Carnegie Mellon University
Research Showcase @ CMU
Tepper School of Business
2006
Objective Standards Matter Too Much: The Use
and Abuse of Absolute and Comparative
Performance Feedback in Absolute and
Comparative Judgments and Decisions
Don A. Moore
Carnegie Mellon University, [email protected]
William M. P. Klein
University of Pittsburgh
Follow this and additional works at: http://repository.cmu.edu/tepper
Part of the Economic Policy Commons, and the Industrial Organization Commons
This Working Paper is brought to you for free and open access by Research Showcase @ CMU. It has been accepted for inclusion in Tepper School of
Business by an authorized administrator of Research Showcase @ CMU. For more information, please contact [email protected].
Absolute and Comparative Performance Feedback
Running Head: ABSOLUTE AND COMPARATIVE PERFORMANCE FEEDBACK
Objective Standards Matter Too Much: The Use and Abuse of Absolute and Comparative
Performance Feedback in Absolute and Comparative Judgments and Decisions
Don A. Moore
Carnegie Mellon University
William M. P. Klein
University of Pittsburgh
Correspondence to:
Don A. Moore
Tepper School of Business
Carnegie Mellon University
5000 Forbes Avenue
Pittsburgh, PA 15213
1
Absolute and Comparative Performance Feedback
2
Abstract
We explored the effects of absolute and comparative feedback on self-evaluations, decisions under
uncertainty, performance attribution, and perceived relevance of the task to one’s self-concept.
Participants (415 undergraduates) were told they had gotten 20% correct, 80% correct, or were not
given then their scores on a practice test. Orthogonal to this manipulation, participants learned that
their performance placed them in the 23rd percentile or 77th percentile, or they did not receive
comparative feedback. They were then given a chance to place bets on two games – one in which
they needed to get more than 50% right to double their money (absolute bet), and one in which they
needed to beat more than 50% of other test-takers (comparative bet). Absolute feedback influenced
comparative betting, particularly when no comparative feedback was available. Comparative
feedback exerted weaker and inconsistent effects on absolute bets. Similar findings emerged on
perceived likelihood of winning and confidence in the bets. Absolute and comparative feedback had
equivalent effects on performance attribution and perceived task importance (such that more
favorable performance increased ability attribution and task importance), but absolute feedback had
much stronger (and more consistent) effects on satisfaction with performance and state self-esteem.
These findings suggest that information about one’s absolute standing on a dimension may be more
influential than information about comparative standing, supporting Festinger’s (1954) assertion that
social comparison was only necessary when objective information was unavailable.
Absolute and Comparative Performance Feedback
Objective Standards Matter Too Much: The Use and Abuse of Absolute and Comparative
Performance Feedback in Absolute and Comparative Judgments and Decisions
In 1954, Leon Festinger coined the term “social comparison” and proposed that people
compare themselves with others to make sense of their own outcomes, abilities, and opinions.
Festinger also argued, however, that comparative social information was not where people
looked first to understand what they had, what they accomplished, and who they were. Instead,
when objective absolute standards were available, Festinger maintained that the use of those
objective standards would obviate the need for comparative social information. Festinger’s
theory has been one of the most long-lived and influential psychological theories. However,
studies building on Festinger’s work have generally not tested the fundamental assumption that
objective absolute information will trump comparative social information when both are
available.
The Power of Comparative Information
Research on social comparison has primarily focused on the effects of different kinds of
social comparisons (Mussweiler, 2003; Suls, Martin, & Wheeler, 2002; Suls & Wills, 1991). A
great deal of evidence emphasizes the tremendous power of social comparisons, and some of it
even suggests that Festinger may have underestimated the degree to which people attend to
social comparisons, even when objective absolute information is readily available. For example,
Blount and Bazerman (1996) report the results of an experiment in which people were offered $7
for 40 minutes’ work, and knew that others would also be paid $7. Seventy-two percent of their
respondents agreed to participate. However, when people were offered $8 but told that other
participants would get $10, participation dropped to 54%. People have a fairly good sense of
3
Absolute and Comparative Performance Feedback
what a dollar is worth. Nevertheless, a reduction in satisfaction brought about by greater wealth
when others are still wealthier has been replicated in studies in the lab (Bazerman, Schroth, Shah,
Diekmann, & Tenbrunsel, 1994; Loewenstein, Thompson, & Bazerman, 1989) and in the field
(Hagerty, 2000; Luttmer, 2005). The implications of these sorts of findings are profound. If
people decide how happy they are by comparing their prosperity with that of others, the result is
what Kahneman (1999) has called a hedonic treadmill, in which everyone can grow materially
better off without improving subjective well-being (see also Frank, 1985; Myers, 2000).
Moreover, research has demonstrated that the power of social comparisons extends well
beyond subjective well-being. For example, Klein (1997) manipulated whether participants
believed that their performance on an aesthetic judgment task was high or low and above or
below average. Participants were then asked to choose between engaging in the same judgment
task again – in which they needed to achieve a high absolute score to win a prize – or a chance
lottery in which they could also win the prize. Those participants who were told their scores
were above average were more likely to choose another trial of the task (over the chance lottery).
They appeared to have inferred that if they were better than average, their future scores would be
high (in absolute terms). They did this despite the fact that rational use of the comparative
feedback they received would dictate they do the opposite: Given uncertainty about how one will
do in the future, knowing that one’s score was far above average in the past suggests that if one’s
future performance regresses to the mean, it is likely to go down. Subjective evaluations of
performance, likewise, were only influenced by comparative feedback. These findings, like the
hedonic treadmill, suggest that comparative feedback influences judgments in contexts where it
should not, normatively speaking.
4
Absolute and Comparative Performance Feedback
Several other studies that did not explicitly manipulate comparative and absolute
feedback in a crossed design have obtained conceptually similar results. Many experiments
testing the self-evaluation maintenance (SEM) model show that people are more likely to define
themselves according to abilities on which they are better than others than for those on which
they have high absolute ability (e.g., Campbell, Fairey, & Fehr, 1986; Tesser & Campbell, 1980).
Additional work shows that social comparisons (and reactions to comparative feedback) are
largely automatic (Gilbert, Giesler, & Morris, 1995; Mussweiler, 2003), suggesting that it is
cognitively effortless to apply comparative information to inferences about absolute selfstanding. All these results suggest that comparative information exerts an inappropriately large
influence on absolute evaluations.
The Power of Absolute Information
On the other hand, other research appears to contradict the findings described above.
This evidence suggests instead that objective information about one’s own absolute standing
exerts too strong an influence on comparative judgments. For example, Moore and Kim (2003)
found that people playing a competitive game were more confident that they would win the game
if the task was easy than if it was difficult, suggesting they failed to adequately consider others,
and instead focused myopically on their own solo performances even when explicitly asked to
compare themselves with others (for similar results, see Chambers, Windschitl, & Suls, 2003;
Hoelzl & Rustichini, 2005; Kruger & Burrus, 2004; Windschitl, Kruger, & Simms, 2003). This
“solo comparison effect” describes the tendency for comparative judgments to be excessively
influenced by perceptions of absolute standing. Camerer and Lovallo (1999) argued that
budding entrepreneurs underestimate their chances of failure because they focus on their own
strengths and abilities, while failing to adequately consider the abilities of their competitors (see
5
Absolute and Comparative Performance Feedback
also Greico & Hogarth, 2004; Moore & Cain, 2005). Related work shows that people may infer
their relative levels of satisfaction and happiness from their absolute levels (Klar & Giladi,
1999).
