 

Syntv

e “my,

V .r (an ..

:2
>
it!

aim.

‘ cl
. .5..."- .5395

 

:

.
a, 4 to! :31"!

V v, I 4 0A
‘1'? .1L{’

a“) .. 3:0 {1‘ .

yﬁmbdnmmig
fii Emun. ,.

96 .1.

v l
ﬁliiﬂﬁur
l..1‘uﬂllﬂvﬂnni.3a
*{vggl .

.. .. . 3A zit?

$4 -- ;
gigs};

; 33.

. (1‘11
L .!!

.n r ’4 11.0.. r.

r .l .téﬁiéli:

1 5i. 3..
. INHHVI.

5...

.'

 

i .

:5:

13:“.-

.S :5;

1
Vt

Willilllilllmllllllll

3 1293 O

 

 

 

This is to certify that the

thesis entitled

Race Effects on Performance Evaluation in a Team
Situation: An Attributional Perspective

presented by

Aleksander P.J. Ellis

has been accepted towards fulﬁllment
of the requirements for

M.A. Psychology

degree in

 

 

 

Majo professor

Date /0/Z.5’/?9

0—7639 MS U is an Affirmative Action/Equal Opportunity Institution

 

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE
2

p33 1 2001

$90401

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

11/00 WM“

RACE EFFECTS ON PERFORMANCE EVALUATION IN A TEAM
SITUATION: AN ATTRIBUTIONAL PERSPECTIVE
By

Aleksander R]. Ellis

A THESIS
Submitted to
Michigan State University
In partial fulﬁllment of the requirements
For the degree of

MASTER OF ARTS

Department of Psychology

1999

ABSTRACT
RACE EFFECTS ON PERFORMANCE EVALUATION IN A TEAM
SITUATION: AN ATTRIBUTIONAL PERSPECTIVE
By
Aleksander PJ. Ellis
This study examined the way in which the race of a team leader inﬂuenced attributions
and evaluations of leader performance on a team task with distributed expertise. There
were two main purposes of this research: to introduce attribution theory as an explanation
of racial differences in performance rating and to examine performance ratings in a team
situation, where subordinates (the raters) interacted with their superior (the ratee).
Research on performance rating and race has, for the most part, ignored both attribution
theory and team settings when attempting to explain racial differences in evaluating an
individual’s performance level. This study focused solely on racial differences in ratings
of White American versus African American team leaders. It was proposed that certain
attributions based on racial stereotypes would be created depending on the team leader’s
level of performance. These attributions would then moderate the direct relationship
between a team leader’s level of performance and performance ratings. In addition, it was
hypothesized that race would also have a direct inﬂuence on the level of performance
ratings. The results did not support the original model, since race was not directly related
to performance ratings. Instead, it was found that race interacted with the team leader’s
actual performance in predicting performance ratings. In the high performance condition,
attributions mediated the relationship between race and performance ratings. The

practical and theoretical implications of these results are discussed.

TABLE OF CONTENTS

INTRODUCTION .......................................................................................................... 1
Race and Performance Ratings .................................................................................. 1
Theories of the Cognitive Processes Underlying Racial Differences in Ratings ...... 5

Tokenism ............................................................................................................ 5
Similar-to-me Effect ........................................................................................... 6
Social Identity Theory ........................................................................................ 6
Intergroup Theory and Positivity Bias ............................................................... 7
Attribution Thoery ..................................................................................................... 9
Attribution Theory and Performance Ratings ......................................................... 11
Attribution Theory, Race, and Performance Ratings .............................................. 12
Teams, Race, Attributions, and Performance Ratings ............................................. 19
Hypotheses .............................................................................................................. 22

Method ......................................................................................................................... 26
Participants .............................................................................................................. 26
Task ........................................................................................................................ 26
Procedures ............................................................................................................... 3 1
Measures and Manipulations ................................................................................... 33

Race .................................................................................................................. 33
Actual Performance .......................................................................................... 33
Causal Attributions ........................................................................................... 33
Performance Ratings ........................................................................................ 38

Results .......................................................................................................................... 38

Manipulation Checks ............................................................................................... 39
Race .................................................................................................................. 39
Team Performance ........................................................................................... 39

Tests of Hypotheses ................................................................................................. 41
Hypothesis 1 ..................................................................................................... 41
Hypothesis 2 ..................................................................................................... 41
Hypothesis 3 ..................................................................................................... 41
Hypothesis 4 ..................................................................................................... 42
Hypothesis 5 ..................................................................................................... 43
Hypothesis 6 ..................................................................................................... 45

Summary of Hypotheses ......................................................................................... 47

Supplemental Analyses ........................................................................................... 48
Subordinate Race .............................................................................................. 48
Subordinate Performance ................................................................................. 48

Post-hoe Analyses ................................................................................................... 49

Discussion .................................................................................................................... 56
Theoretical Implications .......................................................................................... 63
Practical Implications .............................................................................................. 66
Study Limitations .................................................................................................... 67
Future Research ....................................................................................................... 68

REFERENCES ............................................................................................................. 70

APPENDIX A Items Measuring Internal and External Attributions ........................... 80
APPENDIX B Performance Rating Items .................................................................... 82
APPENDIX C Performance Rating Items Scaled l to 100 .......................................... 84
APPENDIX D Power Analyses ................................................................................... 86

LIST OF TABLES
Table 1. Intercorrelation between internal and external attribution items ................... 35
Table 2. Factor categories derived from factor analysis of 8 causal attribution items. 36

Table 3. Intercorrelation between 19-item performance rating scale and 2 rating items
scored 1 to 100 ............................................................................................... 38

Table 4. Means, standard deviations, and intercorrelations among all analyzed
Variables ......................................................................................................... 40

Table 5. Result of regressing internal attributions on leader race, actual team
performance, and their interaction .................................................................. 42

Table 6. Result of regressing external attributions on leader race, actual team
performance, and their interaction .................................................................. 43

Table 7. Result of regressing leader performance ratings on actual team performance,
internal attributions, and their interaction ...................................................... 45

Table 8. Result of regressing leader performance ratings on actual team performance,
external attributions, and their interaction ...................................................... 47

Table 9. Result of regressing leader performance ratings on leader race, actual team
performance, and their interaction .................................................................. 50

Table 10. Means, standard deviations, and intercorrelations among all variables
analyzed post-hoe in the high performance condition .................................. 52

Table 11. Result of regressing leader performance ratings on leader race while
controlling for internal attributions .............................................................. 53

Table 12. Result of regressing leader performance ratings on internal attributions
while controlling for leader race in the high performance condition ........... 53

Table 13. Result of regressing leader performance ratings on leader race while
controlling for external attributions in the high performance condition ...... 55

Table 14. Result of regressing leader performance ratings on external attributions
while controlling for leader race in the high performance condition ........... 55

LIST OF FIGURES

Figure l. Weiner’s emotive theory ............................................................................... 10

Figure 2. Mediation model of the relationship between task and social attraction

and performance feedback ............................................................................ 14
Figure 3. The relationship between race and causal attributions ................................. 15
Figure 4. The relationship between race and performance ratings ............................... 16
Figure 5. The relationship between actual performance and performance ratings ...... 16
Figure 6. The interaction between race and actual performance .................................. 17
Figure 7. The interaction between actual performance and causal attributions ........... 18

Figure 8. Model of the relationship between race and performance ratings in a
team situation ............................................................................................... 23

Figure 9. Hypothesized graph of the interaction between race and actual

performance .................................................................................................. 24
Figure 10. Second hypothesized graph of the interaction between race and actual
performance ................................................................................................. 25
Figure 11. Hypothesized graph of the interaction between causal attributions and
actual performance ...................................................................................... 25
Figure 12. Second hypothesized graph of the interaction between causal attributions
and actual performance ............................................................................... 26
Figure 13. Role requirements ....................................................................................... 28
Figure 14. Measurable attributes .................................................................................. 30
Figure 15. Role structure .............................................................................................. 30
Figure 16. Conﬁrmatory factor analysis of 8 attribution items .................................... 37

Figure 17. Interaction of leader race and actual team performance on external
attributions ................................................................................................... 44

vi

Figure 18. Interaction of internal attributions and actual team performance on leader
performance ratings ..................................................................................... 46

Figure 19. Interaction of leader race and actual team performance on leader
performance ratings ..................................................................................... 51

vii

INTRODUCTION

Over past years there has been a considerable amount of research in the area of
performance appraisal, which has led investigators to look at individual characteristics in
explaining variance in performance ratings (Landy & Farr, 1980). One focus has been on
the characteristic of race. Several early studies investigating the effects of race on
performance ratings found that raters tended to give higher ratings to those of the same
race (Cox & Krumboltz, 1958; DeJung & Kaplan, 1962). Further experimentation
supported these ﬁndings by showing that subjects favored applicants of the same race
even when objective measures were deﬁned (Hamner, Kim, Baird, & Bigoness, 1974).
Yet several studies disagreed with the assertion that African Americans rate African
Americans higher and White Americans rate White Americans higher. For instance, it
was discovered that when performance was high, African Americans and White
Americans were rated similarly, and when performance was low, African Americans
were actually rated higher than White Americans (Bigoness 1976; Rotter & Rotter 1969).
It was then proposed that the proportion of racially similar individuals in an assessment
group would have a signiﬁcant effect on ratings. Although the effects were small, White
American males tended to be rated higher, while African American females tended to be
rated lower, when the proportion of White American males in the assessment group
increased (Schmitt & Hill, 1977). There were also several studies that found no
signiﬁcant race effects on performance ratings (e. g..Bass & Turner, 197 3; Fox &
Lefkowitz, 1974; Schmidt & Johnson, 1973).

With the apparent lack of consistent effects on ratings in many of the

aforementioned ﬁeld studies (e. g. Bass & Turner, 1973), Wendelken and Inn (1981)

argued that the effects found in the early laboratory studies were not generalizable to the
organizational environment. Face-to-face interaction within the workplace, according to
the authors, would dampen the effect of group stereotypes directed at individual group
members. Their ﬁeld data supported this assertion by showing that insigniﬁcant
proportions of the variance in performance ratings were accounted for by the race of the
rater, the individual being rated, and their interaction. Others have argued to the contrary;
their position was that the effects should be less likely in the laboratory because of
college students’ inclination to provide socially desirable responses (Jones & Signal],
1971). In summary, a decisive conclusion with respect to racial bias in performance
ratings could not be reached from the early literature.

Kraiger and Ford (1985) attempted to consolidate these diverse ﬁndings across
settings and populations by conducting a meta-analysis in order to obtain more deﬁnitive
data on the effect of race on performance evaluations. Their review included over seventy
studies, most of which came directly from ﬁeld samples. The results yielded a mean
point-biserial correlation between the race of the ratee and performance ratings of .18
(corrected for unreliability) for White American raters and -.22 for African American
raters. Respectively, these two correlations determined 1.0% and 3.2% of the population
variance for White American and African American raters. In other words, African
Americans gave higher ratings to African Americans and White Americans gave higher
ratings to White Americans. The authors concluded that race has a moderate impact on
performance ratings with both laboratory and ﬁeld samples, directly contradicting the
previously discussed results obtained by Wendelken and Inn (1981) and others (e. g. Bass

& Turner, 1973).

However, the meta-analytic results of Kraiger and Ford still did not allow for
determining whether differences in ratings were due to bias or actual differences in
performance (Kraiger & Ford, 1985; Hunter & Hirsch, 1987). Another problem was the
lack of independence of the rating observations. As in most ﬁeld research on performance
rating measures, one rater evaluated more than one individual. This inevitably
complicated the distinction between within-rater and between-rater variance (Pulakos,
White, Oppler, & Borman, 1989).

Subsequent research by Pulakos et al. (1989) was directed at eliminating previous
problems when examining the role of race in performance evaluations. The data were
collected as part of Project A, a large scale, long term research endeavor conducted in the
military to improve selection and classiﬁcation of new army recruits (Eaton, Goer,
Harris, & Zook, 1984). Data sets were formed where each individual was rated by two
raters- one from the same racial group as the ratee and one from another. In this design,
actual performance was held constant as well as the number of raters assigned to a
particular individual. The results were slightly different from those of Kraiger and Ford
(1985). In this case, race accounted for no more than 1% of the variance in ratings.

However, there were problems with the research done by Pulakos et al. (1989)
that could have shrunk the observed effects by a signiﬁcant proportion. The individuals
who participated in the study also participated in a training program speciﬁcally designed
to eliminate nonperformance factors (e. g. race) from performance ratings. In addition, the
study was based entirely on a sample of military personnel. The percentage of African

Americans in the military was, and still is, much higher than in most organizations

(Pulakos et al., 1989). Perhaps race was not as salient to those participating in the military
study, and thus did not inﬂuence their ratings.

Sackett and DuBois (1991) attempted to translate the ﬁndings of Pulakos et al.
(1989) to the civilian sector by using data from ﬁrms across the United States and
combining it with the military data from Project A in order to expand the data base on
rater-ratee race effects. Similar to previous research, the civilian data employed a design
in which each individual was rated by two raters (i.e. one White American and one
African American) allowing a within subjects analysis. The results obtained from the
military and civilian data were strikingly different from those obtained by Kraiger and
Ford (1985) and Pulakos et al. (1989). In both of those meta-analyses, African Americans
rated African Americans higher and White Americans rated White Americans higher. Yet
that conclusion was not supported by the extremely large amount of data collected by
Sackett and DuBois. In fact, both African Americans and White Americans gave lower
ratings to African Americans and higher ratings to White Americans in both low and high
performance situations.

