Agency Communication, Community Outrage, and Perception of Risk:
Three Simulation Experiments
by Peter M. Sandman, Paul M. Miller,
Branden B. Johnson, and Neil D. Weinstein
Risk Analysis, Vol 13, No. 6, 1993, pp. 585–598
Abstract: Three experimental studies were conducted employing hypothetical news stories to compare the effects on reader risk perceptions of two situations: when agency communication behavior was reported to be responsive to citizens’ risk concerns, vs. when the agency was reported to be unresponsive. In the first two experiments, news stories of public meetings filled with distrust and controversy led to ratings indicating greater perceived risk than news stories reporting no distrust or controversy, even though the risk information was held constant. This effect appeared clearly when the differences in meeting tone were extreme and subjects made their ratings from their recall of the stories, but it was much weaker when the differences were moderate and subjects were allowed to go back over the news stories to help separate risk information from conflict information. In the third experiment, news stories about a spill cleanup systematically varied the seriousness of the spill, the amount of technical information provided in the story, and the agency behavior and resulting community outrage. The outrage manipulation significantly affected affective and cognitive components of perceived risk, but not hypothetical behavioral intentions. Seriousness and technical detail had very little effect on perceived risk.
1. Introduction
Risk assessment and risk management specialists generally agree that the principal standard for
judging and regulating risks should be their relative seriousness – that is, the probability
and magnitude of harm. As laypeople have become increasingly involved in risk decision making, however, they have reached very different judgments than the experts as to which risks most merit public concern and regulatory attention.
Article Table of Contents
One explanation for these public-versus-expert risk controversies holds that they are due to the public’s failure to understand the scientific data. They can thus be avoided by communicating technical information more effectively, especially via the mass media. This view is widely shared by technical experts, and is tacitly accepted by much research documenting the
public’s low “science literacy.” (Note 6) But studies of the link between technical knowledge and support for controversial technologies show mixed results.(2–14)
In stark contrast to the “knowledge breeds support” view is the argument
advanced by Mazur that the more technical information the media provide about a risk
controversy (e.g., nuclear power), the more concerned the audience will be, even if the
information is thought by experts to be reassuring.(15,16) Merely by mentioning potential
problems, the coverage may make those problems seem more likely. A third possibility is that
technical information in the mass media might interact with other attributes of the news story to
affect risk perceptions. For example, technical detail might make a story more credible, thus
heightening the alarm from an alarming story, while confirming the reassurance of a reassuring
story. One test of this hypothesis found no such interaction; it also found no direct effect of
technical detail on readers’ alarm or reassurance.(17)
A competing set of explanations for public-versus-expert risk conflicts holds that the public
responds less to the seriousness of a risk (or its knowledge about seriousness as obtained from the
media) than to such factors as trust, control, fairness, and courtesy. Sandman,(18–20)
Hance et al.,(21,22) and Sandman et al.(23) 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,(24) Bord,(25) Krimsky and
Plough,(26) Johnson,(27) Covello et al.,(28) Covello and Allen,(29) and Slovic.(30)
In Sandman’s terminology, “hazard” is the product of risk magnitude 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.
Supporters of this distinction argue that hazard and outrage are both components of risk
deserving attention, and that laypeople have had as little success communicating what they
consider significant about risks to the experts as the experts have had communicating to the
public. No matter how serious the risk is (in hazard terms), and no matter how much technical
detail is used to explain it, this view maintains that the degree of outrage is likely to
determine much of the public’s response to the risk.
The predominant strategy in much research on risk perception has been to ask people to rate the riskiness of an assortment of hazards, and then to rate the same hazards on several other attributes thought by the investigators to be related to risk perception. Factor analysis or other statistical analysis of the ratings then reveals relationships between risk perception and the hazard attributes under investigation. This methodology omits the social context in which risk judgments are made, although we know that judgments about risk in the abstract can be very different from judgments about specific, personally relevant risk situations.(5, 31,32) Furthermore, when large numbers of risk ratings are factor-analyzed much can be learned about the sources of risk perception, but the imputation of causality is unjustified. Finally, some factors in risk perception, including important outrage variables, are so tied to situations that they simply cannot be studied from lists of hazards. Anecdotal discussions of agency–community and company–community interactions usually focus on such factors as trust, power-sharing, respect for community concerns, openness to community suggestions, and promptness and completeness in releasing risk information. (Note 7) Because they are not characteristics of the hazard itself, but rather of an agency’s or company’s approach to managing the hazard, most of these factors have been difficult to study via the riskiness-ratings methodology.
There are legitimate reasons, of course, why most studies of risk perception have not been
experimental. (There are exceptions, of course; see Slovic.)(33) We cannot experimentally
manipulate the attributes of existing hazardous substances, activities, and technologies. Ethics and
logistics prevent exposing people to hazards varied systematically by attribute – nor do
the environmental problems facing communities cooperate by differing one attribute at a time.
Simulation is one way to take advantage of the inferential power provided by experimental research to study situational variables. In the three studies reported here, an effort was made to create hypothetical hazard situations realistic enough to elicit risk judgments like those that would occur with actual hazards. All three studies examine the same central hypothesis, that manipulation of the reported behavior (in a hypothetical news story) of the organization managing a risk controversy will lead to variations in subjects’ outrage, and therefore to variations in the perceived seriousness of the risk. (Note 8) In the third study, moreover, the outrage effect is experimentally compared with the effects of manipulating hazard seriousness and amount of technical information provided in the story. (Note 9)
2. Study One (Note 10)
2.1. Method
Two mock newspaper stories were written, each with two versions. One story dealt with barrels of chemicals discovered in a community; the other dealt with plans to build a hazardous waste incinerator. In each case a government agency (the state Department of Environmental Protection), rather than a corporation, was responsible for dealing with the issue. In one version of each story, the agency representative willingly shared information, encouraged community residents to form their own opinions, and acknowledged that there was some small risk. In the other version, the agency representative was unwilling to release some types of information, released other facts only after repeated questioning, and suggested that residents were silly to be concerned. Both versions of each story had the same information about the risk itself. The stories were presented in narrow(2–1/2-inch) newspaper columns, two columns to a page. The “barrels” stories were 72 lines long; the “incinerator” stories were 102 lines long.
