Introduction

Social trust is frequently discussed in risk management and with regard to related policy issues (Slovic et al., 1991a; Poortinga and Pidgeon, 2003; Siegrist et al., 2003). It is a precondition for successful communication, probably a necessary condition. But is it also sufficient? In the present paper, additional constructs are proposed, which may play an important role in risk management. They are compared empirically with social trust.Social trust is of interest both for its antecedents or components, and for its consequences such as perceived risk. Both of these lines of research on trust are discussed.

Components of trust

Trust components have been studied by several researchers. Peters et al. (1997) summarized work on the topic and tested a model of trust based on three hypothesized determinants: knowledge/expertise, openness/honesty, and concern/care. In three different sets of data, measuring trust in industry, the Government, and citizen groups, they found that the three factors accounted for 20–40% of the variance of trust. Each of the three categories was operationalized with one concrete example, for example chemical industry in the case of industry. Earlier suggestions about the components of trust (Renn and Levine, 1991; Kasperson et al., 1992) were similar but had not been empirically tested.

A simpler trust structure was suggested by Metlay (1999). He found two factors: competence and affect. They accounted for 50–60% of trust in the US Department of Energy, in two data sets. Poortinga and Pidgeon (2003) argued that there are two components of trust: general trustworthiness, comprising such factors as fairness and competence, and a second factor, which could be called vested interest (Frewer et al., 1996) or (lack of) integrity. They found support for the second factor and labeled it cynicism.

The work by Peters, Covello, and McCallum and by Poortinga and Pidgeon were important contributions to the understanding of the dynamics of trust. However, the explanatory power of their models was not so high as to exclude the consideration of other factors and different models. One such factor, related to cynicism but still distinct from it, is antagonism. The reasoning behind considering it as a factor in trust was as follows.

The conceptualization of trust by Peters et al. is closely related to attitude models (Fishbein and Ajzen, 1975), but it lacks the value component. In expectation models of attitude, the global attitude is broken down into components of belief and value. Value can be negative or positive, and would be expected to vary across individuals and attitude objects. Viklund and Sjöberg (in press) found good fit to such an expectation model of trust with an interesting exception: advertising agencies.

Why should the dynamics of trust be different for advertising agencies, compared to, for example the Government or various authorities? The reason was suggested by Viklund and Sjöberg to be the presence of beliefs about antagonistic relationships. Some people may see advertising agencies as working mainly in their own or their clients' interest, not for the consumers or the public. If they succeed, the consumer fails. In other words, the more competent advertising is, the worse for the consumer.

Risk communication experts have emphasized the importance of hostile feelings in some situations, as testified by the following quotation from a discussion of problematic situations in risk communication: Some people in the community may have decided that you are the enemy; that the company is uncaring, arrogant, and dishonest; or that chemical manufacturing is an industry that does society little benefit at great cost. (Covello et al., 1988, p 34) Beliefs in hostile or antagonistic intentions make things even worse than just a lack of trust. Such beliefs are probably common, to judge from the intense debates on risk issues such as nuclear power or genetically modified food. What are the interests of the nuclear industry, for example? Do they coincide with the interests of the public? Views vary with regard to this question. Many would answer yes, but some would be equally sure of the opposite. Therefore, perceived antagonism should be measured. Such measures may improve on models of trust.

Earle and Cvetkovich (1995) suggested a trust model in which perceived value similarity is hypothesized to be a major determinant of trust. Later work has provided some empirical support for the model (Poortinga and Pidgeon, 2003, 2005, 2006). The notion that value similarity is a factor in trust does not preempt the hypothesis about antagonism proposed here. People may well perceive value similarity with an antagonist; still not trust him or her or the institution, which is believed to work against one's interests.

The reasoning above motivates a study of antagonism as a possible (negative) component of trust. The following hypothesis was formulated:

(1) Perceived antagonism is an important (negative) determinant of trust.

In an antagonistic relationship, the standard trust model breaks down. What is good in a normal cooperative relationship – competence – becomes less positive or even bad in an adversarial relationship. The competence–trust relationship should therefore be moderated by the antagonism, leading to the following hypothesis:

(2) Competence is a positive factor in trust only in an actor perceived as benevolent. In an actor perceived as adversarial or hostile, competence is a negative factor in trust.

People tend to rate trust at different levels, regardless of for which actor or organization they are rating their trust. The average of trust ratings across organizations, here called Average Specific Trust, is a possible explanatory factor for any given trust rating. A third hypothesis was therefore formulated:

(3) The individual level of trust ratings, computed as the average across actors or organizations, accounts for a share of the perceived trust of any rated actor or organization.

Finally, the issue of specific vs general measures of trust was considered (Sjöberg, 1999). Sjöberg suggested a measure of generalized trust in people, organizations, politicians, and common societal value systems. This is an alternative approach to that of Average Specific Trust ratings and it should explain part of the variation of trust in any actor or organization. Hence the following hypothesis:

(4) Generalized trust contributes significantly to the variance of trust in any one actor or organization.

Trust and risk perception

So far, the discussion has focused on social trust and its components. Much interest has also been oriented towards the question of social trust as a determinant of perceived risk. Research on trust and risk perception continues to be dominated by social trust, see, for example Cerully et al. (2006), Bakir (2006), and Poortinga and Pidgeon (2006). According to Bakir (2006), "trust shapes public risk perception" (p 4). Is there a solid empirical basis for such an assertion?

