You are on page 1of 17

The Different Dynamics of Personal and General Risk

Author(s): Lennart Sjöberg


Source: Risk Management, Vol. 5, No. 3 (2003), pp. 19-34
Published by: Palgrave Macmillan Journals
Stable URL: http://www.jstor.org/stable/3867764
Accessed: 04-12-2015 01:06 UTC

REFERENCES
Linked references are available on JSTOR for this article:
http://www.jstor.org/stable/3867764?seq=1&cid=pdf-reference#references_tab_contents

You may need to log in to JSTOR to access the linked references.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.

Palgrave Macmillan Journals is collaborating with JSTOR to digitize, preserve and extend access to Risk Management.

http://www.jstor.org

This content downloaded from 130.237.29.138 on Fri, 04 Dec 2015 01:06:39 UTC
All use subject to JSTOR Terms and Conditions
Risk Management: An International Journal

The Different Dynamics of Personal

and General Risk

Lennart Sjoberg1

Risk perception is oftenstudied withoutfurtherspecifying to whom the risk is supposed


to pertain?the risk target. Such specification is important,in particular as to whether
the risk pertains to the respondents themselves (personal risk) or to other people
(general risk). In the present study,participants judged the personal and general risk
of a large number of hazards. Average personal risk was judged as smaller than
general risk, often much smaller, and the two sets of riskjudgments did not have the
same rank order. Furthermore, they had differentcorrelates. Gender differences in
perceived risk were found to be much more pronounced for general thanfor personal
risk. For personal risk, gender differences were most pronounced for hazards where
there was a small perceived capacity of protection, while gender differences tended
to be more constant for general risk. Economic vulnerability had an effecton both
personal and general risk, but a stronger effect on general risk. Similar effects of
economic vulnerability were found for men and women, refutingthe parallel with a
'white male effect'. Demand for risk mitigation in most cases was more strongly
related to general than to personal risk, but the present results also showed this trend
to be weakened or even reversed in cases where personal risk was perceived to be at
the same level as general risk; hazards where this was the case included nuclear
technology and depletion of the ozone layer. In a second study it was found that non?
specific risk ratings (no risk target specified) were most closely related to general
rather than personal risk, and that one's own responsibilityfor risk management was
regarded as much lower than governmental responsibility. Implications for risk
communication are discussed.

Key Words: Risk perception; personal risk; general risk

Introduction

Risk perception has been a topic of interestto researchers and policy makers at least since the end
of the 1960s, when Starr's seminal paper was published. Starr showed that factors beyond
probability and benefit were important in accounting for social acceptance of risk. His ideas were
extended by several authors in the 1970s, and in 1978 Fischhoff et al published an important
contribution, later called the 'psychometric model' or 'paradigm' (Fischhoff et al, 1978). They
collected several suggestions made about important risk characteristics such as voluntariness or
new vs old risk, and proceeded to a statistical analysis of ratings of perceived and acceptable risk
among a large number of technologies. The approach was reviewed by Slovic (1992) and Slovic
et al (1985); see also a recent excellent volume edited by Slovic (2000a). The psychometric
paradigm illustrates the fruitfulnessof a quantitative approach to risk perception (Sjoberg, 2000b)
and indeed of investigating risk perceptions in general.

The pioneering efforts in work on the psychometric paradigm have been discussed elsewhere
(Sjoberg, 1996b; 2002a; in press a; in press c). In particular, although perceived risk seemed to be

Copyright ? 2003 Perpetuity Press Ltd Page 19

This content downloaded from 130.237.29.138 on Fri, 04 Dec 2015 01:06:39 UTC
All use subject to JSTOR Terms and Conditions
Risk Management: An International Journal

an important dimension in social policy with regard to risky technology such as nuclear power,
the relationship has never been proven to be very strong and in fact has seldom been investigated
empirically in the psychometric tradition. In empirical work, perceived risk has, however, been
found to be less potent a factor than the severity of consequences of an accident or otherwise
unwanted event (Sjoberg, 1998; 2000a).

Perceived risk is still of interest,since risk is a subjective dimension which may have an important
role in several policy contexts, eg with regard to energy production. Data collected in a Swedish
survey concerned with risk perception and attitudes to various energy production technologies
(Sjoberg, 1999c) showed that the risk dimension had considerably larger weight than benefit in
all cases. Similar results have been found in previous work with several sets of data.

Political decision makers have been found to have risk management priorities similar to those of
the public (Sjoberg, 1996c). It is sometimes regretted that the public's risk perception apparently
plays an important role in policy setting, but it is probably unavoidable and quite appropriate in
a democracy (Sjoberg, 2001a; 2001b). Implied life values in various iife saving' programmes or
regulations have been found to vary greatly (Ramsberg and Sjoberg, 1997; Tengs et al, 1995),
and part of the variation appears to be accounted for by variation in perceived risk (Ramsberg
and Sjoberg, 1998). Hence, aspects of perceived risk are important in a number of different
contexts.