So what are we to make of Festinger’s belief in the primacy of absolute information?
Despite a wealth of research on responses to objective absolute feedback (for a review see
Kluger & DeNisi, 1996) and social comparison (for reviews, see Buunk & Gibbons, in press;
Suls & Wheeler, 2000), very little research pits absolute and comparative feedback against each
other in a way that allows us to compare their influences. Research on the effects of comparative
feedback – as in the classic work testing aspects of social comparison theory and its descendant
theories (Festinger, 1954; Suls & Wheeler, 2000) – typically does not employ designs that
include both comparative and absolute feedback, and especially not both types of feedback in a
crossed design. Studies concerned with the effects of absolute feedback are similarly unlikely to
also manipulate comparative feedback in the same design.
Comparison of effect sizes across studies is impossible due to variation in the dependent
measures employed in these studies. Dependent measures in most social comparison studies
include psychic variables such as affect (Taylor & Lobel, 1989) and self-evaluations (Tesser,
1988), with much less attention to behavioral measures. Few studies discriminate between
behaviors that should be differentially influenced by absolute and comparative feedback. There
is no one standard set of dependent measures of either comparative or absolute evaluation; this is
a problem because evidence suggests that the effects of comparative and absolute feedback differ
substantially on slightly different measures of the same construct (Klein, 2003; Moore, 2005).
For example, Klein (1997) found that comparative feedback had a stronger effect on subjective
ratings, whereas absolute feedback had a stronger effect on behavioral measures (i.e.,
6
Absolute and Comparative Performance Feedback
participants’ willingness to bet on their performances). Furthermore, most prior studies show
effects of native beliefs about ability on judgments and decisions, rather than assessing the
effects of performance feedback.
Design and Hypotheses
The obvious missing piece of this puzzle is an experiment that manipulates comparative
and absolute feedback orthogonally. This paper is designed to fill this void. The inclusion of
no-feedback control groups also makes it possible to determine the locus of a given effect. For
example, if individuals given favorable comparative feedback behave differently than those
given unfavorable feedback, it is difficult to ascertain whether the difference is due to one or
both of these groups. No study has crossed relative and absolute feedback in a design that
includes conditions where participants include one type of feedback and not the other (or neither
type of feedback), so this important question has not been examined.
In the study reported here, we gave participants feedback about their performance on a
practice test and then asked them to evaluate their performance, place bets on how well they
would perform on the actual test, and estimate their likelihood of winning. We informed
participants that they earned a low score (20% correct) or a high score (80%) on the practice test,
or we gave them no absolute feedback at all, thereby creating a three-level manipulation of
absolute feedback. Crossed with this variable, we also informed participants that their practice
test score was much better than that of other students (77th percentile), much worse (23rd
percentile), or we gave them no comparative feedback at all. Thus, there were also three levels
of comparative feedback, completing a 3 x 3 factorial design. Importantly, participants placed
bets (and rated the likelihood of and confidence in winning those bets) on two different
outcomes. The first was an “absolute” outcome, in that the participant had to answer over 50%
7
Absolute and Comparative Performance Feedback
of the items correctly on the actual test in order to win. The second was a “comparative”
outcome; in this case, participants would win the prize if their score was higher than at least 50%
of the other participants. By including equivalent decisions that depended on absolute or
comparative ability, it was possible to more directly assess the effects of absolute and
comparative feedback on each type of decision.
Previous studies that have examined the effects of comparative feedback have often
manipulated it by providing participants with feedback about their own absolute performances as
well as the absolute performances of others. This sort of manipulation confounds absolute and
relative feedback because participants must be given information about their own performance in
order for the comparative information to have the intended effect. Instead, our manipulation of
comparative feedback varies what participants are told about what percentile rank they fall in,
thereby providing them with purely relative information that says nothing about absolute
performance.
Our first goal is to replicate the “solo comparison effect” observed by Moore and Kim
(2003). In other words, we expect that when participants get only absolute feedback, their
comparative bets will be higher when the absolute feedback suggests good performance rather
than bad performance. Thus, participants would be inappropriately using absolute information to
make predictions and decisions that should only be influenced by relative feedback. We also
expect to replicate the effects reported by Klein (1997). In particular, when people only have
comparative feedback, we expect them to use it when judging their absolute performance
(particularly their satisfaction with that performance). In keeping with Klein’s (1997) findings,
we expect to find that comparative feedback only influences comparative bets, not absolute bets.
8
Absolute and Comparative Performance Feedback
The more interesting question is how each effect is influenced by variations in the other
type of feedback and whether the effects of absolute feedback or comparative feedback tend to
differ in their magnitude. We can measure the magnitude by contrasting the effect sizes
associated with each type of feedback. To look at the differential effects of absolute and
comparative feedback, we included a wide range of variables in addition to self-evaluations and
decision-making. For example, people exhibit a self-serving bias whereby they attribute
successes to internal factors and failures to external factors (Mullen & Riordan, 1988), and
according to the self-evaluation maintenance model, people tend to ascribe self-importance to
those characteristics on which they have favorable standing (Tesser, 1988). Are they more likely
to engage in either of these processes when receiving favorable feedback of one type (absolute or
comparative)? The full factorial design of this study allows a direct examination of these
questions.
Method
Participants
Participants were 415 undergraduate students at Carnegie Mellon University who
participated in exchange for monetary payment.
Procedure
Upon arrival, participants were given an instruction sheet with the heading “Guessing
weights” and the following instructions:
“How good are you at figuring out how much other people weigh? In this exercise, you
will be shown a series of pictures of other people and your task will be to guess, within 5
pounds, how much they weigh. You will get $4 for participating in this exercise. So that
you can get used to how this game works, we will first have a practice round in which
9
Absolute and Comparative Performance Feedback
you will see 10 pictures and guess the weights of those 10 people. Then you will get
feedback about how you did on the practice test. Then you will take the real test.”
All participants then saw the 10 practice pictures and were asked to guess the weight of the
individuals in the pictures. After having taken this test, participants received feedback about
their performance. In actuality, this feedback was randomly determined and was unrelated to
participants’ actual performance.
Experimental Manipulations
There were three absolute performance conditions. One group was told that they had
answered 2 (or 20%) of the 10 items correctly (low absolute), and another group was told they
had gotten 8 (or 80%) of the items correct (high absolute). A third group did not receive any
absolute feedback (no absolute). Similarly, there were three comparative performance
conditions; participants learned that they had scored better than 23 percent (low comparative) or
77 percent of other test-takers (high comparative), or they received no comparative feedback (no
comparative). Crossing these manipulations created nine cells in a 3 x 3 between-groups design.