Although the research by both Pulakos et al. (1989) and Sackett and DuBois
(1991) focused on solving methodological issues of previous research, they neglected the
greater issue of understanding the cognitive processes underlying race effects on
performance ratings. In support of this proposition, Kraiger & Ford (1985) noted that
further research should concentrate on the cognitive processes involved in evaluating

performance instead of the magnitude of racial bias.

Theories of the Coggitive Processes Underlying Racial Differences in Ratings

Several explanations have been proposed in the literature. For instance, Kraiger

 

and Ford (1985) suggested that tokenism (Kanter, 1977) may explain why studies have
found that, as the proportion of group members possessing a certain characteristic (e. g.
race) decreases, that characteristic becomes easier to process and recall when searching
for information about the individual (Taylor, 1981). These minority characteristics (e. g.
race) “carry assumptions about culture, status and behavior highly salient for majority
group members” (Kanter, 1977, p.968). According to Kanter, the “token” member is
singled out and viewed negatively by the majority. A group member is seen as “token”
when the subgroup to which he or she belongs comprises 35% or less of the work group.
When 15% or less are present, the situation is referred to as “skewed,” while more than
15% results in a “tilted” environment (Kanter, 1977). Research on tokenism applied to
race has found mixed results. Although Ziller, Behringer, and Goodchilds (1960) found
that when African Americans enter a group they tend to be characterized by the single
characteristic of race, their results failed to reach statistical signiﬁcance. However,
similar studies have shown that for most comparisons, the racially distinctive member of
a group was judged signiﬁcantly less positively than the majority members (e. g. Craig,
1996). Firm conclusions on the theory of tokenism applied to performance ratings,
unfortunately, are difﬁcult to ﬁnd due to the nonsigniﬁcant ﬁndings of other researchers.
Sackett, DuBois, and Noe (1991) demonstrated that, when the proportion of African
Americans in a work group increases, there is no signiﬁcant difference in performance
ratings, indicating an absence of group composition effects. Support for the theory

becomes even more difﬁcult to ﬁnd when the data of Sackett and DuBois (1991) is taken

into account. Tokenism cannot explain why African Americans rated other African
Americans lower than White Americans.

Aside from tokenism, other theories have also been applied to the issue of rating
individuals based on race. Two theories, the similar-to-me effect and social identity
theory, have been examined and seem to converge in their apparent explanation of the
process underlying the evaluation of others. The crux of the similar-to-me effect is
interpersonal attraction, where individual’s are attracted to those similar to them (Byme,
1971). And the result is often that higher ratings are given to similar individuals both in
ﬁeld (Lin, Dobbins, & Farth, 1992) and in laboratory (e.g. Rand & Wexley, 1975)
studies.

Social identity theory posits that people inherently wish to sustain a positive self-
regard, which leads them to view others in their same group as positive (Messick &
Mackie, 1989, p.59). As a consequence, those individuals within the group are evaluated
more favorably than outsiders (T ajfel, 1981). To become a member of such a group, the
individual must possess certain characteristics such as belonging to the same social
category (Turner, 1984, p.530).

In essence, both social identity theory and the similar-to—me effect agree that
individuals of the same group (i.e. race) will rate each other in a more positive manner
than those outside the group. This effect will take place only when other members of the
group are present (Doise, 1978; Turner, Hogg, Oakes, Reicher, & Wetherell, 1987). In
the case where the individual is the sole representative from his or her group, the person
would instead opt to identify with the members of the majority (Davis et al., 1995,

p.131). Support has been partially received for this theory. It was found that, in an

interview situation with equal African American and White American panel members,
individuals tended to rate same-race job candidates higher. In addition, when an African
American rater was alone among a group of White Americans, he or she tended to give
higher ratings to the opposite race, supporting the contention that the similar-to-me effect
only takes place with others of the same group. However, although the pattern was
similar for the case with one White American interviewer, the results did not reach
statistical signiﬁcance and thus did not provide strong support for the argument that one
White American will identify with the African American majority. (Prewett-Livingston,
Field, Veres, & Lewis, 1996)

Other interview rating researchers have found no support for the similar-to-me
effect and social identity theory (e. g. Campion, Pursell, & Brown, 1988). In other words,
African Americans and White Americans were rated similarly by the interview panel.
Unfortunately, although the studies mentioned above examined race, they based their
ratings on interviews rather than observations of on-the-job performance. On the other
hand, studies that did use on-the-job performance indices did not examine race (e. g.
Cardy & Dobbins, 1986; Dovido & Gaertner, 1983). More importantly, none of the
studies actually measured perceived similarity to the individual being rated.

Kraiger and Ford (1990) continued the search for the cognitive processes
underlying racial bias in performance ratings by completing a second meta-analysis. They
examined another two theories, intergroup theory (Alderfer, 1986) and positivity bias
(Pettigrew, 1979), as possible explanations for their results. Intergroup theory proposes
that evaluations of performance are inﬂuenced by actual performance level differences,

but are also dependent on the perceptions of that performance. For example, if two

employees performed equally well, one would be perceived as having performed better
because the rater identiﬁed with his or her race. The second theory, positivity bias,
hypothesizes that those in the majority group will bolster the ratings of other majority
group members. In this case, higher ratings are given to those of the majority race instead
of lower ratings being given to the minority group member. The inﬂation of ratings is due
to the addition of job-irrelevant factors when rating majority group members. Kraiger and
Ford (1990) found that majority members (White Americans) tended to focus more on
work performance (job-relevant) measures when rating minority members (African
Americans), but did not focus more on job-knowledge (job-relevant) measures. In other
words, ﬁrm support for the use of intergroup theory in understanding the process of
evaluating performance was not evident in Kraiger and Ford’s (1990) data. And both
intergroup theory and positivity bias are unable to accurately explain why African
Americans would rate other African Americans lower, as Sackett and DuBois (1991)
previously reported.

In sum, tokenism, the similar-to-me effect, social identity theory, intergroup
theory, and positivity bias received only mixed support from the literature on race effects
in performance appraisal. More importantly, none of the research supporting these
theories utilized the technique of meta-analysis to the extent that it was used by Sackett
and DuBois (1991). Their civilian data included 12,022 White Americans rated by White
American raters, 661 White Americans rated by African American raters, 5,972 African
Americans rated by White American raters, and 1,110 African Americans rated by
African American raters for the between subjects analyses. The within subjects analysis,

where each individual was rated by both a African American and a White American rater,

consisted of 286 White Americans and 331 African Americans. The military study, done
by Pulakos et al. (1989), used more than 35,000 individuals in the between subjects
analysis and more than 1,800 individuals in the within subjects analysis. The sum of these
two massive data sets by Sackett and DuBois created a very powerful tool with which to
examine race effects in the process of performance evaluation. The civilian ratings of
overall performance and part of the military results clearly showed that African
Americans were rated lower not only by White Americans, but also by other African
Americans. This result was in direct conﬂict with previous meta-analytic results (e. g.
Kraiger & Ford, 1985) that African Americans and White Americans rate the same race
higher. Another problem this created was that none of the theories previously enumerated
could explain the psychological process behind giving lower ratings to individuals of the
same race, especially since it was often a one on one rating system devoid of any group
(i.e. token) effects. However, it is suggested here that one theory that has not been
examined, attribution theory, could potentially explain variance in the relationship
between race, actual performance, and performance ratings. The next section addresses it.
Attribution Theog

Bernard Weiner (1985) wrote that, when examining the behavior of another
individual, one has an inherent need to know why someone did what they did. Was it due
to ability, hard work, luck, or good looks? “The attributor is not simply an attributor, a
seeker after knowledge; his latent goal in attaining knowledge is that of effective
management of himself and his environment” (Kelly, 1971, p.22). The search for
causality has been conducted in a wide variety of research contexts over many decades

(e.g. Feather & Davenport, 1981; Heider, 1958; Rosenbaum, 1972; Rotter, 1966).

Weiner (1985), in reviewing many of the studies, concluded that past
investigation had shown that there are only a small number of applicable causes that
individuals utilize, and these causes all share three main properties: locus, stability, and
controllability. The ﬁrst property, locus, is used to assign cause either internally (due to
the person) or externally (due to his or her environment). For example, cognitive ability is
an internal cause while the difﬁculty of the task is an external cause, since it is not within
the person. As for the second property, cognitive ability is also viewed as a stable cause
while the difﬁculty of the task is viewed as unstable. Cognitive ability, in reference to the
third property, is seen as uncontrollable since the individual does not have the option of
immediately increasing their cognitive abilities.

Following Schachter and Singer’s (1962) famous work dealing with emotional
states, Weiner (1985) developed his own emotive theory. Experience, similar to
Schachter and Singer’s model, is reﬁned and differentiated as more complex cognitions

enter the emotive process (see Figure 1).

General positive or negative emotions

Outcome —> Outcome evaluatio

Causal attribution and dimensions

1

Distinct emotions

Figure 1. Weiner’s emotive theory ( from Weiner, 1985, p.560)

10

The ﬁrst step in the abbreviated model above is a “primitive” one for the
individual. A general negative or positive reaction occurs depending on whether the event
was a success or a failure. This general reaction is attribution independent and outcome
dependent. Following the immediate reaction, a reappraisal of the event takes place and
the individual begins to search for causal attribution. This step is attribution dependent
and, after the individual locates the appropriate causal ascription, different emotions will
be formed according to the locus, controllability, and stability of the cause.

Attribution Theog and Performance Ratings

Weiner’s ( 1985) emotive theory, as described above, can be directly relevant to
the organizational environment. Often employees are asked to rate the performance of
others within the workplace. When observing an individual performing, a general
reaction takes place depending on whether their effort was a success or failure. Obviously
individuals will be evaluated higher if their behavior was a success than if it was a
failure. However, according to attribution theory, this evaluation is then reﬁned as the
observer searches for the appropriate causal attribution based on the locus, controllability,
and stability of the cause. According to the theory, in the case of success, the best ratings
will be given to the individual who is believed to have performed well because of some
internal attribute such as ability, while those who performed well due to external
attributes such as luck will be rated lower. In the case of failure, the best ratings will be
given to the individual who performed poorly because of external attributes, while lower

ratings will be given to those who performed poorly due to internal attributes.

ll

Attribution Theog, Race, and Performance Ratings

Attribution theory, as described above, has not been used in the literature as a
potential explanation for the effects of race on performance ratings. However, researchers
have suggested that “good performance for minorities might be attributed to external
factors (e. g., luck or the task) or unstable factors (e. g., effort) rather than to ability. Good
performance for majority group members, on the other hand, might be more likely
attributed to ability, which is both internal and stable” (Ilgen & Youtz, 1986, pp. 313-
314).

It has been shown that African Americans, out of all the racial minorities in the
United States, are more negatively stereotyped than anyone else (Smith, 1990).
Attribution researchers propose that stereotypes like this create expectations concerning
individual behavior and thus inﬂuence causal attributions (e.g., Deaux, 1976; Jones &
McGillis, 1976). Behavior that is inconsistent with the stereotype is attributed to external
causes, while consistencies are attributed to internal causes (e.g., Heilman, 1983; Nieva
& Gutek, 1980). The research linking stereotypes and attributions has, for the most part,
supported the assertion that the success of women and African Americans is more likely
to be attributed to external causes such as luck than that of White males. Failure,
however, is attributed more to internal causes such as ability (e.g., Ilgen & Youtz, I986;
O’Leary & Hansen, 1984). These results have important implications because, when the
performer attributes his or her failure to the difﬁculty of the task (external-unstable),
individuals will continue to work towards success. However, attributions of low ability
will tend to lead to feelings of hopelessness in the face of future difﬁculty (Andrews &

Debus, 1978; Chapin & Dyck, 1976; Dweck, 1975).

12

Although the literature appears consistent with an attribution interpretation, there
is research that does not completely agree with the majority position. Jackson, Sullivan,
and Hodge (1993) found that, in reviewing college applications of African Americans
and White Americans, subjects viewed effort as an intemal-stable cause when behavior
was stereotype-inconsistent for a period of one year. Although part of their results
differed slightly from previous research, Jackson and her colleagues’ reached conclusions
similar to the standard position that stereotypes inﬂuenced attributions, stereotype-
inconsistent behavior (high performance for African Americans and low performance for
White Americans) resulted in external attributions, stereotype-consistent behavior
resulted in internal attributions, and attributions inﬂuenced evaluations.

However, the ﬁndings of Jackson et al. (1993), like many of the studies in the
literature, may not relate to the performance rating contexts within organizations. For
example, evaluations given to both African Americans and White Americans resulted
from the examination of paper descriptions, which listed an individual’s background and
personal characteristics (e.g., Whitehead, Smith, & Eichhom, 1982). Within an '
organization, however, there are face to face interactions between individuals that could
inﬂuence performance ratings through a different process and could result in different
conclusions. In fact, Pettigrew (1979) felt that a true test of attributional error could occur
only when individuals interact. The present study was designed to closely mimic an
actual organizational environment. Perhaps the communication between real people
would increase the rating effects found using “paper people.” More importantly, most
studies in the literature do not consist of subjects actually observing an individual’s

performance ﬁrsthand. When subjects participate in research and end up looking at a

13

piece of paper, there is no personal connection to those they are rating. Finally, there are
no consequences for the failure or success of the ratee in much of the literature. If
subjects are working with an individual toward a common goal (e. g. a monetary reward),
they may be more likely to rate that individual as an employee would in a real
organization. In conclusion, face-to-face interaction, observing an individual’s
performance ﬁrsthand, and working toward a common goal are all critical boundary
conditions that are lacking in most research to date.

DeCarlo and Leigh (1996) attempted to expand the attribution literature to the
organization by looking at the impact of a salesperson’s task and social attraction on a
sales manager’s causal attributions. The authors deﬁned task attraction as “the
desirability of the salesperson to the sales manager as a work partner” and social
attraction as “the desirability of the salesperson to the sales manager as a friend and

social partner” (DeCarlo & Leigh, 1996, p.48). Their basic model is pictured in Figure 2.

