On the Relation Between Trust and Fairness in Environmental Risk Management
Abstract
In this study, we empirically examine the relations between trust, fairness, and cooperation within two environmental risk management contexts, one in which the focal issue is of high personal moral importance and the other in which the focal issue is of low moral importance. Using an experimental design embedded in two parallel survey questionnaires, one mailed to residents of Washington State, the other to German-speaking residents of Switzerland, we either manipulated or constructed three factors, issue importance (high/low), procedural fairness (fair/unfair), and policy outcome (risk averse/risk accepting). This design enabled us to compare the predictions of the standard account of procedural fairness, that trust and cooperation are determined by judgments of fairness, with the predictions of an alternative account, that trust and cooperation will be determined by judgments of procedural fairness only when the issue involved is not morally important. Results for the American case showed that under conditions of high issue importance, policy outcome affected judged fairness, trust, and cooperation. Under conditions of low issue importance, policy outcome had no effect on judged fairness or trust but did have a moderate impact on cooperation. Analyses also showed that when issue importance was high, procedural fairness had no effects. When issue importance was low, procedural fairness had moderate effects on judged fairness and trust. Results for the Swiss case replicated the main findings for the American case. Together, these results support the alternative model of the relation between trust and fairness, suggesting that the efficacy of fair procedures is strictly limited.
1. INTRODUCTION
In response to recent calls for increased public participation in risk communication and risk management processes, both practitioners and researchers have identified trust and fairness (in the sense of fair procedures) as being key factors in the development of cooperation between risk management professionals and public groups (Beierle & Cayford, 2002; Carolan, 2006; Charnley, 2000; Gregory, 2002; Gregory & McDaniels, 2005; Gutrich et al., 2005). Trust and fairness are widely assumed to be closely related, with fair treatment of a person or group by an official or institution leading to a relationship of trust, which in turn leads to cooperation (De Cremer, 1999; Konovsky, 2000; Konovsky & Pugh, 1994; Lind, 2001). The intended role of fairness, in both contemporary environmental risk management (Renn et al., 1995) and in classical liberal thought (Sandel, 1996; Verba, 2006), is to provide a common ground that is valued by all sides, and, as such, can provide a neutral basis for the resolution of disputes.
Common experience with controversial issues, such as capital punishment, immigration, abortion, nuclear power, stem cell research, and the sale of human organs, demonstrates, however, that fair processes do not always justify unwanted outcomes. What these issues have in common is that, for many people, they are morally charged—or personally morally important. For these persons, the moral importance of the issue is greater than the moral importance of fairness. A problem with fair procedures, then, is that they may not work in difficult cases. Fairness may not lead to trust and cooperation when most needed, in cases contested by persons with differing morality-based positions on issues they consider to be of central importance.
In this study, we empirically examine the relations between trust, fairness, and cooperation within two environmental risk management contexts, one in which the focal issue is of high personal moral importance and the other in which the focal issue is of low moral importance. To our knowledge, no previous study has investigated the effects of issue importance on trust and fairness in risk management contexts. Using an experimental design embedded in two parallel survey questionnaires, one mailed to residents of Washington State, the other to German-speaking residents of Switzerland, we manipulated three factors, issue importance (whether an issue is judged by pretest respondents to be personally morally important: high/low), procedural fairness (whether a decision-making process is judged by pretest respondents to be fair: fair/unfair), and policy outcome (whether a policy advocates risk mitigation actions or denies the need for such actions: risk averse/risk accepting).3 This design enabled us to compare the predictions of the standard account of procedural fairness—that trust and cooperation are determined by judgments of fairness (Tyler, 2001a)—with the predictions of an alternative account based on our TCC (Trust, Confidence, Cooperation) model (Earle & Siegrist, 2006)—that trust and cooperation will be determined by judgments of procedural fairness only when the issue involved is not morally important. Before turning to the specifics of this study, we first review (1) current research on trust and risk management, (2) studies investigating the role of fairness in risk management, (3) the standard model of trust and fairness, (4) our alternative model of trust and fairness, and (5) the theoretical and empirical rationale for the moderating effect of issue importance on the relation between trust and fairness.
2. TRUST AND RISK MANAGEMENT
The focus of most studies of trust in risk management has been either the relation between trust and risk perception or the relation between trust and some form of cooperation. A recent review indicated that there is little agreement among researchers on how trust in risk management should be studied (Earle et al., 2007). Although most studies of trust in risk management have been ad hoc, atheoretical attempts to generate practical understandings of trust in specific contexts, they have generated an impressively uniform set of core results (Earle et al., 2007). Trust is based on agreement or similarity, and trust affects judgments of risks and benefits, which lead in turn to various forms of cooperation (for a contrary view, see Poortinga & Pidgeon, 2005). Earle et al. (2007) found, however, that the magnitude of the relations between trust, risk perception, and cooperation were highly variable across studies, affected by certain contextual factors. In addition to agreement on values or policies, knowledge of the focal issue appears to be important: all else equal, the relation between trust and risk perception is stronger when knowledge level is low (Siegrist & Cvetkovich, 2000).
Earle et al. (2007) also found that, within the context of risk management (as well as within other fields of application), it is useful to distinguish between two “types of trust,” or pathways to cooperation. One, called trust, is based on a judgment of similarity of intentions or values. The second, called confidence, is based on experience or evidence. This distinction has been supported in a number of empirical studies (Earle & Siegrist, 2006; Siegrist et al., 2003) and is similar to distinctions made by trust researchers in fields other than risk management (e.g., Das & Teng, 1998; Kramer, 1999; Rousseau et al., 1998). Within the context of risk management, Metlay (1999) empirically identified two pathways to cooperation similar to trust and confidence, while both Frewer et al. (1996) and Poortinga and Pidgeon (2003) found two slightly different pathways. This study is based on the dual-pathway TCC model (Earle & Siegrist, 2006; Earle et al., 2007; Siegrist et al., 2003).
2.1. The TCC Model
The TCC model hypothesizes that trust and confidence provide two interacting pathways to cooperation (Earle et al., 2007). Trust is based on social relations—shared values or intentions (e.g., in-group membership, morality, benevolence, caring, etc.). The basis for confidence is past performance, or institutions designed to constrain future performance (e.g., evidence, regulations, rules and procedures, indicators of competence, etc.). Both trust and confidence, in a range of combinations, can lead to various forms of cooperation (e.g., support for a policy, purchase of a product, etc.). The TCC model is shown in Fig. 1, with trust on the upper path and confidence on the lower.4 At the far left of the model, the information perceived by a person is divided into two types, that which is judged to be relevant to “morality” and that which is judged relevant to “performance.” Morality information is equivalent to actions reflecting the values of an agent, and performance information is simply the behavior of an object (i.e., an entity not granted agency). The agent's values define a relationship of trust, and it is within that relationship that the performance information and the confidence to which it leads are judged. Two recent studies using structural equation techniques have confirmed the primary paths described in the TCC model (Earle & Siegrist, 2006; Siegrist et al., 2003).
The TCC model of cooperation based on trust and confidence.
Source: Earle and Siegrist (2006, p. 387).
