How improved models of risk perception can inform risk communication Lennart Sjöberg Center for Risk Research Stockholm School of Economics
Risk perception (RP) and risk communication (RC) Beliefs (”perceptions”) constitute the basis: What RC should be about Beliefs are both the motives for RC and targets of RC The sucess of RC is therefore dependent on the validity of RP models applied Lennart Sjöberg SSE
Received view of risk perception There are only few and generally applicable factors. Novelty and dread with regard to a hazard are the major facors, and explain e g opposition to nuclear power virtually completely Experts are objective and not influenced by the ”subjective” risk factors such as novelty and dread Trust, always social trust, if it can established, will reassure the public about risk and make them believe in and accept the ”objective” risk assessment ”Worldviews”, as specified by Cultural Theory, also have a large impact on risk perception The social dilemma of risk management concerns different views of experts and the public – hence research is concentrated on these two groups, which are treated as homogenous Lennart Sjöberg SSE
Credibility of the received view These notions are widely spread and often cited with little or no critical comment The reasons for this situation are many: persuasive data results, closeness to common-sense, and political expediency perhaps the most important Sandman’s “outrage” model is a market adaptation of the basic model Lennart Sjöberg SSE
Data analysis problems The model is usually tested on averages, not inter-individual variability This means that data are aggregated and variability is hidden (“ecological error”) Hence, very high are produced, around 0.8, and it seems that the model accounts for all variability Many have been satisfied with just statistical significance, which is of course easy to achieve with large or medium size samples Lennart Sjöberg SSE
Dread and Novelty as driving factors behind risk perception Explained variance between individuals of original model is typically only 20%, often less This is mostly due to the Dread factor Novelty has no explanatory power at all ”Tampering with Nature” is a very important additional factor Reactions to new technology are not driven by Novelty per se but by other factors, such as perceived benefit or if the technology brings about unique advantages and is hard to replace In addition, various hazards, some new such as terrorism, require their own specific factors Lennart Sjöberg SSE
Cultural theory, world views and values Extensive research has shown that there is only a very weak relationship between world views and perceived risk Other value based dimensions, such as political ideology and New Age beliefs are related to perceived risk, calling for a different theoretical approach Lennart Sjöberg SSE
Trust and conceptions of knowledge Typically only weak effects of social trust on perceived risk, correlations of 0.3 or less Stronger effects of trust in science, as distinct from social trust (in experts or organizations) –epistemic trust Another important factor is perceived antagonism Lennart Sjöberg SSE
Whose ”risk” – more specifically? Personal and general risk differ both as to level and rank order. General risk is important for lifestyle (smoking etc), personal for environmental risks, and technology hazards This distinction was missed in the Psychometric Paradigm, where one asked for ratings of ”risk” without specifying the target Research shows that such ratings are close to general risk But general risk is not the most relevant one in policy contexts Lennart Sjöberg SSE
Prevalence of optimistic bias Only a few percent in typical studies judge personal risk as larger than general risk People will go ahead and take risks even if they are aware that they take them, because of expected benefits (pleasure, beauty, etc) Lennart Sjöberg SSE
Focus on risk: Traditional approach People are asked to rate the ”risk” It is assumed that perceived risk, defined in this way, is the factor driving risk related behavior, such as demand for risk reduction But the assumption is implict and never regarded as a problem Lennart Sjöberg SSE
The problem What is driving demand for risk reduction? Is perceived risk the important factor? If not, what factor is most important ? Lennart Sjöberg SSE
Example Consider the risk for a Swedish citizen, age 30-45, to 1. Get a severe cold during the next 12 months 2. Be infected with the HIV virus during the same time period Which risk is the largest one? Which risk is more important to be protected from? Lennart Sjöberg SSE
Analysis ”Risk” is dependent upon both probability of injury and severity, should it occur People were therefore asked to rate probability severity risk demand form risk reduction Lennart Sjöberg SSE
Results Many (all) studies where this has been tried have shown: risk and probability are closely related severity and demand for risk reduction are closely related risk and demand for risk reduction are only moderately related Lennart Sjöberg SSE
Note: risk and probability are almost synonyms, Summary of findings on risk, consequences, probability and demand for risk mitigation, for hazards above a probability threshold of concern Probability Risk Consequences Demand for mitigation These findings are very clear and have been replicated several times. Note: risk and probability are almost synonyms, and have little impact on demand for mitigation, above a probability threshold level pf concern Lennart Sjöberg SSE
