terms ‘‘safety culture” and ‘‘safety climate” have often been way, 2006).
used interchangeably, although safety culture is considered to be However, in the petroleum industry there is a lack of knowledge
of a more complex and enduring phenomenon than safety climate, concerning to which extent organizational belonging matters,
reflecting fundamental values, norms, assumptions and expecta- compared to the platform location. Such information is of major
tions (Mearns and Flin, 1999) which is, to some extent, presumably interest in the practical work on HSE in the petroleum industry.
linked to national and societal culture. Several studies have found What is most important: the organizational belonging or the local
safety management, colleague involvement and collaboration to be work place?
important dimensions for safety climate (e.g. Flin et al., 2000;
Rundmo and Hale, 2003; Guldenmund, 2007). Other important 1.3. Aim of the study
dimensions are safety system, risk, work pressure, competence
and procedures/rules (Flin et al., 2000). However, Cooper and Phil- The aim of this study was to examine the health and safety cli-
lips (2004) indicate that ‘‘the climate–behavior–accident path is mate in the petroleum industry in relation to the company belong-
not as clear cut as commonly assumed” and that differences in ing and the local offshore installation. The findings will be of
underlying key structures may reflect methodological differences importance to future planning of the HSE-related work in the
in question generations, sample populations across industries, petroleum industry.
labelling of constructs according to the theoretical model driving
the research or that different instruments measure distinctly dif-
ferent safety climate concepts. On the other hand, safety climate 2. Materials and methods
measures seem to be useful to ascertain employees’ perceptions
of the way in which safety is being operationalized – despite differ- Data from the Petroleum Safety Authorities project ‘‘Trends in
ences in how safety climate is conceptualized (Cooper and Phillips, Risk Levels” carried out in 2005 was used for this study. The ques-
2004; Guldenmund, 2007; Clarke, 2006). tionnaire was called ‘‘the Norwegian Offshore Risk and Safety Cli-
One approach to study variations in HSE-related outcomes in mate Inventory (NORSCI)” and has previously been conducted in
the petroleum industry has been safety climate surveys describing 2001 and 2003. The aim of the survey was to measure health
employees’ perceptions of the priority an organization places on is- and safety climate and risk for occupational health and accidents
sues concerning safety (Zohar and Luria, 2005). The safety climate on Norwegian offshore petroleum installations. The NORSCI ques-
is most often measured by self-administered questionnaires. tionnaire was developed by health and safety researchers, and
Mearns et al. (1998) proposed that questionnaire-based surveys used experts from occupational health and safety in the industry
measuring safety climate are capable of sensing transient surface and representatives from the unions to review, test and examine
features discerned from the workers attitude to safety at a given it (Tharaldsen et al., 2008; Petroleum Safety Authority Norway,
point of time – a snapshot of the prevailing safety culture. 2006). The questionnaire was limited to factors of relevance to
Mearns et al. (2003) indicates that organizations performing safety and working environment, excluding external environment.
well in safety climate surveys in the offshore petroleum indus- It has been described in more detail elsewhere (Tharaldsen et al.,
try in UK have fewer accidents. Similarly, a study based on a 2008).
survey in a large Norwegian oil and gas company stated that
there was a connection between employees’ opinions of man- 2.1. Sample and data collection
agement and safety results, especially regarding accidents
(Høivik et al., 2007). All who attended different installations offshore in Norway dur-
The importance of the local work environment has been studied ing a period from December 2005 to February 2006 were invited to
in the petroleum industry as well as in other industries. Mearns participate in the survey (NORSCI). This included workers on all
and Reader (2008) suggest that improving safety performance production and mobile units on the Norwegian Continental Shelf
may be better delivered indirectly through other sources than di- and workers on vessels inside the safety zone around the installa-
rectly through safety inventions such as company indication of tions. The respondents received the questionnaire through the
safety commitments and safety messages. For example studies companies’ own routines at their workplaces or at the heliports
from the UK offshore industry have found that manager’s positive wherefrom employees are being shuttled to Norwegian offshore
attitude against its workforce (Shannon et al., 1997), managements installations. All employees were encouraged to participate and
concern about their workforce and e.g. health promotion activities either to hand the questionnaire back to a nurse offshore in a
and education at the workplace seems to be effective for safety closed envelope or to return it by mail to the researchers responsi-
performance (Mearns and Hope, 2005). Employee perceptions of ble in Stavanger. Here the answers from NORSCI were put in a
management commitment, social support and subjective evalua- database. The participants did not write their name or birth date
tions of priorities of safety versus production goals, seem to be on the form.
1326 D. Høivik et al. / Safety Science 47 (2009) 1324–1331
According to Petroleum Safety Authority Norway (2006) 9820 health, safety and environment during the 1990s and it is also seen
offshore workers, working on 52 different installations/rigs, com- in the regulations from the Petroleum Safety Authorities Norway.
pleted the questionnaire. This yielded an estimated response rate
of 50%. The responses for different categories of personnel corre- 2.3. Statistical methods
sponded closely to the distribution found in the industry (Petro-
leum Safety Authority Norway, 2006). 2.3.1. Factor analysis
For the purpose of this study we wanted to have information A factorial structure of five dimensions was developed on the
from workers who worked in or for operation companies with safety climate items from similar, previous surveys from 2001
more than one offshore installation because we wanted to study and 2003 (Tharaldsen et al., 2008). The model fit of the dimen-
what was most important: the organizational/company belonging sional structure previously found (Tharaldsen et al., 2008), was
or the local work place/platform. We selected the workers who had tested for the 2005 survey by applying Structural Equational Mod-
their ordinary work offshore, not the ones being on short visits or elling (SEM) and confirmatory factor analyses (CFA). The fit mea-
short stay on the installations. We selected all respondents who sures applied were the following: Comparative Fit Index (CFI),
fulfilled the following criteria: (a) employees from companies Root Mean Square Error of Approximation (RMSEA), Goodness of
operating at two or more offshore oil and gas installations, (b) Fit Index (GFI), Adjusted Goodness of Fit Index (AGFI) and Critical
employees who had spent more than 50% of their work time off- N (CN). The Linear Structural Relation (LISREL) analyses program
shore in 2005 and (c) employees who reported in the survey that version 8.8 was used to estimate the fit of the factor solutions.
they were working all the time or most of the time on one instal- In the 2005 survey two extra items reflecting safety competence
lation. The information was found in the survey questionnaires. were added, giving a new and sixth dimension, ‘‘Competence”
The respondents were categorized into two groups: operator comprising four items. The dimensional structure of safety climate
employees (employed in an operating company) and contractor was then examined by CFA. The items that made up the six dimen-
employees (employed in a supplier company). Since previous stud- sions, as well as the factor loadings from the CFA, are shown in Ta-
ies have shown differences in opinions among these two groups of ble 1. The CFA analysis showed and confirmed the five earlier
workers (Tharaldsen et al., 2008), this categorization was included dimensions for both 2001 and 2003. Four of the loadings were be-
in the analyses of the material. The operating companies are man- low 0.4, but were retained in the model because they all contrib-
aging the day to day activities of the petroleum activities produc- uted well to the fit of the model. The SEM fit statistics indicated
tion. In this study five operating companies were included, based that the suggested model fitted well with the data in our sample
on the condition that the company was operating two or more off- (CFI = 0.98, RMSEA = 0.044, GFI = 0.94, AGFI = 0.93 and
shore oil and gas installations. The contractor category contains CN = 590.41). In order to examine how the theoretically underlying
employees from many different offshore contractor companies like dimensions (safety climate) fitted the model, a second order anal-
drilling, maintenance and catering. We did not have information of ysis was performed. The fit measures held approximately the same
what contractor firms the workers came from: the contractors are level as well as the same loadings (CFI = 0.97, RMSEA = 0.04,
grouped as one group although they came from different contrac- GFI = 0.93, AGFI = 0.92 and CN = 528.99).
tor firms. Operator and contractor employees are working side by To test the internal consistency of the dimensions, Cronbach’s
side and collaborate on the offshore installations in different job alpha was calculated. Cronbach’s alpha, measuring the overall
sections such as catering section, drilling and well maintenance internal reliability of the items, ranged from 0.65 to 0.84, which
section, deck, construction, inspection and maintenance section was regarded as acceptable. Four of the six dimensions reached
and production and process section. In this study only contractors the recommended level of having alphas higher than 0.70
who worked on the installations of the five operating companies (Schmidt, 1996), while two of the dimensions (Individual motiva-
were included. tion and System comprehension) showed lower alphas, 0.678 and
All analyses in the present study are done on the selected pop- 0.654, respectively.
ulation described above.
2.3.2. Statistical analyses
2.2. Measures We used descriptive statistics to describe each company and the
total sample concerning employment status, gender, age and num-
NORSCI consisted of a total of 124 questions divided into six ber of installations. Age was obtained by categories of 20 years or
main parts relevant for health and safety. For the present study less, 21–30 years, 31–40 years, 41–50 years, 51–60 years and 61
only data from the first and second parts were analyzed. The first years or more, and Pearson Chi-square test was applied to test
part asked for information concerning age (six age categories), gen- the age differences among the operator and contractor employees.
der, installation name, company affiliation, work time offshore that Mean scores were calculated for each item and each dimension.
year and if they were working all the time on one installation, most One way analysis of variance (ANOVA) was used to test the differ-
of the time one installation or varying. The second part included 35 ences in mean scores of the dimensions ‘‘HSE prioritisation”,
items measuring perceived safety climate at the work place such as ‘‘Safety management and involvement”, ‘‘Safety versus produc-
safety practice: daily HSE prioritisation and risk communication, tion”, ‘‘Individual motivation”, ‘‘System comprehension” and
own and others’ safety skills and behavior connected to role clarity, ‘‘Competence” among the operator and contractor employees in
safety training and competence and workplace conditions influ- the five companies and on the 41 installations. Effect size was de-
encing health and safety such as management and prioritisation fined by the differences between two mean scores divided by the
and individual motivation and following up systems and average of the standard deviations. This measured the effect of
procedures. the relationship between the scores regardless of the sample size.
The questions were phrased as ‘‘I do not participate actively . . .” ANOVA was also used to estimate adjusted explained variance
or ‘‘It is easy to tell . . .”, and were to be answered on a scale from 1 (R2) for each dimension for operator and contractor employees in
to 5. In the analyses all items were graded by 1 being the lowest the five operating companies and 41 installations separately.
and 5 the highest or most desirable response. One difference be- Mixed model statistics were performed to study the six health
tween this survey and other safety climate surveys was that the and safety climate dimensions and the different levels of manage-
expression ‘‘HSE” was used in the questions, not only ‘‘safety” (Ta- ment of the workers: company and installation. The dimensions
ble 1). The petroleum industry has used this concept for integrated were analyzed in separate models as dependent variables, with
Table 1
Results from factor analyses of six safety climate dimensions from a 2005-survey in five petroleum companies showing dimensions and items, factor loadings and reliability
(Cronbach’s alpha).
Dimensions and items Loadings Cronbach’s alpha

Safety prioritisation
1. Occasionally I’m required to work in a manner that jeopardizes safety 0.64 0.730
2. When it comes to one’s career it is a disadvantage to be too concerned with HSE 0.73
3. I sometimes violate safety rules to get the job done 0.69
4. My lack of knowledge of new technology can sometimes lead to an increased risk of accidents 0.55
5. I do not participate actively at safety meetings 0.36
6. My work site is often untidy 0.45
7. I find it uncomfortable to call attention to violations of safety rules 0.68
8. The regulatory requirements on HSE are not good enough 0.59
Safety management and involvement
1. The company I work for take HSE seriously 0.52 0.835
2. My supervisor is committed to working with HSE on the installation 0.56
3. The safety deputies’ suggestions are taken seriously by the management 0.61
4. Risky work operations are always carefully examined before they are commenced 0.39
5. My colleagues are very preoccupied with HSE 0.43
6. I can influence the HSE-conditions at my workplace 0.48
7. Information about undesirable incidents are effectively used to prevent them from recurring 0.55
8. The emergency preparedness is good 0.56
9. The safety deputies are doing a good job 0.42
10. It is easy to tell the nurse/company health service about worries and sickness related to the work situation 0.48
11. The work permit system is always lived up to 0.48
Safety versus production
1. Lack of maintenance has resulted in reduced safety 0.76 0.735
2. In practice concern for production precedes the concern for HSE 0.95
3. Reports on accidents or dangerous situations are often ‘‘smartened up” 0.88
4. There are often parallel work operations proceeding that leads to dangerous situations 0.67
Individual motivation
1. I report dangerous situations when I see them 0.43 0.678
2. Safety has top priority when I do my job 0.41
3. I ask my colleagues to stop work when I think the task in question is being carried out in a risky manner 0.43
4. I stop working if I think it can be dangerous for me or other to continue 0.28
5. I use personal protective equipment 0.20
System comprehension
1. I think it’s easy to find the right steering document 0.70 0.650
2. I always know which person within the organization to report to 0.62
3. The HSE procedures are suitable for my work tasks 0.69
Competence
1. I have the necessary competence to perform my job in a safe manner 0.53 0.722
2. I have easy access to personal protective equipment 0.40
3. I have received sufficient safety training 0.61
4. I have received sufficient education when it comes to occupational health and work environment 0.53
company as a fixed factor and installation as a random factor. In nies and 1747 (39%) from contractor companies. Because of
addition, an analysis was made by including work in an operating confidentiality-related issues we did not have information about
company or a contractor company as a random factor, with age and the number of contractor firms. They were categorized into one
gender as covariates. The level of significance was set at 0.05. The group. The number of operator and contractor employees per com-
Statistical Package for the Social Sciences (SPSS 13.0) was used for pany varied between 244 and 2390 employees (Table 2). In the to-
analysis. tal population there were 89% men. The percentage of women in
the operating companies were 14% and in the contractor compa-
2.3.3. Ethical consideration nies 5%. There was a significant difference in age for operator and
The present study examined data from NORSCI. The Petroleum contractor employees (chi-square test, p < .001): the contractor
Safety Authorities Norway gave the researchers at the University of employees were on average younger than the operator employees
Bergen their approval to analyze the results. Approval was also gi- (Table 2). No significant age differences among the employees in
ven from each of the operating companies that were studied. It was the operating companies were found.
emphasized that the company identities should not be revealed. To The number of employees in each of the five operating compa-
assure the anonymity of the five operating companies, they are nies varied between 111 and 1415. In the study population 54% of
identified as A, B, C, D and E in Table 2 and as numbers 1–5 in Table the employees were employed in or worked for the largest com-
3. There is no link between the letters (Table 2) and the number as- pany with 19 offshore installations. The second and third largest
signed to the respective companies (Table 3). company employed 21% and 12% of the population, respectively,
and had 10 and 7 offshore installations. The two smallest compa-
3. Results nies, with, respectively, two and three offshore installations,
staffed 8% and 5% of the total workers. All the contractors who
3.1. Demographic data are included in this study have a regular offshore rotation on the
same offshore platform (1333 employees worked on the same plat-
4479 employees fulfilled the inclusion criteria and were inform every work schedule, 414 answered they mostly worked on
cluded in the study: 2662 (59%) came from five operating compa- the same offshore platform).
Table 2
Number of employees, company type, gender and age among responders to a safety climate survey in five petroleum companies offshore, both for workers employed in these and
contractors working at the same installations, 2005.
Variable Category Total Company

N % A B C D E
Number of employees 4479 2390 355 534 244 956
Company type Operator 2662 59 1450 150 334 111 617
Contractor 1747 39 909 196 194 129 319
Missing 70 2 31 9 6 4 20
Gender
Male Operator 2257 86 1213 136 319 107 482
Female 376 14 219 12 13 4 128
Male Contractor 1644 95 866 178 187 107 306
Female 94 5 36 18 6 22 12
Age
<30 Operator 197 7 120 23 14 3 37
31–40 686 26 416 31 33 37 169
41–50 1092 41 578 50 131 52 281
51–60 619 23 305 40 142 19 113
>61 64 3 31 6 14 0 13
Age
<30 Contractor 188 11 92 24 24 11 37
31–40 642 37 321 78 75 60 108
41–50 641 37 340 68 67 41 125
51–60 238 13 129 26 25 17 41
>61 35 2 25 0 3 0 7
Number of installations 41 19 2 7 3 10
Table 3
Mean scores and standard deviations (SD) for six dimensions from a safety climate survey among workers from five petroleum companies offshore 2005. The five companies were
compared using oneway ANOVA. The comparisons were performed separately for the workers in the operating companies and the workers in all the contracting companies
working at the same installation.
Variable Operator Contractor

Company Mean SD P Mean SD P
Safety prioritisation 1 3.96 .68 4.02 .77
2 3.78 .76 4.00 .74
3 4.06 .63 .000 3.91 .67 .272
4 3.91 .68 3.94 .69
5 3.87 .73 3.89 .73
Safety management and involvement 1 4.21 .55 4.26 .61
2 4.23 .58 4.36 .62
3 4.22 .51 .117 4.23 .52 .119
4 4.20 .55 4.28 .54
5 4.21 .53 4.21 .53
Safety versus production 1 3.23 .92 3.28 1.02
2 3.30 1.01 3.69 .99
3 3.52 .95 .000 3.26 .90 .000
4 3.37 .96 3.21 .95
5 3.22 .92 3.20 .93
Individual motivation 1 4.68 .43 4.70 .46
2 4.75 .38 4.77 .32
3 4.72 .39 .000 4.72 .38 .108
4 4.65 .43 4.68 .43
5 4.76 .31 4.71 .42
System comprehension 1 3.80 .83 4.07 .85
2 4.06 .78 4.02 .71
3 3.78 .79 .000 3.77 .81 .000
4 3.44 .85 3.84 .80
5 3.75 .85 3.93 .75
Competence 1 4.37 .64 4.48 .63
2 4.43 .51 4.55 .50
3 4.42 .58 .054 4.44 .55 .413
4 4.39 .60 4.48 .60
5 4.29 .64 4.44 .55
3.2. Differences in HSE dimensions among companies tween the contractor employees working for these five operating
companies. Among the operator employees there was a significant
Table 3 shows the mean scores of the six HSE dimensions difference within ‘‘Safety prioritisation”, ‘‘Safety versus produc-
among the employees in the five operating companies and be- tion”, ‘‘Individual motivation”, ‘‘System comprehension” and a
Table 4
Explained variance (R2) by five operating companies and 41 offshore installations on six different dimensions of health and safety climate. The analyses were performed
separately for the workers in the operating companies and the workers in all the contracting companies at the same installation as the operator companies.
Variable Operator Contractor

