Associations Between Risk Perception and Safety

Pergamon Safety Science Vol. 24, No. 3, pp. 197-209, 1996

0 1997 Else&r Science Ltd. All rights reserved Printed in the Netherlands

0925-7535/97 $17.00 + 0.00

PII: SO9257535(97)00038-6

ASSOCIATIONS AND SAFETY

Torbjo rn Rundmo

BETWEEN RISK PERCEPTION

Department of Psychology, Norwegian University of Science and Technology, 7055 Trondheim, Norway

Abstract-The relationship between perception of risk and involvement in accidents is receiving

increasing attention in the offshore oil industry. A self-completion questionnaire survey was

carried out among employees on twelve offshore oil installations in 1994. ’ The number of

respondents were 1138. Employee evaluations of the status of safety and contingency measures

were affected by physical working conditions, attitudes towards safety and accident prevention

work as well as management commitment and involvement in safety promotion. These factors

were also related to job stress, perceived risk and risk behaviour. There was a significant

positive correlation between perceived risk and risk behaviour, but risk perception was not

found to predict risk behaviour. Risk behaviour affected accidents as well as near-misses. The

possibility that safety cannot be improved by changing individual risk perception is discussed.

0 1997 Elsevier Science Ltd.

1. Associations between risk perception and safety

In Norway safety is given extensive priority in the offshore oil industry through The Petroleum Act of 22 March 1985, The Working Environment Act of 4 February 1977, especially section 1.3 which contains regulations for offshore activities on the Norwegian Continental Shelf, Regulations Concerning Internal Control in the Offshore Oil Industry of 28

June 1985, Regulations Concerning Work Protection and the Work Environment in the

OfSshore Oil Industry of 27 November 1992, Regulations Concerning Systematic Reuision of

the Working Enuironment in The Offshore Industry of 8 March 1995, Regulations Concerning Safety in Offshore Production of 22 June 198.5, Regulations Concerning Implementation and

Use qf Risk Analyses in the Petroleum Activities of 12 June 1990, and several other acts and regulations. The presence of potentially-hazardous risk sources which relate to the risk of ordinary occupational accidents as well as catastrophes and major accidents points to the importance of such regulations.

The majority of the efforts to improve safety in offshore oil production have, naturally enough, focused on measures aimed at reducing and removing ‘objective’ risks. However

’ The study was financed by the Norwegian Petroleum Directorate (NPD). NPD’s constant

interest and support for my research has been a very important motivation factor in my work

with these issues.

197

198 T. Rundmok

employees’ perception of risk as well as their subjective assessments of the working

conditions and work environment can be important for the personnel’s behaviour with regard to risk and hence these factors may also influence ‘objective’ risk or safety. In addition,

perception of risk and dissatisfaction with the working conditions can reduce the personnel’s

well-being and health. A study on risk perception and safety on offshore installations was carried out among

employees on oil platforms in the Norwegian part of the North Sea in 1990 and repeated in 1994. The core aims of the study were to determine employee risk perception and to analyze

the relation between organizational factors, perception of risk and safety. The results of the 1990 study have been published previously (see e.g. Rundmo, 1990a,b,c, 1992a,b, 1994a,b, 1995). This paper focuses on the associations between risk perception, employee risk

behaviour and ‘objective’ risk. How employees perceive the risk they are exposed to during the conduct of their work may

contribute to an understanding of risk management and thereby to the safety of their working

conditions (see Advisory Committee on the Safety of Nuclear Installations, 1993). Biased perception of risk can cause misjudgments of potentially-hazardous risk sources. When the risks are misjudged, this may cause risk behaviour and inappropriate action towards the risk source, inappropriate decisions with regard to safety measures and ordinary occupational

accidents as well as catastrophes. Therefore, if the ‘subjective’ evaluation of potential sources of danger does not correspond to the ‘objective risk’, i.e. “the risk which exists whether we are aware of it or not and regardless of whether we are concerned about it” (Risk Research

Committee, 1980, p. 12>, subjective judgments should be corrected, “since handling risk in an emergency requires a realistic assessment of the situation” (Marek et al., 1985, p. 143).