Personal and general risk


Research on risk perception has often worked with an undifferentiatedconcept of perceived risk.
In other words, respondents have not been given any furtherspecification or clarification of the
concept. Yet people do make a clear distinction between risk to themselves (personal risk) and
risk to others (general risk2) (Drottz-Sjoberg, 1993; Sjoberg and Drottz-Sjoberg, 1994). When
the target is not specified it is unclear which definition of risk people choose to work with. In one
study it was found that respondents rated non-specific risk close to general risk (Drottz-Sjoberg,
1993), but only one study of the question has been reported so far and the finding needs to be
generalised and replicated.

In many studies it has been found that personal risks are judged as smaller than general risks,
especially so-called lifestyle risks, ie the risks of smoking, drinking alcohol and the like (see eg
Drottz-Sjoberg, 1993). Weinstein has described a related phenomenon called 'unrealistic optimism'
(Weinstein, 1984). Personal risks are seen as so small that they can be dismissed or ignored, and
behaviour then proceeds to various forms of pleasant or exciting experiences that can be quite
destructive and dangerous (Sjoberg, in press b). The personal/general distinction is thus crucial
to the understanding of the rationality, or lack of it, involved in various forms of addictive
behaviours. People who smoke may be aware of the risk to others, but see a very small risk to
themselves. Failure to observe this distinction leads to misleading analysis of addictive behaviour
(Sjoberg, in press b).

Besides documenting how common and large the differences are, it was also the purpose of the
present paper to investigate how personal and general risk are related to important background
data?gender and economic vulnerability. Such an investigation is important in order to pinpoint
the wider importance of the distinction between the two types of riskThe differentialconsequences
of the types of risk were also to be studied by relating them to data on demand for risk mitigation.

Gender differences in risk perception are often observed (Davidson and Freudenburg, 1996). The
trend is usually that women perceive risks to be larger than men do. However, it is unclear if this
is true for all risk targets. In particular, it is important to investigate if women perceive both

Page 20 Lennart Sjoberg

This content downloaded from 130.237.29.138 on Fri, 04 Dec 2015 01:06:39 UTC
All use subject to JSTOR Terms and Conditions
Risk Management: An International Journal

general and personal risks as larger than men do. The initial purpose of the first part of the
investigation (Study 1) was to investigate this question.

Another background factor is socio-economic status. An often-cited result attributable to Flynn


et al involves a difference between a white male group and all others (Flynn et al, 1994). The
white males reported lower risks than other groups. It would be of interest to investigate this
finding in another cultural context, and with regard to socio-economic status rather than ethnic
background. Ethnic background may be a proxy for socio-economic status. The second purpose
of Study 1 was to investigate risk perception?general and personal?in relation to economic
vulnerability.

In many previous studies it has been taken for granted that the larger the risk, the more people
want it to be mitigated. This is simply not true. Other factors enter and make it necessary to
investigate the policy attitudes separately. An important factor is the severity of the consequences
of a hazard, should its risk become reality. It has been found repeatedly that this aspect is more
important for policy attitude than the size of the risk, which is most clearly related to probability
of accident or mishap (Sjoberg, 1999a; 2000a). The psychometric paradigm has worked on the
assumption that demand for risk mitigation is related to perceived risk. However, it is not clear if
general or personal risk is the most important factor in demand for risk mitigation. Demand for
risk mitigation needs to be measured separately, then related to independent ratings of personal
and general risk. The differentproperties of general and personal risk will be illuminated in this
way. It was the third purpose of Study 1 to carry out such an analysis. A second study (Study 2)
followed up and extended some of the findings.

Study 1

In this study, the degree of difference between personal and general risk is firstdocumented on a
large representative set of data and related to the perceived ability to protect oneself from the
hazards. Then details of gender differences in risk perception are studied. Next, economic
vulnerability is related to personal and general risk. Finally, the differentialimportance of personal
and general risk in demands for risk mitigation is investigated.

Method
In 1996 an extensive questionnaire dealing with risk and related issues was mailed to three random
samples of the Swedish population, obtained from the central Swedish population register.Thirty-
four hazards were to be rated in terms of risk. Respondents also rated the possibility of protecting
themselves and the demand forrisk mitigation by the government for a subset of 19 of the hazards.

The 34 hazards were as follows (those contained in the subset of 19 where demand for the
possibility of mitigation protection possibility was studied are italicised):

?
smoking; greenhouse effect;

alcohol; ? indoor radon;

vehicle exhaust; ? unsuitable dietary habits;

AIDS; ? irradiated food;

air pollution; ? trafficaccidents;

?
high-tension power lines; lightning;

Lennart Sjoberg Page 21

This content downloaded from 130.237.29.138 on Fri, 04 Dec 2015 01:06:39 UTC
All use subject to JSTOR Terms and Conditions
Risk Management: An International Journal

depletion of the ozone layer; X-ray diagnostics;

Swedish nuclear power plants; solar radiation;

other Western European nuclear radioactive fallout from the


power plants; Chernobyl accident;

Eastern Europe nuclear power plants; war;

natural background radiation; chemical waste;

nuclear waste; transport of dangerous goods;

genetic manipulation; nuclear arms;

spoiled drinking water; flood;

terroristattacks; inadequate medical care when ill;

spoiled food; violence and aggression;

food polluted by radioactive substances; 'mad cow disease' (BSE).