The study questionnaire asked for judgments about the seriousness of the risk on a 6-point scale (1 = no risk at all; 6 = very serious risk). Next, it asked how trustworthy the agency spokesman seemed (1 = very trustworthy; 4 = not trustworthy at all) and whether the spokesman appeared to be withholding important information (1 = definitely is; 4 = definitely not). The fourth question presented a list of words describing how someone might feel if he or she lived in the community described: angry, helpless, frightened, safe, alarmed, relieved, concerned, pleased, confused, and annoyed. Subjects could choose as many items as they liked to describe how they thought they would feel. Finally, the “incinerator” story questionnaire asked whether the facility should be built (1 = definitely yes; 4 = definitely not). Each questionnaire concluded with demographic items. Extensive pilot testing with college students and nonstudents ensured that none of the situations was viewed as presenting risks so high or so low that ceiling or floor effects would make it impossible to observe differences between versions.
Subjects were a cluster sample of adult residents of single-family homes in East Brunswick, New Jersey, a middle-income to upper-middle-income suburban community. Research assistants went door-to-door to recruit individuals. Only adults at least 18 years of age (one person per household) were eligible to participate. People who agreed were given two news stories, one on the “barrels” and one on the “incinerator,” and a stamped envelope addressed to the “Environmental Information Project” at Rutgers University. All four combinations of the outrage manipulation were used: high on the first story and high on the second, high-low, low-high, and low-low. The two stories appeared in random order. Subjects were asked to read the stories and send back the questionnaire in the next day or two. In order to lower refusal rates, subjects’ names and addresses were not requested. This meant, however, that reminders could not be used to increase response rates from those who agreed to take part.
2.2. Results
A volunteer was found in 83% of the houses visited where an eligible respondent was at home; 71% of these volunteers actually returned their questionnaires, yielding a net response rate of 59%. Nearly all of the 86 respondents (93%) owned their home. The sample was 57% male, 63% had completed a four-year college, and 12% belonged to an environmental group. The mean age was 44.2. (Note 11)
Analyses of variance in the data on seriousness, trustworthiness, and secrecy were done for each story separately, using the variables outrage (high or low) and story reading order (first or second). Emotion checklists were compared by Fisher’s exact test. Story reading order showed no main effects or interactions with outrage, and is not discussed further. Table I shows the outrage manipulation results.
Table I. Effect of Outrage Manipulation on Perceived Risk
(Study One) (N = 86)
| Response |
Low
outrage |
High
outrage |
Significance (a) |
|
| “Barrels” Story |
| Seriousness (b) | 4.18 | 4.59 | .08 | |
| Trust (c) | 2.48 | 3.26 | .0001 | |
| Secrecy (d) | 2.41 | 1.48 | .0001 | |
| Emotion checklist: |
| Angry | 42.5% | 84.8% | .0001 | |
| Relieved | 2.5% | 0.0% | NS | |
| Frightened | 37.5% | 52.2% | NS | |
| Safe | 0.0% | 2.2% | NS | |
| Alarmed | 47.5% | 65.2% | NS | |
| Helpless | 27.5% | 43.5% | NS | |
| Concerned | 72.5% | 67.4% | NS | |
| Pleased | 2.5% | 0.0% | NS | |
| Confused | 10.0% | 30.4% | .04 | |
| Annoyed | 32.5% | 58.7% | .02 | |
“Incinerator” Story
|
| Seriousness (b) | 3.19 | 3.44 | NS | |
| Trust (c) | 2.02 | 2.54 | .03 | |
| Secrecy (d) | 2.46 | 2.37 | NS | |
| Build decision (e) | 2.72 | 2.82 | NS | |
Emotion checklist:
|
| Angry | 19.0% | 50.0% | .004 | |
| Relieved | 14.3% | 25.0% | NS | |
| Frightened | 26.2% | 34.1% | NS | |
| Safe | 9.5% | 9.1% | NS | |
| Alarmed | 33.3% | 54.5% | .06 | |
| Helpless | 7.1% | 0.0% | NS | |
| Concerned | 81.0% | 65.9% | NS | |
| Pleased | 4.8% | 2.3% | NS | |
| Confused | 19.0% | 18.2% | NS | |
| Annoyed | 19.0% | 38.6% | .06 | |
Notes:
(a) Probabilities are based on analysis of variance tests for means and on Fisher’s exact test for percentages.
(b) 1 = no risk at all; 6 = very serious risk.
(c) 1 = very trustworthy; 4 = not trustworthy at all.
(d) 1 = definitely is withholding information; 4 = definitely is not.
(e) 1 = definitely should build; 4 = definitely should not.
|
The two “barrels” stories produced significantly different perceptions of agency trustworthiness and secrecy, as intended, as well as significant differences in anger, confusion, and annoyance. However, the perceived seriousness of the risk was only marginally greater in the high-outrage condition (P < .08 (Note 12)). The “incinerator” stories were less successful in producing different ratings for trust, secrecy, and emotional responses, and yielded no significant difference in perceived seriousness.
Correlations between trust and perceived seriousness were .61 and .62 for the “barrels” and “incinerator” stories, respectively (P’s < .0001). Correlations between perceived agency secrecy and perceived seriousness were .53 and .63 for the two stories (P’s < .0001). These high correlations suggest (but do not demonstrate) that a stronger manipulation of trust and secrecy might have had more impact on risk perception.
3. Study Two (Note 13)
The lack of a strong effect on risk perceptions in Study One was a surprise. Pilot tests had shown
large differences in responses to the two story versions. Also, the strong correlations of perceived
trustworthiness and secrecy with risk judgments suggested that agency (or corporate) actions that
successfully build trust and show openness should strongly affect community risk perception. One
likely explanation for the Study One results is that the experimental outrage manipulation
– the difference between the two hypothetical news stories – was too small to
show this impact clearly. Another possibility is that subjects adopted an atypically rational
orientation to the task, looking back at the articles and noting only those sentences directly
relevant to the risk. Both of these possibilities were addressed by the design used in Study Two.
3.1. Method
Subjects in Study Two were a cluster sample of 156 New Jersey residents who lived in Edison
Township and in adjacent areas of Metuchen Borough and Woodbridge Township. These were
primarily areas of single-family houses. If more than one person per household offered to take
part when visited by a researcher, they were given different versions of the news story.
Only one story was used, a revision of the one dealing with barrels of chemical waste. This
story was selected because it had been much more successful in creating different perceptions of
trust and openness than the story about the incinerator.
The questions on perceived risk seriousness, agency trustworthiness, and withholding of
information were unchanged from Study One. The checklist of emotions retained only the choices
of angry, frightened, safe, concerned, annoyed, and alarmed.
The key change in procedure is that study participants were not permitted to review the story
when answering the questions. Subjects were also asked to complete the questionnaire
immediately instead of returning it by mail.
The story was revised to increase the differences between the two versions. In Study One,
the two versions did not diverge until the third paragraph, and even then the differences required
careful reading. In the excerpts in Display 1, the low-outrage condition is on the left.
Display 1.
