Distal factors in risk perception

Journal of Risk Research 6 (3), 187–211 (2003)111123456789101111234567892011112345678930111123456789401111234567181111Distal factors in risk perceptionLENNART SJÖBERGCenter for Risk Research, Stockholm School of Economics, SwedenAbstractThis is an empirical and quantitative study of the validity of four kinds of distal explanatoryfactors in risk perception. In an initial study, personality constructs (Five FactorModel, Myers–Briggs Indicator of Jungian constructs and risk attitudes) were relatedto risk perception data (26 hazards). A relationship was found between emotionalstability and risk perception, but none with Jungian constructs. One risk attitude dimension,‘Macho’ risk willingness, was (negatively) related to demand for governmentalrisk mitigation. In a second study with a different sample, indices were constructed tomeasure the four World Views according to Cultural Theory (CT) as well as Group/Griddimensions, New Age beliefs and the New Environmental Paradigm (NEP) dimensionsof Dunlap et al. Risk perception data were obtained with regard to 37 hazards, bothgeneral and personal risk. The respondents were a large representative sample of theSwedish population. Only about 5% of the variance of perceived risk was accountedfor by Cultural Theory dimensions, considerably more by New Age beliefs and one ofthe NEP scales (eco-crisis). In a third study, data from the five Nordic countries wereused to analyse the relationships between CT dimensions and risk perception. Onlyweak relations were found. The results are discussed in relation to other current workon models of risk perception and the question of what should be considered ‘strong’evidence for a theory.1. IntroductionRisk perception is focused in many current social science investigations and applications(Sjöberg, 1999b; Sjöberg, 2001b). The need for understanding what factors areimportant in accounting for risk perception is obvious (Sjöberg, 2000b). The best knownattempt to model risk perception is the Psychometric Model (Fischhoff et al., 1978);see Sjöberg (2002a) for a discussion. The Psychometric Model attempts to track riskperception to characteristics of hazards, such as voluntariness and new versus old risk.The Psychometric Model is an example of an approach that uses explanatory variableswhich are semantically close to the risk dimensions which it tries to explain. Themodel uses aspects or characteristics of the hazards to account for its perceived levelor risk, and for risk acceptability. It is a general and powerful principle that variableswhich are close in semantic content tend to correlate more than those which are distantin content (Fishbein and Ajzen, 1975; Sjöberg, 1980). Obtaining high correlationsbetween semantically distant measures is more difficult than between those which aresemantically close, but often also more interesting (Slovic and Peters, 1998).Journal of Risk ResearchISSN 1366-9877 print ISSN 1466-4461 online © 2003 Taylor & Francis Ltdhttp://www.tandf.co.uk/journalsDOI: 10.1080/1366987032000088847

188 Sjöberg11112345678910111123456789201111234567893011112345678940111123456711181111Distal explanations of risk perception have, so far, been scarce. In the present paper,two attempts at distal explanation are presented: Cultural Theory (Douglas andWildavsky, 1982) and some personality constructs.Cultural Theory is regarded by some researchers as an important explanatory schemeand very useful for understanding risk perception. Whether this is true depends, to alarge extent, on its power in accounting for empirical data. That power has been questionedin previous theoretical (Boholm, 1996) and empirical (Sjöberg, 1997; Sjöberg,1998a) work. However, the critique has been challenged (Slovic and Peters, 1998; Tanseyand O’Riordan, 1999; Tansey, in press), on a mixture of empirical, theoretical andmethodological grounds. In the present paper, these arguments are dealt with first bypresenting new data bearing on the issues, and finally by discussing the more theoreticaland methodological aspects.Personality is a very important construct (or set of constructs) in the psychologicalstudy of individual differences. There have been a few attempts at relating personalityto risk perception, see e.g. Källmén (2000). Attempts at relating perceived risk toJungian personality concepts have been published (Fritzsche, 1995; Fritzsche, 1996), butno data were presented. The present study does present results on the issue, based onmeasurement of the Jungian typology concepts (Jung, 1971).2. Study 1: Personality and risk perceptionSince the field of risk perception arose in connection with decision and policy studiesit is perhaps understandable that relatively few attempts have been made to relate riskperception to personality, which is otherwise a prime concern of many psychologists.An attempt to find a link to personality is well motivated by the fact that current modelsof risk perception provide only a very incomplete explanation of it (Sjöberg, 2002a).Risk perception is also a factor of interest in accident prone behaviour, which has beenrelated to personality, albeit success has not been great (Lawton and Parker, 1998).One starting point for the present study was work by Fritzsche (Fritzsche, 1995, 1996).He suggested that perceived risk be construed within the framework of Jungiantheory and gave an interesting, though speculative, account of such an approach,based on the concept of archetypes. He presented no data bearing on the issue. TheMyers–Briggs test (Boyle, 1995) is the most common procedure for operationalizingJung’s personality system, and it will be used here to test Fritzsche’s suggestionsthat risk perception is explained by Jungian personality dynamics. A Swedish version ofthe Myers–Briggs scale was available (Mårdberg et al., 1994) and is used here. TheMyers–Briggs scale measures extraversion/introversion, intuition/sensing, feeling andjudging. Its relationship to Jung’s theory is debatable (Barbuto, 1997), but it is acceptedby many, both for practical and theoretical purposes, as a useful way of measuring the centralconcepts of Jung’s approach. The basic four dimensions carry the information of thetest and will be used here; in practice they are often used to derive 16 personality types butthat is not done here since that part of the test is highly controversial (Wiggins, 1989).A further line of work is connected with risk attitudes, as discussed in the analysisof aviation accidents. Risk taking is alleged to be a crucial negative factor in some tasks(Helmreich et al., 1986; Vollrath et al., 1999), and has been related to cognitive styles(Streufert, 1986). Five hazardous thought patterns of pilots in commercial aviation havebeen identified: anti-authority, impulsivity, invulnerability, macho and resignation

Distal factors in risk perception 189111123456789101111234567892011112345678930111123456789401111234567181111(Lester and Bombaci, 1984). A Swedish scale, using an ipsative (comparative judgements)format, was developed by the author.The risk attitude scale measures:● Anti-authoritarian inclination● Impulsivity● Beliefs about invulnerability● Macho attitudes● Resignation● Social desirability response style (lie scale) 1These risk attitudes were originally suggested in the literature on civil aviation accidents,where they have been suggested to account for a large share of pilot error, inso far as accidents are due to some form of pilot risk taking. They appear, however,to be much more generally applicable and comparisons of various groups have givenpromising results (so far unpublished). The present values of Cronbach’s alpha were,in the order of the scales given above, 0.52, 0.63, 0.65, 0.61, 0.53 and 0.65. These valuesare low, but so were the standard deviations of the scores. In a nonselect group ofadults, alphas around 0.75 were recently obtained. In the nonselect group, the standarddeviations were about 30% larger than in the present group.The current emphasis in personality psychology on the ‘Big Five’ personality factors(the Five Factor Model) opens the door for some new developments with regard topersonality, job performance (Goldberg, 1994; Wiggins and Trapnell, 1997) and socialskills (Shafer, 1999). In the present study, use was made of a Big Five questionnaireconstructed by Hendruks (1997) and translated and adapted to Swedish use byEkehammar.2.1. METHODSubjects were instructed to rate 26 hazards in terms of personal risk, general risk,demand for risk mitigation by the government, and personal responsibility for risk mitigationfor each hazard, see Table 1. Category scales going from 0 to 7 were usedthroughout, using 8 discrete steps. (A few more risk ratings were also obtained butthey will not be reported here.) A Big Five questionnaire was also used, having 20balanced items in each factor. 2 The alpha values were 0.8 or better. The Myers–Briggstest was administered in standard format. The Swedish version of the Myers–BriggsType Indicator (Mårdberg et al., 1994) was used, but the typology was not employed,only its informational basis of 4 dimensions.The risk attitude scale used an ipsative format. Subjects were instructed to rank howwell they agreed with the statements in lists of six statements each. Ten such listswere presented, in all 60 items, and each scale score was thus based on 10 items. 3 Therespondents were also asked to agree or disagree with each item, and the final rankscore of each item was negative if the subject disagreed, positive if he or she agreedto it. On the whole, the subjects appeared to be highly motivated for the test.1This lie scale is different from the two other scales of social desirability response scale used here.2Due to a technical mishap one item was missed and one was deleted for other reasons. The response scaleused 4 categories, not 5 as in the standard version. A few items were slightly rephrased.3Factors 1 and 5 used 9 items only.

