Most studies leveraged attributions explicitly on self-defined characterizations of ability, effort, task-difficulty, and luck. But there were some instances where luck was implicitly evoked; Knight et al. (2003) for example, proxied luck by stereotype-consistent and stereotype-contrastive factors beyond the actors’ control. Explicit (i.e., direct) verses implicit (i.e., indirect) inferences of luck have been indicated in Table 1.
Overall, we find that observers tend to ignore the existence of situational factors (i.e., luck), with the belief that they are weighting exclusively on personal qualities (i.e., ability). This reliance on dispositional attributions for what may be largely situational effects creates a fallacy, given that success is fundamentally a weighted byproduct of both dispositional and situational factors. The review also indicates that inter-individual demographics directly influence how attributions are made, with a tendency to disadvantage women and Black individuals. Thus, our findings suggest that differential attributions of luck may contribute to differences in high stakes outcomes such as pay, promotions and opportunity gaps. We argue that the perception of success and how it is attributed is a crucial inclusion and diversity leverage point.
Gender
Gender-role stereotypes are the rudiments of gender discrimination narratives, and are especially tenacious, with a powerful tendency to maintain them (Heilman & Haynes, 2005). Historically, men have been thought to be more competent than women, which is a clear disadvantage to women seeking managerial, executive, or male-dominated roles, despite information contradictory to stereotypes (Heilman & Guzzo, 1978). Psychologists and feminists alike have argued that the concept of success is epitomised through maleness (Jabes, 1980). Therefore, if a woman is successful, this unexpected outcome is explained by luck (Deaux & Emswiller, 1974) or other mutable, unstable causes (Feldman-Summers & Kiesler, 1974). There is substantive evidence to support that these stereotypes persist in occupational settings (e.g., Dodge et al., 1995; Powell, 1999; Schein, 2001). Hill and Augoustinos (1997) describe gender-based attribution differentials as a ‘male-favoring’, ‘female-derogating’ phenomena, where males are favored under circumstances that are both stereotypically masculine and feminine orientated or in neutral conditions. Attributions of success can, therefore, be skewed by an awareness of an individual’s gender, such that a man’s success is normatively attributed to ability without regard to indices of luck, whereas the success of a woman is normatively derogated through dominant attributions of luck (e.g., Deaux & Emswiller, 1974; Etaugh & Brown, 1975; Feather, 1975). This phenomenon provides the basis for gender discrimination reinforced by luck attributions within occupational settings, and the ramifications of these inferences are multidimensional. Males can become disproportionately rewarded for what is thought to be ability, without regard to the ease through which earlier successes may have been afforded. This can also be harmful to men when hired or promoted into roles where failure is foreseeable (Hohman & Brown, 2020). In contrast, women may be overlooked for employment or promotion because of a perceived rather than actual ability-deficit. This error in judgment can cause women to be deemed unworthy of the positions they hold, excluded from key discourse, given less credit, and overlooked for progression into senior roles (Heilman, 2012; Heilman & Haynes, 2005). It can also mean that individual successes in women are thought of as outliers that cannot be generalized to women more broadly. Such pretenses can influence quantifiable metrics, with women less likely than men to be rewarded with equitable pay, merited salary increases, or promotions (e.g., Auspurg et al., 2017; Blackaby et al., 2005; Heilman & Guzzo, 1978). Heterogeneity in attributions between the sexes can drastically shape intrinsic and occupational gendered experiences; with males seen as being instrumental in their success and women seen as being passive (Sieverding & Koch, 2009). Finally, known unmerited reward systems in male-dominated fields can discourage female participation for fear of their progression being hindered for reasons other than a lack of ability. These collectively reduce the chances of equitable representation into roles that have been historically reserved for men (Powell, 1999).
12 studies among 11 papers investigated gender as an influencer of interpersonal attributions of success, with two papers sitting on the intersect with race. Six of those studies were randomized control studies (RCT) and one was a matched case control study. Nine studies were conducted in the laboratory and three were occupational field-based. A result summary can be found in Table 1. A discussion ordered by the date of publication follows. Our focus is on situational attributions (i.e., luck) and dispositional attributions (i.e., ability) of success made for females as compared to males.
The first paper received on this topic was by Feldman-Summers and Kiesler (1974) who found no difference in attributions to luck as a function of gender in two laboratory studies that investigated the effect of gender on interpersonal attributions. Their study involved 114 participants, who self-identified as male and female, with 10 logical and mathematical exercises, and six answer sheets reflecting fictitious actor responses. Participants were advised they were to observe two highly successful, two moderately successful, and two unsuccessful actors. At each level of success, actors identified as either male or female by way of sex-casted names. Attributions were explicitly made on self-defined characterizations of ability, effort, task-difficulty, and luck. Participants were advised that their evaluations would determine the future involvement of each actor. A 2 (observer_sex) x2 (actor_sex) x3 (actor_success) analysis of variance (ANOVA) revealed highly significant effects for the level of success (p<0.001). Perceived success increased as the number of problems increased (F(2,224)=1169.88, p<0.010). As the level of success increased, so did attributions of ability (F=728.64), effort (F=463.77), task-difficulty (F=463.77), and luck (F=30.55). A significant main effect was found for the sex of an actor on effort (F(1,112)=4.75, p<0.050), but not for ability, task-difficulty, or luck.
Feldman-Summers and Kiesler's (1974) second laboratory-based study provided 160 male and female participants with descriptions of highly successful pediatrician or surgeon actors. Actors identified as either male or female using sex-casted names. The pre-test data suggested that males were not expected to succeed as pediatricians over women but were expected to succeed more as surgeons. Attributions were explicitly made on self-defined characterizations of ability, effort, task-difficulty, and luck, by the division of a circle representing the percentage weight of the actors’ success. But to invoke the luck condition, half of the descriptions stated that the actor secured their practice through nepotism. There was a significant interaction between actor sex and the luck condition (F(1,155)=5.572, p<0.050), such that males altered their attributions toward luck more than females did when the luck condition was invoked. The 2 (observer_sex) x2 (actor_sex) x2 (actor_specialty) x2 (acceded practice) ANOVA revealed both sexes were expected to be equally successful as pediatricians. However, males were expected to be more successful as surgeons than females (χ2=36.38, p<0.001). In addition, though results were not significant for luck, male observers attributed ability more to male actors than female actors (F(1,144)=7.952, p<0.010), and attributed the success of female actors to either an ease of task (t(1,144)=3.160, p<0.010) or greater effort (t(1,144)=1.92, p<0.060). Though ability and luck were not significant, female observers attributed an ease of task more to male actors than female actors (F(1,144)=9.099, p<0.010), and attributed more effort to female actors than male actors (F(1,144)=7.540, p<0.010). There were significant interactions between observer sex and actor sex for ability (F(1,144)=7.953, p<0.010) and task-difficulty (F(1,144)=11.533, p<0.010) attributions, but not for effort and luck.