However, as pointed out by Brehmer (1994, p. 82) ‘objective’ measures are “not more objective than any other risk measure”. This is due to the generic complexities which are present in risk calculation, which include uncertainty about how to define the problem,

difficulties in assessing the facts, and difficulties in assessing the values (see Fischhoff et al., 1981, pp. 9-27).

The probability of the occurrence of a hazardous event is often seen as an objective risk measure. In offshore oil production, the accident frequency differs between personnel groups,

due to employment conditions and the type of work tasks an employee conducts (The Norwegian Petroleum Directorate, 1988-1992, Rundmo, 1994b). Therefore, if the purpose is

to reduce the number of accidents, an estimation of the probability of the occurrence of a hazardous event may be an efficient tool in accident prevention work. However, preventing the occurrence of certain consequences may be even more important.

In order to find measures aimed at avoiding catastrophes caused by blow-out, explosion and fire are given priority in the offshore oil industry irrespective of the rather low probability for such events. Therefore, to include ‘consequences’ in a risk estimate may lead to different priorities compared to an estimate based solely on the probability of an accident. Furthermore, risk estimation methods may give different weight to the consequences and hence there is a chance that the same risk source may be judged differently depending on the type of estimation method which is applied.

1.1. Risk perception, objectice risk and the role of risk analysis in safety control

Rasmussen (1994) differentiates between empirical, evolutionary and analytical safety control: Small scale accidents, e.g. traffic and ordinary work accidents, can be controlled through analysis of past accidents. This is empirical sqfety control. It is not possible to carry

Associations between risk perception and safety 199

out empirical safety control when accidents are infrequent. Medium size, infrequent accidents,

such as aircraft crashes and train collisions, may be prevented by euolutionan, safety control,

which is “improvements in response to analysis of the individual, latest major accidents” (Rasmussen, 1994, p. 4). When we assess very low probabilities we need to model the risk. This type of assessment has to rely on predictive analysis of possible accidents. This is

analytical safety control, which has to be applied when analysing e.g. major nuclear and chemical hazards. In the offshore oil industry all three types of control are applied.

In accordance with sections 10 to 16 of the Regulations Concerning Implementation and

Use qf Risk Analysis in the Petroleum Activities the operator companies are charged to carry

out extensive safety control and define safety objectives, to define ‘acceptance criteria’ for risk, i.e. criteria which express an acceptable level of risk in the activities concerned, to plan,

implement and carry out risk analyses, to assess identified risks with regard to acceptance criteria as well as to implement risk reducing measures. The term risk analysis was given a broad definition, covering quantitative (empirical) as well as qualitative (evolutionary and

analytical) safety control methods. The risk analysis activities in the offshore oil industry have hitherto largely been dominated

by ‘experts’ such as engineers and statisticians who like to think that risk can be given an

operational definition and measured in the same way as we measure e.g. ‘length’ and that simple rules can decide when measures should be implemented. However the experts’ judgements of the same risk source may differ and expert judgements are difficult to justify (Brehmer, 1994). Expert judgments of risk also clearly differ from lay people’s judgments (see e.g. Shrader-Frechette, 1994). Furthermore, a large proportion of the population lack confi- dence in information about risk given by experts, especially the information given by public

authorities (Drottz-Sjijberg and Persson, 1993). If ‘objective’ risk calculations can reduce perceived risk and cause personnel to feel safer, it may be useful to continue to carry out such

calculations. However, if the purpose is to enhance safety, it may be equally important to determine employee subjective assessment of risk.