The published data with some bearing on this issue have documented the expected negative correlations between perceived risk and social trust. However, the correlations have varied widely, from weak to strong (Sjöberg, 1999, 2001). Siegrist and Cvetkovich (2000) found correlations between -0.56 and 0.10 between perceived risk and trust in authorities. The correlations were statistically significant for 16 of 25 activities. The average correlation was -0.27. The highest negative correlations were found for activities where respondents rated that they had little knowledge. Siegrist et al. studied perceived risk for mobile phone technologies (Siegrist et al., 2005), and found correlations between -0.31 and -0.13. None of these results suggests that social trust is a strong determinant of perceived risk.

Siegrist et al. (2000) presented several SEM models of perceived risk and trust. They found the trust–risk relationship was statistically significant. The strength of relationship was quite high. For example, for nuclear power trust explained 71% of the variance of perceived risk, corresponding to a correlation of 0.84. This is much higher than other published results. The reason for this deviating finding is probably that both risk and trust were measured with similar Likert-type attitude items, not by ratings of perceived risk and trust on separate scales. Overlapping semantics can easily give rise to such very high correlations when similar scales are used. Sjöberg (2004b) reported similar findings with Likert-type items for measuring risk and trust.

Since social trust accounts for only part of the variance of perceived risk, there must be other determinants as well. Several factors will be proposed here:

• antagonism,
• individual differences in risk sensitivity,
• specific trust,
• individual differences in generalized trust.

Antagonism: Antagonism is a likely candidate for explaining a share of the variance of perceived risk. If there is an agent responsible for risk management, which is seen as indifferent, or even hostile, to the well-being of the public it is reasonable to expect perceived risk to increase. Such a phenomenon has been documented in the case of the perceived risk of nuclear waste (Sjöberg and Wester-Herber, in press) and for perceived risk of terrorism (Sjöberg, 2005). The latter case is extreme, but milder forms of perceived antagonism may well have a similar effect. The following hypothesis was formulated:

(5) Perceived antagonism is positively related to perceived risk.

Individual differences in risk sensitivity: Risk sensitivity, measured as the average risk rating given by respondents to several hazards different from the one, which is being modeled, has been found to be an important factor in perceived risk (Sjöberg, 2004a). People use the risk rating scale in different ways. Some rate risks as low throughout, others rate them as high. Whatever the psychological reason is for such variation among individuals, it is likely to be an important factor in accounting for the level of perceived risk of any given hazard. It should therefore be included in a model of perceived risk. A sixth hypothesis was formulated:

(6) The individual level of risk ratings account for a share of the perceived risk associated with any rated actor or organization.

Individual differences in specific trust: Specific Trust has been claimed to be a major factor in perceived risk, as pointed out in the introduction. The following hypothesis was therefore formulated:

(7) Specific trust ratings account for a share of the perceived risk of any rated actor or organization.

Individual differences in generalized trust were finally considered as a factor also in perceived risk. There is some previous work on this issue. Sjöberg found a relationship between perceived risk and generalized trust (Sjöberg, 1999), and so did Siegrist et al. (2005). Viklund reported that generalized trust was more powerful in explaining perceived risk than was specific trust, that is trust semantically linked to the risk measure used as dependent variable in the model (Viklund, 2002, 2003). Others have found that general trust measures are less efficient than specific trust measures in accounting for perceived risk (Siegrist et al., 2000). It was considered to be of interest to pursue the matter further in the present context, and a fourth hypothesis was therefore formulated:

(8) Generalized trust contributes significantly to the variance of perceived risk.

Summing up, the first four hypotheses pertain to models of trust, the last four to models of perceived risk. The hypotheses are tested on two data sets. In Study 1, a large number of actors or organizations are studied. In Study 2, the perceived risk of spent nuclear fuel is studied in detail. Study 2 also introduces yet another aspect of trust (epistemic trust), to be further explained in the introduction to that study.

Study 1

It was important to include a rather large number of organizations or actors (in the following called actors) since the dynamics of trust may well be different for different actors. The notion of studying antagonism arose in the context of studying a type of commercial actor not usually investigated in risk-oriented trust research (advertising agencies). The possibility to generalize the findings is larger when several actors are studied and the design is not constrained to only a few.

Method

Questionnaire

The respondents were asked to judge 23 actors¹ for several dimensions. The actors to be judged were:

• Banks
• Advertising agencies
• Auto repair shops
• Quality newspapers
• Tabloids
• State television
• Police
• Courts of law
• Medical care
• Universities
• Labor unions
• Grammar school
• Local politicians
• Used car salespersons
• SAS airline
• Drug companies
• Stockbrokers
• Capital owners
• Insurance companies
• National government
• Coop
• National organization of local authorities
• National defense.

Each actor was rated on the following scales:

• Trust, from complete trust to very small trust or none at all.
• Egotism, from working entirely in the interests of the public to working only for their own interests.
• Efficiency, from very efficient in reaching their goals to very inefficient.
• Honesty, from very honest to very dishonest.
• Antagonism, from their being entirely in accordance with the respondent's interests, to being entirely antagonistic.
• Power, from very large power to almost no power.
• Risk of negative impact on the people and society, from none to very large.