The three waves of data collection did not yield any significant differences in the respects
investigated here so they were combined for the present purposes, giving an effective sample size
of n = 2338 for all data, with only one exception. The exception was the BSE hazard, which was
only included in two of the subsamples, so sample sizes for that case were n = 1382. The response
rates for the three waves were 61.5, 55.8 and 54.1 per cent, respectively. The main bias detected
was that of level of education, which was higher among the respondents than the population at
large (about ten per cent more with college education). On the other hand, level of education did
not correlate strongly with the variables investigated here, so the bias was probably of minor
importance.

Personal and general risk were rated on eight-step category scales, going from 0 = 'no risk at all'
to 7 = 'a very large risk'. The mitigation ratings were also made on eight-step category scales,
from 0 = 'no importance at all' to 7 = 'very large importance'. Respondents could also choose the
response alternative that mitigation of a particular risk was not the responsibility of the state or
the municipality.

For personal risk, the question asked was:

How large do you thinkthatthe riskis foryou personallyof the following?

For general risk, the question was:

How large do you thinkthe riskis to people in Sweden of the following?

For demand for risk mitigation, the question was:

How importantdo you thinkit is forthe Swedish State or the municipalitiesto mitigatethe
followingrisks?

The three tasks were given in that order. The questionnaire contained many questions and rating
scales not covered here. Further information about details of the study are provided elsewhere
(Sjoberg et al 2000).

Page 22 Lennart Sjoberg

This content downloaded from 130.237.29.138 on Fri, 04 Dec 2015 01:06:39 UTC
All use subject to JSTOR Terms and Conditions
Risk Management: An International Journal

Results
Personal and general risk. Analysis of mean personal and general risk ratings showed that general
risk tended to be rated as larger than personal risk, and that the differences between general and
personal risk were of varying size across hazards. Many of the differences were quite large, and
in 28 of the 34 cases mean general risk was larger than mean personal risk.. Significance testing
was of less interest than the size of the differences.3 The standard deviations were combined for
general and personal risk, separately for each hazard. The standardised differences between the
means were then calculated as the mean differences divided by the combined standard deviations.
They varied between +1.149 and -0.095, with a mean of 0.271.

The differences between the mean general and personal risk ratings constitute a measure of risk
denial. It was found that risk denial was strongly related to mean rated protection possibility,
implying that the difference is particularly large when people believe that they can easily protect
themselves against a hazard?see Figure 1.

Figure I. Risk denial (general minus personal risk) plotted against rated
possiblity to protect oneself from the risk

c
o
s;
o
Q.
IA
3
C

C
O
O

2 3 4 5
Protection possibility

Gender differences in personal and general riskperception. Differences between mean values of
data from female respondents and male respondents (in that order) were calculated. It was then
found that the gender difference was smaller for personal than for general risk, indeed reversed in
some cases for personal risk with males giving somewhat larger risk ratings than females. The
differences were standardised by division with the combined standard deviations. It was then
found that the mean standardised gender difference was 0.232 for personal risk and 0.416 for
general risk, almost twice as large. The ranges were -0.214 to 0.634 and 0.162 to 0.628 for
personal and general risk, respectively. All the gender differences were positive for general risk,
implying that women rated all general risks as higher than men did, on the average. For both
personal and general risk there were thus substantial gender differences.

The two sets of mean gender differences (personal and general risk, respectively) are plotted
against each other in Figure 2. The reversed differenceswhere males gave somewhat larger personal
risk ratings than females occurred in those cases where there was a small difference only for
general risk ratings, although still positive (see Figures 3a and 3b).

Lennart Sjoberg Page 23

This content downloaded from 130.237.29.138 on Fri, 04 Dec 2015 01:06:39 UTC
All use subject to JSTOR Terms and Conditions
Risk Management: An International Journal

Figure 2. Gender differences, personal risk, plotted against gender


differences, general risk

1.5
r (square) = 0.58
n
c
o * *.
</>
o
a.
u
c 0.5
* */*

i-
<D
?D
C
o / *

-0.5
-0.5 0 0.5 I 1.5
Gender difference, general risk

The mean ratings of the possibility of protecting oneself from the risks differedsomewhat, women
giving lower ratings for this possibility in most respects. However, the two sets of means correlated
strongly (r = 0.99) and were combined into a common measure of protection capacity. Mean
differences between genders in personal and general risk were then related to this measure of
protection capacity. It was found that the gender difference in general risk was unrelated to
protection capacity, but the mean difference in personal risk was strongly and negatively related.