In a RCT conducted in the laboratory, Heilman and Guzzo (1978) also found no impact of luck as a function of gender. A 2x4 factorial design was used to explore whether attribution processes mediate biased reward allocation across the sexes. 15 male and 14 female participants were randomly assigned to a male or female condition. Actor’s sex was denoted male or female by sex-casted names. Varying causal explanations were given for the success of four actors. Conditions were simulated across explicit dimensions of self-defined characterizations of ability, effort, task-difficulty, and luck, scored on a 9-point scale ranging from ‘very appropriate’ to ‘very inappropriate’. Participants were tasked with evaluating the appropriateness of several personnel actions, counting promotions and pay rises, based on a fictitious excerpt with supervisor ratings of the actors’ performance, ability, and effort. There were no significant main effects for ability, effort, task-difficulty, or luck, nor interactions relating to sex biases. Attributions of success that were biased against females impacted the degree to which rewards were viewed as appropriate and it influenced the magnitude of the reward. But males were disadvantaged in the same way when attributions of success were made toward males in stereotype-contrastive roles. Evidence of sex biases dematerialized when similar causal explanations were given for male and female success. Post hoc tests for pairs of factor level means revealed that reward allocation was deemed appropriate when success was predicated on ability and effort, rather than task-difficulty, or luck (p<0.010). Promotion was only considered appropriate when success was ability based (p<0.010), and the preference for reward was greater in the ability over effort condition (χ2 (1)=12.76, p<0.001). Thus, actors high in ability, irrespective of sex, were given ratings substantially positive in valence.
Sherrod and Goodman (1978) did not find an impact of luck as a function of gender in a laboratory study. The authors used a one-way ANOVA with a 2x2x2 design to investigate the role of sex in interpersonal and intrapersonal attributions of success. 24 male and 96 female participants, self-evaluated and evaluated the success or failure of female actors. Actors were observed attending to a series of 15 spatial reasoning tasks that required matching unfolded geometric outlines to three-dimensional figures. Fictitious performance outcomes were disclosed; the success condition with 12:15 correct answers placed some in the 88th percentile; the failure condition with 5:15 correct answers placed others in the 20th percentile. Attributions were measured on two 15-point scales; relating to how well performance reflects general ability and the extent to which performance was affected by extraneous situational factors (i.e., luck). Actual performance did not differ across conditions (F(2,96)=0.42, p<0.050). Actors made more ability-based intrapersonal attributions in the presence of male observers than when observed by females (F(1,44)=8.88, p<0.010). Congruently, an orthogonal contrast determined that male observers made more ability-based attributions than female observers (F(1,44)=4.28, p<0.050). Thus, both interpersonal and intrapersonal ability attributions were influenced by sex. However, there were no significant effects of gender on luck.
In contrast to the four earlier received studies, Jabes's (1980) RCT, occupational field-based study, found an impact of luck as a function of gender, but the bias was opposite to that what was expected. In a 2 (actor_sex) x2 (actor_age) x3 (actor_occupation) factorial design study, 144 female public and private sector managers (Mage=35, range=22-61) with years of managerial experience between them ranging 1-18 years (Mage=4) were recruited. Participants were to explicitly make attributions on self-defined characterisations of ability, effort, task-difficulty, and luck through a four-segment circle divide; each segment therefore varied by the perceived importance of each component. Attributions were predicated on 12 fictitious letters of recommendation and personnel files. The personnel files varied by forename (i.e., David/Joan), age (i.e., 32/46), address, occupation (i.e., Architect/Editor/Social Worker, with equivalent educational details), income range, marital status, and number of children, along with spouse name and occupation. Files were otherwise identical. As compared to males, female success was attributed more to ability (F(1,132)=5.62, p<0.020), and males were thought to be luckier (F(1,132)=3.52, p<0.06).
In a RCT, conducted in the laboratory, Yarkin et al. (1982) found luck attributions were a function of gender, but intersectionality was salient, as attributions of success were evaluated on the intersection of sex and race. (Race differences are later described). 120 participants, male and female, read a description of a highly successful 27-year-old banker, who depending on the condition was male or female, black or white. A cover letter defended the promotion request, highlighting three years of banking experience and the actor’s occupational history. The actor’s sex was denoted by sex-casted names. Explicit attributions were made on self-defined characterizations of ability, effort, task-difficulty, and luck. No significant sex of observer differences were obtained in any of the attribution measures. But main effects for the sex of the actor were found for ability (F(1,112)=4.64, p<0.050), effort (F(1,112)=5.76, p<0.050), and luck (F(1,112)=16.67, p<0.001), though not for task-difficulty. In addition, significant interaction effects were seen between actors’ race and gender on ability (F(1,112)=4.24, p<0.050), effort (F(1,112)=3.96, p<0.050) and luck (F(1,112)=9.35, p<0.010), but not task-difficulty. A Newman-Keuls multiple comparisons test indicated that the success of white male actors was attributed more to ability and less to effort or luck, compared with both female actors, irrespective of race, and black male actors (p<0.050).