The respondents were finally asked to judge 42 Likert-type attitude scale items measuring generalized trust, see Sjöberg (1999). The questionnaire is available on a website.²

Participants

A sample of people, 189 in all, from the general population, was approached with the mailed questionnaire. They had previously responded to another questionnaire and then marked that they were willing to take part in another study. The number of returned questionnaires was 143, a response rate of 76%. Previous work with this type of sample has supported that they do not differ systematically from ordinary respondents, selected at random from the general population (Viklund, 2002).

Two respondents did not state their gender, 68 were men and 73 women. Their median age was 50 years. Thirty percent had college education (14% in the population). This is a bias, which is usually present in survey work and probably not important enough to make the results unrepresentative.

Results

Construction of trust indices

Trust indices were constructed in order to measure basic sources of interindividual variability. Each respondent rated his or her trust in each of the 23 actors. A pooled rating provided a measure of general trust. Its alpha value was 0.89. For modeling the trust in each actor, the index had to be corrected by deleting the trust rating for that particular actor, yielding 23 slightly different measures of Average Specific Trust.

The 44 Likert-type items were pooled to an overall measure of general trust, called Generalized Trust in the following. Its alpha value was 0.93. Average Specific Trust and Generalized Trust correlated 0.44 (P<0.0005).

Average Specific Trust was expected to be a powerful explanatory factor since people differ in their tendencies to rate trust as high or low – such a tendency is manifest in the ratings of each of the actors.

Perceived antagonism as a (negative) component of trust

Analysis at the aggregated level  Psychometric research on risk perception has frequently been carried out at the level of aggregated ratings, usually arithmetic means. Variables have been related across hazards. Such an analysis gives a misleading impression of the strength of relationship at the individual data level but the trends are probably not different. Mean intraindividual correlations among variables have been found to be linearly related to the correlations between means, only smaller (Sjöberg, 2002a).

For the analysis at the aggregated level, mean ratings of each of the 23 actors were computed on each variable. Intercorrelations among the means are given in Table 1.

Table 1 - Correlations among mean ratings.

Full table

The table shows that antagonism had the strongest relationship with trust, r=-0.90, supporting Hypothesis (1). Figure 1 shows the relationship and gives the coordinates of all actors with regard to trust and antagonism.

Figure 1.

Mean ratings of trust plotted against mean ratings of antagonism for 23 actors.

Full figure and legend (29K)

Analysis at the disaggregated level  A series of hierarchical regression analyses was performed, one for trust in each organization/actor.

The following explanatory variables were used:

• Demographics (Gender, age and educational level)
• Average Specific Trust
• Generalized Trust
• Egotism
• Efficiency
• Honesty
• Power
• Antagonism.

Antagonism was added last. The results of all 23 regression analyses are summarized in Table 2.

Table 2 - Trust models.

Full table

The mean overall explained variance was 54.1%. Antagonism added, on the average, 3.5% explained variance to the other nine explanatory constructs. The added power due to antagonism was statistically significant in 18 of the 23 cases, at least at the 0.05 level, usually highly significant.

Average Specific Trust was an important explanatory factor, a finding that supports Hypothesis 3. Antagonism was approximately equally important, supporting Hypothesis (1). Honesty and egotism were also important in accounting for trust, but less so than antagonism. Generalized Trust added very little to the power of the regression models, rejecting Hypothesis 4.

Moderator analysis

To test Hypothesis 2, the respondents were divided with regard to whether they had given ratings of antagonism below or above the mean rating, for all 23 actors combined. Correlations between trust and efficiency (competence) are given for the two groups of respondents, and all 23 actors, in Table 3.

Table 3 - Correlations between trust and efficiency for respondents rating a high or low level of perceived antagonism.

Full table

The correlation was larger in the low antagonism group in 20 of 23 cases. The mean correlation was 0.273 for the former group, 0.069 for the latter. In seven of the 23 analyses, there was a significant difference between the two correlations in the expected direction at the 0.05 level or higher. Hypothesis 2 was therefore supported to rome extent.

Models of perceived risk

Regression analyses were performed for the perceived risk associated with each of the 23 actors. The outcomes of the analyses are provided in Table 4.

Table 4 - Models of perceived risk.

Full table

The mean overall explanatory power was moderately high; R²_adj=0.364. The table shows that risk sensitivity was the most important explanatory factor, mean beta=0.476. It was a significant explanatory variable in all 23 analyses. Antagonism was second in importance with a mean beta coefficient of 0.201. It was significant in 15 of the 23 analyses and added on the average 0.030 to the proportion of explained variance of perceived risk. It added significantly in 16 of the 23 cases. Specific Trust was considerably weaker than antagonism as a determinant of perceived risk, mean beta=0.114, significant in seven of the 23 cases. It can also be noted that Generalized Trust contributed very little to the power of the models, as did demographics.

Hypothesis 5 was supported by the results. In addition, Hypothesis 6 was supported by the findings that risk sensitivity was an important explanatory factor in the models of perceived risk. Hypothesis 7 was only weakly supported and Hypothesis 8 not at all.

Conclusions

Trust was well explained by the models. On average, 58 percent of the variance was accounted for. This is considerably higher than the models investigated by Peters et al. (1997). Part of the reason for the improved fit was probably the inclusion of antagonism among the explanatory variables, as well as the average specific trust rating. When the latter two variables were excluded from the trust models, the explanatory power decreased to 46 percent, on average, which is more in line with previous research. Antagonism emerged as an important factor in accounting for both social trust and perceived risk. Risk sensitivity was an important factor in models of perceived risk. Average Specific Trust was important in trust models. Generalized Trust carried little weight in the models.

