Who Reaps the Benefits, Who Bears the Risks? Comparative Optimism, Comparative Utility, and Regulatory Preferences for Mobile Phone Technology
Abstract
Although the issue of risk target (e.g., self, others, children) is widely acknowledged in risk perception research, its importance appears underappreciated. To date, most research has been satisfied with demonstrating comparative optimism, i.e., lower perceived risk for the self than others, and exploring its moderators, such as perceived controllability and personal exposure. Much less research has investigated how the issue of target may affect benefit perceptions or key outcomes such as stated preferences for hazard regulation. The current research investigated these issues using data from a public survey of attitudes toward mobile phone technology (N= 1,320). First, results demonstrated comparative optimism for this hazard, and also found moderating effects of both controllability and personal exposure. Second, there was evidence of comparative utility, i.e., users believed that the benefits from mobile phone technology are greater for the self than others. Third, and most important for policy, preferences for handset regulation were best predicted by perceptions of the risks to others but perceived benefits for the self. Results suggest a closer awareness of target can improve prediction of stated preferences for hazard regulation and that it would be profitable for future research to pay more attention to the issue of target for both risk and benefit perceptions.
1. INTRODUCTION
1.1. Overview
Sjöberg (2000) reviews a number of factors that are important for understanding risk perception. Among these is the important issue of “risk target,” which is often underappreciated. The term refers to the person or persons who potentially bear a risk, for instance, the self, one's peers, children, and society as a whole. Although most surveys of risk perception now explicitly include a target of some sort, i.e., they don't just say “how risky is this?”, they usually include only one. More often than not this is the risk to “society in general,” which has been used right through from the earliest studies in the field (e.g., Fischhoff et al., 1978) to recent research (e.g., Siegrist et al., 2005). Even where measures of different targets are taken these are sometimes collapsed into a single scale due to high correlations (e.g., Poortinga & Pidgeon, 2005). However, as Sjöberg points out, this is a potential problem because “people do not make the same estimates when they rate the risks to themselves, to their family, or to people in general” (2000, p. 2).
In particular, in line with unrealistic or comparative optimism (Sheppard et al., 2002; Weinstein & Klein, 1996) the perceived risks to the self are often judged to be lower than the perceived risk to one's peers (Weinstein, 1980). Thus even if the correlations between perceptions of risk and benefit for different targets are high, differences in absolute levels across targets may be important in understanding key outcomes such as preferences for regulation. For instance, in some cases people may prefer greater regulation of hazards based on the perceived benefits and risks for the self (e.g., airline safety), whereas in other contexts their preferences may be based instead on the perceived benefits and risks to certain sections of society (e.g., child vaccinations). Thus only by examining the perceptions of risk and benefit separately for these targets will we get a clearer picture about whose risk and benefit people are taking into account when they make judgments about certain hazards. The present research explores these issues in the context of the electromagnetic fields (EMFs) associated with mobile phone technology. This is an especially appropriate context given that unlike many (potential) hazards, the benefits are widely appreciated, regulators have identified certain targets as potentially vulnerable (e.g., children), and there are multiple sources of potential risk associated with the same technology (e.g., handsets vs. transmission masts).
1.2. Comparative Optimism and Regulatory Preferences
The need to clearly distinguish between targets emerges from a closer examination of the risk perception literature. Much of the early research in this area asked about people's perceptions of societal, rather than personal, risks and benefits (e.g., Slovic, 2000). For example, the kind of task participants were presented with was to assess “the total value to society of each item during one year” where “item” referred to various potential hazards (Fischhoff et al., 1978, italics added). The results have sometimes been interpreted as saying something about how people view the risks to themselves, for instance, “despite the expenditure of billions of dollars and steady improvements in health, safety and longevity of life, people view themselves as more rather than less vulnerable to the dangers posed by technology” (Kasperson et al., 1988, p. 177, italics added). In Sjöberg's (2000) survey of 15 different hazards ranging from lightning to AIDS, however, people tended to report higher risk perceptions for “society in general” than for either themselves or their family, suggesting that we should be cautious in generalizing from statements about society to statements about the self.
Similar findings are reported by researchers within the more general field of comparative optimism (Sheppard et al., 2002), also known as optimistic bias, or unrealistic optimism. Concerned that the category of general or societal risk represents a qualitatively different target than personal risk due to a lack of specificity, these researchers tend to define the “other” more tightly by asking about people's peers, e.g., people of the same age and gender as the self. Even using these comparisons, the perceived risks to the self are frequently reported to be lower than those for one's peers across a range of health (e.g., heart attacks, Weinstein, 1980), and environmental hazards (e.g., air pollution, Pahl et al., 2005). Of course, for any given individual, personal risk may be lower than average and thus their comparative optimism may be justified. However, at the aggregate level, societal risk should equate with the aggregate of personal risk, at least in a representative sample. Thus it seems that at least some individuals within this aggregate are being “unrealistically” optimistic because their own comparative risks are actually higher than they think.