Figure 3a. Gender differences related to rated protection possibility, personal risk

c 1.5
#o r (square) = 0.60
e>
u
o
Q.
*
</>
* x-.

o 0.5 #
u
c
* *x

0)
?o
c
0>
-0.5 J_J_
2 3 4 5 6
Mean protection possibility

Women who rated personal risks as larger than men hence did so mainly when the risks were
perceived as hard to protect oneself against. With regard to general risk, the gender difference did
not vary much across hazards.

Risk perception and economic vulnerability. Levels of income were coded in three categories:
lowest 17 per cent, middle 66 per cent and highest 17 per cent.4 Mean personal and general risk
ratings of the 34 hazards were computed for each respondent and subjected to ANOVAs, with

Page 24 Lennart Sjoberg

This content downloaded from 130.237.29.138 on Fri, 04 Dec 2015 01:06:39 UTC
All use subject to JSTOR Terms and Conditions
Risk Management: An International Journal

Figure 3b. Gender differences related to rated protection possibility, general risk

c 1.2
#0 r (square) = 0.05
Q. * *
u
s-
o 0.8 * *
o.

0.6
#c .-*.___
o 0.4
u
c
o 0.2
1.

Li
o
c -0.2
0)
O
-0.4
2 3 4 5 6
Mean protection possibility

gender and income as independent variables, and age as a covariate. Gender, income and age all
yielded statistically significant effects for both personal and general risk. For personal risk, F-
values were 58.875 (1,1816), 9.948 (2,1816) and 25.120 (1,1816) for gender, income and age,
respectively. The corresponding values for general risk, with the same degrees of freedom, were
136.247, 25.988 and 45.404. All these F-values were significant at the 0.001 level. However, the
interactions were not significant. It was found that:

?
personal risk was rated as lower than general risk;

? males gave lower ratings than females; and

?
higher-income groups gave lower ratings of risk than lower-income groups, confirming
the vulnerability hypothesis.

The weak tendency towards interaction was the opposite of that expected from the 'white male
effect', since women showed a somewhat stronger income effect than men. Furthermore, the
effects were stronger for general than for personal risk.5 See Figures 4a and 4b.

Figure 4a. Mean risk ratings for genders and income groups, personal risk

3.6
Men Women
?*? .*??????
3.4

DO
.? 3.2
IS
v.

.*...
C
S 2.8
z
2.6 ?_---*--
*-? -
2.4 -*
_!_
Low Middle High
Income level

Lennart Sjoberg Page 25

This content downloaded from 130.237.29.138 on Fri, 04 Dec 2015 01:06:39 UTC
All use subject to JSTOR Terms and Conditions
Risk Management: An International Journal

Figure 4b. Mean risk ratings for genders and income groups, general risk

Middle
Income level

Demand for risk mitigation in relation to personal and general risk. The judgements of demand
for risk mitigation were analysed by means of regression analysis, with personal and general risk
as explanatory variables. In most cases, general risk had the larger weight in accounting for
demand for risk mitigation. There were exceptions, however, and the difference between the two
beta weights varied across hazards. In Figure 5 there is a plot of the difference in beta weights
(standardised regression weights) against risk denial (difference between mean ratings of general
and personal risk).

Figure 5. Differences in beta weights for general and personal risk obtained
when predicting demand for risk mitigation, plotted against risk denial
(general minus personal risk)

u
c
o

0 0,5 I 1,5 2 2,5

General minus personal risk

Page 26 Lennart Sjoberg

This content downloaded from 130.237.29.138 on Fri, 04 Dec 2015 01:06:39 UTC
All use subject to JSTOR Terms and Conditions
Risk Management: An International Journal

The figure shows that the balance between general and personal risk in influencing the risk
policy attitude was related to the level of risk denial. With a very high level of risk denial, general
risk dominated in risk policy attitude; with a low level of risk denial, personal risk tended to be
more important. Somewhat schematically, it can be said that for lifestyle risks, general risk was
most important in influencing policy attitude while personal risk also entered the picture when it
came to environment and technology risks. This relationship between risk denial and differences
in beta values was not affected to any large extent by differences in standard deviation of personal
and general risk, or the level of perceived protection possibility.

Discussion
Why should people perceive others to be at larger risk with regard to some hazards and not to
others? The results point to perceived control, or the possibility of protecting oneself, as an
explanatory factor.Work on unrealistic optimism has reached the same conclusion (Harris, 1996).
But that explanation gives rise to another question: why should not others be credited with a
similar ability and/or willingness to carry through prudent behaviour? The reason could reside in
attribution processes. The well-known 'fundamental attributionerror' (Ross and Fletcher, 1985)
implies thatothers are seen to be governed more by personality traitsthan oneself, who is governed
more by flexible adaptation to the situation. In a study of the perceived effects of alcohol it was
found that others were believed to be inferior in several aspects of management of alcohol risks.