190 Sjöberg11112345678910111123456789201111234567893011112345678940111123456711181111They were in all 226, who applied for admission to the Stockholm School ofEconomics. (The test situation and procedures are described in full in Sjöberg 2001a.)Their mean age was 21.3 years (range 18–37), 87 (38.5%) were female and 139 male.2.2. RESULTSThe relationships between risk perception and demand for risk mitigation and theJungian dimensions were throughout negligible. An example is given in Table 1 whichgives the results for personal risk correlated with each of the four Jungian dimensions.Very similar results were obtained for general risk and demand for risk mitigation.Only a few (weak) correlations were significant, seemingly at random.The Big Five factors did somewhat better. The pattern of Table 2 (personal risk)tells the story. The factor of emotional stability had a weak but consistent tendency tobe related to perceived risk, both personal and general. However, there was no relationshipbetween the Big Five and demand for risk mitigation.Finally, risk attitudes were also related to risk perception data, and results were interestingin particular for governmental responsibility to mitigate risk, see Table 3. TheTable 1.Personal risk correlated with Jungian dimensions according to the Myers–Briggs scale.Risk no. Extraversion Thinking Judging Intuition1. Smoking 0.05 –0.17* 0.03 0.052. Alcohol 0.14* –0.15* 0.05 –0.023. Vehicle exhausts 0.07 –0.15* –0.05 –0.024. AIDS 0.09 –0.04 –0.03 –0.085. Air pollution –0.02 –0.01 0.03 0.026. High voltage transm. lines –0.06 0.09 0.06 0.107. Greenhouse effect –0.03 –0.00 0.00 0.068. Unsuitable dietary habits 0.02 –0.14* –0.14* –0.139. Traffic accidents –0.02 –0.07 –0.10 –0.1110. Lightning 0.13 0.16* –0.00 0.16*11. Depletion of the ozone layer 0.02 0.03 –0.03 0.0712. Swedish nuclear power –0.02 0.09 0.03 0.18**13. Eastern Europe nuc. power 0.07 0.01 –0.07 0.14*14. Natural background radiation 0.11 0.06 –0.03 0.0715. Nuclear waste –0.01 0.02 0.02 0.14*16. Gene modification 0.02 –0.02 0.09 0.0117. Terrorism 0.02 0.01 0.06 –0.0618. X-ray diagnostics 0.03 0.03 –0.09 0.0419. Sunrays 0.07 –0.11 –0.15* 0.0320. War –0.04 0.00 –0.05 0.0321. Nuclear arms 0.05 –0.02 –0.00 0.1122. Floods 0.11 –0.08 0.01 0.0123. Inadequate medical care 0.01 –0.04 –0.07 –0.1024. Violence and aggression –0.04 –0.13 –0.07 –0.0925. Cellular telephones 0.06 0.04 0.03 0.0926. ‘Mad cow disease’ (BSE) –0.03 0.14* 0.08 0.08* p < 0.05, ** p < 0.01, *** p < 0.001

Distal factors in risk perception 191111123456789101111234567892011112345678930111123456789401111234567181111Table 2.‘Macho’ attitude was clearly related to demand for the government to mitigate risk.The results were very similar for own responsibility to mitigate risk. The risk attitudesdid not correlate with perceived risk per se.The correlations in Tables 1–3 were not appreciably changed when measures of socialdesirability response style were partialled out. The measures used were those suggestedby Crowne and Marlowe (1960) and the MPI Lie scale (Eysenck, 1959).2.3. DISCUSSIONPersonal risk correlated with scales in the Five Factor Model of personality.Agreeable- Extraver- Emot. Openness Consciennesssion stab. tiousness1. Smoking –0.04 –0.02 –0.09 –0.03 –0.052. Alcohol –0.04 0.03 –0.12 0.01 –0.113. Vehicle exhausts 0.06 0.10 –0.12 0.04 –0.044. AIDS –0.14 0.05 –0.14* 0.05 –0.18*5. Air pollution –0.02 0.05 –0.09 –0.01 0.036. High voltage transm. lines –0.02 0.01 –0.11 –0.02 0.097. Greenhouse effect –0.02 –0.01 –0.14* –0.03 0.048. Unsuitable dietary habits –0.07 –0.05 –0.16* –0.07 –0.23**9. Traffic accidents –0.02 –0.09 –0.16* –0.01 –0.0910. Lightning 0.03 0.07 –0.02 –0.02 0.16*11. Depletion of the ozone layer 0.06 0.01 –0.11 –0.07 0.1112. Swedish nuclear power 0.01 –0.04 –0.04 –0.02 0.20**13. Eastern Europe nuc. power 0.03 0.07 –0.12 0.04 0.1014. Natural background radiation 0.06 0.07 –0.13 –0.02 0.1015. Nuclear waste –0.06 –0.04 –0.16* –0.01 0.0916. Gene modification –0.08 0.04 –0.01 0.09 0.0117. Terrorism –0.17* –0.08 –0.13 0.00 –0.0718. X-ray diagnostics –0.03 0.01 –0.16* –0.03 –0.0019. Sunrays –0.05 0.05 –0.20** –0.12 0.0220. War –0.09 –0.18* –0.08 –0.07 –0.0621. Nuclear arms –0.08 –0.09 –0.16* –0.07 0.0122. Floods –0.04 0.06 –0.13 0.01 –0.0123. Inadequate medical care –0.04 –0.03 –0.09 –0.05 –0.0824. Violence and aggression 0.01 –0.08 –0.16* –0.06 –0.1125. Cellular telephones –0.01 0.05 –0.07 –0.05 0.0926. ‘Mad cow disease’ (BSE) 0.05 –0.03 –0.09 –0.07 0.08* p < 0.05, ** p < 0.01, *** p < 0.001The present results show:● Jungian dimensions as measured by the Myers–Briggs scale are probably notuseful for understanding risk perception● Of the Big Five scales, only emotional stability was (negatively) related toperceived risk, both general and personal, but not to demand for risk mitigation● Macho attitude was (negatively) related to demand for risk mitigation of bothtypes studied here.

192 Sjöberg11112345678910111123456789201111234567893011112345678940111123456711181111Table 3.Risk attitude scales correlated with demand for risk mitigation.Impul- Invulner- Macho Anti author- Resignasivityability itarian tion1. Smoking –0.06 –0.09 –0.03 –0.10 –0.17*2. Alcohol –0.13 –0.14* –0.04 –0.11 –0.15*3. Vehicle exhausts –0.14* –0.21** –0.24** –0.14 –0.044. AIDS 0.02 –0.13 –0.19** –0.06 –0.085. Air pollution –0.05 –0.17* –0.20** –0.09 –0.016. High voltage transm. lines –0.18* –0.24** –0.27** –0.13 –0.047. Greenhouse effect –0.04 –0.18* –0.20** –0.06 0.048. Unsuitable dietary habits –0.06 –0.05 –0.04 –0.00 –0.059. Traffic accidents –0.18* –0.15* –0.21** –0.12 –0.0710. Lightning –0.10 –0.06 –0.25** –0.06 0.0311. Depletion of the ozone layer –0.05 –0.11 –0.17* –0.00 –0.0312. Swedish nuclear power –0.08 –0.24** –0.16* –0.07 –0.0113. Eastern Europe nuc. power –0.04 –0.18** –0.20** –0.03 –0.0514. Natural background radiation –0.09 –0.07 –0.08 –0.16* –0.0715. Nuclear waste –0.14* –0.22** –0.24** –0.10 –0.0216. Gene modification –0.07 –0.03 –0.17* –0.07 0.0617. Terrorism –0.07 –0.15* –0.23** 0.03 –0.0018. X-ray diagnostics –0.08 –0.21** –0.23** –0.12 –0.0219. Sunrays –0.09 –0.18** –0.16* –0.12 –0.1020. War –0.17* –0.14* –0.23** 0.01 –0.0421. Nuclear arms –0.18** –0.18* –0.22** –0.05 –0.0722. Floods –0.04 –0.09 –0.19** –0.09 –0.0223. Inadequate medical care –0.11 –0.14* –0.23** –0.04 –0.0824. Violence and aggression –0.09 –0.12 –0.11 –0.09 –0.0425. Cellular telephones –0.05 –0.09 –0.14 –0.04 –0.0426. ‘Mad cow disease’ (BSE) –0.14* –0.19** –0.25** –0.20** –0.06* p < 0.05, ** p < 0.01, *** p < 0.001Previous studies have reported results that are in accordance with the present findingsregarding the relationship between perceived risk and emotional stability (Sjöberg,1998b; Källmén, 2000). The fact that perceived risk has structural properties differentfrom demand for risk mitigation is to be expected from work on demand for risk reduction,which has shown that such demand is only weakly related to size of a perceivedrisk, much more to the severity of consequences should an adverse event occur (Sjöberg,1999a; Sjöberg, 2000a).Personality theories such as the Jungian one lend themselves easily to speculationsabout risk perception. However, what data say is another thing. Risk perception is noteasy to account for with personality concepts and the present results, while consistent,do not show very strong correlations. Maybe other personality constructs will do better.But it is likely that it will be hard to find them. Standard Big Five scales used here docover a wide ground of personality in a succinct manner.It is interesting to note that demand for risk mitigation could be explained to someextent with a risk attitude. The demand variable has turned out to be quite hard toaccount for in previous research. It is possible that it would be fruitful in future work

