III. EVALUATIONS OF AFFIRMATIVE ACTION AND AAPs

Much of the psychological research on affirmative action has dealt with the antecedents of attitudes toward affirmative action, where attitudes can be defined as "evaluative judgments about particular objects, issues, persons, or any other identifiable aspects of the environment" (Baron & Graziano, 1991, p. 197). Presumably, by understanding, predicting, and changing attitudes toward AAPs, one can substantially increase the likelihood of understanding, predicting, and changing AAP-related behaviors (Ajzen & Fishbein, 1977, 1980; Fishbein & Ajzen, 1975; Kraus, 1995). Consistent with this assumption, AAP administrators and personnel officers have identified supportive attitudes among managers and employees as one of the most important factors in determining an affirmative action plan's effectiveness (Hitt & Keats, 1984; N = 31). For this reason we provide an extensive review of research on attitudes toward affirmative action and AAPs. We first describe the ways in which researchers have measured attitudes and have presented affirmative action stimuli. We then discuss research on structural influences, and finally turn to research on individual differences.

A. The Nature of Affirmative Action Attitudes

Researchers have measured attitudes in many different ways. In most studies, measures have been purely evaluative. Thus, respondents might be asked to agree or disagree with such statements as: "Affirmative action is a good policy" (Kravitz & Platania, 1993, p. 937), "After years of discrimination, it is only fair to set up special programs to make sure that women and minorities are given every chance to have equal opportunities in employment and education" (Jacobson, 1985, p. 310), and "Affirmative action programs that help Blacks and minorities to get ahead should be supported" (Kluegel & Smith, 1983, p. 804). Some researchers have obtained ratings of fairness rather than attitudes. Empirical research to be discussed reveals that fairness judgments and attitudes are very closely related. Thus, we treat the fairness ratings as surrogates for attitudes except when explicitly addressing the relation between the two constructs.

As explained below, attitudes are strongly influenced by the precise manner in which the AAP is described. To understand research results, therefore, it is important to know how affirmative action was described to the participants. Insofar as possible, we categorize AAP procedures as follows. This categorization does not always match the terms used by the authors of the research being discussed. In these definitions, decision can refer to any employment decision (e.g., selection, salary, promotion). Under strong preferential treatment, decisions are based solely or primarily on demographic status. For example, merit is not measured, an unqualified minority is selected, or the less qualified minority applicant is favored. Some authors refer to this condition as discrimination in reverse. In most cases quotas would fall in this category because they require the selection of a certain number or proportion of minorities regardless of qualifications. Under preferential treatment (neither weak nor strong), decisions are based on both merit and demographic status, with the relative weighting left unspecified. The preferred minority is not said to be unqualified. Under weak preferential treatment, decisions favor the more qualified applicant unless qualifications are equivalent, in which case the minority applicant is favored. Under merit, decisions are said to be based solely on merit, which presumably is an indication of the individual's ability to perform well. When procedures attempt to eliminate discrimination, they include actions designed to do away with existing barriers to success. Compensatory procedures include other actions designed to help minorities (e.g., training, career guidance), but decisions are based solely on merit. Diversity procedures include efforts to increase the diversity of the workforce (e.g., through recruitment), but decisions are based solely on merit

B. Structural Influences

Much of the research on affirmative action attitudes has emphasized the importance of structural features of AAPs. This research has frequently included a manipulation of some aspect of an AAP. The underlying assumption is that reactions to an AAP will depend on details of the plan, particularly on the weighting of demographic status. This research is relevant to the public debate about affirmative action because much of that debate involves disagreement about what affirmative action means (e.g., preferential treatment versus assurance of equal opportunity). After discussing this research we will turn to work on other structural factors, including the identity of the target group, the organizational setting, and the need for affirmative action.

1. Qualifications of Selected Candidate

In selection situations, fairness ratings are affected by the qualifications of the selected and rejected candidates. Nacoste (1985, 1987; Nacoste & Lehman, 1987) created scenarios in which two people were competing for a fellowship, and asked respondents to play the role of the winning candidate. Fairness ratings were higher when the more qualified candidate was selected. Using a similar experimental procedure, Arthur, Doverspike, and Fuentes (1992) found that fairness ratings were higher when qualifications of the selected minority candidate were equal rather than inferior to qualifications of the rejected majority candidate. In Mann and Fasolo (1992), in contrast, the effect of qualifications on ratings of decision fairness did not quite attain traditional levels of statistical significance (p < .08). In Gilliland and Haptonstahl (1995), fairness ratings were highest when the better qualified majority candidate was selected, intermediate when the majority and minority candidates had equal qualifications (regardless of who was selected), and lowest when the poorer qualified minority candidate was selected. In Heilman, McCullough, and Gilbert (1996), fairness ratings were higher when qualifications of the selected female candidate were equal or superior to those of the rejected male candidate than when they were inferior. Singer (1990, 1992) provided more precise tests of the importance of candidate qualifications by manipulating the test scores of the selected and rejected candidates. Fairness ratings were monotonically related to the difference in scores. That is, the larger the superiority of the selected candidate, the higher the fairness rating. Contrariwise, the larger the superiority of the rejected candidate, the lower the fairness ratings. The relative impact of the difference in qualifications was largest when the difference was smallest (i.e., when qualifications of the two candidates were almost identical); the fairness rating function was steepest near the zero point.