State Says Hazwaste
Barrels Not Health Risk
CANTERVILLE – Several hundred barrels of hazardous waste stored near Mill Road in Canterville pose little threat to public health, officials from the state Department
of Environmental Protection said yesterday.
Although some of the barrels are leaking, DEP spokesperson Thomas Nicholas said the leaking materials would not contaminate nearby wells. The area where the leaking barrels are located has dense clay soil, Nicholas explained.
Nicholas spoke at a meeting of local citizens, organized by DEP to explain to the community about the contents of the barrels, the likely health effects, and the plans for cleaning up the
site.
…
State Says Hazwaste
Barrels Not Health Risk
CANTERVILLE – Several hundred barrels of hazardous waste stored near Mill Road in Canterville pose little threat to public health, officials from the state Department
of Environmental Protection said yesterday.
Although some of the barrels are leaking, DEP spokesperson Thomas Nicholas said the leaking materials would not contaminate nearby wells. The area where the leaking barrels are located has dense clay soil, Nicholas explained.
DEP uncovered the barrels two years ago outside a plant abandoned in 1986 by its former owner, the ARC Chemical Company. But DEP did not announce its discovery at that time. The problem finally became public last week, when Councilwoman Gladys Smith told reporters about the leaking barrels.
…
The two versions used for Study Two (Display 2), by contrast, show their differences earlier and more obviously. Once again, the low-outrage condition is on the left.
Display 2.
DEP Helps Council Understand Problems of Waste Site
NEW FALLS, NJ – In a meeting organized by the Department of
Environmental Protection and New Falls residents, a DEP official said that there was no serious
danger from the leaking barrels found in the abandoned hazardous waste site near Main Street.
At the meeting, citizens were able to ask DEP what was going on at the site in an open forum format. Spokesperson Frank Doll started the meeting by assuring the residents that the leaking materials were safe.
“We have analyzed the chemicals and have found them to be a variety of industrial solvents and intermediate chemical products,” said Doll. “Most of them are similar to everyday laundry detergents and are safe.”
Doll passed out a list which contained the names of all the chemicals present. He also gave out a pamphlet which would help people understand what each chemical was used for.
…
Residents Still Uncertain About Status of Waste Site
NEW FALLS, NJ – Townspeople who attended last night’s council meeting left angry and confused even after a Department of Environmental Protection official said that there was no serious danger from the leaking barrels found in the hazardous waste site near Main Street.
Outraged citizens continuously asked spokesperson Frank Doll to release information about the barrels. He offered no definitive answers however, stating that they were no serious health threat and need not be worried about. The exact chemical composition of the materials inside the barrels is still unknown.
“We have analyzed the chemicals and have found them to be a variety of industrial solvents and intermediate chemical products,” said Doll. “Most of them are similar to everyday laundry detergents and are safe. We are not going to make the chemicals known because you probably wouldn’t understand them anyway.”
…
Note that in Study One, “outrage” was operationalized solely in terms of agency communication behavior thought likely to generate an outraged response. In Study Two, by
contrast, the high-outrage version also presented subjects with an explicitly outraged community,
while the low-outrage version showed the community to be calm and cooperative. Study Two
stories thus included two kinds of reported behavior: of the agency spokesperson and of
neighborhood residents. (Such “person in the street” reactions to government
statements are typical of news stories on environmental issues.) These two sets of behaviors may
have joint, separate, or even offsetting effects on risk perception – but it is useful to
determine whether outrage in general affects risk perception before designing studies to tease
apart its constituents.
3.2. Results
Table II shows the Study Two results. The two versions produced almost totally different
reactions. As desired, and as in Study One, the agency spokespeople were judged to be very
different in trustworthiness and secrecy. Subjects were also more likely to express anger in the
high-outrage condition than the low-outrage condition. More importantly, and unlike Study One,
the manipulation had a powerful impact on subjects’ risk perceptions as well. When the
agency was depicted as untrustworthy and secretive and the community was depicted as outraged,
subjects rated the risk as much more serious and their responses to the risk as much more
frightened and less safe. When the agency and community were depicted as mutually respectful
and cooperative, on the other hand, subjects saw the risk as much less serious, and described their
own reactions as concerned rather than frightened or unsafe. Despite identical technical
information about the risk, “outrageous” agency behavior and an outraged
community strongly influenced perceived risk. (Note 14)
Table II. Effect of Outrage Manipulation on
Perceived Risk (Study Two) (N = 156)
| Response |
Low
outrage |
High
outrage |
Significance (a) |
| Seriousness (b) | 2.65 | 4.96 | .0001 |
| Trust (c) | 1.77 | 3.22 | .0001 |
| Secrecy (d) | 3.06 | 1.64 | .0001 |
Emotion checklist:
|
| Angry | 1.2% | 68.0% | .0001 |
| Frightened | 2.6% | 39.7% | .0001 |
| Safe | 24.4% | 1.2% | .0002 |
| Alarmed | 2.6% | 10.0% | NS |
| Concerned | 69.2% | 39.7% | .0004 |
| Annoyed | 20.5% | 23.1% | NS |
(a) Probabilities are based on analysis of variance tests for
means and on Fisher’s exact test for percentages.
(b) 1 = no risk at all; 6 = very serious risk.
(c) 1 = very trustworthy; 4 = not trustworthy at all.
(d) 1 = definitely is withholding information; 4 = definitely is not. |
4. Study Three (Note 15)
To clarify the impact of the outrage manipulation, Study Three manipulated three experimental variables: outrage, the seriousness of the risk itself, and the amount of technical detail provided in the news story. This article focuses on the results for the outrage manipulation; results for the technical detail manipulation are reported elsewhere.(37) Because Study Three was more elaborate than the two earlier studies, with a bigger sample and a more complex design, it will be discussed in more detail. To aid in the discussion, variable names appear in capital letters.
4.1. Method
4.1.1. Manipulations
The hypothetical news stories developed for this study focused on a spill of perchioroethylene (PERC) and the resulting response by the state Department of Environmental Protection. The story versions were developed in consultation with substantive specialists at the New Jersey Department of Environmental Protection. Care was taken to assure that the differences between the “low” and “high” treatments were great enough to fairly test the impact of each variable, but not so great as to be unrealistic in their simulation of agency, journalistic, or chemical behavior.
The seriousness manipulation varied the estimated toxicity of perchloroethylene, the
estimated exposures resulting from the spill, and the number of people exposed. It is appropriate
to multiply these sources of variation to get the overall difference in seriousness (not necessarily
perceived seriousness) between the two treatments: The high-seriousness condition was about
five orders of magnitude (a hundred thousand times) worse than the low-seriousness condition.
