Abstract: The purpose of the research reported here was to determine how best to explain risk magnitude, and thus improve the correlation between risk and response. Subjects were 1,402 homeowners, who read one of three hypothetical news stories about radiation exposure: a low-outrage, high-risk story; a high-outrage, low-risk story; or a low-outrage, low-risk story. The story was followed by a personal radiation test result and various types of information, and then by a feedback questionnaire focusing on perceived threat and action intentions. There were four treatments:
(1) A ‘base risk’ condition in which subjects were given only the result and an estimate of the associated lifetime cancer risk;
(2) an ‘alternate risk’ condition in which subjects were provided the same information, but with a test result and risk estimate 10 times higher (for the low-outrage, high-risk story) or 10 times lower (for the high-outrage, low-risk story);
(3) a ‘compare to normal’ condition, in which subjects were given the same test result as in the ‘base risk’ condition, but with a comparison to normal background levels provided instead of the risk magnitude information; and
(4) a ‘base risk + chart’ condition, in which subjects received the risk information augmented with a risk ladder, risk comparisons, and a recommended action level.
Both comparisons to normal background and the chart were much more effective than the ‘base risk’ condition in increasing threat perception and action intentions when the risk was high and decreasing them when the risk was low. Outrage substantially increased these responses, but did not diminish the ability of comparisons to background and risk charts to reduce them.
There is considerable agreement about the difficulty of conveying information about risk magnitudes to the public (National Research Council, 1989). As many discouraged policy-makers have noted, citizens often ignore information designed to alert them to significant risks, yet insist on remedial action for other risks that are too small to merit the attention they receive. Even when people obtain their own tests, as with home measurements for airborne radon or lead in drinking water, there may be only a weak relationship between their test results (a measure of risk magnitude) and their responses (Johnson and Luken, 1987; Smith and Johnson, 1988; Sjöberg, 1989). Furthermore, the response to comparable risk numbers often varies from hazard to hazard (Sandman et al., 1994). Some hazards provoke a conservative response, while others elicit much less concern.
Article Table of Contents
A substantial research literature has identified many of the factors other than risk magnitude that seem to explain why some hazards provoke a far greater response than others (Slovic et al., 1979; Slovic, 1987). But relatively little research has attempted to determine how best to explain risk magnitude, and thus to improve the correlation between risk and response (Rohrmann 1990; Weinstein et al., 1994; Weinstein et al., 1996).
In previous research we assigned subjects a hypothetical test result for radon or asbestos contamination of their homes, provided a constant fact sheet about the hazard, and explained the risk magnitude associated with their test results using various techniques. Subjects were then asked questions about the threat, their mitigation intentions, and related perceptions. A significant risk magnitude effect was found; that is, subjects perceived a higher risk when the risk magnitude was higher. Three specific factors were found to improve risk explanations: (1) A recommended action standard; (2) explicit advice on what to do at various concentration levels; and (3) a risk ladder showing a range of possible test results and the risk magnitude associated with each (with ladders designed so that subjects’ test results appeared high on the ladder if the goal was to increase risk perception, and low on the ladder if the goal was to decrease risk perception). Risk comparisons were found to be helpful in some ways, but they did not significantly affect threat perception or action intentions. (Note 1)
The research reported here built on this base in three ways. First, it tested the combined effects of the components found useful in the earlier research – a ‘best shot’ risk explanation that augmented risk probability information for the subject’s assigned test result with a risk ladder, a recommended action guideline, (Note 2) and comparisons to more familiar risks. (Note 3)
Second, it tested an approach in which comparisons to normal background levels were provided instead of illness likelihood information. In many real-world situations (well water testing, factory emissions, etc.), risk data are often not available; concentration information and comparisons to background concentrations are provided instead. We are aware of no prior empirical research exploring the effectiveness of comparisons to background vis-à-vis risk data as a way of providing context for a test finding.
Third, it examined communication in two situations in which the risk itself was quite low, one of them a high-outrage risk controversy. This factor requires amplification.
The effectiveness of risk explanations probably cannot be meaningfully investigated without considering the size of the risk. Different strategies may be required for helping people see that a small risk is small than for helping people see that a large risk is large. Moreover, the policy implications of risk communication failures are quite different in these two situations. Lives can be lost when an agency’s efforts to warn people about serious risks are unsuccessful. When people persist in worrying disproportionately about minuscule risks, in contrast, the costs include unnecessary anxiety, misused environmental protection dollars, public policy gridlock, and reduced agency credibility.
Our earlier research focused on two hazards that both pose serious risks. Residential radon and residential asbestos are not identical hazards, of course. But on the variables that best predict whether a hazard will provoke under-reaction or over-reaction from the public – dread, control, familiarity, trust, fairness, etc. – both tend toward under-reaction. The earlier research thus shed considerable light on how to alert people to serious risks but much less light on how to reassure people about negligible risks.