In summary, research on qualifications clearly reveals that fairness ratings increase along with the superiority of the selected over the rejected candidate. This research relates to the work on AAP structure because underqualified candidates can only be selected if demographic status is given positive weight. Thus, we now turn to research that has directly addressed the effect of AAP structure on evaluations of the affirmative action plan.

2. Structure of the AAP (Weighting of Demographic Status)

a. Conceptualizations

The underlying assumption of research on AAP structure is that an individual's understanding of what affirmative action entails will influence his or her attitude. This point has been made by many theorists and researchers, and recently has been developed by Barnes Nacoste (1994; Nacoste, 1994, 1995). Briefly, Nacoste argued that people have cognitive (policy) schemas that incorporate their beliefs about affirmative action. Beliefs about what constitutes a typical procedure are critically important, with beliefs about the use of universalistic and particularistic contributions playing a central role. Universalistic contributions include merit and other capacities that will influence performance. Particularistic contributions include individual attributes that may be taken into consideration but will not influence performance. Race and ethnicity are the most relevant particularistic contributions in the context of affirmative action. The individual's policy schema will strongly influence his or her reactions to affirmative action; reactions will become increasingly negative as the anticipated weighting of particularistic contributions increases.

Nacoste (1994, 1995) used a combination of interdependence theory and procedural justice theory to explain reactions to affirmative action. He argued that reactions to an AAP (or to the individual's affirmative action policy schema) are based on three comparison level standards. Weighting of demographic status plays a key role in these comparisons. Observers compare the enacted procedure to: (a) the procedure that gives too much weight to group membership, (b) the set of realistic alternative procedures the observer considers to be superior, and (c) the procedure that precisely counteracts the effects of discrimination. If an individual believes that an AAP gives more weight to demographic status than is appropriate, he or she will consider the AAP to be unfair, will dislike the AAP, will not be attracted to the organization, and will stigmatize people who are selected under the AAP. Note that these reactions will be based on the individual's affirmative action policy schema. The validity of the policy schema is irrelevant; perceptions determine reactions.

Other researchers have suggested that AAP structure influences attitudes because it has implications for personal and collective self-interest (e.g., Kravitz, 1995; Veilleux & Tougas, 1989). That is, the greater the weighting of demographic status, the more the procedure will help individuals in that demographic group and hurt individuals in other demographic groups. We discuss competitive tests of the theories elsewhere in this review. We now review the empirical research on AAP structure.

b. Empirical Research

(1) Knowledge of Affirmative Action

The underlying assumption of research on AAP structure is that an individual's understanding of what affirmative action entails will influence his or her attitude. Thus, we begin with a brief review of research on knowledge of affirmative action.

Goldsmith, Cordova, Dwyer, Langlois, and Crosby (1989) interviewed 62 women associated with a liberal arts college for women shortly after an announcement that the college would become an affirmative action institution. They obtained significant positive correlations between two measures of knowledge and two measures of attitude. Stout and Buffum (1993) surveyed 193 members of the Texas Chapter of the National Association of Social Workers. Both positive and negative experiences with affirmative action were positively related to self-reported knowledge about affirmative action and to commitment to affirmative action, but knowledge did not correlate significantly with commitment. In a study to be discussed in more detail later, Bell (1996) surveyed the attitudes of 610 participants, half of whom were full-time employees in a variety of firms. Bell reports that self-reported knowledge of the details of affirmative action policies was related to more negative attitudes. Differences in the measures of knowledge might account for the different results of these three studies; at this time no clear conclusion can be drawn about the relation between knowledge of affirmative action and attitudes.