Distal factors in risk perception 193111123456789101111234567892011112345678930111123456789401111234567181111to further develop the attitude measure, in order to account for at least part of thedemand dimensions.3. Study 2In the present study, three distal approaches to risk perception are studied andcompared: Cultural Theory (Douglas and Wildavsky, 1982), the NEP scales (Dunlapet al., 1992) and New Age beliefs (Sjöberg, 2002c). In Cultural Theory it is assumedthat people ‘choose’ to worry about different hazards on the basis of their social engagementsin a ‘Group/Grid’ pattern. The theory is presented in a text by Douglas andWildavsky (1982), see Boholm for a cogent critique (Boholm, 1996). Most of the empiricalwork on this model is quantitative and uses scales devised by Dake (Wildavskyand Dake, 1990), or similar to them. In the USA, such scales account typically for some10% of the explained variance of perceived risk; but in Europe only for about 5%(Sjöberg, 1997; Brenot et al., 1998; Sjöberg, 1998a).In spite of the weak empirical results of Cultural Theory it seems to be credible insome quarters, see e.g. Adams (1995), and it is used in various applied settings. Furtherempirical work on the theory is therefore called for. A recent discussion by Tansey (inpress), in particular, called for further work on these issues, since it also brought upthe issue of distal explanatory variables more generally.Statistical illusions seem to occur in discussions of Cultural Theory. Peters and Slovicobtained data well in line with a very modest explanatory power (Peters and Slovic,1996) but described their very low correlations, down to values well below 0.05, betweenCultural Theory scales and risk perception as ‘strong’.Almost all of the attempts at quantitative tests of Cultural Theory investigate thelink between World Views and risk perception. However, the most basic concepts inthe theory are those of Group and Grid and they have rarely been tested. ‘Grid’ refersto deference to others, especially to authority. ‘Group’ refers to membership in groupsand freedom of expression of deviant opinions. Here the relationship between riskperception and World Views is investigated once more, now with a large and reasonablyrepresentative sample of the Swedish population. Also the Group/Grid dimensionsare investigated and related to risk perception. Furthermore, New Age beliefs are investigated.In a previous study by Sjöberg (2002c) it was found that such beliefs correlatedmoderately strongly (about 15% explained variance) with perceived level of varioustechnology and environment hazards. Such a relationship was expected on the groundsthat the New Age movement, very strong and of growing importance since the 1970s,at its core includes a set of beliefs which are alien and even hostile to modern technologyand to science.It is therefore interesting to compare and contrast World Views according to CulturalTheory and New Age beliefs. The former are based on an elaborate theory about socialprocesses which are assumed to generate belief structures. The belief contents are quitedistant from risks and hazards (with a few exceptions in some of the scales which havebeen used to measure them). For example, egalitarian beliefs are close to politicalsocialism, and individualistic beliefs are close to classical liberalism. New Age beliefs,on the other hand, tend to be oriented towards various conceptions of epistemologyand ontology. In the previous work cited above, Sjöberg found, using a large numberof items as potential indicators of New Age beliefs, four distinct factors:

194 Sjöberg111123456789101111234567892011112345678930111123456789401111234567111811111. Beliefs in a ‘Higher Consciousness’2. Belief in the physical reality of the soul3. Traditional folk superstition4. Denial of science and analytic thinking as modes of inquiryThe most powerful dimension turned out to be the first one, illustrated by the itemsused (see Appendix). In addition, beliefs in paranormal phenomena such as theBermuda Triangle and ‘The hundredth ape’ were useful as an explanatory factor inaccounting for perceived risk.A final set of items was concerned with environmental beliefs (Dunlap and van Liere,1978; Dunlap et al., 1992; Dietz et al., 1998; Dunlap et al., 2000), and measured fivesuch dimensions:1. Limits to growth2. Anti-anthropocentrism3. Balance of nature4. Rejection of exemptionalism5. Possibility of eco-crisisSome environmental beliefs have been found to correlate moderately strongly with riskperception (Sjöberg, 2002b). The NEP items have been used in many contexts tomeasure environmental concern and were therefore considered to be of interest here.The purpose of the study was, thus, to investigate three sets of potentially importantdistal determinants of risk perception.Most previous work has been carried out with convenience samples or with sampleswith very low response rates, see the methodological discussion in Sjöberg and Drottz-Sjöberg (2001); In the present paper results on these dimensions from a largerepresentative sample of the Swedish population are presented.3.1. METHODA questionnaire was mailed to a random sample of the Swedish population. It wasintended to cover a wide range of issues and the respondents therefore were asked toanswer 313 questions on a total of 37 pages. The questionnaire was sent out in May 1998.The net sample consisted of 1202 respondents. Since we got 797 answers, the responserate was 66.3%. The sample was bought from SPAR (the government’s person andaddress database) and included respondents in the age range 18–75, but three people hadturned 76 when they responded to the questionnaire. The respondents who were notSwedish citizens were excluded from the sample as well as the respondents stating thatthey were not the person to whom the questionnaire was addressed.The objective of the study was to work with a sample that was representative of theSwedish population, in order to be able to generalize the results. According to the resultsregarding background variables, we seem to have reached that objective. The sample isrepresentative of the Swedish population in terms of gender distribution and average age.When it comes to whether the respondents had children or not, it turned out that thepercentage of respondents who had children was slightly higher than the percentageamong the Swedish population. Since the difference was so small, it is reasonable toconclude that it did not affect the results very much. With regard to income, there was a

Distal factors in risk perception 195111123456789101111234567892011112345678930111123456789401111234567181111difference between the sample and the Swedish population. Income is not, however, animportant explanatory variable. Furthermore, the structure of employment status amongthe respondents was quite similar to that of the population, but there were differencesconsisting of fewer students and more retired people in the sample. This is a well-knownphenomenon, which is probably due to the fact that retired people have more time to fillout extensive questionnaires. Moreover, there was a notable difference between theshare of unemployed people in the sample and the share in the population, which is probablyat least partly a consequence of the fact that the unemployment rates had decreasedsomewhat in Sweden since 1997 (the year from which the information on unemploymentrates was collected). The respondents are also representative of the population in termsof type of employer and occupational status. The largest difference is due to educationallevel. The respondents were better educated than the general population. Level of educationtends, however, to be only weakly related to risk perception. Hence, the mainconclusion is that the sample on the whole is representative of the population, especiallywith a response rate as high as 66.3%.In the present study ratings of general and personal risk of 37 hazards are used (seeTable 7), on eight step category scales. A ‘don’t know’ response category was also used,and DK answers were throughout treated as missing. In addition to the risk ratings,we also used new items measuring World Views and Group/Grid dimensions (Lockartet al., 1997). It had been believed in some quarters that the lacklustre effects in manystudies were due to items being too similar to the American ones and not taking specialcultural dimensions into account. Hence, it was of interest to scrutinize the validity ofCultural Theory with items that may fit better in a European context.3.2. RESULTS3.2.1. Cultural TheoryIndices were formed, to measure the six core concepts of Cultural Theory, viz. hierarchy,egalitarianism, individualism, fatalism, group and grid. There were three itemsfor each of these indices. The first task was to study correlations between World Viewsand perceived general and personal risk of the 37 hazards, see Tables 4 and 5.It is immediately clear from the tables that neither general nor personal risk ratingswere strongly correlated with world views. Egalitarianism and fatalism did show sometendency towards correlations about 0.2. Naturally, many correlations were significant,due to the large sample size.Cultural theory makes differential predictions about the relationships betweendifferent hazards and the World Views. However, multiple regression analyses with allWorld Views and the Group/Grid dimensions as explanatory variables yield an upperbound to how much can be accounted for. The average R 2 adj was 2.6% for personal riskand 3.2% for general risk.According to Cultural Theory, the four combinations of high versus low group andgrid should be characterized as follows:●●●●High group, high grid: hierarchyHigh group, low grid: egalitariansLow group, high grid: fatalistsLow group, low grid: individualists