Section three of the Risk Analysis Regulation also emphasizes the complexity of risk

analysis, including technical, human as well as organizational factors. In particular more basic knowledge of the effect on safety of the interaction between these factors is needed. The generic problems of ‘objective’ risk calculation point however to the difficulties of relating organizational and human factors to the apparently precise, but rather abstract calculations of

risk given by engineers and statisticians. The generic problems related to such calculations emerge when the calculation is not

directly linked to the respondents’ own experience of near-misses and injuries, i.e. whether or not the respondents themselves have experienced near-misses or an accident/injury. The judgement of nuclear and other ‘epistemological’ risks are examples of risks which cannot be linked to personal experience. However, for an employee who has suffered an injury/ordinary occupational accident this experience may be ‘real’ enough. It is mainly a semantic problem whether an employee’s own experience of near-misses and occupational accidents is termed ‘objective’ or seen as a part of the person’s ‘subjective world’. However, as with ‘objective’ risk calculations, measurements of risk perception have limited value when they are handled apart from the situation in which the employee experiences and deals with risk.

1.2. The relation between perception, behauiour and accidents

The association between subjective assessment of risk, risk behaviour and injury experience may exert an influence on safety. Risk behaviour includes the extent to which the personnel

200 T. Rundmok

ignore safety regulations in order to get a job done, carry out activities which are forbidden,

perform their work duties correctly, use personal protective equipment, and break procedures to carry out jobs quickly. However, the associations between risk perception, risk behaviour and injury experience are complex. There are at least three different approaches to these associations:

(1) Accidents may cause risk perception: An employee who has experienced an accident

may assess the risk differently after the accident than before. However, the probability of an accident is not greater after such an event than before. Therefore, there is no good reason why

people should become less safe after they themselves have experienced an accident than they were before. The probability of an accident is not enhanced. If the employees’ assessments of risk corresponds to the reality of the situation they should not feel less safe after they

themselves suffered an accident than they did before. Rundmo (1995) found that offshore oil personnel who had suffered an accident themselves,

felt less safe than those who had not experienced an accident. Personnel on high-injury platforms, i.e. platforms where accidents took place frequently, felt less safe than personnel on low-injury platforms, i.e. platforms where accidents happened less often. The difference between low-injury and high-injury platforms was more marked than the difference due to the respondents’ own injury experience. Another interesting result was that there was no signifi- cant difference between risk perception among injured personnel on low-injury and high-in-

jury platforms. In the group of non-injured employees risk perception corresponded well to the ‘real’ risk. However, the results indicated that there was no major effect on risk perception from accidents. Thus, it is more probable that risk perception is a causal factor in accidents

than the other way around. (2) Risk perception may cause accidents: When an employee feels unsafe, this may cause

workload and strain, which enhances the probability of accidents.

The 1990 study of risk perception and safety in the Norwegian part of the North Sea showed that the personnel who experienced most job stress and felt most unsafe, were also more often subject to accidents and near-misses (Rundmo, 1992a,b, 1994a,b). Organizational and social factors explained why the personnel felt at risk and experienced job stress. The

respondents who experienced stress to the greatest extent and also felt most at risk, were most dissatisfied with the status of safety and contingency measures and with management and employee commitment and involvement in safety work. Their attitudes towards safety and accident prevention were also less ideal. In addition, they experienced a greater physical workload. Thus, the same set of organizational, social and physical predictor variables seemed to affect risk perception as well as risk behaviour, near-misses and accidents. Hence, there is

also a correlation between perceived risk and accidents/near-misses. However, these findings do not imply that risk perception is a causal factor in accidents. Because the same set of organizational and social predictor variables can explain a significant proportion of the variance in both these variables, there is a correlation between them. A study carried out on the UK continental Shelf in 1994 came to identical conclusions (Flin et al., 1994, 1995).

(3) Risk perception and accidents/safety are both endogenous L;ariables: Risk perception and accidents can both be effect variables which are independent of each other. However, as I have shown in studies carried out previously (Rundmo, 1992a,b) personnel who felt unsafe also more often experienced an accident than those who felt safe. Therefore, there has to be a connection between these two variables.

Perception of risk is relevant to safety because it may affect the employees’ behaviour and behaviour can exert an influence on the probability of accidents. Therefore, the probability of

Associutions between risk perception and safe0 201

accidents may be influenced by risk perception. When employees feel unsafe this may cause

workload and strain, which cause them to take chances. However, risk perception is not necessarily a significant predictor of risk behaviour. Perception and behaviour may as well be independent variables.