Task and social attraction
causal attributions—p manageri performance feedback

for performance

Figure 2. Mediation model of the relationship between task and social attraction and
performance feedback

Causal attributions enter the picture as a partial mediator between task/social attraction
and feedback. In other words, task and social attraction were hypothesized to directly
affect feedback and to inﬂuence causal attributions which, in turn, were hypothesized to

inﬂuence managerial performance feedback. Regarding the ﬁrst link, those salespeople

l4

high in social and task attraction beneﬁted by receiving external causal attributions from
their managers when they performed poorly and internal causal attributions when they
performed well. On the other hand, those salespeople low in task and social attraction
received internal attributions for poor performance and external attributions for good
performance. Unfortunately, DeCarlo and Leigh did not include actual performance or
performance ratings in their model. However, the inﬂuence of task and social attraction
on causal attributions can be translated to the performance rating context.

Logically, the next step is to combine the previous research on attribution theory
and the literature on race and performance ratings. In doing so, DeCarlo and Leigh’s
(1996) model can be used by replacing task and social attraction with racial stereotypes,
which is supported by a number of researchers (e.g., Ilgen & Youtz, 1986; O’Leary &
Hansen, 1984). It is expected that stereotypes concerning African Americans will
function in much the same way as low task and social attraction. White American
stereotypes, on the other hand, are hypothesized to coincide with high task and social

attraction (see Figure 3).

 

Race ’ Causal Attributions

Figure 3. The relationship between race and causal attributions

Bringing in the work of Schmitt, N ., Rogers, W., Chan, D., Sheppard, L., &

Jennings, D. (1997), race is expected to have a direct effect on performance ratings (see

15

Figure 4). In their meta-analysis, Schmitt et al. found that race was correlated .18 with

performance ratings.

Race + Performance Ratings

\ Attributions

Figure 4. The relationship between race and performance ratings

Hamner, Kim, Baird, and Bigoness’(1974) research can also be added to the
model, since they found that actual performance has a direct effect on performance
ratings (see Figure 5). Poor performance resulted in low performance ratings, while good

performance resulted in high performance ratings.

Race , Performance Ratings

\ Attributions

Actual Performance
Figure 5. The relationship between actual performance and performance ratings
However, the direct effects of race on attributions clearly depend on the actual
performance of an individual on a speciﬁc task, which is a moderated relationship (see

Figure 6). Although there has been little relevant research on this link, some studies have

indicated that racial effects on performance attributions are moderated by actual

l6

performance (Whitehead, Smith, and Eichhom, 1982). Ilgen & Youtz (1986)
hypothesized that, based on previous research (e. g., Deaux & Emswiler, 1974; Heilman,
1983), African Americans would receive internal attributions for failure and external
attributions for success while White Americans would receive external attributions for

failure and internal attributions for success.

Race , Performance Ratings

\ , ,
Attributions

Actual Performance

 

Figure 6. The interaction between race and actual performance

The last step is to connect causal attributions to performance ratings (see Figure
7). In other words, this study adds causal attributions to the model of race, actual
performance, and performance ratings previously investiagted by meta-analytic
researchers (e. g., Sackett & DuBois, 1991) in order to explain the cognitive processes
underlying racial effects on performance ratings. It is proposed that causal attributions,
based on an individual’s race and actual performance, moderate the relationship between
actual performance and performance ratings. Internal attributions for success will result
in the highest performance ratings, with external attributions for success resulting in
signiﬁcantly lower ratings. External attributions for failure will result in even lower

ratings and internal attributions will result in the lowest performance ratings.

17

 

Race

 

 

 

 

 

 

Performance Attributions Performance Ratings

\

 

 

 

 

 

Actual Performance

 

 

 

Figure 7. The interaction between actual performance and causal attributions

The above model was constructed using studies that examined ratings by one
individual, usually a supervisor, of another individual, usually a subordinate. However,
this model can also be applied to team member (i.e., subordinate) ratings of a team leader
(i.e., supervisor). This study introduces teams to the literature on race and performance
ratings for three reasons. First, previous laboratory research concentrated on ratings of
“paper” African Americans and White Americans. These stimuli do not offer any
opportunities for the rater to interact with the ratee as employees would within an
organization. However, if rater and ratee are both working together on a collaborative
task (i.e., as part of a team) they must have at least limited contact with each other.
Second, an individual’s performance within an organization inﬂuences other employees,
including supervisors, subordinates, and peers. Lab research has, for the most part,
focused on “paper people,” where negative or positive performance is of no consequence

to the rater. In a team situation, all participants are trying to reach a common goal (e. g., a

18

monetary reward) and therefore one team member’s performance impacts the
performance of the other team members, which could affect the performance evaluation
process. Third, organizations are currently interested in gathering performance feedback
not only from supervisors and peers, but also from subordinates (e.g., 360O feedback). 80,
in order for lab work to simulate the organizational environment, bias in performance
ratings should be investigated from all angles. In other words, racial differences in
performance ratings should be examined using subordinate raters as well as supervisor
raters. Unfortunately research has failed to address this gap within the literature, which is
why this study looks at performance ratings in team contexts.
T earns, Race, Attributions, and Performance Ratings

The past twenty years have evidenced many distinct changes in the workplace,
including a transition from individual to team work, where several employees are
designated a single task to complete together, often with a designated leader (Ilgen,
1994). In order to be successful, employees must achieve their goals as members of a
cohesive group. But, although teams work as a single unit, they are still composed of
individual members, often with varying levels of expertise and ability. Such individual
differences make it difﬁcult to judge the performance of another team member, which, if
inaccurate, could have deleterious effects on team performance. Team members may not
be willing to participate in additional tasks if not satisﬁed with the performance of one of
their members (Phillips, 1997). The question is, how do individuals rate the performance
of other team members, speciﬁcally the team leader, and is their appraisal equally

accurate across team leaders of the same ability?

l9

Little research has been done in order to examine the effects of race on the
performance evaluations of team leaders. However, several studies have looked at racial
differences in the evaluation of managerial performance (e. g., Davidson, Swigert, and
Ruderrnan, 1998; Mount, Sytsma, Hazucha, and Holt, 1997). Davidson et al. (1998)
found that African American managers were rated higher by both White American and
African American subordinates. On the other hand, Mount et al. (1997) found that
African American subordinates rated managers of the same race higher, while White
Americans did not exhibit a racial preference. Unfortunately several issues inhibit the
interpretation of their data to the team context. For one thing, managers were involved in
their subordinates work to varying degrees depending on the project and the organization.
This variance in involvement may have inﬂuenced the level of ratings in the studies by
both Mount et a1. and Davidson et al. Another inﬂuence may have been the actual
performance level of the managers. Unfortunately the researchers in the aforementioned
studies could not control the actual performance level of ratees, which allowed for
different interpretations of the same data. Finally, employees are often sensitive to racial
issues in the workplace and could have adjusted their written responses accordingly.

Research on gender effects has attempted to more directly examine the rating of
team leaders (e.g., Rice et al., 1980; Rosen & Jerdee, 1973), often with conﬂicting
results. Dobbins and Platz (1986) suggested that, in addition to investigating the
evaluation of team leaders by gender or any other subgroup, researchers should
concentrate on the cognitive process behind the evaluation of different team leaders. In
doing so, one general conclusion emerges from the research on the evaluation of team

leaders based on gender: behaviors which are consistent with the rater’s implicit theories

20

are more likely to inﬂuence evaluations than inconsistent behaviors (e.g., Allen &
Ebbesen, 1981; Hastie, 1981; Wyer & Scrull, 1981). In other words, an individual’s
behavior will be evaluated in accordance with pre-existing subgroup stereotypes, which is
the essence of the model proposed in this study.

Although the literature on gender differences mentioned in the previous paragraph
concedes that team member behavior is evaluated according to its stereotype consistency,
attribution theory has not been introduced as a possible explanation. Conversely, studies
that have introduced causal attributions as a possible explanation of team member
evaluation have not examined any subgroup differences (e. g., Norvell & Forsyth, 1984;
Mitchell & Wood, 1980; Martinko & Gardner, 1987). Attribution researchers have noted
“that a comprehensive perspective of the interactive nature of the leader/member
attributional processes is needed” (Martinko & Gardner, 1987, p.235). To create such an
extensive theory, it is likely that attributions of team leader performance will vary across
subgroups. This variance should therefore be included in such models.

Team member ratings of a team leader introduces an important variable, the
fundamental attribution error, which could inﬂuence the ability to translate attributional
information concerning the rating process collected in this study to previous work on
subordinate rating (e.g., Kraiger & Ford, 1985). The fundamental attribution error has
been extensively documented in the literature (e. g. Ickes & Layden, 1978; Zuckerrnan,
1979). Basically, individuals will take credit for success and will blame external causes
for failure. Although the fundamental attribution error is expected to have an effect, it is
not anticipated to change the original model pictured above because it is only likely to

impact the strength of the attributional effects. In other words, when seeking blame for

21

failure, it is expected that individuals will attribute it to internal causes in the case of an
African American team leader and to external causes in the case of a White American
team leader. On the other hand, when taking credit for the success of the team, it is
proposed that individuals will still attribute some of their good fortune to the high
performance of their team leader. If the team leader is African American, success is likely
to be attributed to external causes, whereas a White American team leader will receive
internal attributions as expected. In other words, these effects are predicted to be in the
same direction as effects for raters and ratees who are not interdependent.

It is not just a question of who gets the credit for success or failure (i.e., the
fundamental attribution error), but also how much they get. In other words, the ratings of
the team leader will also be inﬂuenced by the amount of blame or praise given to the
other team members, not just to the team leader. In other words, raters must apportion
blame or praise between all individuals participating in the team. But again, no matter
how it is allocated, some amount of blame or praise will necessarily be given to the team
leader in the attributional pattern previously described. What it will inﬂuence is the level
of the direct effect of race on performance ratings. Although African Americans will still
be rated lower than White Americans, those individuals who feel the other team members
are more responsible than the team leader will give higher ratings to the leader compared
to individuals who feel the outcome is entirely the team leader’s responsibility.
Hypotheses

The previous model, reproduced below, based on dyadic rating research, is

expected to hold in teams where staff members rate their team leader. My hypotheses are

22

discussed in reference to the links in Figure 8 below. The numbers on the lines of the

ﬁgure refer to the corresponding hypotheses.

 

 

 

 

 

 

 

 

 

 

 

 

Race 1
Performance Attributions Performance Ratings_
3&4 5&6\‘
Actual Performance 2

 

 

 

Figure 8. Model of the relationship between race and performance ratings in a team
situation
Hymthesis 1. There is a relationship between race and performance ratings, such that
African American team leaders will receive lower performance ratings than White
American team leaders.
Hymthesis 2. There is a relationship between actual performance and performance
ratings, such that team leaders in the low performance condition will receive lower
performance ratings than team leaders in the high performance condition.

The relationship between race and performance attributions is moderated by
actual performance in the following ways:
Hypothesis 3. The relationship between race (African American = 0, White American =
1) and internal attributions will be positive in the high performance condition and

negative in the low performance condition (see Figure 9).

23

Hypothesis 4. The relationship between race and external attributions will be negative in
the high performance condition and positive in the low performance condition (see Figure
10).

The relationship between actual performance and performance ratings is
moderated by performance attributions in the following way:
Hymthesis 5. The relationship between actual performance and performance ratings will
be more positive for high internal attributions than for low internal attributions (see
Figure 11).
Hymthesis 6. The relationship between actual performance and performance ratings will

be more positive for high internal attributions than for low internal. attributions (see

 

 

Figure 12).
A
High
Internal White
Attributions Race
Black
Low ,
Low High

Actual Performance

Figure 9. Hypothesized graph of the interaction between race and actual performance

24

High i

 

 

Black
External Race
Attributions White
Low ’
Low High

Actual Performance

Figure 10. Second hypothesized graph of the interaction between race and actual
performance

 

 

A
High High Internal
Performance Ratings Low Internal
Low ,
Low High

Actual Performance

Figure 11. Hypothesized graph of the interaction between causal attributions and actual
performance

25

High A

 

 

Low External
Performance Ratings High External
Low >
Low High

Actual Performance

Figure 12. Second hypothesized graph of the interaction between causal attributions and
actual performance

Method

Participants

One hundred and eighty undergraduates at a large midwestem university ‘
participated as team members for three hours in exchange for $20. Each subject was also
informed that top scoring team members would have the opportunity to earn bonus
money on the basis of their performance. Participants were then placed in a four-person
team consisting of three raters and one confederate team leader.
_T___k

The task the students were presented with was the TIDE2 decision making task

(Team Interactive Decision Exercise for Teams Incorporating Distributed Expertise;

Hollenbeck et al., 1995). TIDE2 is a software program for a decision task that presents

26

participants with a number of attribute values relevant to either a problem or an object. A
judgement regarding that object or problem is then formed based on the values of the
attributes. Then decisions are made by the participants regarding the state of that object
(e. g., the severity of an injury, the value of a piece of merchandise, etc.). In this study, the
software program simulated a naval command-and-control scenario with a team leader
and three staff members. The team leader assumed the role of aircraft carrier
commanding ofﬁcer (Alpha), while the staff members assumed the roles of three
commanding ofﬁcers of different air patrol units: an AWACS radar plane (Bravo), an
Aegis cruiser (Charlie), and a Coastal Air Defense (Delta).