196 Sjöberg11112345678910111123456789201111234567893011112345678940111123456711181111Table 4.World views correlated with general risk ratings.Hierarchy Egalitarianism Individualism Fatalism1. Violence and aggression 0.07 0.04 0.02 0.09*2. Environmental degradation –0.06 0.11** –0.12** 0.063. Deteriorated economy –0.08* 0.07 –0.05 0.014. Unemployment –0.07* 0.08* –0.09* 0.015. Own smoking –0.06 0.05 –0.05 –0.066. Own alcohol consumption 0.02 0.13** –0.03 –0.007. Vehicle exhausts –0.02 0.16** –0.06 0.09*8. Getting AIDS 0.09* 0.09* –0.01 0.13**9. Air pollution –0.07 0.13** –0.08* 0.0210. High voltage power lines 0.06 0.15** –0.12** 0.14**11. The Greenhouse effect –0.06 0.13** –0.10** 0.0412. Radiation from indoor radon –0.07 0.12** –0.12** 0.0713. Inadequate dietary habits –0.13** 0.03 –0.07 –0.12**14. Irradiated food 0.09* 0.19** –0.08* 0.15**15. Traffic accident 0.05 0.09* –0.06 0.0116. Struck by lightning 0.16** 0.14** –0.04 0.20**17. Depletion of the ozone layer –0.07 0.14** –0.11** –0.0018. Domestic nuclear power –0.05 0.29** –0.19** 0.12**19. Western Europe nuclear power –0.00 0.28** –0.15** 0.15**20. Eastern Europe nuclear power 0.00 0.13** –0.07 0.10*21. Natural background radiation 0.07 0.19** –0.10* 0.14**22. Nuclear waste 0.01 0.23** –0.15** 0.18**23. Genetic engineering –0.00 0.14** –0.13** 0.14**24. Polluted drinking water 0.02 0.11** –0.06 0.12**25. Terrorist attacks 0.20** 0.11** 0.03 0.23**26. Food poisoning outside home 0.08* 0.05 –0.05 0.0627. Food poisoning in home –0.06 –0.00 –0.11** –0.0128. X-ray diagnostics 0.08* 0.16** –0.02 0.14**29. Sunrays –0.03 0.12** –0.10** –0.0130. War 0.06 0.15** –0.08* 0.15**31. Chemical waste –0.05 0.12** –0.15** 0.11**32. Transports of dangerous goods 0.04 0.17** –0.10** 0.17**33. Nuclear arms 0.05 0.18** –0.13** 0.20**34. Floods 0.08* 0.09* –0.06 0.10**35. Radioactive fallout from the 0.12** 0.19** –0.06 0.17**Chernobyl accident36. Inadequate medical care when ill –0.01 0.05 0.01 0.11**37. ‘Mad cow’ disease (BSE) 0.07 0.17** –0.12** 0.19*** p < 0.05, ** p < 0.01, *** p < 0.001Correlations between the four World View scales were very low and nonsignificant forthe group variable. For grid, they were all significant at the 0.01 level or better: 0.43, 0.11,0.26 and 0.11, respectively. The sample was split at the median of group and grid, respectively,and these variables were used as independent variables in two-way ANOVA’swith the four World View scales as dependent variables. Grid was significant throughout,group not in any case. No interactions were significant. High grid–low group gave the

Distal factors in risk perception 197111123456789101111234567892011112345678930111123456789401111234567181111Table 5.World views correlated with personal risk ratings.Hierarchy Egalitarianism Individualism Fatalism1. Violence and aggression –0.05 –0.02 –0.03 0.042. Environmental degradation –0.15** 0.06 –0.12** –0.013. Deteriorated economy –0.10** 0.07 –0.05 0.054. Unemployment –0.08* 0.09* –0.05 0.065. Own smoking –0.03 0.04 0.02 0.09*6. Own alcohol consumption –0.04 –0.04 0.04 0.047. Vehicle exhausts –0.08* 0.10** –0.06 0.058. Getting AIDS –0.01 0.01 –0.01 0.059. Air pollution –0.13** 0.09* –0.08* –0.0110. High voltage power lines 0.04 0.16** –0.09* 0.13**11. The Greenhouse effect –0.07 0.13** –0.04 0.0012. Radiation from indoor radon –0.02 0.13** –0.05 0.09*13. Inadequate dietary habits –0.04 –0.03 0.01 0.0014. Irradiated food 0.04 0.17** –0.07 0.14**15. Traffic accident –0.05 –0.01 0.03 –0.0316. Struck by lightning 0.10** 0.10** –0.00 0.16**17. Depletion of the ozone layer –0.05 0.17** –0.05 –0.0118. Domestic nuclear power –0.00 0.29** –0.16** 0.11**19. Western Europe nuclear power 0.04 0.27** –0.13** 0.13**20. Eastern Europe nuclear power 0.03 0.14** –0.04 0.0721. Natural background radiation 0.03 0.20** –0.10* 0.0622. Nuclear waste 0.01 0.28** –0.17** 0.14**23. Genetic engineering 0.03 0.11** –0.08* 0.15**24. Polluted drinking water 0.01 0.08* –0.09* 0.13**25. Terrorist attacks 0.17** 0.06 0.04 0.20**26. Food poisoning outside home 0.01 –0.05 –0.01 0.0327. Food poisoning in home –0.05 0.01 –0.09* 0.0428. X-ray diagnostics 0.05 0.13** 0.03 0.13**29. Sunrays –0.02 0.05 –0.07 0.0330. War 0.06 0.14** –0.09* 0.10**31. Chemical waste –0.08* 0.09* –0.12** 0.08*32. Transports of dangerous goods –0.03 0.11** –0.13** 0.13**33. Nuclear arms 0.06 0.19** –0.14** 0.16**34. Floods 0.09 0.06 –0.06 0.10**35. Radioactive fallout from the 0.06 0.14** –0.05 0.12**Chernobyl accident36. Inadequate medical care when ill –0.02 –0.01 0.04 0.11**37. ‘Mad cow’ disease (BSE) 0.08* 0.10** –0.05 0.23*** p < 0.05, ** p < 0.01, *** p < 0.001highest value in three cases, high grid–high group for hierarchy. Hence, in that case, thetheory had some predictive power while in all other cases the data looked as if thesubjects were fatalists. (By the way, fatalism was the least popular World View.) The relativelyhigh correlation between grid and hierarchy may be due to some semantic overlapbetween the items. Be that as it may, the theory was not supported by the present results.

198 Sjöberg111123456789101111234567892011112345678930111123456789401111234567111811113.2.2. New Age beliefsNew Age and Environmental Beliefs were correlated with perceived general andpersonal risk, see Tables 6 and 7. New Age beliefs had the strongest correlation inTable 6. Correlations between NEP scales, New Age beliefs and general risk (largest correlationin each row in boldface).Limits to Anti Balance Rej. Poss. Newgrowth anthr. nature exemp. eco crisis Age1. Violence and aggression –0.01 0.04 0.08* –0.02 0.12** 0.15**2. Environmental degradation 0.06 0.16** 0.16** 0.10** 0.32** 0.08*3. Deteriorated economy 0.07 0.09* 0.12** 0.08* 0.24** 0.064. Unemployment 0.03 0.07* 0.11** 0.03 0.20** 0.045. Own smoking –0.01 0.11** 0.13** 0.07* 0.14** –0.056. Own alcohol consumption 0.02 0.07* 0.05 0.04 0.12** 0.10**7. Vehicle exhausts 0.06 0.10** 0.11** 0.03 0.28** 0.14**8. Getting AIDS 0.02 0.01 0.07 –0.01 0.11** 0.22**9. Air pollution 0.04 0.12** 0.15** 0.12** 0.26** 0.16**10. High voltage power lines 0.10** 0.07 0.11** 0.03 0.23** 0.33**11. The Greenhouse effect 0.12** 0.13** 0.16** 0.13** 0.27** 0.14**12. Radiation from indoor radon 0.07 0.04 0.08* 0.09* 0.22** 0.18**13. Inadequate dietary habits 0.05 0.11** 0.11** 0.17** 0.23** 0.0314. Irradiated food 0.07 0.06 0.13** 0.07 0.25** 0.24**15. Traffic accident 0.05 –0.00 0.05 –0.02 0.07 0.09*16. Struck by lightning 0.07 –0.12** –0.04 –0.13** –0.03 0.22**17. Depletion of the ozone layer 0.10** 0.11** 0.19** 0.16** 0.31** 0.13**18. Domestic nuclear power 0.05 0.05 0.07 0.03 0.20** 0.27**19. Western Europe nuclear power 0.09* 0.05 0.11** 0.03 0.21** 0.28**20. Eastern Europe nuclear power 0.14** 0.15** 0.18** 0.09* 0.25** 0.22**21. Natural background radiation 0.17** 0.01 0.03 –0.02 0.15** 0.26**22. Nuclear waste 0.11** 0.02 0.08* 0.01 0.22** 0.26**23. Genetic engineering 0.16** 0.07 0.19** 0.06 0.23** 0.22**24. Polluted drinking water 0.11** 0.02 0.09* 0.01 0.14** 0.21**25. Terrorist attacks 0.08* –0.06 0.05 –0.12** 0.01 0.23**26. Food poisoning outside home 0.08* 0.05 0.10** 0.00 0.08* 0.11**27. Food poisoning in home 0.05 0.04 0.07 0.01 0.13** 0.0328. X-ray diagnostics 0.07 –0.03 0.05 –0.03 0.08* 0.25**29. Sunrays 0.07 0.09* 0.16** 0.13** 0.22** 0.10**30. War 0.11** 0.02 0.07 –0.03 0.12** 0.23**31. Chemical waste 0.07 0.09* 0.15** 0.06 0.23** 0.17**32. Transports of dangerous goods 0.05 0.02 0.13** –0.01 0.13** 0.25**33. Nuclear arms 0.08* 0.03 0.09* –0.03 0.15** 0.25**34. Floods 0.06 0.01 0.09* 0.03 0.10** 0.13**35. Radioactive fallout from the 0.08* 0.06 0.12** 0.00 0.16** 0.27**Chernobyl accident36. Inadequate medical care when ill 0.09* 0.03 0.10** 0.07 0.13** 0.16**37. ‘Mad cow’ disease (BSE) 0.08* 0.05 0.08* –0.01 0.09* 0.17*** p < 0.05, ** p < 0.01, *** p < 0.001