Measures of strain and risk behaviour were not included in the 1990 study. However, they were in the 1994 study. Consequently, the specific aims of this study are as follows: 1. To determine whether risk perception affects risk behaviour or if risk perception and risk

behaviour are independent effect variables, i.e. variables which both may be affected by organizational and social predictor variables such as commitment and involvement in safety work, attitudes towards safety and accident prevention and the status of safety and

contingency measures. 2. To determine whether or not risk behaviour influences near-misses and accidents.

2. Methods

2. I. Sample

A self-completion questionnaire survey was carried out among employees on oil platforms in the Norwegian part of the North Sea in February 1994. The study included 1138 respondents on 12 platforms. The response rate was 87 percent. The response rate makes the

results highly representative.

2.2. Content of the questionnaire

The questionnaire contained a total of approximately 250 questions divided into the following information: The respondent’s age and gender, professional experience, type of job

on board, subjective assessment of risk, determination of job-stress, physical working condi- tions, experience of accidents and near-accidents, satisfaction/dissatisfaction with safety and contingency measures, attitudes towards safety, social support from management, supervisors and colleagues, and employee and management commitment and involvement in safety work. The questionnaire also contained information about job-related strain and risk behaviour.

With regard to risk perception the respondents were asked to rate to which extent they felt

safe/unsafe on a five-point evaluation scale ranging from ‘very safe’ to ‘very unsafe’. On the job stress items the respondents were asked to rate to which extent they experienced the problem, also on a five-point rating scale ranging from “Yes - to a high extent” to “No - not at all”. The test items intended to measure risk perception and job stress are shown in

Table 1. The five-point rating scale used to measure satisfaction/dissatisfaction with the safety and

contingency measures ranged from “very satisfied” to “very dissatisfied”. The attitude scale

ranged from “agree strongly” to “disagree strongly”. The test items applied for measuring satisfaction with safety and contingency measures as well as safety attitude are shown in Table

Commitment and involvement in safety work included commitment from platform manage- ment, immediate supervisors trade unions and fellow workers. The respondents were asked to rate to which extent they believed that these groups are concerned with their safety and in participating in accident prevention work. Also for this factor a five-point rating scale was

202 T. Rundmok

Table I List of test items - risk perception potentially-hazardous circumstances and job stress

Area Test item

Risk perception potential explosion

hazardous risk sources tire

blow-out

toxic gas leak (e.g. H,S)

structural failure

electric shock

contact with cold/hot surfaces, pipes etc.

crushing by machinery or machine parts

slipping

falling overboard

falling to a lower level

falling object

weather and wind conditions

helicopter crashing on the platform

vessel hitting platform

sabotage act

Job stress I can do my work independently and according to my own views

I have sufficient amount of freedom to decide on my own pace of work

1 can decide when and how each individual work task shall be implemented

I can take short breaks whenever I wish without having to take account of other people

I know exactly the expectations of me held by others

my immediate superiors ask for my advice before making their decisions

I know what I can expect from others

I have the opportunity to influence the decisions to be made by my superiors

my superiors issue differing and contradicting orders

1 am satisfied with the way I am kept informed of what takes place on the platform

applied. The scale ranged from “very concerned” to “not concerned at all” in improving

safety and involvement in accident prevention work. I measured perception of instrumental support, i.e. aid in the form of changes of

environment; emotional support, i.e. feeling of trust and concern; and informal support, i.e. advice, suggestions and information. The respondents were asked the following questions: “How much support do the following people provide you with to carry out your work?“, “To what extent can you talk with the following people?“, and “How much do you feel you can

trust the following people when things get difficult at your place of work?“. They were asked to give separate ratings for supervisors, fellow workers and trade union representatives in reply to each of the three questions. The five-point rating scale which was used ranged from “very much support” to “very little support”.