The technical detail manipulation consisted of several additional paragraphs of information on
exposure pathways and toxicological studies, absent in the low-technical detail condition and
present in the high-technical detail condition.
The outrage manipulation was more extreme than in Study One, but much less extreme than
in Study Two, with its arguably unrealistic depiction of agency behavior in the high-outrage
condition. As in Study Two, reported community outrage, not just the agency
spokesperson’s behavior, was manipulated, but the manipulations were less extreme. Four
typical paragraphs appear in Display 3; the low-outrage condition is on the left.
Display 3.
“We will certainly want to take another look at the regulations,” Chester said. “Perhaps the agency should consider tougher standards for lightning protection.”
Chester said DEP would be developing plans to test area wells for PERC. “At this point I wouldn’t really expect any wells to be seriously contaminated,” Chester said. “But we still want to test to be sure.”
Clara Stevenson, whose home is the closest one to the site of the spill, said she was
“impressed” by DEP’s promise to test her well. “I’m much less
upset now that I have talked to the DEP people,” she said.
“Soon after I woke up there was a DEP person at my door explaining what happened and what the clean-up crews were doing,” said Maple Ridge resident Alex Sands.
“It looks like a fluke to me,” Chester said. “As far as I know, DEP has no plans to re-examine the regulations. You can’t cover every conceivable
event.”
Chester said DEP had no plans to test area wells for PERC. “At this point I wouldn’t really expect any wells to be seriously contaminated,” Chester said.
“People who want to be sure will have to make their own arrangements.”
Clara Stevenson, whose home is the closest one to the site of the spill, said she was “furious” about DEP’s unwillingness to test her well. “My whole family is upset and the DEP people just don’t seem to care,” she said.
“I have no idea what happened or what they’re doing about it, and nobody from DEP is taking the time to tell me,” said Maple Ridge resident Alex Sands.
The study focused on two questions: the extent to which agency behavior and community outrage increase people’s risk perception of low-risk events, and the extent to which
technical detail decreases people’s risk perception of low-risk events. The outrage and
technical detail manipulations were presented at both low and high levels, in a 2 × 2 design, with
seriousness kept low. In addition, a fifth high-seriousness condition was included (with outrage
and technical detail both kept low) to assess the magnitude of outrage and technical detail effects
compared to seriousness effects.
4.1.2. Instrument
The instrument included 13 questions. (Note 16) One manipulation check was used for each of the three experimental variables. All were 6-point Likert-type items, with a seventh option of “no opinion.” For serious risk (the check on the seriousness manipulation), subjects were asked: “What is your impression of how serious this situation is?” For perceived detail (the check on the technical detail manipulation), they were asked: “How detailed was the information in the story about the health effects of the PERC spill and the ways people might get exposed?” For perceived appropriateness (the check on the outrage manipulation), they were asked, “How appropriate was DEP’s handling of the PERC spill?”
Three response measures were used to assess perceived risk – one affective, one cognitive, and one behavioral. Once again, 6-point Likert-type scales were used, with a “no opinion” option as well. Subjects were asked: “If you lived in the area, how worried would you be about the risk from the PERC spill?” (worry); “How important do you consider the risk posed by this situation?” (important risk); and “If you lived in the area, how willing would you be to spend $500 to have your water tested for PERC after the spill?” (intention to test).
worry and important risk correlated +.72, but the correlations between worry and intention to test and between important risk and intention to test were both only +.21, statistically significant but small. Although serious risk was originally conceived as a check on the seriousness manipulation, correlations between serious risk and worry and between serious risk and important risk were +.65 and +.67, respectively. worry , important risk, and serious risk were therefore collapsed into a single index variable called perceived risk, with intention to test, the behavioral measure, kept separate. Combining worry, important risk, and serious risk into a composite index provided a more reliable (coefficient 
= .78) and therefore more sensitive response measure.
Two pilot studies for Study Three (using university student subjects and just the high-high-high and low-low-low treatments) had shown that a general measure of risk aversion was
significantly related to the risk perception response measures, and that use of risk aversion as a
covariate improved the sensitivity of the analysis. Four items developed by Weinstein(31,38) were therefore included to assess subjects’ risk aversion. Each item consisted of a statement about environmental risk with which subjects were asked to rate their agreement or disagreement on 7-point Likert-type scales. Choices ranged from very strongly disagree to very strongly agree; statements were phrased so that low ratings indicated acceptance of risks and high ratings indicated risk aversion.
As in Weinstein’s studies,(31,38) two dimensions of risk aversion –
societal risk aversion and personal risk aversion – were assessed, using two items each. A risk aversion score was derived for each dimension by adding the ratings for the two items measuring that dimension. The societal risk aversion items asked subjects about their agreement with two statements: “The public has the right to demand zero pollution from industry” and “An industry that pollutes should not be allowed to stay open, no matter how little pollution it produces.” The personal risk aversion items measured people’s agreement with two statements: “If there was even the slightest amount of asbestos in my home, I would remove it” and “I try to avoid all food additives and preservatives.” Although the two risk aversion variables turned out to have a + .54 correlation, their correlations with other variables in the study had quite different patterns, justifying the decision to keep them separate.
Data on three demographic variables, sex, age, and education, were also collected.
4.1.3. Subjects and Procedures
Subjects were 595 New Jersey residents over 18 years old from middle-income residential areas in
Middlesex county. An additional 81 questionnaires were unusable because of incomplete
responses on the dependent measures.
Using a prepared script, trained interviewers obtained a cluster sample, canvassing every
home in an identified area. To make sure the sample was balanced by age and sex, interviewers
alternated the type of subject asked for at the door between oldest/youngest (over 18) and
male/female. Half of the subjects received the story, then the six-item survey instrument, and
finally the risk aversion/demographic questionnaire; the other half received the risk
aversion/demographic questionnaire first, then the story, then the survey instrument. No order
effects were found, and this variable will not be discussed further. All subjects were asked to
return the story before receiving the survey, to prevent them from rereading the story in search of
the “right” answers.
4.2. Results
4.2.1. Effects of Outrage
Table III shows results for the outrage variable. The outrage manipulation was successful.
Subjects who read high-outrage news stories saw agency behavior as much less appropriate than
subjects who read low-outrage stories; the difference was more than a scale point on the 6-point
scale, significant at P < .0001. (note 17) outrage correlated with perceived appropriateness at .41 (P < .0001).