Comparative optimism is not, however, a universal phenomenon and there are instances where, for specific risks or populations, perceived personal risk will be much closer to perceived societal or peer risks. One such moderator is perceived control (Harris, 1996). People are more likely to see their own risks as similar to those of others when they feel they have little personal control over them. Sjöberg (2000), for example, found that where personal control was perceived to be high, for risks such as alcohol and tanning, the difference between perceptions of personal and general risk tended to be higher than for those risks where personal control was thought to be low (e.g., lighting and the greenhouse effect).
Comparative optimism also tends to be attenuated when personal exposure to a risk is high. Thus there are “reality constraints” on the degree to which people are optimistic relative to their peers. For example, comparative optimism for various diseases associated with smoking such as lung cancer is lower for smokers than for nonsmokers, indicating that they are aware that their personal risk is relatively high compared to the population in general (McKenna et al., 1993). However, some degree of comparative optimism may still remain even in these situations. For example, some smokers may still claim that their personal risk is less than that of other smokers (McKenna et al., 1993; for similar findings with sex workers, see van der Velde et al., 1994). Thus the degree to which perceptions of risks to the self and others are likely to be similar will also be affected by exactly who is being asked.
Though these observations may highlight when researchers are likely to find differences in risk perception as a function of target, they say little about the potential impacts of such differences. Of more importance for policy purposes is whether one or other of the measures is a better predictor of important outcomes such as public preferences. One of the key preferences with respect to technological hazards is the amount of regulation members of the public believe the government should impose (Slovic, 1987). Not surprisingly, perhaps, prior research suggests that people generally want higher regulation for hazards that are seen as riskier in terms of a lack of knowledge and feelings of dread (e.g., Slovic, 1987). However, we know of no research that has directly examined and reported the relationships between these regulatory preferences and perceptions of risk in relation to different targets (e.g., self vs. societal). Knowing whether perceptions of self or other risks is the best predictor of regulatory preferences would be useful to policymakers keen to establish where public reactions to hazards—in the form of demands for more stringent regulation through petitions, demonstrations and so on—are likely to occur.
Despite a general lack of prior research examining this issue, a recent investigation into risk perceptions of mobile phone use while driving did report both self and other risk and regulatory preferences (White et al., 2004). Although the original analysis did not include an examination of the current issue, a reanalysis of these data suggests that, for people who reported using a mobile phone while driving (N= 446), regulatory preferences were more influenced by perceived risks to others (β= 0.33, p < 0.001) than the self (β= 0.09, ns.). Since users tended to see the risks as relatively controllable (compared to perceptions of those who did not use mobile phones while driving), this finding suggests their calls for regulation are primarily based on concerns about other people's behavior rather than their own. It seems as if they are saying, “Don't worry about me, I am in control of this risk and don't need my behavior regulated. However, I am not too sure about everyone else and think regulation is needed to protect them from themselves and others.” It is conceivable that this relationship will be even stronger where the risks and regulations relate to a particularly vulnerable group with little perceived control such as children or the elderly. Failing to take into account the distinction between self and others, therefore, potentially presents a misleading picture of the determinants of people's preferences for regulation.
We propose that for controllable risks, the perceived risks to society in general (and vulnerable members in particular) will be a stronger predictor of regulatory preferences than the perceived risks to the self. For uncontrollable risks, by contrast, societal and personal risk perceptions should be equally strong predictors of regulatory preferences. In this case people do not feel they can implement behaviors that will reduce their risk and therefore many will want policymakers to regulate exposure levels both on their own and society's behalf.
1.3. Comparative Utility and Regulatory Preferences
While a considerable amount of work has looked at the issue of target (self, society) in relation to perceived risks, far less has looked at it in relation to perceived benefits. By and large, people are asked to consider either the benefits to society or the self, but not both (although see Poortinga & Pidgeon, 2005). This is unfortunate because the majority of man-made hazards were developed to achieve a benefit of some kind. Nuclear power is used to produce electricity, foodstuffs are genetically modified to increase shelf life, for instance, and mobile phones are used to communicate more flexibly. Thus for many technological risks, perceptions of benefit, in addition to perceptions of risk, are key to understanding people's overall perceptions of the hazard (e.g., Fischhoff et al., 1978; Gaskell et al., 2004; Pidgeon et al., 2005; Slovic, 2000).
Despite a lack of prior research comparing benefit perceptions across target, we propose that the same processes underlying comparative optimism will extend to judgments of self/societal benefit. We will refer to this possibility as comparative utility and define it as the possibility that people see greater benefits of a technology for themselves than for society in general. We could also refer to this as the “better-for-me-than-average effect” as a variant of the “better-than-average effect” where people tend to see themselves more positively compared to others in general (Alicke et al., 1995). We suspect that comparative utility might occur because people have readier mental access to instances of personal than societal benefit when asked to rate them. It is probably relatively easy to bring to mind instances where a mobile phone, for example, was of significant personal benefit. However, when asked about the benefits of mobile phones to society in general, it may be less easy to come up with concrete examples, especially if one thinks of some of the (half) conversations one often hears in public places.
As with comparative optimism, however, comparative utility is likely to be moderated by personal use (i.e., exposure). People who use their mobile phone relatively little are likely to exhibit less comparative utility than those who use it a lot. Thus, reality constraints are also likely to be shown in the degree to which comparative utility will be expressed. Our main focus on comparative utility is, however, concerned with the degree to which personal versus societal benefit perceptions are related to regulatory preferences.