This distinction is especially important for negligible risks that are substantial sources of public concern. Typically, the public responds less to the seriousness of a risk than to such factors as trust, control, fairness, and courtesy. Sandman (1987, 1991, 1993), Hance et al. (1988, 1990), and Sandman et al. (1987) have proposed the labels ‘hazard’ and ‘outrage’ to refer, respectively, to the technical and the nontechnical aspects of risk. Using different vocabulary, many others have also noted and studied the importance of these nontechnical aspects of risk perception, among them Kasperson (1986), Slovic (1987), Krimsky and Plough (1988), Covello and Allen (1988), and Covello et al. (1988). In Sandman’s terminology, ‘hazard’ is the product of outcome severity and probability, while ‘outrage’ is a function of whether people feel the authorities can be trusted, whether control over risk management is shared with affected communities, etc. According to this perspective, no matter how serious the risk is (in hazard terms), and no matter how much effort is put into explaining how serious it is, the degree of outrage is likely to determine much of the public’s response to the risk.
Three experimental studies of outrage (Sandman et al., 1993), employing hypothetical news stories, found higher perceived risk when the manipulated outrage was high than when outrage was low. The outrage manipulation had more impact on risk perception than a 10,000-fold difference in risk seriousness; the amount of technical information in the news stories had no impact on risk perception at all.
In the language of ‘hazard’ versus ‘outrage’, then, our earlier research focused on risks that were moderate to high in hazard and low in outrage. The research reported here addresses the important question of how effective the approaches tested previously are when applied in different contexts: to a risk that is low in hazard and high in outrage, or to a risk that is low in both hazard and outrage. (Note 4)
2. Materials and Methods
2.1. Experimental Conditions
Note 5); (3) a ‘compare to normal’ condition, in which subjects were given the same test result as in the ‘base risk’ condition, but with a comparison to normal background levels provided instead of the risk information; and (4) a ‘base risk + chart’ condition, in which subjects received the risk information augmented with a risk ladder, risk comparisons, and a recommended action level.Three hypothetical news stories were used: a low-outrage, high-risk story (Radon), a high- outrage, low-risk story (Nuclear Waste), and a low-outrage, low-risk story (Radiation). For each of the first two stories, four conditions were used: (1) A ‘base risk’ condition in which subjects were given their hypothetical test result and an estimate of the lifetime cancer risk associated with that result in terms of deaths per 1,000 people; (2) an ‘alternate risk’ condition in which subjects were provided the same information, but with a test result and risk estimate 10 times higher (for the low-outrage, high-risk story) or 10 times lower (for the high-outrage, low-risk story) (
For the third story, only the ‘base risk’ and ‘base risk + chart’ conditions were used. The radiation levels and explanatory information used in each of the ten experimental conditions are summarized in Table 1.
|Low outrage, high risk (Radon)|
|Compare to normal||8||–||20× greater
|Base risk + chart||8||40-in-1,000||–||Yes|
|High outrage, low risk (Nuclear Waste)|
|Compare to normal||0.002||–||200× less
|Base risk + chart||0.002||1-in-100,000||–||Yes|
|Low outrage, low risk (Radiation)|
|Base risk + chart||0.002||1-in-100,000||–||Yes|
|Note: pCiL-1 = picoCuries of radiation per litre of air.|
The sample was recruited from two communities with quite low radon levels: Irmo, South Carolina and Cary, North Carolina. Because several of the questionnaire items asked what decisions subjects would make about home remediation in the hypothetical situation presented, only owners of single family dwellings – who might find it easier to imagine themselves in such a situation – were eligible. Within each such residence the male or female head of household was selected to participate.
Of those households where a contact was made and which met the selection requirements, 65.5% agreed to take part in the study. A total of 1,803 individuals were recruited and 1,402 (77.8%) returned completed questionnaires. This is a net response rate of 51.0%. The return rates were equal across conditions.
Overall, the sample was 53.3% female. As expected, the sample was better educated and wealthier than the communities from which it was recruited, in part because home ownership was a requirement for participation.
Each subject received a four-page, 8-1/2" by 11" pamphlet. The first page of the pamphlet was a cover letter asking participants to ‘tell us how you think you would feel in this situation and what you think you would do.’ Page two and the top of page three contained a news story that the subject was asked to read. The story was followed by various types of information designed to help the subject interpret his or her radiation level.
2.3.1. News stories
Three stories were created to describe situations in which radiation within homes posed a threat of unknown magnitude to residents of ‘the Washburn Circle section of Middletown.’ The stories were presented in newspaper columns and emulated newspaper writing style. In the low-outrage, high-risk story (Radon Risk to Middletown Homes), the radiation came from radon produced by naturally occurring uranium. In the high-outrage, low-risk story (Nuclear Waste Contaminates Middletown Homes), the radiation came from sand that had absorbed radioactivity when it covered a storage site for spent nuclear power fuel rods, sand that had then been used illegally to make concrete for the homes’ foundations. In the low-outrage, low-risk story (Radiation in Middletown Homes), the source of radiation was sand used to make concrete for the homes’ foundations that was only later found to contain naturally occurring uranium.