The order in which participants received absolute and comparative feedback was
counterbalanced.
After participants had received performance feedback on the practice trials, they were
given the option of betting any amount of their $4 earnings on their subsequent performance on
the actual test. If they bet and won, the amount of the bet would be doubled. Half of the
participants were first asked to specify an absolute bet: “To win this bet, you will need to get
more than 5 of the 10 items correct. That is, your guesses will have to fall within 5 pounds of
the person’s real weight on more than 5 of the 10 questions. You may risk up to $2 of your $4 on
10
Absolute and Comparative Performance Feedback
this bet.” Note that this score was 3 points higher than that ostensibly achieved by participants
in the low absolute condition, and 3 points lower than that in the high absolute condition.
The other half of the participants were first asked to specify a comparative bet: “To win
the bet, you will need to do better than average on the test. That is, you will need to score better
than at least 50% of the other test-takers. You may risk up to $2 of your $4 on this bet.” Again,
the criterion (50%) was the midpoint between the comparative scores given to participants in the
two feedback conditions (23% and 77%). After participants had made this first bet they were
then invited to make the other kind of bet. Those who made the comparative bet first made the
absolute bet second, and vice versa. After participants had specified both absolute and
comparative bets, they were invited to bet again:
“Now that you know that there are two types of bets, you may want to re-apportion your
money between the two types. On this page, you may revise your bets if you like. What
you write on this page will supersede the bets you have written on the previous two
pages. Please do not change what you have written on previous pages. You still have
two betting opportunities, but you are not restricted to a maximum of a $2 bet in either
one. You may choose to bet any of your $4 on the two bets. You may even bet all $4 on
one of the bets if you choose. The only restriction is that the amount of money you bet
and the amount of money you keep must add up to $4.”
After recording their bets, participants completed the remaining dependent measures.
They estimated how many of the 10 items they and the average participant in the study would
answer correctly, and guessed the percentage of other participants who would have scores lower
than theirs on the upcoming test. They then estimated the percentage likelihood that they win
each of their bets (and were encouraged to use any number between 0% and 100%). After doing
11
Absolute and Comparative Performance Feedback
so, participants completed a series of items on 5-point scales (from “not at all” [1] to “extremely”
[5]) measuring confidence in their bets; anxiety, concern, satisfaction, and disappointment about
their practice performance; interest in the task and in doing well; and beliefs about the effects of
luck and skill on their performance. Finally, participants completed a state self-esteem scale
(Heatherton & Polivy, 1991) with 13 items such as “I feel inferior to others at this moment” and
“I feel like I’m not doing well” on 5-point scales from “not at all” (1) to “extremely” (5). Finally
participants completed the actual test (which was scored by an experimenter), after which they
were paid, debriefed, thanked, and dismissed.
Results
Overview
Neither the order in which participants got feedback (either absolute or comparative first)
nor the order in which participants made their bets (either absolute or comparative first)
produced significant main or interaction effects, so all the results reported below collapse across
order. What follows is a series of 3 x 3 between-subjects ANOVAs examining each of the key
dependent variables in turn.
Bets
Time of bet (first or second bet) did not moderate the effects of absolute or comparative
feedback, and first bets were highly correlated with second bets (rs = .72 and .73 for absolute and
comparative bets, respectively). Thus, we computed the mean absolute bet (that is, the mean of
the first and the final absolute bet) and mean comparative bet (mean of the first and the final
comparative bet), and conducted analyses with these measures The bets were included in a 3
(absolute feedback: No feedback, 2 items correct, or 8 items correct) x 3 (comparative feedback:
No feedback, 23rd percentile, or 77th percentile) x (2) (type of bet: absolute or comparative)
12
Absolute and Comparative Performance Feedback
mixed analysis of variance (ANOVA). The first two factors were between-subjects
manipulations whereas the third was a within-subjects factor. Mean absolute and comparative
bets are reported in Table 1.
The ANOVA revealed a main effect of absolute feedback, F(2, 405) = 48.46, p < .0001,
such that bets were highest among those informed they scored an 8, followed by those given no
feedback, followed in turn by those learning they scored a 2. Tukey post-hoc tests showed that
all differences among the three conditions were significant (p < .05). The effect of absolute
feedback interacted with type of bet (F[2, 405] = 45.97, p < .0001); whereas absolute bets
conformed to the above stepped pattern among the three conditions (ps < .05), comparative bets
in the 2 and 8 conditions were roughly equivalent (p > .05), and were both lower than in the nofeedback condition (ps < .05). There was also a main effect of bet type on bets more generally
such that participants made higher comparative bets (M = $1.14) than absolute bets (M = $.95),
F(1, 405) = 13.01, p < .0001.
The omnibus ANOVA also yielded a main effect of comparative feedback, F(2, 405) =
12.92, p < .0001. Notably, the effect size for this effect (r = .18)1 is significantly lower than the
effect size of the absolute feedback variable (r = .33), p < .05. Tukey post-hoc tests showed that
bets were higher in the 77th percentile condition than in the 23rd percentile or no feedback
conditions (ps < .05), which did not differ from each other. Evidently, when participants did not
receive any comparative feedback, they assumed that their comparative performance was poor.
The main effect of comparative feedback was qualified by an interaction with type of bet
(absolute or comparative), F(2, 405) = 40.88, p < .0001. Whereas comparative bets conformed
to the above pattern, the only difference among absolute bets was between the 23rd and 77th
percentile conditions. Note that participants placed higher absolute bets when their comparative
13
Absolute and Comparative Performance Feedback
performance was lower. This result contradicts the findings of Klein (1997) but is consistent
with normative logic, since participants should have expected that if everyone did better than
them on the practice trial, that (if their own performance regressed to the group’s mean) they
should expect to have a higher absolute score on the actual test than they did on the practice trial.
Finally, there was an interaction between absolute and comparative feedback (F[2, 405] =
4.55, p < .001), although it was not qualified by bet type (F = 1.1, n.s.). As shown in Table 1,
when absolute feedback was high (8), there was no effect of comparative feedback (p > .10).
When absolute feedback was absent or low (2), bets were higher when comparative feedback
was high (77th percentile) than when it was absent or low (23rd percentile), ps < .05.
When participants did not receive comparative feedback, their comparative bets were
significantly higher when they were told their score was an 8 than when it was said to be a 2 (p <
.001). This suggests that, in the absence of comparative information, participants used their
absolute score as cue for comparative performance – consistent with the solo comparison effect
(Moore & Kim, 2003). When participants received no absolute feedback, comparative feedback
only affected absolute bets when absolute performance was poor or unknown. This finding is
less consistent with that of Klein (1997), who found that comparative feedback had no effect on
betting behavior when it was objectively irrelevant.