Table III. Response Measure Means, Standard Deviations, and Significance Tests for Mean Differences Between Outrage Conditions (Study Three)
| Variable |
Low
Outrage |
High
Outrage |
Mean
difference |
F-value
significance |
| Perceived risk |
| (Mean) | 13.42 | 14.18 | 0.76 | F(1,495) = 6.99 |
| (SD) | 3.30 | 3.06 | | ** |
| (N) | 254 | 245 | | |
| Intention to test |
| (Mean) | 3.16 | 3.36 | 0.20 | F(1,469) = 1.51 |
| (SD) | 1.77 | 1.81 | | NS |
| (N) | 240 | 233 | | |
| Perceived appropriateness |
| (Mean) | 3.71 | 2.50 | -1.21 | F(1,495) = 99.79 |
| (SD) | 1.28 | 1.42 | | **** |
| (N) | 254 | 245 | | |
| Perceived detail |
| (Mean) | 3.69 | 3.30 | -0.39 | F(1,495) = 10.61 |
| (SD) | 1.25 | 1.39 | | ** |
| (N) | 254 | 245 | | |
| *P < .05; **P < .01; ***P < .001; ****P < .0001.
|
As predicted, outrage had a significant, if small, effect on perceived risk (P < .01). (Note 18) Subjects who read high-outrage stories saw the risk as more important, serious, and worrisome than did those who read low-outrage stories. Outrage did not significantly affect intention to test, however.
An interesting and unexpected finding was the small but significant effect of outrage on
perceived detail. (Note 19) Subjects who read high-outrage stories judged that they had significantly less technical detail than subjects who read low-outrage stories (P < .01). As we shall see, the actual amount of technical detail in the stories had no significant effect on perceived detail. This suggests that if an agency or company behaves satisfactorily otherwise, people tend to judge that it is providing enough information as well, while if its behavior is improper or offensive the information given is more likely to be thought insufficient. Perhaps “outrageous” agency behavior makes people distrust the
technical detail coming from the agency, or distracts them from the detail actually present, or
makes them require more detail then they would have required had agency behavior been more responsive.
4.2.2. Effects of Technical Detail
No significant effects of the technical detail manipulation were found for any dependent measure.
(Indeed, technical detail had no significant correlations with any other variable in the study.) Even
the manipulation check, perceived detail, showed no effect. This is consistent with the findings of the two pilot studies, where subjects saw some individual paragraphs as more detailed in the “high” condition than in the “low” condition, but saw no differences in detail, clarity, or informativeness between the overall news stories.
The technical detail manipulation was intentionally kept within the range of journalistic
possibility. More extreme variations might be more visible to readers, and have more impact on
their risk perceptions. (And more extreme variations are feasible in other formats, such as
brochures or interpersonal interactions.) But within the range tested, variation in technical detail
had no effect on perceived risk, intention to test, or even perceived detail. (Note 20)
4.2.3. Effects of Seriousness
In four of the five cells in this study, the “objective” level of risk seriousness was
quite low. To test whether subjects had greater perceived risk and greater intention to test when objective risk was greater, a fifth cell was included with low outrage and low technical detail, but much higher seriousness. Analysis of this fifth cell compared it to the low-outrage, low-technical detail, low-seriousness cell.
Mean ratings for serious risk were marginally higher in the high-seriousness condition than in the low-seriousness condition (the difference between the means was less than a third of a standard deviation, P < .05). (Note 21) The manipulation worked, in other words, but just barely; subjects who read a news story reporting a substantially more serious risk perceived it to be slightly more serious than those who read the low-risk news story. Note that subjects in the two pilot studies accurately reported exposure, toxicity, and related factors to be higher when they were in fact higher (P’s < .0001). Thus, the small effect of manipulated seriousness on perceived seriousness probably is not due to any failure to detect the manipulation. Rather, people apparently see the seriousness of a risk as more than the outcome of such factors as exposure and toxicity. The high correlations of serious risk with worry and important risk, which led to its inclusion in the perceived risk composite index variable, underscore the point. The seriousness manipulation had no effect on worry, important risk, or perceived risk, nor on intention to test.
Mean ratings of perceived appropriateness were somewhat
lower in the high-seriousness than in the low-seriousness condition, a result also significant at P < .05. In other words, when the risk reported was more serious, subjects saw the agency’s behavior as less appropriate than when the risk was lower. Though this finding was not predicted, it is not surprising that people should expect more from the agency when the risk is more serious, and thus find the same agency behavior less acceptable in the high-seriousness condition. (Note 22) The objective risk, in short, has less effect on the public’s perception of risk than it has on the public’s perception of agency response.
4.2.4. Regression Analyses of the Response Variables
Regression analyses were run for perceived risk and intention to test, using data
from the 2 × 2 study (excluding the fifth cell testing the seriousness manipulation). Results of
both analyses appear in Table IV.
Table IV. Standardized Regression Coefficients and Significance Tests for Models Predicting Perceived Risk and Intention to Test (Study Three)
| Overall Model Tests |
Response
measure |
Adjusted squared
multiple correlation |
F-value |
Significance |
| Perceived risk | 0.25 | F(9475) = 19.14 | **** |
| Intention to test | 0.14 | F(10,449) = 8.69 | **** |
| Unique Contribution Tests |
| Test |
Standard regression
coefficient |
Unique
variance |
F-value |
Signif-
icance |
| Perceived risk |
| Age | -0.04 | 0.00 | F(1475) = 0.82 | NS |
| Education | -0.11 | 0.01 | F(1475) = 8.01 | ** |
| Sex | 0.12 | 0.01 |
F(1475) = 8.89 | ** |
| Societal risk aversion | 0.32 | 0.07 | F(1475) = 45.37 | *** |
| Personal risk aversion | 0.07 | 0.00 |
F(1475) = 1.82 | NS |
| Perceived appropriateness | -0.23 | 0.04 |
F(1475) = 23.65 | *** |
| Perceived detail | 0.15 | 0.02 |
F(1475) = 11.15 | *** |
| Outrage | 0.02 | 0.00 |
F(1475) = 0.21 | NS |
| Technical detail | -0.02 | 0.00 |
F(1475) = 0.31 | NS |
| Intention to test |
| Age | -0.05 | 0.00 |
F(1448) = 1.43 | NS |
| Education | 0.27 | 0.07 |
F(1448) = 35.91 | **** |
| Sex | -0.03 | 0.00 |
F(1448) = 0.34 | NS |
| Societal risk aversion | 0.02 | 0.00 |
F(1448) = 0.17 | NS |
| Personal risk aversion | 0.22 | 0.03 |
F(1448) = 17.16 | **** |
| Perceived appropriateness | -0.05 | 0.00 |
F(1448) = 0.79 | NS |
| Perceived detail | 0.03 | 0.00 |
F(1448) = 0.41 | NS |
| Perceived risk | 0.15 | 0.02 |
F(1448) = 8.44 | ** |
| Outrage | 0.03 | 0.00 |
F(1448) = 0.50 | NS |
| Technical detail | -0.01 | 0.00 |
F(1448) = 0.01 | NS |
| *P < .05; **P < .01; ***P < .001; ****P < .0001. |
Both models found significant multiple correlations between response measures and predictor variables, with the strongest relationship for perceived risk (adjusted R2 = .25, P < .0001). This is not a strong relationship; clearly many factors other than those measured in this study affect perceived risk. The relationship for intention to test was still weaker (adjusted R2 = .14, P < .0001).
