You are on page 1of 29

Journal of R isk R esearch 1 (2), 135163 (1998)

Co mparative stu die s o f risk pe rce ptio n: a re vie w o f twe nty ye ars o f re se arch *
SA B O H O L M
Center for R isk R esearch, Stock holm School of E cono mics and CE FO S, Gothenbu rg University, Sweden

A bstract This paper critically reviews a number of cross-national studies of perceptions of risks which have been conducted in accordance with the psychometric paradigm developed by the Oregon research group in the 1970s. It considers attempts to study risk perception comparatively, from other theoretical and methodological perspectives, and discu sses various issues that are highlighted in a comparative framework, relating to distinctions between objective and perceived risk, the role of communication and the media, the political system and various societal determinants such as marginality, gender and ethnicity. One of the main conclusions is that comparative studies of risk perception need to be further re ned, both methodologically and theoretically.

1. Introduction
In the social sciences, comparative cross-national or cross-cu ltu ral research may be u ndertaken for several interrelated reasons (for overviews of the research eld in social psychology and psychology, see for example Triandis and Berry 1980; Kagitcibasi and Berry 1989; Kohn, 1989a, 1989b; Berry et al. 1992; Miller-Loessi 1995; H antrais and Mangen, 1996; Schooler 1996). Investigating systematic variation is a cru cial element in theoretical development, and by su bjecting new samples to replications of earlier stu dies it is possible to test the generality of the ndings. It has been argu ed that su ch tests are instru mental in re ning and su bstantiating theories. But comparisons across national or cu ltu ral bou ndaries are not withou t their complications. If cross-national similarities are indeed encou ntered how might we tell whether they resu lt from common processes or stru ctu res or from speci c historical, sociological or psychological circu mstances? There are similarities in condu ct, ideas and institu tions between societies, bu t this does not signify that their meanings, or the processes which order them, are the same. Qu ite to the contrary; there is a mass of ethnographic evidence, from all kinds of society, to prove that hu man social existence is cu ltu rally variable. A ctions and u nderstandings abou t risks, in ju st the same way as other experiential phenomena, are informed by socially and cu ltu rally stru ctu red conceptions and evaluations abou t the world, what it looks like, what it shou ld be or shou ld not be. Perceptions of events and phenomena are conditioned by valu es which vary according to local bodies
*This report has been fu nded by the Eu ropean Commission Nu clear Fission Safety research programme u nder Contract No. F14PCT954016 (D G 12-WSME ). I wish to thank Professor Lennart Sjberg, Center for R isk R esearch, Stockholm School of E conomics, the anonymou s reviewers of Jou rnal of R isk Research for valu able su ggestions and Patrick Crozier for editing services.

1366-9877 1998 E & FN Spon

136

B oholm

of assu mptions, conventions and practices. Hu man societies constitu te u ltra-complex systems, in that hu mans do not merely respond to the physical impact of measu rable and qu anti able aspects of events. Information abou t events, what is recorded and reported, in what way and by whom, is cru cial to hu man life, as is the way that information is socially processed and morally valu ed whether it is tru sted or probed, esteemed or contested (R appaport, 1996) . It follows from the speci city and variability of hu man social existence that it shou ld not simply be presu med that scores and ratings on identical instru ments have the same meanings in different contexts. The qu estion of whether, and if so to what degree, social and psychological phenomena are comparable and equ ivalent across nations and cu ltu res, presents considerable methodological problems (Berry, 1980).
To compare two phenomena, they mu st share some featu res in common; and to compare them to some advantage, they should u su ally differ on some feature. That is, it must be possible to place two phenomena on a single dimension in order to ju dge them validly in relation to each other; and for the comparative judgement to be of value they shou ld not be identical in all respects. (Berry, 1980 p. 8)

Berrys words of warning that mu ch more is requ ired than to collect data in two cou ntries and then compare the resu lts, remind u s how imperative it is that the methodology, and the design in accordance with which research is condu cted, are sou nd if crossnational research is to be su ccessfu l. O therwise methodological inconsistencies may bring abou t divergent resu lts which stem from the design, rather than being valid indicators of processes or stru ctu res intrinsic to the object of stu dy. That a stu dy needs to be initially designed as comparative, in order to allow comparisons, might appear to be a tru ism. Nevertheless, a common problem when earlier stu dies are replicated is that this requ irement is not met. If the collection of data is not accomplished u nder equ ivalent circu mstances or throu gh the u se of equ ivalent measu res, the comparative valu e of the resu lts will be seriou sly compromised. Theoretical concepts and de nitions mu st also be equ ivalent between one national stu dy and another, and of cou rse the scales adopted for measu rement, and for comparison needs to be the same (Berry et al., 1992, pp. 2378). G iven that these preconditions for sou nd comparative research have been met the next dif cu lty will be encou ntered in the interpretation of the resu lts. If we su cceed in demonstrating valid cross-national differences and similarities, how might su ch resu lts be u nderstood? H ow do we select which, from among myriad national differences, shou ld be adopted as the analytical variables pertinent to u s? A s Kohn (1989a, p. 23) has su ccinctly pu t it, it is far from easy to translate an entire nation into variables. A lthou gh cross-national research is costly in terms of time and money, and raises seriou s methodological and interpretative problems, there are strong argu ments in its favou r (see Kohn, 1989b ). By taking into accou nt inconsistencies and differences, inconspicu ou s in a stu dy of a single nation, bu t which might become more noticeable in a comparative exercise, cross-national research can oblige a discipline to revise previou s assu mptions abou t relationships. O ne su ch example is represented by the stu dies across cou ntries in Poland, Japan and the U SA of relationships between feelings of distress and occu pational self-direction, by which is meant the degree of in u ence and selfcontrol an individu al has over his own work tasks as well as the overall situ ation at work (Kohn, 1989b; Schooler, 1996).

R isk perception

137

There has been an increasing recognition of the valu e of stu dies, across nations and across cu ltu res, of risk perception (Cvetcovich and Earle, 1991; R osa et al., 1995). Motivated by a nu mber of diverging aims, orientations and research interests, the stu dies produ ced so far present a heterogeneou s pictu re. Cross-national research has been instigated becau se it has been recognized that little was indeed known abou t the way in which the pu blic in other cou ntries u nderstood risks, besides the U nited States, where most empirical stu dies of risk perception had been condu cted. Cross-national stu dies of risk perception have been gu ided by explorative as well as more theoretical aims: 1. To learn more abou t the way speci c societal categories or grou ps perceive risks: for instance, women as opposed to men; experts rather than laymen; variou s occu pational categories; people with different edu cational backgrou nd; and ethnic minorities (for example Bastide et al., 1989; Nyland 1993; Flynn et al., 1994; Sjberg et al., 1996). To test the generality of resu lts by the u se of a psychometric paradigm regarding the qu alitative dimensions of risk perception (E nglander et al., 1986; Teigen et al., 1988; Keown, 1989; G oszczynska et al., 1991; Kleinhesselink and R osa, 1991, 1994). To evalu ate the proposition by proponents of cu ltu ral theory, to the effect that risk is cu ltu rally constru ed, so that what people fear, and why, is determined by broader valu es (R ohrmann, 1994) .

2.

3.

This article reviews and evalu ates a selection of pu blished stu dies that in variou s ways concern risk perception from a comparative and cross-national perspective. The aim has not been to make a complete inventory; rather, by making a comparative, critical evalu ation of themes of research, problematization and resu lts, to try to identify ndings and ideas as a gu ide to fu ture cross-national, comparative stu dies of risk perception. The stu dies selected are con ned to pu blished articles and reports, and the selection has been made throu gh a strategy of snowballing: references in the literature on risk perception have been examined, relevant articles have been traced and new searches have then been made from the new references in a continu ou s process. D atabases in sociology and psychology, su ch as SO CIO FILE and PSYCH LIT, were also consu lted. A n appendix su mmarizes the main resu lts of 17 stu dies with regard to methodology, size and characteristics of the samples and what risk factors have been assessed.

2. The psychometric paradigm the background


R esearch by cognitive psychologists has demonstrated that when laypersons make estimates of risk they do not merely calcu late in accordance with statistical, (probabilistic) information. O ther considerations affect u ncertain events as well; u nder experimental conditions, people tend to constru e the world in terms of cau sal or ru le-governed schemata rather than by means of probabilistic calcu lations. H eu ristic cognitive devices are resorted to mental gu idelines with regard to which knowledge abou t any risk is readily accessible, and the manner and style in which the knowledge is represented that do not fu lly explore the information concerning decision alternatives (Tversky and Kahnemann, 1973). A s de ned by cognitive psychologists the availability heu ristic relates to what people rem ember, and not to what actu ally has taken place. It is a cognitive scheme for processing information, and not an imprint of the material world u pon

138

B oholm

the mind. People might, or might not, recall what they experience directly, or indirectly throu gh channels of commu nication and the transmission of information. In the 1970s, with a view to investigating the psychological and cognitive dimensions of the estimation of risks, research centres emerged in a few places in Eu rope and the U nited States. R esearch was initiated by the D ecision R esearch G rou p in O regon to explore how laypersons rate and perceive risk. When su bjects were asked to estimate the actu al nu mber of deaths du e to a range of hazards, inclu ding natu ral disasters, diseases, crime and traf c, it was fou nd that they had a tendency to overestimate u nu su al and spectacu lar cases, su ch as botu lism, tornadoes, and oods, and u nderestimate common ones, su ch as cancer, strokes and heart diseases (Lichtenstein et al., 1978). These ndings were interpreted in the light of what cognitive psychologists had discovered abou t the role of heu ristic schemes in decision-making: hazards that are more dramatic and spectacu lar may be more easily remembered and their higher cognitive availability may thu s exp lain the tendency among su bjects to overrate the risks of su ch hazards (Lichtenstein et al., 1978). A n assu mption that was made early on in psychometric research into perception of risks, was that the way hazards are semantically classi ed, the mental models u sed to stru ctu re information, and estimates of probabilities constitu te interdependent cognitive activities. Interest in variou s qu alitative attribu tes su ch as the immediacy of an adverse effect, choices available, knowledge, familiarity and control of a hazard, initiated fu rther qu estions abou t risk perception (Lowrance 1976; von Winterfeldt and E dwards 1984). In a stu dy by Fischhoff et al. (1978), su bjects were asked to rate 30 hazards on nine seven-point scales indicating qu alitative characteristics of risks: whether they were volu ntary, chronic, catastrophic, common, fatal, immediate, ascertainable, controllable, or novel. Factor analysis of these ratings identi ed two factors: the dread of the risk how u ncontrollable, potentially catastrophic, dangerou s to fu ture generations, and involu ntary and the knowledge of the risk for example how chronic, u nknown to those exp osed, delayed, and new. This stu dy was extended fu rther in 1979 by Slovic and his associates (Slovic et al., 1980), who developed a qu estionnaire listing 90 activities, su bstances and technologies, and to which 175 college stu dents were su bjected. A ll kinds of risks were inclu ded from trampolines, hand gu ns, nu clear reactors, chemical su bstances, smoking, aviation, traf c, dru gs, and home appliances. The su bjects were asked to rate the risk of each item for society at large on a scale extending from 0 (not risky) to 100 (extremely risky). The explicit task for the respondent was to estimate how lethal the risk was in each case for society. R isk was de ned as risk of dying from the hazard. In addition, du ring 1976 and 1978, fou r other samples of respondents, consisting of a) members of the Leagu e of Women Voters (n = 76); b) college stu dents (n = 69); c) bu siness people (n = 47); and d) experts on risk assessment (n = 15), were asked to rate a su bset of thirty hazards on the nine seven-point scales of qu alitative dimensions (Slovic et al., 1980). Ju dgements of many of the qu alitative dimensions were strongly correlated, and the same two factors were identi ed as had emerged in the stu dy by Fischhoff et al. (1978). In addition, a third factor emerged, relating to magnitu de of risk, the nu mber of people affected. H azards considered to be volu ntary were also highly likely to be u nderstood as controllable and well known. Factor analysis condensed the risk characteristics to two or three higher-order dimensions. Factor 1, labelled U nknown R isk, was primarily determined by whether it was u nknown to those exposed and to science, and to a lesser extent by u nfamiliarity, involu ntariness, and