First, in line with prior research we predict that perceived benefits will be negatively associated with regulatory preferences. The more benefits people see, the less they want the technology to be regulated. More importantly for the present discussion, we suspect this general pattern is qualified by the target of the benefit, at least for controllable risks. For controllable risks, perceived personal benefits should be more important for regulatory preferences than perceived societal benefits, because the more benefits people see for themselves the less they want the technology to be regulated. However, just because they see benefits for others, does not necessarily mean they are prepared to accept low levels of regulation, since other's benefits may still entail substantial costs for the self.
1.4. Mobile Phone Technology
Mobile phone technology is a particularly interesting context in which to investigate these issues. First, at the time when, and in the location where (the United Kingdom), the study was carried out there was well-publicized uncertainty among the scientific community about the potential for negative health effects arising from the technology. In the U.K. context this was highlighted in the final report of the “Independent Expert Group on Mobile Phones” (IEGMP), a body set up by the U.K. government to investigate and report publicly on these issues. The authors concluded that “it is not possible at present to say that exposure to RF (radiofrequency) radiation, even at levels below national guidelines, is totally without potential adverse effects,” and that “gaps in knowledge are sufficient to justify a precautionary approach” (IEGMP, 2000, p. 3). This admission of scientific uncertainty (even if the objective uncertainty was no greater or even less than for other risks) potentially allowed members of the public to draw their own conclusions about the benefits and risks (see also Wiedemann et al., 2003).
Second, the technology has two quite separate sources of potential RF radiation risk (hereinafter risk source), namely, mobile phone handsets themselves and the transmission masts or base stations that transmit their messages. For reasons already well researched (e.g., Slovic, 2000) it is likely that the perceptions of the risks of these two sources of RF radiation will be different (Siegrist et al., 2005). For example, we predict that, while exposure from handsets would be perceived as relatively voluntary and controllable, exposure from masts would not. These two aspects of the hazard are important for investigating the moderating role of perceptions of controllability for comparative optimism and subsequent regulatory preferences.
Third, the IEGMP suggested that “if there are currently unrecognized adverse health effects … children may be more vulnerable because of their developing nervous system, the greater absorption of energy in the tissues of the head, and a longer lifetime of exposure” (2000, p. 8). Thus, the report effectively divided “society” into two categories (adults and children) and identified children as being at potentially greater risk. This caveat was widely discussed in the media and had real policy implications in the United Kingdom, resulting, for example, in guidelines relating to the siting of mobile phones masts on or near schools (IEGMP, 2000). For current purposes, this differentiation of risk targets allowed us to extend the more usual comparison between the self and similar others (Weinstein, 1980) to comparisons between the self, similar others, and a potentially vulnerable group (children).
Fourth, personal exposure to handset risks differs in the population and can be relatively easily assessed by individuals themselves. People are likely to have some indication of the amount of time they use their mobile phone thanks in part to the billing system that charges them a certain amount per time of use (i.e., exposure). Although exact estimates of use are likely to be less than fully accurate, people are likely to have some idea about whether their own use is high or low relative to the population in general and thus reality constraints are likely to come into effect.
Finally, the uptake of the technology (76% coverage in the United Kingdom by 2002, for example; OFTEL, 2002) suggests that the public sees many personal benefits. This was confirmed recently by Siegrist et al. (2005) who report that Swiss respondents rated the risks to society as 2.93 for mobiles and 3.26 for base stations but the benefits as 3.36 and 3.32, respectively. This is in contrast to other new technologies such as biotechnology. In many people's minds, at least in Europe, biotechnology's potential for personal risk appears to loom larger than the potential for personal benefit (e.g., Frewer et al., 1997; Gaskell et al., 1999; Pidgeon et al., 2005). Since one of our interests was to compare perceived personal and societal benefit it was important to avoid such a technology and investigate one where personal benefits were tangible (i.e., mobile phones).
1.5. Current Research and Hypotheses
In sum, the current research investigated the perceived risks and benefits of mobile phone technology. Of central concern was the possibility that these estimates would vary as a function of (1) risk target (i.e., the self, society in general, one's peers, or children), (2) risk source (handsets vs. masts), and (3) personal use/exposure. Moreover, for policy purposes, we were especially interested in seeing which of the target-related judgments would best predict regulatory preferences. Such findings could help policymakers design more sensitive measures of where and when the public is likely to react negatively, or positively, to new technologies. These issues were investigated using data from a large-scale public survey in the United Kingdom. Specific hypotheses (H) are detailed below.
-
H1: Perceptions of mobile phone technology— First, we expected that the potential health effects associated with handsets would be seen as more controllable and voluntary than those associated with masts.
-
H2: Comparative risk—Second, we expected to find evidence of comparative optimism, with mobile phone users perceiving risks of the technology in general to be lower for the self than for one's peers, which in turn should be lower than the perceived risks to children (H2a). However, given the predicted differences in perceptions of controllability, we expected this pattern to be greater for handsets than masts (H2b). Moreover, the amount of use should also moderate a user's comparative optimism, with those who use mobiles a lot showing less optimism due to reality constraints (H2c). Non-users were excluded from these analyses because the risks of mobile phones are lower for them in absolute terms.