Physical workload included test items which measured heavy physical workload; bad design of work place; noise; vibration; draughts/cold; polluted air/poor ventilation/hot work climate; and vapours/acids/inflammable or health-hazardous chemicals. Strain included sleeping disorders and stomach trouble caused by stress, lack of privacy, poor insulation between berths, noise from helicopters, changes in shifts and other reasons.

Associations between risk perception and safe9 203

Table 2

List of test items - status of safety and contingency measures and job stress

Area Test item

Status of safety and control and inspection routines

contingency measures safety instructions/training

follow-up measures taken after accidents have taken place

housekeeping at the work place

protective and safety devices on machines and equipment

escape routes on platform

evacuation devices

first aid training

medical services

marking and sign-posting

availability of personal safety equipment

reliability of alarm systems

fire and gas detection systems

deluge system

permit to work system

temporary refuge

safety officer

emergency response training

Safety attitudes “sometimes it is necessary to depart from safety requirements for the sake of production”

“good operational economy is often in conflict with measures to improve personal safety”

“rules and instructions relating to personal safety sometimes make it difficult to keep up

with production targets”

“sometimes it is necessary to take risks to get a job done”

“whenever 1 see safety instructions not being complied with, I call attention to this on

the spot”

“many minor injuries and minor accidents are an indication that serious accidents can

also easily occur”

“safety measures only shift the danger from one area to another”

“occupational accidents are often the result of bad planning and poor management”

“calling attention to breaches of safety can easily be felt as unnecessary hassle”

“good proposals on how to improve safety are often dropped if they cost too much”

“many accidents just happen, there is little one can do to avoid them”

2.3. Stutistical procedures

In this study the main statistical analysis method was path analysis. The LISREL (Analysis

of Linear Structural Relationships) program (JBreskog and %rbom, 1979, 1989) was used to

test direct and indirect effects on risk behaviour from organizational and social factors presented above. One can compare the magnitude of the direct and indirect effects. The higher the path coefficient, the stronger is the effect that a certain variable has on another variable. Path coefficients vary from - 1 to + 1. They are analogous to standardized partial regression coefficients. The standard regression coefficient expresses the change, measured in relation to standard deviation, p, in the expected value of the effect variables, y, when the cause variables, X, are changed according to their standard deviation, s,~. Error terms in the path model are termed e.

Before being entered into the model, sum scores were made for all the main areas which

204 T. Rundmok

were included in the models, one index for each area. Cronbach’s (Y was used to test the

reliability of the indices, which all were found to be satisfactory. The LISREL analysis program was also used to test a MIMIC (Multiple Indicators and

Multiple Causes)-model of the effects of risk behaviour items on near-misses and accidents. A MIMIC-model is a model where a single unobserved latent variable is defined as a criterion which is influenced by several directly observed x-variables. 5 = x (Jbreskog and Siirbom,

1989). A MIMIC-model consists thus of one structural model and one causal model. The

measurement model specifies how the latent or hypothetical (endogenous) q-variable (‘objec- tive’ risk) depends upon the observed T-variables (near-misses and accidents). Correlations

between each of the indicators and the theoretical concept. the h-values, is interpreted in the same manner as ordinary factor loadings. E, designates measurement error in the A,-indica-

tors, i.e. unexplained variances which cannot be attributed to the latent variable. The y-values express the effects of the x-variables on the v-variable. R* indicates the proportion of

variance in the effect variable that is explained by the x-variables. The error term of the variation in the q-variable, which cannot be attributed to other variables in the model, is designated (.

The LISREL models can only be used to search for causal relationships indirectly by eliminating or modifying poor models that give predictions which are inconsistent with the data, i.e. the model indicates causality. How well the path model as well as the MIMIC model

fitted the data was tested by the X2-test and the Goodness of Fit test. Both the models presented in this study satisfied fairly well the demand of non-significance, p > 0.05.

Pearson’s r correlation was used as an auxiliary method to analyze the associations between risk perception and risk behaviour.