3. FAIRNESS AND RISK MANAGEMENT
In contrast to the many studies of trust in risk management, fairness in risk management has been relatively neglected by researchers. Evaluations of public participation processes indicate that stakeholder involvement has been generally successful both in terms of process quality and outcome quality (Beierle, 2000; Beierle & Cayford, 2001, 2002). In addition to large-scale evaluations, a number of case studies have focused specifically on the fairness of processes and outcomes. Some of these studies (e.g., Hunt & Haider, 2001; Lauber & Knuth, 1999; Smith & McDonough, 2001) have based their evaluations on social-psychological fairness research (e.g., Tyler, 2000); other studies have drawn on the work of the philosopher Jürgen Habermas (Webler & Tuler, 2000, 2002). The most sustained investigation of fairness in environmental risk management has been conducted by Syme and colleagues (Syme & Nancarrow, 1997; Syme et al., 1999, 2000). A major focus of Syme's work, however, is on knowledge of fairness norms rather than on actual judgments of fairness (a problem also pointed to by Lerner, 2002). When judgments of fairness are examined, they turn out to be positively related to outcomes: the more favorable the outcome, the fairer the process is judged to be. Thus, in practice as opposed to theory, fairness judgments appear to be instrumental. Syme refers to this contrast as situational fairness versus universal fairness, and he acknowledges that, in the case of important issues, the former tends to dominate the latter.
In psychology, a number of empirical studies have demonstrated the limits of fairness. Rasinski (1987) showed that fairness judgments are ideological expressions of the judge's values. As Rasinski points out, his findings have significant practical implications: “to the extent that opposing values are held by constituents, policies and actions designed to appear fair may be judged as unfair” (Rasinski, 1987, p. 210). Several recent studies of affirmative action programs have documented the rationalizing function of fairness judgments (Bobocel et al., 1998; Federico & Sidanius, 2002; James et al., 2001). Voice, or the opportunity to express one's opinions, has been used by many researchers as an indicator of fairness. Avery and Quiñones (2002), however, have shown that it is voice instrumentality, having one's views positively responded to, that is critical to fairness, not voice opportunity or even voice behavior (for a review of the relations between procedural fairness, outcome fairness, and outcome favorability, see Skitka et al., 2003).
The most general and fundamental critique of fair procedures is that fairness is within group. Peterson, for example, concluded that “people must share a common understanding before the procedures used by supervisors can temper negative outcomes. In the absence of this shared understanding or identification, disputants focus almost exclusively on outcomes” (1999, p. 322). Opotow (1995), Platow and colleagues (Platow et al., 1995, 2000), and Wenzel and colleagues (Mikula & Wenzel, 2000; Wenzel, 2002) have also demonstrated within-group fairness. Opotow and Weiss (2000) apply their notion of moral exclusion to the analysis of environmental conflict.
4. TRUST AND FAIRNESS
In contrast to the critical research reviewed above, which implies that trust within a group leads to judgments of fairness, the standard, widely accepted, and cited model of fairness (e.g., De Cremer, 1999; Konovsky, 2000; Konovsky & Pugh, 1994; Lind, 2001) holds that fair procedures lead to trust. Some fairness researchers, however, seem to find the relation between trust and fairness difficult to pin down. Tyler, for example, has at times said that fairness leads to trust (Tyler, 2001a, 2001b); on other occasions, he has said that trust leads to fairness (Tyler, 2000, 2001c; Tyler & Degoey, 1996). Tyler should not be faulted for inconsistency, however, because, as we will argue, the arrow, depending on context, can point either way. When trust and/or fairness become salient (i.e., within the domain of cooperation), the key contextual factor, as Van den Bos, et al. (2001) have pointed out, is the type of information available. Faced with a decision on whether to cooperate, people make use of whatever information is at hand. When information with value implications is available (e.g., information allowing a relevant categorization), people will use it as a basis to judge trustworthiness (Earle & Siegrist, 2006). Of course, information about fair treatment can (but does not necessarily) have value implications. If it is the only value-implicative information available, or if fairness is the dominant category, then fair treatment will lead to trust (or act as a surrogate for trust). This is the “fairness heuristic” process described by Lind (e.g., Lind, 2001) and generalized by van den Bos and Lind (2002). (For a challenging critique of fairness heuristic theory, see Árnadóttir, 2002, and also Smedslund, 1997.) When no information about fair treatment is available, or when such information is dominated by other value-implicative information, fairness is judged within the group formed by the dominant values. This is the process by which trust leads to fairness.
One way to see the close, integrated relationship between trust and fairness is to compare researchers' accounts of the context of fairness to our account of the context of trust. Lind (2001), for example, describes what he calls the “fundamental social dilemma”: whether to cooperate with others and expose oneself to the risks, as well as the benefits, of doing so. One way to resolve this dilemma is by using fairness as a heuristic for trust. But the more general solution, of course, is trust itself, which can be based on a vast array of values in addition to (or instead of) fairness. In most cases, fairness is an epiphenomenon, a by-product, of trust. We expect fairness from those we trust (Árnadóttir, 2002). Another comparison is provided by Greenberg's (2001) account of the conditions that generate concerns about fairness. When one receives negative outcomes, for example, one becomes concerned about fairness. Similarly, violations of expectations, periods of change, scarce resources, unbalanced power relations—as well as the general need to belong (De Cremer & Blader, 2005)—all these circumstances lead to concerns with fairness. But these conditions—in general, loss of confidence, uncertainty—also lead to concern with the more inclusive relationship of trust. For Luhmann, the subject's leading theorist, the fundamental function of trust is to reduce uncertainty (Luhmann, 1979).
5. INTEGRATING TRUST AND FAIRNESS IN THE TCC MODEL
In the previous sections outlining the conceptual background for this study, we have argued the following points: (1) trust, based on shared values, and confidence, based on past performance, are separate, but interacting, pathways to cooperation; (2) in risk management contexts, trust is within group and based on agreement and similarity; (3) judgments of fairness can lead to trust when fairness is the dominant value or when no other trust-relevant information is available but, more often, trust leads to perceived fairness; (4) in risk management contexts, fair processes are sometimes effective (e.g., in low-conflict, within-group situations) and sometimes ineffective (e.g., in high-conflict, between-group situations). Although current theory and research are far from settled on an understanding of the relations between trust and fairness, the TCC model provides a framework within which some aspects of those relations can be comprehended more clearly.
Procedures and processes are either rules designed to constrain future behavior or behavior guided by such rules. Judgments of the fairness of procedures or processes can refer either to the rules (e.g., prospectively, “Is the procedure fair?”) or to behavior (e.g., retrospectively, “Was the process fair?”). In both cases, procedures and processes play the same role as performance in the TCC model (Fig. 1). A process that is judged to be fair can thus lead to confidence and cooperation. But the evaluation of performance and the judgment of confidence are both affected by trust. In the case of high levels of trust, a process can be judged fair and lead to confidence and cooperation. In the case of distrust, a process cannot lead to confidence and cooperation (De Cremer & Tyler, 2007; Drolet et al., 1998). Distrust occurs between two groups (or members of those groups). Thus, any factors that lead to the creation of two conflicting groups can make fair procedures ineffective. Conversely, factors that lead to the formation of a single group can make fair procedures effective. In this scenario, trust leads to fairness.