R isk perception

139

the delay of effects. Factor 2, labelled D read R isk, tended to be most strongly determined by severity of consequ ences, dread, and catastrophic potential. R isks of nu clear power were ju dged to be extremely catastrophic, likely to be fatal, dreadfu l, u nknown and u nfamiliar. Broadly speaking, nu clear issu es occu pied a position at one extreme of the factor axes: this technology was perceived to be as u nknown, dreaded, u ncontrollable, catastrophic and having delayed adverse effects on fu ture generations (Slovic et al., 1980) . R esearch, not only by the O regon grou p (also for example by Vlek and Stallen, 1980; 1981; O tway and von Winterfeldt, 1982), has provided some evidence that risk perception is in u enced by qu alitative u nderstandings meanings associated with hazards that do not derive from compu tations of statistical nu mbers of fatalities. H owever, how acceptable new technologies are to the general pu blic is not redu cible to its cognitive dimensions (von Winterfeldt and E dwards, 1984). Technology is socially and politically embedded; vested interests of social actors, pu blic debate and dispu te regarding risks and bene ts, may be expected to in u ence the ordinary persons opinions. For an u nderstanding of controversies abou t technology, the perspective therefore needs to be broadened so that it also inclu des valu es and moral considerations, and political, social and ideological concerns. These points have been taken fu rther by advocates of cu ltu ral theory (see D ou glas and Wildavsky, 1982; R ayner and Cantor, 1987; Thompson et al., 1990; Wildavsky and D ake, 1990; D ake, 1991; D ou glas, 1992) as well as by others pu rsu ing constru ctionist approaches to risk perception (H ilgartner, 1992; Short and Clarke, 1992; Clarke and Short, 1993, pp. 37982).

3. Psychometric cross-national studies


3.1. R E PLICA TIO NS O F E A R LIE R WO R K

Several replications of the original stu dy by Slovic et al. (1980) have been made in other cou ntries. O ne aim has been to obtain resu lts comparable to the A merican data. It was hoped that, by carrying ou t research in another cou ntry, broader and more su bstantial theories of the perception of risks might also be formu lated. What has gu ided these comparative attempts has been a mixtu re of ambitions, rst to test general theory and secondly to generate a body of exploratory new knowledge of pu blic opinion in distinct cou ntries. It goes withou t saying that these two aims do not necessarily overlap. The list of 90 hazards, developed in the original A merican research, has been modi ed in most of these replicating stu dies so as to su it speci c national conditions. It has been assu med beforehand that each cou ntry presents its own speci c hazard pro le, so that the qu estionnaires have been adapted in the light of these assu mptions. Strategies for sampling have also diverged, althou gh small, conveniently assembled samples of college or u niversity stu dents have been u sed in most cases. The scales for measu ring the variou s qu alitative criteria have also varied from investigation to investigation. This lack of standardization needs to be kept in mind when comparisons are made between the psychometric research in different cou ntries (see the R oyal Society Stu dy G rou p, 1992) . Moreover, even if similar ratings of perceived risk are demonstrated across samples, we may not assu me that the reasons u nderlying these similar ratings are the same. To take one example, the u ndergoing of radiation therapy might be adju dged risky by artists, who are possibly less informed abou t its basic principles, as well as by scientists who are aware of its limitations (Karpowicz-Lazreg and Mullet, 1993) .

140

B oholm

In 1983 the rst in a series of comparative stu dies of risk perception was carried ou t in Hu ngary (E nglander et al., 1986). Its aim was to compare the stru ctu re of risk perception and the ratings of perceived risk in Hu ngary, with earlier A merican resu lts (reported by Fischhoff et al., 1978; Slovic et al., 1980). Two small samples of Hu ngarian college stu dents were su bjected to two qu estionnaires. The rst listed the 90 activities, su bstances and technologies selected earlier by Slovic et al. (1980), while the second, previou sly u sed by Fischhoff et al. (1978), consisted of a su bset of 30 hazards to be rated on the nine qu alitative dimensions. The factor stru ctu re of the Hu ngarian data was fou nd to be very similar to that of the A merican ones. Two dominant factors strongly correlated with the A merican factors U nknown R isk and D read R isk. H owever, in the Hu ngarian material, the attribu te certain to be fatal was most closely associated with known risks. Comparison between the speci c locations of hazards within the factor space also revealed differences between Hu ngary and A merica. For example, in the Hu ngarian set, nu clear power rated highly as a D read R isk, bu t was not extreme in terms of U nknown R isk. The most striking difference between the two cou ntries was that the mean of ju dgements of risk for the 90 hazards was almost twice as high for the A mericans as it was among Hu ngarians. The speci c concerns were also to some extent different. While A mericans in general were relatively more concerned abou t technological risks resu lting from radiation and chemicals herbicides, pesticides and medically prescribed dru gs the Hu ngarians worried more abou t common, everyday risks su ch as traf c and transport (cars, bicycles, trains and boats), sou rces of danger at home and at work (electricity and gas fu rnaces) and health conditions relevant to pregnancy and childbirth. The conclu sion was that Hu ngarians seemed to be more sensitive to risks associated with the failu re of machines and the people who operate them (E nglander et al., 1986, p. 64), while A mericans tended to be more concerned abou t the delayed effects of su bstances throu gh the failu re of organized safety (Vlek and Stallen, 1981). A cou ple of years later, another replication was made in Norway (Teigen et al., 1988). A gain, this stu dy focu sed on the stru ctu re of risk perception, as well as on the ratings for perceived risk. Two stu dies were condu cted, the rst du ring the winter of 1985/86, and the second in November 1986. In between these two stu dies, the accident at Chernobyl took place. With regard to the qu alitative dimensions of risk, compared with the A merican sample the Norwegian stu dents seemed to be more con dent in science. R isks were rated as older, and with less immediate consequ ences. There was an almost identical ordering of scales in the A merican and Norwegian samples bu t there were some conspicu ou s differences in the way in which the scales were clu stered. U nknown and delayed risks had diverged from the dimension involu ntary/control towards an association with the dimension not necessarily fatal. Norwegian scores on the magnitu de of risks, apart from a handfu l of items, fell below those from A merica, bu t when compared with the corresponding mean ratings from Hu ngary, Norwegian ratings were higher. More speci cally, Norwegians had a risk pro le closer to that of the A mericans, in that they tended to be more concerned than the Hu ngarians with dru gs and narcotics. The main difference between the A merican and the Norwegian ratings concerned chemical su bstances u sed in agricu ltu re, bu t also in food and with regard to medically prescribed dru gs. A third replication came ou t in Poland (G oszczynska et al., 1991), data being obtained this time not from a sample of stu dents bu t from professional people in variou s occupations. The su bjects were selected from two regions in Poland, one highly indu strialized

R isk perception

141

and the other moderately so. O n the criterion of occu pation, each of the two regional grou ps was divided into two su bgrou ps: one technical, inclu ding those engaged as engineers and technicians, and one social, consisting of teachers, jou rnalists and physicians. Correlations among the 15 qu alitative risk dimensions produ ced two main factors, the same as those which had been established in the A merican and Hu ngarian stu dies: a D read R isk factor and an U nknown R isk factor. In comparison with previou s resu lts in other cou ntries, the mean ratings of the estimates of perceived risk magnitu de in Poland were closest to A merican data. Polish ratings with regard to the magnitu de of risks for the 27 hazards common to the qu estionnaires u sed in all fou r cou ntries were somewhat lower than A merican scores, somewhat higher than the Norwegian ones and considerably above the mean ratings in Hu ngary. R isks from warfare, nu clear weapons, alcoholic beverages and railways were ranked higher in Poland than in the other cou ntries. In this stu dy, economic and social hazards the risks of which had not been inclu ded in the qu estionnaires u sed in the other cou ntries were rated higher than those from indu strial activities characteristic of Poland su ch as coal mines, steel mills, and petrochemical plants. The su b-sample of respondents who lived in a more indu strialized area gave higher ratings of perceived risk than respondents from the moderately indu strialized region. H owever, this difference was only tru e for the social professions; the samples from technicians did not vary in relation to the area in which they resided. A replication in France by Karpowicz-Lazreg and Mullet (1993) aimed also at investigating the effects of gender and edu cation on the perception of risk. Stu dent respondents were divided into fou r grou ps differentiated by gender and edu cational orientation: ne arts versu s science. Women gave higher mean ratings than men with regard to the magnitu de of risks. The mean ratings for the French sample were very close to the mean ratings in the A merican stu dy (Slovic et al., 1980). Compared to the resu lts in Hu ngary and Norway, the mean ratings exhibited by the French respondents were considerably higher. Speci c differences between the Norwegian and French samples related to risks associated with violence, high technology, and chemical su bstances which were all considered more risky in France. There was an extremely high nu mber of major differences between it and the Hu ngarian sample, French ratings being higher in each instance. In France, women gave higher ratings of perceived risk than men, regardless of backgrou nd. Stu dents of science feared certain medical techniqu es and toxic su bstances more. There are marked similarities between A merican resu lts and the French ones reported in the stu dy by Karpowicz-Lazreg and Mullet (1993). To a large extent A merican and French su bjects seemed to share the same preoccu pations. French concerns regarding nu clear indu stry were similar to those expressed in A merican stu dies. Similar conclu sions were also reached in another cross-national stu dy of risk perception in France and the U SA (Pou madere et al., 1995). In this stu dy, perceptions of risks showed many similarities between the two cou ntries, althou gh the French were mu ch more tru stfu l of the competence and ef ciency of experts and au thorities. So far, we have focu sed on cross-national research covering Eu ropean cou ntries, which have generated resu lts which are comparable to earlier A merican stu dies. There have also been a few investigations that have been based on samples of A sian su bjects. The rst of these was a stu dy by Keown (1989) who fou nd that the mean ratings of the perceptions of the magnitu de of risks among H ong Kong stu dents were not significantly higher than corresponding A merican ratings. H owever, the ratings of the risks