Gender differences in risk perception have been investigated in previous work (Gustafson, 1998),
but the distinction between personal and general risk has not been adequately attended to in that
work. Yet it is quite clear that gender differences vary as a function of risk target. It would be
interesting to study whether men and women also have differentperceptions of risks to other men
and women, rather than simply to people in general. In some cases it is probably a widespread
notion that men are at larger risk than women, eg when it comes to alcohol consumption, although
such traditional gender differences now seem to be vanishing.

Study 2

The work described so far opened up some interesting new lines of inquiry. First, what happens
when no target is specified? Some previous work with independent groups suggested that people
then rate the risks as they pertain to the general ratherthan personal mode. That finding has so far
not been replicated with a repeated-measures design, and to do so was the firstpurpose of Study
2. Second, policy attitudes have previously been studied with regard to demands on society, but
there is also individual responsibility for risk management. For example, a smoker may expose
others to the risk of passive smoking, a reckless driver is a risk to others as well as to him/herself,
etc. How do people regard individual responsibility as compared to societal responsibility for
risk management?

Method
The subjects were young persons who had applied for admission to the Stockholm School of
Economics. There were 226, 139 male and 87 female. Their mean age was 21.3 years.

The questionnaire asked for ratings of 26 hazards, in the following respects:

?
unspecified target; the question was: 'How large do you think the risk is of the following?';

?
personal risk; see Study 1 for the question asked;

?
general risk; see Study 1 for the question asked;

Lennart Sjoberg Page 27

This content downloaded from 130.237.29.138 on Fri, 04 Dec 2015 01:06:39 UTC
All use subject to JSTOR Terms and Conditions
Risk Management: An International Journal

? demand for risk mitigation by the government; see Study 1 for the question asked; and

?
personal responsibility for risk mitigation; the question was: 'How much do you think
that you yourself should get involved in mitigating the following risks to yourself and
others?'.

The hazards were a subset of the hazards used in Study 1, chosen so as to cover the range in that
study:

smoking (L); natural background radiation (E);

alcohol consumption (L); nuclear waste (T);

vehicle exhausts (E); gene modification (T);

AIDS (L); terroristattacks (C);

air pollution (E); X-ray diagnostics (M);

high-tension lines (T); solar radiation (L);

greenhouse effect and climate change (E); war (C);

unsuitable food habits (L); nuclear arms (C);

trafficaccidents (T); flood (E);

lightning (E); inadequate medical care when ill (M);

depletion of the ozone layer (E); violence and aggression (C);

Swedish nuclear power plants (T); cellular telephones (T);

Eastern European nuclear power plants (T); 'mad cow disease' (BSE) (L).

The hazards have been classified as follows: L- lifestyle; E - environment; T - technology; M -


medical care; C - crime and violence.

Throughout, the risk judgment scales were category scales with eight steps, with one exception
noted above.

Results
Role of the target. The first issue investigated was that of rating 'risk' without specifying the
risk target. At the mean level, non-specific risk ratings were close to general risks (means
2.93 and 2.37 respectively). The mean of the personal risk ratings was 1.86. Correlations
between the series of means of separate hazards (N = 26) told the same story. For non?
specific and general risk ratings the correlation was 0.95, for personal risk and non-specific
risk it was 0.72.

Table 1 gives the correlations between ratings of risk, no target, and personal and general risk,
respectively.

Page 28 Lennart Sjoberg

This content downloaded from 130.237.29.138 on Fri, 04 Dec 2015 01:06:39 UTC
All use subject to JSTOR Terms and Conditions
Risk Management: An International Journal

Table I . Correlations between non-specific risk ratings and general


and personal risk

Hazard Correlation with Correlation with


general risk personal risk

Smoking .50 .11


Alcohol .54 .22
Vehicle exhausts .60 .52
AIDS .60 .51
Air pollution .67 .58

High-voltage power lines .72 .53


Greenhouse effect .75 .68
Unsuitable diet .60 .33
Trafficaccidents .71 .59

Lightning .42 .39

Depletion of the ozone layer .66 .69


Swedish nuclear power plants .77 .77
Eastern European nuclear power plants .78 .73
Natural backgroundradiation .86 .83
Nuclear waste .68 .63
Genetic manipulation .67 .60
Terroristattacks .52 .48

X-ray diagnostics .80 .72


Solar radiation .70 .73
War .37 .38
Nuclear arms .55 .55
Floods .43 .51

Inadequate medical care when ill .59 .61


Violence and aggression .64 .65
Mobile telephones .84 .80
'Mad cow disease' (BSE) .67 .62

It is seen that, in most cases, the correlation between non-specific risk and general risk was larger
than the corresponding correlation for personal risk. However, the difference in correlations
varied greatly, and when related to mean risk denial (difference between general and personal
risk) it was found to correlate strongly, r = 0.87.