Distal factors in risk perception 19911112345678910111123456789201111234567893011112345678940111123456718111120 of the 37 cases, ecological crisis in the remaining 17 cases. The results were similarto those in the case of general risk. New Age beliefs had the largest correlation withpersonal risk in 20 of the 37 cases (there were some ties).Table 7. Correlations between NEP scales, New Age beliefs and personal risk (largest correlationin each row in boldface).Limits to Anti anthr. Balance Rej. exemp. Poss. eco New Agegrowth nature crisis1. 0.04 –0.01 0.04 0.02 0.13** 0.022. 0.07* 0.12** 0.16** 0.11** 0.30** 0.043. 0.01 0.04 0.07 0.04 0.17** 0.064. –0.01 0.05 –0.00 0.03 0.08* 0.08*5. 0.05 0.09* 0.02 –0.01 0.06 0.066. 0.08* 0.01 –0.02 –0.03 0.03 –0.057. 0.05 0.09* 0.08* 0.07 0.21** 0.10**8. –0.01 0.05 0.02 –0.02 0.08* 0.12**9. 0.09* 0.15** 0.18** 0.13** 0.28** 0.08*10. 0.07 0.03 0.07 0.06 0.21** 0.24**11. 0.11** 0.14** 0.21** 0.15** 0.30** 0.12**12. 0.09* 0.01 0.01 0.07 0.09* 0.14**13. 0.06 0.05 0.03 0.04 0.06 0.0014. 0.05 0.06 0.10** –0.01 0.18** 0.19**15. 0.06 0.02 0.06 0.01 0.07 0.0316. 0.00 –0.14** –0.06 –0.12** –0.04 0.22**17. 0.13** 0.15** 0.22** 0.15** 0.34** 0.15**18. 0.07 0.08* 0.09* 0.07 0.22** 0.25**19. 0.10** 0.08* 0.15** 0.01 0.23** 0.28**20. 0.15** 0.13** 0.21** 0.08* 0.25** 0.20**21. 0.14** 0.02 0.06 0.03 0.14** 0.23**22. 0.11** 0.04 0.05 0.01 0.19** 0.26**23. 0.14** 0.05 0.17** 0.06 0.21** 0.19**24. 0.15** 0.05 0.12** 0.01 0.15** 0.15**25. 0.09* –0.08* –0.02 –0.15** 0.00 0.23**26. 0.11** 0.04 0.07 0.03 0.07* 0.0327. 0.05 –0.03 –0.04 –0.03 0.06 0.0428. 0.04 –0.01 0.05 0.00 0.09* 0.19**29. 0.08* 0.06 0.12** 0.10** 0.17** 0.0130. 0.09* 0.01 0.04 –0.03 0.12** 0.19**31. 0.10** 0.06 0.10* 0.06 0.19** 0.11**32. 0.10* 0.06 0.09* 0.03 0.11** 0.14**33. 0.13** 0.01 0.09* –0.02 0.16** 0.25**34. –0.03 –0.10** –0.01 –0.07* 0.00 0.13**35. 0.13** 0.11** 0.08* 0.01 0.13** 0.20**36. 0.08* –0.01 0.05 0.03 0.09* 0.09*37. 0.06 –0.00 0.05 –0.05 0.04 0.13*** p < 0.05, ** p < 0.01, *** p < 0.001

200 Sjöberg111123456789101111234567892011112345678930111123456789401111234567111811113.2.3. Comparing New Age, Environmental Beliefs and World ViewsSix indices of risk perception were formed, measuring personal and general risk ofhazards in general, radiation hazards (non-nuclear) and nuclear hazards. The groupingof hazards was based on both practical and theoretical considerations, since people wereexpected to react in special ways to radiation and nuclear hazards.The indices were subjected to block regression analyses, with the following strategy:●●●●Block 1 – age, sex, educational level and political orientation (left versus right)Block 2 – New Age BeliefsBlock 3 – NEP scalesBlock 4 – World Views according to Cultural TheoryThe amount of total variance explained by each block, as well as the added variancedue the inclusion of each block, are given in Table 8. It is seen that World Views constitutedthe least efficient block of predictors, accounting for a mere 5.5% of the variancein perceived risk, on the average. NEP scales were also not more efficient than that.New Age beliefs accounted for twice that amount. World Views only contributedanother 2.6% explained variance in addition to the other blocks, on the average.Furthermore, the regression coefficients (see Table 9 for personal nuclear risk)showed that New Age beliefs had a dominating position in accounting for perceivedrisk, when all independent variables were included in the regression equation. It waseven more powerful than gender. New Age beliefs had the largest regression weight infour of the six analyses and a large one also in the remaining two.4. Study 3Studies 1 and 2 allow for the analysis of several approaches to distal determinants ofperceived risk. However, the most important one has been Cultural Theory and it istherefore justified to obtain more extensive data than in previous studies on its dimensionsand relation to perceived risk. In a collaborative project, the four CT dimensionsTable 8. Variance accounted for (R 2 adj) by blocks of explanatory variables, for six indices of riskperception.Index Block 1 Block 2 Block 1 Block 3 Blocks Block 3 Full(back- (New + NEP 1–3 (World modelground Age Block 2 items Views) (all 3data) beliefs) blocks)Hazards in general, 0.016 0.069 0.039 0.047 0.072 0.038 0.099personal riskHazards in general, 0.091 0.095 0.110 0.055 0.145 0.038 0.164general riskRadiation, personal risk 0.023 0.055 0.057 0.020 0.068 0.037 0.091Radiation, general risk 0.065 0.109 0.110 0.041 0.136 0.049 0.154Nuclear, personal risk 0.085 0.129 0.130 0.054 0.165 0.075 0.191Nuclear, general risk 0.115 0.138 0.160 0.053 0.193 0.095 0.235Average 0.066 0.099 0.131 0.045 0.130 0.055 0.156

Distal factors in risk perception 201111123456789101111234567892011112345678930111123456789401111234567181111Table 9.were measured in the five Nordic countries (Denmark, Finland, Iceland, Norway andSweden) and related to risk perception of genetic engineering (Grendstad et al., 1999).4.1. METHODThe technical details of this study are available in a separate publication (Grendstad et al.,1999). The data were obtained by professional polling firms and with telephone interviews.The respondents are representative of the respective populations, aged 15 and above, inthe sense of professional standards developed by these firms. The sample sizes obtainedwere Denmark: 1015, Finland: 1003, Iceland: 817, Norway: 997, Sweden: 1000. There werefive items for each of the four CT dimensions. Great care was taken in translating the itemsto make them comparable in content and connotations across countries. There were alsotwo items measuring the perceived personal and general risk of genetic engineering.4.2. RESULTSRegression weights of the full model, personal nuclear risk.Variable B SE B Beta T Sig TSex 0.493 0.100 0.168 40.927 0.000Age 0.001 0.003 0.012 0.351 0.726Educational level –0.051 0.026 –0.076 –20.012 0.045Political att. –0.009 0.053 –0.006 –0.173 0.863Limits to growth 0.138 0.074 0.063 10.857 0.064Anti-anthrop. 0.009 0.077 0.004 0.113 0.910Balance nat. 0.004 0.094 0.002 0.040 0.968Rej. exemp. 0.009 0.078 0.004 0.120 0.905Eco crisis 0.325 0.091 0.152 30.591 0.000New Age beliefs 0.404 0.068 0.219 50.924 0.000Hierarchy –0.089 0.072 –0.052 –10.244 0.214Egalitarianism 0.185 0.068 0.101 20.729 0.006Individualism –0.257 0.074 –0.128 –30.483 0.001Fatalism 0.130 0.066 0.074 10.984 0.048Grid –0.080 0.073 –0.039 –10.106 0.269Group 0.039 0.048 0.027 0.801 0.423(Constant) –0.129 0.652 –0.198 0.843The Cronbach values are given in Table 10, correlations with risk perception in Tables11 and 12. The data show acceptable reliabilities, although in some cases bordering onlow values. This is a fairly common finding in CT work with few items. The correlationswith risk perception within countries were low but in many cases statisticallysignificant. There were some consistent trends in the sense that signs and levels of correlationcoefficients were similar across the five countries.The data set allows for a detailed analysis of the relationships. Personal riskwas selected for the analysis, as well as egalitarianism. The overall correlation was 0.13(p < 0.0005). The sample was divided into 10 deciles with regard to egalitarianism andthe mean standardized risk was computed for each decile, see Fig. 1. The distributions