Perceived Likelihood of Winning
Because it is conceivable that participants’ bets were based on something other than their
perceived likelihood of winning, we computed the correlations between perceived likelihood and
amount of bet. These correlations were, of course, highly significant (rs = .70 and .72 for
comparative and absolute bets, respectively, ps < .0001). The likelihood judgments were then
submitted to a 3 x 3 x (2) mixed ANOVA similar to that used above for bets. Consistent with
14
Absolute and Comparative Performance Feedback
those findings, absolute feedback exerted a significant effect on perceived likelihood of winning
the absolute bet (F[2, 402] = 150.79, p < .0001); as shown in Table 1, likelihood judgments were
highest in the 8 condition, next highest in the no-feedback condition, and least highest in the 2
condition (all between-groups differences were significant by Tukey post-hoc tests, ps < .001).
Absolute feedback also influenced perceived comparative bet likelihood, F(2, 402) = 8.80, p <
.0001; in this case, participants thought they were more likely to win their comparative bets
when they scored an 8 or received no feedback than when they scored a 2 (ps < .05).
On the other hand, comparative feedback had no effect on perceived likelihood of
winning the absolute bet (F = 1.03, n.s.). As would be expected, comparative feedback affected
perceived likelihood of winning the comparative bet, F(2, 402) = 68.47, p < .0001. Participants
thought they were least likely to win in the 23rd percentile condition followed by the no-feedback
condition and then the 77th percentile condition (all between-group differences are significant, ps
< .05, by Tukey post-hoc tests). Although this mostly replicates the findings for bets, it is
interesting that bets did not differ between the 23rd percentile and no feedback conditions,
whereas perceived likelihood did. Participants told they scored in the 23rd percentile thought
they were less likely to win than those receiving no feedback, yet the magnitude of their bets was
similar.
There also emerged a significant absolute x comparative feedback interaction which was
further qualified by a 3-way interaction with type of bet (Fs > 2.60, ps < .05). Because the
interaction with bet type was weak and inconsistent with the lack of interaction in the analyses
on actual bets, we do not discuss this finding further.
Confidence in Bets
15
Absolute and Comparative Performance Feedback
The same 3 x 3 x (2) mixed ANOVA was conducted on participants’ confidence that they
would win their bets. The results of these analyses were similar to those conducted on
participants’ bets; there were significant effects of both absolute and comparative feedback
(Fs[2, 402] = 45.06 and 16.98, respectively, ps < .0001), an interaction between the two (F[4,
402] = 2.99, p < .02), and an interaction between each type of feedback and type of bet (Fs[2,
402] > 31, ps < .0001). Again, as shown in Table 1, there was a stepped effect of absolute
feedback on confidence in absolute bets, and in this case no effect of absolute feedback on
confidence in comparative bets. Comparative feedback influenced confidence in comparative
bets, but the only condition that differed from the others was the 77th percentile condition;
comparative feedback had no effect on confidence in absolute bets. There was no 3-way
interaction, so we do not discuss the weak interaction between absolute and comparative
feedback given that the effect is based on confidence collapsed across both types of bet (absolute
and comparative). Despite the difference in F values, the effect size of absolute feedback in this
analysis (r = .32) was not significantly different than that for comparative feedback (r = .20), p >
.05.
Predicted Performance
Participants also predicted their own scores and the average person’s score on the weight
test, as well as their anticipated performance relative to others (on a percentile scale). Mean
values on each variable appear in Table 2. The two score estimates were included in a 3
(absolute feedback) x 3 (comparative feedback) x (2) (target: self or others) mixed-model
ANOVA, with target as a within-groups factor. As expected, there was a substantial effect of
absolute feedback on predicted scores (F[2, 405] = 100.49, p < .0001), qualified by an
interaction with comparative feedback (F[4, 405] = 2.42, p < .05; comparative feedback F[2,
16
Absolute and Comparative Performance Feedback
405] = 2.76, p = .07). Score estimates were highest among those ostensibly scoring an 8,
followed by those among the no-feedback condition and then the 2 condition (all groups differed
by post-hoc Tukey tests, ps < .05). There was no main effect of target (F = 2.12, n.s.), yet target
interacted with comparative feedback (F[2, 405] = 5.55, p< .01). Participants reasonably
estimated that their score would be lower that that of others when they were told their practice
performance was in the 23rd percentile and higher than others when in the 77th percentile; when
given no comparative feedback, they expected their performance to be equivalent to that of
others.
Finally, there was a three-way interaction among target, absolute feedback, and
comparative feedback (F[4, 405] = 2.57, p < .05). As seen in Table 2, the nature of this
interaction was such that the comparative feedback x target interaction described above only held
in the condition where participants received no absolute feedback. When participants were told
they scored a 2, they concluded that their score would be lower than that of others if they
received no comparative feedback; otherwise, the pattern was the same as in the conditions
where no absolute feedback was received. The greatest discrepancy occurred when participants
were told they scored an 8. In this condition, participants expected both themselves and others to
score highly (and to the same degree) when getting any type of comparative feedback, and they
expected to outperform others when getting no comparative feedback. These findings are
consistent with the solo comparison effect (Moore & Kim, 2003); in the absence of comparative
feedback, participants inferred that they would outperform others when they had achieved a high
absolute practice score and that they would earn a lower score than others when their practice
score was low. However, participants did not appear to use their comparative performance to
17
Absolute and Comparative Performance Feedback
infer their own score, given that estimated scores were equivalent across all three comparative
feedback conditions when absolute feedback was absent.
We then determined whether comparative and absolute feedback influenced percentile
estimates, which are inherently comparative. Predictably, comparative feedback exerted a strong
effect on estimated percentile score (F[2, 405] = 74.03, p < .0001), with participants estimating
their percentile highest in the 77th percentile condition, followed by participants in the nofeedback condition and then those in the 23rd percentile condition (see Table 2; all groups
differed significantly by Tukey post-hoc tests, ps < .05). There was also a main effect of
absolute feedback (F[2, 405] = 5.91, p < .01), showing that participants in the 8 condition
believed their percentile to be higher than those in the 2 condition, p < .05; the no feedback
condition fell in between these two conditions and did not differ significantly from either, ps >
.05. Once again, then, absolute feedback influenced beliefs about comparative standing. Finally,
there was a significant interaction between absolute and comparative feedback (F[2, 405] = 4.41,
p < .01), which emerged because the effects of comparative feedback were strongest when no
absolute feedback was available.
Affective Reactions to Feedback
We created two composites to represent affective reactions to practice performance. The
first reflected negative affective reactions and was composed of the items measuring anxiety and
concern resulting from one’s performance (α = .73). This composite originally included the item
measuring disappointment, but this item was only moderately correlated with anxiety (r = .21)
and reduced the reliability of the composite below acceptable levels, so this item was excluded.
The second composite represented positive affective reactions and included the items measuring
pride, satisfaction, and confidence based on one’s performance (α = .81). The latter item was the
18
Absolute and Comparative Performance Feedback
first item on the state self-esteem scale but was more highly related to the present items than to
the items on that scale. These two composites were unrelated (r = .06, n.s.), consistent with
other work suggesting that positive and negative affective variables need to be treated separately
(Russell & Carroll, 1999).