There is some evidence that the public has a poor understanding of affirmative action, and some argue that opposition to affirmative action is based in part on this misunderstanding (e.g., Crosby, 1994; Eberhardt & Fiske, 1994). Kravitz and his colleagues (Kravitz & Platania, 1992, 1993; Kravitz, Stinson, & Mello, 1994) presented their participants with 10 to 25 potential components of an AAP. Respondents in Kravitz and Platania (1992, 1993) were undergraduates; those in Kravitz et al. (1994) were community residents contacted via telephone. All three samples were multi-racial (White, Black, and Hispanic). The potential AAP components included several questions dealing with preferential treatment, compensatory actions, diversity efforts, the elimination of discrimination, and other actions. Participants evaluated each component and estimated the likelihood that it would be included in an AAP. The likelihood ratings revealed that respondents had a less than perfect understanding of what affirmative action entails. For example, although affirmative action regulations emphasize recruitment and the elimination of discrimination, these potential components received neutral likelihood ratings. Respondents also did not know which organizations are required to have AAPs. Kravitz (1994) briefly reported the results of content analyses of open-ended definitions of affirmative action provided by respondents in Kravitz and Platania (1993), Kravitz et al. (1994), and a third unpublished study that sampled employees of a firm without an AAP. He reported that approximately 40% of the respondents were completely unfamiliar with the concept of affirmative action. Many of those who claimed familiarity provided vague or inaccurate definitions.

In 1989, Pace and Smith (1995) surveyed 1,075 municipal and county chief financial officers, and asked them to identify which of three statements most closely matched their understanding of federal affirmative action requirements. Almost all respondents believed the requirements involved weak preferential treatment (48.3%) or recruitment (43.8%); only a few respondents (7.8%) thought the requirements involved strong preferential treatment.

A poor or undeveloped understanding might imply the possibility of attitude change in response to new information. This possibility was addressed by Fletcher and Chalmers (1991), who questioned a representative sample of Canadian citizens and decision makers about their affirmative action attitudes. This study is discussed in more detail later. For the moment, we note that more than half the citizen respondents indicated their opinions would change when given an opposing argument. Consistent with this result, Kinder and Sanders (1990) categorized their respondents as less- or more-informed about public affairs. They found that report that the manner in which affirmative action was framed generally had a larger influence on reactions to a variety of questions among the less-informed respondents than among the more-informed respondents.

In summary, although theorists and others (e.g., Collison, 1992) have argued that public opposition is based in part on a poor understanding of what affirmative action entails, empirical research is limited. The studies of public beliefs have all been reported by a single researcher who obtained the data in a single geographical area, which suggests that these results need to be replicated. Other research suggests that attitudes toward affirmative action are related to knowledge about affirmative action and that people will change their support when given information inconsistent with their initial opinion. There is a clear need for more research on public knowledge of affirmative action and how that knowledge relates to attitudes. Research on reactions to specific AAPs provides more information about the relation between knowledge and attitudes. We shall turn to that research after briefly reviewing some conclusions drawn from opinion polls.

(2) Opinion Polls

Lipset and Schneider (1978) examined nearly 100 opinion polls completed between 1935 and 1977. They concluded that Americans strongly support equality of opportunity and the elimination of discrimination, moderately support compensatory action (e.g., extra training for minorities), and oppose preferential treatment and the use of quotas. Sigelman and Welch (1991) provided an analysis of many national surveys of attitudes toward affirmative action. They concluded that both Blacks and Whites support equal opportunity and affirmative action in general, and oppose preferential treatment and quotas. In all cases, support was somewhat higher among Blacks than Whites. Lynch and Beer (1990) cited public opinion polls, Lynch's in-depth interviews, and other commentators to make the point that Whites dislike affirmative action that involves quotas or preferential treatment, though they support equality and compensatory action.

Kluegel and Smith (1983) report data from a representative sample of 1,596 English-speaking Whites in the U.S. They reported support for affirmative action programs that "help Blacks and other minorities to get ahead," with 76% agreeing or strongly agreeing. Opinions were divided on the use of quotas in colleges and universities (59% agree or strongly agree) and in businesses (51% agree or strongly agree). Two-thirds of the respondents felt that such preferential treatment for Blacks is unfair. Fine (1992a, 1992b) analyzed opinion survey data from a 1986 national probability sample of U. S. adults (319 Blacks and 1,790 Whites). Support for preferential hiring and promotion of Blacks varied with question phrasing. It ranged from 11% to 24% among Whites, and from 61% to 75% among Blacks. In addition, 46% of Black respondents agreed that it is not the government's job to guarantee equal opportunity, and 43% agreed that Blacks should overcome prejudice and work their way up without any special favors.

In summary, opinion polls results revealed that White Americans strongly support equality and the elimination of discrimination, but oppose preferential treatment. Reactions to compensatory actions are less clear. Polls of Black Americans revealed somewhat higher support for preferential treatment, but a clear preference for equal opportunity. Experimental research provides more detailed information about reactions to specific AAPs, and is described below.

(3) Experimental Research

Most laboratory studies have been limited to two or three different types of AAPs, and we begin by reviewing that research. Except where otherwise noted, all the following studies used undergraduates as respondents.