-
H3: Comparative benefit—Third, we expected to see comparative utility, with users rating their own benefits higher than those for society in general, and children in particular (H3a). Like perceived risk, we expected amount of use to moderate this effect, though in the opposite direction, with higher level users showing greater comparative utility (H3b).
-
H4: Regulatory preferences—Finally, we predicted that regulatory preferences for the uncontrollable risk (masts) would be higher than for the controllable risk (handsets) (H4a), and that regulatory preferences for handsets would be higher for children than adults (H4b). Moreover, due to comparative optimism and comparative utility for handsets, we predicted that personal benefit but risks to others would be the strongest predictors of regulatory preferences for adult use (H4c). Since comparative optimism was not predicted for masts, we expected perceived risks for the self, others, and children to be equally strong predictors of regulatory preferences (H4d).
2. METHOD
2.1. The Survey
The survey was sent to a total of 8,000 households in a medium-sized U.K. city during the autumn of 2001. Households were selected according to postcode subsectors with high demographic heterogeneity and representativeness of the general U.K. population, as defined in the Royal Mail (“ Cenario”) database. Every household within the chosen postcode subsectors received a single delivery envelope containing a questionnaire and a prepaid reply envelope. To conform to the system offered by the postal service for mass mailing, the envelopes were not personally addressed. Hence, we had no control over which individual in a household completed the survey. Thus although the survey employed a sample including members of the general public, it is not presented as an “opinion poll.” Despite the diversity of our sample in terms of age, gender, and use of mobile phones, our concern was not with public opinion per se but rather with how a heterogeneous sample of respondents viewed the risks and benefits of the technology in comparison to others.
2.1.1. Participants
The total number of questionnaires returned was 1,320, a response rate of 16.5%. There were 629 males (48%), 671 females (52%), and the mean age of respondents was 48.07 years old (SD= 16.05, range 12–93 years). Since 26 respondents provided only “additional comments,” our sample for quantitative analysis consisted of 1,294. Analyses with Ns lower than this are due to specific instances of missing data and exclusion of children's responses (age < 18 years, N= 22) from the comparative risk scores. This was needed to avoid children comparing their own risks to those of “other adults.” Due to the large sample size, significance for statistical tests was set at p < 0.001. Although low in absolute terms, the response rate exceeds the distributor's (i.e., Royal Mail) normal estimate of about 5% for this type of un-addressed delivery. Clearly, our method does not allow us to identify factors that distinguish respondents from nonrespondents.
2.1.2. Questions
Exposure. There were three personal-exposure-related questions, two related to handsets and one to masts. Specifically, Ps were asked (1) whether they had a mobile phone (e.g., own phone, share one with partner); (2) if yes, roughly how much time they use it for making calls (hrs & mins), receiving calls (hrs & mins), and text messaging (hrs & mins) per week; and (3) as far as they were aware did they live near an existing or proposed mobile phone mast—Yes (existing), Yes (proposed), No, Don't know. The questions relating to risk perceptions, perceived benefit, and regulatory preferences are presented in Table I.3
| Question | Response Format | |||
|---|---|---|---|---|
| Low Anchor | Mid Point | High Anchor | ||
| 1 | Perceived controllability/voluntariness: | |||
| To what extent do you agree with the following statements about the potential health risks from a) regular use of a (hand-held) mobile phone; b) living near a mobile phone mast? | Strongly disagree (−3) | Unsure (0) | Strongly agree (+3) | |
| i) Users (residents) can control the level of risk they are exposed to | ||||
| ii) The only people at risk are those who use mobile phones | ||||
| 2 | Perceived benefit: | |||
| Generally speaking do you think mobile phones are: | Definitely not (0) | / | Definitely yes (6) | |
| a) A benefit to you | ||||
| b) A benefit to society | ||||
| c) A benefit to children | ||||
| 3 | Perceived risk: | |||
| How likely do you think it is that the following would experience any negative health effects due to a) regular use of a hand-held mobile phone; b) living near a mobile phone mast? | Not at all (0) | / | Extremely (+6) | |
| i) You personally | ||||
| ii) The average person of your age and sex | ||||
| iii) The average child | ||||
| 4 | Regulatory preferences: | |||
| To what extent do you think the following should be restricted for potential health and safety reasons: | No restrictions (0) | / | Total ban (+6) | |
| a) Use of mobile phone handsets for adults | ||||
| b) Use of mobile phone handsets for children | ||||
| c) The building of base stations/masts | ||||
3. RESULTS
3.1. Preliminary Findings
3.1.1. Self-Reported Handset Use
The majority of respondents, 1,070 (81%), said they had a mobile phone. These results were similar to the U.K. average at the time (76%) (OFTEL, 2002). From the response to the three questions relating to use (i.e., make, receive, and text) we calculated a total time use per week for each respondent. Mean total reported use was just over two hours per week (M= 124.67 minutes, SD= 282.54) though a large negative skew (a few people reported very large amounts of use) meant that the median amount of time at 40 minutes per week is a more appropriate average. Due to the skew, we carried out a log transformation to create a more normal distribution suitable for further analysis.