Each of the composites was then included in a 3 x 3 ANOVA including absolute and
comparative feedback as independent variables. The analysis on negative affect revealed only a
main effect of absolute feedback (F[2, 405] = 4.03, p < .02; other Fs < 1, n.s.) such that
participants receiving a score of 2 experienced more anxiety and concern about their
performance (by a Tukey post-hoc test, p < .05) than those getting a score of 8 and those getting
no feedback (which did not differ themselves, p > .05). Participants were disturbed if their score
was low, but seemed unfazed if they were told that their score was below that of other students
(as suggested by the lack of an effect of comparative feedback).
Both absolute and comparative feedback strongly influenced positive affective reactions
to performance (Fs[2, 405] = 26.33 and 92.27 for comparative and absolute feedback,
respectively, ps < .0001). Notably, the effect size for absolute feedback (r = .43) is significantly
greater than that for comparative feedback (r = .25), p < .01. Tukey post-hoc tests showed that
participants receiving an 8 were more pleased with the performance than those receiving no
absolute feedback, who in turn were more pleased than those receiving a 2 (ps < .05). Thus, as is
true of many of the variables reported here, the no-feedback group fell in between the two
feedback groups, suggesting these participants assumed moderate performance. The effects of
comparative feedback were identical; those learning their performance placed them in the 77th
percentile were more pleased than those receiving no feedback, who in turn were more pleased
than those told they placed in the 23rd percentile (ps < .05). There was also an interaction
19
Absolute and Comparative Performance Feedback
between absolute and comparative feedback (F[4, 405] = 5.75, p < .001), such that the effects of
comparative feedback were strongest when no absolute feedback was available, and absent when
participants learned they had scored an 8 on the practice test (in which case their reactions were
equivalently high regardless of comparative feedback).2
Task Importance
Based on the self-evaluation maintenance model (Tesser, 1988), we should find that
participants who get favorable feedback will find the task more interesting and will consider
successful performance to be more personally important. We combined the interest in task and
importance to self measures (r = .45, p < .01), and included them in a 3 x 3 ANOVA as before.
This analysis yielded significant main effects of both absolute and comparative feedback (Fs[2,
405] > 5.30, ps < .01), and a significant interaction (F[4, 405] = 4.43, p < .01). As predicted,
participants rated interest and importance more highly when getting favorable rather than
unfavorable comparative or absolute feedback. As seen in Table 2, the pattern differed slightly
for the two effects; for absolute feedback, the mean in the no feedback condition was not
significantly different than the poor feedback condition (score of 2), and each was significantly
lower than the good feedback condition (score of 8), ps < .05. With regard to comparative
feedback, ratings in the no-feedback condition and the 77th percentile condition were
significantly greater than those in the 23rd percentile condition. The magnitude of the effects of
absolute and comparative feedback did not differ (p > .05).
Attribution
Consistent with research on self-serving attribution (Mullen & Riordan, 1988), we
expected that internal attributions for performance would be more apparent among participants
who received positive feedback (either absolute or comparative). We tested this hypothesis by
20
Absolute and Comparative Performance Feedback
combining the two items assessing the perceived effects of luck (reverse-scored) and ability on
performance (r = .46, p < .01) and submitting them to the same 3 x 3 ANOVA as above. Indeed,
there were main effects of both absolute and comparative feedback (Fs[2, 405] = 5.44 and 3.48,
respectively, ps < .05) in the expected direction. The magnitude of these effects did not differ (p
> .05), and there was no interaction (F = 2.09, n.s.). Tukey post-hoc tests showed that
participants were more likely to make ability attributions in the 8 condition than in the 2
condition, and in the 77th percentile condition than in the 23rd percentile condition (ps < .05).
The means in the no-feedback conditions fell in between these values in both cases and were not
significantly different than those values.
State Self-Esteem
Twelve of the 13 items on the state self-esteem scale were collapsed (α = .73) into a
composite measure (the exception was the item measuring confidence in one’s ability, which was
included in the earlier composite of positive affective responses). All items were coded such that
higher values signified higher state self-esteem. This composite was then submitted to the same
3 x 3 ANOVA as above. As seen in Table 2, variability was limited; most means were over 4 on
a 5-point scale. Nevertheless, there was a significant effect of absolute feedback (F[2, 405] =
8.32, p < .001); a Tukey post-hoc test showed that those told they scored an 8 had higher state
self-esteem than those told they scored a 2. The mean in no-feedback condition was in between
these two and was not significantly different from either, ps > .05. These findings once again
demonstrate the impact of absolute feedback. On the other hand, and in contrast to Klein’s
(1997) findings, there was no effect of comparative feedback on state self-esteem (nor an
interaction, Fs < 1.8, n.s.).
Discussion
21
Absolute and Comparative Performance Feedback
Past work in disparate areas has shown that absolute standing on a dimension may
influence judgments that should be based on comparative standing (e.g., Moore & Kim, 2003),
and that comparative standing may influence judgments that should be based on absolute
standing (Klein, 1997). The present study examined the magnitude of each of these effects in the
same context, and did so across a range of self-evaluation, affective, and behavioral variables.
We found that absolute feedback was particularly influential across many of these variables.
When participants learned that they had achieved a very low score on the practice test, they not
only predicted lower scores on the actual test but also expected to perform worse than other
participants. Not surprisingly, then, they not only bet less money on the absolute task (where the
outcome depended on absolute performance), but also the comparative task (where the outcome
depended on comparative performance) – although only when no comparative feedback was
available. When comparative feedback was available, they were less likely to rely on their
absolute performance to formulate their bets. Similar patterns emerged for perceived likelihood
of winning these bets, which is not surprising given that perceived likelihood was a highly
significant predictor of bets. Absolute feedback influenced confidence in the absolute bet, as it
should, but not confidence in the comparative bet.
Those participants who received low absolute scores also appeared higher on negative
affective responses to the feedback (anxiety and concern) and lower on positive affective
responses (pride, satisfaction, and confidence), as well as on state self-esteem. Such participants
found the task less interesting and less personally important, and were less likely than other
groups to attribute performance on this task to ability rather than luck. The no-feedback control
group was sometimes significantly different than both feedback groups and other times only
different than one group, but in all cases this group fell in between the group that was given a
22
Absolute and Comparative Performance Feedback
low score and the group that was given a high score. Thus, the locus of these effects seemed to
occur in both feedback groups, an observation that would not have been possible without the
inclusion of this no-feedback group.
Our results are consistent with recent research showing the power of absolute outcomes
to drive social comparisons. For instance, even when they know it will be graded on a forced
curve, students’ expectations for getting an A on a test go down significantly as the test gets
more difficult (Windschitl et al., 2003). When an event is highly unlikely (such as being struck
by lightening or living past 100), people assume that they will be below average (Chambers et
al., 2003; Kruger & Burrus, 2004). People believe that they are below average on difficult tasks
and above average on simple ones (Moore & Kim, 2003). Note that participants in the current
study did not use absolute feedback indiscriminately; on the contrary, effects of absolute
feedback on comparative bets and inferences about comparative performance were muted, at
least relative to absolute bets and inferences about absolute performance.