It was also found that antagonism, as expected, functioned as a moderator of the relationship between trust and competence/efficiency. A high level of perceived antagonism tended to be associated with a weak or negative relationship between trust and competence.

Study 2

The results of Study 1 suggest that perceived risk is partially determined by social trust and that antagonism enters as a powerful additional factor. However, only a rather modest 36% of the variance of perceived risk was accounted for by the explanatory variables, on the average. What other factors could be of importance in risk perception? Sjöberg found that epistemic trust, or trust in the science on which the risk analysis of a technology is based (Sjöberg, 2001, 2004b; Sjöberg and Wester-Herber, in press), is an important factor in risk perception. Consequently, it was hypothesized that

(9) Epistemic trust is an important factor in perceived risk.

The application investigated in the present study is that of siting a repository for spent nuclear fuel, a controversial issue in many countries. Extensive research on nuclear waste risk perceptions and attitudes has been carried out in the USA and testifies to the difficulties in managing repository siting (Slovic et al., 1991a, 1991b and 1991c; Dunlap et al., 1993). Sweden has been among the few successful countries when it comes to managing the search for a nuclear waste repository site (Sjöberg, 2004c; Dawson and Darst, 2006).

Model

Perceived risk is a central variable in much of the discussion of risk management and risk communication (Slovic, 1987, 2000; Renn, 2004). It is, however, desirable to broaden the scope of study beyond perceived risk because it has been found to be weakly related to such dimensions as demand for risk mitigation (Sjöberg, 2000, 2002b). In the context of siting a nuclear waste repository, a variable of direct policy relevance is the intention to vote for or against a siting proposal. Intention to vote has been found, in many studies, to be a good predictor of voting (Pieters and Verplanken, 1995; Visser et al., 1996), even with a long time interval before the election takes place (Randall and Wolff, 1994). Voting is a way of expressing acceptance of a facility. A number of related items measuring acceptance were also included in the design of the present study.

The purpose of Study 2 was to investigate the relations between perceived risk and acceptance on the one hand, trust (both social and epistemic), and antagonism on the other. These relations were investigated by formulating and estimating SEM models.

The relationship between social and epistemic trust was specified as follows. Social trust was assumed a determinant of epistemic trust. Further, social trust was expected to have no independent effect on perceived risk or acceptance. Its effect was assumed to be mediated by epistemic trust. The relations between the two types of trust and risk perception could be construed in other ways. It is well known, however, that the issue of disposal of high-level radioactive waste is considered by the public to be very important (Sjöberg, 2004c). It is therefore to be expected that communication on the issue tends to follow a "central" route, to use the terminology and model of Petty and Cacioppo (1986). In that case, people will listen to and actively evaluate the contents of a message rather than be influenced by peripheral cues connected to the source of the message.

Note, however, that epistemic trust probably also has other determinants. It is not assumed to be fully determined by social trust. Antagonism is assumed one such additional determinant of epistemic trust. Antagonism should also affect social trust directly, as well as perceived risk and acceptance.

Risk sensitivity should be an independent determinant of perceived risk. It was less obvious that it should be a factor in acceptance. The proposed model is illustrated in Figure 2 for the case of perceived risk of nuclear waste. The model for acceptance of a local nuclear waste repository is similar (Figure 4).

Figure 2.

Proposed model of perceived risk.

Full figure and legend (38K)

Method

Respondents

A random sample of 1,000 people living in Sweden was approached with a mailed questionnaire.³ After two reminders, 508 respondents had returned filled out questionnaires. The response rate was 52.6%, considering that some people had moved without giving a forwarding address to the post office, and that some were unable to answer because of illness or old age. The demographic properties of the respondents resembled closely those of the population at large, except for of their having a higher level of education than people have in general. Among the respondents, 24% had a college degree as compared to 13% in the population at large. For the age distributions of the respondents and the population, see Table 5. There were somewhat too few respondents in the youngest and oldest groups. As to gender, 51.8% of the respondents were women, compared with 49.6% in the population.

Table 5 - Age distribution among the respondents and the population according to official data from Statistics Sweden.

Full table

Questionnaire

The questionnaire was extensive and only part of it will be used in the present context. The entire questionnaire is posted on the Internet (see footnote 2). The following data were used for the analyses presented here:

Demographic variables:

• Gender
• Age
• Level of education

Dependent variables:

• Perceived risk of a local nuclear waste repository, personal and general. Personal risk refers to the risk as perceived by the respondent to him- or herself. General risk is the risk as perceived to others.
• Acceptance measured six items, among them the stated intention to vote in a future local referendum about siting a repository, from "surely pro" to "surely con". In addition, there were five more items measuring various aspects of acceptance, see Appendix.

Nineteen items were used to measure the explanatory variables. They were:

• Social trust, rated in five steps from "very large" to "very small", with regard to the Swedish Nuclear Fuel and Waste Management Co (SKB), the nuclear power inspectorate, the radiation protection institute and the respondent's municipality. The ratings were made of trust in how well these organizations handled questions regarding the management of spent nuclear fuel.
• Epistemic trust, measured by four items. Examples, in condensed form are: "How well does science know about risks of spent nuclear fuel"? "Do you consider the technical and scientific questions about spent nuclear fuel to be solved?", and "Is current scientific knowledge sufficient for building a repository for spent nuclear fuel?"
• Antagonism, measured by ratings of four organizations: the radiation protection institute, the Swedish Nuclear Fuel and Waste Management Co (SKB), the nuclear power inspectorate, and the nuclear industry. They were rated with regard to whether they acted in accordance with the respondent's interests or against them, on a 7-step Likert scale.
• Risk sensitivity, measured by ratings of the general risk of seven hazards, excluding nuclear waste.