All four team members sat at separate computer stations, which were networked
together. Alpha, either the African American or White American confederate, was seated
at a computer station in an adjacent room. This was done for several reasons. First, the
confederate team leader entered team decisions from a script, and it was imperative that
the staff members were not aware that his behavior was scripted. Second, by having the
confederate in another room, his staff members hopefully felt less self-conscious about
assigning blame for failure or selﬁshly taking credit for success. Third, since no verbal
communication was allowed during the game, separating the players most likely helped
to reduce the amount of talking.

The object of the game was to monitor the airspace surrounding the team. Each
team member gathered information regarding particular attributes of an incoming aircraft
that entered the airspace monitored by the team. Bravo was in charge of the aircraft’s

size, speed, radar, and frequency. Charlie was responsible for the aircraft’s direction,

27

speed, angle, and range. And Delta was in charge of the aircraft’s radar, altitude, corridor

status, and range (see Figure 13).

 

 

 

Bravo Charlie Delta

Size Direction Radar

Speed Speed Altitude

Radar Angle Corridor Status
Frequency Range Range

 

 

 

 

Figure 13. Role requirements

Bravo, Charlie, and Delta were all instructed that their area of expertise only
included their four characteristics, while Alpha’s area of expertise included all nine
characteristics. In order to make a decision regarding the threat of the aircraft, each staff
member had to have information on all four of their characteristics. However, on each
target, the staff members could measure four attributes, only one of which was necessary
for their rule, while Alpha could measure all nine (See Figure 14). That meant that each
staff member had to get three characteristics from his or her teammates. As shown in
Figure 15, the staff members had several options to choose from when requesting
characteristics from their teammates. However, the option of requesting information from
Alpha was discouraged, since he had to measure and evaluate all nine attributes and
would not have much time to help his staff members.

After the staff members sent their recommendations to Alpha, he entered the ﬁnal
team decision. All ﬁnal team decisions and staff member recommendations varied in

aggressiveness on a 7-point continuum from Ignore (1) to Defend (7), with defend

28

indicating the highest level of threat. A clock on the screen counted down the time before
the team needed to make a decision. At thirty seconds, the computer began to beep
indicating that time was running out. The staff members were told that Alpha could either
follow the advice of his team members, or he could make decisions on his own. But
instead of combining the three subordinates’ recommendations, Alpha made decisions
based on prearranged script so as to manipulate the performance level of the leader as
seen by the subordinates. Thus, leader behavior was manipulated to create either high or
low performance across the experimental sessions.

After the team leader entered the ﬁnal team decision, a feedback screen appeared
on all four computers comparing the team’s decision to the correct decision. The team
either saw a hit (2 points), a near miss (1 point), a miss (0 points), an incident (-1 points),
or a disaster (-2 points). Team members also saw how each of their individual
recommendations compared to the correct decision. The team’s overall score (an average
of all decisions) was given on the bottom of the screen next to the team goal score. The
goal score was entered based on the average performance of 90 teams from LePine
(1998). All 90 experienced the same target set that was used in this study. The goal score
remained constant across both high and low performing teams. Finally, the feedback
screen included the team’s performance history (i.e., the number of hits, misses, etc.) on

the right hand portion of the screen.

29

 

 

 

Bravo Charlie Delta Alpha
Size Radar Frequency Frequency
Direction Frequency Speed Speed
Range Angle Direction Direction
Corridor Status Corridor Status Altitude Altitude
Radar
Angle
Corridor Status
Size
Range

 

 

 

 

 

Figure 14. Measurable attributes

Bravo

  

Range
Corridor Statu

     

Corridor Status
Charlie ‘ 'Delta
Speed

Direction

 

Figure 15. Role structure

30

Procedure
Participants were recruited via advertisements posted in dorm rooms, classrooms,

and the university newspaper. Then they were scheduled to time slots to create teams and
randomly selected into conditions. Upon arrival to the laboratory, participants (including
the confederate team leader) signed a consent form while they waited for other team
members to arrive. After three subjects arrived, they were taken into a separate room
along with the confederate team leader. They were then instructed to take a seat at one of
the three computer consoles. The confederate team leader walked in last in order to
ensure that the subjects occupied all the computers. It was announced that, since he was
the last person to enter the room, the confederate would be the team leader and would
play the game in another room because verbal communication was not allowed. Once the
confederate was escorted out of the room, the participants were welcomed and were
given a general task overview booklet and an individual role responsibility sheet. The
experimenter then left the room for 8 minutes in order to let them read through the
materials. Upon returning, the experimenter answered any questions before beginning the
“hands-on” portion of the training where the subjects were presented with ﬁve practice
trials.

During the ﬁrst practice trial, the experimenter explained the mechanics involved
in gathering and sharing information about target attribute using the pull down menu on
their computer screens. The mechanics that were included were (a) measuring the
attributes of the aircraft, (b) querying others for attribute values, (c) directly transmitting
attribute values to others, and (d) communicating via sentence-long free form text-

messages.

31

The ﬁrst practice trial also included instruction on individual role responsibilities,
which were provided on the role responsibility sheet. The instruction included (a) the
speciﬁc attributes needed for each role, (b) how to translate raw data on targets into
judgments about how threatening the target is based on one attribute, and (c) how to
combine all three attributes into a recommendation about a rule. It was also explained
that Alpha could measure all nine attributes and could make decisions on each rule
without help from his staff members.

After the staff members made their recommendations regarding the ﬁrst target,
the confederate team leader combined their decisions (only for the ﬁve practice trials)
and made a ﬁnal team decision. When the feedback screen appeared, the experimenter
pointed out (a) how to read the screen (i.e., whether they got a hit, miss, etc.), (b) their
overall score, (c) their goal score, ((1) how their, decisions relate to the actual score, and
(e) their performance history.

The practice trials were also used to practice communicating with one another
through the text mode and other pre-programmed messages in order to get the
information they needed for their role. The experimenter was present to answer any
questions that the subjects might have had. All three participants were also reminded that
any communication must occur over the computer. After the last practice trial, the
experimenter wished the team “good luck” and started the game clock. An “aircraft” soon
appeared in their airspace, and the team began to gather and process the relevant

information.

32

Measures and Manipulations

193:}, The team leader’s race was manipulated by randomly assigning either the
African American or the White American confederate team leader to either the high or
low performance condition. Both team leaders were male and similar in age, height, and
weight.

Actual Performance. The actual performance of the team was manipulated. The
goal score for teams in both the high and low performance conditions was equal to the
average score of the 90 teams from LePine (1998). Individuals in the high performance
condition scored two standard deviations higher than the average, while individuals in the
low performance condition scored two standard deviations lower than the average. The
performance levels were then piloted using 18 subjects so as to have levels that were
believable but also clearly high or low.

Often there are real differences in performance between African American and
White American employees. For example, African Americans tend to score around one
standard deviation lower than White Americans on standard cognitive ability tests
(Jensen, 1980). This issue will not be addressed by this study, since I am interested in
racial bias in ratings and not actual performance differences.

Causal Attributions. In the literature, researchers have used a variety of attribution
measures. Unfortunately many of them have exhibited poor reliability and construct
validity. Based upon Weiner’s (1985) model of performance attributions, an attribution
measure was constructed with items speciﬁcally referenced to the TIDE2 task features
experienced by the participants. Using an 8-item scale, internal attributions were assessed

by asking participants whether their team leader’s behavior was due to his effort or his

33

intelligence in the high performance case, or his lack of effort or lack of intelligence in
the low performance case. External attributions were assessed in the high performance
condition by asking subjects whether the team leader’s behavior was caused by luck or
the ease of the game. In the low performance condition they were assessed by asking
whether the team leader’s behavior was due to bad luck or the difﬁculty of the game.
This left four items which measured internal attributions and four items which measured
external attributions (see Appendix A). Responses were on a ﬁve point, Likert-type scale.
Intercorrelations between the eight items are presented in Table 1. Since the internal
items correlated more highly with themselves rather than with the external items, two
scales of four items each were constructed by summing the four z-scores. Internal
consistency for the internal scale, measured using coefﬁcient alpha, was .81. Alpha for
the external scale was .67. An exploratory factor analysis on the eight items revealed two
factors with eigenvalues greater than 1.0. The four internal items loaded highest on the
ﬁrst factor that explained 35.5% of the variance. The four external items loaded highest
on the second factor and explained 23.1% of the variance (see Table 2). A conﬁrmatory
factor analysis was also completed. The internal and external latent variables had four
indicators each, as shown in Figure 16. The model was tested using AMOS. The ﬁt
indices were as follows: GFI=.91; AGFI=.84; RMSEA=.118. The results indicated a fair
ﬁt of the data to the hypothesized model. Finally, the two scales evidenced good
divergent and convergent validity. The external scale correlated -.18 (p<.05) with the
internal scale and -.18 (p<.05) with three intemality items from Russell’s (1982)
attribution scale. A high score on Russell’s three-item scale would indicate that the

participant believes that the team leader’s behavior was due to internal factors. This was

34

 

.5. v a .1. .3. v a .. .EEonuo .REBENA ow_uZ ”202

 

 

- aim. tom. sewn. 3.3...- semmr *2: tram.- N_._ cod #0 .w
- *2. gnu—N. SEN.- er mu.- N_.- 84 voﬁ mo .5

- 1.3... @9- Nb.- nsvn. we. SA wc.N No .0

- *2: no.- No. cor an: 0N6. 3 .m

- 3.8. the. *smm. vm. Gd E .v

- *Lv. 1.5m. ma. wmﬁ mm .m

- Elm. hm. wad Q .N

- mm. mod 2 A

V0 mo No #0 3 mm mm 3 Gm 53: 03353!

 

gas???

_ 039—.

35

 

was. we

 

 

mwm. no

«me. we

awn. #0
£89552 .5:me “N .803

in. vm

can. 3

$5. Q

S w. :
822552 12:85 3 ~88”—
wﬁgoq cocoa Eu: Baum

 

 

 

mac: =on=n§< .330 m we av. :22 88am Eat uo>tun more 030 .583

N 033.

36

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

i1
i2
i3 1
i4
ei
ea 1
e4

 

 

 

Figure 16. Conﬁrmatory factor analysis of 8 attribution items

37

supported by the fact that the internal scale constructed in this study correlated .29
(p<.01) with Russell’s attribution scale. Overall, the data support the use of a separate
internal and external scales.

Performance Ratings. A 19-item performance measure was constructed in order to
evaluate the team leader’s performance on TIDEZ. It included items such as, “Alpha was
a good team leader” and “Our team leader performed very well on this task.” Responses
were on a ﬁve point, Likert-type scale (see Appendix B). Coefﬁcient alpha for the scale
was .95. Two additional items asked the subordinates, on a scale of l to 100, how good
their team leader was (see Appendix C). Since the two items scaled from 1 to 100
correlated highly with each other and the 19 item scale (see Table 3), these two plus the
sum of the 19 were transformed into z-scores and combined into one leader performance
rating scale.

Table 3

Intercorrelation Between 19 Itpm Performance Rating Scale and 2 Rating Items Scored 1
to 100

 

 

Variable Mean so i1 i2 i3
1- Pr19 58.53 15.32 -—
2. prl 63.43 24.92 .846M --
8- M 63.89 25.93 .815" .902M --

 

Note: N=l 80. pr19=19 item performance rating scale. prl=ﬁrst rating item scaled l to
100.
pr2=second rating item sealed 1 to 100. * p < .05. ** p < .01.

Results
Table 4 provides the means, standard deviations, and intercorrelations among all

the variables used in testing the hypotheses. Table 4 also includes two variables,

38

subordinate race and subordinate performance, which were not part of the hypotheses but
were used in supplemental analyses.

The analyses, as described in detail below, were completed in four distinct steps. First,
data gathered as manipulation checks are presented for the team leader’s race and the
team’s actual performance. Second, the analyses used to test each of the hypotheses
presented earlier are provided. Third, subordinate race and subordinate performance are
added to the analyses to test for possible confounds. Fourth, the results from additional
analyses that were performed in order to test the ﬁt of alternative models are presented.
Manipulation Checks

hep. The team leader’s race was manipulated depending on whether the subjects
were randomly selected into the African American or White American team leader
condition. Upon completion of the experiment, subjects were given a two-item
questionnaire asking whether their team leader was male or female and African American
or White American. Ninety-nine percent of the subjects correctly identiﬁed their team
leader as an African American or White American male, indicating that the manipulation
was effective. Those subjects who incorrectly identiﬁed the team leader were included in
the analyses to be conservative.

Team Performance. The team’s performance was also manipulated depending on
whether the team was randomly selected into the low performance or high performance
condition. At the end of the experiment, subjects completed a one-item questionnaire
asking whether the team performed well or not well. Ninety-ﬁve percent of the subjects

correctly identiﬁed their performance as high or low, indicating that the performance

39

dog—Boron :3: c8 w 98 cognac-tom 32 c8 o R68 83 Sggotom :83 383».

 

.393. snow—29% BE? 05 com _ Ea Bunc— amotoca Swot-.2 05 Em C v38 83 3am 530-— .3. v a .1. .8. v a .. .ow_uZ “202

 

 

-- .2 .- 8. S.- 3. 8. E. SE 8.8 aaaaéotaa 226.8% s
-- 8.- 3. 3.- 8. 8. am. 2 8am saseoaam s
- ten.- ..2. 8. :2. as as mass... 85:83.. .m
-- a2.- 8- ram.- mwa as 82.332 seesaw a.
- a:- ......a.. a; as 22335... 3525 .m
-- 8. on. on. 83. .233 .N
.i OW. CW. COED-tom Each.- Rﬂuo< A
a a n a m N _ cm 5a: ass-5

 

 

1E<=< 5:2 mucus—8.8285 28 2.2350 235% 232

 

v 033.