3.1.2. Self-Reported Mast Proximity
Eighty-nine respondents (7%) reported living near an existing mast, 133 (11%) reported living near a proposed mast, 533 (44%) said they did not live near a mast. Four-hundred-and-sixty-nine (38%) said they did not know. Due to town planning records, we know that all respondents lived within 2 miles of a mast. However, it is not clear how this objective distance relates to the subjective distance of “near.” Preliminary analysis found no important differences in risk perception as a function of this question so future analyses do not differentiate people according to this variable.
3.2. Perceptions of Mobile Phone Technology
Our first general prediction was that the potential health risks associated with handsets would be perceived to be more controllable and voluntary than those associated with masts. To investigate we carried out a 2 (risk source: handsets/masts) × 2 (risk dimension: controllable/voluntary) within-participant analysis of variance (ANOVA). There was a large main effect of risk source, F(1,1263) = 1044.54, p < 0.001, η2= 0.45, but not of risk dimension, F(1,1263) = 0.63, ns. In line with predictions, handset risks were perceived as both more controllable (M= 0.20, SD= 1.82) than mast risks (M=−1.74, SD= 1.62), F(1,1263) = 1003.99, p < 0.001, η2= 0.44) and more voluntary, or at least less involuntary (handsets: M=−0.30, SD= 1.66; masts: M=−1.31, SD= 1.61), F(1,1263) = 356.24, p < 0.001, η2= 0.22. The correlation between controllability and voluntariness was r= 0.33, p < 0.001 for handsets, and r= 0.50, p < 0.001 for masts, suggesting the two dimensions were positively related such that a lack of voluntariness was associated with a lack of control and vice versa. For future analysis we collapse the scores for each of the hazards together to form a single “controllability” score.
3.3. Comparative Risk
Second, we predicted that mobile phone users would show comparative optimism for the technology in general (collapsing across handsets and masts) and that children would be seen as particularly at risk (H2a). We investigated this using a 2 (risk source: handsets/masts) × 3 (risk target: self/peer/children) within-participant ANOVA with planned contrasts (Table II). Supporting predictions, self-risk (M= 2.81) was perceived to be lower than peer risk (M= 3.08), F(1,1026) = 153.90, p < 0.001, η2= 0.13, for the technology as a whole (planned comparison CO1, Table II). Moreover, children were perceived to be at more risk (M= 4.07) from the technology in general than other adults (M= 3.08), F(1,1026) = 698.50, p < 0.001, η2= 0.41 (planned comparison CO2, Table II).4
| Perceived Risksa | Perceived Benefitsb Handsets | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Handsets | Masts | ||||||||
| Self | Others | Children | Self | Others | Children | Self | Society | Children | |
| Msc | 2.46 | 2.99 | 4.27 | 3.23 | 3.35 | 4.09 | 4.58 | 3.36 | 2.82 |
| (SDs) | (1.71) | (1.50) | (1.49) | (1.62) | (1.51) | (1.55) | (1.81) | (1.81) | (2.09) |
| Msd | 2.46 | 2.91 | 4.15 | 3.15 | 3.25 | 3.99 | 5.10 | 3.61 | 2.97 |
| (SDs) | (1.60) | (1.44) | (1.50) | (1.57) | (1.47) | (1.55) | (1.19) | (1.73) | (2.06) |
| CO1e | 1 | −1 | 0 | 1 | −1 | 0 | - | - | - |
| CO2e | 0 | 1 | −1 | 0 | 1 | −1 | - | - | - |
| CO3e | 1 | −1 | 0 | −1 | 1 | 0 | - | - | - |
| CU1e | - | - | - | - | - | - | 1 | −1 | 0 |
| CU2e | - | - | - | - | - | - | 0 | −1 | 1 |
- aFrom 0 “Not at all” to 6, “Extremely, likely”; bFrom 0 “Definitely not” to 6 “Definitely yes”; cTotal sample means; dMeans for users only (N= 1,027). Planned contrasts are based on these means; eRows contain the planned contrast coefficients for the specific predictions about comparative optimism (CO) and comparative utility (CU). See text for more details.
In line with H2b, however, the overall level of optimism for the self over one's peers was moderated by risk source. Specifically, comparative optimism was higher for the more controllable risk, handsets (MDiff= 0.45), than for the less controllable risk, masts (MDiff= 0.10), F(1,1026) = 104.16, p < 0.001, η2= 0.09 (see CO3 Table II). Moreover, supporting H3b, the correlation between reported use and comparative optimism (i.e., the difference between self and other ratings) was r=−0.20, p < 0.001. This suggests that people who used their mobiles a lot were less optimistic than those who used it less.
3.4. Comparative Benefit
To investigate predictions regarding comparative utility (H3a) we used a one-way ANOVA (risk target: self/peer/children) with planned contrasts (Table II). As predicted, users rated their own benefits as higher (M= 5.10) than those to society in general (M= 3.61), F(1,1019) = 782.64, p < 0.001, η2= 0.43 (planned comparison CU1). In turn, the benefits to society were rated as higher than those to children (M= 2.97), F(1,1019) = 100.21, p < 0.001, η2= 0.09 (planned comparison CU2). Also in line with predictions (H3b), comparative optimism increased as level of self-reported use increased, r= 0.25, p < 0.001. When combined with the findings for comparative optimism, this suggests that people who use their mobiles a lot believe they are at more risk but that this is to some extent compensated by higher levels of benefit. For ease of comparison, the whole sample means for risk and benefit by target (which were very similar to those for just users) are presented in Fig. 1.