Company Installation Company Installation
R2 R2 R2 R2
Safety prioritisation .010 .036 .001 .025
Safety management and involvement .001 .042 .002 .025
Safety versus production .008 .100 .015 .057
Individual motivation .007 .012 .002 .010
System comprehension .048 .066 .011 .039
Competence .002 .015 <.001 .020
borderline significant difference for the dimension ‘‘Competence”. A mixed model analysis using companies and installations as fixed
Effect size analyses, underlined the findings of this five dimensions factors and installation as random factor produced the same re-
(0.40, 0.32, 0.30, 0.76 and 0.24, respectively). The dimension, sults: installation explained more than company. Including age
‘‘Safety management and involvement”, yields an effect size of and gender in the model did not modify the results.
0.05. Among the contractor employees there was a significant
difference in ‘‘Safety versus production” and ‘‘System comprehen-
sion” (effect size 0.51 and 0.36, respectively). 4. Discussion
Contractor employees answered more positively on the health
and safety dimensions than operator employees, except for the The offshore installation the employees worked on seemed to
dimensions ‘‘Safety versus production” and ‘‘System comprehen- be of greater importance for several safety climate dimensions
sion” in 3 and 2 operating companies, respectively. The highest than did the company they were employed in or worked for (con-
mean scores were found for the dimensions ‘‘Individual motiva- tractors). This result was found both for workers in operating com-
tion” and ‘‘Competence”, and the lowest for ‘‘Safety versus produc- panies and for contractors. Differences were found in safety
tion” and ‘‘System comprehension”, both for operator and for climate dimensions both when comparing the scores from the
contractor employees. operating companies and when comparing scores obtained at the
installation level, but the main variance was related to work at
3.3. Differences in HSE dimensions among offshore installations the installations.
To our knowledge, no other studies have performed a similar
In this study, the number of participating installations was 41. analysis of the relationship between the local belonging offshore
The total number of employees on each installation varied between on the installations and the company belonging as employees or
12 and 270 employees. Oneway ANOVA analysis between the 41 working for an employer as contractor among workers in petroleum
installations in the 5 operating petroleum companies, analyzing companies. However, the findings are in line with a previous Nor-
operator and contractor employees separately, showed that the wegian study showing that the installation the employees worked
installations differed significantly in mean scores on all the health on was more important for the safety climate than their work area,
and safety dimensions (p < .001) for both operator and contractor company type, customer or contractor company and installation
employees. type, or production platform or mobile rig (Tharaldsen et al.,
Among all installations the range of mean scores for ‘‘Safety pri- 2008). This study indicates that the local belonging on the installa-
oritisation” was 3.4–4.2, for ‘‘Safety management and involve- tion is important for perceived health and safety. Also, Mearns et al.
ment” 3.9–4.5, for ‘‘Safety versus production” 2.8–3.9, for ‘‘HSE (2004) found by similar analyses, in a study of Norwegian and UK
behavior” 4.5–4.8 and for ‘‘System comprehension” 3.3–4.2. For offshore employees, that installation as a factor explains more of
‘‘Competence” the range was 4.1–4.4. the variance in safety climate than national belonging.
Summarized, the results of the mean scores of the safety cli- The adjusted R2 was low, indicating that there are other vari-
mate dimensions displayed the greatest variations for all employ- ables that explain the results as well, apart from the company
ees on the 41 installations on the dimensions ‘‘System belonging or the installation. This might for instance be accident
comprehension” and ‘‘Safety versus production”, and it displayed involvement, job experience and individual factors.
the smallest variations on the dimensions ‘‘HSE behavior” and The contractor employees were grouped separately, as we had
‘‘Competence”. no information of their company belonging. They were influenced
by their own company’s cultures and this may be one explanation
3.4. Comparisons between companies and offshore installations why the R2 was very low in Table 4 when analyzing company and
the six safety climate variables for the contractors.
To compare the effects of company belonging and to be working Zohar and Luria (2005) studied the cross-level relationship be-
on a specific installation, univariate analyses of variance were tween climate at organizational-level and group-level in manufac-
used. For all six dimensions of the health and safety climate, the turing plants in metal, food, plastics and chemical industries. The
installation factor had a higher adjusted R2 (explained a greater results indicated that safety climate at both the organizational-le-
percentage of the total variance) than company belonging had. vel (company, in our study) and group-level (installation, in our
The findings were similar when operator and contractor employees study) were globally aligned: organizational climate predicts group
were analyzed separately (Table 4). climate. This is in line with the findings in our study. Glendon and
The adjusted R2 related to the installation was highest for oper- Litherland (2001) identified differences in the safety climate of
ator employees for the dimensions ‘‘Safety versus production” (ad- sub-groups in Australian road workers in two different districts
justed R2 (installation) = .100) and ‘‘System comprehension” of construction and maintenance departments. A difference was
(adjusted R2 (installation) = .066). The adjusted R2 related to com- found only between job types, and not between districts, and in
pany belonging among the operator employees showed the highest two of his six dimensions: ‘‘Relationships” and ‘‘Safety Rules”.
level for ‘‘System comprehension” (adjusted R2 (company) = .048). However, the differences between his results and the results in
the present study might be caused by differences in items and in and Budworth (1997) identified differences between the safety cli-
methodology. mate profiles of senior and junior staff within a single organization.
The findings in this study may be explained by the fact that peo- This should be further studied. Another explanation may be that
ple respond to the safety climate survey based on their own expe- contractor employees occasionally works in land based industries
rience regarding health and safety, expressed through own and where the health and safety focus is less expressed.
colleagues’ behavior, local health and safety systems and prioriti-
sation from the installation management. Company policy may 4.2. Strengths and limitations
be viewed as distant from the installation and hence not being
especially influential on their attitudes. On the other hand, super- Several questionnaires have been developed in attempts to
visors explain, justify and apply higher management’s policies to determine the key factors that compromise safety climate (Flin
the workforce and are thus responsible for the notions that live et al., 2000; Guldenmund, 2000; Glendon and Litherland, 2001;
among the workers (Guldenmund, 2007). Top managers make pol- Fullarton and Stokes, 2007). Guldenmund (2000) identified the
icies and establish procedures, whereas supervisors at lower hier- most frequently employed safety climate factors to include: man-
archical levels execute these procedures by turning them into agement, risk, safety arrangements, procedures, training, compe-
situation-action directives. tence and work pressure. Our instrument has items including all
these factors. However, the findings can be caused by the question-
4.1. Differences among companies and among installations naire itself, and its validity may be considered, as it depends on the
way the safety climate questions are asked. In the present ques-
The differences in perceived health and safety climate between tionnaire many items are targeted towards the local environment.
the companies may be explained by different safety cultures in the This problem is of less importance with grouping the questions as
different companies. Some of the companies are Norwegians while in Table 4 underlining that the local working environment is more
others are of other nationalities. International differences were not important than the company belonging for all the dimensions.
analyzed in this study, but are known to influence the safety cli- Coyle et al. (1995) argue that safety climate factors are not univer-
mate (Mearns et al., 2004). However, although the company own- sally stable, and differences in safety climate among groups have
ers might differ with regards to nationalities, most of the workers been identified by several researchers (Mearns et al., 1998; Glendon
in the Norwegian sector are Norwegians and the companies have and Litherland, 2001; Guldenmund, 2000; Flin et al., 2000). However,
to follow the same Norwegian HSE requirements in addition to the present study shows that the health and safety climate dimen-
company rules and procedures. sions ‘‘Safety prioritisation”, ‘‘Safety management and involvement”,
Two of the dimensions: ‘‘Safety management and involvement” ‘‘Safety versus production”, ‘‘Individual motivation”, ‘‘System com-
and ‘‘Competence”, differed from the other dimensions by showing prehension” and ‘‘Competence” were found to be stable through
no significant differences among the companies for neither opera- 2001, 2003 and 2005 in the Norwegian offshore population. This
tor employees nor contractor employees. These dimensions re- may be seen as an expression of a general high focus on HSE in the
ceived high scores from the employees, both operators and offshore petroleum industry in Norway with no major changes in this
contractors. Management commitment remains as a key compo- period. The results indicate that many of the employees are aware of
nent of contemporary safety climate research. Flin et al. (2000) re- the focus on the HSE-issues from the top management and the
ported senior safety management’s attitudes to safety as the factor authorities, and have adopted it as a part of their own attitudes and
appearing most frequently in papers dealing with safety climate/ behavior. On the other hand, this finding must be interpreted with
culture (13 of 18 papers). In our study the uniformly and relatively caution, due to a relative low response rate in this study.
high reporting on the ‘‘Safety management and involvement” The instrument used was validated in a former study (Tharald-
dimension may be seen as an expression of the high focus on sen et al., 2008). Five of the six dimensions were stable through
HSE in the offshore petroleum industry in Norway. It indicates that 2001 and 2003 (ibid), the sixth dimension, ‘‘Competence”, was
many of the employees are aware of the focus on the HSE-issues integrated in the 2005 solution and showed a good fit with the to-
from the top management and the authorities, and have adopted tal model. The Chronbachs alpha was also satisfactory (0.722).
it as a part of their own attitudes and behavior. This study has a cross sectional design which makes it impossi-
Differences among offshore installations have been identified in ble to evaluate causal relationships. However, the important find-
other studies (Mearns et al., 2003). Although a given company has ings in this study are not affected by causality, as it is the
the same HSE policy, health and safety climate were experienced different variances of the findings in relation to other factors which
and reported differently by employees on different platforms. Zohar are in focus. The level of the safety climate itself in relation to the
and Luria (2005) assumes that supervisory discretion within a com- different companies and installations is not the main issue.
pany may stem from several sources such as interpretation of pro- We had no objective data on health and safety results. However,
cedures, tasks not covered by the procedures and possible the safety climate measures used in this study have earlier been
conflicts between performance quantity versus quality (e.g. produc- shown to be linked to objective data like accident rates on platform
tion speed versus safety precautions). Leader–member exchanges level, even if the links were weak (Tharaldsen et al., 2008). Whether
involve interpersonal dynamics that are only partially governed good results in a safety climate survey would give good outcomes
by formal procedures, and individual beliefs and attributions influ- such as reduced work related disorders or injuries cannot be con-
ence supervisory interpretation and the implementation of formal cluded. However, as previous studies have shown a relationship be-
procedures. The managers positive attitude against its workforce tween self reported climate data and objective measures, we argue
are influencing the health and safety performance and the employ- that this type of measure might be useful (Mearns et al., 2003; Coo-
ees perception of safety climate (Mearns and Hope, 2005; Mearns per and Phillips, 2004; Høivik et al., 2007). Similar results have also
and Reader, 2008). Installation management teams are responsible been found for other industries; Johnsen (2007) showed a relation-
for local work environment which seems to be important for health ship between safety climate and injury frequency through a rela-
and safety performance and perceived safety climate. tionship mediated by safety behavior in a manufacturing
Contractor employees answered more positively on most of the company. Due to this, we suggest that the findings can be important
health and safety dimensions than operator employees. The expla- for future work on health and safety on the installations.
nation for this is not known. It might be related to the fact that the This study is quite unique, including data from the main part of
contractor employees were younger. Mason and Simpson (1995) the petroleum industry in Norway, five operating companies and a
large number of contractor companies. The response rate in the Coyle, I.R., Sleeman, S.D., Adams, N., 1995. Safety climate. Journal of Safety Research
26, 247–254.
study was relatively low. However the responses for the different
Flin, R., Mearns, K., O’Connor, P., Bryden, R., 2000. Measuring safety climate:
categories of personnel corresponded closely to the distribution identifying the common features. Safety Science 34, 177–192.
found in the industry (Petroleum Safety Authority Norway, Fullarton, C., Stokes, M., 2007. The utility of a workplace injury instrument in
2006). Hence, we think it might be possible to generalize from this prediction of workplace injury. Accident Analysis and Prevention 39,
28–37.
study to the Norwegian offshore industry as a whole and it should Gardner, R., 2003. Overview and characteristics of some occupational exposures and
also have an impact in other types of industry with multiple loca- health risks on offshore oil and gas installations. Annals of Occupational
tions. Similar studies ought to be performed in other countries as Hygiene 47, 201–210.
Glendon, A.I., Litherland, D.K., 2001. Safety climate factors, group differences and
well, to see if the findings are the same. The measures were self-re- safety behaviour in road construction. Safety Science 39, 157–188.
ported, and although no names were written on the question- Guldenmund, F.W., 2000. The nature of safety culture: a review of theory and
naires, and assurances of confidentiality were given, some Research. Safety Science 34, 215–257.
Guldenmund, F.W., 2007. The use of questionnaires in safety culture research – an
persons might have avoided answering to protect themselves from evaluation. Safety Science 45, 723–743.
any risk of insight into their personal opinions. Høivik, D., Baste, V., Brandsdal, E., Moen, B.E., 2007. Associations between self-
The company belonging and the installations are factors which reported working conditions and registered health and safety results. Journal of
Occupational and Environmental Medicine 49 (2), 139–147.
are not possible to separate. In our study both mixed model and Johnsen, S.E., 2007. The predictive validity of safety climate. Journal of Safety
univariate analysis of the six dimensions were performed, and both Research 38, 511–521.
methods indicate that the variances were mostly caused by the off- Lauridsen, O., Tonnesen, T., 1990. Injuries related to the aspects of shift working – a
comparison of different offshore shift arrangements. Journal of Occupational
shore installations. However, the results must be interpreted with
Accidents 12, 167–176.
some caution due to the close relation among these two factors. Mason, S., Simpson, G., 1995. Measuring safety attitudes to target management
The findings from the present study suggest that the local work- actions. The Safety and Health Practitioner (Nov), 17–20.
place might be of more importance for the workers’ perception Mearns, K.J., Flin, R., Gordon, R., Fleming, M., 1998. Measuring safety climate on
offshore installations. Work Stress 12, 238–254.
of health issues, working environment and safety conditions than Mearns, K.J., Flin, R., 1999. Assessing the state of organizational safety – culture or
the superior policy of the company itself, expressed by company climate? Current Psychology 18, 5–17.
belonging. However, as these factors are difficult to separate com- Mearns, K.J., Whitaker, S.M., Flin, R., 2003. Safety climate, safety management
practice and safety performance in offshore environments. Safety Science 41,
pletely, further studies are needed before any clear conclusions can 641–680.
be given. Mearns, K., Rundmo, T., Gordon, R.F.R., Fleming, M., 2004. Evaluation of psychosocial
and organizational factors in offshore safety: a comparative study. Journal of
Risk Research 7, 545–561.
4.3. Impact on the industry Mearns, K.J., Hope L., 2005. Health and Well-being in the Offshore Environment: The
management of Personal Health. Research Report 305. HSE books, Norwich.
These findings suggest that both local HSE work on the installa- Mearns, KJ., Reader, T., 2008. Organizational support and safety outcomes: an un-
investigated relationship? Safety Science 46, 388–397.
tions and the HSE work by the company, have importance for the Ministry of government administration and reform, 2002. Helse, Miljø og Sikkerhet
safety climate. Although a good policy for HSE must be set up by i Petroleumsvirksomheten. (Stortingsmelding nr 7 (2001–2002). [In Norwegian,
the top managers to initiate the HSE work, this is not enough to Health, Environment and Safety in Petroleum Activities]. Oslo. <http://
www.regjeringen.no/nb/dep/fad/dok/regpubl/stmeld/20012002/Stmeld-nr-7-
produce good health and safety results on the installations. Local
2001-2002-/1.html?id=134388>.
management teams must be aware of the importance of their roles Ministry of Petroleum and Energy, 2005. Facts 2005. Ministry of Petroleum and
in the HSE work. Management’s commitment, prioritisation and Energy. The Norwegian Petroleum Sector, Oslo. <http://odin.dep.no/oed/
follow up of the local work environment are important for the english/doc/reports/026031-120027/dok-bn.html>.
Morken, T., Bratveit, M., Moen, B.E., 2005. Reporting of occupational hearing loss in
safety climate. Leadership training focusing on HSE risks and man- the Norwegian offshore industry 1992–2003. The Journal of the Norwegian
agement’s competence about important safety climate factors are Medical Association 125 (23), 3272–3274 (in Norwegian).
necessary for taking the best decisions to raise the HSE level at Mhlum, I.S., Kjuus, H., 2006. Omfang og Konsekvenser av Arbeidsskader og
Arbeidsbetinget Sykdom på Norsk Kontinentalsokkel [In Norwegian, Extent and
the workplace. This may have major importance for the workers Consequences of Work-related Injuries and Work-related Illness on Norway’s
and the companies. Continental Shelf]. National Institute of Occupational Health and the Norwegian
Petroleum Safety Authority, Oslo.
Parkes, K.R., 1998. Psychosocial aspects of stress, health and safety on North Sea
Acknowledgments installations. Scandinavian Journal of Work, Environment and Health 24 (5),
321–333 (Review).
The Research and Development Program Health, Safety and Parkes, K.R., 1999. Shiftwork, job type, and the work environment as joint predictors
of health-related outcomes. Journal of Occupational Health Psychology 4, 256–
Environmental (HSE) Culture in StatoilHydro ASA initiated and 268.
supported this research. The project is part of the program ‘‘HSE Parkes, K.R., 2003. Shiftwork and environment as interactive predictors of
in the Petroleum Industry in Norway” supported by the Research work perceptions. Journal of Occupational Health Psychology 8, 266–
281.
Council of Norway. We are thankful for support from Øyvind Lau-
Petroleum Safety Authority Norway, 2006. Trends in Risk Levels 2005. <http://
ridsen, The Petroleum Safety Authorities Norway, and for valuable www.ptil.no/English/Helse+miljo+og+sikkerhet/Risikonivaa+paa+sokkelen/RNNS+
comments from Stein Atle Lie and Trond Riise at the University of Fase+6+%282005%29+Sammendragsrapport/coverpage.htm>.
Rundmo, T., 1994. Associations between safety and contingency measures and
Bergen.
occupational accidents on offshore petroleum platforms. Scandinavian Journal
of Work Environment and Health 20 (2), 128–131.
Rundmo, T., Hale, A.R., 2003. Managers’ attitude towards safety and accident
References prevention. Safety Science 41, 557–574.
Shannon, H.S., Mayr, J., Haines, T., 1997. Overview of the relationship between
Budworth, N., 1997. The development and evaluation of a safety climate measure as organizational and workplace factors and injury rates. Safety Science 26, 201–
a diagnostic tool in safety management. Journal of the Institution of 217.
Occupational Safety and Health 1, 19–29. Steinsvag, K., Bratveit, M., Moen, B.E., 2007. Exposure to carcinogens for defined job
Chen, W.Q., Yu, I.T., Wong, T.W., 2005. Impact of occupational stress and other categories in Norway’s offshore petroleum industry, 1970 to 2005. Journal of
psychosocial factors on musculoskeletal pain among Chinese offshore oil Occupational and Environmental Medicine 64, 250–258.
installation workers. Journal of Occupational and Environmental Medicine 62, Schmidt, N., 1996. Uses and abuses of coefficient alpha. Psychological Assessment 8
251–256. (4), 350–353.
Clarke, S., 2006. The relationship between safety climate and safety performance: a Tharaldsen, J.E., Olsen, E., Rundmo, T., 2008. A longitudinal study of safety climate
meta-analytic review. Journal of Occupational Health Psychology 11 (4), 315– on the Norwegian continental shelf. Safety Science 46, 427–439.
327. Zohar, D., Luria, G., 2005. A multilevel model of safety climate: cross-level
Cooper, M.D., Phillips, R.A., 2004. Exploratory analysis of the safety climate and relationships between organization and group-level climates. Journal of
safety behavior relationship. Journal of Safety Research 35, 497–512. Applied Psychology 90, 616–628.
Article III
Journal of Risk Research
Vol. 12, Nos. 3–4, April–June 2009, 375–388
Culture and behavioural perspectives on safety – towards a balanced