The TCC model also accommodates the fairness leading to trust scenario. When information about procedures or processes is the only information available (i.e., there is no category-relevant information about the other), or if it is the most value-relevant information available, then that information will be assessed for its value implications and used as the basis for judging trust. In addition, fairness as a value could be the dominant value in a situation in which other value-implicative information is available. In that case, fairness will also lead to trust and cooperation.
5.1. Issue Importance
Throughout our discussion of trust, fairness, and risk management, we have seen that trust seems to dominate fairness in contexts characterized by conflict and issues with high value implications, morally important issues. Recently, Skitka and colleagues (Skitka, 2002; Skitka et al., 2005; Skitka & Houston, 2001; Skitka & Mullen, 2002) have investigated the relations among procedural fairness, outcome preferences, and issue importance. According to Skitka's value protection model, “people's outcome preferences will shape how they reason about fairness when they are rooted in moral conviction” (Skitka & Mullen, 2002, p. 1421). When outcome preferences are rooted in moral conviction, they are, by definition, important values and, as such, form the basis for trust relations. Those trust relations provide the context within which procedural fairness is judged. When, on the other hand, outcome preferences have no moral implications, they are a weak basis for trust. In those cases, procedures may offer the most relevant value-implicative information available, leading to judgments of fair procedures and of trust. A number of other research programs, such as the work of Tetlock and colleagues on “taboo trade offs” (Fiske & Tetlock, 1997; Tetlock, 2000) and the studies of mortality salience by Solomon et al. (2000), also demonstrate the effects of issue importance on subsequent judgments.
For personally important risk issues, outcomes, expressed or interpreted in terms of values, dominate the effects of procedures, and cooperation is based on trust. Strong emotional reactions can be evoked by personally important risk issues. An archetypal personally important risk issue is gun ownership, as demonstrated by a recent energetic exchange among gun-risk researchers (Kahan & Braman, 2003a, 2003b). The key point made by Kahan and Braman is that there are contexts in which empirical evidence is useful in producing cooperation (i.e., is persuasive to researchers on opposing sides of an issue) and there are contexts in which it is not useful. The two types of situations are distinguished by issue importance, which is low in the first case and high in the second. Thus, pro-ownership gun-risk researchers are unconvinced by the evidence produced by anti-ownership researchers, and vice verse. In our TCC model, empirical evidence occupies the same confidence path to cooperation as procedures. The findings of Kahan and Braman thus are in accord with the TCC model and opposed to the standard model of trust and fairness (e.g., Tyler, 2000), which does not distinguish between issues that are personally important and those that are not. For risk issues that are not personally important, and that do not evoke strong emotional reactions, procedures can play a role in generating confidence and cooperation.
6. THE PRESENT STUDY
This study employed an experimental design embedded in survey questionnaires mailed to residents of Washington State (the American case) and to residents of the German-speaking part of Switzerland (the Swiss case). In the American case, the study design was based on scenarios that described two risk management issues, one that was determined by pretesting to be highly morally important to respondents and the other that was identified as being not morally important. The Swiss case employed a single risk management issue, and a median split on the judged personal importance of the issue was used to create two groups of respondents. In both cases, the scenarios also described two sets of procedures, one that pretests showed that respondents considered to be fair and the other that respondents considered to be unfair. Finally, the scenarios described two policies designed to manage each of the two hazards (single hazard in the Swiss case). One policy was risk averse, increasing controls over the hazard; the other policy was risk accepting, minimizing controls over the hazard. The study design, then, was a 2 (issue importance: high/low) × 2 (procedures: fair/unfair) × 2 (policy outcome: risk averse/risk accepting) factorial design. In this study, we test the implications of the TCC model using analysis of variance (ANOVA) techniques: under conditions of low issue importance, fair procedures can determine judgments of fairness and trust; under conditions of high issue importance, judgments of fairness and trust are determined by outcome preferences.
7. PRETESTS
Before turning to the specifics of this study, we first describe the pretesting work that was necessary to establish the relative fairness of procedures used in the study and the relative moral importance of the risk management issues. Separate pretesting was done in the United States and in Switzerland.
7.1. Procedural Fairness
7.1.1. Method
7.1.1.1. Participants. In the United States, 146 participants were recruited from university students taking courses in psychology and business. In Switzerland, 175 social psychology students volunteered to participate in the pretests.5 In both cases, participants responded to the research materials in groups.
7.1.1.2. Materials. The same materials, presented as questionnaire booklets, were used in the United States and (translated into German) in Switzerland. Under the heading “The Fairness of Risk Management Procedures,” participants read the following introduction: “In developing risk management policies, public officials follow procedures designed to produce effective results. These procedures can vary, depending on the issue being addressed and on the specific goals the officials are trying to achieve. We are interested in your views on the fairness of a particular approach to risk management procedures. Following a brief description of that procedural approach, below, a set of questions asks for your judgments on its overall fairness.” Next, half the participants read about a procedural approach that, on the basis of full public participation, was designed to be fair: “There are two overriding goals to the process: technical soundness and full participation by concerned citizens. Decisions will be made by public officials who will be guided by the views of prominent experts on the issue. In addition, an open public participation process will solicit the opinions of the general public. Although time consuming and costly, the public process is designed to include the views of broadly diverse groups of citizens.” The remainder of the participants read about a procedural approach that, on the basis of an absence of public participation, was designed to be unfair: “There are two overriding goals to the process: technical soundness and efficiency. Decisions will be made by public officials who will be guided by the views of prominent experts on the issue. This process will assure the decisions will be based on the best information available, made in a timely manner, and at reasonable cost.”
7.1.1.3. Measures. Participants made their fairness judgments by responding to eight questions about the procedural description they read. For example: “Is the procedural approach unbiased or biased?” Responses were made on a five-point scale anchored by “Unbiased” (1) and “Biased” (5). The other items, with similar response scales, were: “Is it reasonable or unreasonable?,”“Is it impartial or partial?,”“Is it balanced or slanted?,”“Is it even-handed or one-sided?,”“Is it open-minded or closed-minded?,”“Is it objective or subjective?,” and “Is it fair or unfair?” Since lower scale scores indicated higher judgments of procedural fairness, responses were reverse scored for the analyses. Our aim was to reduce this set of items to four, if the reliability of the scale was not negatively affected.
7.1.2. Results
7.1.2.1. Reliability. The reliability of the eight-item procedural fairness scale was satisfactory for both the American (α= 0.89) and Swiss cases (α= 0.85). Further reliability analyses showed that a four-item scale would perform as well as the eight-item scale. The items selected for the final procedural fairness scale were: “Was the process unbiased or biased?,”“Was the process balanced or slanted?,”“Was the process even-handed or one-sided?,” and “Was the process fair or unfair?”
7.1.2.2. Relative fairness. For the United States, the fair procedure was judged to be fairer (M= 3.39, SD= 0.89) than the unfair procedure (M= 2.83, SD= 0.73), t(144) = 4.14, p < 0.001 (one-tailed). The results were similar for Switzerland: (M= 3.42, SD= 0.69) for the fair procedure and (M= 3.06, SD= 0.65) for the unfair procedure, t(173) = 5.64, p < 0.001 (one-tailed).