4O

manipulation was effective. Again the conservative position was taken by including all
subjects in the analyses.
Tests of Hymtheses

Hywthesis l. The ﬁrst hypothesis predicted that race would be related to
performance ratings, such that African Americans would receive lower ratings than
White Americans. Examining the correlation matrix presented in Table 4 shows that race
was uncorrelated (r = .00, p=ns) with performance ratings. Hypothesis 1 was not
supported.

Hypothesis 2. The second hypothesis asserted that leader performance ratings
would be related to the team’s actual performance, such that team leaders in the low
performance condition would be rated lower than team leaders in the high performance
condition. Table 4 shows a correlation of .76 (p<.01) between actual performance and
leader performance ratings. Clearly subordinates rated team leaders in line with the
performance information they received on them. Hypothesis 2 was strongly supported.

Hyp_othesis 3. The third hypothesis predicted that the relationship between race
and internal attributions would be moderated by the team’s actual performance. To test
this hypothesis, a two-step hierarchical regression was performed entering race and actual
performance in the ﬁrst step, and their interaction in the second. Actual performance did
have a direct effect on internal attributions (13:.56, p<.01), such that subordinates in the
high performance condition attributed more of their team leader’s performance to internal
attributions. However, team leader race was not directly related to internal attributions
(B=-. 10, p<.26) and the interaction between team leader race and actual performance

explained an insigniﬁcant portion of the variance in internal attributions (AR2=.01, p=ns).

41

Thus, Hypothesis 3 was not supported. The results of the regression analysis are
presented in Table 5.
Table 5

Result of Reggessing Internal Attribution; on Leader Race. Actual Team Perform’ance.
and Their Interaction.

 

 

Variable B R2 AR2
Step 1
Leader Race -.10
Actual Team Performance .56** .27**
Step 2
Leader Race X Actual Team --.13 .27** .01
Performance

 

M; N=180. B is the standardized regression coefﬁcient. Increments for variables
entered at the R2 signiﬁcance levels are based on F tests for that step. * p < .05. ** p <
.01. Hymthesis 4. Hypothesis 4 proposed that the relationship between race and
external attributions would be negative in the high performance condition and positive in
the low performance condition. Once again, a two-step hierarchical regression was
performed entering race and actual performance in the ﬁrst step, and their interaction in
the second. Actual performance did have a direct effect on external attributions (B: -.73,
p<.01). As performance increased, subordinates attributed less of their team leader’s
behavior to external attributions. In addition, race was directly related to external
attributions (B=-. 17, p<.05). Speciﬁcally, subordinates were more likely to attribute the
leader’s performance to external causes if the team leader was African American. Finally,
the interaction between race and actual performance explained a signiﬁcant portion of the

variance in external attributions (AR2=.03, p<.01). The results of the regression analysis

are presented in Table 6.

42

Table 6

Result of Reggessing External Attributions on Leader Race. Actual Team Performpnce.
and Their Interaction.

 

 

Variable B R7 AR2
Step 1
Leader Race -. 17*
Actual Team Performance -.73** .32**
Step 2
Leader Race X Actual Team .28 .35** .03**
Performance

 

& N=180. B is the standardized regression coefﬁcient. Increments for variables
earltered at the R2 signiﬁcance levels are based on F tests for that step. * p < .05. ** p <
However, although the interaction was signiﬁcant, Hypothesis 4 was not supported. As
illustrated in Figure 17, as performance increases, the slope of the African Americans’
regression line was signiﬁcantly more negative than the line for White Americans. In
other words, as performance increased, subordinates were more likely to attribute their
team leader’s performance to external factors if he were White American.

Hymthesis 5. Although it was expected (and observed) that actual team
performance would have a major impact on leader ratings, Hypothesis 5 predicted that
the size of the impact would vary depending upon attributions for leader performance.
Speciﬁcally this hypothesis predicted that the effects of actual performance on
performance ratings would be moderated by internal attributions. Again, a two-step
hierarchical regression was performed entering internal attributions and actual

performance in the ﬁrst step, and their interaction in the second. As previously noted,

actual performance did have a direct effect on performance ratings (B=.84, p<.01).

43

gunﬁre: 352$ 29 uogobum :32 3:83 ES 83 @303th 539325 .2 059m

 

 

coca—Eaton. Eco... .mBo<
tea :9: ten. 26..

9E; l I I“
cosmic

 

suolrnqlrnv lame-1x3

 

 

44

Furthermore, internal attributions were directly related to performance ratings (B=-.53,
p<.Ol), such that, as actual performance increased, subordinates were less likely to highly
rate their team leader if they attributed his behavior to internal causes. Finally, the
interaction between internal attributions and actual performance explained a signiﬁcant
portion of the variance in performance ratings (AR2=.O9, p<.01). The results of the
regression analysis are presented in Table 7.

Table 7

Result of Regressing Leader Performance Ratings on Actual Team Performance, Internal
Attributions and Their Interaction.

 

 

 

Variable B R2 AR2
Step 1
Actual Team Performance .84"
Internal Attributions -.53** .63"
Step 2 .
Internal Attributions X Actual .41 .72" .09"

Team Performance
Note: N=180. B is the standardized regression coefﬁcient. Increments for variables

egltered at the R2 signiﬁcance levels are based on F tests for that step. * p < .05. ** p <
Figure 18 shows that, as performance increases, the slope of the regression line for high
internal attributions was signiﬁcantly more positive. But for low internal attributions,
performance ratings were unrelated to actual performance. Thus, Hypothesis 5 was
supported.

mm. Hypothesis 6 suggested that the effects of actual performance on

performance ratings would be moderated by external attributions. As in the previous

hypotheses, a two-step hierarchical regression was performed entering external

45

$2.23 oo§§§b2~ x333 no wogobum 3:83 33 neocaﬁxuo ~322wa $5585 .w_ 233m

 

_c
_

 

Eﬂioullllll
E. ;m__._1.ll.

oocueﬁoton. E3... 333‘
Ed :9...

tom 30..

 

sﬁunea
aouewiouad iapeen

 

 

46

attributions and actual performance in the ﬁrst step, and their interaction in the second.
Actual performance again had a direct effect on performance ratings (B=.64, p<.01). In
addition, external attributions had a direct effect on performance ratings (B=-.27, p<.01),
such that subordinates were less likely to rate their team leader higher as performance
increased if they attributed his behavior to external causes. However, the interaction
between external attributions and actual performance did not explain a signiﬁcant portion
of the variance in performance ratings (AR2=.00, ns). The results of the regression
analysis are presented in Table 8. Thus, Hypothesis 6 was not supported.

Table 8
Result of Regressing Leader Pert'grmang Ratings on Actual Team Performance, External

Attributions and Their Interaction.

 

 

 

Variable B . R’ AR2
Step 1
Actual Team Performance .64"
External Attributions -.27** .60"
Step 2
External Attributions X Actual .08 .60" .00

Team Performance

Note: N=180. B is the standardized regression coefﬁcient. Increments for variables
entered at the R2 signiﬁcance levels are based on F tests for that step. * p < .05. ** p <
.01.

Summg of Hymtheses

In conclusion, only two of six hypotheses were supported by the data. One
possible explanation could have been the inﬂuence of extraneous variables. The
subordinate’s race and the subordinate’s level of performance could have had an effect on
the results presented above. In particular, the subordinate’s race could have effected the

relationship between leader race and performance ratings. It is reasonable to suggest that

47

African Americans rate team leaders of their own race differently than if they were part
of the majority. It is also reasonable to suggest that the effect of race on causal
attributions will depend not only on the team’s actual performance level, but also on the
ratee’s race in relation to that of the rater. African Americans may attribute behavior to
different causal loci than White Americans when the team leader is a minority.

In addition, since both the rater and the ratee were doing similar tasks, it is
possible that the raters’ own performance may have inﬂuenced their responses on the
attribution and performance rating measures. For example, someone who performed very
poorly on the game could have given the team leader positive ratings even if they were in
the low performance condition. In addition, someone who had difﬁculty with TIDE2
might be more likely to attribute low performance to external causes than someone who
did very well at the game. As a result, the subordinate’s performance level and the
subordinate’s race were entered into the supplemental analyses carried out on the data.
Supplemental Analyses

Subordinate Race. It was thought that subordinate race might inﬂuence leader
performance ratings. However, due to the small sample of African American
subordinates (n=15), the data lacked sufﬁcient power to test this well. In fact, race did not
relate to performance ratings (r=-.05, ns). It also did not inﬂuence the relationship
between the race of the leader and the leader’s performance ratings (B=-.35, ns). Finally,
the subordinate’s race did not interact with the leader’s race and actual performance to
effect either external (B=.35, ns) or internal (B=.23, ns) attributions.

Subordinate Performance. The second variable, the subordinate’s level of

performance, was explored as it related to performance ratings. Again, there was no

48

evidence of a direct effect (r=.07, ns). And the subordinate’s performance did not
inﬂuence the relationship between the team’s actual performance and performance
ratings (AR2=.003, ns). Lastly, the subordinate’s level of performance did not interact
with the leader’s race and actual performance to impact either external (B=.04, ns) or
internal (B=-.07, ns) attributions.

To sum up, the supplemental analyses of the subordinate’s race and performance
level did not help to clarify the previous results. However, the analyses were all based on
the model presented in the introduction (see Figure 7). It could be that an alternative
model explains the data much better. As described in the introduction, many of the
studies examining the link between race and performance ratings have found conﬂicting
results (e. g., Kraiger & Ford, 1985; Bass & Turner, 1973). However, none of the
aforementioned researchers attempted to collect laboratory data where the performance of
the ratee could be manipulated. Therefore, their results are subject to actual differences in
performance, which could effect the relationship of race and performance ratings (e. g.,
Kraiger & Ford, 1985).

Post Hoc Analyses

To investigate whether the relationship between leader race and performance
ratings depended on the team’s actual performance, a two-step hierarchical regression
was set up entering leader race and actual performance in the ﬁrst step, and their
interaction in the second. As stated in earlier analyses, actual performance effected
performance ratings (B=.86, p<.01) and leader race did not effect performance ratings

(B=.10, ns). However, the interaction between leader race and actual performance did

49

explain a signiﬁcant portion of variance in performance ratings (AR2=.01, p<.05). The
results of the regression analyses are presented in Table 9.
Table 9

Result of Regressing Leader Performance Ratings on Leaderﬁce, Actual Team
Performance and Their Interaction.

 

 

 

Variable B R2 AR2
Step 1
Actual Team Performance .86"
Leader Race .10 .57**
Step 2
Leader Race X Actual Team -.18* .58“ .01*
Performance

 

1193;: N=180. B is the standardized regression coefﬁcient. Increments for variables
earltered at the R2 signiﬁcance levels are based on F tests for that step. * p < .05. ** p <
However, the interaction was not as expected. As shown in Figure 19, African American
team leaders’ ratings were higher than White American team leaders’ ratings under high
performance. In the low performance condition, African Americans were rated slightly
lower than White Americans were. In the high performance condition, on the other hand,
African Americans were rated much higher than White Americans were.

The next step in the post hoc analyses involved the addition of causal attributions
to the model. It is possible that African Americans were rated higher in the high
performance condition because their behavior was attributed internally, and this in turn
could have contributed to their level of performance ratings. So internal attributions as
mediators of the relationship between race and performance ratings were explored in the

high performance condition. In order to test for mediation, it is necessary to demonstrate

that (a) both the independent (leader race) and the mediating (internal attributions)

50

$53» 9385.593 .833 .8 mogobom E38 35.3» 38 83 Love». Kc =o.5~2&=~ .2 oSmE

 

 

oucuELoton. E3... .932
:9. :9... tea 25..

emiﬂlbll
.. .ewm Ill

sﬁunea
aoueuuouad lapea1

 

 

5|

 

variables relate to the dependent variable (performance ratings), (b) the independent
variable relates to the mediating variables, (c) the relationship between the independent
variable and the dependent variable becomes negligible or is reduced signiﬁcantly when
controlling for the mediating variables, and (d) the relationship between the mediator
variables and the dependent variable is still signiﬁcant when controlling for the
independent variable (Baron & Kenny, 1986; James & Brett, 1984).

Both race (r=-.23, p<.05) and internal attributions (r=.32,p<01) were related to
performance ratings, which satisﬁed the ﬁrst condition of mediation (see Table 10). Race
was also related to internal attributions (r=-.36, p<.01), which satisﬁed the second
condition (see Table 10).

Table 10

Means, Standard Deviations, and Intercorrelations Among All Variables Analyzed Post
Hoc in the High Performa_n_ce gondition

 

 

Variable Mean SD 1 2 3 4
1. Leader Race .50 .50 --
2. Internal Attributions 1.54 2.30 -.36** ~-
3. External Attributions -l.60 2.21 .19 -.07 --

4. Leader Performance Ratings 2.15 1.22 -.23* .32** -.26* --

 

lime; N=90. * p < .05. ** p < .01.

Hierarchical regression was employed to test the remaining conditions. As shown in
Table 11, when internal attributions were entered in the ﬁrst step, leader race was no
longer signiﬁcantly related to performance ratings, which passes the third mediation
hurdle. Finally, the fourth condition was also satisﬁed because the relationship between
internal attributions and performance ratings remained signiﬁcant even when controlling

for leader race (see Table 12). So, in the high performance condition, internal attributions

52

Table l 1

Result of Regressing Leader Performance Ratings on Leader Race while Controlling for
Internal Attributions in the High Performance Condition

 

 

 

Variable B R2 ARZ
Step 1
Internal Attributions .27* .10**
Step 2
Leader Race -.14 .12” .02

 

Note: N=90. B is the standardized regression coefﬁcient. Increments for variables entered
at the R2 signiﬁcance levels are based on F tests for that step. * p < .05. ** p < .01.