approach
Jorunn-Elise Tharaldsena* and Knut Haukelidb
a
University of Stavanger, Risk Management and Societal Safety, 4036 Stavanger, Norway;
IRIS, PO Box 8046, 4068 Stavanger, Norway; bCenter for Technology, Innovation and Culture,
University of Oslo PO Box 1108 Blindern, 0317 Oslo, Norway
The core aim of this paper was to examine differences or similarities in cultural
and behaviour-based approaches to safety. The theoretical bases for culture
theories and behaviourism are very different and, at times, mutually excluding.
Downloaded By: [Swets Content Distribution] At: 10:03 26 August 2009
While many culture theories focus on cognitive aspects like values, ideas and
thoughts – in short, patterns of meaning – behaviourism tends to neglect the
cognitive aspects of human life and focus strictly on behaviour. We find cultural
perspectives to be poorly treated in safety research and assert a need for cultural
approaches in general and a better understanding of them in particular. In order
to do so, we introduce a four-field model which balances between tacit or explicit
sides of culture and whether the approach focuses on behavioural or cultural
aspects. Time, stability and change are also integrated in the discussion of the
model; whether risks are treated as sudden, observable events or as developing
over a longer time period, and not-so-easily observable phenomena.
Keywords: safety culture; behavioural safety; organizational culture; cultural
approaches; safety intervention
Introduction
In the field of safety research there is an ongoing debate about the difference between
behavioural and cultural approaches to safety and how to perform effective
interventions in order to improve safety-related aspects in an organization. This
debate is also echoed in the Norwegian petroleum industry and tends to result in
trench-war-like positioning of the approaches as opposites and as excluding each
other. However, we often tend to use insights from both perspectives, but a
distinction between them is needed.
In this paper we shall concentrate on the scientific roots of behaviour-based
safety (BBS), the concept of culture and organizational (safety) culture. Our claim is
that many safety researchers treat the concept of culture in an instrumental and
technical manner – even if a more interpretative approach is asserted. The scientific
roots of the culture concept are often superficially treated and much of the
organizational (safety) culture research is dominated by functional safety manage-
ment approaches, which might be considered one out of many perspectives. Also, we
find that attempts to undermine or bypass conflicting issues often lead to a
superficial handling of both.
Safety researchers lately (DeJoy 2005) tend to treat cultural approaches as top-
down strategies and behavioural approaches as bottom-up. This manner of naming
and understanding cultural and behavioural approaches, we claim, resides in an
*Corresponding author. Email: jorunn.tharaldsen@iris.no
ISSN 1366-9877 print/ISSN 1466-4461 online

# 2009 Taylor & Francis
DOI: 10.1080/13669870902757252
http://www.informaworld.com
376 J.-E. Tharaldsen and K. Haukelid
incorrect equation between safety management approaches and cultural approaches.

This, we feel, is a simplification. Differences in perspectives could rather be found
when sorting out the following questions; how are the concepts of culture or
behaviour defined? What kind of methods are used? Are the outcomes of culture
(behaviour, practices, interactions etc.) in focus or the underlying taken-for-granted
patterns of ideas? How is culture or behaviour analyzed? Are concepts like tacit or
explicit knowledge taken into account? And is culture looked upon as a noun or as a
process? The kinds of strategies that are chosen to solve these problems will reflect
the stand of the researcher and the perspective of the intervention.
In order to discuss and visualize the dynamics between different approaches, we
introduce a model, which balances tacit (hidden) and explicit (visible) sides of culture
and behavioural versus cultural aspects. Time, stability and change are also
integrated in the discussion of the model; whether risks are treated as sudden,
observable events or as developing over a longer time period, and not-so-easily
observable phenomena. The time aspect is important because it defines what is

treated as risky or safe and how to deal with it. Finally, the implications of the model
for the understanding and implementation of safety interventions will be discussed.
Theoretical framing and roots

BBS – applied behaviour analysis
The main principles of BBS can be traced back to early behavioural theory starting
with controlled laboratory experiments on animals. It is associated with Pavlov in
Russia (classical conditioning) and Thorndike, Watson and Skinner in the USA
(operant conditioning). It makes use of stimuli – response models and principles of
operant conditioning and reinforcement. BBS also has much in common with safety
management perspectives, which often make use of goal theory (Krause, Seymour
and Sloat 1999; Depasquale and Geller 1999; Geller 2001; Johnson 2003; Cox, Jones,
and Rycraft 2004; Seo 2005). A distinction is often made between behaviour
modification and applied behaviour analysis, reserving the latter term for
applications in natural settings, such as work places (DeJoy 2005).
BBS is characterized as an analytic, objective and data-driven approach focusing
on safety-related behaviours performed by first-line personnel. Work-related risks
and critical behaviour are identified and targeted for change. Systems of
performance tracking and performance goals are supplemented by motivational
activities and (positive) feedback systems. The right behaviour is reinforced and the
consequences controlled. The goal is to change risky behaviour into safe behaviour
and, through this, build a culture which supports acceptable behaviour. As a
contrast to the analytic and objectively driven BBS approach, cultural methods are
being understood as being of a more intuitive and subjective kind (DeJoy 2005) and
as being top-down driven.
Several studies have shown that in order to succeed with BBS, various criteria
should be fulfilled; like safety commitment, trust and involvement (Depasquale and
Geller 1999). Cox, Jones, and Rycraft (2004) summarize following critical areas from
a BBS intervention in the nuclear industry. The BBS process would fail if (1) there
was no trust between management and employees; (2) the intervention was used as a
management tool to spy on the employees; (3) the process was used as a weapon
towards the employees; and (4) the employees experienced lack of correspondence
Journal of Risk Research 377
with the project and their understanding of a just culture (Reason 1997). Other
obstacles may be employee rejection of the performance goals; the efforts may not
give the expected improvements or the expected improvements may not be valued by
the employees. Lack of risk awareness and recognition of risks may also affect the
process, i.e. if the employees don’t recognize the behaviour as problematic or risky –
even if it is evidently understood like this by the safety experts (Lingard and
Rowlinson 1997). Another limitation is linked to the old, but nevertheless important,
argument of a too one-sided focus on the individual, resulting in ‘blindness’ towards
other more fundamental social and organizational factors that also may affect safety
and risk.
All in all, the focus is on (observable) behaviour, behavioural change techniques
and the expected effects of them. This implies a view of humans as mostly controlled
by stimuli in the environment and even if the focus is positive, it means a view of
humans as controlled by external rewards and free will as a figment of our
imagination (Skinner 1971; cited in Deaux and Wrightsman 1984).
The concept of culture

While behaviourism has its roots in psychology, the concept of culture has its roots
in other scientific traditions and disciplines, such as anthropology and sociology.
The definitions of the concept tend to differ, but most of them are related to some
kind of sharedness: symbolic meaning systems (Rohner 1984), ways of thinking,
feeling and reacting, meanings and identities, common ways in which technology is
used etc. (House et al. 2004; Alvesson 2002). Such commonality characterizes groups
of people and guides their actions. Methods used are often qualitative in character
and the epistemological roots are hermeneutic.
Unlike behavioural approaches, cultural analysts see mind and culture as a
necessity for human nature: ‘There is no such thing as a human nature independent
of culture… They would be unworkable monstrosities with very few useful instincts,
fewer recognizable sentiments, and no intellect: mental basket cases’ (Geertz 1973,
49). In short, we need culture to create order in the world, and the way we think and
the way we act are to be interpreted as culturally mediated.
In this paper we will address four themes which distinguish cultural from
behavioural perspectives: (1) culture as construction of meaning; (2) sub-cultures and
power; (3) the importance of tacit and embodied knowledge; and (4) culture as a
process – not as a noun.
Culture as construction of meaning

Culture is often treated as a separate system of ideas or as an ordered system of
meanings, which ‘have a minimal degree of coherence, else we would not call them
systems’ (Geerts 1973, 7). Culture is regarded as ‘the fabric of meaning in terms of
which human beings interpret their experience and guide their action’ (Geertz 1973,
145). Identification and analysis of meaning construction become the main point of
any cultural analysis. One of Geertz’s most famous remarks is that ‘Man is an animal
suspended in webs of significance he himself has spun; I take culture to be those
webs’ (Geertz 1973, 5). Webs of significance understood as cultural meanings
become essential in our understanding of ourselves and others, our contextual
frames and our daily handling of challenges – also related to safety and work.
Culture as power
Geertz has been criticized by, among others, Keesing (1987, 1994), who argues that
culture is not a homogenous whole. There are always conflicts between individuals
and subcultures; conflicts over power, values, knowledge and ‘truth’. Keesing also
points out that few do the actual spinning while the majority is simply caught. Both
Keesing and Scholte (1986) assert that culture is not something people ‘make’, but
rather something we are captured in and by – that is, ideologies. Following Keesing,
as well as authors like Foucault (1977) and Bourdieu (1979), power has become a
central issue in many cultural studies.
Tacit and embodied sides of culture

Is culture always linked to what is explicitly expressed through language and
observable practice or is it to be understood as a more embodied and tacit
phenomenon? Brown and Duguid (1991) introduced the concept of communities of

practice (CoP) in order to bring us towards ‘a unified view of working, learning and
innovation’ (Brown and Duguid 1991, 40). CoP refers to the places and processes
where people work, learn and innovate and has much in common with traditional
cultural perspectives. Gherardi and others (Gherardi, Nicolini, and Odella 1998;
Gherardi 1999, 2000; Gherardi and Nicolini 2000a, 2000b) bring the discussion
further into the area of safety culture and emphasize the need to learn safe work
practices – rather than safety as a specific topic. This perspective counters the more
common understandings of work and practice; their tendency to value ‘abstract
knowledge’ and devalue the more practical kind, and to separate learning from
practice. An aspect of this, they argue, is that too much weight is put on the
organizations’ formal descriptions of work, overlooking the evolving communities-
of-practice as significant sites of innovating, through their ‘constant adapting to
changing membership and changing circumstances’ (Brown and Duguid 1991, 41).
To get a grip of the difference between abstract and practical knowledge they
introduce the concepts of ‘canonical’ and ‘non-canonical’ practice. Canonical
practice refers to the formal practice often made explicit in Total Quality
Management (TQM) systems, company visions and values, while non-canonical
refers to actual practice often tending to be tacit and embodied, difficult to make
explicit, but evolving in the working communities. The non-canonical practice – and
the knowledge and experience related to it – play an essential role when employees
have to deal with unexpected events or risky situations, which often is not made
explicit in company procedures or rules.
Culture as process – not as noun

Finally, most modern cultural theory tends to treat culture as a process (or
processes), rather than a noun (Alvesson 2002; Keesing 1994); i.e. noun refers to a
view of culture as something you possess and imply a static view of. The view of
culture as a noun is quite common within some organizational culture perspectives,
especially within approaches which focus on unity, integration and strength.
Another variant is to view ‘safety culture’ related to some kind of maturity level
which characterizes the way an organization deals with safety aspects (Westrum
1993; Reason 1997; Parker, Lawrie, and Hudson 2006). Cultural approaches
focusing on process, however, treat culture as always being present – unfolding its
dynamic in everyday practice. However, this does not mean that the culture
automatically should be regarded as sound, safe or ‘healthy’.
Organizational culture
The concept ‘organizational culture’ gained a lot of attention in the 80s – referred to
as the corporate culture boom (Haukelid 1996; Alvesson 2002). Corporate cultures
were seen as a prerequisite for performance, growth and success (Peter and
Waterman 1982). However, excellent companies experienced a decrease in
performance, which challenged this view. The corporate perspectives were then
replaced by more rational business concepts focusing on – among other aspects –
individual behaviour (Alvesson 2002).
Both corporate and other ‘new’ perspectives, also within safety research, put
strong emphasis on management and leaders as important role models (Zohar 1980,
2003; Zohar and Luria 2005; Flin et al. 2000; Guldenmund 2000). A leader-driven
approach emphasizes management’s ability to influence others and seems to have a
strong belief in charismatic leaders – often without taking into consideration the
context of where leadership is performed or the meaning workers attribute to
leadership. There are also a lot of examples of how leaders come and go, but the
organizational culture remains the same (Haukelid 1996; Alvesson 2002). All in all,
the management literature often implies that culture is a variable or an
organizational aspect that can be managed, changed or manipulated according to
whim. This kind of instrumental approach to culture is seen as problematic by others
who claim that: ‘Culture cannot be managed, it emerges’ (Martin 1985, 95).
Advocates of this latter view treat leadership more as culture influencing activity;
leaders then work through culture and have to tune themselves according to other
members of the organization, external and internal processes and conditions
influencing the organizational culture.
Several authors have pointed out that organizations consist of systems of
meaning which are common in varying degrees (Frost et al. 1991). Frost et al. (1991)
divided organizational culture into three main perspectives: integration, differentia-
tion and fragmentation. Subcultures and conflicting interests are often treated as a
threat, especially within the integration perspective, which asserts the importance of
unity, strength and agreement; i.e. a corporate culture. The differentiation
perspective put into focus the existence of several sub-cultures within one single
company, often living harmonically side by side within the same organization. In the
fragmentation perspective, sub-cultures, ambiguity and fragmentation are given
more weight. The same event or act can be interpreted differently by different people,
and conflicting interpretation frames constitute an important part of the company’s
real life – dividing the organization into various sub-cultures.
Decision-making approaches to safety

As a part of organizational safety studies, there are different socio-technical
approaches to organizational safety under headings like High Reliability Theory
(HRT), Normal Accidents Theory (NAT) named by some as the ‘Technical route to
safety’ (Gherardi and Nicolini 2000a). They differ in some important ways; among
others on how they view the possibility of preventing accidents or if accidents
constitute a ‘normal’ part of everyday organizational practice. According to Weick

(1987), one has to distinguish between a view of organizations as decision makers
and organizations as interpretation systems that generate meaning. And in the
management of organizations one has to be concerned with both.
Metaphors much used within the ‘technical routes’ are often (very roughly)
related to individual mental states being lifted or transformed into collective level
variables describing what it takes to create reliable, collective action; e.g. collective
mind, heedful interrelating (Weick and Roberts 1993; Weick, Sutcliffe, and Obstfeld
1999) or according to Reason (1997), a sound safety culture is described as a
reporting, just, flexible and learning culture. In summary, these approaches have
much in common with ideas from the corporate culture literature, making culture a
matter for management to change, improve or influence by structural, technical or
‘soft’ human barriers.
Summary – theoretical approaches