The formulations of all items used in Study 2 are available in the Appendix.

Results

The same model was estimated for both the perceived risk and the acceptance of a local nuclear waste repository, see Figures 3 and 4. The model analyses were done by means of the LISREL program, version 8.80. In preliminary analyses, a direct link from social trust to perceived risk and acceptance was included in the models, but the resulting weight parameter was close to zero and non-significant.

Figure 3.

Estimated model of perceived risk.

Full figure and legend (47K)

Figure 4.

Estimated model of acceptance of a nuclear waste repository.

Full figure and legend (46K)

To make the figures more readable, they do not include the weight coefficients of the manifest indicators of each of the latent constructs. The chi-squares indicate that there is a statistically significant lack of fit. For the model in Figure 3, =.00, P<0.005. For the Figure 4 model ² (266)=427.14, P<0.005.⁴

All weight parameters in Figures 3 and 4 are significant at the 0.10 level, most of them at the 0.05 level. In both models, the effect of social trust was mediated by epistemic trust, which in turn was the most important trust dimension. Antagonism played its most important role in accounting for a sizable share of social trust. In addition, perceived risk, acceptance, and epistemic and social trust were all explained to some extent by risk sensitivity.

General discussion

Eight hypotheses were proposed in the introduction to Study 1, and an additional one in Study 2. Five of them were strongly supported by the results of Study 1, and one by Study 2, viz.

• Perceived antagonism was an important (negative) determinant of trust (Hypothesis 1).
• Antagonism moderated the relationship between competence and trust (Hypothesis 2).
• Perceived antagonism was positively related to perceived risk (Hypothesis 5).
• The overall level of risk or trust ratings accounted for a share of the perceived risk or trust of any rated actor or organization (Hypotheses 3 and 6).
• Epistemic trust was important explanatory factor in perceived risk (Hypothesis 9), more so than social trust.

Generalized Trust contributed little to models of perceived risk and specific trust. Hence, Hypotheses 4 and 8 were rejected. Specific Trust did not account noticeably to the variance of perceived risk, rejecting Hypothesis 7.

The regression models (Study 1) showed antagonism to be a more important determinant of perceived risk than social trust, which only gave a minor contribution to the explanatory power of the models.

In Study 2, a further hypothesis was tested, viz. that epistemic trust is a determinant of perceived risk. It was strongly supported. Social trust also had an effect, but it was mediated by epistemic trust. Epistemic trust, on the other hand, was only partially explained by social trust. It was also affected by risk sensitivity and antagonism, and it was accounted for, in the models, at a high level.

Hence, epistemic trust is supported as an important factor in perceived risk and repository acceptance by the present results. It should be noted, that acceptance was reasonably well accounted for in the present model (Figure 4).

Social trust seemed to play a rather small role in risk perception and the acceptance of hazardous technologies and facilities.

Both epistemic risk and antagonism call for different approaches to communication than the case where only social trust is considered. Epistemic concerns are about the validity of the scientific basis for risk regulation. The concerns could take many forms, including "New Age" beliefs about the nature of the world and how knowledge can be gained (Sjöberg and af Wåhlberg, 2002). One of the factors found by Sjöberg and af Wåhlberg measured denial of the validity of science.

Science is never 100% certain and it is necessary to deal openly with scientific uncertainty (Short and Rosa, 2004). When there is uncertainty of risk estimates, people probably want to be informed about it (Frewer et al., 2002). Meeting epistemic concerns requires a demonstration of the validity of scientific theories and research – a challenging task in risk communication.

The precautionary principle stresses the importance of prudence in cases of scientific uncertainty (Klinke et al., 2006). The effect on public attitudes of the implementation of such a principle may lead to a lower rather than a higher trust and acceptance, however (Wiedemann et al., 2006). Perceived risk may be amplified (Johnson and Slovic, 1995, 1998; Johnson, 2003). It is clearly quite difficult to communicate about scientific uncertainties – yet it is necessary (Breakwell, 2000).

Epistemic trust should be distinguished from trust in experts, which is a form of social trust. Different types of social trust exist, depending on who is the target. Trust in accumulated systematic knowledge, science, is something different. People may feel that experts are honest and competent, still not trust their message because they argue that science has not yet developed to the point that sufficiently reliable knowledge is available about an issue.

Communication presumes mutual understanding (Sjöberg et al., 2000). If it is assumed that the public just needs to be more trusting in the people and organizations responsible for risk management, risk communication may fail. The present data underscore the very real possibility that people question science itself, and that they construe of mutually antagonistic relationships, a situation where the goals are not in common but opposite to each other. The modest success of risk communication (Fischhoff, 1995; McComas, 2006), let alone the outright failures (Freudenburg, 2004), may be due, among other things, to too much trust in social trust (Sjöberg and Wester-Herber, in press).

Notes

¹ Most of the actors were not primarily risk management or regulatory organizations which made it necessary to formulate instructions and trust components somewhat differently than was done in some previous work, for example the study published by Peters et al. (1997).

² http://www.dynam-it.com/institute/.