7.2. Issue Importance
7.2.1. Method
7.2.1.1. Participants. The same persons who participated in the procedural fairness pretest also participated in the issue importance pretest.
7.2.1.2. Materials. As with procedural fairness, the same materials were used in the United States and in Switzerland (translated into German). After the heading “The Personal Importance of Risk Management Issues,” participants read these instructions: “An important function of government is the protection of citizens and the environment from possible harmful effects of industrial products and processes. Our government must deal with many of these risk management issues. We are interested in how important to you, in a deep personal sense, a number of these issues are. For each of the issues identified in the first column of this table, please indicate your degree of agreement with each of the statements in the first row of this table. Do this by writing a number between 1 (Don't agree at all) and 5 (Agree entirely) in the appropriate box. Thus, for each issue you will make 5 judgments, entering a number between 1 (Don't agree at all) and 5 (Agree entirely) in each of the 5 boxes to the right of the issue.” Twelve risk management issues, selected to represent both emerging and established issues as well as a broad range of personal importance, were identified in the first column of the table: depletion of natural resources, such as fish stocks and forests; contamination of food supply in production, processing, or distribution; medical use of X-rays and radiation therapy; occupational health and safety; development, production, and sale of genetically engineered food; electromagnetic fields from high-voltage power lines; global climate change; structural safety of buildings; safety of municipal water supply; nuclear waste disposal; automobile accidents; and contamination of pharmaceutical drugs in production or distribution.
7.2.1.3. Measures. Participants made their judgments of personal importance for each of the 12 issues by indicating their degree of agreement with the following five statements: “This is a central concern of mine,”“This issue threatens values that are very important to me,”“My attitude toward this issue is a matter of personal conscience,”“This issue strikes to the heart of who I am,” and “This is a fundamentally important issue to me.” All judgments were made on the same five-point scale anchored by “Don't agree at all” (1) and “Agree entirely” (5).
7.2.2. Results
7.2.2.1. Reliability. The reliability of the five-item personal importance scale was satisfactory for both American (average α= 0.90) and Swiss cases (average α= 0.86). Deletion of none of the items would consistently improve reliability. Thus, all five items were retained for the final scale.
7.2.2.2. Relative personal importance. For the American case, the risk management issue with the lowest-judged personal importance was electromagnetic fields from high-voltage power lines (M= 1.88, SD= 1.02); the issue judged to be of highest personal importance was automobile accidents (M= 3.34, SD= 1.19). Mean personal importance for the 12 issues was 2.71. For inclusion in our survey study, we required two issues that differed in judged personal importance. In addition, the two issues had to be associated with plausible alternative management strategies, one that was risk accepting and one that was risk averse. Finally, the two issues had to be of current interest to potential study participants. For these reasons, we selected medical use of X-rays and radiation therapy as our low importance issue (M= 1.99, SD= 1.06) and depletion of natural resources, such as fish stocks and forests, as our high importance issue (M= 3.02, SD= 1.03). For the Swiss case, the risk management issue with the lowest-judged personal importance was medical use of X-rays and radiation therapy (M= 1.95, SD= 0.88); the issue judged to be of highest personal importance was global climate change (M= 3.75, SD= 0.84). Mean personal importance for the 12 issues was 2.70. We selected a single, high importance issue: development, production, and sale of genetically engineered food (M= 3.22, SD= 0.88). This issue had the advantage over others of being of highest current interest to potential study participants. Instead of basing the personal importance between-participants factor on separate issues, in the Swiss case we based it on the judged personal importance of a single issue, dividing participants equally into high and low groups.
7.3. Summary of Pretests
The pretests in the United States and Switzerland served two functions, the development of scales to measure procedural fairness and personal issue importance, and the construction of two between-participant factors for use in our survey study—procedural fairness (fair/unfair) and personal issue importance (high/low). The third between-participant factor, policy outcome (risk averse/risk accepting), required no formal pretesting since interviews with potential study participants indicated that its clear, straightforward meaning—whether a policy called for increased or decreased controls—was easily grasped. Given these scales and experimental factors, we were ready to proceed with the study.
8. THE AMERICAN CASE
8.1. Method
8.1.1. Participants
Our mail-survey sample was a randomly drawn list of 1,500 addresses in Washington State purchased from Survey Sampling International (Fairfield, Connecticut). Addresses were instructed to have the person over 18 years of age with the next birthday complete the survey. After reminder letters, a total of 525 usable questionnaires were returned, for a response rate of 35%. Sixty-two percent of the study participants were male (N= 323) and 36% were female (N= 190). Twelve persons did not report gender. The mean age was 55.60 (SD= 16.39). Nineteen persons did not report age. The survey was conducted during the fall months of 2004.
8.1.2. Materials
Questionnaires were presented to participants in the form of a printed booklet accompanied by a cover letter that explained the background and purpose of the study. Each booklet began with a risk management scenario (described in the Appendix), which was followed by a series of questionnaire items.
8.1.3. Measures
The questionnaire consisted of items that measured, first, two of the experimental factors—the personal importance of the issue and judged procedural fairness. These measures were developed in pretests for this study. Sets of items then measured five factors from the TCC model—value similarity, trust, past performance, confidence, and cooperation. These measures were developed in previous studies. Since three of these factors, value similarity, past performance, and confidence, are not directly relevant to this study, they will not be discussed further.
8.1.3.1. Judged personal importance. Participants used a five-point scale anchored by “agree entirely” and “disagree entirely” to indicate their agreement with each of the following five statements: “This issue is a central concern of mine,”“This issue threatens values that are very important to me,”“My attitude toward this issue is a matter of personal conscience,”“This issue strikes to the heart of who I am,” and “This is a fundamentally important issue to me.”
8.1.3.2. Judged procedural fairness. Four items, each with five-point response scales, were used: “Was the process unbiased or biased?” (Completely unbiased/completely biased), “Was the process balanced or slanted?” (Completely balanced/completely slanted), “Was the process even-handed or one-sided?” (Completely even-handed/completely one-sided), and “Was the process fair or unfair? (Completely fair/completely unfair).
The remaining two factors—trust and cooperation—used the same five-point response scale anchored by “agree entirely” and “disagree entirely.” Trust was measured by four items, cooperation by three.
8.1.3.3. Trust. “The decision makers are too busy looking out for selfish interests to be helpful in dealing with this issue” (reversed scored), “I couldn't trust the decision makers to manage this issue” (reverse scored), “In working on this issue, the decision makers can be counted on to do the right thing,” and “In working on this issue, the decision makers will make a good-faith effort to treat everyone even-handedly.”
8.1.3.4. Cooperation. “I would support the efforts of the decision makers on this issue in any way I can,”“If we are ever going to make progress in dealing with this issue, it will be through the efforts of the decision makers and others like them. I support their efforts,” and “It's about time that some people who know what they're doing—such as the decision makers and others like them—tried to do something constructive about this issue. I'm with them all the way!”