The patterns of prediction were substantially different. For perceived risk, the strongest predictor in terms of uniquely contributed variance was societal risk aversion (about 7% of the variance, P < .001), followed by perceived appropriateness (4% of the variance, P < .001). (Note 23) The higher the societal risk aversion and perceived outrage (lower perceived appropriateness of the agency response), the higher was the perceived risk. In addition, higher perceived risk was associated with significantly higher perceived detail, lower education, and female rather than male subjects.
Neither outrage nor technical detail made a significant unique contribution to the variance in perceived risk, although their respective manipulation checks, perceived appropriateness and perceived detail, did. When the two manipulation checks were dropped from the regression analysis, technical detail still had no significant unique effect on perceived risk. But outrage then was a significant predictor (P < .05), with a regression coefficient of .10, accounting uniquely for just under 1% of the variance.
For the intention to test variable, only three predictors made significant unique contributions: education (7% of the variance, P < .0001), personal risk aversion (3% of the variance, P < .0001), and perceived risk (2% of the variance, P < .01). The more educated and the more averse to personal risk-taking individuals were, the greater their inclination to test. (Note 24) In addition, those who saw the risk in the story as more serious, important, and worrisome (the three components of the perceived risk composite index variable) were more inclined to test.
5. Discussion
The three studies reported here show that the outrage variable has a significant but not overwhelming effect on risk perception.
Study One found a very weak relationship for one news story and none for the other. But the
sample was small, the dependent variable was a single question, and the differences between the
stories were subtle. There was a strong correlation for both stories between perceived agency
trustworthiness and secrecy and the perceived seriousness of the risk. But the outrage
manipulation did not strongly affect perceived trustworthiness and secrecy. In Study Two, the
sample was larger and the story differences much more extreme. In addition, subjects were not
permitted to review the story to determine their answers. Strong relationships emerged between
the experimental outrage manipulation and perceived seriousness.
Study Three had a much larger sample and a much more sensitive design, with risk aversion as a covariate and a 3-item index of perceived risk (seriousness, worry, importance). The stories were less obviously different than in the second study, but more so than in the first. Here the outrage manipulation significantly affected perceived risk, though not the one-item measure of intention to test. By contrast, a seriousness manipulation of roughly five orders of magnitude barely affected perceived seriousness and did not affect other components of perceived risk. And experimental manipulation of the amount of technical detail in the story did not significantly affect any dependent variables.
Among the three variables examined in these studies, in other words, outrage was the most
powerful in its impact on risk perception. The studies suggest that an agency or company that
deals responsively, openly, and respectfully with concerned citizens, and succeeds in avoiding
hostile public reactions, is likely to reduce risk perceptions by doing so – much more than
by providing technical information or even by reducing the technical risk by several orders of magnitude.
Nonetheless, the regression analysis in Study Three shows that outrage is a significant but by
no means a strong predictor of risk perception, much less of self-protective behavior. Education, sex, and risk aversion – all factors beyond the control of the agency or corporate communicator– are more potent still. And all the factors assessed in the research reported here together accounted for relatively small percentages of the variance in perceived risk, and still smaller percentages of the variance in behavioral intentions. Clearly, many other factors, as yet unknown, are at work.
In fact, the outrage manipulation was not significantly related to perceived risk in the
regression analysis except when perceived appropriateness was omitted. Whether people
consider an agency’s or a company’s behavior outrageous seems to matter a good deal in risk perception. How closely the public’s view of agency behavior tracks
actual agency–community interactions remains to be determined.
The use of hypothetical news stories adds three more caveats. Study Three compared effects
of outrage as reflected in news stories with technical detail as given in news stories. Other, more
personal vehicles might work very differently. People who attend a public meeting, receive an
informational brochure, or telephone an agency with questions can acquire far more technical
detail than the few extra paragraphs of the high-technical detail condition in Study Three –
and they acquire it in a very different setting. Similarly, each of these settings might convey
agency responsiveness or unresponsiveness and community acceptance or outrage very
differently. The effects of outrage vis-à-vis technical detail and other variables
need to be studied in contexts other than newspaper journalism.
The second caveat concerns the fact that the news stories in all three studies were
hypothetical. Subjects were asked to imagine that the stories had appeared in their local
newspapers and that their own communities were faced with the situations described. 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 on risk perception were suppressed
by the hypothetical nature of the study, while the effects of seriousness and technical detail were
more likely augmented – that is, we would expect subjects to be more attentive to data
and less liable to outrage in these studies than they would be in a real situation. But no research
findings back this supposition.
Finally, real community hazard situations develop over days, months, or even years; the simulations compress these histories into written materials that take only a few minutes to read. Once again, we consider it likely that the distortion is conservative, that prolonged exposure to a risk controversy makes people more responsive to outrage and less responsive to seriousness and technical detail than they were in this research. Yet no studies demonstrate or dispute this
point either.
Note also that the research reported here treats outrage as a single variable, though it is in fact a cluster of related – and perhaps not so closely related – variables. (Note 25) Among the factors varied in the hypothetical news stories were agency secretiveness/openness, agency courtesy/contemptuousness, agency responsiveness/unresponsiveness to community input, etc. In Study Two and Study Three, the community’s reported response (angry, suspicious, and frightened or grateful, trusting, and calm) was also varied. These factors are all conceptually distinguishable from one another. Furthermore, Sandman and colleagues have applied the term “outrage” to a far wider range of variables, including less interactional ones like voluntariness, familiarity, dread, and the like.(18–22) To develop a powerful explanatory model of the effects of outrage on risk perception, these variables must be teased apart experimentally (not just through factor analysis of survey data) to measure their effects independently.
Nonetheless, the evidence so far suggests strongly that the outrage cluster (communicator
behavior, community response, and the communicator–community interaction) has a
substantial impact on people’s perception of risk. As government agencies and
corporations struggle to reassure communities about risks that represent small threats to health
and environmental quality, much that determines the public response is beyond the risk
manager’s control: risk aversion, demographics, etc. But how risk managers interact with
communities is very much in their control. Further research is needed to guide this interaction, to
help risk managers avoid exacerbating outrage in the public’s response to low-consequence hazards.