By contrast, our results suggest that the effects of comparative feedback were weaker
than the effects of absolute feedback. Although comparative feedback had predictable effects on
the variables it should have influenced – namely comparative bets, perceived likelihood of
winning the comparative bet, and estimated percentile (which represents one’s comparative
standing) – it had weaker and sometimes no effects on many other variables. Comparative
feedback rarely influenced absolute bets, and had no effects on perceived likelihood of or
confidence in winning the absolute bets. These results are consistent with those of Klein (1997),
who found that comparative feedback was more likely to influence self-evaluative and affective
variables than behavioral variables. However, in the current study, even the latter variables were
not clearly affected by comparative feedback. Comparative feedback had no effect on negative
23
Absolute and Comparative Performance Feedback
affect (anxiety and concern) or on the composite self-esteem scale, which consisted mostly of
negatively-valenced items. This type of feedback did influence positive reactions such as
satisfaction with performance, but the effect size was significantly lower than that for absolute
feedback. The only variables for which comparative feedback yielded significant effects of the
same magnitude as those for absolute feedback were the measures of task interest and attribution
for performance. As was true for absolute feedback, favorable comparative feedback increased
task interest, self-relevance, and the perceived role of ability (vis-à-vis luck) in determining task
performance, consistent with past research.
Apart from the direct contrast permitted between the effects of these two types of
feedback, the crossed design also allowed for an assessment of interactions between them.
Several interesting interactions emerged. Comparative performance was more likely to influence
betting and positive affect when absolute performance was low or absent. Evidently, when
people perform well (in absolute terms), their comparative standing is less important to them –
they are happy with their excellent performance and do not consider comparative information
very important. When absolute performance was low or absent, comparative information
became more influential. This represents a caveat to Festinger’s (1954) theory, which argued for
a uniform disinterest in social comparison when absolute information was available, irrespective
of the valence of that information. For self-serving reasons, it makes sense for someone with
poor performance to seek out comparative information given the possibility that if other people
have also performed at a low level, the sting of the negative feedback can be lessened.
Why did absolute feedback produce much stronger and more consistent effects than
comparative feedback? Prior findings suggest two clear possibilities. The first is that
comparative judgments overweight information about the self, perhaps due to egocentrism
24
Absolute and Comparative Performance Feedback
(Kruger, 1999; Windschitl et al., 2003). When people are led to focus on others, the standard
effect of absolute feedback is weakened or eliminated (Moore & Kim, 2003; Windschitl et al.,
2003). Purely comparative feedback, such as the percentile rank feedback we provided our
participants, weighs self and others appropriately, and is not open to egocentric weighting.
The second possibility is that when people have absolute information only about
themselves, they infer estimates of others’ performances that are regressive (Moore & Small,
2005). If people have no clear expectations of performance, it might be sensible for them to
assume that all possible scores are equally likely (Fox & Rottenstreich, 2003). When they then
find out that their own score was fairly extreme (either good or bad) then it makes sense for them
to assume that others scores are unlikely to be as extreme. The result is that people believe that
others will be better than them when they do poorly and worse than them when they do well.
This explanation is consistent with our findings regarding participants’ estimates of absolute
scores by self and others.
We purposely chose a task for which most people have little prior experience (guessing
other people’s weights) in order to maximize the import of the feedback they received.
Consequently, participants were unlikely to have prior beliefs one way or the other about their
skills in the tested domain. This is exactly the situation that Festinger (1954) believed would
instigate the most social comparison, given that such comparison can disambiguate personal
ability in an unfamiliar domain. However, later work suggests that much of social comparison is
directed at confirming cherished positive views of the self, and that people are persuaded more
by favorable comparative feedback than by unfavorable comparative feedback (e.g., Dunning &
Hayes, 1996; Weinstein & Klein, 2001; Wood, 1989). If self-enhancement is a primary goal of
social comparison, it makes sense that comparative feedback will have less impact in cases
25
Absolute and Comparative Performance Feedback
where self-relevance is low. In the current study, participants who received no comparative
feedback did not believe they were better than average, suggesting that participants did not hold
favorable views regarding this skill that they were motivated to protect. Instead, the task in
Klein’s (1997) studies concerned esthetic judgment, a skill that participants considered desirable
and for which they believed they possessed above average skills. Clearly, it will be important to
replicate the current findings in a context where self-relevance of the ability varies orthogonally
with feedback on the associated task.
Another possibility is that our manipulation of comparative feedback was inherently
weaker than our manipulation of absolute feedback. In particular, the difference between being
in the 23rd and 77th percentile may have been perceived to be smaller than the difference between
scoring a 2 or an 8 on the practice test. On the other hand, given that the comparative feedback
told participants something meaningful about where their performance on an unfamiliar task
stood relative to that of others, it is easy to argue that it should still have been more influential.
The absolute feedback manipulation was primarily a function of the difficulty of the test, which
had been arbitrarily based on a 5-pound difference between a participant’s answer and the correct
weight. The task would have been substantially more difficult if that range had been 1 pound,
making the scores lower, and the task would have been substantially simpler if that range had
been 30 pounds. Nevertheless, absolute feedback had a more potent and consistent effect across
a range of dependent variables, consistent with Festinger’s (1954) original theory.
One important conclusion to be taken from the current data is that people are quite adept
at distinguishing between the implications of absolute and comparative feedback. Comparative
feedback yielded strong effects on comparative bets and judgments (such as estimated
percentile), and little or no effect on absolute bets and judgments. Absolute feedback had much
26
Absolute and Comparative Performance Feedback
stronger effects on absolute than on comparative bets and judgments. The evidence that people
are attuned to this difference suggests that past work may have exaggerated the extent to which
people make inappropriate inferences from performance feedback. Such careful distinctions may
be less likely in situations characteristic of high cognitive load. On the other hand, they may be
more likely in cases where outcomes are highly consequential or where participants are
answering questions on vague scales that make it easy to conflate comparative with absolute
evaluation (Moore, 2005). Future research needs to explore these possibilities.
The current study is the first to address the effects of absolute and comparative feedback
on both absolute and comparative judgments, and in the same context. The inclusion of nofeedback groups is also a unique feature of the study, and allowed for an examination of how
people respond to only one type of feedback when forced to make judgments and decisions that
are associated with the other type of feedback. As a result, the various conditions we created in
this experiment represent many of the feedback scenarios people face in everyday life.
Sometimes they may know their absolute standing on a dimension (e.g., running ability) and
must make inferences about how they will fare in a competitive setting (e.g., a marathon). Other
times they may be aware only of their comparative standing – knowing for example, that they
were selected above others for a promotion at work – and must make inferences about the
underlying absolute performance that put them ahead of others. In many other situations, both
types of feedback are available and must be weighed accordingly. To the extent that people
make these inferences carefully and rationally, decision-making will lead to favorable
consequences. Research such as the study presented here will not only enhance extant theories
of how people respond to feedback and engage in social comparison but also help in the
development of aids to promote rational decision making.