BBS implies little or no focus on the cognitive aspects of human mind, sense
making and meaning systems – as opposed to cultural approaches. According to
DeJoy (2005) the strengths of the BBS approach can be related to its specific
technology, the systematic collection and analysis of observed data, its focus on
frontline personnel, a participatory approach and its positive focus on desired
safety behaviours. DeJoy characterizes BBS as a bottom-up strategy, much because
of its involvement of frontline personnel, and cultural approaches as top-down.
However, it could be argued that most BBS interventions are defined from ‘the
top’ and might foremost be anchored among the management. Also, BBS
interventions are mostly driven by safety consultants or experts and do not
originate from the frontline personnel themselves – even if frontline personnel
participate and are involved in the intervention. In contrast, DeJoy (2005) defines
cultural approaches as top-down driven. However, theoretically, a cultural safety
approach might have as its main departure the ‘bottom’ or at least start with an
understanding of the main challenges characterizing everyday safety practice and
sense-making processes. However, DeJoy’s point and observations might well
correspond with most of the cultural interventions described, accomplished and
published within the safety literature.
As an alternative to the model described by DeJoy, we argue for a perspective
which takes as its point of departure the real communities of practice, the employees’
construction of meaning towards work-related risks and safety, to put more weight
on tacit/embodied knowledge, to balance the formal and informal sides of the
organizational culture and also to take a look at the external and internal conditions
influencing the employees’ possibility of behaving safe or unsafe. When trying to
understand cultural processes, one also has to take into account aspects of power.
Some people have more influence than others, and some are closer to the risk than
others; safety and risk are not equally distributed (Tharaldsen and Haukelid 2007;
Tharaldsen, Olsen, and Rundmo 2008). A cultural approach often has a holistic, and
inter-subjective starting point and differs in this respect from BBS approaches.
The BBS upholds as a strength the methodological ‘technology’ in use. We would
rather argue for methodological diversity and combination of methods useful for
giving a better understanding of various safety phenomena. And it is about time to
eliminate delusions of qualitative methods as being subjective and intuitive, which

are pointed out as the weak sides of cultural approaches (DeJoy 2005).
Keeping the balance

We will now introduce our four-field model, ‘A balanced approach to safety’, which
focuses on the dynamics between two axes: (1) tacit–explicit sides of culture; and (2)
cultural versus behavioural aspects, illustrated in Figure 1.
The model is meant to comprise both behavioural and cultural approaches to
safety and the differences between them – even if they at times might have similarities
and are complimentary to each other. We also introduce this model in order to
understand the strengths and weaknesses of both perspectives. One of our main
concerns was to avoid a narrow focus on sudden, observable, countable risks at the
cost of long-term influence, slow change and not-so-easily observable phenomena.
Another intention with the model is to avoid a too one-sided focus on individual
behaviour, which may result in a blaming culture, i.e. where accidents, near misses or
events are explained by individual mistakes.
This section starts with the placement of cultural and behavioural approaches to
safety in relation to the model. Locating such comprehensive theoretical perspectives
in a four-field model implies a coarse simplification, but will hopefully bring some
clarity as a result. Thereafter, each of the four fields in the model will be explained
and illustrated by empirical examples.
Cultural and behavioural aspects in the model

As referred to in the theoretical framing earlier, a cultural perspective often promotes
a holistic approach and aim to providing understanding of different aspects related
to the tacit, hidden and habitual sides of culture, but also keeps an eye on the more
visible and explicit aspects. A cultural analyst will be less willing to treat culture as
manageable matter and look upon culture as stable, related to processes of meaning
construction, sometimes operating beneath awareness and hard to change. A BBS
Figure 1. A balanced approach to safety.

perspective focuses on explicitly expressed, observable acts and how these acts cause
desired or undesired behaviour. A much-used slogan is ‘If you can measure it, you
can change it!’. A BBS programme will not be concerned with tacit, embodied
knowledge, history, socialization or external factors; in short conditions that may
influence or help us understand human behaviour in relation to its contextual
frames. A BBS project will look into the specific context of the employees, diagnose
and treat the critical behaviour and, hence, work with the outcomes of culture. In
this respect, behavioural and cultural analyses have different starting points. Both
will often start with what is observable, explicitly expressed and try to get an
understanding of ‘What is going on here?’. They often use similar methods in order
to examine an addressed phenomenon or the problem(s) in focus, and to get a sound
picture of safety challenges in the organization.
Thereafter their ways depart. In the cultural approach data is interpreted in
relation to the dynamics of the context; the relation between the actor, the situation,
the technology in use etc., searching for the driving forces influencing ‘what’s going
on’; like economical frames, project organization and rewards, values in use,
strengths and weaknesses, what and how to improve, etc. Important success criteria
in a cultural approach would also be related to the involvement of management and
employees in the intervention to come. Hence, the top-down mark on cultural
approaches may be unsuitable or at least dependent on how the actual intervention is
carried out. In a BBS project the data gathered serve as input for the behavioural
intervention to be implemented and only what is observable, measurable and
possible to change is of interest. This implies a more limited locus of interest, which
also might make the intervention easier to follow and to document the results from.
This is well in accordance with DeJoy’s outlining of the methodological strengths:
simplicity and documented results on employees’ safety behaviour. However, some
of the weaknesses or problems associated with BBS interventions may have been
avoided by using a combined approach. As mentioned in the theoretical section, BBS
intervention may be at risk of not taking into consideration important social and
organizational factors such as trust, organizational justice, recognition of goals by all
involved employees and at times a too one-sided focus on the individual behaviour.
We would add the risk of not taking into consideration the dynamics between the
tacit and explicit sides of culture, also in order to understand how the intervention
could be integrated in the organizational working practice, with its formal and
informal sides.
Field I: ‘Cultural and Tacit’

The Cultural and Tacit (CT) dimension is related to social and organizational
stability, history, tradition, learning processes and socialization. Processes that
build organizational identity and influence the organizational member’s perception
of themselves in relation to their work; their roles and responsibilities and how
they deal with safety, in short: ‘The way we do things around here’. It involves the
more tacit and hidden sides of culture, traits that are not easily observed and often
taken for granted or just regarded as common sense. Cultural blindness and
ethnocentricity may be problems associated with this phenomenon, so will
concepts like embodied and tacit knowledge. We have several types of words and
phrases for this, which relate to different social levels; individual, organization or
societal: It sits in ‘your fingers’, ‘your hearts and minds’, ‘your head’, ‘in the walls’,
‘under your skin’ etc.
Processes of power also form part of this. Symbolic power are characterized by
being silent, not made explicit and taken for granted by its members (Bourdieu
1979). When seen in relation to safety, we find that risk and dangers are unequally
distributed on different levels; individual, work group, organizational, regional
and societal (Douglas and Wildavsky 1983). According to Douglas and Wildavsky
(1983) we neither know the risks we face nor do we perceive them in the same way:
we neither agree on what is risky, how risky it is or how to deal with it.
Catchwords for this field in the model may be slow change, slow learning and slow
fix.
To illustrate this dimension we have chosen enforcement of new national laws in
order to change risky behaviour, improve public health and reduce accidents.
Authorities, for instance, try to change citizen behaviour by enforcing new laws and
regulation; the wearing of seat belts when driving, prohibition of smoking in

restaurants or public buildings, national and international traffic rules etc. Such laws
and regulations might meet a lot of resistance in the beginning. However, resistance
often diminishes or vanishes after a period of adjustment and slowly becomes an
intrinsic and automatic part of individual behaviour and culture. When travelling or
experiencing different manners of conduct, we become surprised, annoyed or
provoked when others have another way of thinking or behaving. The smell of
smoke and cigarettes may be experienced physically differently today than not more
than a few years ago. We automatically put our seatbelts on today, but drove
ignorantly around without them on not many decades ago. If we relate this example
to our model, we see that all dimensions are affected; long-term cultural change, tacit
and explicit sides of culture and behavioural aspects.
Field II: ‘Cultural and Explicit’

The Cultural and Explicit (CE) dimension points to observable, explicitly expressed
and often formal sides of culture. It refers to the culturally influenced, observable
and emerging work practice and symbols related to it. Expressions of CE might be
visualized in company vision and values, logos, flags, TQM systems, work permit
system, procedures of ‘safe job analysis’ etc., and ought to be in correspondence with
the tacit sides of culture and the actual work practice. The concepts of canonical and
non-canonical practice discussed earlier in the theoretical section illustrate this
dynamic. If we hold on to the seatbelt example, new national laws and regulations
may be regarded as a tool trying to guide human behaviour in desired directions, but
not as the actual practice. Formal and explicitly formulated laws intend to guide
action and might have to be monitored or kept under surveillance in order to be
followed or fully integrated as common practice.
Management commitment to safety has been shown to have an impact on the
organizational safety climate (Zohar 1980, 2003). Hence, managers are supposed to
play an important role as models towards their ‘subordinates’. Sayings like ‘walk the
talk’ express an expectation towards managers as performing according to rules and
procedures. Their behaviour is supposed to be in correspondence with company
vision and values, and they ought to demonstrate correspondence between what they
say they should do and what they actually do.
Field III: ‘Behavioural and Tacit’

The Behavioural and Tacit (BT) aspects of the model refer to behaviour which is
anchored in habits and the locus of interest is on the driving forces guiding individual
and group behaviour. External incentives, sanctions or goals influence individual
behaviour and might lead to habitual behaviour or a work practice which is safe or
risky. When the goal is to improve safety, it becomes crucial to understand the
dynamics between actual behaviour and how it is influenced by incentives in the
employees’ daily work situation. Well-intended incentives or systems might produce
paradoxical consequences on individual and group behaviour – slowly becoming a
part of bad habits. A safety intervention on the Norwegian Continental Shelf looked
into different types of not-intended consequences or side-effects from originally well-
intended systems sometimes causing a worsening of originally harmless situations
(Tharaldsen and Johannessen 2006; Tharaldsen, Olsen, and Eikeland 2003).
Monetary bonuses combined with the vision of zero accidents, could lead to
underreporting of serious accidents or incidents. Positive intentions expressed in

phrases like ‘always take the time for safety’, ‘safety first’ etc. might be disturbed by
shortage of time, contractual issues, production pressure, lack of knowledge and
experience, unsuitable design, bad technology etc. If important background factors
are not taken into consideration when carrying out safety interventions, desired
goals would be difficult to reach. In this field (BT), the time aspect becomes
important. Habits are developed over time and are not to be considered as one-time
effects. However, these kinds of habits are not considered as stable and deeply rooted
as in the CT dimension.
Field IV: ‘Behavioural and Explicit’

Field IV, Behavioural and Explicit (BE) aspects, implies attention towards observable
single acts or events, events which often are sudden or dramatic in their character.
These kinds of acts or events attract our attention. They are observable, easier to
measure or count and hence easier to fix – sometimes at the sacrifice of slowly
developing and sneaking risks. This dimension relates both to micro- and macro-
accidents, events or incidents. Major accidents have dramatic effects on future safety
prioritization and might totally alter company or national strategy as a consequence.
Major accidents in the oil and gas industry like Piper Alpha on the UK Shelf or
Aleksander Kielland at the Norwegian Continental Shelf had severe outcomes, both
materially and fatally. However, later on they contributed to a more humble approach
and heavier focus on risk and safety for the whole industry both in the UK and
Norway.
Not so easily observable and long-term developing risks, like dangers threatening
employees’ health or the environment, have not easily gained much attention.
However, both on a societal level and in work life we seem to be more engaged in
these kinds of risks than before; the hidden, unpredictable ones, developing over
time; e.g. chemical exposure, organizational complexity, numerous organizational
interfaces, complex supply chains, technical vulnerability etc., finding their way
behind protected or unprotected lines. It is never easy to find ‘quick fixes’ and
change is hard to measure. The relation between cause and consequence is also often
blurred. Attention towards observable, sudden, countable, micro- or macro-events
tends to be reactive in character and might win at the sacrifice of long-term planning
and awareness of macro, slow, industrial, organizational or external changes, which

directly or indirectly influence the work practice and the overall picture of risk and
safety.
Discussion: the model and its implications for safety interventions

As mentioned earlier, to grasp all sides of a cultural approach to safety in a four-field
model implies a gross simplification and should be treated as such thereafter.
However, we hope our model provides an outlining of distinctions between cultural
and behavioural approaches and gives analytic insight into their characteristics.
They both have their strengths and weaknesses, and they might be used as
complimentary or corroboratory to each other. The model is meant to bring us
towards a more unified picture of the differences and uniqueness of the different
methods, not to point at similarities.
In a balanced approach to safety, both tacit and explicit sides of culture are
embraced; for example, shared cultural traits, group characteristics, social (formal
and informal) structure, symbols, processes of power etc. Common practice,
‘sharedness’ or diversity is understood in relation to external framework conditions
or expectations directed towards the individual or groups of individuals. Observable,
safe or risky behaviours are understood in relation to the tacit and taken-for-granted
sides of culture, and the underlying reasons for acting in certain manners are
searched for.
A balanced approach concentrates on how canonical (formal) practice is related
to non-canonical (informal) practice: does the map correspond with the terrain? Do
formal procedures reflect best practice often formulated in management systems,
procedures, rules, company values etc.? In order to get a better understanding of the
dynamics or gaps between such formal, systemic, desired practice and the actual
practice, multi-method approaches are needed (Brannen 1992, 2004). Then a fuller
picture of why rules and procedures are not always followed, or why well-intended
incentives might have paradoxical, undesired consequences might be better
perceived.
A balanced approach searches for strengths and weaknesses in the safety culture
under examination in order to identify improvement areas, while a BBS analyst will
look for critical behaviour, set up a system of behaviour tracking, monitor the
consequences and reinforce desired behaviour. Hence, BBS results are easily
measured and improvements come quicker. According to DeJoy (2005), BBS
interventions may have long-term effects on safety culture, or on the other hand,
they might lose their effect when the intervention stops if cultural aspects are not
understood properly. A balanced approach sees culture as more stable, reproduced
through learning and socialization and sometimes hard to change. Long-term effects
may be linked to macro-changes like use of new technology – for instance more use
of integrated operations by the use of Information Communication Technology
offshore, hydraulic-based technology at the expense of manual work etc. Such
changes will slowly affect the cultural framing of organizational activities, and hence
what is regarded as safe or risky work practice. Short-term safety interventions might
contribute to a higher consciousness towards specific risks, which on the other hand
might have long-term effects on culture. Various types of safety tools like
observation and reporting systems are often used in high-risk organizations in
addition to culture programmes or BBS interventions. These kinds of tools will in

different ways focus on long-term change or sudden events, and they influence in
complex ways both the tacit and explicit sides of culture. Finally, they might become
an intrinsic part of the work culture.
There is no such thing as a perfect recipe or quick-fix solution for safety
interventions. Both technical and human vulnerability have to be dealt with –
preferably combined with sensitivity towards cultural, long-term changes and how they
influence the level of risk and safety. As highlighted by Douglas and Wildavsky (1983)
and their cultural theory of risk perception, risk is a socially constructed phenomenon
and our understandings of risk and safety should never be taken at face value.
Acknowledgements
This project would not have been possible without financial support from the Petromax
programme in the Norwegian Research Council. We are also grateful for the cooperative
attitude we have met from several petroleum companies when examining safety aspects within
the industry – with informants willing to spend time and share knowledge with us. Also,
special thanks to Professor Karina Aase (University of Stavanger) for careful reading and
comments on the paper.
References
Alvesson, M. 2002. Understanding organizational culture. London and Thousand Oaks, CA:
Sage Publications.
Bourdieu, P. 1979. Outline of a theory of practice. Cambridge: Cambridge University Press.
Brannen, J. ed. 1992. Mixing methods: Qualitative and quantitative research. Aldershot:
Avebury.
———. 2004. Working qualitatively and quantitatively. In Qualitative research practice, ed.
C. Seale, G. Gobo, J.F. Gubrium, and D. Silverman. London and Thousand Oaks, CA:
Sage Publications.
Brown, J.S., and P. Duguid. 1991. Organizational learning and communities-of-practice:
Toward a unified view of working, learning and innovation. Organization Science 2, no.
1: 40–57.
Cox, S., B. Jones, and H. Rycraft. 2004. Behavioural approaches to safety management within
UK reactor plants. Safety Science 42, no. 9: 825–39.
Deaux, K., and L.S. Wrightsman. 1984. Social psychology in the 80s. 4th ed. Monterey, CA:
Brooks/Cole Publishing Company.
DeJoy, D.M. 2005. Behaviour change versus culture change. Divergent approaches to
managing workplace safety. Safety Science 43: 105–29.
Depasquale, J.P., and E.S. Geller. 1999. Critical success factor for behaviour-based safety: A
study of twenty industry-wide applications. Journal of Safety Research 30, no. 4: 237–49.
Douglas, M., and A. Wildavsky. 1983. Risk and culture. An essay on the selection of technical
and environmental dangers. Berkeley, LA and London: University of California Press.
Flin, R., K. Mearns, P. O’Connor, and R. Bryden. 2000. Measuring safety climate: Identifying
the common features. Safety Science 34, nos 1–3: 177–92.
Foucault, M. 1977. Discipline and punish: The birth of the prison. Oslo: Gyldendal.
Frost, P.J., L.F. Moore, M.R. Louis, C.C. Lundberg, and J. Martin, eds. 1991. Reframing
organizational culture. London and New Delhi: Sage.
Geertz, C. 1973. The interpretation of culture. London: Fontana Press.
Geller, S. 2001. Behaviour-based safety in industry: Realizing the large-scale potential of
psychology to promote human welfare. Applied and Preventive Psychology 10, no. 2: 87–105.
Gherardi, S. 1999. Learning as problem-driven in the face of mystery? Organization Studies 20,
no. 1: 101–24.
———. 2000. Practice based theorizing on learning and knowing in organizations.
Organization 7, no. 2: 211–23.
Gherardi, S., and D. Nicolini. 2000a. The organizational learning of safety in communities of
practice. Journal of Management Inquiry 9, no. 1: 7–18.
———. 2000b. To transfer is to transform: The circulation of safety knowledge. Organization
7, no. 2: 329–48.
Gherardi, S., D. Nicolini, and F. Odella. 1998. Toward a social understanding of how people
learn in organizations. The notion of situated curriculum. Management Learning 29, no.
3: 273–97.
Guldenmund, F.W. 2000. The nature of safety culture: A review of theory and research. Safety
Science 34, nos 1–3: 215–57.
Haukelid, K. 1996. Kultur og bedriftskultur i antropologi og ledelsesteori. In I teknologienes
tegn [In the character/sign of technology], ed. E. Lie, S. Myklebust, and T. Hviid Nielsen,
137–59. Norway: Universitetsforlaget.