Table 12

Result of Reg; ssing Leader Performance Ratings on Internal Attributions while
Controlling for Leader Race in the High Performance Condition

 

 

Variable B R2 AR2
Step 1
Leader Race -.14 .05*
Step 2
Internal Attributions .27** .12" .06*

 

m N=90. B is the standardized regression coefﬁcient. Increments for variables entered
at the R2 signiﬁcance levels are based on F tests for that step. * p < .05. ** p < .01.

53

mediated the relationship between leader race and performance ratings as initially
predicted across all levels of performance. Subordinates were more likely to attribute the
performance of the African American team leader to internal attributes and this in turn
increased their performance ratings.

It is also possible that White Americans could have been rated lower in the high
performance condition because their behavior was attributed externally, and this in turn
contributed to their level of performance ratings. So it was further hypothesized that
external attributions mediated the relationship between race and performance ratings in
the high performance condition. As previously noted, leader race was related to
performance ratings. And since external attributions were also related to the dependent
variable (r=-.259, p<.05), the ﬁrst condition was satisﬁed (see Table 10). The second
condition required external attributions to correlate with the leader’s race (1:.193, p<.07),
which passed (see Table 10). Hierarchical regression was again used for the remaining
steps. When external attributions were entered in the ﬁrst step, leader race was no longer
signiﬁcantly related to performance ratings, which satisﬁed the third condition (see Table
13). The data also passed the fourth and ﬁnal condition, since the relationship between
external attributions and performance ratings remained signiﬁcant even when controlling
for leader race (see Table 14). Therefore, in the high performance condition, external
attributions mediated the relationship between leader race and performance ratings.
Subordinates were more likely to attribute the performance of the White American team

leader to external attributes and this in turn decreased their performance ratings.

54

Table 13

Result of Reggessing Leader Performance Ratings on Leader Race while Controlling for
External Attributions in the High Performance Condition

 

 

 

Variable B R2 AR2
Step 1
External Attributions -.22 .07*
Step 2
Leader Race -. 19 .10” .04

 

Note: N=90. B is the standardized regression coefﬁcient. Increments for variables entered
at the R2 signiﬁcance levels are based on F tests for that step. * p < .05. ** p < .01.

Table 14

Result of Remssing Leader Performance Ratings on External Attributions while
Controlling for Leader Race in the High Performance Condition

 

 

Variable . B R2 AR’
Step 1
Leader Race -. 19 .05*
Step 2
External Attributions .22* .12” .05*

 

m N=90. B is the standardized regression coefﬁcient. Increments for variables entered
at the R2 signiﬁcance levels are based on F tests for that step. * p < .05. ** p < .01.

55

In the low performance condition, the data did not support the mediation
predictions. Since leader race was not signiﬁcantly related to performance ratings
(r=. 126, ns), the ﬁrst condition of mediation was not satisﬁed. As a result, further
analyses on subsequent steps would have been meaningless.

Discussion

The main goal of this study was to examine the relationship between race and
performance ratings from a different perspective than has been the norm in the literature.
Most research has focused on whether African American and White American
subordinates are evaluated equally by their superiors (e. g., Kraiger & Ford, 1985; Sackett
& DuBois, 1991). Overall, it appears that African American subordinates are rated lower
than their White counterparts (Sackett & DuBois, 1991).

However, within the organization there are alternative sources of feedback that
are now being used instead of, or supplemental to, the standard performance appraisal by
a superior (e. g., Miles & Snow, 1995). Researchers have found that subordinates and
peers also have critical information to offer an employee regarding his or her
performance (e.g., Borman, 1974).

In addition, there has also been a transition within the workplace from individual
to team work, where a task is completed by a group of individuals, often with a
designated team leader (Ilgen, 1994). In order to evaluate the effectiveness of a team
leader, it is essential that as much information as possible is collected regarding his or her
effectiveness, including performance ratings from the members of the team.

Finally, most of the literature is based on data for which there is little or no ability

to assess the impact of the performer’s actual performance on ratings. Since this study

56

was able to control the level of performance of the team leader, it offered an advantage
by examining ratings given to a very poor versus very successful team leader.
Researchers have noted that, in order to better estimate the population correlation
between race and performance ratings, it is necessary to separate out the ratee’s actual
level of performance (e.g., Pulakos et al., 1989).

Since the literature offered little empirical work regarding the relationship
between team leader race and subordinate performance ratings, it was assumed that race
would effect ratings in a similar fashion to Sackett and DuBois (1991). It was expected
that African American team leaders would get rated lower than their White counterparts.
However, despite the control offered by the lab, this study found team leader race was
unrelated to performance ratings across all performance conditions (r==.00, p=ns). Clearly
the results described by Sackett & DuBois (1991) did not translate to this team
environment where subordinates evaluated the performance of their team leader.

The second goal of this study was to investigate the role of performance
attributions in possible cognitive biases present in the evaluation of African American
and White American team leaders. It has been suggested that attribution theory (W einer,
1985) could help explain why African American subordinates are rated lower than White
American subordinates when their performance is evaluated by a superior within the
workplace (Ilgen & Youtz, 1986). Therefore it was hypothesized that attribution theory
may also inﬂuence the relationship between race and performance ratings when the ratee
is the superior and not the subordinate. It was expected that African Americans would be
rated lower in the low performance condition because their behavior would be attributed

internally, while they would be rated lower in the high performance condition because

57

their behavior would be attributed externally. However, no relationship was found
between team leader race and the evaluation of his performance. This negated any further
investigation of the relationship using attribution theory. Our model did not ﬁt the data.

There are several possible explanations why the model presented in the
introduction was not supported. First, it could be that the effects of race on performance
ratings wane when raters and ratees work collectively on a designated task. By interacting
with one another for an extended period of time, the leader’s race could have lost its
impact when each team member rated their leader’s performance. But if the team
members did not evaluate their leader on the basis of his race, then their behavior should
be attributed to similar causes. For example, poor performance should be attributed to
external factors for both African American and White American team leaders. However,
the data regarding external attributions did not support such an assertion. In the low
performance condition, African Americans were more likely to have their behavior
attributed to external causes. And, in the high performance condition, White Americans
were more likely to have their behavior attributed to external causes-

The data regarding internal attributions were a little more encouraging, since race
did not have an impact on whether the leader’s behavior was attributed internally or
externally. And when the team leader’s poor performance was highly attributed to
internal causes, he was rated much lower. Although that difference dissipated when the
team was successful, the team leader still received lower ratings when his behavior was
highly attributed to internal causes.

In conclusion, causal attributions offered little explanation as to the reason why

African Americans and White Americans were rated similarly. African American and

58

White American team leaders did not always receive similar causal attributions for their
performance and those attributions did not effect performance ratings as expected.

A second explanation of why the hypothesized model was not supported involves
the level of analytical power available in this study. It could be that a sample size of 180
subjects was simply too small to obtain a signiﬁcant relationship between race and
performance ratings. Much of the latest research on race and performance ratings has
relied on the technique of meta-analysis, which combines the data from a variety of
studies completed in the ﬁeld and in the laboratory in order to come up with a better
estimation of the population correlation (Hunter, Schmidt, & Jackson, 1982). By doing
so, researchers have been able to analyze the relationship between race and performance
ratings using thousands of subjects (e.g., Sackett & DuBois, 1991; Kraiger & Ford,
1985). This study, on the other hand, dealt with only a small sample of undergraduate
students and that could have affected my ability to ﬁnd small effect sizes. However, in
this study the relationship between race and performance ratings was almost nil, which
means that a substantial raise in power would probably not have helped the correlation
reach signiﬁcance.

Since power was not likely an issue, a third possible explanation was the
existence of a moderator. In this study, one potential moderator of the relationship
between race and performance ratings was the actual performance of the team. As stated
earlier, laboratory research offers the advantage of increased control. By using ﬁeld data,
researchers lose the ability to control and manipulate variables. Much of the literature on
the relationship between race and performance ratings has dealt with samples gathered

from the ﬁeld. For example, in Kraiger and Ford’s (1985) meta-analysis, they note that

59

their results do “not preclude the possibility that actual performance differences between
races existed” (pp. 62). For many of the early studies, it was impossible to determine
whether African Americans were rated differently due to bias or due to the fact that they
actually performed their job less effectively. As a result, some of the studies found a no
relationship between race and performance ratings, while others found African
Americans and White Americans rated individuals of the same race higher (e. g., Hamner
et al., 1974; Bass & Turner, 1973). Pulakos et al. (1989) proposed that any bias in ratings
given to African American and White American employees had to be separated from
their actual level of performance. In fact, researchers have shown that the relationship
between race and performance ratings depends on the actual performance of the ratee
(Bigoness, 1976; Rotter & Rotter, 1969). Following along those same lines, this study
proposed post hoc that the absence of a statistically signiﬁcant relationship between race
and performance ratings may have been due to the moderating effect of ratee
performance.

A new model was tested investigating whether the actual performance of the team
leader moderated the relationship between leader race and performance ratings. As can be
seen in Figure 19, there was a signiﬁcant interaction between the leader’s race and his
actual performance. The African American team leader was rated higher than the White
American team leader in the high performance condition and lower in the low
performance condition. However, in the low performance condition, the relationship
between leader race and performance ratings did not reach signiﬁcance (r=.13, p=ns).
Based on past research (e. g., Sackett & DuBois, 1991), it was expected that African

Americans would be rated lower due to the stereotypes overarching the African American

60

community (e.g., Smith, 1990). Speciﬁcally, stereotypes for African Americans are of
less ambition and lower intelligence than White Americans. This View of African
Americans creates expectations concerning their behavior, and if an African American
performs poorly on a certain task, the behavior is consistent with the stereotype. Yet in
this case both team leaders were rated almost equally, with the African American team
leader receiving slightly lower ratings.

One possible explanation for this could be a lack of power. By splitting the
sample into only the low performance condition, the power to ﬁnd a signiﬁcant
relationship dropped by ﬁfty percent. Perhaps if the number of subjects doubled, the
relationship would have been signiﬁcant.

Another explanation might be the students’ inclination to give socially desirable
responses (Jones & Signal], 1971), thereby ameliorating the amount of blame given to the
African American team leaders. Most people are aware of the stereotypes assigned to
African Americans, and, in an effort to stay away from such negative responses, the staff
members could have unconsciously or consciously raised their ratings of the African
American team leader. And since the consequences for poor performance were relatively
minor (i.e., everyone got paid $20 no matter how well they performed), it may have been
easier to tone down the ratings given to the African American team leader.

The reverse may also have been true. The stereotypes of high intelligence and
effort given to White culture could have severely impacted the performance ratings given
to the White team leader after he mishandled target after target on TIDEZ. It was
proposed that White American team leaders would receive the beneﬁt of the doubt in the

low performance condition, and their performance would be attributed to bad luck or the

61

difﬁculty of the game. In that way, their performance ratings would not suffer as much as
the African American team leaders. But in this case, the staff members may have found
the White American team leader’s poor performance on such an easy task so salient that
the only explanation was complete incompetence.

Although African Americans were only rated slightly lower than Whites in the
low performance condition, race was signiﬁcantly correlated with performance ratings
(r=-.233, p<.05) in the high performance condition. African Americans were rated higher
than White Americans when they led their team to success. Originally it was proposed
that the stereotypes surrounding African Americans would prevent them from receiving
any credit for success while playing TIDEZ. The African American team leaders’ success
would simply be blamed on luck or the ease of the game. As a result, by trivializing the
excellent performance of the African American team leader, the team members would
have enough cause to effectively lower their performance ratings. However, in this study,
the performance of the African American team leader was not trivialized, but was
applauded.

Nevertheless, it is still possible that the team members did not expect their
African American team leader to do so well on TIDEZ, since it contradicted their
stereotypes of African American culture. Instead of trying to suppress conﬂicting
information by attributing the behavior to external causes, the team members may have
resolved the conﬂict by viewing his performance as an exception to the rule. As a result,
the subordinates might have attributed the success of the African American team leader
to internal factors such as intelligence and effort, and that in turn bolstered their

performance ratings. In fact, the results of the mediation analyses supported this

62

assertion. Internal attributions mediated the relationship between race and performance
ratings, such that race was negatively correlated with internal attributions and positively
correlated with performance ratings. Therefore, African Americans received internal
attributions for their behavior in the high performance condition, and that resulted in
higher performance ratings.

If the subordinates were expecting their African American team leader to fail to a
greater extent than those with a White team leader, then it is also possible that they were
expecting the White American team leader to succeed. If so, in the high performance
condition, the White American team leader does just as expected by hitting target after
target on TIDEZ. But fulﬁlling expectations does not usually reap the same reward as
surpassing expectations. Maybe the subordinates felt that their White American team
leader did not accomplish anything remarkable during the game and thus did not deserve
to receive all the credit for his performance. The results do, in fact, support this
contention, since the successful performance of the White team leader was more likely to
be attributed to external causes such as luck or the difﬁculty of the game. And the
directions of the correlations imply that that may be why the White American team leader
was rated lower in the high performance condition.

Theoretical Implications

The results of this study have both theoretical and practical implications.
Theoretically, it is clear that male ratees’ level of performance inﬂuences the relationship
between their race and their performance ratings, as other researchers have suggested
(e.g., Pulakos, 1989). This helps to explain why some researchers investigating the

ratings of managers by their subordinates ﬁnd conﬂicting results. For example, Davidson

63

et al. (1998) found that African American managers were rated higher than White
American managers, while Mount et al. (1997) found that White subordinates rated
African American and White American managers similarly. It could have been the actual
performance of the managers that inﬂuenced the differential effects.