The key actors in all three stories were Charles Schmidt of the State Department of Environmental Protection (DEP), at whom mistrust was aimed in the high-outrage story; Dr. Susan Baxter, Director of the Middletown University Health and Safety Program, who offered to do independent testing and was highly trusted in all stories; and Harriet Mossman, chair of the Washburn Circle Neighborhood Association. All three stories explained that the problem had been discovered accidentally by a sixth grade student working on a science project, that Dr. Baxter had followed up with measurements of individual home radiation levels, and that the results were about to be mailed to homeowners.
2.3.2. Risk information
Immediately after the news story, instructions asked subjects to imagine living in one of the homes in Washburn Circle and to read their home radiation report, which followed. The report described ‘the level of extra radiation in your principal living area from the (experimental source)’ as ‘– picocuries of radiation per litre of air (pCiL-1).’ This information was provided in all conditions.
The ‘base risk’ treatment then translated the radiation level into a risk estimate. In the 8 pCiL-1 conditions, for example, a paragraph stated: ‘To interpret this result, it may help you to know the health risk caused by being exposed to this amount of radiation. If 1,000 people lived for 70 years in homes with 8 picocuries of radiation per litre of air, about 40 of them would be expected to contract lung cancer as a result of this exposure. In other words, for every 1,000 people exposed to this level of radiation over a lifetime, 40 more of them, on average, would get lung cancer than if they were not exposed to the radiation.’
Comparisons to normal background levels took the following form: ‘To interpret this result, it may help you to know that the average outdoor background level of breathable radiation in the United States, from all sources, is approximately 0.4 picocuries of radiation per litre of air. The radiation exposure in your house from (the experimental source) is thus 20 times greater (or 200 times less) than the average outdoor background level.’
The ‘base risk + chart’ treatment included the risk paragraph plus a chart containing the following information: (1) A ladder of concentrations that started at 0.0 pCiL-1 and stopped at 10 pCiL-1; (2) Information about the cancer risk at 11 different levels; (3) The EPA action guideline for radon (‘at 4 pCiL-1 EPA recommends that you reduce your home radiation level’ ); (4) Comparisons to three other causes of death (stroke, diabetes, and colon/rectal cancer); and (5) Comparisons of three radiation levels to normal background (average outdoor level, 15 times average outdoor level, and 25 times average outdoor level). A rubber stamp was prepared that contained an arrow and the phrase ‘YOUR LEVEL.’ When subjects received a chart, the stamp pointed to the spot on the risk ladder that represented their radiation report, in red ink. For subjects assigned 8 pCiL-1, the red arrow pointed four-fifths of the way up the ladder; for subjects assigned 0.002 pCiL-1, it pointed just above the 0.0 point at the bottom of the ladder.
2.3.3. Feedback questionnaire
The feedback questionnaire appeared on a separate sheet and was identical in all conditions. The questionnaire asked how clear the newspaper article was (1 = very confusing; 4 = very clear) and how clear the information was about the seriousness of the radiation level (1 = very confusing; 4 = very clear).
Three questions assessed the effects of the outrage manipulation: ‘How angry would you feel to find this level of radiation?’ (1 = not at all angry; 5 = extremely angry); ‘From what you have read, do you think you could trust the risk information provided to you by Dr. Baxter of the University Health and Safety Program?’ (1 = definitely could trust; 5 = definitely could not trust); and ‘From what you have read, do you feel that Charles Schmidt, the State DEP spokesperson quoted in the newspaper article, cares about the health and safety of your neighborhood?’ (1 = cares a lot; 5 = doesn’t care at all).
Four questions were employed to measure perceived threat: ‘How would you describe the danger from the radiation level found in your Washburn Circle home?’ (1 = no danger; 6 = very serious danger); ‘If you continued to live in your Washburn Circle home and did nothing about the radiation, what is your impression of the chance that the radiation would give you lung cancer?’ (1 = no chance; 7 = certain to happen); ‘How concerned would you feel finding this level of radiation?’ (1 = not at all concerned; 5 = extremely concerned); and ‘How frightened would you feel finding this level of radiation?’ (1 = not at all frightened; 5 = extremely frightened).
Four questions measured action intentions: ‘Given what you have learned about the risk, do you think it would be worth your spending $300 to reduce the risk to zero?’ (1 = definitely would not spend $300; 5 = definitely would spend $300); ‘Given what you have learned about the risk, do you think it would be worth your spending $3000 to reduce the risk to zero?’ (1 = definitely would not spend $3000; 5 = definitely would spend $3000); ‘If you learned that it wasn’t possible to reduce the radiation in your home, would that make you want to move away?’ (1 = would not feel any interest in moving away; 5 = would insist on moving away); and ‘Imagine that you were looking for a new home in a new neighborhood, and found that it had this level of radiation (the level we told you was found in your home). Would this reduce your interest in buying this new home?’ (1 = would not be at all reluctant to buy a home with this level of radiation; 5 = definitely would not buy a home with this level of radiation).