These materials generated our 2 (issue importance: high/low) × 2 (procedures: fair/unfair) × 2 (policy outcome: risk averse/risk accepting) experimental design, and the measures provided the means for testing the predictions of the standard account of procedural fairness against the predictions of the TCC model. The TCC model predicts an interaction between issue importance and policy outcome, but the standard model does not. For the TCC model (but not the standard model) policy outcome should have greater effects on all the dependent variables, including judged procedural fairness, when issue importance is high than when it is low. Also, whereas the standard model predicts that procedural fairness will have significant positive effects on trust and cooperation, the TCC model predicts that procedural fairness will affect only judged procedural fairness (with possible effects on other dependent variables, particularly trust, in the low issue importance condition). After considering the reliability of measures and a manipulation check, the following results address these hypotheses, first across all conditions and then for the high- and low-issue importance conditions separately.
8.2. Results
8.2.1. Reliability of Measures
All of the measures had satisfactory levels of reliability: personal importance (α= 0.88), procedural fairness (α= 0.95), trust (α= 0.94), cooperation (α= 0.94).
8.2.2. Manipulation Check
8.2.2.1. Issue importance. Our experimental design required that study participants judge the high importance issue (resource depletion) to be of higher personal importance than the low importance issue (medical radiation). Results confirm that this was the case: resource depletion was judged to be more important (M= 3.62, SD= 1.01) than medical radiation (M= 2.54, SD= 0.92), t(518) = 12.81, p < 0.001 (one-tailed).
8.2.3. Analyses of Variance
Means and standard deviations for all dependent variables are shown in Table I. A series of 2 (issue importance: high/low) × 2 (procedures: fair/unfair) × 2 (policy outcome: risk averse/risk accepting) between-participants ANOVA showed strong effects of issue importance on all the dependent variables: judged fairness, F(1, 517) = 70.02, p < 0.001, η2= 0.12; trust, F(1, 517) = 130.63, p < 0.001, η2= 0.20; and cooperation, F(1, 517) = 75.95, p < 0.001, η2= 0.13.6 Policy outcome had modest effects on all the dependent variables with the exception of judged fairness: judged fairness, F(1, 517) = 1.67, n.s.; trust, F(1, 517) = 7.25, p < 0.01, η2= 0.01; cooperation, F(1, 517) = 5.67, p < 0.05, η2= 0.01. The effects of both issue importance and policy outcome were qualified by the predicted interaction between the two factors for all dependent variables: judged fairness, F(1, 517) = 6.90, p < 0.01, η2= 0.01; trust, F(1, 517) = 18.10, p < 0.001, η2= 0.03; cooperation, F(1, 517) = 25.58, p < 0.001, η2= 0.05. Also as predicted, the procedural fairness factor affected only judged fairness, though weakly: judged fairness, F(1, 517) = 4.66, p < 0.05, η2= 0.01; trust, F(1, 517) = 2.74, n.s.; cooperation, F(1, 517) = 0.42, n.s. The analyses revealed no other significant effects. For clarity of presentation, the interaction between issue importance and policy outcome, as well as the effects of procedural fairness, are explored further in the separate results for the high- and low-issue importance conditions.
| Dependent Variable | Issue Importance | Procedural Fairness | Policy Outcome | Mean | Standard Deviation | N |
|---|---|---|---|---|---|---|
| Judged fairness | Low | Low | Averse | 3.62 | 0.89 | 68 |
| Accepting | 3.75 | 0.95 | 63 | |||
| High | Averse | 3.83 | 0.92 | 49 | ||
| Accepting | 3.92 | 0.80 | 83 | |||
| High | Low | Averse | 3.17 | 0.97 | 57 | |
| Accepting | 2.86 | 1.00 | 66 | |||
| High | Averse | 3.34 | 1.09 | 67 | ||
| Accepting | 3.01 | 0.81 | 72 | |||
| Trust | Low | Low | Averse | 3.70 | 0.98 | 68 |
| Accepting | 3.75 | 0.91 | 63 | |||
| High | Averse | 3.82 | 0.81 | 49 | ||
| Accepting | 4.03 | 0.83 | 83 | |||
| High | Low | Averse | 3.14 | 1.05 | 57 | |
| Accepting | 2.58 | 0.99 | 66 | |||
| High | Averse | 3.22 | 1.09 | 67 | ||
| Accepting | 2.64 | 0.72 | 72 | |||
| Cooperation | Low | Low | Averse | 3.59 | 0.87 | 68 |
| Accepting | 3.80 | 0.96 | 63 | |||
| High | Averse | 3.62 | 0.92 | 49 | ||
| Accepting | 3.86 | 0.90 | 83 | |||
| High | Low | Averse | 3.27 | 1.13 | 57 | |
| Accepting | 2.61 | 1.00 | 66 | |||
| High | Averse | 3.31 | 1.12 | 67 | ||
| Accepting | 2.71 | 0.82 | 72 |
Under conditions of high issue importance, the TCC model (but not the standard model) predicts that policy outcome should significantly affect judged fairness as well as the remaining dependent variables. Analyses of variance confirmed this prediction: judged fairness, F(1, 254) = 9.47, p < 0.01, η2= 0.04; trust, F(1, 254) = 29.54, p < 0.001, η2= 0.10; cooperation, F(1, 254) = 39.78, p < 0.001, η2= 0.14. Means for all dependent variables were higher for the risk-averse outcome. Also as predicted, procedural fairness affected none of the dependent variables, including judged fairness: judged fairness, F(1, 254) = 0.74, n.s.; trust, F(1, 254) = 0.00, n.s.; cooperation, F(1, 254) = 0.00, n.s. Under conditions of low issue importance, both the TCC model and the standard model predict that policy outcome should have no significant effects on judged fairness or any of the remaining dependent variables. With the exception of cooperation, analyses of variance confirmed this prediction: judged fairness, F(1, 255) = 1.17, n.s.; trust, F(1, 255) = 1.71, n.s.; cooperation, F(1, 255) = 4.82, p < 0.05, η2= 0.02. For participants in the low issue importance condition (medical radiation), the risk-accepting policy outcome evoked slightly more cooperation than the risk-averse outcome. Partially as predicted, procedural fairness affected (though only marginally) judged fairness as well as trust, but not cooperation: judged fairness, F(1, 255) = 3.15, p < 0.10, η2= 0.01; trust, F(1, 255) = 3.91, p < 0.05, η2= 0.02; cooperation, F(1, 255) = 0.34, n.s. High procedural fairness produced slightly higher judgments of fairness and trust than low procedural fairness. The analyses revealed no other significant effects.
8.3. Conclusion
Data analyses showed a strong interaction between issue importance and policy outcome. Under conditions of high issue importance (resource depletion), policy outcome affected judged fairness, trust, and cooperation, with the risk-averse policy producing higher judgments. Under conditions of low issue importance (medical radiation), policy outcome had no effect on judged fairness or trust but did have a moderate impact on cooperation. Analyses also showed that issue importance moderated the effects of procedural fairness. When issue importance was high, procedural fairness had no effects. When issue importance was low, procedural fairness had modest effects on judged fairness and trust. These results demonstrate the pivotal role that issue importance can play in determining the effects of both policy outcome and procedural fairness on judgments of fairness, trust, and cooperation. In all, the results of the American case support the account of the relation between trust and fairness provided by the TCC model as opposed to that of the standard model.