27
Absolute and Comparative Performance Feedback
References
Bazerman, M. H., Schroth, H. A., Shah, P. P., Diekmann, K. A., & Tenbrunsel, A. E. (1994).
The inconsistent role of comparison others and procedural justice in reactions to
hypothetical job descriptions: Implications for job acceptance decisions. Organizational
Behavior and Human Decision Processes, 60(3), 326-352.
Blount, S., & Bazerman, M. H. (1996). The inconsistent evaluation of absolute versus
comparative payoffs in labor supply and bargaining. Journal of Economic Behavior and
Organization, 30(2), 227-240.
Buunk, B. P., & Gibbons, F. X. (in press). Social comparison: The end of a theory and the
emergence of a field. Organizational Behavior and Human Decision Processes.
Camerer, C. F., & Lovallo, D. (1999). Overconfidence and excess entry: An experimental
approach. American Economic Review, 89(1), 306-318.
Campbell, J. D., Fairey, P. J., & Fehr, B. (1986). Better than me or better than thee? Reactions to
intrapersonal and interpersonal feedback. Journal of Personality, 54, 479-493.
Chambers, J. R., Windschitl, P. D., & Suls, J. (2003). Egocentrism, event frequency, and
comparative optimism: When what happens frequently is "more likely to happen to me".
Personality and Social Psychology Bulletin, 29(11), 1343-1356.
Cohen, J. (1992). A power primer. Psychological Bulletin, 112(1), 155-159.
Festinger, L. (1954). A theory of social comparison processes. Human Relations, 7, 117-140.
Fox, C. R., & Rottenstreich, Y. (2003). Partition priming in judgment under uncertainty.
Psychological Science, 14(3), 195-200.
Frank, R. H. (1985). Choosing the right pond : human behavior and the quest for status. New
York: Oxford University Press.
28
Absolute and Comparative Performance Feedback
Gilbert, D. T., Giesler, R. B., & Morris, K. A. (1995). When comparisons arise. Journal of
Personality and Social Psychology, 69(2), 227-236.
Greico, D., & Hogarth, R. M. (2004). Excess entry, ambiguity seeking, and competence: An
experimental investigation. Unpublished manuscript.
Hagerty, M. R. (2000). Social comparisons of income in one's community: Evidence from
national surveys of income and happiness. Journal of Personality and Social Psychology,
78(4), 764-771.
Heatherton, T. F., & Polivy, J. (1991). Development and validation of a scale for measuring state
self-esteem. Journal of Personality and Social Psychology, 60(6), 895-910.
Hoelzl, E., & Rustichini, A. (2005). Overconfident: Do you put your money on it? Economic
Journal, 115(503), 305-318.
Kahneman, D. (1999). Objective happiness. In D. Kahneman, E. Diener & N. Schwarz (Eds.),
Well-being: The foundations of hedonic psychology (pp. 3-25). New York, NY, USA:
Russell Sage Foundation.
Klar, Y., & Giladi, E. E. (1999). Are most people happier than their peers, or are they just
happy? Personality and Social Psychology Bulletin, 25(5), 585-594.
Klein, W. M. P. (1997). Objective standards are not enough: Affective, self-evaluative, and
behavioral responses to social comparison information. Journal of Personality and Social
Psychology, 72(4), 763-774.
Klein, W. M. P. (2003). Self-prescriptive, perceived, and actual attention to comparative risk
information. Psychology and Health, 18, 625-643.
29
Absolute and Comparative Performance Feedback
Kluger, A. N., & DeNisi, A. S. (1996). Effects of feedback intervention on performance: A
historical review, a meta-analysis, and a preliminary feedback intervention theory.
Psychological Bulletin, 199(2), 254-284.
Kruger, J. (1999). Lake Wobegon be gone! The "below-average effect" and the egocentric nature
of comparative ability judgments. Journal of Personality and Social Psychology, 77(2),
221-232.
Kruger, J., & Burrus, J. (2004). Egocentrism and focalism in unrealistic optimism (and
pessimism). Journal of Experimental Social Psychology, 40(3), 332-340.
Loewenstein, G., Thompson, L., & Bazerman, M. H. (1989). Social utility and decision making
in interpersonal contexts. Journal of Personality and Social Psychology, 57, 426-441.
Luttmer, E. F. P. (2005). Neighbors as negatives: Relative earnings and well-being. Cambridge:
Harvard University.
Moore, D. A. (2005). When good = better than average. Pittsburgh: Carnegie Mellon, Tepper
Working Paper 2004-E38.
Moore, D. A., & Cain, D. M. (2005). Overconfidence and underconfidence. Pittsburgh: Carnegie
Mellon University, Tepper Working Paper Number 2003-E76.
Moore, D. A., & Kim, T. G. (2003). Myopic social prediction and the solo comparison effect.
Journal of Personality and Social Psychology, 85(6), 1121-1135.
Moore, D. A., & Small, D. A. (2005). Error and bias in comparative social judgment: On being
both better and worse than we think we are. Pittsburgh: Carnegie Mellon University,
Tepper Working Paper 2004-E1.
Mullen, B., & Riordan, C. A. (1988). Self-serving attributions for performance in naturalistic
settings: A meta-analytic review. Journal of Applied Social Psychology, 18(1), 3-22.
30
Absolute and Comparative Performance Feedback
Mussweiler, T. (2003). Comparison processes in social judgment: Mechanisms and
consequences. Psychological Review, 110(3), 472-489.
Myers, D. G. (2000). The funds, friends, and faith of happy people. American Psychologist,
55(1), 56-67.
Rosenthal, R., & Rosnow, R. L. (1991). Essentials of behavioral research: Methods and data
analysis (2nd ed.). New York: McGraw-Hill.
Russell, J. A., & Carroll, J. M. (1999). On the bipolarity of positive and negative affect.
Psychological Bulletin, 125(1), 3-30.
Suls, J. M., Martin, R., & Wheeler, L. (2002). Social comparison: Why, with whom, and with
what effect? Current Directions in Psychological Science, 11(5), 159-163.
Suls, J. M., & Wheeler, L. (Eds.). (2000). Handbook of social comparison: Theory and research.
New York: Kluwer Academic/Plenum Publishers.
Suls, J. M., & Wills, T. A. (Eds.). (1991). Social comparison: Contemporary theory and
research. Hillsdale, NJ, England: Lawrence Erlbaum Associates, Inc.
Taylor, S. E., & Lobel, M. (1989). Social comparison activity under threat: Downward
evaluation and upward contacts. Psychological Review, 96(4), 569-575.
Tesser, A. (1988). Toward a self evaluation maintenance model of social behavior. In L.
Berkowitz (Ed.), Advances in experimental social psychology (Vol. 21, pp. 181-227).
New York: Guilford.