142

B oholm

of speci c hazards differed a great deal. In the H ong Kong sample some items, su ch as food colou ring, heroin, caffeine, space exploration, crime, non-nu clear electrical power, commercial aviation, food preservatives, railways and bicycles, were rated considerably higher. But, for example, risks from D D T, research into D NA , pesticides, and alcoholic beverages were rated mu ch lower. Factor scores for the su bset of 15 hazards revealed, as in the A merican stu dy, two factors: U nknown R isk and D read R isk. Keown postu lates that perception of risk is likely to vary depending on a wide array of factors: what the media report, what people discu ss, what cu ltu ral norms are predominant and what the technical and legal opportu nities are for the control and regu lation of risk. The stu dies referred to have deliberately u sed qu estionnaires which have inclu ded qu ite different sets of risks, and in which the de nitions of risk have not been consistent. It is far from clear how respondents have u nderstood the standard instru ction that they shou ld focu s on the risk of dying, and on risk to society as a whole or in general. H ow do people in Norway, Hu ngary, the U SA or H ong Kong, conceive society as a whole? Clearly it is wrong simply to assu me that the constru ct society in general, de ned as a national arena, is u niformly u nderstood in all these contexts, and not conditioned by speci c cu ltu ral perceptions. In addition, the particu lar focu s on the risk of dying presents dif cu lty. Shou ld dying be a direct or an indirect consequ ence of a particu lar risk? The risk of a hand gu n might not primarily be u nderstood as du e to the weapon itself, bu t to the intentions of the person who might u se it. Shou ld the risk (as an estimation of probability) be u nderstood to derive from an ef cient cau se, a material cau se, or an intended cau se, of death? A fu rther dilemma is that, althou gh death might seem to be a clear-cu t and u ncontroversial physical condition, matters concerning death its morality, exegesis, and ontology take u s into a eld of hu man existence that is highly determined by cu ltu ral conventions, practices and valu es (see Bloch and Parry, 1982; Hu ntington and Metcalf, 1979). Fu rthermore, samples have not been comparable. Systematic controls have not been introdu ced for su ch variables, among the samples u sed, as age, gender, edu cation, income and occu pation. There has also been a considerable time lag between stu dies so that Slovics A merican stu dy in 1979 has served as a point of comparative reference for stu dies made in other cou ntries decades later. A s a resu lt of these objections, it is open to qu estion how valid comparisons between all the resu lts might be.
3.2. CO MPA R A TIVE STU D IE S

There have also been psychometric cross-national stu dies of the perception of risks which at the ou tset have been designed to be comparative (Kleinhesselink and R osa, 1991; 1994; H inman et al., 1993). The means for collecting data were similar, the samples u sed had equ ivalent characteristics, u sing the same measu rements, and the stu dies were condu cted at rou gh ly the same ju nctu re in time. Kleinhesselink and R osa (1991) aimed at evalu ating and testing two alternative perspectives on risk perception: 1) the psychometric paradigm which claims that there are pancu ltu ral cognitive patterns, and 2) the anthropological su ggestion that perceptions of risk are determined by cu ltu re, and are therefore highly variable. They focu sed on two indu strial nations with similar levels of indu strialization, bu t with divergent cu ltu ral traditions: Japan and the U SA . To provide some control with respect to other social characteristics, so as to allow cu ltu ral backgrou nd to act as a variable maximu m in u ence on the ou tcome of the responses, the

R isk perception

143

su bjects chosen were stu dents; if cu ltu re does indeed in u ence the perception of risk, then the responses from two samples of stu dent, one from each cou ntry, wou ld be expected to differ considerably. The qu estionnaire on risks derived from the one originally u sed by Slovic et al. (1980). The list of hazards was modi ed so as to be better adapted to Japanese respondents, and it came to inclu de 70 items which were rated on nine qu alitative dimensions. No ratings were made regarding the perceived magnitu de of risks. A dditional qu estions referred to self-ef cacy, general political ef cacy, and perceived general vu lnerability to risks. Considerable work was called for when translating the E nglish version of the qu estionnaire into Japanese. The A merican sample of college stu dents rated higher when it came to political ef cacy, ef cacy of response, and self-ef cacy. H owever, the factor scores for the qu alitative dimensions were rather similar. The two higher-order characteristics for stru ctu ring risk perception appeared to be the D read R isk factor and the U nknown R isk factor. The most dreaded activity in both samples was nu clear war. A ll seven nu clear issu es were closely clu stered together and dreaded. A n important difference between the two samples was that, along the dimension relating to the factors known-u nknown, risks were differently constru ed. In Japan, nu clear risks were regarded as well-known by both individu al and society, while in contrast the A mericans saw these risks as being more u nknown. Crime was mu ch more dreaded by the A merican stu dents, while the Japanese rated a nu mber of items related to safety of foods, dru gs, and transport as more dreaded and catastrophic. A mericans reported higher self-ef cacy and, it is su ggested, this may have cau sed them to report less dread in general than the Japanese. A follow-u p of this stu dy was made the following year, u sing the same qu estionnaire on a larger sample of stu dents, and this time the Japanese su bjects completed the qu estionnaire in Japan (Kleinhesselink and R osa, 1994). A gain, nu clear risks were viewed as well known by Japanese su bjects and as moderately u nknown by A mericans. The seven nu clear risks were regarded as being some of the most well known to the Japanese, older and more well known both to individu al and to society. A ll nu clear risks were more dreaded in Japan than in A merica, althou gh nu clear risks were highly dreaded in both samples. The nu clear risks were viewed as being highly u ncontrollable, highly involu ntary, and highly catastrophic. The most dreaded risk in both samples was nu clear weapons (war). Kleinhesselink and R osa point to a plau sible explanation for this Japanese tendency to regard nu clear risks as more well-known: du ring World War II Japanese civilians experienced a tremendou s nu clear catastrophe when H iroshima and Nagasaki were bombed. Thu s, rather than re ecting a cu ltu ral difference in the stru ctu ring of risk perception, these divergent u nderstandings of nu clear risks in A merica and in Japan might be u nderstood as deriving from historical conditions. O ne nding in the variou s psychometric stu dies is that the qu alitative risk characteristics are rated fairly mu ch the same, whether Japanese, A merican, Norwegian, French or Polish respondents were u sed. A lthou gh slight differences in factor scores have been reported, hazards have in general been grou ped together along certain common identi ed factors relating to dread and knowledge. Su ch resu lts have come to be u nderstood as evidence that the cognitive stru ctu ring of risk is similar across nations irrespective of cu ltu ral backgrou nd. H owever, the conclu sion that the cognitive stru ctu ring of risk shou ld be regarded as u niversal seems to reach too far, since it is hardly likely that the psychometric scales adopted by reasons of their u se in earlier stu dies, wou ld exhau st all possible meanings of risks.

144

B oholm

4. Explanations for the perceptions of risk in a comparative perspective


O ne lesson from cross-national research is that the prominence attribu ted to speci c risks might vary from one cou ntry to another. Whereas stu dents in Hu ngary in 1983 worried more abou t risks arising, for instance from smoking, alcohol, crime and motor vehicles, in the U SA , France and Norway concern was stronger abou t hidden chemicals in the environment and in food (herbicides and pesticides), dru gs and narcotics. In Brazil, in contrast to Sweden, people thou ght that, for example, dynamite, reworks, and res in skyscrapers, were big risks (Nyland, 1993). In H ong Kong people worried abou t heroin, crime and chemical su bstances in food and they regarded bicycles and trains as qu ite hazardou s means of transport. D ata from Japan gave a similar risk pro le. A nother nding is that the mean ratings regarding the magnitu de of risks vary considerably from one cou ntry to another, A merican respondents tend to give high mean ratings, as do Bulgarian, Japanese, H ong Kongese, French, Brazilian and Polish su bjects. Low ratings have been encou ntered among Ru ssians, Hu ngarians, R omanians, and Swedes. D espite su ch variation in perceptions of the magnitu de of risks, the correlations between rank orders of hazards across national samples tend to be qu ite strong. Similar rank orders have been obtained in Brazil and in Sweden (Nyland, 1993) , in Bulgaria, R omania and Sweden (Sjberg et al., 1996), in Norway and the U SA , in the U SA and Poland, in Poland and Norway, and in Poland and Hu ngary (G oszczynska et al., 1991). O ne explanation proposed for the different mean ratings concerning the perception of the magn itu de of risk in variou s cou ntries, is that the sheer size of a cou ntry constitutes an independent explanatory variable (G oszczynska et al., 1991, p. 191). It is argu able that the size of a cou ntry brings this effect to bear throu gh interaction with the availability heu ristic; in a large cou ntry, there will be a larger absolu te nu mber of accidents so that, when asked to rate the risk of variou s hazards, people will recall these nu mbers, and this magnitu de will be re ected in their ratings. This theory gives rise to several problems. What is to be taken as the size of a cou ntry? Is it area, measu red in squ are kilometres? O r is size a combination of the nu mber of inhabitants and the area? If we are seeking a cru dely materialistic explanation for the perception of risks, shou ld not a cou ntrys infrastru ctu re and geographical featu res be taken into accou nt as well? Some nations su ch as Au stralia, contain regions that are virtu ally u ninhabited and others that are densely popu lated. H ow, in su ch a case shou ld rates of fatalities be measu red? In terms of the absolu te nu mber of fatal incidents, or as a relative nu mber? A nd if the second, relative to what? In her stu dy of the perception of risks in Sweden and Brazil, Nyland (1993) develops an argu ment along similar lines. The su bstantial difference between Brazil and Sweden in mean ratings of perceptions of magnitu de of risks, as well as in ratings on speci c risks, is u nderstood to derive from actu al characteristics of each cou ntrys environment. Thu s factu al risk is seen as the main determinant of perceived risk. But what is to be taken by factu al risk, and how might it be assessed? Nyland su ggests that factu al risk shou ld be de ned as annu al inju ry and fatality rates (p. 90). That statistics of fatalities measu re objective levels of risk was an assu mption made early in stu dies on risk, bu t it has been severely criticized (see R oyal Society Stu dy G rou p, 1992, pp. 948; A dams, 1995). Nu mbers of fatalities du e to a particu lar hazard, for example per 1000 persons in a given period, do indeed constitu te an objective measu re of risk. H owever, aggregate measu res of risks su ch as annu al rates of fatalities cannot be experienced