Hence, the non-specific risk ratings were least loaded in personal risk with regard to lifestyle
risks, and for environment and technology risks they tended to be equally loaded in personal and
general risk. There were no cases where personal risk was strongly dominant; hence the general
picture was that non-specific risk ratings were most closely related to general risk.

Own versus society's responsibility for risk management. The mean ratings of demand for risk
reduction of the two kinds are plotted against each other in Figure 6.

Lennart Sjoberg Page 29

This content downloaded from 130.237.29.138 on Fri, 04 Dec 2015 01:06:39 UTC
All use subject to JSTOR Terms and Conditions
Risk Management: An International Journal

Figure 6. Mean ratings of own responsibility for risk mitigation plotted


against government responsibility

c
o
Q.
</>
O
i-

12 3 4 5 6 7
Government reponsibility

It is clear that in most cases the government was seen to be much more responsible for mitigating
the risk than oneself. The differences were in many cases very large. The mean difference
between the government's and one's own responsibility, on the one hand, and the difference
between general and personal risk, on the other, were correlated -0.44. In other words, the
government's and one's own responsibilities were most similar when there was a large difference
between general and personal risk. The latter cases pertain to lifestyle risks, such as drinking
alcohol.

Demand for risk mitigation and perceived risk. Mean demand for mitigation by the government
was regressed on mean risk and mean risk denial, across hazards and hence with N = 26. Both
beta values were significant: 0.73 for risk and -0.55 for risk denial, R2a(ij= 0.218. For own
responsibility for risk reduction the corresponding betas were 1.02 and -0.31 respectively, while
R2adj= 0.675. These results show that:

?
personal and general risk contributed to the explanation of policy attitude beyond the
information given by a non-specific risk judgment;

? the larger the personal component in the risk, the more risk reduction was demanded;

? one's own responsibility was more strongly related to perceived risk than was demand for
governmental risk reduction.

General discussion

The Study 1 data came from a large and representative sample, and the results may be regarded as
stable and reliable. The data analysed in Study 2 came from a small sample which was not
representative of the general population; however, there is no strong reason to believe that the
results are misleading for that reason. Several of the findings replicated earlier results, and the
new results were in accordance with expectations.

Page 30 Lennart Sjoberg

This content downloaded from 130.237.29.138 on Fri, 04 Dec 2015 01:06:39 UTC
All use subject to JSTOR Terms and Conditions
Risk Management: An International Journal

General and personal risk, and their differences, appear to be very relevant to the understanding
of risk attitudes. Lifestyle risks almost always involve a strong tendency to deny personal risk.
New information about lifestyle risks is absorbed as information pertinent to other people.
Campaign techniques need to be devised which decrease the tendency to dismiss information as
personally irrelevant.

Policy with regard to lifestyle risks cannot be based on the assumption that people make rational
decisions with regard to these hazards. Their personal risk perception is distorted. They will,
however, be likely to support risk management measures as necessary for others, and there are
many examples of public acceptance of such measures. In Sweden, alcohol policy has been very
stringent for almost a century, bringing down the consumption level to a relatively modest per
capita amount. This has been achieved by a draconic price policy in combination with restrictions
on the availability of alcohol?it can only be sold in rather few special stores, with restricted
business hours, and not to persons under 21. No major political party has seriously proposed any
important liberalisation of these restrictions.6

Risk communication is a relatively new, and quite difficult,field of research. It has two different,
and somewhat independent, subfields of application: technology policy and health behaviour.
The typical problems are very differentin the two cases. Health behaviour is typically seen as not
prudent enough, and experts attempt to stimulate people towards quitting smoking, healthier
food habits, testing their homes for radon, etc. The problem here is that people see the risk as true
mostly for others, and campaigns as less relevant for themselves. With technology risks experts
often assume a reassuring role, but these risks tend to be seen as more personal by members of the
public, and the idiosyncratic approach needed to address personal concerns is hard or impossible
to administer on a large scale. Hence the dilemma of risk communication: when individual
behaviour is in focus, people are mostly worried about others, and when they are worried for
their own sake, as with some technology concerns, reassuring measures can hardly be fitted to
the situation of each individual.

Environment and technology risks tend to be seen as not under one's control, and here the personal
risk factor becomes more important when it comes to policy attitudes. At the same time, these
hazards are seen as mostly the responsibility of the government, while lifestyle risks are seen as
more a question of personal responsibility. The contrast between the two types of hazards can be
illustrated, as in Table 2.

Table 2. Contrast between lifestyle and environmental risks

Lifestyle Environmental

Risk denial Large Small

Responsibilityforrisk management Self Government

Importantforpolicy attitude General risk Personal risk

An important implication of the present results is that the design of risk surveys needs to attend
closely to the selection of a risk target. If none is selected, respondents are likely to rate general
risk. This, in turn, is most appropriate if the goal is to understand the role of risk in policy
attitudes with regard to lifestyle risks?contrary to much previous work on risk perception and
on technology and environment risks. In such work personal risk is important, and it would be
preferable to include both personal and general risk.