202 Sjöberg11112345678910111123456789201111234567893011112345678940111123456711181111Table 10.Cronbach values for the CT dimensions in the Nordic countries.Norway Sweden Denmark Finland IcelandEgalitarianism 0.63 0.58 0.62 0.70 0.69Fatalism 0.63 0.59 0.56 0.62 0.57Hierarchy 0.67 0.56 0.62 0.61 0.58Individualism 0.66 0.58 0.58 0.60 0.55Table 11.personal.Correlations between CT dimensions and perceived risk of genetic engineering,Norway Sweden Denmark Finland IcelandEgalitarianism 0.08* 0.13** 0.22** 0.15** 0.14**Fatalism 0.01 0.08* 0.09* 0.00 0.06Hierarchy –0.04 –0.01 0.03 0.07* 0.06Individualism –0.12** –0.03 –0.11** –0.14** –0.03* p < 0.05, ** p < 0.01Table 12.general.Correlations between CT dimensions and perceived risk of genetic engineering,Norway Sweden Denmark Finland IcelandEgalitarianism 0.13** 0.23** 0.24** 0.18** 0.19**Fatalism 0.13** 0.11* 0.10** 0.07* 0.11**Hierarchy 0.05 0.00 0.05 0.10** 0.09*Individualism –0.10** –0.05 –0.12** –0.14** –0.08** p < 0.05, ** p < 0.01of risk ratings for the 20 per cent lowest and the 20 per cent highest in egalitarianismare given in Fig. 2. The figures show that the strongest relationship between perceivedrisk and the CT dimension are found for low values of egalitarianism (Fig. 1).Furthermore, there is a group of low egalitarians who also are extremely low inperceived risk. It is this group which accounts for the relationship.A further analysis of the low-risk and high-risk judges was of interest. The subgrouprating both personal and general risk as = 1 (‘insignificant’) was singled out (N = 636),and standardized values of the CT dimensions computed and compared to all otherrespondents (N = 3632), see Fig. 3. In the same figure, data are also provided for thesubgroup rating both risks as ‘very large’, a group consisting of 125 (3%) risk alerters.The high-risk raters may be termed risk alerters and the low-risk raters risk deniers.The risk deniers turned out to be especially low in egalitarianism but also in fatalismand high in individualism. The hierarchy dimension did not differentiate them from allothers. The risk alerters were high especially in egalitarianism but also to some extentin hierarchy and low in individualism. The low-risk group contained 64.2% of men, for

Distal factors in risk perception 203111123456789101111234567892011112345678930111123456789401111234567181111Fig. 1.Fig. 2.Mean risk rating as a function of egalitarianism (10 deciles).Distribution of risk ratings for the lowest and highest 20% in egalitarianism.

204 Sjöberg11112345678910111123456789201111234567893011112345678940111123456711181111Fig. 3.Mean Cultural Theory dimensions for risk deniers and risk alerters.all others the figure was 47.0%. The risk alerters had a slight majority of women, 56%.If risk deniers and risk alerters are deleted, all correlations between perceived risk andCT dimensions fell below 0.1, except for general risk and egalitarianism and fatalismwhich correlated 0.13 and 0.12 in the group that remained. Note that even a low correlationof 0.13 looks like a strong relationship in this kind of plot.4.3. DISCUSSIONThe significant values in the present study were found with large sample sizes. Hence,values down to 0.07 were significant at the 0.05 level. The absolute size of the correlationswere, however, quite modest and in line with other empirical tests of CT.Egalitarianism seemed to be most clearly related to perceived risk. The findings supportthe notion that there exists a weak but consistent relationship between perceived riskand at least some CT dimensions.A further analysis suggested that the crucial distinction is that between low-risk judgesand others. The low-risk judges did have a deviating CT profile with a quite low valuein egalitarianism and a relatively low value in fatalism and high in individualism. Theytended to be men. This is a pattern remindful of the Flynn and Slovic findings that lowrisk ratings were given by white men, while women and black men gave higher ratings(Flynn et al., 1994). In a recent study we also found some evidence of a tendencyto mock the risk of genetic engineering, even seeing it as morally condemnable not tosupport it (Sjöberg, 2002b). It may be concluded that correlations between CT dimensionsand perceived risk are largely due to the existence of a group of risk deniers (inthe present data 15% of all respondents), but also a group of risk alerters (much smaller)plays a role in boosting the relationship.

Distal factors in risk perception 2051111234567891011112345678920111123456789301111234567894011112345671811115. General discussionThe present results agree well with the previous study of risk perception and New Agebeliefs (Sjöberg, 2002c). Since the New Age movement has developed and grown inparallel with technology opposition and worries about technology risks, and since themovement espouses many beliefs alien or even hostile to science and technology,it seems reasonable to conclude that it may be an important component in modernrisk discourse.The problem of finding distal determinants of risk perception is of course not solvedby the use of New Age beliefs. Much remains to be explained. It is, however, possiblethat New Age beliefs constitute a promising start towards the understanding of riskperception on the basis of a variable that does not have a close conceptual relationshipto perceived risk.The present findings may seem to be hard to reconcile with Cultural Theory.However, some authors argue that a pattern of correlations in accordance with a theoryis the important type of finding and that strength of relationship is an inconsequentialaspect (Marris et al., 1998); see Sjöberg (2002a) for a discussion. However, weak correlationscan easily arise due to the effect of confounding factors, such as demographicdimensions. In addition – and more important – a theory whose proponents are satisfiedwith explaining only a very small share of the phenomenon it sets out to accountfor is of marginal interest. After all, we should explain as much as possible of the truevariance of risk perception.Other types of value dimensions do not fare very well either. In other work, theWorld Views suggested by Buss and Craik (1983) have been found to be equally, infact even more, ineffective than Cultural Theory variables in accounting for risk perception(Sjöberg, 2002b), and the same can be said for more general value dimensionsystems (Sjöberg, 1998a). Hence, Cultural Theory does not come out very well in thepresent study, neither do World Views nor the Group/Grid dimensions. In this sense,the present results are in good agreement with previous research, with the possibleexception of Dake’s dissertation (Dake, 1990; Wildavsky and Dake, 1990) wherestronger correlations were reported. It is unclear what the reason could have been forthe more positive results obtained by Dake.Now, given the very weak findings, what – if anything – can save the theory? Onestrategy is to do qualitative case studies (Tansey and O’Riordan, 1999). One can alwayshope to find single individuals who appear to function according to the theory. This istrue in clinical psychology as well, to take another example, and little of interest seemsto be coming out of such an approach. In medical research, a similar warning againstreliance on single case studies was issued in a paper by Kline (Kline, 1962), who calledthis strategy ‘individualizing’ and grouped it with other methods for dodging from negativeconclusions on the basis of empirical evidence. It is just too easy to bend and twistthe ‘evidence’ to fit to the desires of the investigator, who often has a vested interestin demonstrating the truth of his theory.Another strategy is simply to deny that the theory was intended to explain risk perceptionin the first place. Given published articles with this very purpose, it is a surprisingclaim, but nonetheless Marris et al. (1998) write that ‘cultural theory does not reallyclaim to explain such abstract ratings of risk’ (p. 645).Still another way out is to state that better measures of cultural biases would yieldstronger results (Peters and Slovic, 1996). Yes, provided that the theory holds. The logic

206 Sjöberg11112345678910111123456789201111234567893011112345678940111123456711181111of the argument is circular. If the theory holds can only be tested by means of operationalizingsuch improved measures. The hope that they may one day exist is no supportfor the theory.In a paper by Slovic and Peters, a further attempt was made to salvage CulturalTheory (Slovic and Peters, 1998), now based on statistical argumentation. They notedmy critique of the empirical power of Cultural Theory, as it had been operationalizedin attitude-type items (Sjöberg, 1997). I had pointed out that the scales derived in thisway usually correlated only 0.2 – 0.3 with risk ratings, 4 and that Peters and Slovic hadobtained even lower correlations, still claiming them to be ‘strong’. They asserted that‘The squared correlation has long been known to be a misleading indicator of the importanceof a relationship’ (p. 166). This seems to overstate their position. The squaredcorrelation has been used traditionally as a measure of strength of relationship and is,for example, recommended in the 1994 edition of the APA publication manual whereeffect size measures are strongly recommended for general use (American PsychologicalAssociation, 1994). Cohen’s widely cited and used categorization of effect sizes requiresa much higher level of correlation than 0.1 for deserving to be called ‘strong’ (Cohen,1988).The squared correlation is the first measure of effect size mentioned in the APAmanual (p. 18). It is true that some authors have criticized it and proposed othermeasures, among them Rosenthal and Rubin’s Binomial Effect Size Display (Rosenthaland Rubin, 1982). This method is said to yield results more in line with intuitive notionsas to effect size, but it is by no means uncontroversial, see the cogent critique publishedby Thompson and Schumacker (1997). The Binomial Effect Size Display may be usefulin a situation where the effects of experimental interventions are under study; that isnot at all the case in evaluating the validity of Cultural Theory in tests based whollyon correlational data. The same argument pertains to the paper by Funder and Ozer(1983), cited by Slovic and Peters. Various interventions may yield what appears to beimpressive results while the effects described by means of correlation statistics appearto be small. A similar intervention based argument has been made by Prentice andMiller (1992).Slovic and Peters furthermore cited D’Andrade and Dart (1990) and their argumentagainst r 2 which says that r 2 should not be used because it uses a different metric thanthe original r coefficient. This argument is superficially correct, of course, but it isunclear what should be concluded from it. They then went on to demonstrate that evenwith small correlations, extreme cases in one variable will be quite different also in thecorrelated variable. Of course, this will always happen as long as the correlation is notzero and the effect depends on how extreme the selected cases are. With cases extremeenough, any nonzero correlation can be made out to look very impressive.Why is this discussion important? A correlation of 0.1 is a correlation of 0.1. If somebodywants to call it ‘strong’ this is his or her prerogative. The important thing aboutthe discussion is the rhetorical and persuasive uses of statistics that are revealed bysuch verbal habits. The reader is given the impression that most of the story has beentold, when a predictor correlating 0.33 with a dependent variable is even called‘extremely strong’ (p. 168) by Slovic and Peter. 5 Furthermore, readers are discouraged4This empirical statement has not been contested, as far as I know.5The ‘extremely’ strong correlation of 0.33 reported by Slovic and Peters is, by the way, the standard level obtainedin most reasonably well done psychometric work, such as studies of personality and behaviour, see Mischel (1968).