3. Results

3. I. Associations between risk perception and risk behaciour

Table 3 shows that there are significant correlations ( p < 0.001) between risk perception caused by or related to ordinary occupational accidents and all the six indicators of risk

Table 3

Correlations (Pearson’s r) between risk perception and risk behaviour

Risk behaviour

Ignore Forbid. Duties Take Prot. Break Total regul. activ. incorr. chance equipm. procedure

Ordinary occ. accid. I = 0.26 r = 0.22 r = 0.24 r = 0.27 i-=0.12 r = 0.23 r = 0.31

n= 1001 n= 1002 n = 988 n= 1001 n = 992 n = 1001 n=981 p = 0.000 p=o.ooo p= 0.000 p=o.oOo p= 0.000 p= 0.000 p= 0.000

Catastrophes major accid. r = 0.05 Y = 0.04 i-=0.13 r = 0.09 r=0.04 i- = 0.07 r = 0.09 n = 1001 n=1002 n = 989 n = 1001 )I = 992 n = 1001 n = 982 p=O.lll p=O.226 p=O.OOO p=O.O04 p=O.207 p=O.OOO p=O.O02

Post accid. measures r-=0.16 r = 0.11 r=0.19 r=0.15 r = 0.08 r=O.l6 i- = 0.20 n = 1002 n=lOo3 n = 989 n=loO2 n = 993 n=1002 n = 982 p = 0.000 p = 0.00 I p = 0.000 p = 0.000 p=O.Oll p=o.OOo p=o.OOo

Associations between risk perception and safe0 205

behaviour. The correlation between the index of risk behaviour made up of all the six

indicators and risk perception related to ordinary occupational accidents was also significant.

The correlation between risk perception related to catastrophes and major accidents was not significant, with the exception of the test items “I cannot always carry out my work duties

correctly”, p < 0.001, and “I break procedures to carry out jobs quickly”, p = 0.028. There were also significant correlations between risk perception with respect to post-accident

measures and all the test items measuring risk behaviour (see Table 3). The correlations were positive, showing that the less safe the employees felt, the more risky was their behaviour. Thus, these results show that there is an association between risk perception and risk behaviour. However,

behaviour.

these results do not show whether or not risk perception affects

3.2. Factors affecting risk behaciour

Figure 1 is a path-model which shows how organizational and social factors, as well as workload affect risk behaviour. The LISREL-program was used for this analysis. The diagram

applies to the 1994 data. This is because risk behaviour was included only in the 1994 questionnaire. The model fit was acceptable, x2(2/1020) = 3.44, p = 0.170. The figure

shows the following: (I) The effects on status for safety and contingency factors, job stress and perceived risk of

management and employee commitment and involvement in safety work, attitudes towards safety and accident prevention on the platform, and the physical working environment.

(2) The effects on strain and risk behaviour of safety and contingency measures, job stress, and perceived risk.

(3) The direct effects on strain and risk behaviour of the predictor variables, which are

y42= IO \

Xl Management

y41-.08

and Employee \ y, I=.41 \

Risk ex.78 Behaviour

f

y43=.17 / WOr

Conditions fl33.17 x 2 (2/1020)=3.44. p..170

Goodness of Fit index (GFI)=.999

Adjusted GFI=.985

Fig. 1. Associations between organizational and physical working conditions, safety and contingency measures.

stress, risk perception, strain and risk behaviour.

job

206 T. Rundmok

commitment and involvement in safety work, safety attitudes, and physical working condi-

tions. As can be seen from the model:

(1) The three exogenous variables, management and employee commitment and involve- ment in safety work, safety attitudes and physical working conditions contributed significantly to the respondents’ satisfaction/dissatisfaction with safety and contingency measures. The error term (e) was 0.78 which means that the squared regression coefficient was 0.22. In other words, these three predictor variables accounted for 22 percent of the variance in

satisfaction/dissatisfaction, i.e. R* X 100%. (2) Management and employee commitment was definitely the most important predictor of

satisfaction/dissatisfaction with the safety and contingency measures, y,, = 0.4 1.