³ For the full project report, see Sjöberg (2006).

⁴ The outcome of the chi-square tests should not be a source of concern since the sample was large and the measures of absolute fit are at a satisfactory level.

References

1. Bakir, V. (2006). Policy Agenda Setting and Risk Communication – Greenpeace, Shell, and Issues of Trust. Harvard International Journal of Press-Politics. Vol. 11, No. 3, pp 67–88. | Article |
2. Breakwell, G.W. (2000). Risk Communication: Factors Affecting Impact. British Medical Bulletin. Vol. 56, No. 1, pp 110–120. | Article | PubMed | ChemPort |
3. Cerully, J.L., Klein, W.M.P. and McCaul, K.D. (2006). Lack of Acknowledgment of Fruit and Vegetable Recommendations among Nonadherent Individuals: Associations with Information Processing and Cancer Cognitions. Journal of Health Communication. Vol. 11, Supplement 1, pp 103–115. | Article | PubMed |
4. Covello, V.T., Sandman, P.M. and Slovic, P. (1988). Risk Communication, Risk Statistics, and Risk Comparisons: A Manual for Plant Managers. Washington, DC: Chemical Manufacturers Association.
5. Dawson, J.I. and Darst, R.G. (2006). Meeting the Challenge of Permanent Nuclear Waste Disposal in an Expanding Europe: Transparency, Trust and Democracy. Environmental Politics. Vol. 15, No. 4, pp 610–627. | Article |
6. Dunlap, R.E., Kraft, M.E. and Rosa, E.A. (eds) (1993). Public Reactions to Nuclear Waste. Durham: Duke University Press.
7. Earle, T.C. and Cvetkovich, G.T. (1995). Trust: Toward a Cosmopolitan Society. Westport, CT: Praeger.
8. Fischhoff, B. (1995). Risk Perception and Communication Unplugged: Twenty Years of Process. Risk Analysis. Vol. 15, No. 2, pp 137–145. | Article | PubMed | ChemPort |
9. Fishbein, M. and Ajzen, I. (1975). Belief, Attitude, Intention, and Behavior: An Introduction to Theory and Research. Reading, MA: Addison-Wesley.
10. Freudenburg, W.R. (2004). Can We Learn from Failure? Examining us Experiences with Nuclear Repository Siting. Journal of Risk Research. Vol. 7, No. 2, pp 153–169. | Article |
11. Frewer, L.J., Howard, C., Hedderley, D. and Shepherd, R. (1996). What Determines Trust in Information About Food-Related Risk? Risk Analysis. Vol. 16, No. 4, pp 473–486. | Article | PubMed | ChemPort |
12. Frewer, L.J., Miles, S., Brennan, M., Kuznesof, S., Ness, M. and Ritson, C. (2002). Public Preferences for Informed Choice under Conditions of Risk Uncertainty. Public Understanding of Science. Vol. 11, No. 4, pp 363–372. | Article |
13. Johnson, B.B. (2003). Further Notes on Public Response to Uncertainty in Risks and Science. Risk Analysis. Vol. 23, No. 4, pp 781–789. | Article | PubMed |
14. Johnson, B.B. and Slovic, P. (1995). Presenting Uncertainty in Health Risk Assessment: Initial Studies of Its Effects on Risk Perception and Trust. Risk Analysis. Vol. 15, No. 4, pp 485–494. | Article | PubMed | ChemPort |
15. Johnson, B.B. and Slovic, P. (1998). Lay Views on Uncertainty in Environmental Health Risk Assessment. Journal of Risk Research. Vol. 1, No. 4, pp 261–279. | Article |
16. Kasperson, R.E., Golding, D. and Tuler, S. (1992). Social Distrust as a Factor in Siting Hazardous Facilities and Communicating Risks. Journal of Social Issues. Vol. 48, No. 4, pp 161–187.
17. Klinke, A., Dreyer, M., Renn, O., Stirling, A. and Van Zwanenberg, P. (2006). Precautionary Risk Regulation in European Governance. Journal of Risk Research. Vol. 9, No. 4, pp 373–392. | Article |
18. McComas, K.A. (2006). Defining Moments in Risk Communication Research: 1996–2005. Journal of Health Communication. Vol. 11, No. 1, pp 75–91. | Article | PubMed |
19. Metlay, D. (1999). Institutional Trust and Confidence: A Journey into a Conceptual Quagmire. In Cvetkovich, G. and Löfstedt, R.E. (eds) Social Trust and the Management of Risk. London: Earthscan, pp 100–116.
20. Peters, R.G., Covello, V.T. and McCallum, D.B. (1997). The Determinants of Trust and Credibility in Environmental Risk Communication: An Empirical Study. Risk Analysis. Vol. 17, No. 1, pp 43–54. | Article | PubMed | ChemPort |
21. Petty, R.E. and Cacioppo, J.T. (1986). Communication and Persuasion: Central and Peripheral Routes to Attitude Change. New York: Springer-Verlag.
22. Pieters, R.G.M. and Verplanken, B. (1995). Intention Behavior Consistency – Effects of Consideration Set Size, Involvement and Need for Cognition. European Journal of Social Psychology. Vol. 25, No. 5, pp 531–543. | Article |
23. Poortinga, W. and Pidgeon, N.F. (2003). Exploring the Dimensionality of Trust in Risk Regulation. Risk Analysis. Vol. 23, No. 5, pp 961–972. | Article | PubMed |
24. Poortinga, W. and Pidgeon, N.F. (2005). Trust in Risk Regulation: Cause or Consequence of the Acceptability of Gm Food? Risk Analysis. Vol. 25, No. 1, pp 199–209. | Article | PubMed |
25. Poortinga, W. and Pidgeon, N.F. (2006). Prior Attitudes, Salient Value Similarity, and Dimensionality: Toward an Integrative Model of Trust in Risk Regulation. Journal of Applied Social Psychology. Vol. 36, No. 7, pp 1674–1700. | Article |
26. Randall, D.M. and Wolff, J.A. (1994). The Time-Interval in the Intention Behavior Relationship – Metaanalysis. British Journal of Social Psychology. Vol. 33, No. 4, pp 405–418.
27. Renn, O. (2004). Perception of Risks. Toxicology Letters. Vol. 149, No. 1–3, pp 405–413. | Article | PubMed | ChemPort |