Distal factors in risk perception 207111123456789101111234567892011112345678930111123456789401111234567181111from pursuing the matter further also by claims that lack of reliability and othermeasurement details, such as a limited number of scale steps, severely limit just howmuch we can increase the correlations. Yet, models have been devised that approach60% explained variance with these types of dependent variables (Sjöberg, 2000b). Suchmodels typically use more proximal variables, i.e. variables closer in content 6 but thatneed not at all imply semantic overlap, as alleged by Slovic and Peters.How small can a correlation be, and still be called ‘strong’? It is unclear if there isa lower limit different from zero, or perhaps nonsignificance. Anything achieving statisticalsignificance may well be accepted as very important in this school of thought. Thistakes us to the discussion of statistical significance which cannot be pursued here. I willonly point out that accumulation of a coherent and reasonably simple set of empiricallybased principles is virtually impossible when researchers take this relaxed attitude.The research fields will be more and more chaotic and complex as time goes by andnew hunches are found to yield small, albeit statistically significant, effects. The searchfor a few important and powerful factors should replace significance and neglect ofeffects size as ideals and habits of researchers.Slovic and Peters show the same kind of goals in their use of statistics as, for example,Rosenthal in defending clinical psychology (Rosenthal, 1995). It is implied that weshould be content with what we have. However, scientific progress is about not beingcontent with what we have but trying to improve on it. Statistics should give a truepicture of the state of the power of our models, and not be used to hide the fact thatmost of our job is still to be done.Tansey (in press) repeated the arguments made by Slovic and Peters. He alsocriticized my work for not doing full justice to Cultural Theory, since it assumes thatpersons can be characterized by a system of dimensions measuring the four basic ‘types’of the theory, called World Views by Slovic and Peters and others. However, theapproach I discussed has been used by several authors, including Dake and Wildavsky(1990) and Peters and Slovic (1996), and it gave rise to considerable interest andattempts at cross-cultural replications of Dake’s scales in the 1990s, see e.g. Brenot etal. (1998). It seemed to me to be an appropriate and even important task to study justhow well this approach functioned, and I found that it did not do very well. I personallyfind it unlikely that such general tendencies as postulated in this approach exist atall, or that they would have a major influence on risk perception if they did exist.In this sense I agree with Tansey. I think it would be more interesting to specify thetheory so as to define ‘types’ in a contextual way. This would make the operationalvariables less distal and more proximal, which is no drawback as long as there is nosemantic overlap, i.e. as long as empirical convergence is not brought about by variablesmeasuring identical concepts. It can be noted that the use of proximally relatedvariables abounds in social psychology, see e.g. Fishbein and Ajzen’s work on attitudesand planned behaviour (Ajzen, 1985, 1991).Marris et al. conclude a recent article on the Psychometric Model (Marris et al., 1997)by stating their belief that ‘a more thorough analysis’ of the psychometric data wouldyield better understanding of risk perception. This may sound reasonable, but is, inreality, very uncertain. It is more likely that new information is needed and that no6Compare the present paper’s use of specific trust vs. general trust. Specific trust is closer in content to perceivedrisk because it involves specifying the hazard in question, which is not done in general trust measures. Yet, thereis no semantic overlap. Semantic overlap occurs when the denotations of the concepts responded to are the same.

208 Sjöberg11112345678910111123456789201111234567893011112345678940111123456711181111statistical twisting and bending of the data, however sophisticated, will save the modelas long as powerful explanatory variables simply have not been measured. The sameargument can be made for Cultural Theory.6. ConclusionThe studies reported here have shown that it is hard to devise distal explanations ofrisk perception, but also that some variables show promise. The finding that emotionalstability has some relationship to risk perception is in accordance with some other workby Källmén and Sjöberg. ‘Macho’ risk taking propensity is a promising factor for understandingdemand for risk mitigation. The latter dimension has usually turned out to bequite hard to account for (Sjöberg, 1999a).The present findings also give clear support to the stability of New Age attitudes inaccounting for perceived risk. The New Age movement is contemporary with technologyopposition, it involves ontological and epistemological concerns leading toanti-science and anti-technology standpoints, and is hence a very reasonable distaldimension in accounting for the perceived risk of technology.All explanations call for further explanations. If New Age in part is responsible forperceived risk and technology opposition, what caused New Age? This interesting sociologicalquestion is discussed elsewhere (Sjöberg, 2002c).AcknowledgementsThis is a study within CEC project RISKPERCOM (Contract F14PCT950016),supported also by the Swedish Council for Planning and Coordination of Research(FRN), the Swedish Council for Humanistic and Social Science Research (HSFR), theSwedish Nuclear Power Inspectorate (SKI), and the Swedish Radiation ProtectionInstitute (SSI). Mattias Viklund analysed and reported the background data of therespondents. Gunnar Grendstad of the University of Bergen provided Cultural Theoryitems and was helpful in discussing the manuscript, but should not be held responsiblefor the views expressed herein. He also coordinated the Nordic project which providedthe data reported here as Study 3.ReferencesAdams, J. (1995) Risk, London: UCL Press.Ajzen, I. (1985) From intentions to actions: a theory of planned behavior, in J. Kuhl andJ. Beckmann (eds) Action control: from cognition to behavior, New York: Springer. pp. 11–39.Ajzen, I. (1991) The theory of planned behavior, Organizational Behavior and Human DecisionProcesses, 50, 179–211.American Psychological Association (1994) Publication manual of the American PsychologicalAssociation, 4th edn, Washington, DC: American Psychological Association.Barbuto, J. E., Jr. (1997) A critique of the Myers–Briggs Type Indicator and its operationalizationof Carl Jung’s psychological types, Psychological Reports, 80(2), 611–25.Boholm, Å. (1996) The cultural theory of risk: an anthropological critique, Ethnos, 61, 64–84.Boyle, G. J. (1995) Myers–Briggs Type Indicator (MBTI): some psychometric limitations,Australian Psychologist, 30, 71–74.Brenot, J., Bonnefous, S. and Marris, C. (1998) Testing the cultural theory of risk in France, RiskAnalysis, 181, 729–40.