A final section asked subjects their sex, education, and household income, and the amount of time they spent reading the booklet and filling out the questionnaire.
The study was described on the telephone as focusing on how homeowners make decisions about environmental risks. The caller told the homeowner that people were being asked ‘if they would read a brief, one-page news article about an environmental problem and answer a short questionnaire to tell us what they would do if they were in that situation.’ The sum of $1 was offered as a thank you for their help.
The four-page pamphlet, two-page questionnaire, and a stamped, self-addressed envelope were mailed to each subject who agreed to participate. Reminder calls (and additional mailings if needed) were made to increase response rates. All data collection took place in June 1993.
3. Results (Note 6)
3.1. Scale Construction
To measure perceived threat and action intentions, answers to the four threat questions were added together and answers to the four action questions were added together. Scale reliabilities were calculated after standardizing the variables to the same mean and variance within each condition. The reliability coefficient (Cronbach’s alpha) of the threat perception scale was found to be 0.90 and the reliability of the action intentions scale was 0.86.
3.2. Analysis Strategy
Because the experimental design was not a complete 3 (stories) × 4 (treatments) factorial, analyses could not be based on a simple, two-factor analysis of variance. Instead, the analysis strategy was to look first at the low-outrage, high-risk and high-outrage, low-risk stories in a simple 2 (stories) × 4 (treatments) analysis of variance. Differences among conditions – either across both stories when no story × treatment interactions were found or within each story when significant interactions were present – were examined with post-hoc tests using Tukey’s HSD criterion, with a significance criterion of 0.02 to minimize false positives. Additional calculations examined the effect of the outrage manipulation at a constant low-risk level by comparing the ‘base risk’ and ‘base risk + chart’ conditions for the high-outrage, low-risk story and the low-outrage, low-risk story in a simple 2 (stories) × 2 (treatments) analysis of variance. These two analyses will be referred to as the 2 × 4 analysis and the 2 × 2 analysis, respectively. The mean values of the main dependent variables are shown in Table 2 for each condition.
|Condition||Article clear||Risk report clear||Threat||Action intentions||Anger||Distrust Baxter||Distrust Schmidt|
|Low outrage, high risk (Radon)|
| Base risk|
(N = 131)
| Alternate risk|
(N = 135)
| Compare to normal|
(N = 136)
| Base risk + chart|
(N = 139)
|High outrage, low risk (Nuclear Waste)|
| Base risk|
(N = 149)
| Alternate risk|
(N = 153)
| Compare to normal|
(N = 131)
|Base risk + chart (N =137)||3.42||3.48||10.45||12.64||2.91||2.24||3.13|
|Low outrage, low risk (Radiation)|
| Base risk + chart|
(N = 143)
3.3. Manipulation Checks
If the outrage manipulation were successful, distrust of the Department of Environmental Protection spokesman, Schmidt, and anger at the situation would be much greater in the high-outrage conditions than in the low-outrage conditions. The 2 × 4 analysis confirms that distrust was much greater in the high-outrage conditions. The effect of story was highly significant, F(1, 1074) = 177.7, p < 0.0001. There was no treatment effect nor any interaction between story and treatment (p’s > 0.2).
Anger revealed a more complicated pattern. In the 2 × 4 analysis, there was a large effect of story, F(1, 1102) = 68.8, p < 0.000l, with more anger overall in the high-outrage conditions. There was no treatment effect, F(3, 1102) = 1.8, p = 0.15, but there was a highly significant story × treatment interaction, F(3, 1102) = 10.4, p < 0.0001. Further comparisons among the treatments associated with each story revealed that while the high-outrage stories produced a high level of anger, the ‘compare to normal’ and ‘base risk + chart’ treatments, which did a good job of convincing subjects that the risk was low (as described later), decreased this anger substantially.
As intended, distrust of the independent expert, Dr. Baxter, was low. The 2 × 4 analysis showed a very small difference between the low-outrage, high-risk and the high-outrage, low-risk stories, F(1, 1098) = 4.05, p < 0.05, with mean values of 2.21 and 2.31 in these two stories, respectively. There were no significant treatment effects or story × treatment interactions, p’s > 0.15. The difference in distrust between the high-outrage, low-risk and the low-outrage, low-risk conditions was not statistically significant, p > 0.1.
Other calculations examined the perceived clarity of the news stories themselves and of the risk information provided. The 2 × 4 analysis showed that the low-outrage, high-risk story was rated as clearer than the high-outrage, low-risk story, F(1, 1092) = 8.80, p = 0.003. The difference among the means, however, was quite small: 3.48 vs. 3.37. Significant differences were found among treatments in how clearly the seriousness of the risk was explained, F(3, 1103) = 9.28, p < 0.0001. The ‘base risk + chart’ condition (mean = 3.46) was judged to be clearest. Overall, it was rated significantly higher than both the ‘compare to normal’ condition (mean = 3.10) and the ‘base risk’ condition (mean = 3.23), p’s = 0.02.