House, R.J., P.J. Hanges, M. Javidan, P.W. Dorfman, and V. Gupta, eds. 2004. Culture,
leadership, and organizations. The GLOBE study of 62 societies. Thousand Oaks, CA:
Sage.
Johnson, S.E. 2003. Behavioural safety theory. Understanding the theoretical foundation.
Professional Safety October: 39–44.
Keesing, R.M. 1987. Anthropology as an interpretive quest. Current Anthropology 28: 161–76.
———. 1994. Theories of culture revisited. In Assessing cultural anthropology, ed. R.
Borofsky. USA: McGraw Hill.
Krause, T.R., K.J. Seymour, and K.C.M. Sloat. 1999. Long-term evaluation of behaviour-
based method for improving safety performance: A meta-analysis of 73 interrupted time-
series replications. Safety Science 32, no. 1: 1–18.
Lingard, H., and S. Rowlinson. 1997. Behaviour-based safety management in Hong Kong’s
construction industry. Journal of Safety Research 28, no. 4: 243–56.
Martin, J. 1985. Can organizational culture be managed. In Organizational culture, ed.
P. Frost. et al. Beverly Hills, CA: Sage.
Parker, D., M. Lawrie, and P. Hudson. 2006. A framework of understanding the development
of organizational safety culture. Safety Science 44: 551–62.
Peters, T., and R. Waterman. 1982. In search of excellence. New York: Harper and Row.
Reason, J. 1997. Managing the risk of organizational accidents. Aldershot: Ashgate.
Rohner, R.P. 1984. Toward a conception of culture for cross-cultural psychology. Journal of
Cross-Cultural Psychology 15, no. 2: 111–38.
Scholte, B. 1986. The charmed circle of Geertz’s hermeneutics. Critique of Anthropology 6,
no. 1: 5–15.
Seo, D.C. 2005. An explicative model of unsafe work behavior. Safety Science 43, no. 3:
187–211.
Tharaldsen, J., and K. Haukelid. 2007. The underdogs on the Norwegian Continental Shelf –
High risk, high exposure and high satisfaction! Abstract presented at NoFs, Nordic
conference on safety, 13–15 June, in Tampere, Finland.
Tharaldsen, J., and I. Johannessen. 2006. Helse og miljø for sikkerhets skyld? Policy og praksis
i Vetco Aibels HMS-kultur. Sluttrapport NFR. IRIS rapport 2006/013.
Tharaldsen, J.E., E. Olsen, and T. Eikeland. 2003. A comparative study of safety culture.
ABB offshore systems compared with other companies on the Norwegian Continental
Shelf. Proceedings of EAOHP, November 20–21, in Berlin, Germany. I-WHO
Publications.
Tharaldsen, J., E. Olsen, and T. Rundmo. 2008. A longitudinal study of safety climate on the
Norwegian Continental Shelf. Safety Science 46: 427–39.
Weick, K.E. 1987. Organizational culture as a source of high reliability. California