R isk perception

145

directly. Factu al risk is not a phenomenon bu t a constru ct one that can be represented by statistical gu res. But it might also be represented by a condensed and powerfu l image, su ch as A merican war cemeteries. These, with the endless rows u pon rows of graves, or long lists of names of fallen soldiers, speak graphically u pon the factu al risk of war. Factu al risk is mediated by knowledge su ch knowledge being disseminated throu gh many channels regarding su ch representations and what they might mean. O ne lesson from cross-national research is that perception of risks is both u niform and variable; cross-national resu lts contain mixed bags of similarities and differences (Kleinhesselink and R osa, 1991, p. 22). A lthou gh hu mans live in a mu ltifaceted world they, to some extent, share similar concerns. Modernization and new information systems mean that new technological achievements, lifestyles, modes of thinking and perceiving are rapidly disseminated globally, to an extent hitherto u nknown in the history of mankind. D epending on a complex interplay of many technological and societal factors, su ch as organized safety systems, infrastru ctu re of a society, living and hou sing conditions, traf c, levels of pollu tion, as well as natural, climatic and geographic conditions, people in variou s parts of the world will be exposed to distinct kinds of danger, and dangers of different orders of magnitu de. For example, it is plau sible that motor traf c is regarded relatively more dangerou s in Brazil, H ong Kong and Hu ngary than in the U nited States and Sweden, becau se traf c conditions are not the same. Traf c is indeed more aggressive and crowded, speed higher, regu lations fewer, cars older and more u nsafe, the drivers less aware and roads more dangerou s in some cou ntries than in others. Su ch objective factors might therefore be re ected in the local perception of risks, which can be expected to vary cross-nationally, du e not to cu ltu ral constraints bu t to objective criteria (Kleinhesselink and R osa, 1991, 1994). The environment, althou gh objecti ed, external, and historically given, is socially produ ced and reprodu ced over the generations. Why, we might ask, are vehicles and roads of a poorer standard, and traf c more u nsafe, in some cou ntries than in others? This qu estion will give rise to other qu estions abou t au thority, morality, equ ity, rights and du ties, ju stice and honou r, and also to qu estions abou t the meanings of symbols and valu es that su stain and promote one kind of world rather than another.

5. Perceived risk and coverage in the media


With respect to many everyday hazards, people acqu ire information to a large extent from direct personal experience. But other hazards are encou ntered only indirectly, throu gh statements made by experts and risk management institu tions, news media, pu blic agencies, political pressu re grou ps, or informal networks of friends and family. If the availability heu ristic is to be taken seriou sly as a theoretical framework for u nderstanding the perception of risk, attention shou ld be directed towards the way variou s hazards are socially represented, for example in the media, rather than towards statistics regarding rates of accidents. From E nglander et al. (1986) we learn that the observed differences in the perception of risks between A merican and Hu ngarian su bjects might be conditioned by the role of the news media in each cou ntry. They su ggested, as a possible explanation, that for the signi cantly lower level of perceived risk in Hu ngary was that the news media, controlled by the commu nist go vernment, gave very little coverage to domestic hazards. For political reasons the regime played down problems inside the cou ntry. A tentative

146

B oholm

estimate of coverage in media of the two cou ntries indicated that A merican newspapers inclu ded to a greater extent reports of domestic dangers and of cau ses of death of all kinds, whereas the Hu ngarian press, apart from traf c accidents, chie y reported deaths occu rring ou tside the cou ntry. Hu ngarian news, in Kasperson et al.s (1988) terminology thu s attenu ated risks, and the A merican press ampli ed them. Thu s, information abou t hazards obtained throu gh variou s means may be expected to be qu alitatively processed, evalu ated and ju dged (Kasperson et al., 1988; R enn et al., 1992). E valu ations of risks in the media may be expected to in u ence pu blic attitu des and responses as to whether risks are accepted, rejected, modi ed, tolerated or eliminated. The media not only bu ild complex messages abou t risks and hazards; while some risks might be ampli ed others might be attenu ated by the presentation in the media. It is worth emphasizing that this model presu pposes a distinction between objective risk and perceived risk. The risk object is taken for granted as a given objecti ed entity, the qu estion being how su ch objects are perceived throu gh a complex process of transmission of information throu gh variou s channels. A nother approach is to regard the risk object as a semantically constru ed entity, a bu ndle of meanings bou nd together by associative and distinctive characteristics pertaining to the object (H ilgartner, 1992; and speci cally with respect to the mass media, Stallings, 1990). The u nderstanding of a risk object is an interpretative process, so that we ou ght to enqu ire what view it might be expert opinion or of cial statistics in relation to which the perception of risks are to be ampli ed or attenu ated? (see Clarke and Short, 1993). Seeking for explanations of cross-national variation in mean ratings of the perception of the magnitu de of risks, Teigen et al. (1988) discu ss a combination of size of cou ntry and impact from the media. Members of a large society might feel more u nprotected and su bjected to objective remote dangers since, in the U nited States for example, people will be more exposed to a greater nu mber of reports abou t risks, and also to more divergent opinions abou t what is dangerou s and what is not. A spate of relevant reports in the media will convey more, and more frequ ent, messages; the probability of information being ambigu ou s is enhanced, and this might resu lt in estimations of risk becoming exaggerated. This argu ment presu pposes a direct relationship between, on a macro level, the total national ou tpu t of news media and, at a micro level what people actu ally take an interest in. The habits of the media consu mer cannot be dedu ced simply from the su m of information in the media, however. A stu dy by Sjberg et al. (1996) in R omania and Bulgaria inclu ded a detailed analysis of relevant media ou tpu t daily newspapers in 1985 and 1993. In both cou ntries reports abou t risks increased du ring the period, althou gh in Bulgaria the increase was considerably greater. The mean ratings of the perceptions of magnitu de of risks were also higher in Bulgaria. In 1985, in both cou ntries, the main focu s in the media was on hazards in foreign cou ntries, similar to what E nglander et al. (1986) had fou nd in Hu ngary. In 1985, the risk most commonly reported in both cou ntries was political instability, whereas in 1993 the problems most frequ ently mentioned in Bulgaria were accidents and in R omania the focu s was still on political instability. Coverage in the media was analysed in detail and correlated with the scores regarding mean perceived risks, and with the rank order of speci c hazards. R esu lts were somewhat negative; they did not su pport the idea that coverage in the media in u ences the perception of risks, or at least not directly in the form of a qu antitative correlation between assessments of levels of risk from variou s cau ses, and the coverage that each received in the media. It is only to be expected that the relationship between the media and the perception

R isk perception

147

of risks will be complex; the role of the media in society, and their different roles in different political and economic systems, mu st be taken into accou nt, and it cannot simply be assu med that the media mediate risk in an invariant way across all societies (Sjberg et al., 1996, pp. 678). Fu rthermore, the media may be expected to have different nationally and traditionally determined interests; these will in u ence the selection of what is newsworthy and also the manner in which whatever material is selected is presented and evalu ated. O f interest in this context is a recent cross-national project that has explored attitu des to risk, natu re and environment in the U K, G ermany, France, Italy and Portu gal (LvyLeboyer et al., 1996). The analysis of the content of three daily leading newspapers in each cou ntry has revealed su bstantial national differences in press coverage. Nu clear pollu tion received almost the same amou nt of attention in all cou ntries except for Italy, where it was not referred to. In the U K, 20% of all reports in media on environmental issu es dealt with the equ ilibriu m of the fau na. In the other cou ntries, neither the equ ilibriu m of fau na nor that of ora received special attention. The G erman press was more concerned with environmental pollu tion of all kinds, while in Portu gal the degradation of natural sites was accorded high signi cance. Little interest in this issu e was expressed in the French press, which was more preoccu pied with natural events, atmospheric conditions and natural calamities. In Italy, it was the degradation of landscape and bu ildings, which received the most attention in the press. In addition, more work needs to be directed towards the way in which people actually u se the mass media. News media shou ld not be assu med to operate in the same way all over the world; there is, for example, considerable cross-national variation in the frequ ency of, and the time spent on, reading newspapers, in the contexts in which they are read, and what social categories of persons read them (Gu stafsson and Weibu ll, 1996). People do not ju st absorb the content of the media throu gh systematically reading page after page of written material. The ordinary reader glances at headlines, skims over certain articles, reads others more carefu lly, while some material remains u nnoticed. Visu al information in newspapers, advertising and television might be exp ected to play a large role in conveying messages abou t risks, especially when it comes to synthesizing information into condensed messages that are more readily remembered (G raber, 1996) . News media are not digested in isolation by individu als. What people have read in the papers or seen on televison tends to be discu ssed at work, with the family, and among friends and neighbou rs, and it might be wondered how these informal exchanges in u ence the impact of coverage by the media. Messages are circu lated in social contexts, and stu dies in social memory indicate that peoples capacity to retrieve information, and to recollect experiences depends to a large extent on social stru ctu res. Memory is not merely a psychological process of information retrieval bu t an active social and discu rsive process that constru cts relevancy and meaning within the context of a social fabric (Middleton and E dwards, 1990; Fentress and Wickham, 1992; H albwachs 1992 [1952 and 1941]; H arr and G illett 1994). It is from this theoretical perspective that social anthropologist Franoise Z onabend (1993) condu cted an important and novel stu dy of the perception of risks at La H agu e, the site of a vast nu clear indu strial complex on the north-west French coast. This work, which adopts qu alitative research methods and interpretations, focu ses on spontaneou s discou rse, emotional and intellectu al ways of dealing with nu clear risks, and their cu ltu ral symbolization by inhabitants and employees at La H agu e.

148

B oholm

More than ten years ago , one vital qu estion for E nglander et al. (1986), while condu cting research into the perception of risks, was how people ju dge the relevance to their own lives of the exp erience of others. What is the role of political, social and geographical distance in determining how relevant they feel messages to be to them, which relate to events that take place in another cou ntry, in another region, or in another social environment? H ow is information abou t hazards in the news media interpreted by the pu blic? These pressing qu estions still await answers. In a recent review of research on the in u ence of pu blic perception of risks Whlberg and Sjberg (1997) conclu de that su ch in u ence has to a large extent been taken for granted, as an established fact. H owever, empirical stu dies by cognitive psychologists and researchers into the mass media, among others have indicated that the impact of media is varied and complex. New qu estions will therefore need to be asked in fu tu re.

