Support Theory and the Above-Average Effect

ASSESSING THE BALANCE OF EVIDENCE: SUPPORT THEORY AND THE “ABOVE-AVERAGE EFFECT”

The intuition guiding support theory’s representation of subjective probability in terms of the balance of evidence between two hypotheses may be applied to issues beyond likelihood judgment. To illustrate, we consider as a case study the well-documented above-average effect, in which people tend to rate themselves more positively than others with respect to personality traits, skills, or susceptibility to risks. Most people, for example, consider themselves to be happier with life (Klar & Giladi, 1999), better drivers (Svenson, 1981), and less likely to suffer a heart attack (Weinstein, 1980) than the average person.

Support theory is well suited for representing comparative judgments in which the judge evaluates his or her own standing relative to a group. One natural way to represent such judgments using support theory is to treat the focal hypothesis as concerning the self, and the alternative hypothesis as concerning others in the group. Increasing the support for the focal hypothesis produces judgments favoring the self over others, and increasing the support for the alternative hypothesis produces judgments favoring others over the self.

Support theory’s emphasis on description-dependence suggests a consideration of how the focal and alternative hypotheses are represented by the judge. Based on the circumstances of the task or the framing of the query, there is substantial room for variability in how the hypotheses are represented. When asked, for example, to rate your driving skills relative to other people of your age, the focal hypothesis may take the form of something like, “I am a good driver.” The alternative hypothesis may assume a number of forms – perhaps “The average person of my age is a good driver,” “All people of my age are good drivers,” or “My spouse (or some other salient person) is a good driver.” Different representations of the alternative hypothesis call to mind different evidence, accrue different amounts of support, and yield different judgments.

Indeed, the selection of the alternative hypothesis has been shown to influence the magnitude of the above-average effect. The effect is larger when people evaluate themselves relative to a generic “average peer” than when they evaluate themselves relative to a specific peer (Klar et al., 1996; Perloff & Fetzer, 1986), even when they have little or no additional information about the specific peer (Alicke et al., 1995). A natural interpretation from the support theory perspective is that the alternative hypothesis recruits greater evidential support when it is represented as a concrete, specific individual than when it is represented as an abstract, conceptual entity.

Based on this view, any specific individual ought to be at an advantage when pitted against the generic average peer in terms of recruiting support. Consistent with this, specific others from one’s group (and not just oneself) are systematically rated above the group average on positive traits (Klar & Giladi, 1997). This presupposes a positive attitude toward the group as a whole, which is then conveyed to any individual belonging to the group. Systematic “below average” effects are also observed – in which an individual belonging to a disliked group is rated below the group average on positive traits (Klar & Giladi, 1997). According to the support theory account, given a generally positive impression of the group and its members, one can more readily recruit support favoring individual group members than the group as a whole. Likewise, given a negative impression, support favoring a negative evaluation is more readily recruited regarding an individual than regarding the group as a whole.

Klar and Giladi (1997, 1999) offer a similar interpretation using somewhat different terms. They posit a tendency to base judgments pitting a “singular” target against a “distributional” comparison group almost exclusively on the features of the singular target. Consistent with this interpretation, Klar, and Giladi (1997) found that providing individuating information about the comparison group or generic average peer reduced the magnitude of the above-average effect.

Klar and Giladi (1997) argue that these findings reflect the “noncomparative nature of comparative judgments.” Thus, people claim to be happier than others because, in absolute terms, they are happy with their lives in general (Klar & Giladi, 1999; cf. Fox & Kahneman, 1992). Along these lines, Kruger (1999) offered an anchoring-and-adjustment account of the above-average effect in which the absolute judgment serves as an anchor subject to subsequent (typically insufficient) adjustment, with the result that people tend to rate themselves above average on traits or skills for which their absolute standing is high. Most drivers, for example, are highly skilled in an absolute sense due to years of experience behind the wheel. An assessment of absolute level of skill may be a natural starting point for a comparative judgment, with insufficient adjustment for the skills of others producing an above-average effect. One straightforward prediction from this account is that the opposite pattern should be found when absolute standing is low – for example, in the case of playing chess or juggling. Indeed, participants in Kruger’s (1999) studies gave themselves low absolute ratings with respect to these skills, and also rated themselves as systematically below average in a comparative judgment.

From the perspective of support theory, the findings of Kruger (1999) and Klar and Giladi (1999) imply that greater weight is placed on the focal, self-based hypothesis than on the alternative, other-based hypothesis in comparative judgments. One observation highlighted by a support theory analysis is that the self-based hypothesis typically occupies the focal position in the comparative judgment, making it difficult to determine if the resulting above-average effect reflects a tendency to overweight the self-based hypothesis specifically, or whichever hypothesis happens to be in the focal position more generally. The literature provides some evidence for both accounts. Klar and Giladi (1997) report that the tendency to rate a specific peer as above the group average on positive traits persists even when the direction of the comparison is reversed, such that the participant is asked to evaluate whether the generic “average peer” falls above or below the specific peer group member. The researchers, however, did not directly compare the magnitude of the effect with that found using the usual direction of comparison, which leaves the possibility that hypothesis position also plays a role. Consistent with this possibility, Hoorens (1995) reports that a larger above-average effect is found for positive attributes when people rate themselves relative to the average peer than when they rate the average peer relative to themselves. The evidence to date, then, suggests that the above-average effect reflects both position dependence and a tendency to place greater weight on an individuated target (typically the self) than on the comparison group.