3.4. Effects on Perceived Threat
The 2 × 4 analysis of the low-outrage, high-risk and the high-outrage, low-risk stories revealed a significant story effect, F(1, 1087) = 253.0, p < 0.0001, a significant treatment effect, F(3, 1087) = 14.12, p < 0.0001, and a small but significant story × treatment interaction, F(3, 1087) = 3.02, p < 0.05.
Although the Radon story was correctly judged to present a greater threat than the other two stories (see Figure 1), the difference was not as great as might be desired. The perceived threat in the Nuclear Waste (high-outrage, low-risk) ‘base risk’ condition was just 1.3 scale points less than the perceived threat in the Radon (low-outrage, high-risk) ‘base risk’ condition, despite the fact that the lung cancer risk was described as being 1-in-100,000 in the former and 40-in-1,000 in the latter, a 4000× difference. The difference in perceived threat between these two base risk conditions was not significant, F(1, 273) = 3.50, p < 0.l.
Figure 1. Effects of condition on perceived threat.
Within the low-outrage, high-risk Radon story, the ‘base risk’ condition was viewed as significantly less threatening than all the other conditions, which did not differ from one another. Comparing the ‘base risk’ condition (with a lung cancer risk of 40-in-1,000 or 4%) and the ‘alternate risk’ condition (with a lung cancer risk of 400-in-1,000 or 40%), we see an increase in the perceived threat scale from 13.9 to 16.2. Thus a 10-fold increase in risk produced an increase on the perceived threat scale of 2.3 points. Roughly the same increase in perceived threat was accomplished merely by describing the radon level as 20 times normal background levels (without providing any risk statistics) or by adding the chart to the risk statistics in the ‘base risk’ condition.
For the high-outrage, low-risk Nuclear Waste story, there was no significant difference between the ‘base risk’ condition and the ‘alternate risk’ condition. People reacting to a risk of 1-in-100,000 did not respond differently than those reacting to a risk of 1-in-1,000,000. However, the conditions that compared the risk to normal levels or added the chart to the risk probability information were viewed as presenting significantly less risk than the ‘base risk’ condition. The mean on the perceived threat scale declined from 12.6 in the ‘base risk’ condition to 10.9 for the ‘compare to normal’ condition and to 10.4 for the ‘base risk + chart’ condition, decreases of 1.7 points and 2.2 points, respectively. In a high-outrage, low-risk situation, in other words, both the comparison to background and the risk chart reduced perceived threat more than a 10-fold decrease in the actual probability of lung cancer.
The 2 × 2 analysis involving the high-outrage, low-risk and the low-outrage, low-risk stories showed a strong story effect, F(1, 568) = 68.4, p < 0.0001, a strong treatment effect, F(1, 568) = 38.9, p < 0.0001, and no interaction, p > 0.9. There was a 2.9-point decrease in perceived threat when outrage elements were absent from the story, an effect far greater than the effect of decreasing the actual risk by a factor of ten. The treatment effect was also substantial. Adding a risk chart to the ‘base risk’ condition reduced the perceived threat scale from 9.8 to 7.6 for the low-outrage, low-risk story, and from 12.6 to 10.4 for the high-outrage, low-risk story, reductions of 2.2 points in both cases. The absence of an interaction effect shows that the outrage effect and the treatment effect were independent and additive. With or without the risk chart, subjects who read the low-outrage story perceived less threat than those who read the high-outrage story, though the actual risk level was the same. And regardless of which story they read, high-outrage or low, subjects who received the risk chart perceived less threat than subjects who received just their own risk probability information.
3.5. Effects on Action Intentions
The findings for action intentions were similar to those for perceived risk. The 2 × 4 analysis of the low-outrage, high-risk story and the high-outrage, low-risk story revealed a significant story effect, F(1, 1081) = 76.2, p < 0.0001, a significant treatment effect, F(3, 1081) = 11.22, p < 0.0001, and a significant story × treatment interaction, F(3, 1081) = 5.00, p < 0.002.
Figure 2 shows that the low-outrage, high-risk story produced greater action intentions overall than the other two. But in the ‘base risk’ conditions – that is, with risk numbers only – there was no difference between subjects who read the low-outrage (Radon) story and faced a high risk of 40-in-1,000 and those who read the high-outrage (Nuclear Waste) story and faced a low risk of 1-in-100,000.
Figure 2. Effects of condition on action intentions.
Within the low-outrage, high-risk story, the ‘base risk’ condition led to significantly lower action intentions only in comparison with the 10-fold greater ‘alternate risk’ condition. The comparisons to normal and the risk chart yielded only slightly lower action intentions than the ‘alternate risk’ treatment. These were higher, but not significantly higher (at p < 0.02) than the ‘base risk’ condition. For subjects who read the low-outrage, high-risk story, in other words, no explanation significantly increased action intentions above the action intentions produced by the bare-bones risk information – though an increase in the risk from 40-in-1,000 to 400-in-1,000 did increase action intentions.