Reid et al. (1986) found an impact of luck as a function of gender, by exploring the mediating role of attributions in sex biased interviewing processes in a laboratory-based RCT. 89 male and 91 female participants were randomly assigned to one of six conditions, in which they were to evaluate an interview transcript of male or female actors applying for teaching roles in English, History or Mathematics (sex-typed feminine; neutral; masculine respectively). This was to be assessed against an explicit quadchotomy of self-defined characterizations of ability, effort, task-difficulty, and luck. A 2 (observer_sex) x2 (actor_sex) x3 (actor_occupation) three-way ANOVA revealed no main effects or interactions on the overall evaluations. However, a stepwise regression for the overall evaluation of all attributions was significant. Significant three-way interactions were found for effort (F(2,168)=4.64, p<0.010); task-difficulty (F(2,168)=4.41, p<0.010), and luck (F(2,168)=4.21, p<0.020), but not for ability. In the luck condition, as compared to males (M=2.53; ±1.41), the past successes of female actors (M=3.67; ±1.50) in the English role received greater attributions to luck. However, in the effort condition, male observers attributed more effort to female actors (M=4.71; ±1.64) than male actors (M=3.27; ±1.58) for the Mathematics role. Thus, in the field that was stereotype-contrastive (e.g., a female in a Mathematics role), attributions were biased toward effort, but in a field that supports a priori beliefs about traditional gender roles attributions were biased toward luck. These effects were a function of the observer’s sex, as there were no significant differences among female observers, while male observers differentially attributed the success of male and female actors as a function of teaching role, on dimensions of effort and luck. However, as it pertains to task-difficulty, both observers, male (female actor - History: M=3,31; ±1.25; Mathematics: M=2.21; ±0.80) and female (male actor - History: M=3.27; ±1.22; Mathematics : M=2.71; ±1.05), agreed that it was easier for actors to achieve success in neutral fields than in stereotype-contrastive roles.
Almeida and Kanekar (1989) found an impact of luck as a function of gender in a laboratory-based study. Authors used a 2 (observer_sex) x2 (actor_sex) x2 (actor_occupation [bank manager (high-status role)/personal assistant (low-status role)]) x2 (actor_outcome [success/failure]) factorial design, to test whether attributions of success varied according to sex-typed occupations. 336 participants, male and female, were provided with fictitious bibliographical information of actors, and were to explicitly attribute actor success on a four-point scale that reflected self-defined characterizations of ability, effort, task-difficulty, and luck. The product-moment correlations for pairs of causal elements between ability and luck were 0.14 (df=334, p<0.050). Female actors in low-status roles were attributed less to ability by male observers than female observers (t=2.62, p<0.010). Male actors in high-status roles were attributed more to ability than female actors in the same role (t=2.38, p<0.020), but male actors in low-status roles were attributed less to ability than female actors in the same role (t=2.31, p<0.050). Congruently, male (High-status: M=6.10; Low-status: M=5.36; t=3.26, p<0.005) and female (High-status: M=5.38; Low-status: M=6.05; t=3.18, p<0.005) actors received higher ability attributions in stereotype-consistent roles. Overall, male success was more highly attributed to effort (M=6.02) than female success (M=5.95; t=2.62, p<0.010) and males in low-status roles (M=5.55), but women in high-status roles (M=5.17) were perceived to have applied less effort than males in low-status roles. Male actors in a high-status role (M=3.26) received less attributions to task-difficulty than female actors in the same role (M=4.14; t=2.52, p<0.020). But the success of female actors was attributed more to task-difficulty when in high-status roles (M=4.14) than when in a low-status role (M=3.36; t=2.25, p<0.050). Critically, females in high-status roles (M=3.41) were considered luckier than males in the same role (M=2.48; t=3.18, p<0.005). But males in low-status roles (M=3.10) were considered luckier than females in the same role (M=2.26) and males in high-status roles (M=2.48; t=2.86, p<0.005). In addition, female actors were thought to be luckier in high-status roles (M=3.41) than when in low-status roles (M=2.26; t=3.92, p<0.001). But males in low-status roles (M=3.10) were thought to be luckier than males in high-status roles (M=2.48; t=2.12, p<0.050).
An occupational field-based study, using a matched case control design, conducted by Greenhaus and Parasuraman's (1993) found an impact of luck as a function of gender only on the intersect of race. The paper was part of a larger study on the career experience of 317 male, 423 female, and eight non-gender stipulated black and white American managers (Mage=38.76, ±8.20) from the fields of banking, communication, and electronics. (Race differences are later described). A matched sampling process was used to match participants by age, organizational tenure, job function, and seniority, together with their supervisors’ details, to insure against confounding. Performance was rated ‘satisfactory’ or ‘unsatisfactory’ to invoke the success and failure conditions. Success was explicitly attributed to ability, effort, task-difficulty, support, and luck on a five-point scale. The authors determined that the success of females (M=3.97) was less likely to be attributed to ability than the performance of males (M=3.90; t=2.47, p<0.020). However, this result was true only where it related to highly successful managers. Among moderately successful managers, the performance of females was as likely to be attributed to ability as that of males. No gender differences in attributions were found for effort, task-difficulty, support, or luck, but an interaction between gender and race was observed for task-difficulty (r=0.66, p<0.010) and luck attributions (r=0.39, p<0.050).
Taylor et al. (1993) did not support that gender biases luck attributions of success in an occupational field-based study. Each of the 60 males (aged 15-17) and 60 females (aged 14-50) were provided with two scenarios: a success and failure condition that was factorially manipulated to account for sex and career sex-typing. Explicit attributions were made on self-defined characterizations of ability and luck. Results determined that success was mostly attributed to ability irrespective of the actor’s sex or the sex-typing of roles. There were gender differences in ability and luck attributions, but only as it pertained to failure. Significant interactions between the actor’s gender, the sex-typed role, and success or failure, revealed that the actor’s gender influenced attributions when the failed task was stereotype-consistent (F(1,232)=13.44, p<0.001) and when the succeeded task was stereotype-consistent (F(1,232)=11.37, p<0.001). No other main effects or interactions reached significance.