These two influences require different explanations. Position dependence can be seen as an attentional effect; the focal hypothesis is considered more extensively, giving it greater weight in the final judgment. Indeed, the accounts of the above-average effect offered by Kruger (1999) and by Klar and Giladi (1999) are quite similar to Macchi et al.’s (1999) account of position dependence in probability judgment. The notion of position dependence in the support theory framework may help to clarify questions about an unresolved issue in the study of the above-average effect, which concerns the role of the alternative hypothesis. Do people give judgments that merely underweight the alternative hypothesis, or do they entirely fail to treat the judgment as comparative at all, instead reporting an absolute judgment based exclusively on an evaluation of the focal hypothesis? In the terminology of support theory, this is analogous to asking whether participants are reporting a probability judgment (pitting focal against alternative hypothesis) or an assessment of the support for the focal hypothesis alone.

One method of addressing this question is to separately elicit absolute assessments of the self and of the comparison group, then predict the comparative judgment from the two separate assessments (e.g., Klar & Giladi, 1997, 1999; Kruger, 1999). Such studies have consistently found greater weight is placed on the self-based hypothesis (in the focal position) than on the other-based hypothesis (in the alternative position). Results have been mixed with respect to the precise role of the alternative hypothesis, with some studies (Kruger, 1999) reporting a negative regression coefficient (indicating that the greater the support for the alternative hypothesis, the lower the comparative judgment) and others (Klar & Giladi, 1997, 1999) reporting coefficients near zero. It is worth noting that regression coefficients of zero do not necessarily imply a completely noncomparative process, but are also consistent with the possibility that support for the alternative hypothesis has little variance.

If the judgment is truly comparative, then manipulations that facilitate recruitment of support for the alternative hypothesis ought to have some influence. Several findings are consistent with this claim, including Alicke et al.’s (1995) observation that providing information about the comparison individual decreases the size of the above-average effect. Weinstein (1980), Weinstein and Lachendro (1982), and Rothman, Klein, and Weinstein (1996) attribute people’s tendency to underestimate their vulnerability to negative events in part to overestimation of the risks faced by others. They show, for example, that provision of accurate information regarding others’ risks can reduce people’s biased sense of invulnerability (Rothman et al., 1996). Even when no new information is provided, manipulations that prompt greater consideration of others’ circumstances were sufficient to reduce the bias (Weinstein & Lachendro, 1982). These observations suggest that judgments in studies of the above-average effect may be comparative in a manner that could be captured by support theory, even if the alternative hypothesis is frequently underweighted.

The other apparent source of the above-average effect is a tendency to exaggerate the evidential support for the singular target (the self) in the comparative judgment. This may reflect both informational and motivational influences, which support theory might be useful in separating more clearly. The informational influence can arise because more is generally known about oneself than about specific others, about which in turn more is known than about the group as a whole or a generic average peer. A specific other may also serve as a better retrieval cue than a group, such that evidence known to characterize the entire group may nonetheless come more readily to mind when thinking about a specific group member. In terms of support theory, such influences generally yield greater support for the self or a specific other than for the group as a whole, regardless of whether the trait being evaluated is negative or positive; in either case, the judge should more readily recruit evidence supporting the hypothesis regarding oneself than the hypothesis regarding the average peer. This claim is consistent with research demonstrating that people asked whether they are characterized by a given attribute tend to rate themselves as higher on that attribute than when they are asked whether they are characterized by an opposite attribute (Kunda, Fong, Sanitioso, & Reber, 1993), presumably because they have so much self-relevant knowledge that some evidence supporting either attribute can be retrieved. Along these lines, Downs and Shafir (1999) showed that pairs of opposing traits are not only ascribed more frequently to oneself than to others, but also are more frequently ascribed to well-known celebrities than to less well-known individuals.

Motivational influences, by contrast, should be directed such that support for positive hypotheses regarding well-liked targets (e.g., self or close friend) should be more readily recruited and support for negative hypotheses regarding those targets should be less readily recruited. There is some evidence consistent with this possibility. Perloff and Fetzer (1986) reported that college students’ self-ratings of susceptibility to risk exhibited a less pronounced illusion of invulnerability when the students compared themselves to a close friend than when they compared themselves to a less close friend or the average college student, suggesting a motivated tendency to underestimate the risks faced by friends. From the support theory perspective, such influences could arise from a tendency to exaggerate support for hypotheses one wishes to believe.

Such motivational influences ought to be limited under conditions that make it difficult to exaggerate support for the desired hypothesis. Consistent with this perspective, Dunning, Meyerowitz, and Holzberg (1989) showed that the tendency to rate oneself above average on positive attributes is substantially reduced or even eliminated when the attributes are clearly defined. Only when the attributes are relatively ambiguous do pronounced above-average effects emerge, which these researchers attribute to a tendency to interpret ambiguous traits in a manner that best fits one’s own description; thus, the analytic individual considers herself a good leader due to her planning skills, and the friendly individual considers himself a good leader due to his communication skills.

The support theory framework can accommodate this finding: Motivational factors influence comparative judgments by maximizing support for desirable hypotheses and minimizing support for undesirable hypotheses. Circumstances that constrain the kind of evidence that can be recruited and the manner in which it can be interpreted necessarily limit the extent to which one can construct a judgment favoring the desirable hypothesis. One additional finding (e.g., Klar et al., 1996; Weinstein, 1980) consistent with this approach is that above-average effects in judgments of risks are more pronounced for controllable future outcomes (e.g., heart attack, for which exercise and diet play a large role) than for uncontrollable future outcomes (e.g., leukemia). In the case of uncontrollable future outcomes, it is presumably more difficult to recruit evidence supporting one’s own relative invulnerability.