Lennart Sjoberg Page 31

This content downloaded from 130.237.29.138 on Fri, 04 Dec 2015 01:06:39 UTC
All use subject to JSTOR Terms and Conditions
Risk Management: An International Journal

Notes

Lennart Sjoberg is in the Center for Risk Research, Stockholm School of Economics; email:
Lennart.Sjoberg@hhs.se. This study is part of the RISKPERCOM project (contract
F14PCT950016), sponsoredalso bytheSwedish Council forPlanningand Coordinationof Research
(FRN), the Swedish Council for Humanistic and Social Science Research (HSFR), the Swedish
Nuclear Power Inspectorate(SKI) and theSwedish RadiationProtectionInstitute(SSI). The partners
in the project were IPSN, France (Dr JeanBrenot), NRPA, Norway (Dr JonReitan), IFR, UK (Dr
Lynn Frewer) and CIEMAT, Spain (Dr Ana Prades).
Some authorspreferthe term'social risk', but thattermhas connotationsnot intendedhere.

With the very large sample size, even verytrivialdifferenceswill be statisticallysignificant.For


thisreason, data analysis in the presentpaper will be mainlydescriptive,not inferential.It can be
noted thatthe mean differencesbetweenpersonal and generalrisks were t-tested,and foundto be
highlysignificant(p<0.0005 in all cases).
Othercut-offsand statisticalmodels could have been chosen, but the presentone was preferred,
firstbecause fairlyextremegroups were needed in orderto obtain data relevantto a comparison
with ethnic backgroundin the US (Swedish society has a much smaller spread of income), and
second because they were not so extremethat the resultinggroups would be too small for the
analysis.

Separate analyses at the level of single hazards showed that income effectswere consistently
differentin respectof personal and general risk,formen and women. However, we did not locate
any variable thatcould explain this variation.

However, membershipin the European Union bringsin new perspectiveson many issues of risk
management.A more liberal alcohol policy is already being implementedin Sweden.

References

Baric, L. (1969) Recognition of the 'At-risk' Role. International Journal of Health Education. Vol. 12,
No. l,pp 24-34.

Davidson, D.J. and Freudenburg,W.R. (1996) Gender and EnvironmentalRisk Concerns: A Review and
Analysis of Available Research. Environmentand Behavior. Vol. 28, No. 3, pp 302-39.

Drottz-Sjoberg,B.-M. (1993) Risk PerceptionsRelated to VariedFrames ofReference.Paper presentedat


the Third Conference of SRA Europe, 'Risk Analysis: UnderlyingRationales', Paris.

Fischhoff,B., Slovic, P., Lichtenstein,S., Read, S. and Combs, B. (1978) How Safe is Safe Enough? A
PsychometricStudy of AttitudestowardsTechnological Risks and Benefits.Policy Sciences. Vol. 9, No.
2, pp 127-52.

Flynn,J.,Slovic, P. and Mertz,C.K. (1994) Gender,Race, and Perceptionof EnvironmentalHealth Risks.


Risk Analysis. Vol. 14, No. 6, pp 1101-8.

Gustafson,P. (1998) Gender Differencesin Risk Perception:Theoreticaland Methodological Perspectives.


Risk Analysis. Vol. 18, No. 6, pp 805-12.

Harris,P. (1996) SufficientGrounds forOptimism?The Relationshipbetween Perceived Controllability


and OptimisticBias. Journal of Social and Clinical Psychology.Vol. 15, No. 1, pp 9-52.

Ramsberg,J. and Sjoberg, L. (1997) The Cost-effectivenessof Lifesaving Interventionsin Sweden. Risk
Analysis. Vol. 17, No. 4, pp 467-78.

Ramsberg,J.and Sjoberg, L. (1998) The Importanceof Cost and Risk CharacteristicsforAttitudestowards


Lifesaving Interventions.Risk - Health, Safety& Environment.Vol. 9, No. 3, pp 271-90.

Ross, M. and Fletcher,G.J.O. (1985) Attributionand Social Perception. In Lindzey, G. and Aronson, E.
(eds) Handbook of Social Psychology. Vol. II. New York: Random House.

Page 32 Lennart Sjoberg

This content downloaded from 130.237.29.138 on Fri, 04 Dec 2015 01:06:39 UTC
All use subject to JSTOR Terms and Conditions
Risk Management: An International Journal

Sjoberg, L. (1989) Radon Risks: Attitudes,Perceptions and Actions. EPA-230-04-89-049. Washington,


DC: Office of Policy Analysis, US EnvironmentalProtectionAgency.

Sjoberg, L. (1996a) Alkohol som valdsforklaringoch riskfaktor. (Alcohol as an Explanation of Violence


and a Risk Factor). F-serie 9. Stockholm: Folkhalsoinstitutet.