6. Risk perception and social background factors


6.1. MA R G INA LITY, PO VE R TY A ND O CCU PA TIO N

A stu dy by the French team of Bastide et al. (1989) explored connections between the perception of risks and sociological and psychological backgrou nd. When the respondents estimates of the frequ ency of 30 cau ses of death from variou s diseases and accidents, were compared with actu al statistics regarding cau ses of death in France, it appeared that the most overestimated cau ses were motor vehicle accidents, indu strial inju ries, homicides, A ID S, leu kaemia, and drowning. The most u nderestimated cau ses were asthma and bronchitis, pneu monia, cerebrovascu lar diseases, and accidental falls. R espondents that had a tendency towards systematic overestimation of the frequ ency of all cau ses of death shared certain socio-demographic featu res. They were more likely than the average to be divorced, have low incomes, be u nemployed, and live in a mediu m-sized town. O ther respondents, who tended rather to u nderestimate the frequ ency of cau ses of death were more likely to have higher edu cation, better income, be resident in a big city, or a farmer living in a village. When it came to estimating rates of death respondents did not differ signi cantly with respect to age, gender or size of family. People who stated that they were depressed, who had themselves su ffered a physical accident or had a relative who had su ffered one, or who declared personal discomfort over pollu tion in daily-life, tended to overestimate cau ses of death. A respondent who had overestimated or u nderestimated cau ses of fatalities wou ld also tend to do the same when rating the danger of hazards from a nu mber of activities, su bstances and technologies. Women were more sensitive than men to the risks from technology. Factor analysis of the ratings regarding these hazards revealed a qu alitative distinction between hazards that were considered volu ntary and of which the bene ts ou tweighed the risks vaccines, railways, and blood transfu sions and those that were not hand gu ns, dru gs, smoking, and alcohol. The factor scores were also taken to indicate a contrast between the legitimacy or illegitimacy of perceived risks. The resu lts were u nderstood to be consistent with those of psychometric stu dies of the perception of risks in other cou ntries. The legitimacy of activities and technologies, as well as their ideological and ethical dimensions, appear to in u ence individu als perceptions of risks (see Sjberg and Winroth, 1986) . The conclu sion drawn from these resu lts is that risk perception is in u enced not only by the activity, su bstance or technology of concern bu t also by a more general feeling

R isk perception

149

of secu rity which a society as a whole procu res to its members. Bastide et al. (1989) base this idea on Du rkheims (1951 [orig. 1897]) classic theory of su icide as a social fact, according to which lack of social control and ru les of condu ct in a society make individu als disoriented and poorly integrated. In a state of anomie, du e to a vu lnerable and insecu re social position, which might be triggered off by divorce, poverty, illness, or u nemployment, individu als can experience a sense of hopelessness which gives rise to a tendency to overestimate risks. In those cases in which stu dies of the perception of risks have inclu ded respondents with marginal social statu s, living in extremely poor conditions, their ratings with regard to perceived risks have indeed tended to be very high. In stu dies by Nyland (1994), and by Sjberg et al. (1996), slu m dwellers in Sao Paolo and Bucharest (bu t not in So a) both rated highly when it came to the perceived magnitu de of risks. Poverty might therefore be considered to be a determinant with respect to the perception of risks. If one mu st stru ggle for su rvival, and be su bjected to constant threats on a daily basis, perhaps this might also increase ones general perception of risks at a more abstract level (Nyland, 1994, p. 90). A merican stu dies of the perception of risks among ethnic minorities also point in the same way. In their valu able review of this eld of research, Vau ghan and Nordenstam (1991) conclu de that for many environmental risks, there are signi cant differences in ju dgement correlating with ethnicity, gender, socio-economic statu s, and edu cation. Concerning technological risks, the literatu re reports signi cantly higher concern among ethnic minorities abou t nu clear power and nu clear waste. In A merica, a vast majority among ethnic minorities are u nsu pportive of this technology, and a greater proportion than among whites want the phasing ou t of the operation of existing plants (G allu p, 1979; 1986). R esu lts indicating the higher perception of risks by ethnic minorities shou ld take into accou nt the physical conditions of the neighbou rhoods in which many of them live. To a larger extent than white A mericans, ethnic minorities live in u rban commu nities characterized by pollu ted air and by solid waste. Poverty was been fou nd to be the primary demographic variable associated with high perceptions of risks, and this relationship was independent of the actu al levels of pollu tion in a commu nity (Cu tter, 1981). It does appear that poor economy, bad hou sing, and instability in a neighbou rhood, have more explanatory valu e than ethnicity in the variable levels of perception of risks between social categories. In A merica, minorities on low income live in older, poorly bu ilt hou ses, and closer to hazardou s chemical or toxic waste sites. E thnic minorities are also more at risk at work, which is often in indu stry, agricu ltu re, and mining, characterized by poor working conditions, higher exposu re to chemicals, insu f cient ventilation, du st and toxic materials. A ll these circu mstances are likely to contribu te to an increase in sensitivity and a likelihood of perceiving life as inherently risky (Vau gh an and Nordenstam, 1991) . H owever, theoretical and methodological dou bts have been raised (see Vau gh an and Nordenstam, 1991, pp. 534) abou t ethnicity as a determinant in the perception of risks. Members of ethnic minorities have often been lu mped together as non-white, and there is need for a mu ch greater differentiation. Statu s, economic and edu cational, tends to be con ated with ethnicity, on the grou nds that ethnic minorities do tend to have lower edu cation, lower income and lower socio-economic status. Becau se su ch factors are often not controlled for, it is far from proven that ethnicity explains variation in the perception of risks between A nglo-A mericans and minorities. The heterogeneity within ethnic categories shou ld also be taken into accou nt (different levels of edu cation, income and so on). U nderprivileged people of colou r also exp erience greater

150

B oholm

exposu re to dru gs, violence, and environmental pollu tion, and ru n a higher risk of mortality than the more well-to-do white part of the popu lation (see also Flynn et al., 1994). Stu dies that have u sed respondents differentiated according to occu pation have come to the conclu sion that occu pation partly determines perception of risks (G oszczynska et al., 1991; Nyland 1993; Sjberg et al., 1996). In two stu dies, one comparing Brazil and Sweden (Nyland, 1993), and the other comparing R omania and Bu lgaria (Sjberg et al., 1996), samples of nu rses in all fou r cou ntries yielded high ratings regarding the perception of risks. In all fou r cou ntries, low ratings regarding the perception of risks were obtained for engineers and manu al workers.
6.2. G E ND E R A ND TH E PE R CE PTIO N O F R ISKS

A mong recu rrent ndings (by for example Teigen et al., 1988; Bastide et al., Kleinhesselink and R osa, 1991, 1994; Karpowicz-Lazreg and Mullet, 1993; Flynn et al., 1994; Sjberg et al., 1996; see also R oyal Society Stu dy G rou p, 1992, p. 109) there are differences in sensitivity to risks among women in contrast to men. A recent overview of A merican research into this topic conclu des that women tend to express higher concern over risks from technology and threats to the environment. This tendency is particu larly strong with regard to nu clear technologies, and concerning pollu tion and risks to health from local facilities (D avidson and Freu denbu rg, 1996). Su ch an effect with respect to gender has been noticed with regard to ratings of risk: rst the perceived magnitu de of risks; secondly hazards from speci c items, women being more sensitive to technological hazards; and thirdly the ordering in which hazards are placed qu alitatively. A n A merican stu dy (Flynn et al., 1994) explored the effects of ethnicity and gender on perceptions of environmental risks to health. Intrigu ingly, mean ratings among white men were considerably lower than those among the three other categories: non-white men, white women and non-white women. Non-white men and women showed only one statistically signi cant difference: when it came to perceptions of the personal risks, men rated lower as regards to stress. When the low-scoring white men were treated as a special su b-sample, and examined with respect to societal characteristics, it was fou nd that they were generally better edu cated, had higher incomes, were politically more conservative, and were also more likely to express greater tru st in government, au thority and indu stry than the other categories of respondent. The cross-national stu dy of the perception of risks in Bulgaria and R omania by Sjberg et al. (1996) came to similar conclu sions: women were more sensitive to risks than (some) men. In the Bulgarian sample, there was only a small difference between men and women both men and women regarded their environment as being highly insecu re. H owever, in R omania, there was a su bstantial difference between men and women; generally, women rated risks to be greater than did men. For both samples, the three variables (gender, cou ntry and grou p) interacted. These resu lts su ggest that the effects of gender vary cross-nationally. The stu dy by Kleinhesselink and R osa (1994) reports signi cant general differences relating to gender when it comes to technological risks. Women in both the A merican and the Japanese samples perceived more risks as dreaded and catastrophic, while men to a higher extent perceived risks as u ncontrollable, newer, and lacking scienti c knowledge. A merican women dreaded social (crime and gu ns) and medical hazards

R isk perception

151

more than did men. Japanese men were more likely to perceive some risks as involu ntary, bu t for A mericans the pattern was reversed, with women being more likely to perceive risks as involu ntary. In this respect, Japanese women were more like A merican men, and A merican women were more like Japanese men. Both Japanese men and A merican women were less likely to consider that individu als have knowledge abou t risks. Japanese men focu sed on lack of knowledge on medical risks, and on some non-nu clear technologies (dams and fossil-fu el electrical power), while A merican women thou ght that individu als lack knowledge abou t environmental risks and nu clear technology. Irrespective of nationality, men were more likely than women to perceive risks to be newer. There were no differences relating to gender with respect to self-ef cacy. There was a noticeable interaction between gender and cu ltu re, since Japanese men showed signi cantly higher vu lnerability to risks, bu t the opposite was tru e of the A merican sample. For the involu ntary dimension, a pu zzling reversal of genders appeared: Japanese men, and A merican women, viewed more risks as being involu ntary. A higher sense of vu lnerability to risks and higher political fatalism, may contribu te to a feeling that more risks are involu ntary. These resu lts point to an intrigu ing cross-cu ltu ral, cross-gender reversal, which su ggests that the effects of gender on the perception of risks bring into play a su btle interaction between factors that are cu ltu rally and socially speci c. Why shou ld women perceive risks as being higher? A biological explanation, refu ted by Flynn et al. (1994), su ggests that, since women bear children, and tend to bear the main responsibility for raising them, they will be more concerned abou t health and safety. Women also tend to be physically more vu lnerable, and therefore might be expected to be more sensitive to hazards in general. E xplanations drawing on combinations of biology and social experience shou ld be considered. In A merican society, women in general have less familiarity with science, and might therefore distru st technology, which is u nderstood as alien and male-dominated (see D avidson and Freu denbu rg, 1996 for an evalu ation of theory in this eld). R esu lts from the research referred to do not su pport a biological explanation for differences in the perception of risks between men and women; biological factors shou ld be expected to apply across social strati cation, and over ethnic and national bou ndaries. In the stu dy by Flynn et al. (1994), it was not the responses of women that stood ou t as pecu liar, bu t those of white men with power, and with good income and edu cation, whose ratings in this respect were very low. White men control, manage and bene t from the world, which they also see as less risky. Fu rther research therefore needs to be focu sed on the role of power, statu s, alienation, tru st, and other sociopolitical factors (Flynn et al., 1994, p. 1107).
6.3. WO R LD VIE WS A ND G E NE R A L VA LU E S