Management Review 24, no. 2: 628–52.
Weick, K.E., and K.H. Roberts. 1993. Collective mind in organizations. Heedful interrelating
on flight decks. Administrative Science Quarterly 38: 330–58.
Weick, K.E., K.M. Sutcliffe, and D. Obstfeld. 1999. Organizing for high reliability: Processes
of high reliability. Research in Organizational Behaviour 21: 81–123.
Westrum, R. 1993. Cultures with requisite imagination. In Verification and validation in
complex man-machine systems, ed. J. Wise, P. Stager, and J. Hopkins. New York:
Springer.
Zohar, D. 1980. Safety climate in industrial organizations: Theoretical and applied
implications. Journal of Applied Psychology 65, no. 1: 96–102.
———. 2003. Safety climate: Conceptual and measurement issues. In Handbook of
occupational psychology, ed. J.C. Quick, and L.E. Tetrick, 96–123. Washington, DC:
American Psychological Association.
Zohar, D., and G. Luria. 2005. A multilevel model of safety climate: Cross-level relationships
between organization and group-level climates. Journal of Applied Psychology 90, no. 4:
616–28.
Article IV
Safety Science 48 (2010) 1062–1072
Contents lists available at ScienceDirect
Safety Science
journal homepage: www.elsevier.com/locate/ssci
Perspectives on safety: The impact of group membership, work factors and trust
on safety performance in UK and Norwegian drilling company employees
J.E. Tharaldsen a,*, K.J. Mearns b, K. Knudsen c
a
University of Stavanger, Faculty of Social Science, Societal Risk and Safety, Norway and International Research Institute of Stavanger,
Work Life and Business Development, Professor Olav Hanssensvei 15, 4017 Stavanger, Norway
b
University of Aberdeen, Industrial Psychology Research Centre, Scotland, United Kingdom
c
University of Stavanger, Faculty of Social Science, Change Management, Norway
a r t i c l e i n f o a b s t r a c t
Article history: This study is anchored in a contractor company providing well services for platform drilling on the Nor-
Received 6 January 2009 wegian (NCS) and the UK Continental Shelves (UKCS). The research project has as its point of departure
Received in revised form 18 May 2009 the potential influences of group level characteristics, structural work factors, trust, and safety behaviour
Accepted 26 June 2009
on safety performance. Do perceptions and performance differ across Shelves? Are ‘‘nomadic” groups or
employees that have more unpredictable shift rotations more exposed to accidents than others? Is high
trust and sound safety behaviour enhancing good safety performance? The results are based on question-
Keywords:
naire data from two samples of personnel distributed across three installations on the UKCS and nine on
Relations between group membership
Work factors and trust on safety
the NCS with a response rate of 67%: N = 170 (UKCS) and N = 621 (NCS). In addition, two focus group
performance interviews were held in each country, with 15 participants in each. The results show that our model
Oil and gas industry offshore makes sense. Shelf shows a significant influence on safety performance in all but the final stage in our
Comparative analysis five-step logistic regression model, indicating that the effect may be mediated by safety compliance
Self reported safety performance and safety participation. Installations and different work teams have different exposure and structural
Functional aspects of trust and distrust with work factors matter significantly. Somewhat counter-intuitively, employees who have a ‘‘nomadic” status
regards to safety and who hold the least regular shift rotations appear to have a lower risk of being involved in incidents.
High trust in workmates buffers against incident involvement and the same applies for high safety com-
pliance. The results, challenges and implications for research and safety practitioners are discussed.
1. Introduction 2003), reduced transaction costs (Dyer and Chu, 2003), increased
communication and knowledge exchange (Andrews and Delahaye,
The role of trust and distrust has been argued to play an increas- 2000), enhanced mutual learning (Bakker et al., 2006; Gubbins and
ingly important role in modern, global, complex, and ambivalent MacCurtain, 2008; Nonaka and Takeuchi, 1995) and learning from
risk societies (Barber, 1983; Beck, 1992; Giddens, 1991). Research accidents (Reason, 1997), and to have a positive impact on safety
on trust and distrust has a long history. The concepts are regarded culture and safety performance (Burns et al., 2006; Conchie et al.,
to be complex, multidisciplinary, multifaceted, and multilevel 2006; Conchie and Donald, 2006, 2008; Hale, 2000; Reason, 1997).
(Deutsch, 1958; Doney et al., 1998; Lewicki et al., 1998, 2006; Luh- The general view of trust and distrust as a bipolar construct and
mann, 1979; Rousseau et al., 1998). Despite differing conceptual- the normative view of trust–distrust as respectively good and bad
izations there seems to be some agreement on the meaning of have been attacked both within organizational (Adler, 2001; Dirks
trust and the conditions that must exist for trust to arise. It is trea- and Ferrin, 2001; Lewicki et al., 2006) and safety research (Burns
ted as a cornerstone in the construction of social order, implies po- et al., 2006; Conchie et al., 2006; Conchie and Donald, 2006,
sitive expectations about others’ intentions and behaviour, it 2008; Conchie and Burns, 2008; Hale, 2000; Jeffcott et al., 2006;
reduces complexity and conflict, involves vulnerability and risk, Poortinga and Pidgeon, 2004; Reason, 1997). In the bipolar per-
and it involves interdependence between different types of actors. spective trust and distrust are conceived of as existing at opposite
Research on trust has commonly been related to positive organiza- ends of a single trust–distrust continuum (Lewicki et al., 1998).
tional outcomes, like increased competitiveness (Seppänen et al., Low trust expectations generally become indicative of high distrust
2007), higher organizational performance (Child and Möllering, and distrust is being treated as a psychological (or organizational/
societal) disorder that should be corrected.
* Corresponding author. Tel.: +47 51 87 51 58. In this study we lean towards Luhmann’s (1979) reflections on
E-mail address: jorunn.tharaldsen@iris.no (J.E. Tharaldsen). trust and distrust. Luhmann treated trust in the broadest sense as
doi:10.1016/j.ssci.2009.06.003
‘‘confidence in one’s expectations”. Showing trust then means The influence of culture on trust building processes has also
anticipating the future, even though it is not done blindly and only been touched upon (Doney et al., 1998; Madhok, 2006; Sullivan
in certain directions: . . .trust reduces social complexity by allowing et al., 1981; Zaheer and Zaheer, 2006). Doney et al. (1998) came
specific undesirable conduct to be removed from consideration (. . .) up with 15 propositions regarding combinations of cultural traits
and by allowing desirable conduct to be viewed as certain. Similarly, (from Hofstede, 1980, Hofstede, 1984) and their anticipated influ-
distrust functions to reduce complexity by allowing undesirable con- ence on trust formation processes. Proposition one and six from
duct to be seen as likely – even certain (Luhmann, 1979, from Lew- this study suggested that more Masculine cultures and higher
icki et al., 1998, p. 444). Power Distance would be linked to a more calculative based trust
Hence, both trust and distrust depends on risks being kept under building compared to their counterparts Feminine and low Power
control, which is of crucial importance in high-risk systems and Distance cultures.
organizations. Highly complex systems are further supposed to Research on safety culture/climate and empirical results from
need both more trust and distrust, often in the case of institutional- various studies suggest that the phenomena should be examined
ized distrust (Luhmann, 1979). Hence, reliability in high-risk organi- as close to the group level or the ‘‘communities of practice” as pos-
zations – as within the oil and gas industry – might be enhanced by sible (Zohar, 1980, 2003; Zohar and Luria, 2005; Gherardi and Nicc-
creating collective requisite variety, which according to Weick olini, 2000). Risk and safety perceptions offshore have for instance
(1987) is higher when people trust each other. However, when been found to be framed by the installation you work on (hereafter
diversity increases, trust also becomes more difficult to sustain. A referred to as ‘‘installation membership”), your work area on the
solution to this could be to treat both confidence and doubt as platform, whether you belong to a client or contractor company
important in the production of reliability and to keep an eye on (Tharaldsen et al., 2008a,b), and that installation membership ex-
the limits to faith (Weick, 1987) and the benefits of doubt or distrust. plains a greater percentage of the variance on various safety cli-
If we follow the Luhmanian approach, and move towards safety mate aspects than your corporate ‘‘company membership”
specific trust and distrust – safety specific trust becomes a mean of (Høivik et al., 2009). There might be several explanations for this.
reducing social complexity by allowing undesirable conduct to be Employees are often working on the same platform for a long per-
removed from consideration, and safety specific distrust as allow- iod of time and develop a strong sense of identity towards the spe-
ing undesirable conduct to be seen as likely – even certain. Conchie cific platform or facility and the crew. Offshore workers are also
and Donald (2006) develop a model of the functions of safety spe- stuck on their installation, placed out in open sea, the same people
cific trust and distrust based on insights from Pidgeon et al. (2003, often work together for several years and they spend most of their
in Conchie and Donald, 2006) Critical Trust Model. This model is free time together offshore. Conchie and Donald (2006) found in a
also in line with the Luhmanian approach and points out that both representative study of offshore workers on the UKCS that on the
trust and distrust is necessary for effective risk regulation. The installation level the best predictors of safety performance were
division of trust and distrust into functional and dysfunctional as- attitudes towards workmates and contractors. At an industry level,
pects leads to a scheme where dysfunctional trust and dysfunc- attitudes towards offshore management were the best predictor of
tional distrust is hypothesized to give poor safety performance, safety performance. Collinson (1999) compared platform cultures
while functional trust and functional distrust should lead to sound and found they have similarities with Goffman’s total institutions
safety performance. Hence, functional distrust or ‘‘creative” mis- (Goffman, 1961); the incarceration is recurrent, but temporary.
trust could be treated as a necessary part of a sound safety culture This closeness and enforced privacy might cultivate identity based
and, conversely, ‘‘blind trust” or too much distrust would be detri- trust (Coleman, 1990) and friendly trust relations might presum-
mental for safety (Conchie and Donald, 2006; Conchie et al., 2006; ably be more important to build here than in other work settings.
Hale, 2000). The above mentioned conditions may also enhance platform and
Major oil and gas companies operate on a global basis and have crew membership and make fertile conditions for the building of
to deal with cultural differences in everyday business. Globalisa- installation specific safety cultures.
tion, cultural differences and how these processes may influence In phases of organisational change and restructuring it has also
on occupational safety have recently been reviewed and studied been found that attempts to split or break up stable work groups or
within the oil and gas industry (Mearns and Yule, 2008). With re- crews offshore meets with employee resistance (Tharaldsen et al.,
gards to safety, it is assumed that organisational safety culture may 2007). Employees who work regularly on an installation, tend to
be embedded in deeper rooted values (as cultural values) and that talk about their offshore co-workers as their ‘‘second family” (even
safety is a ‘‘universal value” (Helmreich and Merritt, 1998). How- ‘‘first family”) or of the installation as their ‘‘second home”. On the
ever, others hold the view that cultural values may have an impact other hand, other groups of employees experience a more
on perceptions of what is regarded as important organisational ‘‘nomadic” work life resulting in a more loosely coupling to specific
attributes and practice and that this may vary across cultures platforms and work environments. Mostly, nomads are to be found
(House et al., 2004). Mearns and Yule’s (2008) study used the Hof- among contractor staff, and the question arises whether these
stede cultural value dimensions (Hofstede, 1984) and their results work groups are more or less exposed to injuries than more stable
suggested that more proximal influences such as perceived man- work groups.
agement commitment to safety and the efficacy of safety measures Contractors often find themselves at the sharp end with regards
exerted a higher impact on workforce behaviour and accident rates to several aspects; having more unpredictable working hours, high
than fundamental cultural values (Mearns and Yule, 2008). Previ- production demands, being close to the sources of risk, and moving
ously, only one comparative analysis of offshore safety climate more between platforms (nomads). This involves constant adjust-
across the UK (UKCS) and Norwegian Continental Shelves (NCS) ment to various clients and local platform cultures (Tharaldsen
has been performed (Mearns et al., 2004). The results showed clear and Haukelid, 2007). Other studies also suggest that permanent
differences in how UK and Norwegian workers evaluate various so- workers (client staff) invest less trust in temporary (contractor)
cial and organisational aspects that might have an impact upon workers (Moreland and Levine, 2002), that contractors are being
safety. For most of the scales, a greater percentage of the variance treated unfairly and as inferior B-team members, and that such as-
was explained by installation rather than sector (Mearns et al., pects might create undesired side-effects (Collinson, 1999). How-
2004). Another study found differences in safety performance be- ever, stronger awareness might have developed, at least in the
tween the Nordic countries within the building and construction Norwegian petroleum industry, and some contractors now experi-
sector (Spangenberg et al., 2003). ence that their competence is being more highly valued today than
before (Tharaldsen and Haukelid, 2007). With regards to client- commitment to safety scales (Conchie and Donald, 2004) and
contractor relations Conchie and Donald (2008) show that low GLOBE organizational culture scales (House et al., 2004). Scoring
trust between client and contractors seem to be most important on all dimensions followed a Likert scale from 1 to 5 or 1 to 7
in the initial phases of a relationship. In the course of time, positive expressing individual’s perceptions and attitudes to different state-
trust relations between the customer company and its supplier ments. Certain items were recoded so that all the responses could
might be built, depending on demonstration of ability and good be compared in the same direction of one being the lowest score
workmanship. and five being the highest.
This study examines how specific group membership character-
istics influence perceptions of trust and self reported safety behav- 2.2.1. Self reported safety behaviour
iour on exposure to involvement in incidents (safety performance) Safety compliance reflects compliance with rules and proce-
across two shelves. Our dependent measure will be involvement in dures; taking shortcuts which involve some risk, ignoring safety
incidents over the past year on a specific installation. The results rules to get the job done, breaking of work procedures. Safety par-
are based on both quantitative and qualitative data. The quantita- ticipation refers to personal safety engagement and engagement
tive analysis focuses on causal relations between our measures, among employees; making suggestions to supervisors or manage-
while qualitative data are used to complement our statistical find- ment about ways to improve safety and taking personal initiative
ings and are mostly integrated into the discussion section. for improving safety at work.
Our respondents and informants are offshore workers selected
from a contractor company which provides well services in plat- 2.2.2. Trust dimensions
form drilling, engineering and well intervention on fixed installa- In this paper we use two scales measuring trust in colleagues
tions on both Continental Shelves. The largest portion of and management’s commitment to safety: Trust in workmates
employees is situated on the NCS, workers belong to different dis- commitment to safety and trust in first line supervisor commit-
ciplinary teams, experience both nomadic and stable work condi- ment to safety. The workmate trust scale is built up of items
tions and have different shift rotation patterns. reflecting your belief in your colleagues; whether they care about
your safety, if they can be trusted to support you if you had a com-
plaint about safety, etc. The next trust scale is directed at trust in
2. Methods
your first line supervisor. It contains statements related to your
supervisor’s safety values reflecting how deeply embedded ‘‘true
Several authors have urged for the use of combined methods
priority of safety” is in the supervisor’s behaviour, by being actually
and touched upon the differences or similarities between qualita-
reflected in practice.
tive and quantitative methods (Blaikie, 2000; Brannen, 1992,
Safety performance was in this study operationalized as subjec-
2004; Denzin, 1970; Hammersley, 1992). Concepts like triangula-
tive reporting of involvement in incidents and near misses (yes/no)
tion, facilitation, initiation, corroboration, sequencing, etc. have
on the installation in the last year. The Conchie and Donald study
been used in order to illustrate mixed-methods design – illustrat-
from 2006 used a similar self reporting procedure, but the exposure
ing that different types of data are brought into play in different
period and categorization were slightly different and might there-
phases of a study.
fore not be compared directly with our measure. Formulations in
The first phase of this study is based on self-completion ques-
our study were tied as closely as possible to the manner accidents
tionnaire data from two samples of personnel on offshore oil plat-
are being reported on the installations on both shelves. This was
forms on the NCS and UKCS. The second phase makes two focus
done both in order to use a measure with which the respondents
group interviews comprising of 15 participants in each. The focus
are familiar and to avoid (cultural) misinterpretations. Incidents re-
groups were organized as full day seminars – one in the UK, the
fer to involvement in undesired events (least serious) and near
other in Norway. The motivation for combining methods was both
misses reflecting situations which almost developed into accidents
to cross-validate our findings and to seek the actor’s point of view
(most serious). A control measure was also included trying to check
(Blumer, 1956) on our quantitative findings.
for underreporting or examine how the medical services are used or
Participation was done voluntarily, respondents were promised
not offshore: Have you suffered any injury on this installation in the
confidentiality and information about the study and data handling
last year and did not seek medical attention? In addition three
procedures were given to all participants. The methodological ap-
items on reporting climate were included reflecting peoples’ will-
proach of the project was approved by the Norwegian Social Sci-
ingness to report hazards, near misses and incidents.
ence Data Services (NSD).
The use of self reported safety has been questioned (Marotti
et al., 1997), but is still considered to be a useful and reliable
2.1. Sample way to collect information regarding safety performance (Hurst
et al., 1996). In a former study, validating a safety climate question-
The number of participating platforms in the survey was three naire developed in the Norwegian oil and gas context, we used a
in the UK and nine in Norway. Data collection was carried out dur- weighted version of accidents/incidents (per million man hours
ing Summer and early Autumn, 2007. The response rate was esti- per installation) reported to the Norwegian Petroleum Safety
mated at 67% on both shelves with N = 170 in the UK and Authorities. The results showed quite weak, but significant and
N = 621 in Norway. The difference in number of respondents re- logical correlations with our safety climate measures and employ-
flects the relative share of employees the company has on the ee judgements of the risks of major and minor accidents. However,
two shelves. This has to be taken into consideration when under- the measure did not show predictive qualities (Tharaldsen et al.,
standing the results. 2008a), and a self reported safety performance measure seemed
like a sound alternative, also based on the findings by Hurst et al.
2.2. Questionnaire measures (1996). The reliability of our safety measures will also be discussed
in relation to other scales (reporting climate) and reflections on
In order to examine the main goals in the study, measures from this issue from our focus group interviews.
the following instruments were included in the questionnaire: 2. Six GLOBE organizational culture scales (House et al., 2004)
self reported safety behaviour scales, self reported safety perfor- were used reflecting organizational climate on: Future and
mance (Mearns et al., 2001), trust in colleagues and management Humane orientation, Power Distance, Collectivism, Performance
and Assertiveness. Two of the dimensions will be discussed for our 2.4. Hypotheses
purposes: (1) Power Distance focusing on social distance between
superiors and subordinates and the tendency to question manage- 2.4.1. H1. Influence of cultural and group related factors
ment authority or not, and (2) Assertiveness (corresponds more or Both installation membership and Shelf will have an impact on
less with the Masculine/Feminine dimension in Hofstede, 1984) safety performance. However, we anticipate that the effect of Shelf
mirroring the degree of aggressiveness and dominance in the orga- should mainly work indirectly, through installation membership,
nization. The GLOBE study focuses on cultural differences among group identity (team/work area), trust and self reported safety
global clusters and in this study we used the Anglo (UK) and Nordic behaviour.
(Norway) cluster as a frame of reference. According to the GLOBE
study the Anglo cluster is characterized by higher Power Distance 2.4.2. H2. Stability versus flexibility
and Assertiveness than the Nordic; i.e. a larger social distance be- We anticipate that employees who are mostly or always work-
tween superior and subordinates, a lower tendency to question ing on the same installation and who have regular shift rotations
manager’s authority and a more Aggressive/dominant climate in will be less exposed to incident involvement than the nomadic
the Anglo cluster than in the Nordic. ones and employees having unpredictable rotations. These struc-
tural group characteristics are related to flexibility versus stability.
2.2.3. Groups’ specific characteristics Employees who tend to move between installations or who work
Shelf, Installation, Work Area (1 = Driller Team, 2 = Drill Floor more irregular rotations will to a greater extent have to adjust
Team, 3 = Deck Team and 4 = Technical Team), type of shift rota- themselves to new surroundings, safety systems, working cultures
tion (permanently day or night shifts, Night or day every second and sleeping patterns. We expect such continuous adjustment to
tour or Shift vary) and status on the installation (works perma- increase exposure to incident involvement.
nently or mostly on this installation or nomadic; moving between
installations). 2.4.3. H3. The influence of trust
Trust in workmates and (first line) supervisor commitment to
2.3. Statistical analysis safety will have a positive significant impact on safety perfor-
mance. We expect high trust in both workmates and supervisors
The dimensional structure and the model fit was worked out to buffer against incident involvement.
and reported in a former paper (Tharaldsen et al., 2008a,b) using
Lisrel 8.80, Confirmatory Factor Analysis (CFA). In sum; all of the
2.4.4. H4. Safety behaviour
scales showed satisfactory internal consistency, except for safety
Safety compliance and safety participation will have a signifi-
participation (a = .584). However, this factor was kept because of
cant positive impact on safety performance. Hence, high safety
the overall fit of the model, but the low internal consistency will
compliance and high safety participation should lead to less expo-
be taken into consideration when interpreting the results. Also,
sure to incident involvement.
Structural Equation Modelling indicated that the suggested factor
models fitted the data.
To use the Ordinary Least Squares (OLS) which is the basis of 2.5. The focus groups
linear regression a number of assumptions should be met (Menard,
2002; Eikemo et al., 2007). When the dependent variable is dichot- The focus group interviews were held separately in each coun-
omous, as our performance measure is, a series of assumptions are try. Desired work professions were defined by the researchers,
broken. These problems are solved by using logistic regression while the final choice of participants was done by the company.
which calculates the regression coefficients on the basis of Maxi- Mainly offshore workers were chosen, covering all relevant disci-
mum Likelihood. In sum; logistic regression predicts the probabil- plines, but also rig managers and employees within Human Re-
ity of a case to fall into the higher of the two categories on our source (HR), safety management and safety delegates were
dependent variable. In our case, how exposed you are to being in- included. The seminars where conducted as structured discussions
volved in an incident or not. Involved in incidents are coded like between researchers (team of two senior researchers and a masters
this; yes = 1 and no = 0. student in change management) and the informants. Results from
In the logistic regression the odds ratio (Exp(B)), corresponding the quantitative study were reflected upon, divided along three
confidence intervals (CI = .95), the significance levels and R-square main themes: (1) Self reported safety behaviour and safety perfor-
are reported. An odds ratio of 1 indicates a 50% probability of ‘‘suc- mance, (2) functions of safety specific trust and distrust during dai-
cess” or ‘‘failure”. If the probability of being involved in an incident ly operations (reflections upon the model from Conchie and
is higher than 50% the odds ratio will be larger than 1 and, vice ver- Donald, 2008) and (3) Power Distance and Assertiveness from the
sa, if the probability is less than 50% the odds ratio will be smaller GLOBE study (House et al., 2004). Two of our main goals were to
than 1. Hence, when the CI includes 1.0, the results will tend to be get informants reflections on the results and to make cultural com-
non-significant. It should be noted that the R-square reported in lo- parisons. The trust/distrust reflections were taped and transcribed,
gistic regression represents a pseudo-explained variance and while the other information was summarized in detailed minutes,
hence should be interpreted differently than in linear regression. distributed among the research team members for checking and
The R-square in logistic regression presupposes a two-divided sent back to the company.
dependent variable (involved or not), while linear regression pre-
supposes linear dependency. Discriminant and criterion validity 3. Results
was analysed by Pearson’s r, and independent samples t-test was
used to test whether or not there was an overall difference in em- In Norway 92% of the respondents were distributed across nine
ployee perceptions across shelves on all dimensions. Mean scores installations on the NCS and in the UK 66% of the respondents
on the scales, tests of differences between Shelves and Installations worked on three installations. The rest of the respondents in the
and a frequency distribution of the safety performance measures UK belonged to the Roving crew (16%) and non-operating installa-
are provided in ‘‘Appendix” – as are the correlations between the tions (21%). Forty-three percent of the UK employees report to
dimensions. Dummy variables were made for all group specific work each tour on the same installation, against 69% of the
variables. employees in Norway. Hence, a significantly higher portion of the
employees on the UKCS have a more nomadic work life compared (which our company is) and its employees, this issue constitutes a
to their Norwegian colleagues. Both in the UK and Norway most of matter of concern. The zero vision is an outspoken goal in the over-
the respondents work Fixed shifts with 14 nights/14 days every all industry, on both shelves, and injury statistics are being used in
second tour, 56% in the UK and 52% in Norway. Twenty-six percent contract negotiations.
of the workforce has permanent dayshifts in UK, against 10% in
Norway. Two percent of the respondents on both shelves work per- 3.2. Mean scores on all dimensions across shelves and installations
manent night shifts. For 13% of the respondents in the UK shift
arrangements vary, against 11% in Norway. A fairly high proportion Table A3 shows the mean scores and test of differences on trust
of the employees in Norway work Roll over shifts (25%), with 7 scales, safety behaviour and GLOBE dimensions/items across
nights/7 days or reverse. This shift rotation does not exist in the shelves. Trust in workmates shows a slightly higher, but not signif-
UK. icant, score among the Norwegian workers than their UK col-
In the UK 18% of the group belong to the Driller Team (Toolpus- leagues. On the other side, trust in first line supervisor
hers and Drillers) against 12% in Norway. In the UK 25% are placed commitment to safety is rated significantly higher among the UK
in the Rig Floor Team (Assistant driller, Derrickman, Floorman, respondents than the Norwegians.
Roughneck) against 31% in Norway. On the UKCS 23% are in the The self reported safety behaviour scales show that UK workers
Deck Team (Deck foreman, roustabout) against 18% on the NCS. report significantly higher safety compliance, while the safety par-
29% in the UK are in the Technical Team (Maintenance, electrician, ticipation dimension is significantly higher rated among the Nor-
mechanic) against 21% in Norway. Four percent in the UK belong to wegian employees. Examining the two GLOBE items on Power
the Service team (Materials, Rig safety training, Other) against 18% Distance and Assertiveness, differences are all significant. The ten-
in Norway. dency to question their boss when in disagreement is higher
among the Norwegian workers, the social distance among superior
3.1. Safety performance and reporting climate and subordinate is lower and Assertiveness is rated as lower than
in the UK sample.
For involvement in incidents, we found that 34% of all employees When checking for differences across installations (ANOVA)
had experienced such an event and the difference between shelves only three out of seven dimensions/items discriminate in a signif-
(UKCS = 21%, NCS = 37%) and installations is significant with icant manner: Trust in supervisor commitment to safety
v2 = 0.000 and v2 = 0.002, respectively A total share of 12% had (sig. = 0027), safety compliance (sig. = 0.000) and Assertiveness
been involved in a near miss, with 8% of the respondents on the (sig. = 0.000).
UKCS and 13% on the NCS. The difference between shelves on this
indicator was not significant (v2 = 0.104) while installation vari- 3.3. Correlations
ance was (v2 = 0.023). Further, 5% of the total sample had suffered
an injury without seeking medical attention, 4% among the UK The next step was to examine the correlations between the
employees and 6% of the Norwegians. Differences across shelves dimensions. The analysis was performed separately for the two
were not significant (v2 = 0.223), neither were differences across shelves in order to check that our dimensions and items performed
installations (sig. = 0.978). Complemented with results on the approximately similarly on both. Table A4 shows the correlations
reporting climate items, the results show a slightly, but significant and their corresponding significance level for the UK respondents
higher score on the reporting willingness scale for the Norwegian above the diagonal and for the Norwegians beneath the diagonal.
offshore workers (4.41) compared to the UK sample (4.28). How- The correlates range from 0.355 to +0.590. In the Norwegian sam-
ever, when looking at single items, only the willingness to report ple all correlations are significant. Both of the trust scales and the
near misses is significantly higher in the Norwegian sample. When self reported safety behaviour scales are positively correlated with
checking for differences across installations, none were significant. each other, while Assertiveness (aggressiveness and dominance)
In comparison, the Conchie and Donald (2006) study on a rep- and the Power Distance items are positively correlated with each
resentative sample of offshore workers on the UKCS (N = 203), other, but negatively related to all other scales and items.
showed that 43% of the employees reported involvement in near In the UK sample, the trust scales and safety behaviour scales
misses while working on a gas installation. The lower share of follow the same pattern as the Norwegian sample with significant,
workers exposed in our study may be explained by the fact that positive correlations, except that the relationship between safety
the exposure periods are different – reflecting a specific period participation and safety compliance is not significant. The ten-
(the last year) in our study, while referring to total offshore work dency to question your boss when in disagreement is significantly
exposure in the Conchie and Donald study. and negatively correlated with both trust and safety behaviour
From discussions in the focus groups, it was further argued in dimensions, while the second Power Distance item (distance be-
both groups that the reporting focus has changed in a positive tween superior and subordinate) is negatively related to the trust
manner in the last few years towards a higher openness on both and safety behaviour scales, but only significantly with safety par-
shelves and that the overall prioritization of safety had become ticipation. Assertiveness is negatively correlated with both trust
higher – compared to earlier days when safety seemed to be over- scales and safety participation, but only significantly with trust
whelmed by production demands: ‘‘Production used to be more re- in supervisor – and unexpectedly positively related to safety com-
sult driven before. Now safety is the number one priority. None of the pliance. The correlation pattern among the GLOBE dimensions is
guys feel uncomfortable about stopping the work” (UK participant, almost the same among UK respondents, except for Social distance
UKP). It was also commented that incentives of ‘‘hiding” or under- between superiors and subordinates being weakly, but negatively
reporting injuries or accidents might be stronger on the UK shelf, related to Assertiveness.
due to differences in the regulation of payment. UK workers lose
approximately half of their payment if they have a lost time acci- 3.4. Logistic regression
dent, while Norwegian workers keep full payment while they are
sick or recovering from accidents. A new trend was the reporting Stepwise logistic regression has been argued to be useful in
of positive (safe) behaviour/events. However, the reporting of more exploratory research, where theory construction rather than test-
serious injuries or accidents involving lost time or medical treating is the primary goal (Menard, 2002). In the analysis, only
ment still seems to be challenging – on both shelves. For suppliers personnel situated on the platforms were selected (N = 680). The
variables introduced sequentially may have a mediating or a mod- ship to the platform do not become significant before the safety
erating role. Mediation of effects involves one or more intervening behaviour scales are integrated in model 5 – showing another sup-
variables presumed to ‘‘transmit” some of the causal effect of prior pressor effect. And, surprisingly, employees who have nomadic
variables onto subsequent variables (Kline, 2005 from Larsson, work life show lower risk of being involved in incidents than those
2008). In some cases a variable may appear to have a statistically being mostly or always on the same installation. Hence, H2 has to
significant effect only when another variable is controlled or held be partly rejected. Stable or flexible work conditions matters for
constant. This is called a suppressor effect (Agresti and Finlay, exposure to incidents, but in the opposite direction than expected.
1997, p. 368 from Menard, 2002, p. 64). While other relations Step 4 in our model integrates trust in workmate and supervisor
may be moderated or disappear when controlling for other commitment to safety. Only trust in workmates exerts a signifi-
variables. cant, positive influence on safety performance, while supervisor
Table A5 shows the results from our stepwise logistic regression trust is not showing a significant impact. The trust scales show
model. Our dependent safety performance measure was given the the same influence pattern also after the safety behaviour scales
following coding: Involved in incidents = 1 and Not involved in have been included in the final model. So to conclude, H3 is partly
incidents = 0. Involvement in near misses and suffering injuries accepted, with trust in workmates showing a positive impact on
without seeking medical attention were also tested. However, safety performance.
none of these measures played a significant role in a similar model. The last model 5 includes our two self reported safety behaviour
The building of the model is based on the logic of our four scales. Both safety compliance and safety participation show a sig-
hypotheses and how they might influence employees’ involvement nificant impact on safety performance. However, safety participa-
in incidents. The most general and basic contextual characteristics, tion indicates an almost contradictorily positive relation which
Shelf and installation membership are placed in the first step, should be interpreted as the higher safety participation the higher
thereafter we move towards group specific characteristics (step exposure to incidents. As mentioned earlier, the safety participa-
2), stable/flexible membership to the installation and rotation pat- tion scale showed low internal consistency and should therefore
tern (step 3), trust scales (step 4) and, finally, the safety behaviour be treated with care. The same scale also showed a non-significant
scales (step 5). The results of the total model would have been the correlation with safety compliance among the UK respondents in
same even if the variables were introduced in a more random way. Table A4 (‘‘Appendix”). However, it might also be linked to the fact
However, the suppressor/moderator effects would, naturally, have that employees on the NCS showed significantly higher safety par-
turned out differently. ticipation than their UK workers, while at the same time having
All in all, the total model explains 14.9% of employees’ exposure worse safety performance. Another challenge that was raised in
to involvement in incidents. The R Square increases gradually from the focus groups is that offshore workers seem to be overwhelmed
5% in the first step, to 9.3%. 11.1%, 12.2% and reaches 14.9% in the with behavioural safety initiatives. Safety participation in general
final stage. Employees on the NCS show a higher risk of being in- may then be associated with top–down decided company or client
volved in incidents, however, only until the safety behaviour scales based initiatives – which might have counter-effects. As men-
are introduced. With regards to work area, the results show that tioned earlier, the influence of shelf disappears with the introduc-
our reference category ‘‘Driller Team” shows a significantly higher tion of these measures. On behalf of this, H4 also has to be partly
exposure than employees in the Deck and the Technical Team. Fur- accepted. Only high safety compliance lowers employee exposure
ther, the impact of shelf membership becomes gradually moder- to incident involvement significantly.
ated by the introduction of work area, nomadic or stable
position, rotation patterns and trust; the Odds Ratios (Exp (B))
show a decreasing trend, the CI becomes larger and finally Shelf 4. Discussion
membership loses its significance with the integration of safety
compliance and safety participation. Hence, the influence of shelf This study is built up around four basic theoretical assumptions
might be mediated or explained by the influence of the self re- regarding the potential influence group level characteristics, struc-
ported safety behaviour scales introduced in the last step. Some tural work related factors, trust and individual safety behaviours
installations are found to have significantly lower exposure than may have on self-rated-safety-performance. Cultural differences
others, however, only two out of twelve in the final model. A low in safety performance have earlier been found in the building
number of respondents on some of the installations, especially and construction industry (Spangenberg et al., 2003). In a study
for the UK sample, might explain some of this effect and might comparing 18 offshore installations on the UK and Norwegian Con-
have been more thoroughly examined with larger groups. Hence, tinental Shelves, Mearns et al. (2004) found that general safety atti-
H1 is accepted: the effect of shelf on safety performance works tudes were explained both by the installation employees worked
indirectly, through installation membership, group characteristics, on and the Continental Shelf on which the installation was based
trust and self reported safety behaviour. suggesting a cultural influence on safety attitudes. In comparison,
Our second hypothesis is grounded in the anticipation that social and organizational aspects were only explained by ‘‘installa-
structural conditions regarding your work matters for your expo- tion”. Also, in other recent studies, installation membership has
sure to risk. In model 3 we integrate two variables which are linked been shown to exercise similar differentiation patterns, at least
to: (1) employee membership to the installation (nomadic or sta- on the NCS (Tharaldsen et al., 2008a; Høivik et al., 2009). Our re-
ble) and (2) type of shift rotations (predictable/non-predictable). sults partly show a verification of this pattern, but due to a low
The results in step 3 show a suppressor effect for The Drill Floor number of respondents on some of the installations, predictive
Team, now also turning up as having significant lower exposure qualities may have been statistically difficult to verify. However,
compared to our reference category, and for the two other work tests of means show significant differences between installations
areas (Deck and Technical), the effect is being strengthened. The with regards to Involvement in incidents, Trust in supervisor com-
reason for this may be linked to the fact that some of the employ- mitment to safety and Assertiveness. In the logistic regression, 4
ees in all teams tend to move between installations and might have out of 12 installations are, in different stages, found to have signif-
differing shift rotations. Also, employees with varying shift rota- icantly lower exposure to incident involvement compared to our
tions seem to be less exposed for involvement in incidents than reference category.
those having more predictable rotations. The influence of being The Norwegian sample shows an overall, significantly higher
regularly on the same facility or having a more flexible member- self reported involvement in incidents. On the other hand, Norwe-
gian respondents are found to have better scores on reporting client if whole crews are moved from one installation to another
mate. In our five-step logistic regression model the influence of compared to employees working or moving alone. With crew rota-
shelf seems to work indirectly, through – through installation tions you will, to a larger degree, have greater continuity both in
membership as well as group and individual factors. However, leader and workmate relations, but you will still have to deal with
when checked for separately, t-tests of mean differences across different client leaders and have to create new relations with other
shelves show significant differences on six out of seven scales; people both in own and cooperating companies. Hence, an antici-
Trust in supervisor commitment, safety compliance, safety partic- pated consequence would be that people being close or known to
ipation and even larger differences on all our GLOBE indicators you will play a more important role in trust building processes
Power Distance (two items) and Assertiveness (two items). The and enhance ‘‘Relational trust” (Rousseau et al., 1998): We trust
correlation analysis, performed separately for each shelf, show al- the persons most that we know best (Norwegian Participant, NP).
most the same pattern on both shelves, however, some of the cor- You care most for the people in the crew. Trust takes time and needs
relations are weaker and some non-significant in the UK sample. to be built (UK Participant, UKP). This type of trust is characterized
Among Norwegian respondents high Power Distance and Asser- by reliability and interdependence in long-term interactions, recip-
tiveness are negatively correlated with all other scales, indicating rocated care and concern and gives rise to positive expectations
that high trust, high compliance, etc. go together with low Power about the trustee’s intentions – also referred to as ‘‘affective”
Distance and a less Assertive organizational climate. This pattern (McAllister, 1995) or ‘‘identity-based” trust (Coleman, 1990). It
is different among UK workers with a positive non-significant asso- may be argued that offshore work in itself promotes relational
ciation between Assertiveness and safety compliance, and a non- trust, but it might play an especially important role for the nomads
significant negative relation between Assertiveness and the cli- and employees having unpredictable work conditions – enabling
mate of questioning your boss when in disagreement. This result them to adjust to and control shifting environments. Another the-
indicates that among UK workers the climate may be quite asser- oretical assumption can be that the nomads become careful navi-
tive, but still safety compliant, and that an assertive climate may gators – in order to stay safe and to compensate for lack of
be combined with low social distance between superior and knowledge or lack of ‘‘familiarity” with new circumstances. Hence,
subordinates. being in a nomadic position may enhance both personal and orga-
As an overall judgement of the self reported performance mea- nizational alertness aimed at reducing complexity and risk. A third
sure, it seems to have quite reasonable predictive qualities. Com- issue relates to the client’s (license operator) responsibility of
pany accident reports to the British and Norwegian Safety securing safe operations, including all contractors operations and
Authorities show that the risk of having a fatal accident is four personnel. Client companies often offer their contractors specific
times higher on the UKCS than on the NCS in the period from year courses, demands cultural accommodation (vision and values),
2000 to 2008. If we anticipate that there is a relation between self use of specific behavioural systems, etc. aiming at ‘‘clan control”.
reporting of less serious incidents and company reports of fatal Such expectation of ‘‘harmonized” practices could be regarded as
accidents, our results do differ from this overall tendency. Hence, a soft control element, aimed at complexity reduction and trust
it is important again to emphasize that this study involves only building by an explicit local framing of behavioural expectancies
one company – even if the sample size is statistically sound. A between the actors. In correspondence with this, qualitative find-
one-organization study puts certain limitations regarding the pos- ings indicate that employees or crews coming to new platforms,
sibility to draw representative conclusions, which would require experience an at times frustrating demand to prove their ability,
another sampling technique. being stopped or slowed down by the client. Seen in the light of
Our qualitative findings indicate a potential bias towards safety and work performance such a need to prove your compe-
underreporting among UK workers (losing payment when having tence might be an expression of functional distrust and actually
a lost time accident), however, without checking or concluding contribute to a reduced risk level – even if perceived as tiring
on the amount or size of this problem we cannot say whether this and controlling by the workers themselves. At the start of a rela-
is a genuine effect or not. An overall concern on both shelves, are tionship, the client might find it necessary to make more use of for-
the potential side-effects of using injury statistics in contract nego- mal control systems in order to get proof of trustworthiness. When
tiations. It seems that on both shelves, there are ‘‘negotiations” on trust relations grow in bandwidth and richness, formal control
how to report accidents – towards making them fit with the least might be replaced by a higher degree of flexibility, and the rela-
serious category. Such undesired consequences might strengthen tionship could be ‘‘loosened up” (Lewicki et al., 1998).
the arguments and the reliability of a self report measure of less The Norwegian workers estimate workmate trust (mean = 4.28)
serious events, probably being less biased than company statistics. generally as higher than supervisor trust (mean = 4.18). Taken to-
Also, in a statistical sense a self report measure may constitute a gether with results on Power Distance, Assertiveness and safety
more sound measure, because they happen more frequently and, compliance this might reflect culturally rooted differences in lea-
as a consequence, also makes group level examination possible. der–worker relation and variations in compliance patterns across
Interesting results regarding the trust measures are both related shelves. The UK workers seem to hold a less questioning attitude
to the non-significant difference across shelves on trust in work- towards their bosses when in disagreement, stick more to the
mate commitment to safety and a significantly higher rating of (safety) rules than their Norwegian colleagues (even if the overall
supervisor trust in the UK sample. The very fact that the UK sample compliance level is high), have greater social distance between
is much smaller than the Norwegian, may have contributed to the superiors and subordinates and a climate that is more Assertive
non-significant influence of trust in supervisors in our logistic anal- (aggressiveness and dominance). These tendencies were attested
ysis. However, if these findings are valid, how can they be inter- in both focus groups and may reflect a more rule-based trust
preted? A fairly high share of the employees on both shelves is among UK workers, while being more based on equality and dem-
found to move between installations; i.e. 67% on the UKCS and ocratic values among the Norwegians. Further, some of the UK par-
31% on the NCS. The regression analysis also shows that the effect ticipants claimed that the old Assertive climate was changing:
of our group level variable, work area (team), is strengthened by ‘‘There used to be no trust in leaders and no cooperation. This has
the introduction of the flexibility/stability variables. In itself, work- changed due to leader training and a great focus on values. Manage-
ing on various installations over different periods of time, may en- ment has changed their attitude and do now reflect new values. We
hance the importance of trusting workmates at the expense of respect each other and that has increased mutual trust” (UKP). ‘‘There
managers or supervisors. The consequence can, however, be differ- is no more bullying culture. Well, it is different from platform to
platform. There is still shouting and screaming on Platform X” (UKP). aspect and (cultural) contextual factors. As a positive side-effect it
‘‘In Norway the leaders are easier going, and equal to workers. Here in also made our communication of the results to our contributors,
UK we look more up to leaders. There is not such a gap in Norway” the public and, hopefully, also researchers and safety practitioners,
(UKP). easier.
As an overall impression; trust, but with an element of control The study is anchored in one company having operations on
and functional distrust was expressed in the focus groups as being both the Norwegian and the UK Continental Shelves, which puts
good for safety performance. ‘‘If there is too much trust, people are certain limitations on our possibility of drawing more general con-
not doing what they are supposed to do. Then they are taking a chance, clusions. However, some findings are of a more general character
a risk. It is abdication of responsibility” (UKP). ‘‘Too much trust can and some challenges need to be further investigated. The use of
lead to disaster, in worst cases death” (NP). ‘‘The downside to trust various safety performance measures constitutes such an issue.
is if you admire someone. Then you can lean back too much, you are The same performance measure may be difficult to use across
not alert” (NP). ‘‘I trusted someone too much once, and it caused an industries. However, within one industrial setting, comparisons
accident because I did not check. I was taking over for someone, and are made difficult if wording, exposure time and categorization
all I had to do was to press that button. I did not check. I trusted are different. Hence there is a need to make an evaluation of mea-
him. It caused basically a spill. I always did check myself, but for some sures in use, their predictive qualities and reliabilities in order to
reason I just trusted him on that occasion” (UKP). agree upon or make recommendations of which to use.
On both shelves, the challenge of following rules or procedures Some of our findings would be worth investigating in a larger,
blindly, having too much respect, not checking, not being alert, etc. representative sample. It would, for instance, be interesting to
sum up potential downsides to trust, expected to give poor safety examine more thoroughly how client – contractor relations are
performance. From the regression model, trust in workmates built, for example what is the relation between institutional trust,
seems to have a positive function and a significant impact on safety distrust and control elements in securing safety in complex supply
performance. This tendency may also partly be explained by the chains? In an industrial setting like the oil industry, or similar set-
nomadic work our contractor employees experience. All in all, tings characterized by complex client-supply chains, organiza-
interpersonal trust relations seem to be influenced and restrained tional borders become too limited. Safety seems rather to be
by factors at various levels; societal, structural, organizational built in both horizontal and vertical interfaces, which also makes
and interpersonal elements. it difficult to grasp.
This study has attempted to look into the potential conse-
5. Conclusions and implications quences of cultural differences and how they might influence orga-
nizational performance, trust building relations and safety
Our main results highlight interesting theoretical dynamics be- performance. However, there seem to be some culturally rooted
tween group level characteristics, structural work factors and their differences which overwhelm organizational membership. For
interplay with trust relations, organisational safety and safety per- organizational decision makers operating in global environments
formance. To move as close as possible towards ‘‘natural units” or and organizations, this constitutes a matter of concern. How is
working groups seem to make sense and stepwise logistic regres- for instance the company vision and values communicated across
sion constitutes a sound method of investigating such relations. national borders, how are leader-styles being ‘‘transported” or
Combining quantitative and qualitative methods has been ‘‘translated” in different cultural settings? How do British and Nor-
shown to be useful, giving valuable information on how to under- wegian offshore workers cooperate despite of differences in Power
stand statistical results and to throw light on more dynamic work Distance and Assertiveness?
Table A1
Safety performance across shelf and installations (Pearsons Chi-square below).
Safety perform Shelf