Notes
Notes 1 through 5 were old addresses and contact information for the authors, and were
deleted from this version.
6. See Miller,(1) for example.
7. See for example Refs. 19–22.
8. Note that the term “outrage,” used strictly, should refer to the public’s response to a risk or to the behavior of risk managers; it should not refer to characteristics of the risk or the management behavior themselves. Nonetheless, throughout this paper the variations in reported agency behavior (open vs. secret, compassionate vs. contemptuous, etc.) will be referred to as the outrage manipulation.
9. Both the notion that technical information in the media affects audience risk perception and the notion that media “outrage information” affects risk perception share a focus on how journalists approach risk controversies. The research literature on media coyerage of risk is beyond the scope of this article. For a “manual” on how sources attempt to influence risk coverage, see Sandman et al.(34)
10. This study was conducted by Neil D. Weinstein and Peter M. Sandman. The assistance of Hannah Vo Dinh and Katherine Curcio in collecting the data is gratefully acknowledged. For a more complete analysis and copies of the materials and instrument, see Weinstein.(35)
11. The study reported here was conducted simultaneously with a study assessing the effects on risk perception of individual vs. societal responsibility for a hazard and of existing vs. newly proposed hazards. Subjects received either one set of materials or the other at random. The data on response rate and demographics apply to the combined samples for the two studies.
12. All statistical analyses reported in this paper were two-tailed.
13. This study was carried out by Patrick H. Bivona, David P. Cho, John D’Angelo, and Christine D. Garcia under the direction of Peter M. Sandman. For a more complete analysis and copies of the materials and instrument, see Ref. 35.
14. It is of course possible that in the high-outrage condition readers were less inclined than in the low-outrage condition to believe the technical information provided by the agency. This replicates a familiar pattern in risk controversies, where the key technical information often comes from sources who are also managing the risk, and whose courtesy, compassion, openness, and the like may determine whether the technical information is accepted. A study in which the risk information came from neutral third parties would be useful.
15. This study was carried out by Peter M. Sandman, Paul M. Miller, and Branden B. Johnson. Grateful acknowledgement is made to: Caron Chess and Kandice L. Salomone, who provided critical assistance in the development of the news stories and the design and interpretation of two pilot studies; JoAnn M. Valenti, who assisted in the development of the news stories; Neil D. Weinstein, who advised on the development of the instruments; and Jennifer Field, who coordinated the data collection and data entry. An advisory committee at the New Jersey Department of Environmental Protection provided counsel on technical accuracy and realism in the depiction of agency behavior. For a more complete analysis and copies of the materials and instruments used, see Sandman and Miller.(36)
16. Two pilot studies were conducted, using student subjects and only two versions of the story (high-high-high and low-low-low). Results of the pilot studies led to changes in the text of the story, in the conceptualization of the independent variables, and in the measurement of both independent and dependent variables. For a complete discussion of the pilot study methods and findings, see Ref. 36, pp.15–41.
17. Cohen(39) and Cohen and Cohen(40) note that the size of an effect can be measured by dividing the difference between the means by the standard deviation; they suggest the convention that a quotient of .2 represents a small effect, .5 is a medium effect, and .8 is a large effect. By this standard, the effect of the outrage manipulation on perceived appropriateness is large.
18. By the standard suggested by Cohen(39) and Cohen and Cohen(40) (described in the previous note), this effect is small.
19. This effect, too, would be considered small by the standard of the two previous notes.
20. See Ref. 37 for a fuller discussion of alternative hypotheses and research topics on technical detail.
21. See note 17.
22. There was no high-seriousness high-outrage condition in the study reported here, so it is impossible to determine whether there is an interaction effect of seriousness and outrage. It is conceivable that the failure to prevent a serious risk is seen by many people as inappropriate and outrageous agency performance by definition. Even if the agency handles a high-risk incident superbly, the mere fact that the incident occurred may lead to low perceived appropriateness; this would constitute, in essence, a floor for outrage when seriousness is high.
23. The unique proportion of variance accounted for by each independent variable was computed from standardized partial regression weights using the formula provided in Cohen and Cohen,(40) p. 483.
24. It is interesting that societal risk aversion was so closely connected to perceived risk, while personal risk aversion played an important role in intention to test. Considering an environmental risk serious, important, and worrisome is apparently tied to values such as corporate environmental accountability. Actually intending to do something about the risk, however, seems to have a closer tie to other self- protective behaviors, such as avoiding food additives and cleaning up home asbestos. To some extent self-protective behavior may be a personal characteristic that cuts across the many distinctions among risks. Radon research, for example, has found a stronger relationship between radon testing and personal risk aversion than between radon testing and societal risk aversion – or, indeed, between radon testing and radon knowledge.(41) See also Wildavsky and Dake(42) for a finding that people with “egalitarian” views were societally more risk-averse than others, but personally tended to be risk-takers.
25. Note that seriousness and technical detail are also clusters. Seriousness, for example, includes probability, magnitude, exposure, etc. Technical detail includes various sorts of content (detail on exposure, toxicity, epidemiology, etc.), as well as variations in tone, clarity, and the like. See Ref. 37.
Acknowledgments
The authors are grateful to the Division of Science and Research of the New Jersey Department of Environmental Protection (now the New Jersey Department of Environmental Protection and Energy) and the New Jersey Agricultural Experiment Station for funding the research discussed here. The views expressed in this article are those of the authors and do not necessarily reflect the views of the funding agencies.
References
1. J. D. Miller, “Scientific Literacy: A Conceptual and Empirical Review,”
Daedalus (Spring 1983), p. 29.
2. B. D. Melber, S. M. Nealey, J. Hammersla, and W. L. Rankin, Nuclear Power and the
Public: Analysis of Collected Survey Research (Seattle, Battelle Human Affairs Research
Center, 1977).
3. J. H. Kuklinski, D. S. Metlay, and W. D. Kay, “Citizen Knowledge and Choices on
the Complex Issue of Nuclear Energy,” American Journal of Political Science (1982), pp. 615–642.
4. S. M. Nealey, B. D. Melber, and W. L. Rankin, Public Opinion and Nuclear Energy (Lexington, Massachusettes, D. C. Heath, 1983).
5. R. J. Bord and R. E. O’Connor, “Risk Communication, Knowledge, and
Attitudes: Explaining Reactions to a Technology Perceived as Risky,” Risk Analysis 10, 499–506 (1990).
6. R. J. Bord and R. E. O’Connor, “Determinants of Risk Perceptions of a
Hazardous Waste Site,” Risk Analysis 411–416 (1992).