Tesser, A., & Campbell, J. (1980). Self-definition: The impact of the relative performance and
similarity of others. Social Psychology Quarterly, 43(3), 341-346.
31
Absolute and Comparative Performance Feedback
Windschitl, P. D., Kruger, J., & Simms, E. (2003). The influence of egocentrism and focalism on
people's optimism in competitions: When what affects us equally affects me more.
Journal of Personality and Social Psychology, 85(3), 389-408.
32
Absolute and Comparative Performance Feedback
Footnotes
1
Effect sizes in this paper are reported in terms of r as recommended by Rosenthal and Rosnow
(1991). Effect sizes of .10, .30, and .50 are generally characterized as small, medium, and large,
respectively (Cohen, 1992).
2
In exploratory analyses, we assessed whether these affective reactions mediated the effect of
feedback on betting behavior. Although affective reactions and bets were significantly
correlated, the effect of feedback on bets was unchanged when including affect in the model,
suggesting no mediation. Details of these analyses are available from the authors.
33
Absolute and Comparative Performance Feedback
Author Note
We thank Lauren DeVito, Wemi Peters, Sapna Shah, Kavita Shah, and Sam Swift for their
assistance in collecting the data reported herein, and Rebecca Miller and Jennifer Cerully for
their assistance in preparation of this manuscript. Correspondence concerning this manuscript
may be sent to Don A. Moore, Tepper School of Business, Carnegie Mellon University, 5000
Forbes Avenue, Pittsburgh, PA 15213 (e-mail: [email protected]) or William M. P. Klein at
Department of Psychology, University of Pittsburgh, 210 South Bouquet Street, Pittsburgh, PA
15260 (e-mail: [email protected]).
34
Absolute and Comparative Performance Feedback
Table 1
Bets, perceived probability, and confidence in each bet, by condition
Absolute feedback
None
Comparative Feedback:
2
8
None
(n = 49)
$0.86
(0.72)
23
(n = 43)
$0.78
(0.73)
77
(n = 46)
$0.76
(0.65)
None
(n = 48)
$0.41
(0.70)
23
(n = 45)
$0.48
(0.69)
77
(n = 46)
$0.36
(0.52)
None
(n = 46)
$1.74
(0.89)
23
(n = 46)
$1.82
(0.84)
77
(n = 45)
$1.30
(0.82)
Comparative bets
(out of $4)
$1.15
(0.87)
$0.94
(0.94)
$1.97
(0.87)
$0.57
(0.69)
$0.69
(0.76)
$1.65
(0.91)
$1.20
(0.83)
$0.70
(0.65)
$1.38
(0.83)
Absolute bet likelihood
(0-100)
43.37
(23.86)
35.79
(25.38)
50.48
(24.98)
25.19
(23.52)
22.91
(20.73)
18.27
(18.85)
71.78
(19.98)
72.61
(19.69)
61.89
(23.33)
Comparative bet likelihood
(0-100)
52.39
(22.65)
37.42
(24.77)
75.41
(15.47)
37.67
(22.20)
34.96
(23.14)
62.04
(22.16)
56.73
(21.41)
37.61
(24.24)
64.98
(18.07)
Absolute bet confidence
(0-5)
2.53
(1.16)
2.36
(1.03)
2.78
(1.19)
1.71
(1.03)
1.95
(1.24)
1.74
(0.93)
3.50
(0.98)
3.61
(0.95)
3.33
(1.00)
Comparative bet confidence
(0-5)
2.96
(1.08)
2.37
(1.07)
3.78
(0.84)
2.33
(0.98)
2.53
(1.10)
3.50
(0.78)
3.07
(1.18)
2.46
(1.26)
3.49
(0.66)
Absolute bets
(out of $4)
___________________________________________________________________________________________________________
Note: Standard deviations appear in parentheses.
35
Absolute and Comparative Performance Feedback
Table 2
Other dependent measures, by condition
Absolute feedback
None
Comparative Feedback:
2
8
None
(n = 49)
4.80
(1.57)
23
(n = 43)
4.42
(1.64)
77
(n = 46)
5.57
(1.31)
None
(n = 48)
3.58
(1.64)
23
(n = 45)
3.64
(1.37)
77
(n = 46)
3.41
(1.29)
None
(n = 46)
6.83
(1.18)
23
(n = 46)
7.02
(1.27)
77
(n = 45)
6.40
(1.45)
Predicted average score
4.51
(1.12)
4.79
(1.06)
4.09
(1.36)
4.24
(1.33)
4.29
(1.16)
2.33
(1.54)
5.98
(1.26)
7.10
(1.97)
6.40
(8.22)
Predicted percentile rank
39.96
(20.60)
34.13
(16.48)
61.17
(12.30)
34.79
(16.28)
35.44
(12.50)
58.63
(14.41)
52.39
(19.29)
37.96
(19.51)
58.67
(15.06)
Negative affect
2.60
(1.08)
2.40
(0.99)
2.50
(0.98)
2.26
(1.04)
2.20
(1.10)
2.16
(0.91)
2.43
(1.07)
2.71
(1.01)
2.42
(0.92)
Positive affect
2.80
(0.76)
2.25
(0.82)
3.49
(0.76)
2.03
(0.62)
2.07
(0.88)
2.56
(0.83)
3.53
(0.72)
3.32
(0.76)
3.59
(0.71)
Task interest
3.77
(0.79)
2.99
(0.99)
3.84
(0.87)
3.01
(1.01)
3.22
(1.16)
3.24
(0.95)
3.46
(0.95)
3.58
(0.84)
3.81
(0.72)
Prediction of own score
(continued on next page)
36
Absolute and Comparative Performance Feedback
Absolute feedback
None
Comparative Feedback:
2
8
None
(n = 49)
23
(n = 43)
77
(n = 46)
None
(n = 48)
23
(n = 45)
77
(n = 46)
None
(n = 46)
23
(n = 46)
77
(n = 45)
Attribution to ability
2.96
(0.95)
2.78
(0.87)
3.35
(0.84)
2.71
(0.99)
2.83
(0.99)
3.08
(0.77)
3.35
(0.86)
3.15
(0.89)
3.17
(0.72)
State self-esteem
4.36
(0.52)
4.32
(0.49)
4.40
(0.48)
4.16
(0.48)
4.16
(0.56)
4.33
(0.57)
4.52
(0.41)
4.39
(0.59)
4.49
(0.43)
Absolute bet outcome
$1.19
(0.88)
$1.30
(0.84)
$1.45
(0.83)
$1.65
(0.69)
$1.58
(0.74)
$1.69
(0.48)
$0.10
(1.21)
$-0.16
(1.22)
$0.72
(1.16)
Comparative bet outcome
$2.15
(1.67)
$1.88
(1.58)
$2.02
(2.67)
$2.02
(0.99)
$1.89
(1.11)
$1.97
(2.23)
$1.98
(1.65)
$1.93
(1.00)
$1.88
(1.82)
___________________________________________________________________________________________________________
Note: Standard deviations appear in parentheses.
37