9. THE SWISS CASE
9.1. Method
9.1.1. Participants
Our sample was a randomly drawn list of addresses in the telephone directory for the German-speaking part of Switzerland. Addresses were instructed to have the person over 18 years of age with the next birthday complete the survey. After reminder letters, a total of 759 usable questionnaires were returned, for a response rate of 32.2%. Sixty percent of the study participants were male (N= 442) and 40% were female (N= 297). Twenty persons did not report gender. The mean age was 49.73 (SD= 15.88). Nineteen persons did not report age. The survey was conducted during the fall months of 2004.
9.1.2. Materials
The questionnaires for the Swiss case were modeled after those used in the American case. The primary difference was that the Swiss case employed a single risk management issue and a median split on judged personal importance of the issue to create two groups of respondents. The issue, determined by pretesting to be of high personal importance in the German-speaking part of Switzerland, was the development, production, and sale of genetically engineered food. Otherwise, the materials used in the Swiss case were the same as those used in the American case, translated into German.
9.1.3. Measures
The measures were the same as in the American case, translated into German.
9.2. Results
9.2.1. Reliability of Measures
All of the measures had satisfactory levels of reliability: personal importance (α= 0.91), procedural fairness (α= 0.84), trust (α= 0.77), cooperation (α= 0.89).
9.2.2. Issue Importance Groups
A median split on judged issue importance was used to create two issue importance groups, those for whom genetically engineered food was a highly important issue and those for whom this issue was of lesser importance.
9.2.3. Analyses of Variance
Means and standard deviations for all dependent variables are shown in Table II. A series of 2 (judged issue importance: high/low) × 2 (procedures: fair/unfair) × 2 (policy outcome: risk averse/risk accepting) between-participants ANOVA showed moderate effects of judged issue importance on trust and cooperation, but not on judged fairness: judged fairness, F(1, 685) = 1.68, n.s.; trust, F(1, 685) = 19.30, p < 0.001, η2= 0.03; cooperation, F(1, 685) = 8.80, p < 0.01, η2= 0.01. Similarly, policy outcome affected trust and cooperation, but not judged fairness: judged fairness, F(1, 685) = 2.24, n.s.; trust, F(1, 685) = 12.04, p < 0.01, η2= 0.02; cooperation, F(1, 685) = 12.40, p < 0.001, η2= 0.02. The effects of both judged issue importance and policy outcome were qualified by the predicted interaction between the two factors for all dependent variables: judged fairness, F(1, 685) = 23.23, p < 0.001, η2= 0.03; trust, F(1, 685) = 28.38, p < 0.001, η2= 0.04; cooperation, F(1, 685) = 24.27, p < 0.001, η2= 0.04. Further analyses showed that, when judged issue importance was high, means for all dependent variables were higher for the risk-averse outcome; when judged issue importance was low all means were higher for the risk-accepting outcome. Also as predicted, the procedural fairness factor affected only judged fairness: judged fairness, F(1, 685) = 13.33, p < 0.001, η2= 0.02; trust, F(1, 685) = 0.02, n.s.; cooperation, F(1, 685) = 0.06, n.s. The analyses revealed no other significant effects.
| Dependent Variable | Judged Issue Importance | Procedural Fairness | Policy Outcome | Mean | Standard Deviation | N |
|---|---|---|---|---|---|---|
| Judged fairness | Low | Low | Averse | 3.16 | 0.83 | 83 |
| Accepting | 3.50 | 0.97 | 76 | |||
| High | Averse | 3.43 | 0.85 | 78 | ||
| Accepting | 3.56 | 0.84 | 75 | |||
| High | Low | Averse | 3.34 | 0.91 | 100 | |
| Accepting | 2.97 | 1.01 | 106 | |||
| High | Averse | 3.75 | 0.83 | 81 | ||
| Accepting | 3.24 | 0.94 | 94 | |||
| Trust | Low | Low | Averse | 3.20 | 0.79 | 82 |
| Accepting | 3.40 | 0.83 | 73 | |||
| High | Averse | 3.35 | 0.78 | 78 | ||
| Accepting | 3.40 | 0.84 | 73 | |||
| High | Low | Averse | 3.41 | 1.01 | 98 | |
| Accepting | 2.72 | 0.94 | 105 | |||
| High | Averse | 3.26 | 0.93 | 80 | ||
| Accepting | 2.75 | 0.92 | 95 | |||
| Cooperation | Low | Low | Averse | 3.18 | 0.96 | 82 |
| Accepting | 3.29 | 1.03 | 72 | |||
| High | Averse | 3.29 | 0.91 | 76 | ||
| Accepting | 3.41 | 0.96 | 74 | |||
| High | Low | Averse | 3.51 | 1.06 | 99 | |
| Accepting | 2.65 | 1.26 | 106 | |||
| High | Averse | 3.28 | 1.06 | 80 | ||
| Accepting | 2.74 | 1.19 | 94 |
9.3. Conclusion
The results of the Swiss case replicated the two main findings of the American case: (1) issue importance interacted with policy outcome to produce significant effects for all the dependent variables, and (2) procedural fairness affected judged fairness but neither trust nor cooperation. As in the American case, these results demonstrate the central significance of issue importance in determining the effects of policy outcome and procedural fairness on judgments of fairness, trust, and cooperation, thereby lending further support for the TCC model's account of the relation between trust and fairness.
10. DISCUSSION
In this study, we tested the implications of two alternative models of the relation between trust and fairness. One alternative, which we call the standard model, holds that procedural fairness will determine judgments of fairness, trust, and cooperation. The other alternative, the TCC model, claims that the relation between trust and fairness will be moderated by the personal moral importance of the issue being considered: under conditions of high issue importance, procedural fairness should have no effect on judgments of fairness, trust, and cooperation; under conditions of low issue importance, procedural fairness could significantly affect judgments of fairness, trust, and cooperation. The results of two parallel surveys, one conducted in Washington State and the other in Switzerland, strongly supported the TCC model over the standard model by demonstrating, within the context of risk management issues, that issue importance is a critical factor moderating the relation between trust and fairness.
Our findings regarding judgments of trust and fairness in risk management contexts parallel those of researchers who have recently investigated the effects of personal moral importance on other types of judgments in a variety of contexts. Skitka and colleagues have shown—in studies dealing with criminal justice (Skitka & Houston, 2001), Supreme Court decision making (Skitka, 2002), child custody disputes (Skitka & Mullen, 2002), and abortion (Mullen & Skitka, 2006)—that outcome preferences will determine fairness judgments under conditions of high moral importance but have little effect when moral importance is low. Similarly, Kahan and colleagues (Braman et al., 2005; Kahan & Braman, 2003a, 2003b), addressing the gun control issue, showed that the presentation of empirical evidence, a form of fair procedure that is often persuasive under conditions of low issue importance, loses it force when the issue is of high moral importance to disputants. Our results, as well as those of these studies, call into question the standard model of trust and fairness (Tyler, 2000), which does not distinguish between contexts of low- and high-issue importance. We found that fair procedures will be judged to be fair and lead to judgments of trust and cooperation only when fairness is the dominant value in play. The value of fairness, however, can be, and often is, trumped by the value of outcome alternatives (decisions, policies, actions, etc.). Our results suggest that the efficacy of fair procedures is strictly limited.