Few attempts have been made to test cu ltu ral theory cross-nationally. O ne reason for this might be that what variou s schools of thou ght among cu ltu ral theorists actu ally propose is vagu e and ambigu ou s, so that it does not translate very readily into a rigorou s design for measu ring data (for a critiqu e of cu ltu ral theory, thou gh from diverging points of view, see Boholm, 1996; and Sjberg, 1995; 1996). A ccording to cu ltu ral theory, the perception of risks is cu ltu rally constru ed, from general orientations or world views, hence national bou ndaries and distinct world views will not coincide. People have different perceptions of risks, not primarily becau se they belong to particu lar

152

B oholm

nationalities bu t becau se they have separate world views (on cu ltu ral theory, see D ou glas and Wildavsky, 1982; R ayner and Cantor, 1987; Thompson et al., 1990; Wildavsky and D ake, 1990; D ake, 1991; D ou glas, 1992). O ne stu dy of the perception of risks which combines cross-national comparisons with comparisons of su b-grou ps (social categories) within a single cou ntry, is partly informed by this theoretical perspective (R ohrmann, 1994). H is stu dy focu ses on differences between, what is called, fou r societal grou ps/cu ltu ral orientations, designated as technological, ecological, monetarian and feminist, in each of three cou ntries: G ermany, New Z ealand, and Au stralia. H e provides no information on the way in which respondents were recru ited, or what criteria were u sed for sampling, in each cou ntry. The stru ctu re of ju dgements regarding risks, with respect to a nu mber of hazards, on a wide range of evalu ative criteria, was analysed by means of LISR E L modelling of cau sal relationships between cognitive elements and characteristics of social categories. The acceptance of risks was fou nd to be determined by the perceived magnitu de of a risk and by considerations of bene t. Concern abou t impacts on health carried more weight than probability of dying. A ttitu des su ch as concern abou t the environment, negative evalu ations of technology, and post-material orientations with regard to valu es, had considerable in u ence. The cross-national comparisons revealed differences between the Au stralian and G erman respondents: the Au stralians were more accepting when it came to risks regarding sport, or those from u nhealthy private habits, rated lower with regard to risks from conventional technologies. O ccu pations exp osed to risks of the pollu tion of the environment, and large-scale technology su ch as nu clear energy, were given more negative evalu ations in Au stralia. The resu lts from Au stralia and New Zealand were fairly similar, apart from slight discrepancies regarding earthqu akes, a hazard mu ch more familiar to New Z ealanders. R espondents with ecological and feminist orientations evalu ated risks mu ch higher in connection with most sou rces of risks. Feelings of anxiety were also higher among those categories, while ju dgements on bene t and of the acceptance of risks were lower. R espondents with technological orientation yielded the lowest ratings, seeing more bene ts and being more ready to accept the attendant risks. The ju dgements of the monetarian grou p fell in between the extremes encou ntered among, respectively, the technologists and feminists/ecologists. The largest differences between categories appeared in relation to the item living near a nu clear plant, or chemical indu stry. H owever, this pattern of differences in the perception of risks between the fou r categories (technological, monetarian, ecological and feminist) was reversed in ratings on risks from consu mption smoking, tranqu illisers, and overeating. H ere, it was rather the engineers and stu dents of technology who provided the lowest ratings of acceptance, while respondents with ecological and feminist orientations were more tolerant. There was little difference in sensitivity to risks between Au stralia, New Z ealand and G ermany. D ifferences between categories were stronger in G ermany than in Au stralia, indicating that polarization in categories is lower in Au stralia. The resu lts lend su pport to the idea that societal categories with particu lar professional, cu ltu ral and political orientations differ considerably in their ju dgement of, and evalu ation of, risks. In addition to differentiation between cu ltu ral orientations or world views, the stu dy revealed variation across national bou ndaries. The resu lts are said to su pport the approach advocated by cu ltu ral theory to issu es regarding risk. Since very little information is provided on the way these fou r societal categories were de ned, and how they theoretically

R isk perception

153

relate to the so-called grid-grou p stru ctu re on which cu ltu ral theory is u ltimately based, it is hard to assess the validity of R ohrmanns conclu sion. Most comparative stu dies on the perception of risks have focu sed on long lists of issu es relating to risks. A nother attempt to stu dy the perception of risks cross-nationally gave attention instead to a single issu e. It was condu cted in 198687 in Au stralia, E ngland, France, G ermany and the Netherlands, regarding attitu des to nu clear power in the light of the accident at Chernobyl (E iser et al., 1990). U sing the theory of cognitive dissonance (Festinger, 1957), to the effect that people will strive to redu ce inconsistencies between contradictory cognitions, this stu dy aimed at examining attitudes to nu clear power and the Chernobyl accident. In all ve cou ntries the rate regarding the seriou sness of the accident related to attitu des on other topics related to nu clear ssion. Variations in patterns of decision-making were related to the degree of anxiety and attention evoked by the accident. G enerally, there was a strong relationship between more pro-nu clear attitu des and more right-wing political leanings. The resu lts su ggest that differences in attitu de on nu clear issu es form part of systems of valu es and of ideologies, the scope of which are mu ch broader.

7. Conclusions
H aving su rveyed the evidence for or against variou s determinants of the perception of risks, I will end with a few re ections. What comparative stu dies on the perception of risks have in common is that they tend to su bscribe to the psychometric paradigm; that apart, however, they form a heterogeneou s eld of research, with little agreement on basic theoretical issu es, and low consensu s on problems in their research or on methodology. This is scarcely to be wondered at, since we are dealing with a new eld of research to which have been attracted scholars from many disciplines, each adhering to its own academic tradition. H ence, as a matter of u rgency, we shou ld be systematizing resu lts and theoretical frameworks, thu s generating new research which will lead to more powerfu l statements abou t the perceptions of risks as social and psychological phenomena. O ne lesson, slightly disconcerting perhaps, is that su ccessfu l cross-national comparisons of perceptions of risks are far from straightforward. A lthou gh the collection of data in two or more cou ntries, comparing the resu lts, and conjectu ring abou t plau sible explanations, may present no special problem, even so su ch interpretations tend to be problematic. A recu rrent theme emphasized in the literature is that there are objective risks and dangers in societies, and it is assu med that people, being realists, are aware of these conditions; hence this awareness is expected to be re ected in ratings derived from qu estionnaires which they ll ou t. Social scientists proffer su ch commonsense interpretations; bu t they will be of limited valu e, since they are brou ght into accou nt as post hoc explanations of resu lts that have not been informed by prior theory. A s researchers in the eld of the perception of risks, we shou ld be striving to bring to the fore any remaining u nresolved problems encou ntered in the cou rse of obtaining ou r resu lts, not merely to assert that any particu lar nding is precisely what we might have expected, taking into accou nt su per cial characteristics of the society in qu estion. A s an instance of this, to discover that H ong Kongese perceive high risks in riding a bicycle throu gh city traf c, and then to explain that nding by the objective characteristics of traf c in that same city (Keown, 1989), tells u s little abou t the perception of risks. In the light of one of the cornerstones of anthropological theory abou t hu man

A ppendix
Sample, size and characteristics replication
l l

154

Stu dy college stu dents two samples n =30, n =29 qu estionnaires psychometric
l

Cou ntry risk magnitu de: 90 hazards nine qu alitative dimensions: 30 hazards

D esign

R isk dimensions

R esu lts

E nglander, Farago, Slovic and Fischhoff (1986)

Hu ngary

media analysis

Low mean scores on risk magnitu de as compared to A merican scores More concern with failu re of machines and their operators R isk perception is in u enced by media reporting; in comparison with the American press, the Hu ngarian news media which was u nder censorship reported very little on domestic hazards

Teigen, Bru n and Slovic (1988)


l l l l

Norway qu estionnaires psychometric nine qu alitative dimensions: 30 hazards individu al fear

stu dents several samples n =35, n =23, n =64

replication

risk magnitu de: 90 hazards

Higher mean scores on risk magnitu de than in Hu ngary bu t lower than in the USA Less concern with chemical su bstances in agricu ltu re, in food and prescription dru gs Test of individu al fear revealed gender difference, women reporting more fear Fatal R isk rather than D read Risk and Involu ntary/Uncontrollable R isk rather than Unknown R isk had more effect on qu alitative dimensions
l

Goszczynska, Tyszka and Slovic (1991)

Poland

risk magnitu de: 40 hazards

l l

professions replication technicians, teachers, jou rnalists, and qu estionnaires physicians n =140 psychometric two geographical areas

fteen qu alitative dimensions: 40 hazards

R isk magnitu de close to American resu lts bu t lower when only the risks common in both estimates were considered Warfare, nu clear weapons, alcohol and railroads rated high Su bjects of social professions living in more highly indu strialized area had higher level of perceived risk, there was no su ch difference for technicians
l l l

Karpowicz-Lazreg and Mu llet (1993)

France

u niversity stu dents, replication science and ne arts qu estionnaires n = 107 psychometric

risk magnitu de: 90 hazards

Mean rating very close to A merican scores Similar concerns Women gave higher ratings than men

B oholm

Stu dy qu estionnaires
l l

Cou ntry n =24, researchers n =24, stu dents in technology rank correlations factor analysis ranking: 13 risks on fou r dimensions risk magnitu de: 30 hazards
l l

Sample, size and characteristics risk magnitu de: 75 hazards Very low scores on risk magnitu de Compared to A merican corresponding data, handgu ns, terrorism, crime, nu clear war and asbstos were rated mu ch lower

D esign

R isk dimensions

R esu lts

R isk perception

Mechitov and R ebrik (1990)

Ru ssia

Keown (1989) qu estionnaires psychometric


l

Hong Kong six qu alitative dimensions: 15 hazards

bu siness stu dents n =65

replication

Overall mean ratings not signi cantly higher than corresponding A merican scores Three areas of risks were rated higher: crime and narcotics, traf c by railroads and bicycles and food additives and preservatives A lcohol was rated lower

Kleinhesselink and R osa (1991)


l l

Japan qu estionnaires psychometric dimension: 70 hazards

stu dents

comparative

seven qu alitative

Kleinhesselink and R osa (1994)

USA

1988 stu dy Japanese n =69 American n =62 1989 follow u p u niversity stu dents (psychology & sociology) Japanese n =273 Americans n =152

Overall cognitive stru ctu ring of risks was mu ch the same for both national samples While Japanese su bjects regarded nu clear risks as well-known A mericans u nderstood them as largely u nknown Women in both samples u nderstood more risks as dreaded and catastrophic while men regarded more risks as u ncontrollable, newer and lacking scienti c knowledge Japanese women responded more similarly to A merican men in that they perceived more risks as volu ntarily; Japanese men and A merican women regarded more risks involu ntary E vidence for cross-gender, cross-cu ltu ral reversals of u nderstandings comparative ve qu alitative dimensions: 30 hazards mailed qu estionnaires correlations
l