In a RCT laboratory-based study, Hill and Augoustinos (1997) were unable to support that luck attributions depended on gender. 26 male and 78 female participants (Mage=21.59, ±5.12) explicitly evaluated attributions on self-defined characterizations of ability, effort, task-difficulty, and luck, through 12 random scenarios. The scenarios indicated male and female success and failure, four attribution questions, and a question to gauge social desirability. A 2 (observer_sex) x2 (actor_sex) x3 (actor_course: nursing [feminine]/engineering [masculine]/law [neutral]) x2 (actor_outcome: success/failure) x4 (actor_attribution: ability/effort/task-difficulty/luck]) repeated measures ANOVO revealed a main effect for attribution (F(3,312)=225.16, p<0.001, η2=0.68) and outcome (F(1,104)=14.43, p<0.001, η2=0.12). The interaction between actor sex, attribution, and outcome was not significant (F(3,312)=2.47, p=0.062). Luck was the least made attribution (M=3.90, ±0.64), followed by task-difficulty (M=3.38, ±0.58), then ability (M=2.71, ±0.64), with the most prevalent attributions being made for effort (M=2.01, ±0.55). The interaction between observer and actor sex was significant (F(1,104)=6.19, p<0.050, η2=0.06); revealing that female observers made stronger attributions overall than did male observers. Ability attributions were more readily made for success in law (M=2.25, ±0.82) and engineering (M=2.32, ±0.83), rather than for success in nursing (M=2.39, ±0.77). Remarkably, attributions for task-difficulty were made more for success in nursing (M=3.68 ±0.72) than for engineering (M=3.81, ±0.76) and law (M=3.83, ±0.79). Gender associations with attributions of effort and luck were not significant.
Sieverding and Koch (2009) did not find that luck attributions were a function of gender, in a laboratory-based RCT involving computer related tasks. 99 male and 107 female participants (Mage=23.7, ±4.2) were randomly assigned to one of two experimental conditions. They evaluated actors who were either male (49:53 male to female observers) or female (50:54 male to female observers). Participants watched a video of actors solving a complex, time-restrictive computer task, and were advised that while 66% were successful, the actor they were to evaluate performed well above average. After an explicit attribution of the actors’ success to either self-defined characterizations of ability or luck, participants had to judge their own hypothetical competence, as compared to actors. Irrespective of the observers’ gender actors were expected to be successful (χ2 (2,206)=0.18, p=0.910). There were also no significant gender effects seen when the actors’ sex differed (χ2 (2,206)=1.4, p=0.510). The 2x2 multivariate analysis of variance (MANOVA) revealed that attributions of success for female actors was not attributed more to luck than male actors (F(2,202)=1.24, p=0.270, η2=0.01), even when accounting for the observers gender (F(2,202)=0.24, p=0.620, η2=0.00). Therefore, the posit that female success would be attributed to luck over ability was not supported, nor was there any evidence that the competence of a female would be underestimated.
Gender Discussion
Early reports of gender biases in attribution have ignited efforts to further egalitarian attitudes toward women (Taylor et al., 1993), but laboratory research findings may not accurately capture and reflect these issues. The gender-levied interpersonal attribution studies within our review spanning 1974 to 2009 suggest few instances of an invocation of luck for women, and mixed results for ability with effect sizes relatively small in magnitude. There was a center cluster of five associations relating to luck between the 80’s and the early 90’s, although three of those had unexpected findings. First, where luck was more salient to men (Jabes, 1980), and second, where the intersect of gender and race was critical to the outcome (Greenhaus & Parasuraman, 1993; Yarkin et al., 1982). Therefore, only two out of a total of 12 studies support that the success of women is attributed more to luck than men, and of these, one is a RCT and both are laboratory-based studies. This indication of change toward more egalitarian attitudes is promising but could equally reflect social adaptation and a concealment of ideals that are no longer socially acceptable, rather than a change in attitude or reduced biases. In addition, though there was an even divide, we were not able to extrapolate from the survey of evidence a contrast between early and more recent studies as it pertains to ability attributions. Four laboratory-based and two occupational field-based studies found effects on ability. Of these, three were RCT and one was a matched case control study. Overall, results weighed toward no gender effects for luck and ability attributions together, but design differences between studies, especially relating to participant characteristics, methodology, and time, likely account for much of the heterogeneity in results.
A compelling amount of extant literature has converged on the understanding that the success of women is, by and large, undermined by a prevalence of attributions toward luck over ability (e.g., (Deaux & Emswiller, 1974; Heilman & Haynes, 2005; Reid et al., 1986), but the present review reveals that context matters. Even though associations with luck, specifically, were not significant, Heilman and Guzzo (1978) highlighted the liability of situationally derived causal attributions of success beyond the laboratory environment. Where the preference toward rewarding dispositional factors, without regard to situational factors has implications on sex differentials in ability attributions. This was evidenced by half of our review studies and has been demonstrated elsewhere (e.g., Sieverding & Koch, 2009; Yarkin et al., 1982). In such instances, women were seen as less able, and were more likely to be overlooked for the most meaningful rewards. By contrast, men were seen as being instrumental in their own success, so attributions leant toward the dispositional, with consequential rewards being more likely. Indeed, although we were unable to evidence gender differences in attributions of success – pointing to a shift towards more egalitarian perceptions – absence of evidence is not evidence of absence, and numerous studies contrast this position (e.g., Biernat & Kobrynowicz, 1997; Greenhaus & Parasuraman, 1993; Sherrod & Goodman, 1978), suggesting again that context matters.
Curiously, two of the five studies with an effect of gender on luck attributions (in either direction) had predominant or exclusive female participation. The remaining three had an equal male to female ratio, as did five of the seven studies that indicated no effect. This could point toward variations in luck attributions being influenced by the gender differences of those observing, rather than of those being observed. However, this should be interpreted with caution as most results were not stratified by participant gender, which would be required to substantiate this observation. It is equally plausible that luck attributions made by women are less prone to socially desirable responses (Hill & Augoustinos, 1997) or, as was shown by Sherrod and Goodman (1978), men generally attribute success more to ability than women do. Equally, male participants may have sought to demonstrate politically correct attitudes and, therefore, refrained from making discriminatory assertions relating to a woman’s performance (Taylor et al., 1993). This would suggest that a latent bias still exists but that it is hard to measure.
Congruent with the dominant view that sex biases ability attributions (e.g., Cohen & Bunker, 1975; Schein, 1973; Taylor & Ilgen, 1981), Sherrod and Goodman (1978) and Greenhaus and Parasuraman (1993) found effects for both luck and ability. However, neither Reid et al. (1986) or Almeida and Kanekar (1989) found main effects nor interactions between gender and ability attributions, despite the significant interactions found for luck. This is of particular interest, as it puts into question the salience of luck across studies and could underlie null findings within the laboratory settings. Ability is more ubiquitously used in discourse than notions of luck, but most studies allowed for luck to be explicitly self-defined, which may not have adequately evoked the condition in the same way that ability was. Indeed, Sherrod and Goodman (1978) submit that men attribute success more to ability than women, while Taylor et al. (1993) found that irrespective of gender success was attributed more to ability than other conditions.