10.1. Limitations
Our study is limited in several respects. Only one of a number of ways of defining fair and unfair procedures was explored, that is, on the basis of the inclusion or exclusion of broad public participation. Additional manifestations of fairness/unfairness should be investigated. The effect sizes found in our study were in some cases modest. Different instantiations of the experimental factors may lead to more robust effects. In addition, although we examined three diverse risk management issues, future studies should broaden this range further. Finally, the relations between the level of importance attributed to an issue and the specific emotions evoked by that issue in different individuals is a critical question not addressed in this study.
10.1.1. Trust, Fairness, and Affect
What is it about the personal importance of an issue that makes fair procedures irrelevant? It is tempting to think that judged issue importance and affect are strongly related: an issue that is judged to be personally important (resource depletion, for example) elicits strong positive or negative affect depending on the consistency/inconsistency of values. People who think an issue is important think that something should be done about it; they are affectively involved. In our study of issue importance, however, we identified a number of respondents who (on our measure of personal importance) indicated that the issue was not important to them but who nonetheless were clearly (based on their responses to TCC model items) affectively involved. People who think an issue is not important think that no action should be taken. Some of these people are affectively involved and some are not. Those who are involved are strongly opposed to taking action (e.g., protecting endangered species). Affective involvement may be a better indicator of issue importance than the personal importance scale that we used in our study. Indeed, Skitka and colleagues have recently demonstrated that feelings of anger play a central role in the moral mandate effect (Mullen & Skitka, 2006). It seems likely that specific emotions also mediate the relations between trust and fairness under conditions of high moral importance.
In a recent study of the role of incidental affect in judgments of fairness (i.e., affect not integral to the object being judged), van den Bos and colleagues (van den Bos et al., 2003; see also van den Bos, 2003) observed that there had been few studies of the effects of affect on fairness, and he attributed that lack of attention to the failure of researchers to recognize information uncertainty as a crucial moderating factor. In conditions of information uncertainty (i.e., no information on procedures), people would use incidental affect in making judgments of fairness, according to van den Bos; in information-certain conditions, incidental affect would not be used, and fairness judgments would be based on the cognitive aspects of the procedural information. Although the results of van den Bos's experiments supported his hypothesis, his approach—due to the issue discussed below—is unlikely to prove useful to understanding emotions and fairness in risk management contexts. The issue centers on the crucial distinction between incidental and integral affect. Most studies of the effects of affect on risk and fairness judgments focus only on incidental affect, feelings that are not associated with the object of the risk and fairness judgments. (One study that claims to deal with the integral affect associated with biotechnology—Townsend et al., 2004—appears to address it only on a semantic level.) Lerner and Keltner (2001), for example, studied the effects of dispositional or primed emotions on risk judgments. Importantly, they found that an incidental emotion had strong effects on risk judgments only when the object of those judgments was ambiguous with respect to the dimension of the emotion (i.e., the cognitive appraisal) that was primed. In other words, incidental emotions have strong effects when there are no competing sources of affect. That is the case in van den Bos's information-uncertain condition. In contrast to van den Bos, our study of fairness judgments did not isolate fair procedures from outcomes that are emotionally significant. Whereas van den Bos (2003, studies 2 & 3) provided his study participants with no outcome information, respondents to our survey received full descriptions of both policy outcomes and procedures, as would be the case in naturally occurring conditions. In studies of emotions and fairness, outcomes and the affect integral to them should be judged in concert with the procedures that produced them. Although we did not measure or manipulate affect in our study, we did provide respondents with full access to material that may be affectively involving to them. In that situation, we found that issue importance—which should be strongly related to integral affect—was the key moderating factor. Thus, we would predict that, given strong integral affect, trust, based on outcome preferences, would dominate fairness judgments even when full fairness information is provided.
10.2. Practical Significance
The findings of this study indicate that fair procedures may be of limited value in resolving disputes involving risk management issues that are of high personal moral importance. Related research on the morally charged issue of gun control suggests that arguments based on empirical evidence would also be ineffective in such situations, that “culture is prior to facts” (Braman, et al., 2005, p. 285; see also Kahan & Braman, 2003). Kahan and colleagues argue that the legitimacy of information is judged relative to whether it supports or opposes one's position on the issue. When an issue is not considered to be morally important, then the norm valuing technical expertise will determine judged legitimacy of evidence.
Contrary to these cautionary conclusions, there has been an increased demand for fair procedures and, in particular, public participation in environmental decision making (Yosie & Herbst, 1998). At the same time, others have called for greater reliance on empirical evidence (Baron, 1998, 2000; Baron et al., 2006; Bazerman et al., 2001; Sunstein, 2005). Decisions, it is claimed, should be made rationally, on the basis of the numbers, not on false intuitions. Emotions should play no part in these decisions. As an antidote to affect, Baron (2006) calls for more education in economics.
It is likely, however, that, in the future as in the past, many risk management issues will evoke conflicting emotional responses from different segments of the public and be considered to be of high moral importance. If, in such an environment, disputes cannot be resolved by fair procedures and empirical evidence, what can be done? Traditional risk communication, which depends on judicious combinations of public involvement and technical analysis, evidently cannot be relied on (Charnley, 2000; Gregory, 2002; Gregory & McDaniels, 2005; National Research Council, 2005). Our study (and the work of others, e.g., Perelman, 1963; Tyler, 2000) suggests that the fundamental problem derives from the within-group nature of both trust and fairness: we tend to form groups based on what we have in common (however sensible or arbitrary that might be), and we tend to trust and treat fairly members of our groups, while we distrust and treat unfairly persons who are nonmembers (Hogg, 2007). If this analysis is correct, then a possible pathway to the resolution of emotionally charged risk management conflicts is clear: the competing sides should somehow be induced to redefine themselves as being members of a single, inclusive group (Brewer, 2000; Hogg, 2007; Tyler, 2000). Such a group would provide the basis for judgments of trustworthiness, fairness, and trust-based legitimacy, all of which are needed for reasoned argument to proceed under conditions of high moral importance. Group formation, based on morally significant commonalities, must precede deliberation.
How to do this? As Brewer (2000) and Hogg (2007) point out, members of individual groups tend to resist the formation of superordinate groups that threaten the distinctiveness of their group identities. What is needed, then, is an inclusive group that preserves the particularity of its constituent subgroups. Take the classic example of nuclear power. In Sunstein's (2002) terms, many technocrats feel positively about it, while many populists are concerned. Sunstein raises the following question: “If people are now alarmed about a low-probability hazard, is there anything that government can do to provide assurance and to dampen concern? This is an unanswered question … There appears to be no evidence that any particular strategy will succeed” (2002b, p. 95). Our study suggests that a technique we call “moral reframing,” which is similar to Tetlock's (2003) idea of rhetorical reframing, might be helpful. The focal issue, which evokes negative moral intuitions in some, is reframed within a larger narrative concerning an issue with greater moral implications than the focal issue. In that narrative, the focal issue plays a positive role to reduce concern about the larger issue. In short, moral reframing can effect a change of valence on the focal issue. In the case of nuclear power, one reframing issue might be global climate change (Leiserowitz, 2006; Lorenzoni & Pidgeon, 2006; Lorenzoni et al., 2005; Moser & Dilling, 2004). This is a reframing narrative suggested by prominent environmentalists and by scientists studying both global warming and nuclear power. Thus, erstwhile adversaries might join forces to work toward a common goal while maintaining their separate group identities.