Norwegian Continental Shelf UKC Shelf
P1 P2 P3 P4 P5 P6 P7 P8 P9 P10 P11 P12
Involved incident 42% 36% 40% 40% 33% 24% 28% 21% 49% 30% 34% 12%
(yes) Freq 36 35 33 36 7 12 8 11 23 9 12 5
Involved near miss 7% 11% 12% 19% 38% 9% 10% 10% 17% 10% 17% 7%
(yes) Freq 6 11 10 17 8 5 3 5 8 3 6 3
Suffered injury 6% 7% 7% 4% 5% 4% 3% 6% 2% 7% 6% 2%
(yes) Freq 5 7 6 4 1 2 1 3 1 2 2 1
Involved in an incident: Shelf diff. v2 = 0.000 and installation diff. v2 = 0.005. Involved in a near miss: Shelf diff. v2 = 0.104 and installation diff. v2 = 0.023. Suffered an injury
without seeking medical attention: Shelf diff. v2 = 0.329 and installation diff. v2 = 0.978.
Table A2
Reporting climate – peoples’ willingness to report hazards, near misses and incidents across shelves.a
Reporting climate and items Shelf

UKCS NCS
N Mean St D N Mean St D
Reporting climate (index) 170 4.28 0.83 610 4.41 0.72
. . .Willing to report hazards on this installation 169 4.40 0.83 608 4.50 0.76
. . .Willing to report near misses on this installation 169 4.14 0.98 606 4.32 0.83
. . .Willing to report incidents on this installation 168 4.33 0.88 607 4.40 0.78
a
Differences across shelves: Willingness to report near misses. . . sig. = .014, p 6 0.05. All tests of differences across installations were non-significant.
For safety practitioners this study emphasizes the importance posed differently – even if they work in approximately similar
of knowing and understanding how different work groups are ex- work environments. As a consequence, safety improvements ought
Table A3
Mean scores and test of differencesa on Trust, Safety Behaviour, Power Distance items and Assertiveness across shelves.
Descriptives Shelf
UKCS NCS
N Mean St D N Mean St D
Trust in colleague commitment to safety 169 4.28 0.67 603 4.24 0.61
Trust in supervisor commitment to safety 165 4.36 0.75 604 4.18 0.74
Safety compliance 170 4.73 0.55 611 4.39 0.65
Safety participation 170 4.08 0.87 611 4.30 0.64
Company climate of questioning or obeying boss 179 3.09 1.91 616 2.62 1.71
Social distance between superior and subordinate 167 3.82 1.40 611 3.35 1.41
Assertiveness (aggressiveness and dominance) 170 4.12 0.72 618 3.55 0.81
a
p 6 0.001, p 60.01, and p 6 0.05.
Table A4
Associations (Pearson’s r) between Trust, Safety and GLOBE dimensions/items with UKCS above the diagonal and NCS below.
Dimensions TWM TSV SC SP PD_1 PD_2 Ass

Trust in colleagues’ commitment to saf – .518** .450** .397** .210** .037 .110
Trust in supervisor’s commitment to saf .590** – .451** .331** .323** .023 .165*
Safety compliance .372** .436** – .125 .160* .052 .104
Safety participation .427** .399** .254** – .282** .213** .054
Questioning boss when in disagreement .355** .313** .188** .237** – .175* .221**
Soc distance – superior and subordinate .252** .295** .127** .200** .347** – .027
Assertiveness .305** .285** .142** .167** .374** .368** –
**
p 6 0.01.
*
p 6 0.05.
Table A5
Logistic regression with Involvement in incidents as dependent variablea.
Model 1b Model 2b Model 3b Model 4b Model 5b