In line with Sieverding and Koch (2009), Taylor et al. (1993) submitted that their null findings in attribution differences between genders, suggest a shift in the perception of women, specifically, that the concept of women as skillful professionals may have gained traction as a norm. However, there is nuance to these findings. Reid et al. (1986) submit that non-significant ability results may have been a consequence of setting ability levels so high that ability attributions would be difficult to deny. This supports a view that women are only credited with their success at exceptionally high levels of performance and achievement. An alternative interpretation to this phenomenon is that they must work harder to have their success credited to them over being considered lucky. Biernat and Kobrynowicz's (1997) findings are an adjunct to this narrative, where lower minimal competency standards but higher ability standards were set for women over men. This can result in an enigma, consistent with the status characteristic perspective, where women experience an ease in meeting low standards and must do more to prove their ability, but have a difficulty in proving their success was a result of ability over luck – a “twice as hard to be considered half as good” effect (Biernat and Kobrynowicz, 1997, pg. 550). It also contradicts Greenhaus and Parasuraman (1993) and prior studies (e.g., Cash et al., 1977; Deaux & Emswiller, 1974; Etaugh & Brown, 1975) who determined that the success of women was likely attributed to luck when assessing highly successful managers. There, ability was a less salient cue in the consideration of promotion, so women were less likely to be advanced in their career - metaphorically conceived as the ‘glass ceiling’. Intriguingly, even among moderately successful managers, the success of women was as likely to be attributed to ability as men. These together suggest that gender biases are evoked when a women’s success violates (exceeds specifically) expectations and contradict a priori beliefs, where adverse attributions are exploited as a dissonance reduction technique.
Females were evaluated more positively for feminine type-casted roles, with a greater salience for males and masculine type-casted roles. The inference is that women were considered good relative to other women (Jabes, 1980), but not so on a scale with men (Reid et al., 1986). But, while gender was not associated with luck attributions in Feldman-Summers and Kiesler's (1974) study, it is intriguing that they were unable to determine a single occupation in which women were expected to be more successful than men. A women’s success is, it seems, less anticipated and more likely to be explained through alternative attributional terms, such as luck. In line with predominant sex-casted stereotypes, Almeida and Kanekar (1989) revealed that women were believed to be able in low-status occupations, whereas high-status roles were at the exclusive preserve of men. Jabes (1980) findings might seem evidence to the contrary, but it cued more to individuals being in management, and was not concerned with the peculiarities or the prestige of the occupation.
Race
Ethnic groups, particularly black members, are systematically disadvantaged in the workforce, and remain notably underrepresented in corporate environments in senior positions. Organizations tend to be, instead, distributed in such a way that these groups are predominately relegated to subordinate positions, with little prospect of upward mobility (Knight et al., 2003; Orpen, 1981). The proliferation of workforce inclusion initiatives has made attributions of success potentially more vulnerable to implicit and explicit biases because the working environment has become more diverse. Biases that can negatively impact chances of employment and opportunities for advancement., with the potential inadvertent diminishing of performance and future success (Ellis et al., 2006; Knight et al., 2003). Ingroup-outgroup biases are such that members tend to view others within their group in favorable terms, but then look unfavorably toward individuals outside of their group (Whitehead et al., 1982). Race is one distinguishing feature that is particularly conducive to ingroup-outgroup bias. Whether this bias is a result of negative disposition toward others unlike us, or stereotypes that have integrated into our belief system is critical to intervention strategy and efficacy. We explore a number of studies that investigate whether a homogeneity exists between attribution outcomes made by different racial groups. In particular, we focus on situational attributions (i.e., luck) and dispositional attributions (i.e., ability) made for white individuals as compared to other ethnic individuals.
We survey seven studies in total, five of which exclusively investigated race as a key influencer of interpersonal attributions of success, with an additional two studying the intersection of sex. Of those, five were RCT and one was a matched case control study. Five studies were conducted within a laboratory setting and two were conducted in an occupational field-based setting. We follow this up with a summary of evidence, ordered by publication date, as a way to understand the psychological mechanisms that underlie attribution errors in success as they pertain to race. A result summary can be found in Table 1.
The first study to consider race and luck attribution is owed to Greenberg and Rosenfield (1979). Overall, these authors found that luck attributions were a function of race in a RCT laboratory-based study. They examined the moderating role of ethnocentrism[1] over stereotype utility in racially driven interpersonal attributions of success. Ethnocentrism and discrimination were highly correlated (r=0.54, p<0.010). 116 white male participants were shown an audiovisual of two actors performing well and two performing poorly on a test. The test consisted of ten trials where the actor guessed the shape or line count on hidden cards, followed by a questionnaire with scales on dimensions of ability, effort, task-difficulty, and luck that were explicit and self-defined. One MANOVA revealed a significant main effect for actor outcome on all four attribution dimensions (p<0.001), but success was attributed less than failure. A second MANOVA revealed a significant main effect (F(1,45)=5.63, p<0.050) where participants high in ethnocentricity attributed the success of black actors less to ability (M=1.78) and more to luck (M=1.22) than the success of white actors (Ability: M=2.35; Luck: M=0.22). Participants low in ethnocentricity attributed the success of black actors more to ability (M=2.46) and less to luck (M=0.29) than the success of white actors (Ability: M=2.13; Luck: M=0.63). The interaction between ethnocentrism, actor outcome and actor race was significant on attributions of ability (F(1,46)=8.67, p<0.005), but not so for attributions of luck (F(1,46)=3.11, p<0.090).