As a dispute resolution technique in contexts of high moral importance, moral reframing is untested and may prove to be ineffective. The key practical implication of this study, however, is that any approach to cooperation between conflicting sides to morally charged risk management issues should begin, not by implementing fair procedures or marshalling factual arguments, but by establishing trust based on morally important, commonly held values.
Footnotes
ACKNOWLEDGMENT
This study was supported by National Science Foundation Grant SES-0323451.
Appendix
APPENDIX: RISK MANAGEMENT SCENARIOS FOR THE AMERICAN CASE
There were eight forms of the questionnaire, one for each experimental condition. All forms began with the same introductory page, which included a description of the structure of the questionnaire. Participants were told that they would read a brief hypothetical scenario “describing a risk management issue, a decision making process, and a risk management policy,” and that the scenario would be followed by a series of questions asking about their thoughts on each element of the scenario. The first experimental factor, issue importance (high/low), was manipulated in the description of the risk management issue. Half the participants read a description of the resource depletion issue (high issue importance) containing these key points:
- •
Across the United States and throughout the world, numerous animals and birds are at risk. Their populations and habitat are endangered; in many cases, their numbers are dropping dangerously low.
- •
In recent years, fish populations around the world, including popular restaurant species, have declined dramatically. One reason is overfishing—fishing faster than a population can replenish itself. Other factors include pollution, habitat degradation, the damming of rivers, and the diversion of water for agricultural irrigation.
- •
Half the natural forest cover worldwide has already disappeared, 13% in the last 30 years.
- •
Whether certain natural resources are in fact endangered and, if so, what action, if any, should be taken are matters of some controversy.
The remaining participants read a description of the medical radiation issue (low issue importance) containing these points:
- •
In addition to its widespread use in X-ray diagnoses, radiation is also used to treat cancers.
- •
Ionizing radiation—similar to X-rays—can penetrate tissue and alter the part of the cell that regulates its growth and reproduction. Noncancerous cells can recover from this damage, while cancer cells cannot.
- •
There are two types of radiotherapy—delivered from outside the body by a machine, and using radioactive implants placed inside the body.
- •
Researchers are working to increase the effectiveness of radiotherapy by targeting the beam of energy more precisely, and making the cancer cells more sensitive to it.
- •
Although the treatment itself is painless at the time, the cumulative effect of many sessions does produce side effects.
- •
The degree to which medical radiation should be used in diagnoses and therapy—and which safeguards should be employed—are matters of some controversy.
The second experimental factor, procedural fairness (fair/unfair), was manipulated in the description of the decision-making process. Half the participants read a description of a fair process (as determined in the pretests), which contained these key points:
- •
There were two overriding goals to the process:
- ○
technical soundness and
- ○
full participation by concerned citizens.
- ○
- •
Decisions were made by public officials who were guided by the views of prominent experts on resource depletion (medical radiation).
- •
In addition, an open public participation process solicited the opinions of the general public.
- •
Although time consuming and costly, the public process was designed to include the views of broadly diverse groups of citizens.
The remaining participants read a description of an unfair process, with these key points:
- •
There were two overriding goals to the process:
- ○
technical soundness and
- ○
efficiency.
- ○
- •
Decisions were made by public officials who were guided by the views of prominent experts on resource depletion (medical radiation).
- •
This process assured that decisions were based on the best information available, made in a timely manner, and at reasonable cost.
The third experimental factor, policy outcome (risk averse/risk accepting), was manipulated in the description of the management policy. For the resource depletion issue, half the participants read a description of a risk-averse policy, which contained these key points:
- •
Our natural heritage is vital to our future, and worldwide legal protections to defend threatened animals and birds must be strengthened and enforced. To halt the habitat destruction that has caused so many populations to plummet, unwise logging and development must be controlled.
- •
The root causes of this destruction, such as inefficient wood use and sprawling home and industrial building, must also be targeted. And consumers must be encouraged to demand environmentally sensitive products.
- •
We must work to restore fish and their habitat by strengthening laws and treaties and ensuring that they're applied effectively. And we must press for sustainable ocean fishing practices.
The remaining resource depletion participants read a description of a risk-accepting policy, with these key points:
- •
There is now no clear case for any expensive species-safeguarding policy without further evidence. The existing data on the observed rates of species extinction are inconsistent with the doomsters' claims of rapid disappearance, and they do not support the various extensive and expensive programs they call for.
- •
Furthermore, recent scientific and technical advances—especially seed banks and genetic engineering—have diminished the importance of maintaining species in their natural habitat.
- •
We should not ignore possible dangers to species. Species constitute a valuable endowment, and we should guard their survival just as we guard our other physical and social assets. But we should strive for a clear and unbiased view of the gains and losses to help judge how much time and money to spend guarding our biological assets.
For the medical radiation issue, half the participants read a description of a risk-averse policy, which contained these key points:
- •
Radiation can produce a sunburn-like effect on the skin as it passes through. The extent of this depends on the number and intensity of treatments. There can be hair loss in the area being treated—which is usually temporary. The treatment can also leave the patient feeling fatigued and generally lethargic.
- •
Ionizing radiation produces changes within the genetic structure of the body's cells, and there is a small risk that an increased radiation dose can lead to cancer-causing changes in healthy cells.
- •
Medical exposure to radiation must show a sufficient net benefit compared with the individual detriment that the exposure might cause, taking into account the benefits and risks of available alternative techniques.
- •
All doses due to medical exposure must be kept as low as reasonably achievable and must be consistent with the purpose of the exposure.
- •
Wherever possible, ultrasound or MRI, which involve no hazardous radiation, should be used.
The remaining medical radiation participants read a description of a risk-accepting policy, with these key points:
- •
The radiation doses from X-ray examinations or isotope scans are small in relation to those we receive from natural background radiation, ranging from the equivalent of a few days worth to a few years.
- •
The risks of medical radiation exposure are miniscule when compared to the risks to the patient's health of not having the treatment.
- •
When too little radiation is used for diagnosis or therapy there is an increase in risk due to other factors. Too low an amount of radiation in diagnosis will result in an image that does not have enough information to make a diagnosis. In therapy, not delivering enough radiation will result in increased mortality because the cancer being treated will not be cured.
- •
The aim of managing radiation exposure should be to minimize the risk without sacrificing, or unduly limiting, the obvious benefits in the prevention, diagnosis and also in effective cure of diseases.
All eight risk management scenarios were followed by three questions that tested participants' comprehension of, in turn, the issue, the process, and the policy. Participants who failed any of these tests were eliminated from the study.