28. Renn, O. and Levine, D. (1991). Credibility and Trust in Risk Communication. In Kasperson, R.E. and Stallen, P.J.M. (eds) Communicating Risks to the Public. Dordrecht: Kluwer, pp 175–218.
29. Short, J.F. and Rosa, E.A. (2004). Some Principles for Siting Controversy Decisions: Lessons from the Us Experience with High Level Nuclear Waste. Journal of Risk Research. Vol. 7, No. 2, pp 135–152. | Article |
30. Siegrist, M. and Cvetkovich, G. (2000). Perception of Hazards: The Role of Social Trust and Knowledge. Risk Analysis. Vol. 20, No. 5, pp 713–719. | Article | PubMed |
31. Siegrist, M., Cvetkovich, G. and Roth, C. (2000). Salient Value Similarity, Social Trust, and Risk/Benefit Perception. Risk Analysis. Vol. 20, No. 3, pp 353–362. | Article | PubMed |
32. Siegrist, M., Earle, T.C. and Gutscher, H. (2003). Test of a Trust and Confidence Model in the Applied Context of Electromagnetic Field (Emf) Risks. Risk Analysis. Vol. 23, No. 4, pp 705–716. | Article | PubMed |
33. Siegrist, M., Earle, T.C., Gutscher, H. and Keller, C. (2005). Perception of Mobile Phone and Base Station Risks. Risk Analysis. Vol. 25, No. 5, pp 1253–1264. | Article | PubMed |
34. Siegrist, M., Gutscher, H. and Earle, T.C. (2005). Perception of Risk: The Influence of General Trust, and General Confidence. Journal of Risk Research. Vol. 8, No. 2, pp 145–156. | Article |
35. Sjöberg, L. (1999). Perceived Competence and Motivation in Industry and Government as Factors in Risk Perception. In Cvetkovich, G. and Löfstedt, R.E. (eds) Social Trust and the Management of Risk. London: Earthscan, pp 89–99.
36. Sjöberg, L. (2000). Consequences Matter, "Risk" Is Marginal. Journal of Risk Research. Vol. 3, No. 3, pp 287–295. | Article |
37. Sjöberg, L. (2001). Limits of Knowledge and the Limited Importance of Trust. Risk Analysis. Vol. 21, No. 1, pp 189–198. | Article | PubMed |
38. Sjöberg, L. (2002a). Are Received Risk Perception Models Alive and Well? Risk Analysis. Vol. 22, No. 4, pp 665–670. | Article |
39. Sjöberg, L. (2002b). Attitudes toward Technology and Risk: Going Beyond What Is Immediately Given. Policy Sciences. Vol. 35, No. 4, pp 379–400. | Article |
40. Sjöberg, L. (2004a). Explaining Individual Risk Perception: The Case of Nuclear Waste. Risk Management: An International Journal. Vol. 6, No. 1, pp 51–64.
41. Sjöberg, L. (2004b). Gene Technology in the Eyes of the Public and Experts. Moral Opinions, Attitudes and Risk Perceptions. SSE/EFI Working Paper Series in Business Administration 2004:7. Stockholm: Stockholm School of Economics.
42. Sjöberg, L. (2004c). Local Acceptance of a High-Level Nuclear Waste Repository. Risk Analysis. Vol. 24, No. 3, pp 739–751.
43. Sjöberg, L. (2005). The Perceived Risk of Terrorism. Risk Management: An International Journal. Vol. 7, No. 1, pp 43–61.
44. Sjöberg, L. (2006). Opinion Och Attityder Till Förvaring Av Använt Kärnbränsle. (Opinion and Attitudes to a Repository for Spent Nuclear Fuel). Research Report R-06-97. Stockholm: SKB. Svensk Kärnbränslehantering AB.
45. Sjöberg, L. and af Wåhlberg, A. (2002). New Age and Risk Perception. Risk Analysis. Vol. 22, No. 4, pp 751–764. | Article | PubMed |
46. Sjöberg, L. and Wester-Herber, M. (in press). Too Much Trust in (Social) Trust? The Importance of Epistemic Concerns and Perceived Antagonism. International Journal of Global Environmental Issues.
47. Sjöberg, L., Truedsson, J., Frewer, L. and Prades, A. (2000). Through a Glass Darkly: Experts' and the Public;s Mutual Risk Perception. In Cottam, M.P., Harvey, D.W., Pape, R.P. and Tait, J. (eds) Foresight and Precaution. Volume 1. Rotterdam: A.A. Balkema, pp 1157–1162.
48. Slovic, P. (1987). Perception of Risk. Science. Vol. 236, No. 4799, pp 280–285. | Article | PubMed | ISI | ChemPort |
49. Slovic, P. (ed.) (2000). The Perception of Risk. London: Earthscan.
50. Slovic, P., Flynn, J.H. and Layman, M. (1991a). Perceived Risk, Trust, and the Politics of Nuclear Waste. Science. Vol. 254, No. 5038, pp 1603–1607. | Article |
51. Slovic, P., Layman, M. and Flynn, J. (1991b). Risk Perception, Trust, and Nuclear Waste: Lessons from Yucca Mountain. Environment. Vol. 33, No. 3, pp 6–11, 28–30.
52. Slovic, P., Layman, M., Kraus, N., Flynn, J., Chalmers, J. and Gesell, G. (1991c). Perceived Risk, Stigma, and Potential Economic Impact of a High-Level Nuclear Waste Repository in Nevada. Risk Analysis. Vol. 11, No. 4, pp 683–696. | Article | ChemPort |
53. Viklund, M. (2002). Risk Policy: Trust, Risk Perception, and Attitudes. Stockholm: Economic Research Institute.
54. Viklund, M. (2003). Trust and Risk Perception in Western Europe: A Cross-National Study. Risk Analysis. Vol. 23, No. 4, pp 727–738. | Article | PubMed |
55. Viklund, M. and Sjöberg, L. (in press). An Expectancy-Value Approach to Determinants of Trust. Journal of Applied Social Psychology.
56. Visser, P.S., Krosnick, J.A., Marquette, J. and Curtin, M. (1996). Mail Surveys for Election Forecasting? Public Opinion Quarterly. Vol. 60, No. 2, pp 181–227. | Article |
57. Wiedemann, P.M., Thalmann, A.T., Grutsch, M.A. and Schutz, H. (2006). The Impacts of Precautionary Measures and the Disclosure of Scientific Uncertainty on Emf Risk Perception and Trust. Journal of Risk Research. Vol. 9, No. 4, pp 361–372. | Article |