Perceived likelihood of negative health effects to the self, other adults, and children from handsets and masts and perceived benefits for the self, society, and children (full sample).
3.5. Regulatory Preferences
The correlations between perceptions of risk, benefit, and regulatory preferences as a function of target are shown in Table III. Before discussing the regulatory preference results we want to draw attention to two things. First, the correlations between the three targets in terms of benefit (rs ≥ 0.30), mobile risk (rs ≥ 0.33), and mast risk (rs ≥ 0.64) are all moderate to high. Second, the negative correlations between mobile phone risk and benefit were r=−0.16, p < 0.001 for peers/society, r=−0.18, p < 0.001 for children, but only r=−0.03, ns for the self. The implications of these results are examined in Section 4.
| (1) | (2) | (3) | (4) | (5) | (6) | (7) | (8) | (9) | (10) | (11) | |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Benefits | |||||||||||
| Self (1) | - | - | - | - | - | - | - | - | - | - | - |
| Society (2) | 0.46* | - | - | - | - | - | - | - | - | - | - |
| Child (3) | 0.31* | 0.47* | - | - | - | - | - | - | - | - | - |
| Mobile risks | |||||||||||
| Self (4) | −0.03 | −0.06 | −0.03 | - | - | - | - | - | - | - | - |
| Peer (5) | −0.16* | −0.16* | −0.11* | 0.70* | - | - | - | - | - | - | - |
| Child (6) | −0.14* | −0.18* | −0.18* | 0.33* | 0.45* | - | - | - | - | - | - |
| Mast risks | |||||||||||
| Self (7) | −0.10 | −0.13* | −0.08 | 0.49* | 0.52* | 0.44* | - | - | - | - | - |
| Peer (8) | −0.14* | −0.32* | −0.10 | 0.46* | 0.58* | 0.49* | 0.89* | - | - | - | - |
| Child (9) | −0.13* | −0.15* | −0.13* | 0.26* | 0.37* | 0.67* | 0.64* | 0.70* | - | - | - |
| Regulation | |||||||||||
| Adult (10) | −0.30* | −0.24* | −0.12* | 0.12* | 0.23* | 0.20* | 0.21* | 0.25* | 0.20* | - | - |
| Child (11) | −0.25* | −0.22* | −0.37* | 0.06 | 0.16* | 0.31* | 0.18* | 0.21* | 0.30* | 0.49* | - |
| Masts (12) | −0.16* | −0.18* | −0.11* | 0.18* | 0.27* | 0.37* | 0.39* | 0.42* | 0.43* | 0.28* | 0.33* |
- *p < 0.001.
Our first predictions regarding regulatory preferences suggested that preferences would be higher for the uncontrollable risk (masts) than the controllable risk (handsets) (H4a), and that regulatory preferences for handsets would be higher for children than adults (H4b). To investigate we used a one-way ANOVA with three levels (adult handsets, children handsets, society masts). A significant main effect was found, F(2,2430) = 1085.74, p < 0.001, η2= 0.47, and planned contrasts suggested that, as predicted, the regulatory preferences were higher for masts (M= 4.12, SD= 1.44) than handsets when aggregated across regulation for both adults and children (M= 2.74, SD= 1.53, F(1,1215) = 867.09, p < 0.001, η2= 0.42). Moreover, regulatory preferences over the use of handsets was higher for child (M= 3.64, SD= 1.82) than adult use (M= 1.79, SD= 1.63, F(1,1215) = 1347.21, p < 0.001, η2= 0.53). Qualitative evidence in support of this finding also came from additional comments made at the end of the survey. For example, one respondent said: “Use of mobile phones for young children under 10 should be banned completely until we know for certain there are no health risks” (44-year-old female). Another argued: “The long-term effect of using mobile phones on a regular basis is not known at present, and will not be known for many years. This is why I think children should be restricted in their use even though for personal safety reasons they could be useful” (59-year-old male).
We also predicted that due to comparative optimism and comparative utility for handsets, personal benefit but risk to others would be the strongest predictors of regulatory preferences for adult use (H4c). We examined these predictions by regressing regulatory preferences on all four variables simultaneously (self-risk, other risk, self-benefit, societal benefit; Table IV). In line with predictions, self-benefit (β=−0.23, p < 0.001) was more important for regulatory preferences than societal benefit (β=−0.10, ns), and other risk (β= 0.21, p < 0.001) was more important than self-risk (β=−0.02, ns). In sum, while higher perceived risks to others were associated with greater preferences for regulation, higher benefits for the self were associated with lower regulatory preferences.5
| Handset Use Regulation | Mast Construction Regulation | ||
|---|---|---|---|
| Adult | Children | ||
| Risk | |||
| Self | −0.02 | - | 0.03 |
| Peers | 0.21* | - | 0.22* |
| Children | - | 0.22* | 0.24* |
| Benefit | |||
| Self | −0.23* | - | −0.07 |
| Society | −0.10 | - | −0.08 |
| Children | - | −0.32* | −0.01 |
| R 2 adj | 0.14* | 0.21* | 0.26* |
- *p < 0.001.