For the high-outrage, low-risk story, in contrast, there was no difference in action intentions between subjects in the 1-in-100,000 ‘base risk’ condition and those in the 1- in-1,000,000 ‘alternate risk’ condition. However, the comparisons to normal and the risk chart substantially decreased action plans. Action plans in these two treatments were significantly lower than both the ‘base risk’ and the ‘alternate risk’ condition, and not significantly different from one another.
The 2 × 2 analysis involving the high-outrage, low-risk and the low-outrage, low-risk conditions again showed a strong story effect, F(1, 562) = 78.5, p < 0.0001, a strong treatment effect, F(1, 562) = 43.3, p < 0.0001, and no interaction, p > 0.4. There was a 3.2-point decrease in action intentions when outrage elements were absent from the story, an effect far greater than decreasing the actual risk by a factor of ten. And there was a 2.4-point decrease in action intentions when a risk chart was added to the risk probability information provided to subjects. The absence of interactions between story and treatment shows that the chart was just as influential in reducing action intentions for the high-outrage story as for the low-outrage story. Once again the two effects were independent and additive.
The research reported here used responses to hypothetical situations. It is impossible to say how realistic subjects found these simulations and how realistically they responded to them. It seems likely that the effects of outrage were diminished by the hypothetical nature of the hazards, and that the effects of risk magnitude and of various ways of explaining risk magnitude – especially in the presence of high outrage – were augmented. But no research findings back this supposition.
In addition, real community hazard situations develop over days, months, or even years; the study compressed these histories into written materials that take only a few minutes to read. Prolonged exposure to a risk controversy may make people more responsive to outrage factors than they were in this research. We do not know whether it might make them more or less responsive to explanations of risk magnitude.
4.1. The Effects of Comparisons to Normal Background
Despite having no information at all about the likelihood of harmful consequences, subjects responded strongly to the information that their radiation exposure was 20× higher or 200× lower than normal background. Compared to the ‘base risk’ condition, comparisons to normal increased perceived threat for the high-risk Radon story and decreased both perceived threat and action intentions for the low-risk Nuclear Waste story. Comparisons to background thus did a better job than risk information itself in helping subjects respond in proportion to the actual risk. Phrased another way, comparisons to normal background helped subjects respond more to the high-risk situation even though the outrage was small, and respond less to the low-risk situation even though the outrage was great. The effect was equal to the effect of a 10-fold increase in risk for Radon; it was greater than the effect of a 10-fold decrease in risk for Nuclear Waste. (Note 7)
The potency of the comparison to normal background and its symmetry (that is, its effectiveness in both low-outrage, high-risk situations and high-outrage, low-risk situations) suggest that it may be a valuable piece of information to include whenever this information is available. Although no treatment combined comparisons to normal background with ‘base risk’ information, it seems likely that the former adds value even when the latter is provided as well.
The study leaves unclear the impact of comparisons to normal background for risks that are less serious than 20× greater than background but more serious than 200× smaller than background. Suppose a neighborhood’s exposure to effluent from a nearby factory is equal to normal background for a particular chemical; the factory thus doubles the neighborhood’s total exposure to that chemical. Deciding that 20× background is serious or that 200× less than background is trivial is a comparatively easy decision, but how would citizens interpret an exposure that was roughly the same as background?
It should not be forgotten that a comparison to background does not constitute risk information. For some hazards, normal background levels are sufficient to constitute a meaningful health risk, and even a small increment would be unwise if it were preventable. For other hazards, the risk due to normal background exposure is negligible, and an exposure many times background would still be negligible. Thus, comparisons to background can give misleading impressions contrary to the actual risk magnitudes.
Nonetheless, the finding is clear. When a risk adds only a very small percentage to normal background, people readily conclude that it is not too serious; when it adds a large multiple of normal background, they readily conclude that it is quite serious. At least with the large multiples of ‘normal’ studied here, information about how a target risk compares to normal background levels has more impact on perceived threat and action intentions than numbers describing the odds of experiencing harmful effects.
4.2. The Effects of The Risk Chart
Similar effects were achieved by supplementing the ‘base risk’ condition with a chart that included a risk ladder (with the target risk high on the ladder for the Radon story, low on the ladder for the Nuclear Waste and Radiation stories), risk comparisons, and a recommended action standard. The chart decreased perceived threat in the two low-risk situations, while it increased perceived threat in the high-risk situation. The effects of the chart were roughly equal to the effects of comparisons to normal background – equally effective as a 10× increase in actual risk for Radon, and more effective than a 10× decrease in actual risk for Nuclear Waste. Similarly, the chart decreased action intentions for Nuclear Waste and Radiation, though the effect on action intentions for Radon was not significant.