Orpen (1981) found that luck attributions were a function of race within a RCT, occupational field-based study. The author evaluated whether attributions of success, made by 136 white middle-managers (Mage=32.6) on four black and 26 white subordinates was influenced by race. The racial divide of the actors corresponded to the racial balance in South Africa. Managers were asked to evaluate performance information indicated on six indices (i.e., output quality and quantity; turnover and absenteeism; subordinate grievances; financial/budget control) for actors over the last year. Attributions were explicitly made on self-defined characterizations of ability, effort, task-difficulty, and luck. Actual performance varied widely, except in the case of four actors; two black and white, who had identically good performance, and two other black and white actors with identically poor performance. A comparison was made by t-tests for independent samples between the observers’ mean ratings of the causal factors in the success condition for black and white actors. Ability (M=4.86) and effort (M=5.02) were given less credence for the success of black actors than their white counterparts (Ability: M=6.10; Effort: M=6.04, p<0.01). While task-difficulty (M=8:10) and luck (M=8.08) were deemed more critical to the success of black actors than white actors (Task-difficulty: M=4.87; Luck: M=5.51, p<0.001). Failure followed a divergent pattern of attributions, with dispositional characteristics being most salient among black actors.
In a laboratory-based RCT, Whitehead et al. (1982) found that luck attributions were a function of race. 77 black and 43 white participants were randomly provided with a brief description of 156 black and 208 white actors in college who were to be rated on a series of characteristics, including intelligence and athleticism. Attributions were made explicitly on self-defined characterizations of ability, effort, task-difficulty, and luck. A multivariate analysis revealed a significant interaction between success/failure, observer race and actor race (F(4,170)=2.39, p<0.052). In the academic condition, univariate tests (df=1,173) showed that observers attributed success more to ability (M=5.17, F=56.29, p<0.001) and effort (M=5.63, F=50.32, p<0.001) than task-difficulty (M=2.91, F=5.39, p<0.030), but not to luck (M=2.35, F<1). In the athletic condition, observers attributed success more to ability (M=5.51, F=56.68, p<0.001) and effort (M=5.46, F=48.32, p<0.001), than task-difficulty (M=3.13, F=7.95, p<0.005) or luck (M=2.86, F=4.78, p<0.030). Though not significant in the former, a 2x2 ANOVA indicated that black actors were thought to be less intelligent (Black: M=59.84; White: M=66.95, F(1,114)=2.92, p<0.090), but more athletic (Black: M=70.33; White: M=57.37, F(1,114)=12.15, p<0.001) than their white counterparts. Also borne out of the data, black observers more generously attributed academic success to ability when the actor was part of ingroup (Black: M=5.84; White: M=4.93, F(l,173)=2.98, p<0.086). This result was not replicated when observers were white who, regardless of race, attributed to academic ability equally (Black: M=5.11; White: M=4.94, F<1). Attribution differences in the athleticism success condition were not seen irrespective of whether the observer or actor was black or white (Black/White: M=5.96; Black/Black: M=5.85; White/Black: M=5.11; White/White: M=5.19, F(l,173)=2.98, p<0.086).
Yarkin et al. (1982) could only support that luck attributions for success were a function of race on the intersect of gender, in their RCT laboratory-based study. 120 male and female American participants were randomly assigned to each condition using a 2 (actor_gender) x 2 (actor_race) x 2 (observer_gender) between-subjects factorial design. Attributions were made explicitly on self-defined characterizations of ability, effort, task-difficulty, and luck. Race was denoted by résumé information, including affiliations (e.g., associations to historically black or white American universities). Main effects for the race of the actors were seen for ability (F(1,112)=10.13, p<0.001) and effort (F(1,112)=5.12, p<0.05), but not so for task-difficulty or luck. Observers attributed success significantly more to ability and effort when the actor was a white male, relative to being a white female, or a black person irrespective of gender (p<0.05 in all cases). Interactions between the actors’ race and gender were significant for ability (F(1,112)=4.24, p<0.050), effort (F(1,112)=3.96, p<0.050), and luck (F(1,112)=9.35, p<0.010), but not for task-difficulty.
In an occupational field-based, matched case control study, Greenhaus and Parasuraman (1993) found an impact of luck attributions as a function of race. 322 black and 426 white American managers (Mage=38.76, ±8.20) in banking, communication, and electronics participated. There was a higher proportion of women among black participants comparted to white participants. (Gender differences earlier described). Performance was rated ‘satisfactory’ or ‘unsatisfactory’ to invoke the success and failure conditions. Success was attributed to ability, diligence, luck, task-difficulty, and support on a five-point scale. The authors determined that the promotional prospects of black managers were less promising than their white counterparts (r=0.06, p<0.050). Overall ability attributions were weaker for black (M=4.10) than white managers (M=4.24). A similar patten was observed between black (M=3.96) and white (M=4.14; t=2.14, p<0.050) managers with limited experience (≤1 year). But among experienced managers (>1 year), ability attributions did not significantly differ. Effort attributions were weaker for black (M=4.13) than white (M=4.27) managers, but no differences were found regarding managerial experience. Congruently, black managers (M=2.35) were thought to have received more support than white managers (M=2.17). Significant gender and race interactions were found for attributions to task-difficulty (p<0.010). Overall, black male managers (M=1.70) were thought to have easier work than white male mangers (M=1.33). But white female mangers (M=1.45) were thought to have easier work than black female managers (M=1.32, p<0.050). Among mangers with limited experience, black male managers (M=1.64) were also thought to have easier work than white male mangers (M=1.40, t=2.86, p<0.010). But task-difficulty attributions did not differ among experienced managers. Significant gender and race interactions were found for luck attributions (p<0.050). Overall, the success of black male (M=1.40) and black female managers (M=1.27) was more likely to be attributed to luck than their white male counterparts (M=1.23; p<0.050), but no differences were found by managerial experience. Negative attributions were reduced when the supervisor-supervisee relationship was strong and longer than a year.