Since we did not expect comparative optimism for masts, we predicted that the risks and benefits to all three targets could potentially play a role in predicting regulatory preferences for masts (H4d). Results, however, found significant effects only for peer (β= 0.22, p < 0.001) and child risk (β= 0.24, p < 0.001), but not self-risk (β= 0.03, ns) and no additional impact of benefit perceptions (all βs < −0.08, ns). However, due to the high interitem correlations between target risk perceptions for masts (all rs > 0.64) it may be that some of the effects of self-risk are being masked by this covariation.
Finally, although we had no specific hypothesis for regulation of children's handset use, we found that these were predicted by perceptions of both child risk (β= 0.22, p < 0.001) and benefit (β=−0.32, p < 0.001). As with adults, preferences were higher when the risks to children were perceived to be high, and lower when the benefits were perceived as high.
4. DISCUSSION
The present research investigated perceptions of the risks and benefits of mobile phone technology with regard to different targets (i.e., the self, peers, society in general, and children) and the relations of these perceptions to regulatory preferences. Understanding such preferences is important for policymakers responsible for the risk management of new technologies since it can aid their prediction of when and where negative public reactions are likely to occur.
First, we found comparative optimism, i.e., the belief that one is less at risk than one's peers (Weinstein, 1980), for the potential risks of mobile phone technology. However, in line with earlier findings (e.g., Harris, 1996) this optimism was moderated by the perceived controllability of the risks (handsets vs. masts) and reality constraints concerning the amount of risk exposure (i.e., the amount of self-reported use, e.g., McKenna et al., 1993). Second, we also found evidence of comparative utility, i.e., the belief among users that the benefits of the technology are larger for the self than society in general. To our knowledge, this is the first time that anyone has reported such a finding in the context of risk perception. Third, we identified a particularly vulnerable group of the population, i.e., children, and asked about perceptions of risk and benefit for this subgroup. In general, the risks were perceived to be higher and the benefits lower for children, compared to the self and one's peers. Fourth, and most important for policy, the findings relating to different targets directly affected regulatory preferences.
Specifically, regulation preferences for the more controllable aspect of the technology (i.e., handsets) were best predicted by perceptions of the risks to others, but perceived benefits to the self. The more people perceived the risks to others and the less they saw benefits to the self, the stricter the regulation they wanted. Although the data were less clear for the uncontrollable risk (masts), it seemed that regulatory preferences were primarily driven by perceptions of the risks to others (peers and children) rather than possible risks to the self. Nevertheless, the high correlations between risk targets for this aspect of the hazard mean that greater caution is required in interpreting these findings. The results with respect to children are intuitively appealing because much of the public demonstration in the United Kingdom focused around the siting of masts in and around schools and hospitals (i.e., vulnerable others). Again, the findings offer support for the central message of the current article, namely, that differentiating between risk and benefit targets may aid the understanding of public reactions to hazards.
The findings have a number of implications for risk communication strategies. For instance, communicators who want to raise public awareness of a hazard may find it better to focus on the risks to others in general, and vulnerable others in particular, rather than targeting individuals themselves since calls for regulation were primarily driven by the perceived risks to others. This may be important for hazards about which much is already known as well as relatively new ones. Much of the focus on smoking risk communication, for example, has been to inform smokers of the risks they personally face (e.g., risk labeling on packets). However, as noted above, even smokers appear to attenuate their risks relative to other smokers so this may result in little behavioral change (McKenna et al., 1993). Focusing instead on the risks smoking has on other people (i.e., passive smoking) may help promote behavioral change and greater acceptance of anti-passive smoking regulation. Recent research commissioned by the U.K.'s Health Promotion Agency (2005) supports this suggestion. From November 2004 to February 2005 the agency sponsored an advertising campaign in Northern Ireland explaining the risks of passive smoking. Following the campaign, roughly 40% of smokers said the campaign had changed their attitude toward passive smoking and that they would be more careful in future about smoking near others. Warning smoking mothers that they may be harming their (vulnerable) unborn babies is another strategy whereby the risks to others rather than the self are being emphasized. Further research could investigate these possibilities more systematically by, for instance, providing people with risk and benefit information relating to different targets and observing attitudinal and behavioral shifts as a consequence.
In addition to our main analyses we also replicated two findings reported elsewhere in the literature. First, the correlations between the three targets in terms of risk and benefit perceptions were relatively strong (rs ranged between 0.31 and 0.89; see also Poortinga & Pidgeon, 2005). Does this mean that distinguishing between targets is therefore unnecessary? We believe not. Similarly high correlations are found elsewhere in the literature on optimistic bias but the absolute differences remain potentially important (e.g., Pahl et al., 2005). Imagine, for instance, that a person's threshold to act on a risk lies at the point where they mark a two on a six-point scale from zero to six. Below this point they believe no defensive action is necessary, above this point action is needed. If they rate their own risk as a one and the risks to the average person as a three, reflecting comparative optimism, it suggests they believe defensive actions are only necessary for others and the opportunity to engage in self-protective behaviors may be foregone. Just because people's judgments of the risks to the self and others are correlated does not mean we should simply collapse these judgments (cf. Sjöberg, 2000).