Note that the ‘base risk + chart’ treatment included some of the comparison information that proved so effective in the ‘compare to normal’ treatment. Comparisons to normal background were included on the chart, along with comparisons to the risk from stroke, colon/rectal cancer, and diabetes. At the top of the risk ladder, the chart showed ‘25 times average radon (or radiation) level’; a little over halfway up the ladder, the chart showed ‘15 times average radon (or radiation) level’; near the bottom of the ladder, it showed ‘average outdoor radon (or radiation) level.’ Thus, a subject who studied the chart carefully could deduce that his or her risk was far above normal background for the radon situation, and far below normal background for the other two. Although this information was far less emphasized than it was in the ‘compare to normal’ treatment, the fact that ‘base risk + chart’ and ‘compare to normal’ performed equally well leaves open the possibility that the effectiveness of the former may be due to its inclusion of comparison-to-normal information. No experimental treatment combined the ‘compare to normal’ and the ‘base risk + chart’ treatments.
The ‘base risk + chart’ condition constituted a significant improvement over the ‘base risk’ condition. For all three hypothetical situations, the chart significantly influenced threat perceptions – decreases when the actual risk was low, increases when it was high. It similarly decreased action intentions when the actual risk was low, though the effect on action intentions for the high-risk Radon story was not significant.
4.3. Outrage Effects and Outrage Reduction
Subjects in the high-outrage situation (Nuclear Waste) were of course much angrier than in the two low-outrage situations, and they were much more distrustful of the DEP spokesperson, Schmidt. It is noteworthy that they were not more distrustful of the neutral information source, Dr. Baxter. This suggests that the distrust that typically characterizes high-outrage risk controversies need not contaminate all actors; an independent expert who is not affiliated with the distrusted authorities can sometimes be trusted.
Not surprisingly, outrage affected threat perceptions and action intentions. That is, subjects in the high-outrage, low-risk situation reported much higher perceived threat and higher action intentions than subjects in the low-outrage, low-risk situation, although the actual risk was identical. When subjects received only risk numbers, the effect of outrage was practically as large as the effect of the 4000-fold difference in risk between the high-risk and low-risk conditions. When communication was improved by comparisons to normal levels or by the risk chart, the outrage effect, though still quite substantial, was less than half the effect of the 4000-fold difference in risk.
Most encouraging was the apparent ability of both comparisons to normal background and the risk chart to reduce threat perception and action intentions even in the presence of high outrage. Many practitioners have suggested that when people are outraged, explanations of the risk data are unlikely to prove fruitful. When people are upset about a high-outrage, low-risk situation, explanations coming from the distrusted source of the trouble may not help much; merely providing risk probability data also may not help much, even if the source is trusted. But considerable reductions in threat perception and action intentions are possible when a trusted, neutral source offers a comparison to background or a chart with a risk ladder, risk comparisons, and an action standard.
(1) For the complete reports of these findings, see Weinstein et al. (1989), Communicating Effectively about Risk Magnitudes and Weinstein et al. (1991), Communicating Effectively about Risk Magnitudes, Phase 2. Both are available from the Center for Environmental Communication, Cook College, Rutgers University, New Brunswick, NJ 08903. For a shorter summary of part of this research, see also Sandman et al. (1994).
(2) Although our earlier research found that action advice at several levels is helpful, the present study included only a single recommended action level. Advice was kept off the chart partly to avoid excessive clutter and partly because action advice for radon is largely a function of mitigation feasibility, a variable that we did not wish to add to the study.
(3) Our prior research was inconclusive on the effectiveness of risk comparisons, but did find that people believe risk explanations are clearer and more helpful when comparisons are given. A study focusing explicitly on risk comparisons is reported in Weinstein and Sandman (1991), ‘Evaluating risk comparisons for use in A Citizen’s Guide to Radon,’ available from the authors. It found virtually no differences among radon risk explanations with various sorts of risk comparisons or no risk comparisons at all. Nonetheless, the earlier decision to include risk comparisons in this study was not reversed.
(4) This article will use Sandman’s term ‘outrage’ to cover the nontechnical aspects of risk. However, the technical aspects (in this case, the quantitative likelihood of illness) will be referred to by the conventional terms ‘risk,’ ‘risk magnitude,’ and ‘risk seriousness,’ rather than Sandman’s ‘hazard.’ ‘Hazard’ will be used instead in the conventional way, to describe the substance or situation that poses a risk. Thus, what Sandman would call a ‘high-outrage, low-hazard risk,’ we will call a ‘high-outrage, low-risk hazard.’
(5) One of the strengths of the prior research design was the use of two risk levels to establish a risk magnitude effect on threat perception and action intentions. The size of this effect could then be directly compared with the size of the other experimental effects – establishing, for example, that the difference between a location one-quarter of the way up the risk ladder and one three-quarters of the way up the ladder had roughly the same effect on perceived threat as a two-fold difference in actual risk. To continue this advantage, we included an ‘alternate risk’ treatment, identical in form to the ‘base risk’ treatment, but with a 10-fold difference in the risk magnitude itself.