But even if the actual performance of the ratee moderates the relationship between
race and performance ratings, it is clear that it depends on the rating situation. For
example, several studies have shown that, when performance is low, African Americans
are rated higher than White Americans are, and, when performance is high, there is no
relationship between race and performance ratings (Bigoness, 1976; Rotter & Rotter,
1969). In addition, other studies controlling for the actual performance of the ratee have
found that African Americans are actually rated lower than White Americans (e. g.,
Sackett & DuBois, 1991). No studies have shown that African Americans are rated
higher than White Americans when they are successful and insigniﬁcantly lower when
they are unsuccessful. However, all of the studies mentioned above investigated the
relationship between subordinate race and their superiors’ ratings of their performance.
This study differed in two respects: the superior was the ratee and not the rater, and the
rater and ratee collaborated on a team task with distributed expertise. Perhaps the ratee’s
leadership position and interaction with the raters inﬂuenced the moderated relationship
between race and performance ratings.

The results of this study also add to the literature by helping to uncover the
cognitive processes underlying the ratings of African American and White American
team leaders by their subordinates. Kraiger and Ford (1985) and Dobbins and Platz

(1986) suggested that researchers should not simply study whether there is a relationship

between race and performance ratings. They should also investigate why there is a
relationship between race and performance ratings. Most of the theories that have
subsequently been proposed were not supported by the results of this study. Tokenism,
the similar-to—me effect, positivity bias, social identity theory, and intergroup theory all
say pretty much the same thing- that African Americans will be rated lower than White
Americans because individuals will rate their own race higher. But the results presented
here suggest that the cognitive processes involved in rating ones team leader are more
involved than simply identifying him as the same race.

In this study, the race of the team leader did not directly affect performance
ratings. Instead, the race of the team leader impacted causal attributions, which then
inﬂuenced performance ratings. As Ilgen and Youtz (1986) hypothesized, the
performance of minorities was attributed to different causes than the performance of the
majority. But 1] gen and Youtz also thought that good performance would be attributed to
external factors for minorities and internal factors for the majority. This was not the case
here. Good performance was attributed internally for African American, and not White
American, team leaders. However, those attributions did inﬂuence performance ratings in
the expected direction. Higher internal attributions were related to higher performance
ratings and higher external attributions were related to lower performance ratings. The
results suggest that African American team leaders are rated differently not simply

because the color of their skin is different, but because their performance is attributed to
different factors than the performance of White Americans when they are successful.

The results of this study also cast doubt on the applicability of gender research

regarding the relationship between team leader gender and performance ratings. A

65

number of studies have shown that behaviors consistent with a certain gender stereotype
are more likely to inﬂuence evaluations than behavior that are inconsistent (e.g., Allen &
Ebbesen, 1981; Hastie, 1981). Race, in this study, did not operate in a similar fashion. It
was the inconsistent behavior of the African American team leader that really inﬂuenced
his performance ratings. Perhaps consistent and inconsistent behavior elicit different
emotional reactions when race, and not gender, is the salient subgroup being stereotyped.
Practical Implications

Aside from the theoretical implications, there are also some practical implications
that result from the ﬁndings presented in this study. For one thing, it is clear that bias in
ratings does not remain constant across ratees. Organizations are not simply using
subjective ratings of subordinates by their superiors as performance barometers. The
introduction of 360—degree feedback presents the relationship between race and
performance ratings from a variety of angles (e. g., Borman, 1974). It is possible that,
when minorities are rated by subordinates, the literature regarding the relationship
between race and performance ratings is overestimating the bias present in their
evaluations. Bias against minorities, as suggested in this study, may only present a
problem if those minorities perform poorly.

Another practical implication is that there simply is no practical application at all.
Like many of the studies in the literature, the effect sizes presented here are incredibly
small (e.g., Pulakos et al., 1989). In fact, the actual performance of the team leader
explained much of the variance in performance ratings. Race only had a minor impact on

the relationship and effects were only found in the high performance case. It is likely that,

66

if an organization gathers developmental feedback from a variety of individuals
surrounding the ratee, the effects of race on ratings will probably be non-existent.
Study Limitations

This brings me to the limitations of my study, many of which could have
inﬂuenced the results or lack thereof. As noted above, this study was completed in a
laboratory using undergraduate students and not employees. Although every attempt was
made to simulate an actual workplace, there were clear differences that could have
effected the data. For one thing, the consequences of the students’ evaluation of their
team leader were not equal to the consequences present in the workplace. The students
knew that they were not negatively or positively inﬂuencing their leader's life, since he
could not get ﬁred, promoted, or demoted. In addition, although they worked together,
the team members got paid no matter how well they performed. The bonus of $20 to the
top performing subordinates probably did not have the same effect on their emotions as
the possibility of getting ﬁred would. Furthermore, the students only worked together for
about two hours and only ﬁve minutes of that time were spent face-to—face. In the
workplace, subordinates would probably work much longer and more interactively with
their team leader before ratings his or her performance. Finally, as mentioned earlier, the
performance of the team leader was manipulated in this study such that the team
performed very well or very poorly. It is likely that the performance of team leaders
within the workplace is actually much more moderate than what was created here;
certainly the distribution would not be dichotomous. As a result, the strength of the

manipulation could have inﬂuenced the causal attributions ascribed to the leader’s

67

behavior. For instance, more moderate levels of performance ﬂuctuation could elicit
attributions in accordance with the literature on African American stereotypes.

Another limitation of this study has to do with the measurement of causal
attributions. Causal attributions are one of the more difﬁcult constructs to measure, which
is clearly evidenced by the lack of established measures in the literature. For this study, I
attempted to construct a measure tailored to the task of TIDEz. Although the internal
attribution measure evidenced solid internal consistency, convergent, and divergent
validity, the external measure was less promising. The alpha for the external scale did not
reach .80. And, although the exploratory factor analysis revealed two distinct factors, the
conﬁrmatory factor analysis only evidenced a fair ﬁt of the model to the data.

Yet another limitation involves the minority representation in my sample of
participants. As the workplace becomes more diversiﬁed, minorities will become
employed in a variety of jobs across all status levels. Clearly, with a sample of ﬁfteen
African Americans, the power to ﬁnd signiﬁcant effects utilizing subordinate race was
less than optimal.

Future Research

Future research is clearly needed to address some of these limitations. First and
foremost is the need to develop a widely applicable measure of causal attributions that
can accurately assess whether an individual’s behavior is attributed to internal or external
causes. The measure of attributions developed by Rusell (1982) was inadequate for its
sole use in this study. The intemal-extemal scale, which consisted of three items

evidenced poor internal consistency. Part of that was likely due to the number of items.

68

But another part could also have been due to the ambiguity of the items. The students
were often unclear of what exactly the question was asking.

Perhaps an investigation using ﬁeld data could also improve upon the sample size
used in this study. In the low performance condition, African Americans were rated
slightly lower than White Americans were, but attributions could not be investigated as
potential mediators of a non-signiﬁcant relationship. Perhaps, with a sample size greater
than ninety, the correlation between race and performance ratings would reach
signiﬁcance.

An increased sample size could also introduce subordinate race into the model. In
this study, there were only ﬁfteen African Americans. It would be interesting to see
whether African Americans viewed their team leaders just as White Americans view their
team leaders. Maybe African American subordinates would be less surprised that an
African American team leader was performing well than a White American subordinate
would.

Other laboratory studies also need to be done to see if, when controlling the level
of performance of the subordinates, superiors bias their ratings in a similar fashion to
Sackett and DuBois (1991). If so, attribution theory needs to be investigated as a possible
inﬂuence on that relationship. So far researchers have mostly been proposing that
theories such as intergroup theory have the potential to explain why there is bias in
ratings of subordinates. But no one has investigated whether attribution theory is the

underlying cognitive bias.

69

References

Alderfer, CR (1986). An intergroup perspective on group dynamics. In J. Lorsch (Ed.),
Handbook of organizational behavior. En glewood Cliffs, NJ: Prentice-Hall.

Allen, R.B., & Ebbesen, EB. (1981). Cognitive processes in person perception: Retrieval
of personality trait and behavioral information. Journal of Experimental Social
Psychology, 17, 119-141.

Andrews, G.R., & Debus, R.L. (1978). Persistence and causal perceptions of failure:
Modifying cognitive attributions. Journal of Educational Psychology, 70, 154-166.

Bass, A.R., & Turner, J .N . (1973). Ethnic group differences in relationships among
criteria of job performance. Journal of Applied Psychology, 57, 101-109.

Bigoness, WJ. (1976). Effect of applicant’s sex, race, and performance on employers’
performance ratings: Some additional ﬁndings. Journal of Applied Psychology, 61,
EKL84.

Byme, D. (1971). The attraction paradigm. New York: Wiley.

Campion, M.A., Pursell, E.D., & Brown, B.K. (1988). Structured interviewing: Raising
the psychometric properties of the employment interview. Personnel Psychology, 41,
25-42.

Cardy, R.L., & Dobbins, OH. (1986). Affect and appraisal accuracy: Liking as an
integral dimension in evaluating performance. Journal of Applied Psychology, 71, 672-
678.

Chapin, M., & Dyck, DO. (1976). Persistence in children’s reading behavior as a
function of N length and attribution retraining. Journal of Abnormal Psychology, 85,

511-515.

70

Cohen, J. (1988). Statistical Power Analysis for the Behavioral Sciences. Hillsdale, N.J.:
Lawrence Erlbaum Associates, Publishers.

Cox, J .A., & Krumboltz, J .D. (1958). Racial bias in peer ratings of basic airmen.
Sociometry, 21, 292-299.

Craig, K.M. (1996). Are all newcomers judged similarly? Distinctiveness and time of
entry in task-oriented groups. Small Group Research, 27, 383-397.

Davidson, M.N., Swigert, S., & Ruderrnan, M. (1998). A matter of perspective: The
effect of race of rater and ratee on managerial performance feedback. Paper presented
at the I 998 Academy of Management Meetings in the Careers Division.

Davis, L.E., Strube, M.J., & Cheng, L. (1995). Too many African Americans, too many
White Americans: Is there a racial balance? Basic and Applied Social Psychology, 17,
119-135.

Deaux, K. (1976). Sex: A perspective on the attribution process. In H. Harvey, WJ.
Ickes, and R.F. Kidd (Eds), New directions in attribution research (Vol. 1, pp. 1-26).
Hillsdale, NJ: Erlbaum.

Deaux, K., & Emswiller, T. (1974),. Explanations for successful performance on sex-
linked tasks: What is skill for the male is luck for the female. Journal of Personality
and Social Psychology, 29, 80-85.

DeCarlo, T.E., & Leigh, T.W. (1996). Impact of salesperson attraction on sales
managers’ attributions and feedback. Journal of Marketing, 60, 47-66.

DeJung, J .E., & Kaplan, H. (1962). Some differential effects of race of rater and ratee

on early peer ratings of combat aptitude. Journal of Applied Psychology, 46, 370-374.

71

Dobbins, G.H., & Platz, SJ. (1986). Sex differences in leadership: how real are they?
Academy of Management Review, 1 l, l 18- 127.

Doise, W. (1978). Groups and Individuals: Explanation in social psychology.
Cambridge, England: Cambridge University Press.

Dovidio, J .F., & Gaertner, S.L. (1983). The effects of sex, status, and ability on helping
behavior. Social Psychology Quarterly, 44, 192-203.

Dweck, CS. (1975). The role of expectations and attributions in the alleviation of learned
helplessness. Journal of Personality and Social Psychology, 31, 674-685.

Eaton, N.K., Goer, M.H., Harris, J .H., & Zook, L.M. (1984). Improving the selection,
classiﬁcation, and utilization of Army enlisted personnel: Annual report, I 984 ﬁscal
year (Tech. Rep. No. 660). Alexandria, VA: US. Army Research Institute for the
Behavioral and Social Sciences.

Feather, N.T., & Davenport, P. (1981). Unemployment and depressive affect: A
motivational and attributional analysis. Journal of Personality and Social Psychology,
41, 422-436.

Fox, H., & Lefkowitz, J. (1974). Differential validity: Ethnic group as a moderator in
predicting job performance. Personnel Psychology, 27, 209-223.

Hamner, W.C., Kim, J.S., Baird, L., & Bigoness, W.J. (1974). Race and sex as
determinants of ratings by potential employers in a simulated work-sampling task.
Journal of Applied Psychology, 59, 705-711.

Hastie, R. (1981). Schematic principles in human memory. In E. Higgens, C. Herman, &
M. Zanna (Eds), Social cognition: The Ontario symposium. Hillsdale, N.J.: Erlbaum.

Heider, F. (1958). The psychology of interpersonal relations. New York: Wiley.

72

Heilman, ME. (1983). Sex bias in work settings: The lack of fit model. In L.L.
Cummings & B.M. Straw (Eds.), Research in organizational behavior (Vol. 5, pp.
269-298). Greenwich, CT: JAI Press.

Hollenbeck, J .R., Ilgen, D.R., Sego, D.J., Hedlund, J ., Major, D.A., & Phillips, J.
(1995). Team decision making accuracy under difﬁcult conditions: Construct
validation of potential manipulations using the TIDE2 simulation. In M.T. Brannick, E.

Salas, and C. Prince (Eds.), Team Performance, assessment and measurement: Theory,
research and applications. Hillsdale, NJ: Erlbaum.

Hunter, J.E., & Hirsch, HR. (1987). Applications of meta-analysis. In C.L. Cooper and
LT. Robertson (Eds.), International review for industrial and organizational
psychology. New York: Wiley.

Ickes, W., & Layden, M.A. (1978). Attributional styles. In J. Harvey, W. Ickes, & R.
Kidd (Eds.), New directions in attributional research (V 01. 2, pp.119-152). Hillsdale,
NJ: Lawrence Erlbaum.