(3) Status of safety work/safety and contingency measures had a direct effect on perceived risk, &, = 0.36, and,

(4) Also indirectly affected perceived risk through the effect on job stress. (5) Physical working conditions had a strong effect on job stress, yz3 = -0.26 as well as

perceived risk, yj3 = -0.28. Taken together these variables accounted for 33 percent of the variance in perceived risk.

El=.88 e2=.51

/ ? I

All=.18 /X21=12/ A31=.“2/ h41=.10\ hM\ M1=.0;\

~2 (5/1020)=10.72, ~~057

Goodness of R Index (GFI)=.997

Adjusted GFk.997

Fig. 2. Associations between risk behaviour and ‘objective’ risk (near-misses and accidents).

A.ssociations betieen risk perception and safety 201

(6) Furthermore, perceived risk was the variable which had the most significant effect on

strain, p,j = 0.11, and strain affected risk behaviour, & = 0.10. As can be seen, there was a small and insignificant effect of risk perception directly on behaviour, & = -0.02. The effect operates through strain.

3.3. Risk behauiour and objectice risk

Risk behaviour also affected objective risk, as shown in Fig. 2. These six predictors variables are the test items of risk behaviour. The two most important predictors of accidents

and near accidents were when an employee tended to:

- ignore safety regulations to get a job done, y, , = 0.18; and - carry out forbidden activities, y?, = 0.12.

There are also strong correlations between these two factors and the other predictor

variables. The six predictors accounted for less than 20 percent of the variance in objective risk,

R* = 0.16, which means that there may also be a significant direct effect of organizational and

physical factors on objective risk.

4. Discussion

Several risk perception studies have been carried out because it was believed that risk perception may affect behaviour with respect to potentially-hazardous risk sources and, consequently, also safety. This study has shown that risk perception and risk behaviour are

significantly correlated. However, these two variables were both found to be effect or criterion variables, which are relatively independent of each other.

Risk perception was not found to predict risk behaviour. Thus, safety cannot be improved by changing individual risk perception. The association between risk perception and behaviour is caused by the fact that the same predictor variables affect both these factors. Therefore, it is the factors which cause variations in risk perception as well as risk behaviour and safety which

should be the focus of these efforts.

The results also indicate that employee risk perception and other subjective assessments may be good indicators of the safety level. Perhaps one should start looking at the possibility of changing ‘objective’ risk estimates when they do not correspond well to employee risk perception instead of the other way around. Consequently, subjective assessments should be made part of every safety control system and included as an additional requirement in the regulations concerning risk analysis in the petroleum activities. Further development of proper instruments of subjective assessments of risk as well as of other aspects of the working

environment is therefore needed. Management and employee commitment is very important for the status of the safety work

and the safety work affects perceived risk directly and via job stress. Perceived risk exerts an influence on risk behaviour to some extent. Perceived risk seems to affect behaviour via the effect on strain. Physical working conditions also affect strain. However, the direct effects on risk behaviour of the organizational and physical working conditions are stronger than the effects arising from job stress and risk perception. These associations also indicate which factors one should concentrate on in accident prevention work.

The results did not give any support to the assumption that biased risk perception is a major

208 T. Rundmok

safety problem. On the contrary, the results indicate that employee behaviour to a great extent

is constrained by the working conditions under which employees work. When these condi- tions are not judged to be at a satisfactory level, the employees know that the risk of suffering

an accident is enhanced. Consequently, they feel unsafe. Therefore, there is also a significant correlation between risk perception and behaviour. However, these perceptions are not causal factors in occupational accidents.

When an employee feels at risk he or she also is at risk. Bad working conditions may

cause job stress as well as insecurity. Furthermore, these factors result in strain, which may reduce his or her capability to deal with emergency situations correctly. Consequently, the risk of accidents are increased not because of a biased perception of risk, but rather because risk is perceived ‘correctly’, i.e. is in accordance with objective risk.

References

Advisory Committee on the Safety of Nuclear Installations, Humman Factors Study Group, 1993. Third Report. Organising for Safety. HMSO Publication Centre, London

Brehmer. B., 1994. Some notes on psychological research related to risk. In Future Risks and Risk Management, ed.