Knight et al. (2003) found that race influenced luck attributions in their laboratory-based study of aversive racism[2], by way of stereotype-consistent and stereotype-contrastive patterns in attributions. 156 white males participated in this 2 (actor_race: black/white) x2 (actor_status: subordinate/manager) x2 (actor_mistake: small/large) between-subjects study. Status represented dimensions of success or failure. Participants were to act as middle-level managers who reported to higher-level managers and managed subordinates. Actors varied only by status and the magnitude of the mistake made. Luck was leveraged on stereotype-consistent or contrastive factors beyond the actors’ control. Ambivalent information was included to increase the possibility that participants would pursue schema-consistent evidence in their decision process. Evaluations were made on several dimensions, including work quality, stress tolerance, leadership, potential, interpersonal skills, conceptual skills, administrative skills, and promotion. Using a between-subjects ANOVA with a sequential Bonferroni correction to control the Type-I error rate, a two-way interaction was found for effort (F(1,145)=7.62, p=0.007, η2=0.050). A three-way interaction between race, status, and mistake made was not significant, but the multivariate analysis of variance (MANOVA) revealed a two-way interaction between race and status (F(9,137)=2.20, p=0.026, η2=0.130). Specifically, negative attributions of black leaders and white subordinates were stereotype-contrastive, but positive attributions for black subordinates and white leaders were stereotype-consistent. Stereotype-contrastive patterns in attributions were attributed to luck. Congruently, white managers (M=3.67, ±1.06) received stronger ability attributions than white subordinates (M=3.20, ±1.08; t(74)=-1.92, p=0.029). But this normative pattern was reversed for black managers (M=3.37, ±0.85) and subordinates (M=3.86, ±1.07; t(76)=2.23, p=0.015).
In a laboratory-based RCT, Ellis et al. (2006) determined that luck attributions by subordinates were not impacted by leader race in the expected direction; a reversal of anticipated effects was seen, with attributions toward black leaders being more ability-based than white leaders. 123 male and 54 female participants were randomly assigned to four-person groups, with one leader and three subordinates. 162 participants were white and 15 did not disclose race. In an airspace simulated decision-making task, subordinates independently collated information to determine the level of threat that an aircraft represented. Said information was relayed to the leader who made an overall decision that was judged by the subordinates. Group members were informed of the quality of that decision. On the basis of their performance, the top performing team members could earn a bonus. Attributions were explicitly made on characterisations of intelligence denoting ability, along with effort, task-difficulty, and luck on a five-point scale. Consistent with the fundamental attribution error, as leader performance increased, observers were more likely to attribute outcomes internally (e.g., ability; r=0.47, p<0.010) and less likely to attribute toward external causes (e.g., luck; r=-0.59, p<0.01). In the case of high-performance teams, a three-step hierarchical regression revealed that black leader performance was attributed to ability (M=2.35, ±2.34), more than white leaders (M=0.72, ±1.98; t(88)=3.56, p<0.010). But no significant differences in ability attributions were found between the races when team performance was low. Although the interaction between leader race and performance explained a significant proportion of the variance in ability attributions (ΔR2=0.02, p<0.010), in the case of low-performance teams, black leader performance was attributed to luck (M=2.08, ±2.20) more than for white leaders (M=1.25, ±2.34; t(85)=1.73, p<0.05). But in high-performance teams, the success of black leaders was less attributed to luck (M=-2.03, ±2.33), than white leaders (M=-1.18, ±2.03; t(88)=-1.85, p<0.050). Irrespective of race, high-performing managers received higher ratings when their performance was attributed to ability over luck, but low-performing managers received higher ratings when their performance was attributed to luck.
Race Discussion
All seven race-related studies within this review, from 1979-2006, provided evidence of attribution errors, but two studies had unexpected findings. As it pertains to luck and ability attributions, with one exception (Ellis et al., 2006), attribution errors were in the expected direction. In addition, one study only demonstrated attribution errors with respect to luck on the intersect of race and gender (Yarkin et al.'s, 1982). Thus, five out of a total of seven papers suggest that the success of black individuals is attributed more to luck and six suggest their success is attributed less to ability than their white counterparts. While results are weighted heavily in favor of attribution errors being impacted by race – an immutable factor beyond one’s control – and most were in the expected direction, one late study found that attributions were favorable toward black individuals over their white counterparts (Ellis et al., 2006). The only RCT and matched case control studies that were occupationally field-based were Orpen's (1981) study of white managers, and Greenhaus and Parasuraman's (1993) study of black and white managers; both that revealed attribution errors as it relates to luck and ability. Though several other laboratory-based studies followed the same trajectory (Greenberg & Rosenfield, 1979; Whitehead et al., 1982; Knight et al., 2003).
There are important implications of Greenberg and Rosenfield (1979), Greenhaus and Parasuraman (1993), Knight et al. (2003), Orpen (1981) and Whitehead et al.'s (1982) findings for organizations. They converge on the understanding that attributions of success are negatively influenced by race and yet, within this context, they each highlight distinctions in the obstacles faced between the races, including whether black individuals are seen as being worthy drivers of their own success, particularly as seniority increases. These offer a partial explanation for an inequity in reward allocation within many organizations (Amis et al., 2018), since rewards are afforded on the basis of individuals being credited for their own success (Wittig et al., 1981). These biases appear to persist across a black person’s career, existing on an increasing level of liability with progression into more senior roles despite evidence of strong ability (Orpen, 1981). However, Ellis et al. (2006) found a reversal of effects, since black individuals were thought to have higher ability and were rated lower in luck than their white counterparts. This purported shift toward more egalitarian values runs counter to expectations and suggests that context matters. The authors argue post hoc that, despite this sociopolitical transformation, there remains an inequity in society that develops into “White guilt”, and it is this that positively influences attributions toward black individuals in an attempt at restitution (Ellis et al., 2006, pg. 312). Also of note, it is possible that untested intersectionality has confounded the results, because while gender did not affect accuracy-linear consistency or variability bias, it did significantly impact the decision-making error-mean bias. Still, having controlled for gender, the findings remained the same. Interestingly, despite the demonstration of effect modification by gender, Yarkin et al. (1982) found that race influenced ability attributions, but not luck attributions, which suggests that luck indices, that were explicit and self-defined, were likely not as salient as ability.