Hinman, Rosa, Kleinhesselink and Lowinger (1993) USA n =747, response rate 52.6%

Japan USA

random selection of hou seholds

155

A ll nu clear items were more dreaded by the Japanese respondents than by the A merican Japanese respondents consider the nu clear items to be better known than A merican su bjects

156

Stu dy
l

Cou ntry

Sample, size and characteristics

D esign

R isk dimensions

R esu lts

Japan n =290, response rate 29%


l l

A ll other energy technologies are considered more u nknown by the Japanese than by A mericans Overall Japanese su bjects regard hazards as older than A mericans Both A mericans and Japanese very mu ch dread all nu clear risks, and feel little personal control over them

Bastide, Moatti, Pages and Fagnani (1989)


l

France factor analysis rate 52 hazards on a ve-point scale of danger


l l

statistically representative national sample n =1000

personal interviews

rate frequ ency of 30 mortality cau ses

R espondents who overestimated cau ses of death were more often divorced, had low income, were u nemployed, depressed, or worried abou t pollu tion R espondents who u nderestimated cau ses of death were more likely to have higher edu cation, live in a big city, or be a farmer in a village Su bjects who over- or u nderestimated risks of death wou ld do the same when estimating the danger of variou s hazards Women perceived more risks associated with technology
l l l

Flynn, Slovic and Mertz (1994) total n =1.512 white n =1.275 non-white n =214 telephone interviews correlations

USA

random sample

comparative

magnitu de of health risk: 25 hazards

Pou madere, Mays, Slovic, Flynn and Johnson (1995)

USA France

representative sample

comparative

worldviews tru st attitu des risks in environment magnitu de of health risks:

Mean scores for risk magnitu de lower for white men Non-white men and women, irrespective of ethnicity, gave the same means White men with low scores on risk magnitu de were better edu cated, had higher income, more conservative political orientation and reported more tru st in government and indu stry than average Both popu lations were qu ite sensitive to health risks

B oholm

Stu dy telephone interviews


l

Cou ntry

Sample, size and characteristics USA n =1.500 France n =1.500 worldviews tru st attitu des risks in environment
l

D esign

R isk dimensions

R esu lts

R isk perception

25 hazards

Both A mericans and French expressed strongest concern abou t nu clear waste, A IDS, street dru gs and cigarette smoking. The French expressed mu ch more tru st in experts and au thorities and worldviews were mu ch less convinced that they cou ld themselves exercise control over risks to their health

R ohrmann (1994)
l

comparative qu estionnaries LISR E L modelling


l l l

Au stralia Germany New Z ealand

societal grou ps/ cu ltu ral orientations: technological, monetarian, ecological and feminist Germans n = 217 Au stralians n =339 New Z ealanders n =278 u niversity stu dents (psychology and social sciences) qu estionnaires factor analysis total n =840 comparative attitu des: environment and technology social valu es

ju dgments of 24 hazards according to 11 evalu ative criteria

R espondents with technological orientation gave lowest risk ratings, saw more bene ts and were more ready to accept risks E cological and feminist orientations reported higher anxiety and gave higher risk ratings for most hazards Intra grou p differences were more su bstantial than cross-national differences R isk perception is explained by world view rather than by national identity

E iser, H annover, Mann, Morin, van der Pligt and Webley (1990)

Au stralia UK France Germany Netherlands

political decision making economic expectations Chernobyl accident attitu des on nu clear power

The ratings of seriou sness of the Chernobyl accident related to attitu des on nu clear topics In all national samples it was a strong correlation between pronu clear attitu des and right-wing political valu es Variation with respect to attitu des, reported anxiety and attention evoked by the Chernobyl accident was stronger within than between the national samples

157

158

Stu dy comparative
l

Cou ntry magnitu de of risk: 100 hazards


l

Sample, size and characteristics

D esign

R isk dimensions

R esu lts

Nyland (1993) qu estionnaries correlations


l

Brazil Sweden personal risk societal risk de nitions of risk

Five grou ps: nu rses, u niversity stu dents in technology u nskilled workers, and slu m dwellers (in Brazil only).

Brazilians n =150 Swedes n =120


l

Considerably higher means for risk magnitu de in Brazil as compared to Sweden Brazilian slu m dwellers reported the highest means of perceived risk and economy, followed by Brazilian nu rses Striking differences betwen the ratings of the diverse grou ps, nu rses giving high ratings in both cou ntries A lthou gh the perceived risk magnitu de was mu ch higher in Brazil the rank order between the hazards was rather similar In both cou ntries societal risk was rated higher than personal risk

Bu lgaria Romania
l l l l

Sjberg, Kolarova, Ru cai, Bernstrm and Flygelholm (1996) qu estionnaries correlations analysis of media content

Six grou ps: nu rses, teachers, u niversity stu dents in technology and economy, u nskilled workers, and slu m dwellers Romanians n =192 Bu lgarians n =240

comparative

magnitu de of risk: 100 hazards

personal risk societal risk risk mitigation and compensation personal protection anxiety political issu es and sense of control

Higher means for risk magnitu de in Bu lgaria as compared to R omania Nu rses gave the highest ratings in both cou ntries Interaction between gender and other variables su ch as cou ntry and su b-grou p In Bu lgarian sample, where perceptions were qu ite high, there was little mean difference between men and women. In the R omanian sample women generally rated risks as higher than men A lthou gh there were some differences the rank order between hazards in both cou ntries were rather similar Qu antitative analysis of media content in both cou ntries did not correlate with the mean scores for perceived risk nor with the rank order of hazards

B oholm

Stu dy comparative
l l

Cou ntry teachers in the natu ral or hu man sciences, engineers qu estionnaries correlations
l

Sample, size and characteristics

D esign

R isk dimensions

R esu lts

R isk perception

Lvy-Leboyer, Bonnes, Chase, Ferreira-Marqu es and Pawlik (1996) total n = 742 analysis of media content
l

France Italy Germany Portu gal UK

attitu des on environmental issu es information on environment tru st in media knowledge su bjective and objective risk perception
l

Su bstantial national differences in press coverage of environmental issu es Perceived risk was an important predictor of concern for the environment in three cou ntries: U K, Italy and Portu gal German respondents differed from from all others in several respects having other environmental predictors Highest ranking predictor for risk perception in all cou ntries except Italy was attitu des Variou s degrees of environmental concern in each cou ntry; strongest pro-environmental concern in G ermany, followed by Italy and Portu gal, the French ranking lowest

159

160

B oholm

society the idea that hu man societies constitu te u ltra-complex systems, depending on higher-order, conventionalized systems for processing information we shou ld clearly not merely be posing qu estions abou t the factu al, measu rable properties of the environment or its speci c instru mental aspects. Stu dies of the perception of risks condu cted from a cross-national and cross-cu ltu ral comparative perspective need to address qu estions abou t the way in which risks are embedded in the social fabric, taking into accou nt conceptions of morality, equ ity, ju stice, and honou r; religiou s doctrine; ideas concerning sovereignty; property, and rights and du ties; and aesthetic valu es and what constitu tes qu ality in life (R appaport, 1996: 65). If we are to avoid the fallacy of u nrestrained empiricism, produ cing new empirical resu lts accompanied by trivial explanations simply for the sake of it (Fau cheau x, 1976), bu t be gu ided by relevant front-line research in behaviou ral and social disciplines, the comparative, exp lorative stu dies of the perception of risks shou ld be striving to formu late theoretically well-fou nded designs for research which are amenable to more profou nd interpretations of the resu lts. Fields for cross-national research into the perception of risks that au gu r well for the fu tu re, appear to be the role of backgrou nd factors su ch as gender and social marginality, edu cation and occu pation. Broad ideological issu es, especially those relating to u nderstandings of political matters and tru st in the orders of dominance and au thority, might also be predicted to attract increasing attention. O ne way to proceed methodologically, and one that I personally wou ld favou r, wou ld be by way of intensive qu alitative case stu dies based on strategically chosen data (see for example Vari et al., 1991). Contrasting resu lts from these might well prove illu minating, and serve as sou rces for fertile problem-solving exercises for the fu ture. The role of the media in variou s cou ntries, and how they contribu te to the way the general pu blic constru es risks, may be expected to become a rapidly growing eld of research yet we need to know more abou t the objects themselves that are u nderstood to give rise to risks, and how people socially and cu ltu rally constru e all su ch trou blesome and feared realities (H ilgartner, 1992; Zonabend, 1993).

References
A dams, J. 1995. R isk . London : U CL Press. Bastide, S., Moatti, J-P., Pages, J-P., and Fagnani, F. (1989) R isk perception and the social acceptability of technologies: the French case, R isk A nalysis 9, 21523. Berry, J.W. (1980) Introduction to Methodology, in H .C. Triandis and J.W. Berry (eds) H andbook of Cross-Cu ltu ral Psychology, M ethodology , vol. 2. Boston, London : A llyn and Bacon. Berry, J.W., Poortinga, Y.H ., Segall, M.H . and D asen, P.R . (1992) Cross-Cu ltu ral Psychology: Research and A pplications, Cambridge: Cambridge U niversity Press. Bloch, M. and Parry, J. (eds) (1982) D eath and the R egeneration of L ife, Cambridge: Cambridge U niversity Press. Boholm, . (1996) R isk perception and social anthropology: critiqu e of cu ltu ral theory, E thnos 61, 484. Clarke, L. and Short, J.F. (1993) Social organization and risk: some cu rrent controversies, A nnu al Review of Sociology 19, 37599. Cu tter, S. (1981) Commu nity concern for pollu tion: social and environmental in u ences, Jou rnal of E nvironm ental Psychology 13, 10524. Cvetcovich, G .T. and E arle, T.C. (eds) (1991) Special issue on R isk and Cu ltu re, Jou rnal of Cross-Cu ltural Psychology 22.