Second, we found low but significantly negative correlations between risk and benefit judgments for society in general (r=−0.16) and children (r=−0.18). These correlations echoed the findings of Siegrist et al. (2005) for the same technology (r=−0.18). This negative correlation is in keeping with the notion of an “affect heuristic,” suggesting that people may draw their risk/benefit judgments from a basic attitude toward the hazard (e.g., Finucane et al., 2000). While this idea has received considerable attention recently, the correlations reported by Finucane et al. (2000, Study 1) in the no time pressure condition (similar to our study) were also small (mean r=−0.12). Indeed, only two of the 24 hazards had a significant inverse relationship, mobile phones and cars. The effects in the time pressure condition were much stronger (mean r=−0.37), suggesting that people may in part draw more on this general evaluative framework when less cognitive resources are available.
Contrary to the affect heuristic hypothesis, we did not find a significant inverse relationship between mobile phone risk and benefit perceptions for the self, r=−0.03, ns. These effects are mirrored for masts (peers r=−0.32, p < 0.001; self r=−0.10, ns). Although it is not fully clear why, it is conceivable that the judgment processes involving estimates for the self are qualitatively different from those concerning judgments of others. For instance, while the self is a concrete and familiar target where many of the personal benefits, and at least some of the risks, can be assessed quite well, society in general (the target generally used in this research) is more ambiguous, potentially resulting in more heuristic processing. That is, just as time pressure may enhance the use of the affect heuristic, so might the use of a target about which one has relatively little direct knowledge. Future research could examine whether the inverse correlations between risk and benefit are consistently smaller across a range of hazards for the self compared to others, potentially adding to our understanding of the processes underlying the use of this heuristic.
Some notes of caution regarding the results of the current research are necessary. First, since we only directly asked about benefits of mobile phones, rather than for masts separately, the lack of significant effects of benefits for masts may be due to this lack of specificity rather than any lack of importance of perceived benefits for this aspect of the technology. Second, the high intertarget correlations for mast risk perceptions means that the importance of self-risks for regulatory preferences may be underestimated for statistical reasons rather than because they were unimportant per se. Examination of other hazards is therefore needed to see whether the findings are replicable in other contexts. Third, while there was consistency between risk and benefit judgments for both the self and children, there were differences for others in general, i.e., “peers” for risk and “society” for benefits. Future research may benefit from making these targets consistent and by asking about benefits for each aspect of a hazard (e.g., Siegrist et al., 2005). Fourth, the issue of mobile phones is probably unique in terms of the generally positive reception it has received from the public. As such, many of the issues examined here need to be explored for other hazards to establish generalizability. Despite this concern, however, our findings of comparative optimism, and the factors that moderate it, are consistent with a host of other risk-related issues, suggesting the technology is not so different from other risks (e.g., Weinstein & Klein, 1996).
Finally, the amount of variance in regulatory preferences explained by our regression models was moderate. There are a number of possible reasons for this. First, the models contained very few explanatory variables. Adding further variables, such as demographics, would have increased the amount of variance explained but this was not our primary purpose (see footnote 3). Rather than trying to build the model that best predicts regulatory preferences, our aim was to show the value of differentiating between targets for this purpose. Second, ultimately our predictors were based on single items, which are less reliable than scales due to the greater variance associated with just a single item. Given the high correlations between perceptions of risk and benefit we could have created scales as other researchers have done. However, our aim here was to deliberately avoid collapsing responses across targets. Future research would profit from the creation of scales for the perceived costs and benefits for different targets, and indeed scales for outcomes such as regulatory preferences.
To conclude, the current article echoes Sjöberg's (2000) call for greater attention to be paid to exactly who the risk target is in any judgment of risk. The article extends this call further by showing that differentiating between targets should also extend to perceived benefits, and that doing so has implications for predicting key outcomes such as stated preferences for hazard regulation. While risk perception researchers may have accepted Sjöberg's message in general terms, they may have not been fully aware of its implications. We hope the current article has drawn attention to just some of the potential that differentiating between targets has for both applied and theoretical purposes.
Footnotes
ACKNOWLEDGMENTS
The research was funded by the U.K.'s Health and Safety Executive (HSE), Contract: 4104/R71.046. As a result, they have a keen interest in understanding risk and risk perception, especially in work-related environments and new technologies. The funding was secured by Eiser for a two-year project, following a call for original research. The HSE waived all rights to review or approve the manuscript. Their primary interest was in the research reports submitted as part of the project which were submitted some years earlier.
Since the study involved human participants, the form asked that we detail the ethics procedures that were followed. The study was conducted at the University of Sheffield when an internal panel assesses all proposed research to check that it will be carried out in accordance with the British Psychological Society's ethical charter.
We would like to thank Wouter Poortinga and two anonymous reviewers for helpful comments on an earlier draft.