(7) In results not reported here, comparisons to normal background were also able to diminish anger. Subjects assigned the high outrage Nuclear Waste conditions were understandably angry. Their anger was reduced far more by the knowledge that the situation posed a risk 200× less than normal background than they were by the knowledge that the risk posed was a mere 1-in-100,000 or even a mere 1-in-1,000,000.
The research summarized here was sponsored by the Office of Policy, Planning and Evaluation and the Office of Radiation Programs, both of the U.S. Environmental Protection Agency. This research is one portion of Part Two of Phase III of Cooperative Agreement CR-814506, entitled ‘Communicating Effectively about Risk Magnitudes.’ The complete report of Phase III, Part Two is available from Peter M. Sandman, as are reports on Phase I, Phase II, and Phase III, Part One.
Covello, V.T. and Allen, F.W. (1988) Seven Cardinal Rules of Risk Communication (pamphlet), Washington DC: U.S. Environmental Protection Agency.
Covello, V.T., Sandman, P.M. and Slovic, P. (1988) Risk Communication, Risk Statistics, and Risk Comparisons, Washington DC: Chemical Manufacturers Association.
Hance, B.J., Chess, C. and Sandman, P.M. (1988) Improving Dialogue with Communities: A Risk Communication Manual for Government, Trenton NJ: Division of Science and Research, New Jersey Department of Environmental Protection.
Hance, B.J., Chess, C. and Sandman, P.M. (1990) Industry Risk Communication Manual, Boca Raton FL: CRC Press/Lewis Publishers.
Johnson, F.R. and Luken, R.A. (1987) Radon risk information and voluntary protection: evidence from a natural experiment, Risk Analysis 7, 97–107.
Kasperson, R.E. (1986) Six propositions on public participation and their relevance for risk communication, Risk Analysis 6, 275–81.
Krimsky, S. and Plough, A. (1988) Environmental Hazards: Communicating Risks as a Social Process, Dover MA: Auburn House.
National Research Council (1989) Improving Risk Communication Washington DC: National Academy Press.
Rohrmann, B. (1990) Analyzing and evaluating the effectiveness of risk communication programs, in Studies on Risk Communication 17, Jülich, Germany: Programmegruppe Mensch-Umwelt-Technik (MUT), Forschungszentrum Jülich.
Sandman, P.M. (1987) Risk communication: facing public outrage, EPA Journal 37, 21–2.
Sandman, P.M. (1991) Risk = Hazard + Outrage: A Formula for Effective Risk Communication (videotape), Fairfax VA: American Industrial Hygiene Association.
Sandman, P.M. (1993) Responding to Community Outrage: Strategies for Effective Risk Communication , Fairfax VA: American Industrial Hygiene Association.
Sandman, P.M., Weinstein, N.D. and Klotz, M.L. (1987) Public response to the risk from geological radon, Journal of Communication 13, 93–108.
Sandman, P.M., Miller, P.M., Johnson, B.B. and Weinstein, N.D. (1993) Agency communication, community outrage, and perception of risk: three simulation experiments, Risk Analysis 13, 589–602.
Sandman, P.M., Weinstein, N.D. and Miller, P. (1994) High risk or low: how location on a ‘risk ladder’ affects perceived risk, Risk Analysis 14, 35–45.
Sjöberg, L. (1989) Radon Risks: Attitudes, Perceptions and Actions (EPA-230-04-89-049), Washington DC: Office of Policy, Planning, and Evaluation, US Environmental Protection Agency.
Slovic, P. (1987) Perception of risk, Science 236, 17 April, 280–5.
Slovic, P., Fischhoff, B. and Lichtenstein, S. (1979) Rating the risks, Environment 21(3), 14–20; 36–9.
Smith, V.K. and Johnson, F.R. (1988) How do risk perceptions respond to information: the case of radon, Review of Economics and Statistics 70(1), 1–8.
Weinstein, N.D., Sandman, P.M. and Hallman, W.K. (1994) Testing a visual display to explain small probabilities, Risk Analysis 14, 895–6.
Weinstein, N.D., Kolb, K. and Goldstein, B.D. (1996) Using time intervals between expected events to communicate risk magnitudes, Risk Analysis 16, 305–8.
Weinstein, N.D. and Sandman, P.M. (1991) Evaluating risk comparisons for use in A Citizen’s Guide to Radon, Washington DC: Office of Policy, Planning and Evaluation and Office of Radiation Programs, US Environmental Protection Agency.
Weinstein, N.D., Sandman, P.M. and Roberts, N.E. (1989) Communicating Effectively About Risk Magnitudes (EPA-230-08-89-064), Washington DC: Office of Policy, Planning and Evaluation, US Environmental Protection Agency.
Weinstein, N.D., Sandman, P.M. and Miller, P. (1991) Communicating Effectively About Risk Magnitudes, Phase 2 (EPA-230-09-91-003), Washington DC: Office of Policy and Evaluation, US Environmental Protection Agency.
Copyright © 1998 by Risk Decision and Policy