Qualifications Some very low risks with large potential consequences in everyday life tend to be ignored This could be construed as a threshold function; VERY low probabilities IN EVERYDAY LIFE are ignored But this is probably not true in policy contexts Lennart Sjöberg SSE
Implications In risk communication it should be clear that the public wants to hear about severity of consequences, not so much about probabilities: Probability is hard to understand Precise estimates of very small probabilities must rely on many assumptions and are seldom very credible In risk perception research, it is necessary to broaden the scope; just studying ”risk” is not sufficient Lennart Sjöberg SSE
My recent work on a nuclear waste repository included actual emotional reactions Several emotions were rated, not only one Not only negative but also positive emotions About 800 respondents from two communities where site studies are now carried out Two candidate municipalities, one control and a national sample Response rate 50% Lennart Sjöberg SSE
”Emotion balance” in the four samples (mean standardized values) Lennart Sjöberg SSE
Regression analysis of attitude, variance accounted for by emotions = 0.61 Lennart Sjöberg SSE
Regression analysis of intended vote, variance accounted for by emotions = 0.70 Lennart Sjöberg SSE
Model of the attitude to a repository Lennart Sjöberg SSE
Own emotional reaction to nuclear power Table 1. Correlations between emotional reactions and the attitude to nuclear power. Emotion Own emotional reaction to nuclear power The anticipated emotional reaction of others to nuclear power Anger -0.62 -0.08 Contempt -0.55 -0.10 Fear -0.65 -0.06 Interest 0.28 0.17 Sadness -0.58 -0.12 Satisfaction 0.57 0.21 Guilt -0.25 0.03 Shame -0.29 0.04 Worry -0.61 Lennart Sjöberg SSE
Mean emotional reactions attributed to others vs. own reactions. Lennart Sjöberg SSE
Conclusions about emotions Specific and current emotional reactions do seem to explain much of attitudes and policy behavior Compare these strong effects with the almost zero importance of anticipated dread of others Both positive and negative emotions are important Note that “worry” contributes beyond the effect of “fear” Anger seems to be more important than fear in policy contexts Lennart Sjöberg SSE
Experts vs the public Original work claiming that experts make ”correct” and ”objective” risk judgments made used a very small group of ”experts” with questionable competence Later work with substantive experts has shown that they have similar structure of risk perception, but lower level Risk perception is related to experts’ field of responsibility, not to knowledge Lennart Sjöberg SSE
Experts and the public: general food risks Lennart Sjöberg SSE
Experts and the public: personal food risk Lennart Sjöberg SSE
Conclusion These results exemplify a common trend: Experts judge risks as smaller when they are within their general area of responsibility Lennart Sjöberg SSE
Study the perceived risk of genetic engineering, public and experts Lennart Sjöberg SSE
Conclusion These data (genetic engineering study) suggest that experts judge personal risk in a manner similar to the public However, their judgments of general risk seem to be less correlated with the ”subjective” factors Lennart Sjöberg SSE
Categories of the public Public: alarmists and risk deniers as extremes Risk deniers ”in bed with experts”, about 4 times as many as alarmists in typical cases, but SILENT Lennart Sjöberg SSE
Typical distribution of risk ratings Lennart Sjöberg SSE
Typical distribution of risk ratings Lennart Sjöberg SSE
Conclusions about individual variability The distributions of risk judgments are typically skew, with many people judging risks to be very small and few judging them to be large However, there is a “silent majority” phenomenon – those who perceive large risks are more active and are heard more often in the debate arenas This is one important reason for studying empirically what views people have about risks Experts differ widely depending on the promoter/protector role, and organizational basis (industry, government, NGO’s) Lennart Sjöberg SSE
Conclusions about RP models Novelty and dread have not been found to be of primary importance, especially not novelty Other factors, such as interfering with Nature and moral values, are more important World Views (Cultural theory) have also failed – they are only weakly related to perceived risk Epistemic trust is important, social trust more marginal Emotions are important but need to be operationalized as personal and a broad spectrum must be covered, not only strong fear (dread) Perceived risk must be studied as both personal and general – they give partly different information ”Risk” is not the most useful concept to focus on, people are much more concerned about consequences ”The public” is not a homogenous group – there are both alarmists and risk deniers, the latter are much more numerous Lennart Sjöberg SSE
What does it all mean for risk communication? Emotions are important to take into account – but not only strong fear Several cognitive and ideological factors play a role – but “World Views” are a dead end People’s understanding and trust in science is very important “Risk” and probability are marginal to people who respond to notions about anticipated consequences There are very different reactions to hazards – risk neglect is a more common and important problem than excessive alarm Experts are not that different from the public in how they react to hazards, outside their specific area of responsibility Lennart Sjöberg SSE
For more information… See my homepage http://www.dynam-it.com/lennart/ Many papers and reports can be downloaded from that site Lennart Sjöberg SSE