Ilgen, DR. (1994, April). Jobs and roles: Accepting and coping with changing structure
in organizations. Paper presented at the Army Research Institute Selection and
Placement Conference, Alexandria, VA.

Ilgen, D.R., & Youtz, M.A. (1986). Factors affecting the evaluation and development of
minorities in organizations. In K. Rowland & G. Ferris (Eds.), Research in personnel
and human resource management: A research manual (pp. 307-337). Greenwich, CT:

JAI Press.

73

Jackson, L.A., Sullivan, L.A., & Hodge, C.N. (1993). Stereotype effects on attributions,
predictions and evaluations: No two social judgements are quite alike. Journal of
Personality and Social Psychology, 65, 69-84.
Jensen, AR. (1980). Bias in mental testing. New York: Free Press.
Jones, E.E., & McGillis, D. (1976). Correspondent inferences and the attribution cube: A
comparative reappraisal. In J .H. Harvey, W.J. Ickes, & R.F. Kidd (Eds.), New
directions in attribution research (Vol. 1, pp. 389-420). Hillsdale, NJ: Erlbaum.
Jones, E.E., & Sigall, H. (1971). The Bogus Pipeline: A new paradigm for measuring
affect and attitudes. Psychological Bulletin, 76, 349-364.

Kanter, RM. (1977). Men and women of the corporation. New York: Basic Books.

Kelley, H.H. (1971). Attributions in social interactions. Morristown, NJ: General
Learning Press.

Kraiger, K., & Ford, K.J. (1985). A meta-analysis of ratee race effects in
performance ratings. Journal of Applied Psychology, 70, 56-65.

Kraiger, K., & Ford, J.K. (1990). The relation of job knowledge, job performance,
and supervisory ratings as a function of ratee race. Human Performance, 3(4), 269-
279.

Landy, F.J., & Farr, J .L. (1980). Performance Rating. Psychological Bulletin, 87, 72-
107.

Lin, T.R., Dobbins, G.H., & Farth, J .L. (1992). A ﬁeld study of race and age similarity

effects on interview ratings in conventional and situational interviews. Journal of

Applied Psychology, 77, 363-371.

74

Martinko, M.J., & Gardner, W.L. (1987). The leader/member attribution process.
Academy of Management Review, 12, 235-249.

Messick, D., & Mackie, D. (1989). Intergroup relations. Annual Review of Psychology,
40, 45-81.

Mitchell, T.R., & Wood, RE. (1980). Supervisor’s responses to subordinate’s poor
performance: A test of an attributional model. Organizational Behavior and Human
Performance, 25, 123-128.

Mount, M.K., Sytsma, M.R., Hazucha, J .F., & Holt, K.E. (1997). Rater-ratee race effects
in developmental performance ratings of managers. Personnel Psychology, 50, 51-69.

Nieva, V.F., & Gutek, B.A. (1980). Sex effects on evaluation. Academy of Management
Journal, 5, 267-276.

Norvell, N ., & Forsyth, DR. (1984). The impact of inhibiting or facilitating causal
factors on group members’ reactions after success and failure. Social Psychology
Quarterly, 47, 293-297.

O’Leary, V.E., & Hansen, RD. (1984). Sex as an attributional fact. In D. Levine (Ed.),
Nebraska Symposium on Motivation (pp. 133-177). Lincoln: University of Nebraska.

Pettigrew, 'I‘.F. (1979). The ultimate attribution error: Extending Allport’s cognitive
analysis of prejudice. Personality and Social Psychology Bulletin, 5, 461-476.

Phillips, J .M. (1997). Antecedents and consequences of leader utilization of staff
information in decision making teams: Addressing a leadership dilemma. Doctoral

dissertation, Michigan State University, Department of Psychology, East Lansing, MI.

75

Prewett-Livingston, A.J., Field, H.S., Veres, J .G., 111, & Lewis, RM. (1996). Effects of
race on interview ratings in a situational panel interview. Journal of Applied
Psychology, 81, 178-186.

Pulakos, E.D., White, L.A., Oppler, S.H., & Borman, W.C. (1989).

Examination of Race and Sex Effects On Performance Ratings. Journal of Applied
Psychology,74, 770-780.

Rand, T.M., & Wexley, K.N. (1975). Demonstration of the effect “similar to me” in
simulated employment interviews. Psychological Reports, 36, 535-544.

Rice, R.W., Bender, L.R., & Vitters, AG. (1980). Leader sex, follower attitudes toward
women, and leadership effectiveness: A laboratory experiment. Organizational
Behavior and Human Performance, 25, 46-78.

Rosen, B., & Jerdee, T.(1973). The inﬂuence of sex-role stereotypes on evaluations of
male and female supervisory behavior. Journal of Applied Psychology, 57, 44-48.

Rosenbaum, R.M. (1972). A dimensional analysis of the perceived causes of success and
failure. Unpublished doctoral dissertation, University of California, Los Angeles.

Rotter, J .B. (1966). Generalized expectancies for intemal-stable versus extemal-unstable
control of reinforcement. Psychological Monograph, 80, 1-28.

Rotter, N., & Rotter, GS. (1969, April). Race, work performance and merit rating: An
experimental evaluation. Paper presented at the meeting of the Eastern Psychological
Association, Philadelphia.

Russell, D.W. (1982). The causal dimension scale: A measure of how individuals

perceive causes. Journal of Personality and Social Psychology, 42, 1 137-1 145.

76

Russell, D.W., McAuley, E., & Tarico, V. (1987). Measuring causal attributions for
success and failure: A comparison of methodologies for assessing causal dimensions.
Journal of Personality and Social Psychology, 42, 1248-1257.

Sackett, P.R., & DuBois, C.L.Z. (1991). Rater-ratee race effects on performance
evaluation: Challenging meta—analytic conclusions. Journal of Applied Psychology,
76, 873-877.

Sackett, P.R., DuBois, C.L.Z., & Noe, A.W. (1991). Tokenism in performance
evaluation: The effects of work group representation on male-female and White
American-African American Differences in performance ratings. Journal of Applied
Psychology, 76, 263-267.

Schachter, S., & Singer, J .E. (1962). Cognitive, social, and physiological determinants
of emotional state. Psychological Review, 69, 379-399.

Schmidt, F.L., & Johnson, RH. (1973). Effect of race on peer ratings in an industrial
setting. Journal of Applied Psychology, 57, 237-241.

Schmitt, N., & Hill, TE. (1977). Sex and race composition of assessment center
groups as a determinant of peer and assessor ratings. Journal of Applied
Psychology, 62, 261-264.

Schmitt, N., Rogers, W., Chan, D., Sheppard, L., & Jennings, D. (1997). Adverse impact
and predictive efﬁciency of various predictor combinations. Journal of Applied

Psychology, 82, 719-730.

77

Silver, W.S., Mitchell, T.R., & Gist, M.E. ( 1995). Responses to successful and
unsuccessful performance: The moderating effect of self-efﬁcacy on the relationship
between performance and attributions. Organizational Behavior and Human Decision
Processes, 62, 286-299.

Smith, T.W. (1990). Ethnic images (GSS Topical Report No. 19). Chicago: University of
Chicago, National Opinion Research Center.

Tajfel, H. (1981). Human groups and social categories: Studies in social psychology.
Cambridge, England: Cambridge University Press.

Taylor, SE. (1981). A categorization approach to stereotyping. In D.L. Hamilton (Ed.),
Cognitive processes in stereotyping and intergroup behavior (pp.- 83-114). Hillsdale,
NJ: Erlbaum.

Turner, LC. (1984). Social identiﬁcation and psychological group formation. In H.Tajfel
(Ed.), The social dimension: European developments in social psychology (pp. 518-
538). Cambridge, England: Cambridge University Press.

Turner, J.C., Hogg, M.A., Oakes, P.J., Reicher, S.D., & Wetherell, MS. (1987).
Rediscovering the social group: A self-categorization theory. New York: Basil
African Americanwell.

Weiner, B. (1985). An attributional theory of achievement motivation and emotion.
Psychological Review, 92, 548-573

Wendelken, D.J., & Inn, A. (1981). Nonperformance inﬂuences on performance

evaluations: A laboratory phenomenon? Journal of Applied Psychology, 66, 149-158.

78

Whitehead, G.I. III, Smith, S.H., & Eichhom, J .A. (1982). The effect of subject’s race
and other’s race on judgments of causality for success and failure. Journal of
Personality, 50, 193-202.

Wilder, DA. (1986). Social categorization: Implications for creation and reduction of
intergroup bias. In L. Berkowitz (Ed.), Advances in experimental social psychology
(Vol. 19, pp. 291-355). San Diego, CA: Academic Press.

Wyer, R.S., & Scrull, T.K. (1980). The processing of social stimulus information: A
conceptual integration. In R. Hastie et al. (Eds.), Person memory: The cognitive basis
of social perception. Hillsdale, N .J .: Erlbaum.

Ziller, R.C., Behringer, R.D., & Goodchilds, J .D. (1960). The minority newcomer in
open and closed groups. Journal of Psychology, 50, 75-84.

Zuckerman, M. (1979). Attribution of success and failure revisited, or the motivational

bias is alive and well in attribution theory. Journal of Personality, 47, 245-287.

79

APPENDIX A

Items Measuring Internal and External Attributions

80

5- Strongly agree

4- Agree

3- Neither Agree nor Disagree
2- Disagree

1- Strongly Disagree

i1. Our team leader is (not) very smart, and that contributed to our performance.

i2. Alpha’s (low) high intelligence greatly affected our score.

i3. Our team leader (did not) put a lot of effort into the game, and that inﬂuenced our
performance.

i4. Our team leader worked (not) very hard to accurately assess all nine attributes.
e1. Our team leader was (not) very lucky on many of the trials, and that affected our
average score.

e2. Our performance was due to a streak of (bad) luck by Alpha.

e3. This game was (not) easy, and that contributed signiﬁcantly to our team leader’s
performance.

e4. It was (not) easy to determine the threat of the targets, and that is why our team leader
did (not) do so well.

81

APPENDIX B

Performance Rating Items

82

5- Strongly agree

4- Agree

3- Neither Agree nor Disagree
2- Disagree

1- Strongly Disagree

ﬂ

. Our team leader performed very well on this task.

2. Without our team leader, we would have had little chance at earning the bonus
money.

3. Our team leader was skilled at sorting all nine attributes into the three rules (i.e.,

category, location, and motion).

Our team leader could have earned the bonus money without my input.

The number of hits we got was mostly due to Alpha.

Alpha was a good team leader.

Our team leader was very good at combining the attributes within each of the rules.

Alpha could not have performed any better.

If I were the team leader, I could not have done any better.

10. Alpha was very good at quickly measuring all nine attributes.

11. Alpha was very good at comparing the general values on his/her individual sheet with
the values in his/her measure summary box.

12. Our team leader made this game look easy.

13. Our team leader was very good at combining the rules into one ﬁnal team decision.

14. Our team leader did not measure all nine attributes quick enough.

15. Alpha did not perform very well on this task.

16. Alpha had difﬁculty combining the three rules into one overall team decision.

17. Alpha made a lot of mistakes during the game.

18. Alpha did not remember that one “safe” attribute meant that the entire rule was safe.

19. Alpha could have done a lot better.

©¢SQMP

83

APPENDIX C

Performance Rating Items Scaled l to 100

84

1. On a scale of l to 100, with 100 being perfect performance, how well do you think
Alpha performed?

2. On a scale of l to 100, how good of a team leader was Alpha?

85

APPENDIX D

Planning for Participants: The Power Analysis

86

Planning for power is an effective tool that can, if utilized properly, reduce the
probability of making a Type 11 error when conducting research. Statistical power is
deﬁned as the probability of correctly rejecting the null hypothesis given (a) the sample
size, (b) the effect size, (c) alpha, and (d) the power of a statistical test. Power, for this
study, was set at the conventional .80 for each of the hypotheses.

Hypothesis 1. In a meta-analysis completed by Schmitt et al. ( 1997), the population
correlation between race and performance ratings was estimated to be .18. Using Cohen’s
(1988) power tables, 160 subjects would be needed to achieve a power of .80.
Hymthesis 2. Hamner, Kim, Baird, and Bigoness (1974) found that thirty percent of the
variance in performance ratings was due to the individual’s actual level of performance.
Using this estimate of r (.50 to be conservative), and the power tables from Cohen, 23
subjects would be needed to achieve a .80 power level.

Hypotheses 3 & 4. Whitehead et al. (1982) examined the three-way interaction between
ratee race, rater race, and actual performance on performance attributions. Their results
evidenced an interaction effect F(4,170)=2.39 p<.052, which was on the small side. For a
small effect size, I would need 700 subjects for a power of .80 according to the tables in
Cohen. For 200 subjects, my power level would be .41. But, since my hypothesis is only
a two-way interaction, my effect size should be larger, thereby reducing the amount of
subjects that I would need. For a small-to-moderate effect size (R2=.04), 160 subjects
would be enough for a power level of .80. Unfortunately my literature search did not turn
up any articles with relevant two-way interactions.

Hypotheses 5 & 6. Since my ﬁnal hypothesis has not, as of yet, been investigated, I must

infer correlation values based on either a small or a moderate effect size. For a small

87

effect size (R2=.01), I would need 700 subjects for a power of .80. For 200 subjects, my
power level would be .41. For a small-to-moderate effect size (R2=.O4), 160 subjects
would be enough for a power level of .80.

Therefore, given the assumptions in this analysis, a minimum sample size of 160 is

thought to be adequate to test the hypotheses.

88

"Illlll'lllllllllllll