B. Brehmer and N.E. Sahlin, pp. 71-91. Kluwer Academic Publishers, Dordrecht.

Drottz-Sjoberg, B.M. and Persson, L., 1993. Public reaction to radiation: fear, anxiety, or phobia. Health Physics 64, 223-23 1.

Fischhoff, B., Lichtenstein, S., Slavic, P., Derby, S.L. and Keeney, R.L., 1981. Acceptable Risk. Cambridge

University Press, New York.

Flin, R., Mearns, K., Fleming, M. and Gordon, R., 1994. Risk Percepfion in UK Offshore Workers. Final Report for

the Oil Industry/Offshore Safety Division, HSE (OTH 94454). HSE Books, London.

Flin, R., Meams, K., Fleming, M. and Gordon, R., 1995. Risk Percepfion in UK Offshore Workers. Paper presented at

The Understand Risk Perception Conference. Robert Gordon University, 2 February 1995.

Joreskog, K.G. and Sorbom, D., 1979. Advances in Factor Analysis and Structural Equation Models. Abt. Books,

Cambridge, MA.

Joreskog, K.G. and Siirbom, D., 1989. LISREL 7. Users Reference Guide. Scientific Software Inc., Moresville.

Marek, J., Tangenes, B. and Hellesoy, O.H., 1985. Experience of risk and safety. In Work Enrironment Sratiord Field. Health and Safe@ on a North Sea Oil Platjonn, ed. O.H. Hellesoy, pp. 142-174. Norwegian University

Press. Oslo.

Rasmussen, J., 1994. Risk management, adaptation, and design for safety. In Future Risks and Risk Managemenr, B.

Brehmer and N.E. Sahlin, pp. l-36. Kluwer Academic Publishers, Dordrecht.

Risk Research Committee, 1980. Accidents in Norway: How Do We Perceive and Handle Risk? The Norwegian

Research Council, Oslo.

Rundmo, T., 1990a. Opplevd risiko og sikkerhet. Teknisk Dokumentasjonsrapport (Risk perception and safety.

Technical Report). (Report, NPD-order No. 533Al.: The Norwegian Petroleum Directorate (NPD), Stavanger. (in

Norwegian) Rundmo, T., 1990b. Opplevd risiko og sikkerhet. Resultatrapport (Risk perception and safety. Main Report). (Report,

NPD-order No. 533Bl.: The Norwegian Petroleum Directorate. Stavanger. (in Norwegian)

Rundmo, T., 1990~. Opplevd risiko og sikkerhet. Sammendragsrapport (Risk perception and safety. Summary

Report). (Report, NPD-order No. 533Cl.: The Norwegian Petroleum Directorate, Stavanger. (in Norwegian)

Rundmo, T., 1992a. Risk perception and safety on offshore petroleum platforms - Part I: Perception of risk. Safety Science 15, 39-52.

Rundmo, T., 1992b. Risk perception and safety on offshore petroleum platforms - Part II: Perceived risk, job stress

and accidents. Safety Science 15, 53-68.

Rundmo, T., 1994a. Associations between organizational factors and safety and contingency measures on offshore

petroleum platforms. Scandinarian Journal of Work, Emironmenr and Health 20, 122-127.

Rundmo, T.. 1994b. Occupational accidents and objective risk on North Sea offshore installations. Sqfefy Science 17,

103-116.

Associutions between risk perception and safety 209

Rundmo, T., 1995. Perceived risk, safety status and job stress among injured and non-injured employees on offshore

petroleum installations. Journal of Safety Research 28, 87-97. Shrader-Frechette, K.S., 1994. Lay risk evaluation and the reform of risk management. In Future Risks and Risk

Management, ed. B. Brehmer and N.E. Sahlin, pp. 183-251. Kluwer Academic Publishers, Dordrecht.

The Norwegian Petroleum Directorate, 1988-1992. Annual Reports. The Norwegian Petroleum Directorate, Sta-

vanger.