Different to Ellis et al., (2006) and Yarkin et al. (1982), Greenberg and Rosenfield (1979), Knight et al. (2003) and Orpen (1981) had an exclusively white participant pool that may have influenced luck and ability attributions. It is conceivable that reduced social desirability was observed in participant responses in these latter studies, because of a diminished pressure to conform to politically correct attitudes and ideals among ones ingroup. Ellis et al. (2006) suggest that an awareness of racial biases has important implications for attributions, in that it can create a level of discomfort when making attributions that would privilege certain groups. While not wanting to over-simulate the psychological processes at play, these authors also muse over the potential for social desirability as a way to explain their null results. However, Greenhaus and Parasuraman (1993) and Whitehead et al. (1982) submit that black individuals are not necessarily internalizing these oversaturated messages of racial biases, such that it is implicitly informing their attributions. This was seen in an attributions study that evaluated the level of success, rather than ability versus luck (Hassan, 2017), were race was not a determinant in the interpersonal attributions of success among a pool of exclusively black participants. This suggests that it is not only the race of the actor that shapes attributions, but also the race of the observer, which is an important consideration for focused interventions. Curiously, there was no available evidence on the interpersonal attributions of exclusively Asian or other ethnic individuals on black individuals, nor a direct contrast between these groups, nor a comparison of these groups against white individuals. In spite of one gender-related study having a South Asian sample, this may be subject to the majority of studies being conducted in America (Table 3), where race is primarily dichotomized between black and white because of the population sociodemographic distribution. These documented differences in attributions between participants who are all white and those who are all black suggest that results cannot be legitimately generalized to other racial groups without further evidence.
Greenberg and Rosenfield (1979), Greenhaus and Parasuraman (1993), Knight et al. (2003), Orpen (1981) and Whitehead et al. (1982) demonstrated that the attributions of success of both black and white individuals are subject to positive in-group bias. Positive ability attributions toward ingroup members could be suggestive of a self-serving bias; used to protect ones self-esteem by evoking empathy toward the group that one belongs. This is theorized by Kelley (1967) and Regan and Totten (1975) who shared the view that an observer’s emotional disposition toward an actor is essential to the orientation of the attribution. It also distally aligns with Charles and Littig (1982) and Jones and Nisbett (1987) who stated that the intrapersonal attributions of individuals are skewed toward the dispositional, which is relevant if observers are aligning themselves with the actor. Although Linville & Jones's (1980) polarized appraisal hypothesis purports that information on success has greater impact on outgroup than ingroup evaluations. This would mean that the complex schema that individuals hold on ingroups is weighed by positive and negative features of the ingroup, leading to more moderate evaluations and a realization that the outcome could be attributed to multiple causes.
Throughout this review we see stereotype-consistent and stereotype-contractive patterns of influence. Stereotype utility was greater in Whitehead et al.'s (1982) athleticism condition, where black participants attributed the performance of white actors more than that of black actors to the relative ease of the athletic task. The success of a white actor in the athletic condition created a juxtaposition between the result and the specious stereotype that white people have less athletic ability than black people, whereas failure in this remit would be congruent with said stereotype. Still, irrespective of race, the evocation of luck indices was more prevalent in the athleticism condition than the academic, suggesting that attributions are also task-relative. A similar stereotype consistency paradigm was also observed in Knight et al.'s (2003) study, where black leaders and white subordinates received aversive attributions when compared to white leaders and black subordinates. In both scenarios, stereotype-consistent contexts do not violate a priori beliefs as stereotype-contrastive contexts would. Both are in line with the social dominance orientation paradigm, as observers appeared to have a partiality toward preserving an imbalanced societal power distribution, rewarding white leaders for academic success and black subordinates for athletic success, yet penalizing the inverse. As a plausible alternative to the attribution theory, Ellis et al. (2006) proposed the expectancy-violation theory (Jussim, 1986) as a justification of this phenomenon, where stereotype-contrastive behaviors are evaluated more extremely in the direction of the violation. However, it has not yet been possible to support this theory empirically. In addition, successful managers have been found to receive higher evaluations, irrespective of race, when their performance was attributed to ability over luck (Ellis et al., 2006), but ineffective managers received higher ratings when their performance was attributed to luck, which contrasts the dominant position in the present review, but is congruent with some earlier literature (Knight et al., 2003; Whitehead et al., 1982).
Ellis et al. (2006) uniquely explicated on the basis of Weiner's (1980) causal attribution model with a focus on the locus of control. It was determined that two discrete attributional processes were occurring; the first, where success is perceived over failure; the second, where race was perceived. The influence of the former process was said to prevail over the latter process, to the extent that the influence of race was moderated and ran counter to view that negative race stereotypes drive negative attributions toward luck and for ability. One could submit that this illustrates a positive shift in the prejudicial attitudes of society and the possible attempts at restitution, or more cynically, a realization that occupational diversification is no longer optional. But Greenhaus and Parasuraman (1993) and Orpen's (1981) found that the effects of race on attribution career advancement prospects were often implicit and surreptitious, materializing by way of biased attributions of ability and performance evaluations – again, a ‘glass ceiling’ effect – such that black managers, male and female, were penalized twice. Prejudice in both the attributions of success (i.e., assessments of low ability and greater luck) and the evaluations of performance (i.e., harsher appraisals) meant black managers were seen as less promotable than their white counterparts.
Racial heterogeneity is expected to increase as more organizations come to realize the moral responsibility and competitive advantages of doing so (Jackson & Joshi, 2001). The shift toward equality will invariably lead to more ethnic individuals securing positions over white individuals, such that evaluations of success may become increasingly more vulnerable to implicit biases (Ellis et al., 2006). However, Greenhaus and Parasuraman (1993) found that negative attributions were less prevalent when the supervisor-supervisee relationship was stronger and longer-term. Thus, relationship and familiarity may prove to be an effective prescription to attenuate biases in interpersonal attributions of success. This temporal effect was said to be a result of the supervisor having sufficient information to come to an authentically impartial judgement on ability. They were also able to observe consistency in performance, such that it mediates the association. However, an alternative explanation is that this is merely a selection effect because greater lengths of employment suggest that the employee is liked.
[1] Ethnocentrism – using cultural preconceptions originating in the standards and customs of one's own culture, as a frame of reference to judge other cultures, practices, behaviours, beliefs, and people, instead of using the standards of that particular culture. Ethnocentrism may result in a dislike for the other cultures assessed from that viewpoint.
[2] Aversive racism is a subtle, yet insidious form of prejudice that emerges when discriminatory views can be justified by factors other than race where situational norms are ambiguous (Gaertner & Dovidio, 2000).