Sjoberg, L. (1996b) A Discussion of theLimitationsof thePsychometricand CulturalTheoryApproaches


to Risk Perception. Radiation ProtectionDosimetry.Vol. 68, Nos. 3-4, pp 219-25.

Sjoberg, L. (1996c) Risk Perceptions by Politicians and the Public. Rhizikon: Risk Research Report No.
23. Stockholm: Center forRisk Research.

Sjoberg, L. (1998) Why do People Demand Risk Reduction? In Lydersen,S., Hansen, GK. and Sandtorv,
H.A. (eds) ESREL-98: Safetyand Reliability.Trondheim:A.A. Balkema.

Sjoberg, L. (1999a) Consequences of Perceived Risk: Demand forMitigation.Journal of Risk Research.


Vol. 2, No. 2, pp 129-49.

Sjoberg, L. (1999b) Risk Perception by the Public and by Experts: A Dilemma in Risk Management.
Human Ecology Review. Vol. 6, No. 2, pp 1-9.

Sjoberg, L. (1999c) Risk Perceptionin WesternEurope. Ambio. Vol. 28, No. 6, pp 543-9.

Sjoberg, L. (2000a) Consequences Matter,'Risk' is Marginal. Journalof Risk Research. Vol. 3, No. 3, pp
287-95.

Sjoberg, L. (2000b) The Methodologyof Risk PerceptionResearch. Quality and Quantity.Vol. 34, No. 4,
pp 407-18.

Sjoberg, L. (2001a) Political Decisions and Public Risk Perception.ReliabilityEngineeringand Systems


Safety.Vol. 72, No. 2, pp 115-24.

Sjoberg, L. (2001b) Whose Risk Perception Should Influence Decisions? Reliability Engineering and
SystemsSafety.Vol. 72, No. 2, pp 149-52.

Sjoberg, L. (2002a) Are Received Risk PerceptionModels Alive and Well? RiskAnalysis. Vol. 22, No. 4,
pp 665-70.

Sjoberg, L. (2002b) Policy Implications of Risk Perception Research: A Case of the Emperor's New
Clothes? Risk Management: An InternationalJournal. Vol. 4, No. 2, pp 11-20.

Sjoberg, L. (in press a) Attitudesto Technology and Risk: Going Beyond What is Immediately Given.
Policy Sciences.

Sjoberg, L. (in press b) Neglecting the Risks: The Irrationalityof Health Behaviour and the Quest forLa
Dolce Vita. European Psychologist.

Sjoberg, L. (in press c) Risk Perception,Emotion,and Policy: The Case of Nuclear Technology.European
Review.

Sjoberg, L. and Drottz-Sjoberg,B.-M. (1994) Risk Perception.In Lindell, B., Malmfors,T., Lagerlof, E.,
Thedeen, T. and Walinder,G (eds) ComprehendingRadiation Risks: A Report to the IAEA. Vienna:
InternationalAtomic EnergyAgency.

Sjoberg, L., Jansson, B., Brenot, J., Frewer, L., Prades, A. and Tonnesen, A. (2000) Radiation Risk
Perception in Commemorationof Chernobyl: A Cross-national Study in Three Waves. Rhizikon: Risk
Research Report No. 33. Stockholm: Center forRisk Research.

Slovic, P. (1992) Perception of Risk: Reflections on the PsychometricParadigm. In Krimsky,S. and


Golding, D. (eds) Social Theories of Risk. Westport,CT: Praeger.

Slovic, P. (ed.) (2000) The Perception of Risk. London: Earthscan.

Slovic, P., Fischhoff,B. and Lichtenstein, S. (1985) Characterizing Perceived Risk. In Kates, R.W.,
Hohenhemser,C. and Kasperson, J.X. (eds) Perilous Progress: Managing the Hazards of Technology.
Boulder, CO: Westview.

Lennart Sjoberg Page 33

This content downloaded from 130.237.29.138 on Fri, 04 Dec 2015 01:06:39 UTC
All use subject to JSTOR Terms and Conditions
Starr,C. (1969) Social Benefit versusTechnological Risk. Science.Vol 165, No. 3899, pp 1232-8.

Tengs, O.T., Adams, M.E., Pliskin, J.S., Safran, D.G., Siegel, J.E., Weinstein,M.C. and Graham, J.D.
(1995) Five-hundredLife Saving Interventionsand Their Cost Effectiveness.RiskAnalysis.Vol. 15, No. 3,
pp 369-90.

Weinstein,N.D. (1984) Why itWon't Happen to Me: Perceptionsof Risk Factors and Illness Susceptibility.
Health Psychology.Vol. 3, No. 5, pp 434-57.

This content downloaded from 130.237.29.138 on Fri, 04 Dec 2015 01:06:39 UTC
All use subject to JSTOR Terms and Conditions

You might also like