Variable OR CI (95) Variable OR CI (95) Variable OR CI (95) Variable OR CI (95) Variable OR CI (95)
constant 0.91 constant 1.2 constant 2.5 constant 15.3 constant 23.3
Shelfc UKCS 0.20 0.1–0.5 UKCS 0.3 0.1–0.7 UKCS 0.3 0.1–0.8 UKCS 0.4 0.1–0.9 UKCS 0.5 0.2–1.2
Inst1 0.8 0.4–1.5 Inst1 1.1 0.6–2.0 Inst1 0.7 0.3–1.4 Inst1 0.7 0.4–1.4 Inst1 0.7 0.3–1.4
Installationd Inst6 0.3 0.2–0.7 Inst6 0.5 0.2–1.1 Inst6 0.4 0.2–0.9 Inst6 0.4 0.2–1.0 Inst6 0.4 0.2–1.1
Inst10 2.9 1.1–7.8 Inst1 0 2.6 0.9–7.1 Inst10 1.7 0.6–4.7 Inst1 0 1.6 0.5–4.6 Inst1 0 1.6 0.5–4.7
Inst11 0.8 0.2–2.6 Inst1 1 0.7 0.2–2.2 Inst11 0.4 0.1–1.3 Inst1 1 0.3 0.1–1.2 Inst1 1 0.3 0.1–1.0
Work areae Drill Fl 0.7 0.5–1.0 Drill Fl 0.5 0.3–0.9 Drill Fl 0.5 0.3–0.8 Drill Fl 0.5 0.3–0.7
Deck 0.4 0.3–0.7 Deck 0.3 0.2–0.6 Deck 0.3 0.2–0.6 Deck 0.3 0.2–0.6
Techn 0.3 0.3–0.4 Techn 0.2 0.1–0.3 Techn 0.2 0.1–0.3 Techn 0.2 0.1–0.3
Inst belongingf Nomad 0.6 0.4–1.1 Nomad 0.6 0.3–1.0 Nomad 0.6 0.3–1.0
Shift rotationsg Day/NP 0.7 0.4–1.3 Day/NP 0.7 0.4–1.2 Day/N P 0.8 0.4–1.3
Vary 0.4 0.2–0.8 Vary 0.4 0.2–0.8 Vary 0.4 0.2–0.8
Trust in wm TWM 0.7 0.5–0.9 TWM 0.7 0.5–0.9
Trust in sv TSV 0.9 0.7–1.2 TSV 1.0 0.8–1.4
Safety compl Compl 0.6 0.4–0.8
Safety part Partic 1.5 1.1–2.1
p 6 0.001, p 6 0.01, p 6 0.05.

a
R-Square (Cox and Snell): Model 1 = 0.51. Model 2 = 0.93. Model 3 = 0.110. Model 4 = 0.122. Model 5 = 0.149.
b
c
Reference category: ‘NCS’ having the highest portion of involvement in incidents.
d
Reference category: The installation with the highest portion of involvement in incidents. 12 installations in total.
e
Reference category: ‘Driller Team’ having the highest portion of involvement in incidents. Others: Drill floor, Deck and Technical team.
f
Reference category: Permanent – Employees working mostly or always on the same installation having the highest portion of involvement in incidents. Other: Nomadic.
g
Reference category: Fixed rotations – day/night every second tour having the highest portion of involvement in incidents. Others: Day/night shift permanent and Shift
vary.
to be a combination of systemic and ‘‘tailored” approaches. This Doney, P.M., Cannon, J.P., Mullen, M.R., 1998. Understanding the influence of
national culture on the development of trust. Academy of Management Review
makes it important to not only acquire local knowledge, but also
23 (3), 601–620.
to understand how systemic elements are being anchored in actual Dyer, J.H., Chu, W., 2003. The role of trustworthiness in reducing transaction costs
work practice. Another important issue makes the delicate relation and improving performance: empirical evidence from the United States, Japan
between trust, distrust and control; when does functional trust be- and Korea. Organization Science 14 (1), 57–68.
Eikemo, T.A., Høyvarde Clausen, T. (Eds.), 2007. Kvantitativ analyse med SPSS. En
come dysfunctional, or when do control elements meant to protect praktisk innføring i kvantitative analyseteknikker. (Quantitative analysis with
become destructive? In many ways this study highlights the SPSS. A practical introduction to quantitative analysis techniques.) Tapir
importance of understanding how policies and procedures dissem- akademisk forlag.
Gherardi, S., Niccolini, D., 2000. The organizational learning of safety in
inated at a company level are translated into practices by front-line communities of practice. Journal of Management Inquiry 9 (1), 7–18.
employees with due cognisance of the situational and cultural con- Giddens, A., 1991. Modernity and Self-Identity: Self and Society in the Late Modern
texts in which employees are carrying out their work tasks. Safety Age. Polity Press, Cambridge.
Goffman, E., 1961. Asylums: Essays on the Social Situation of Mental Patients and
continues to be a challenge but by using both quantitative and other Inmates. Doubleday, New York.
qualitative data to achieve a greater understanding of the interac- Gubbins, C., MacCurtain, S., 2008. Understanding the dynamics of collective
tion of relevant factors progress can made, particularly for indus- learning: the role of trust and social capital. Advances in Developing Human
Resources 10 (4), 578–599.
tries operating on a global scale. Hale, A., 2000. Editorial: cultures confusion. Safety Science 34, 1–14.
Hammersley, M., 1992. Deconstructing the qualitative-quantitative divide. In:
Brannen, J. (Ed.), Mixing Methods: Qualitative and Quantitative Research,
Acknowledgements
Gower, London, pp. 39–57.
Helmreich, R.L., Merritt, A.C., 1998. Culture at Work: National, Organizational and
We are grateful for the financial support to this project from the Professional Influences. Ashgate, Aldershot, United Kingdom.
Norwegian Research Council and Seawell AS. A special thanks to Hofstede, G., 1984. Culture’s Consequences: International Differences in Work
Related Values. Sage, Beverly Hills, CA.
the company for opening their doors and spending valuable time House, R.J., Hanges, P.J., Javidan, M., Dorfman, P.W., Gupta, V. (Eds.), 2004. Culture,
and information with us. Leadership, and Organizations. The GLOBE Study of 62 Societies. SAGE
Publications.
Hurst, N.W., Young, S., Donald, I., Gibson, H., Muyselaar, A., 1996. Measures of safety
Appendix management performance and attitudes to safety at major hazard sites. Journal
of Loss Prevention. Process Industry 9, 161–172.
Høivik, D., Tharaldsen, J.E., Baste, V., Moen, B.E., 2009. Does the location on offshore
See Tables A1–A5. installations have higher impact on safety climate than the company
belonging?. (Safety Science) 47, 1324–1331.
Jeffcott, S., Pidgeon, N., Walls, J., 2006. Risk, trust and safety culture in the UK. Train
References operating companies. Risk Analysis 26 (5), 1121–1205.
Larsson, S., 2008. Constructing safety. In: Workplace Safety in Construction Industry
Adler, P.S., 2001. Market, hierarchy and trust: the knowledge economy and the with Emphasis on Safety Climate and Empirical Base in a Large Construction
future of capitalism. Organization Science 12 (2), 215–234. Project, Doctoral Thesis, Gøteborg, Sweden.
Agresti, A., Finlay, B., 1997. Statistical methods for the social sciences, 3rd ed. Upper Lewicki, R.J., Mcallister, D.J., Bies, R.J., 1998. Trust and distrust: new
Saddle River, Prentice-Hall, NJ. relationships and realities. Academy of Management Review 23 (3), 438–
Andrews, K.M., Delahaye, B.L., 2000. Influences of knowledge process in 458.
organizational learning: the psychological filter. Journal of Management Lewicki, R.J., Tomlinson, E.C., Gillespie, N., 2006. Models of interpersonal trust:
Studies 37 (6), 797–811. theoretical approaches, empirical evidence, and future directions. Journal of
Bakker, M., Leenders, R.T.A.J., Gabbay, S.M., Van Engelen, J.M.L., 2006. Is trust really Management 32 (6), 991–1022.
social capital? Knowledge sharing in product development projects. Learning Luhmann, N., 1979. Trust and Power. Wiley, Chichester, England.
Organization 13 (6), 594–605. Madhok, A., 2006. How much does ownership really matter? Equity and trust
Barber, B., 1983. The Logics and Limits of Trust. Rutgers University Press. relations in joint venture relationships. Journal of International Business Studies
Beck, U., 1992. Risk Society. Towards a New Modernity. Sage, New Dehli. 37, 4–11.
(Translated from the German Risikogesellschaft published in 1986). Marotti, R.A., Coney, L.M., Tinetti, ME., 1997. Self report versus state records
Blaikie, N., 2000. Designing Social Research. The Logic of Anticipation. Polity Press, for identifying crashes among older drivers. Journal of Gerontology 52,
in association with Blackwell Publishers. 184–187.
Blumer, H., 1956. Sociological Analysis and the ‘‘Variable”. vol. 21, No. 6, American McAllister, D.J., 1995. Affect- and cognition-based trust as foundations for
Sociological Review, pp. 683–690. interpersonal cooperation in organizations. Academy of Management Journal
Brannen, J. (Ed.), 1992. Mixing Methods: Qualitative and Quantitative Research, 38, 24–59.
Aldershot, Avebury. Mearns, K.J., Whitaker, S., Flin, R., 2001. Benchmarking safety climate in hazardous
Brannen, J., 2004. Working qualitatively and quantitatively. In: Seale, C., Gobo, G., environments: a longitudinal, inter-organizational approach. Risk Analysis 21
Gubrium, J.F., Silverman, D. (Eds.), Qualitative Research Practice. Sage (4), 771–776.
Publications. Mearns, K., Rundmo, T., Flin, R., Gordon, R., Fleming, M., 2004. Evaluation of
Burns, C., Mearns, K., McGeorge, P., 2006. Explicit and implicit trust within safety psychosocial and organizational factors in offshore safety: a comparative study.
culture. Risk Analysis 26 (5), 1139–1150. Journal of Risk Research 7 (5), 545–561.
Child, J., Möllering, G., 2003. Confidence and active trust development in the Mearns, K., Yule, S., 2008. The role of national culture in determining safety
Chinese business environment. Organization Science 14 (1), 69–80. performance: challenges for the global oil and gas industry. Safety Science 47
Coleman, J.S., 1990. Foundations of Social Theory. Belknap Press of Harvard (6), 777–785.
University, Cambridge, MA. Menard, S., 2002. Applied Logistic Regression Analysis. SAGE University Papers.
Collinson, D.L., 1999. Surviving the rigs: safety and surveillance on north sea oil Series/Number 07/106.
installations. Organization Studies 20 (4), 579–600. Moreland, R.L., Levine, J.M., 2002. Socialization and trust in workgroups. Group
Conchie, S.M., Donald, I., 2004. Trust Climates in the Offshore Industry: the Processes and Intergroup Relation 5 (3), 185–201.
Coexistence of Trust and ‘‘Distrust”. In: Paper Presented at the BPS Social Nonaka, I., Takeuchi, H., 1995. The Knowledge Creating Company. Oxford University
Psychology Conference, September. The University of Liverpool, UK. Press, New York.
Conchie, S.M., Donald, I.J., Taylor, P.J., 2006. Trust: missing piece(s) in the safety Pidgeon, N., Walls, J., Weyman, A., Horlick-Jones, T., 2003. Perceptions of and Trust
puzzle. Risk Analysis 26 (5), 1097–1104. in the Health and Safety Executive as a Risk Regulator. Research Report 2003/
Conchie, S.M., Donald, I.J., 2006. The role of distrust in offshore safety performance. 100. HSE Books, Norweich.
Risk Analysis 26 (5), 1151–1160. Poortinga, W., Pidgeon, N.F., 2004. Trust, the asymmetric principle, and the role of
Conchie, S.M., Donald, I.J., 2008. The functions and development of safety-specific prior beliefs. Risk Analysis 24 (6), 1475–1486.
trust and distrust. Safety Science 46 (1), 92–103. Reason, J., 1997. Managing the risks of Organizational Accidents. Ashgate, England,
Conchie, S.M., Burns, C., 2008. Trust and risk communication in high-risk Aldershot.
organizations: a test of principles from social risk research. Risk Analysis 28 Rousseau, D.M., Sitkin, S.B., Burt, R.S., Camerer, C., 1998. Not so different after all: a
(1), 141–149. cross-discipline view of trust. Academy of Management Review 23 (3), 393–
Denzin, N., 1970. The Research Act in Sociology. Butterworths, London. 404.
Deutsch, M., 1958. Trust and suspicion. Journal of Conflict Resolution 2, 265–279. Seppänen, R., Blomqvist, K., Sundqvist, S., 2007. Measuring inter-organizational
Dirks, K.T., Ferrin, D.L., 2001. The role of trust in organizational settings. trust – a critical review of the empirical research in 1990–2003. Industrial
Organization Science 12 (4), 450–467. Marketing Management (36), 249–265.
Spangenberg, S., Baarts, C., Dyreborg, J., Jensen, L., Kines, P., Mikkelsen, K.L., Tharaldsen, J., Mearns K.J., Knudsen, K., 2008b. A Comparative Study of HSE-Culture
2003. Factors contributing to the differences in work related injury rates on the Norwegian and UK Continental Shelves. SPE Paper no. 111, 589,
between Danish and Swedish construction workers. Safety Science 41, 517– Presented in Nice, 2008.
530. Weick, K.E., 1987. Organizational culture as a source of high reliability. California
Sullivan, J.H., Peterson, R.B., Kameda, N., Shimada, J., 1981. The relationship between Management Review 24 (2), 112–127.
conflict resolution approaches and trust – a cross cultural study. Academy of Zaheer, S., Zaheer, A., 2006. Trust across border. Journal of International Business
Management Journal 24 (4), 803–815. Studies 37, 21–29.
Tharaldsen, J., Høydal, R., Junge, A., Westnes, P., 2007. Ny vår på feltet. Vurdering av Zohar, D., Luria, G., 2005. A multilevel model of safety climate: cross-level
den nye organisasjonen på XX. [A new spring on the field. Evaluation of the new relationships between organization and group-level climates. Journal of
organisation at XX.] IRIS Report 2007/052. Confidential. Applied Psychology 90 (4), 616–628.
Tharaldsen, J.E., Haukelid, K., 2007. The underdogs on the Norwegian continental Zohar, D., 2003. Safety climate: conceptual and measurement issues. In: Quick, J.C.,
shelf – high risk, high exposure and high satisfaction! In: Nordic Safety Tetrick, L.E. (Eds.), Handbook of Occupational Health Psychology. American
Conference, Tampere, Finland. Psychological Association, Washington, DC, pp. 96–123.
Tharaldsen, J., Olsen, E., Rundmo, T., 2008a. A longitudinal study of safety climate on Zohar, D., 1980. Safety climate in industrial organizations: theoretical and applied
the Norwegian continental shelf. Safety Science 46, 427–439. implications. Journal of Applied Psychology 65 (1), 96–102.

5.-In Safety We Trust

Uploaded by

Copyright:

Available Formats

5.-In Safety We Trust

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

5.-In Safety We Trust

Uploaded by

Copyright:

Available Formats

“In Safety We Trust”

Safety, Risk and Trust in the Offshore

Thesis submitted in fulfillment of

Faculty of Social Sciences

© 2011 Jorunn Elise Tharaldsen

In addition to providing financial funding, I would also like to thank Seawell

Stavanger, August 2010

Jorunn Elise Tharaldsen

 Jorunn Elise Tharaldsen, University of Stavanger, Faculty of

 Dordi Høivik, Bente E. Moen, Valborg Baste: University of

 Espen Olsen, University of Stavanger, Faculty of Social Sciences,

 Kathryn Mearns, University of Aberdeen, School of Psychology,

 Knud Knudsen, University of Stavanger, Faculty of Social

 Knut Haukelid, University of Oslo, Centre for Technology,

 Torbjørn Rundmo, Norwegian University of Science and

The thesis draws upon theory from different disciplines – mainly

A model envisaging dynamics between functional and dysfunctional trust

The epistemological roots of cultural and behavioral perspectives are

The industrial culture seems to be geared towards unity, shared norms,

Scientific Environments – Co-authors................................................ v

Summary .............................................................................................. vi

Part I .................................................................................................... 15

1 Introduction ............................................................................ 15

2 Industrial context ................................................................... 21

3 Theoretical framing and conceptual clarifications ............. 29

4 Methodology ........................................................................... 45

5 Results ..................................................................................... 61

6 Discussion ................................................................................ 67

7 Limitations and implications ................................................. 81

8 Conclusions ............................................................................. 93

References ........................................................................................... 95

List of Articles................................................................................... 111

Article I.............................................................................................. 113

Article II ............................................................................................ 115

Article III .......................................................................................... 117

Article IV ........................................................................................... 119

That assembly of characteristics and attitudes in organizations and

1.1 The role of trust with regards to safety

1.2 Main research problems and research

 Establishing valid and reliable safety climate dimensions:

 What theoretical concepts best explain safety performance and

 How do these concepts relate to each other and to accident

 What makes a sound indicator of safety performance?

 The distributed character of Safety climate:

 What underlying dynamics contribute to explanations of

 How are safety climate, risk perception and accident exposure

 What factors could explain such a distribution and what

 Culture and behavioral perspectives on safety:

 What are main roots of cultural and behavioral perspectives on

 How do these approaches differ and/or how may they be said

 Where do differing approaches lead us and what kind of

 The links between trust, distrust and safety:

 What are the functions or dysfunctions of trust and distrust

 What kind of impact may trust at different levels have on

 Cross cultural comparison of offshore safety:

 Do safety climate perceptions differ across the UK and

 How may cultural differences influence trust relations,

1.3 Aims of the articles

1. A longitudinal study of safety climate on the Norwegian Continental