R isk perception

161

D ake, K. (1991) O rientating dispositions in the perceptions of risk: an analysis of contemporary worldviews and cu ltu ral biases, Jou rnal of Cross-Cu ltu ral Psychology 22, 6182. D avidson, D .J. and Freudenbu rg, W.R . (1996) G ender and environmental risk concerns. A review and analysis of available research, E nvironm ent and B ehavior 28, 302339. D ou glas, M. 1992. R isk and B lam e, London : R ou tledge. D ou glas, M. and Wildavsky A . (1982) R isk and Cu ltu re: A n Essay on the Selection of T echnological and E nvironm ental D angers, Berkeley, U niversity of California Press. Du rkheim, E . (1951) [orig. 1897] Su icide, translated by John A . Spalding and G eorge Simpson, Illinois: The Free Press. E iser, J.R ., H annover, B., Mann, L., Morin, M., van der Pligt, J. and Webley, P. (1990) Nu clear attitu des after Chernobyl: a cross-national stu dy, Jou rnal of E nvironm ental Psychology 10, 101110. E nglander, T., Farago, K., and Slovic, P. A . (1986) Comparative analysis of risk perception in Hu ngary and the U nited states, Social B ehaviour 1, 5566. Faucheau x, C. (1976) Cross-cu ltu ral research in experimental social psychology, Eu ropean Journal of Social Psychology 6, 269322. Fischhoff, B., Slovic, P., Lichtenstein, S., R ead, S., and Combs, B. (1978) H ow safe is safe enough? A psychometric stu dy of attitu des towards technological risks and bene ts, Policy Sciences 9, 12752. Fentress, J. and Wickham, C. (eds) (1992) Social M em ory , O xford: Blackwell. Festinger, L. (1957) A T heory of Cognitive D issonance, E vanstone, Ill.: R ow, Peterson. Flynn, J., Slovic, P. and Mertz C.-K. (1994) G ender, race and environmental health risks, R isk A nalysis 14, 11018. G allup O rganization (1979) Nu clear power plants, G allup R eport 167 , 812. G allup O rganization (1986) Nu clear power plants, G allup R eport 250 , 1619. G oszczynska, M., Tyszka, T., and Slovic, P. (1991) R isk perception in Poland: a comparison with three other cou ntries, Jou rnal of B ehaviou ral D ecision M ak ing 4, 17993. G raber, D .A . (1996) Say it with pictu res, A nnals of the A m erican A cadem y of Social Sciences 546, 8596. Gu stafsson, K.-E . and Weibu ll, L. (1996) Eu ropean newspaper readership an overview, in Eu ropeans R ead N ewspapers, Brussels: E NPA . H albwachs, M. (1992) [1952, 1941] O n Collective M em ory , Chicago: Chicago U niversity Press. H antrais, L. and Mangen, S. (eds) (1996) Cross-N ational R esearch M ethods in the Social Sciences, London , New York: Pinter. H arr, R . and G illett, G . (1994) T he D iscu rsive M ind , London: Sage. H ilgartner, S. (1992) The social constru ction of risk objects, in J.F. Short and L. Clarke (eds) O rganiz ations, Uncertainties and R isk . Bou lder, San Francisco, O xford: Westview Press, 3953. H inman, G .W., R osa, E .A ., Kleinhesselink, R .R . and Lowinger, T.C. (1993) Perceptions of nu clear and other risks in Japan and the U nited States, R isk A nalysis 13, 44955. Hu ntington, R . and Metcalf, P. (1979) Celebrations of D eath: T he A nthropology of M ortu ary R itu al, Cambridge: Cambridge U niversity Press. Kagitcibasi, C. and Berry, J.W. (1989) Cross-cu ltural psychology: cu rrent research and trends, A nnual R eview of Psychology 40, 493531. Karpowicz-Lazreg, C. and Mullet, E . (1993) Societal risk as seen by the French pu blic, R isk A nalysis 13, 25358. Kasperson, R .E., R enn, O . Slovic, P., Brown, H .S., E mel, J., G oble, R., Kasperson, J.X. and R atick, S. (1988) The social ampli cation of risk: a conceptual framework, Risk A nalysis 8, 17787. Keown, C.F. (1989) R isk perceptions of H ong Kongese vs. A mericans, R isk A nalysis 9, 4015. Kleinhesselink, R .R . and R osa, E.A . (1991) Cognitive representations of risk perceptions: a comparison of Japan and the U nited States, Jou rnal of Cross-Cu ltu ral Psychology 22, 1128.

162

B oholm

Kleinhesselink, R .R . and R osa, E .R . (1994) Nu clear trees in a forest of hazards: a comparison of risk perceptions between A merican and Japanese university students, in T.C. Lowinger and G .W. H inman (eds) Nu clear Power at the Crossroads: Challenges and Prospects for the T wentyrst Centu ry , International R esearch Center for E nergy and E conomic D evelopment: Washington State U niversity, 101119. Kohn, M.L. (ed.) (1989a) Cross-N ational R esearch M ethods in Sociology , London: Sage. Kohn, M.L. 1989b. Cross-national research as an analytic strategy, in M.L. Kohn (ed.) CrossN ational R esearch M ethods in Sociology , London : Sage. Lvy-Leboyer, C., Bonnes, M., Chase, J., Ferreira-Marqu es, J., and Pawlik, K. (1996) D eterminants of pro-environmental behaviors: a ve-cou ntries comparison, Eu ropean Psychologist 1, 12329. Lichtenstein, S., Slovic, P., Fischhoff, B., Layman, M. and Combs, B. (1978) Ju dged frequ ency of lethal events, Jou rnal of Experim ental Psychology (Hum an L earning and M em ory) 4, 55178. Lowrance, W.W. (1976) O f A cceptable R isk: Science and the D eterm ination of Safety , Los A ltos, CA : W. Kau fman. Mechitov, A .I. and R ebrik, S.B. (1990) Stu dies of risk and safety in the U SSR , in K. Borcherding, O .I. Larichev and Messick D .M. (eds) Contem porary Issues in D ecision M ak ing, pp. 26170, A msterdam: E lsevier Science Pu blishers B.V. Middleton, D . and E dwards, D . (eds) (1990) Collective R em em bering, London : Sage. Miller-Loessi, K. (1995) Comparative Social Psychology. Cross-cultu ral and cross-national, in K.S. Cook, G .A . Fine and J.S. H ou se (eds) Sociological Perspectives of Social Psychology , Boston: A llyn and Bacon, 397420. Nyland, L.G . (1993) R isk Perception in B raz il and Sweden , Stockholm: Center for R isk R esearch. O tway, H .J. and von Winterfeldt, D . (1982) Beyond acceptable risk: on the social acceptabilities of technologies, Policy Sciences 14, 24756. Pou madere, M., Mays, C., Slovic, P., Flynn J., and Johnson S. (1995) What lies behind pu blic acceptance? Comparison of U S and French perceptions of the nu clear power option, in T he Nu clear Power O ption. Proceedings of an international conference on the nu clear power option organized by the international atomic energy agency, Vienna September 1994, Vienna, 393405. R appaport, R .A . (1996) R isk and the hu man environment, A nnals of the A m erican A cadem y of the Political and Social Sciences 545 , 6474. R ayner, S. and Cantor, R . (1987) H ow fair is safe enough. The cu ltu ral approach societal technology choice, R isk A nalysis 7, 39. R enn, O ., Bu rns, W.J., Kasperson, J.X., Kasperson, R .E ., and Slovic, P. (1992) The social ampli cation of risk: theoretical fou ndations and empirical applications, Jou rnal of Social Science Issues 48, 13760. R ohrmann, B. (1994) R isk perception of different societal grou ps: Au stralian ndings and crossnational comparison, A u stralian Journal of Psychology 46, 15063. R osa, G ., R enn, O ., Jaeger, C. and Webler, T. (1995) R isk as a challenge to cross-cu ltu ral dialogu e. Paper presented at the XXXII Congress D ialogues Between Cu ltu res and Changes in Eu rope and the World of the International Institute of Sociology, Trieste 37 Ju ly 1995. R oyal Society Stu dy G rou p, (1992) R isk : analysis, perception and m anagem ent, R eport of a R oyal Society Stu dy G rou p, London : The R oyal Society. Schooler, C. (1996) Cu ltu ral and social-structu ral explanations of cross-national psychological differences, A nnu al R eview of Sociology 22, 32349. Short, J.F. Jr. and Clarke, L. (eds) (1992) O rganizations, Uncertainties and R isk , Bou lder, CO : Westview Press. Sjberg, L. (1995) E xplaining R isk Perception: A n E m pirical and Qu antitative E valuation of Cu ltu ral Theory , Stockholm: Centre for R isk R esearch. Sjberg, L. (1996) A discu ssions of the limitations of the psychometric and cu ltu ral theory approaches to risk perception, R adiation Protection D osim etry 68, 21925.

R isk perception

163

Sjberg, L., Kolarova, D ., Ru cai A-A ., Bernstrm M- L. and Flygelholm H . (1996). R isk Perception and M edia R eports in Bu lgaria and R om ania, Stockholm: Center for R isk R esearch. Sjberg, L. and Winroth, E . (1986) R isk, moral valu e of actions, and mood, Scandinavian Journal of Psychology 27, 191208. Slovic, P., Fischhoff, B. and Lichtenstein, S. (1980) Facts and fears: u nderstanding perceived risk, in R .C. Schwing and W.A . A lberts Jr. (eds) Societal R isk A ssessm ent: H ow Safe is Safe E nou gh?, New York: Plenu m Press. Stallings, R .A . (1990) Media discou rse and the social constru ction of risk, Social Problem s 37, 8095. Teigen, K. H ., Bru n, W. and Slovic, P. (1988) Societal risks as seen by the Norwegian pu blic, Jou rnal of B ehaviou ral D ecision M ak ing 1, 11130. Thompson, M., E llis, R . and Wildavsky A . (1990) Cu ltu ral T heory , Bou lder, San Francisco: Westview Press. Triandis H .C. and Berry, J.W. (1980) H andbook of Cross-Cu ltu ral Psychology, M ethodology , vol. 2, Boston, London: A llyn and Bacon. Tversky, A . and Kahnemann, D . (1973) A vailability: a heu ristic for ju dging frequ ency and probability, Cognitive Psychology 5, 20732. Vari, A ., Kemp, R . and Mumpower, J.L. (1991) Pu blic concerns abou t LLR W facility siting: a comparative stu dy, Jou rnal of Cross-Cu ltural Psychology , 22, 83102. Vau ghan, E. and Nordenstam, B. (1991) The perception of environmental risks among ethnically diverse grou ps, Jou rnal of Cross-Cu ltu ral Psychology , 22, 2960. Vlek, C.J.H . and Stallen, P.J.M., (1980) R ational and personal aspects of risk, A cta Psychologica 45, 273300. Vlek, C.J.H . and Stallen, P.J.M. (1981) Ju dging risks and bene ts in the small and in the large, O rganiz ational B ehaviour and Hu m an Perform ance 28, 23571. Wildavsky, A . and D ake, K. (1990) Theories of risk perception: who fears what and why, D aedalu s 112 , 410. von Winterfeld, D . and E dwards, W. (1984) Patterns of con ict abou t risky technology, R isk A nalysis 4, 5568. af Whlberg, A . and Sjberg, L. (1997) R isk Perception and the M edia. A R eview of R esearch on M edia In u ence on Pu blic R isk Perception , Stockholm: Center for R isk R esearch. Z onabend, F. (1993) T he Nu clear Peninsu la, translated from French by J.A . U nderwood, Cambridge: Cambridge U niversity Press.

You might also like