Appendices

Appendix

Items measuring the latent constructs in the models of Study 2 were as follows.

The perceived risk of a repository for spent nuclear power

Would a repository bring about risks for the municipality? (Five response categories, from "Yes, absolutely" to "No absolutely not".)

Would a repository bring about risks for you personally? (Five response categories, from "Yes, absolutely" to "No absolutely not".)

Acceptance of a repository:

Five items to be judged on five-category scales from "Agree absolutely" to "Absolutely disagree":

• A nuclear waste repository in my municipality would be totally unacceptable.
• A nuclear waste repository in my municipality would certainly make me move somewhere else, if I had a chance to do so.
• A nuclear waste repository in my municipality would make me vote for a party which wanted to change the decision.
• It would be irresponsible not to take the chance to site a nuclear waste repository in our municipality.
• I cannot under any circumstances accept a nuclear waste repository in our municipality.

In addition, the respondents were asked to indicate, on a five-category scale, how they would vote in a possible future local referendum about a repository: pro or con.

Social trust

What level of trust do you have for the organizations and authorities below when it comes to the disposal of spent nuclear fuel? (Each item was to be judged with five response categories, from "Very large" to "Very small". "Don't know" was also an explicit response category, treated as missing data.)"

• The Swedish Nuclear Fuel and Waste Management Co (SKB)
• The radiation protection institute
• The nuclear power regulatory agency
• Your municipality

Epistemic trust

Evaluate the following statements about the possible risk associated with a repository for spent nuclear fuel in your municipality. (The following item was to be judged with five response categories, from "Absolutely not" to "To a very large extent".)

Science gives no definite answers to the question about how spent nuclear fuel should be handled in a safe manner.

How complete is scientific knowledge about all aspects of importance when it comes to assessing the risks of handling and handling spent nuclear fuel? (Five response categories, from "Everything or almost everything of importance is known" to "There is very little knowledge about this topic".)

Do you believe that the technological and scientific questions regarding final storage of the spent nuclear fuel are solved in a satisfactory manner? (Six response categories, from "No, absolutely not" to "Yes, absolutely".)

Do you think that current scientific knowledge is sufficiently reliable to be the basis of construction and building of a repository for spent nuclear fuel? (Six response categories, same as in previous item.)

Antagonism

To what extent do you think that the following corporations, authorities, organizations, or persons in Sweden carry out activities, which are contrary to your interests? (Each item was to be judged with seven response categories from "Completely in harmony with my interests" to "Completely contrary to my interests".)

• The Swedish Nuclear Fuel and Waste Management Co (SKB)
• The radiation protection institute
• The nuclear power regulatory agency
• The nuclear power industry

Risk sensitivity

Rate the risk for you personally of the following. Choose the alternative which best corresponds to your opinion. Try to judge all risks. If you cannot answer in some cases, check the square "Don't know". (Each item was to be judged with seven response categories, from "None at all" to "Very large". "Don't know" responses were treated as missing data.)

• Air pollution
• Living close to high voltage power transmission lines
• Climate changes due to the greenhouse effect
• Inadequate food habits
• Genetically modified food
• Cellular telephones
• Plant protection chemicals (pesticides) in agriculture.

Acknowledgements

Study 2 was supported by Swedish Nuclear Fuel and Waste Management Co (SKB).