Distal factors in risk perception 209111123456789101111234567892011112345678930111123456789401111234567181111Buss, D. M. and Craik, K. H. (1983) Contemporary worldviews: personal and policy implications,Journal of Applied Social Psychology, 131, 259–80.Cohen, J. (1988) Statistical Power Analysis for Behavioural Sciences, 2nd edn, Hillsdale, NJ:Erlbaum.Crowne, D. P. and Marlowe, D. (1960) A new scale of social desirability independent ofpsychopathology, Journal of Consulting and Clinical Psychology, 24, 349–54.Dake, K. (1990) Technology on trial: orienting dispositions toward environmental and healthhazards, dissertation, University of California, Berkeley.D’Andrade, R. and Dart, J. (1990) The interpretation of r versus r 2 or why per cent of varianceaccounted for is a poor measure of size of effect, Journal of Quantitative Anthropology, 2, 47–59.Dietz, T., Stern, P. C. and Guagnano, G. A. (1998) Social structural and social psychologicalbases of environmental concern, Environment and Behavior, 30, 450–71.Douglas, M. and Wildavsky, A. (1982) Risk and Culture, Berkeley, CA: University of CaliforniaPress.Dunlap, R. E. and van Liere, K. D. (1978) The ‘New Environmental Paradigm’: a proposedmeasuring instrument and preliminary results, Journal of Environmental Education, 9, 10–19.Dunlap, R. E., Liere, K. D. V., Mertig, A. G., Catton, W. R., Jr. and Howell, R. E. (1992)Measuring endorsement of an ecological worldview: a revised NEP scale, Annual Meeting,Rural Sociological Society. State College, PA.Dunlap, R. E., Van Liere, K. D., Mertig, A. G. and Jones, R. E. (2000) Measuring endorsementof the new ecological paradigm: a revised NEP scale, Journal of Social Issues, 56, 425–42.Eysenck, H. J. (1959) Manual of the Maudsley Personality Inventory, San Diego, CA: Educationaland Industrial Testing Service.Fischhoff, B., Slovic, P., Lichtenstein, S., Read, S. and Cambs, B. (1978) How safe is safe enough?A psychometric study of attitudes towards technological risks and benefits, Policy Sciences, 9,127–52.Fishbein, M. and Ajzen, I. (1975) Belief, Attitude, Intention, and Behavior: An Introduction toTheory and Research, Reading, MA: Addison-Wesley.Flynn, J., Slovic, P. and Mertz, C. K. (1994) Gender, race, and perception of environmental healthrisks, Risk Analysis, 14, 1101–08.Fritzsche, A. W. (1995) The role of the unconscious in the perception of risks, Risk: Health,Safety & Environment, 6, 15–40.Fritzsche, A. W. (1996) The moral dilemma in the social management of risk, Risk: Health, Safety& Environment, 7, 41–45.Funder, D. C. and Ozer, D. J. (1983) Behavior as a function of the situation, Journal of Personalityand Social Psychology, 44, 107–12.Goldberg, L. R. (1994) Basic research on personality structure: implications of the emergingconsensus for applications to selection and classification, in M. G. Rumsey, C. B. Walker andJ. H. Harris (eds) Personnel Selection and Classification, pp. 247–57, Hillsdale, NJ: Erlbaum.Grendstad, G., Jensen, L., Kristinsson, G.-H., Sjöberg, L. and Sundback, S. (1999) Nordic cultures:measurement, consequences and comparisons. Study description and frequencies. Bergen, LOSSenteret. University of Bergen.Helmreich, R. L., Foushee, H. C., Benson, R. and Russini, W. (1986) Cockpit resource management:exploring the attitude-performance measure, Aviation, Space, & EnvironmentalMedicine, 57, 1198–200.Hendrunks, A. A. J. (1997) The Construction of the Five-Factor Personality Inventory (FFPI),Groningen, University of Groningen, the Netherlands: University Press.Jung, C. G. (1971) Psychological Types, Princeton NJ: Princeton University Press.Källmén, H. (2000) Manifest anxiety, general self-efficacy and locus of control as determinantsof general and personal risk perception, Journal of Risk Research, 3, 111–20.Kline, N. S. (1962) Factifuging, The Lancet, 30 June, 1396–99.

210 Sjöberg11112345678910111123456789201111234567893011112345678940111123456711181111Lawton, R. and Parker, D. (1998) Individual differences in accident liability, Human Factors, 40,655–71.Lester, L. F. and Bombaci, D. H. (1984) The relationship between personality and irrationaljudgment in civil pilots, Human Factors, 26, 565–72.Lockart, C., Coughlin, R. M., Ellis, R. J. and Grendstad, G. (1997) Measuring and applyinggrid–group theory: an alternative to liberal–conservative ideology and Inglehart’s schema.Marris, C., Langford, I., Saunderson, T. and O’Riordan, T. (1997) Exploring the ‘psychometricparadigm’: comparisons between aggregate and individual analyses, Risk Analysis, 17, 303–12.Marris, C., Langford, I. H. and O’Riorddan, T. (1998) A quantitative test of the cultural theoryof risk perceptions: comparison with the psychometric paradigm, Risk Analysis, 18, 635–47.Mischel, W. (1968) Personality and Assessment, New York: Wiley.Mårdberg, B., Nienmaa, P., Hillström, B. and Carlstedt, B. (1994) Myers–Briggs Type Indicator.Manual, Stockholm. Psykologiförlaget.Peters, E. and Slovic, P. (1996) The role of affect and worldviews as orienting dispositions inthe perception and acceptance of nuclear power, Journal of Applied Social Psychology, 26,1427–53.Prentice, D. A. and Miller, D. T. (1992) When small effects are impressive, Psychological Bulletin,112, 160–64.Rosenthal, R. (1995) Progress in clinical psychology: is there any? Clinical Psychology: Scienceand Practice, 2(2), 133–50.Rosenthal, R. and Rubin, D. B. (1982) A simple, general purpose display of magnitude of experimentaleffect, Journal of Educational Psychology, 74, 166–69.Shafer, A. B. (1999) Relation of the big five and factor V subcomponents to social intelligenceEuropean Journal of Personality, 13, 225–40.Sjöberg, L. (1980) Similarity and correlation, in E.-D. Lantermann and H. Feger (eds) Similarityand choice, Bern: Huber.Sjöberg, L. (1997) Explaining risk perception: an empirical and quantitative evaluation of culturaltheory, Risk Decision and Policy, 2, 113–30.Sjöberg, L. (1998a) World views, political attitudes and risk perception, Risk: Health, Safety andEnvironment, 9, 137–52.Sjöberg, L. (1998b) Worry and risk perception, Risk Analysis, 18, 85–93.Sjöberg, L. (1999a) Consequences of perceived risk: demand for mitigation, Journal of RiskResearch, 2, 129–49.Sjöberg, L. (1999b) Policy implications of risk perception research: a case of the emperor’s newclothes? Proceedings of the 1998 Annual Conference. ‘Risk analysis: opening the Process’, P.Huberg and C. Mays (eds), pp. 77–86. Paris: IPSN.Sjöberg, L. (2000a) Consequences matter, ‘risk’ is marginal, Journal of Risk Research, 3(3),287–95.Sjöberg, L. (2000b) Factors in risk perception, Risk Analysis, 20, 1–11.Sjöberg, L. (2000c) Perceived risk and tampering with nature, Journal of Risk Research, 3, 353–67.Sjöberg, L. (2001a) Emotional intelligence: a psychometric analysis, European Psychologist, 6(2),79–95.Sjöberg, L. (2001b) Political decisions and public risk perception, Reliability Engineering andSystems Safety, 72, 115–24.Sjöberg, L. (2002a) Are received risk perception models alive and well? Risk Analysis, 22, 665–70.Sjöberg, L. (2002b) Attitudes to technology and risk: Going beyond what is immediately given,Policy Sciences, 35, 379–400.Sjöberg, L. (2002c) New Age beliefs and risk perception, Risk Analysis, 22, 751–64.Sjöberg, L. (in press) The psychometric paradigm revisited, in M. Bertilsson, P. Lindquist andL. Sjöberg (eds) Risk, Risk Research and Scientific Controversies. Stockholm: Stockholm Schoolof Economics.

Distal factors in risk perception 211111123456789101111234567892011112345678930111123456789401111234567181111Sjöberg, L. and Drottz-Sjöberg, B.-M. (2001) Fairness, risk and risk tolerance in the siting of anuclear waste repository, Journal of Risk Research, 4(1), 75–102.Slovic, P. and Peters, E. (1998) The importance of worldviews in risk perception, Risk Decisionand Policy, 3(2), 165–70.Streufert, S. (1986) Individual differences in risk taking, Journal of Applied Social Psychology,16, 482–97.Tansey, J. (in press) Risk as politics, culture as power, Journal of Risk Research.Tansey, J. and O’Riordan, T. (1999) Cultural theory and risk: a review, Health, Risk & Society,1, 71–90.Thompson, K. N. and Schumacker, R. (1997) An evaluation of Rosenthal and Rubin’s binomialeffect size display, Journal of Educational and Behavioral Statistics, 22, 109–17.Wiggins, J. (1989) Review of the Myers–Briggs Type Indicator, Mental Measurements Yearbook,10, 537–38.Wiggins, J. S. and Trapnell, P. D. (1997) Personality structure: the return of the Big Five, in R.Hogan, J. Johnson and S. Briggs (eds) Handbook of personality psychology, pp. 737–66, SanDiego, CA: Academic Press.Wildavsky, A. and Dake, K. (1990) Theories of risk perception: who fears what and why?Daedalus, 119(4), 41–60.Vollrath, M., Knoch, D. and Cassano, L. (1999) Personality, risky health behaviour, and perceivedsusceptibility to health risks, European Journal of Personality, 13, 39–50.AppendixHigher consciousness factor1. The task of man is to attain a higher state of consciousness2. If you search for your inner being you will live in the light3. Humanity is developing a higher consciousness4. You must be aware of and work with your body’s energy flows5. We all have a spark of divinity inside us6. If you have the right thoughts they will make the world positive, to conformto your desires7. You get true knowledge by turning inward, towards yourself8. The soul continues to exist though the body may die9. There is a meaning in all that happens10. I eat health food11. We all have a pre-determined fate12. You should trust your intuition above all else13. We are facing a new era: the Age of Aquarius