7. B. R. N. Baird, “Tolerance for Environmental Health Risks: The Influence of
Knowledge, Benefits, Voluntariness, and Environmental Attitudes,” Risk
Analysis 425–435 (1986).
8. B. B. Johnson and B. Baltensperger, “Community Risk Perception: A Pilot
Study,” in L. B. Lave (ed.), Risk Assessment and Management (New York,
Plenum, 1987), pp. 337–344.
9. D. Golding, S. Krimsky, and A. Plough, “Evaluating Risk Communication: Narrative vs. Technical Presentations of Information about Radon,” Risk Analysis 27–35 (1992).
10. A. J. Wyner and D. E. Mann, Seismic Safety Policy in California: Local Governments and Earthquakes (Report to the National Science Foundation, Washington, D.C., U.S. Department of Commerce, 1983).
11. B. B. Johnson, “Advancing Understanding of Knowledge’s Role in Lay
Risk Perception,” RISK: Issues in Health and Safety 189–212 (1993).
12. F. R. Johnson and A. Fisher, “Conventional Wisdom on Risk Communication and
Evidence from a Field Experiment,” Risk Analysis 209–213 (1989).
13. N. D. Weinstein and P. M. Sandman, “A Model of the Precaution Adoption Process: Evidence from Home Radon Testing,” Health Psychology 170–180 (1992).
14. N. D. Weinstein and P. M. Sandman, “Predicting Homeowners’ Mitigation
Response to Radon Test Data,” Journal of Social Issues 63–83 (1992).
15. A. Mazur, “Media Coverage and Public Opinion on Scientific
Controversies,” Journal of Communication 106–115 (1981).
16. A. Mazur, “Nuclear Power, Chemical Hazards, and the Quantity of
Reporting,” Minerva 294–323 (1990).
17. K. L. Salomone, News Content and Public Perceptions of Environmental Risk: Does
Technical Risk Information Matter After All? (New Brunswick, New Jersey, Environmental
Communication Research Program, 1992).
18. P. M. Sandman, “Risk Communication: Facing Public Outrage,” EPA
Journal 21–22 (1987).
19. P. M. Sandman, “Risk = Hazard + Outrage: A Formula for Effective Risk
Communication” (videotape) (Akron, Ohio, American Industrial Hygiene Association,
1991).
20. P. M. Sandman, Responding to Community Outrage: Strategies for Effective Risk Communication 
(Fairfax, Virginia, American Industrial Hygiene Association, 1993).
21. B. J. Hance, C. Chess, and P. M. Sandman, Improving Dialogue with Communities:
A Risk Communication Manual for Government (Trenton, New Jersey, Division of Science
and Research, New Jersey Department of Environmental Protection, 1988).
22. B. J. Hance, C. Chess, and P. M. Sandman, Industry Risk Communication Manual (Boca Raton, Florida, CRC Press/Lewis Publishers, 1990).
23. P. M. Sandman, N. D. Weinstein, and M. L. Klotz, “Public Response to the Risk
from Geological Radon,” Journal of Communication 93–108 (1987).
24. R. E. Kasperson, “Six Propositions on Public Participation and Their Relevance
for Risk Communication,” Risk Analysis 275–281 (1986).
25. R. J. Bord, “Public Cooperation as a Social Problem: The Case of Risky
Wastes” (Annual Meeting of the American Association for the Advancement of Science,
February 14–18, 1987, Chicago).
26. S. Krimsky and A. Plough, Environmental Hazards: Communicating Risks as a
Social Process (Dover, Massachusetts, Auburn House, 1988).
27. B. B. Johnson, “Accounting for the Social Context of Risk Communication,” Science and Technology Studies 103–111 (1987).
28. V. T. Covello, P. M. Sandman, and P. Slovic, Risk Communication, Risk Statistics, and Risk Comparisons (Washington, D.C., Chemical Manufacturers Association, 1988).
29. V. T. Covello and F. W. Allen, “Seven Cardinal Rules of Risk Communication” (U.S. Environmental Protection Agency, Washington D.C., April 1988).
30. P. Slovic, “Perception of Risk,” Science 280–285 (1987).
31. N. D. Weinstein, Public Perception of Environmental Hazards: Attitudes of the Public and the Department of Environmental Protection Toward Environmental Hazards (Trenton, New Jersey, Division of Science and Research, New Jersey Department of Environmental Protection, 1989).
32. L. Sjöberg, “Perceived Risk, Risk Object, and Perceived Control” (annual
meeting of the Society for Risk Analysis, December 6–9, 1992, San Diego).
33. P. Slovic, “The Role of Trust in Risk Perception and Risk Management”
(annual meeting of the Society for Risk Analysis, December 6–9, 1992, San Diego).
34. P. M. Sandman, D. B. Sachsman, and M. R. Greenberg, The Environmental News
Source: Providing Environmental Risk Information to the Media (New Brunswick, New
Jersey, Environmental Communication Research Program, Rutgers University, 1992).
35. N. D. Weinstein, “Simulation Studies of Hazard Perception” (Division of
Science and Research, New Jersey Department of Environmental Protection, Trenton, New
Jersey, March 1, 1989).
36. P. M. Sandman and P. M. Miller, “Outrage and Technical Detail: 'The Impact of
Agency Behavior on Community Risk Perception” (Division of Science and Research, New
Jersey Department of Environmental Protection, January 1991).
37. B. B. Johnson, P. M. Sandman, and P. M. Miller, “Testing the Role of Technical
Information in Public Risk Perception,” RISK – Issues in Health and Safety 341–364 (1992).
38. N. D. Weinstein, Public Perception of Environmental Hazards: Statewide Poll of Environmental Perceptions (Trenton, New Jersey, Division of Science and Research, New Jersey Department of Environmental Protection, 1987).
39. J. Cohen, Statistical Power Analysis for the Behavioral Sciences, 2nd. ed. (Hillsdale, New Jersey, Lawrence Erlbaum, 1988).
40. J. Cohen and P. Cohen, Applied Multiple Regression/Correlation Analysis for the
Behavioral Sciences, 2nd. ed. (Hillsdale, New Jersey, Lawrence Erlbaum, 1983).
41. N. D. Weinstein, P. M. Sandman, and N. E. Roberts, Public Response to the Risk
from Radon, 1988-1989 (Trenton, New Jersey, Division of Environmental Quality, New
Jersey Department of Environmental Protection, November 1989).
42. A. Wildavsky and K. Dake, “Theories of Risk Perception: Who Fears What and
Why?” Daedalus (Fall 1990), pp. 41–60.
Copyright © 1993 by Risk Analysis