Many studies have found that people prefer merit-based decisions to any type of preferential treatment. Research on French Canadian women (Tougas & Veilleux, 1988) and on French Canadian managers and professional men (Tougas & Veilleux, 1989; Veilleux & Tougas, 1989) revealed a preference for the elimination of discrimination over weak preferential treatment. Brutus and Ryan (1994) found that female undergraduates rated merit-based selection more positively than preferential treatment based on gender. In a third condition the experimenter stated that selection was based on merit, but an experimental confederate argued that preferential treatment was being used. Respondents also evaluated this condition lower than merit-based selection. In Nacoste and Hummels (1994), 186 undergraduates rated the fairness of a weak (equal opportunity) or strong (pressured to hire more women) AAP. Ratings of these plans did not differ significantly. In Hattrup (1994), 266 female undergraduates evaluated merit selection more positively than strong preferential treatment based on gender. In addition, fairness ratings of merit selection (but not of strong preferential treatment), were positively related to respondent self-efficacy. Heilman, Simon and Repper (1987) found that 140 male and female undergraduates considered merit selection fairer than strong preferential treatment. Heilman et al. (1996) replicated this difference in a sample of male undergraduates. The effect was moderated, however, by information about qualifications. The difference was not significant if the woman selected in the preferential treatment condition was more qualified and therefore would also have been selected on the basis of merit. In Experiment 1 of Heilman, Rivero and Brett (1991), using 60 male and female undergraduates, fairness ratings varied with selection procedure, respondent gender, and the Procedure X Gender interaction. Men and women responded equally positively to merit selection, but men responded more negatively than women to strong preferential treatment. In Experiment 2, 36 female undergraduates evaluated merit-based selection more positively than strong preferential treatment, regardless of whether they were told their scores on the pretest. Singer (1996) obtained ratings of merit selection and preferential selection from 312 adult residents in New Zealand. He found that 143 preferred merit selection and only 19 preferred preferential selection; the remaining 150 did not evaluate the two programs differently. In sum, these studies reveal that people generally evaluate merit selection more positively than preferential treatment. Evaluations of merit selection are almost universally positive, but evaluations of preferential treatment are moderated by such factors as respondent gender, self-efficacy, and knowledge about application qualifications.

The importance of merit is further revealed in a study of 96 male francophone managers in Canada (Joly, Pelchat, & Tougas, 1993). The managers evaluated a weak preferential treatment procedure, in which gender was to be considered only if competing candidates had equivalent merit. The description of the procedure varied. In one condition the procedural description emphasized the important role of merit; in a second condition respondents were given information about the percentage of women in the workforce; in the third condition no additional information was given. Evaluations were most positive when merit was emphasized. Thus, evaluations of weak preferential treatment depend on how it is described. This study did not include a pure merit procedure, so it was not possible to know whether emphasizing the role of merit in weak preferential treatment will increase evaluations to the level of a merit procedure alone. Tougas, Crosby, Joly, and Pelchat (1995) reported two similar studies of male francophone Canadian managers and professionals. In Study 1 (N = 96) , the affirmative action conditions included a control condition (details not described), statistics condition (included data demonstrating women's underrepresentation) and a merit condition (emphasized role of merit in affirmative action selection decisions). Study 2 (N = 131) included the control and merit conditions, a third condition describing the weighting of gender in selection decisions, and a fourth condition providing the ultimate numerical goal of the program. Analyses revealed no effects of condition on attitudes. Although condition did affect ratings of fairness, interpretation of this effect is complicated because the fairness measure included wording identical to that used in the description of the merit condition.

Several papers compared strong preferential treatment to a weaker version of preferential treatment in which the precise weighting of merit and demographic status was not specified. In three studies (Nacoste, 1985, 1987; Nacoste & Lehman, 1987; Ns = 96, 106, and 106, respectively), female undergraduates rated fairness of the selection procedure and decision more negatively when the procedure involved strong preferential treatment than when it involved preferential treatment. In Arthur et al., (1992), respondents in the strong preferential treatment condition rated the procedure as less fair than those in the preferential treatment condition. In summary, these studies revealed that evaluations of preferential treatment are inversely related to the weighting of demographic status.

In a series of three experiments, Matheson, Echenberg, Taylor, Rivers, and Chow (1994) studied women's reactions to four AAPs that varied in the weight given to demographic status. In Study One, 60 undergraduates gave positive evaluations to the elimination of discrimination, negative evaluations to weak preferential treatment, and very negative evaluations to strong preferential treatment and complete discrimination. In Study Two, 43 undergraduates evaluated the elimination of discrimination positively and the three versions of preferential treatment equally negatively. In Study Three, 19 police trainees responded negatively to a "no action" program and equally positively to four AAPs. The lack of significant differences among the AAPs may have been due to the low statistical power due to the tiny sample size. It is interesting to note that all AAPs receive negative evaluations in Studies 1 and 2 and positive evaluations in Study 3; this might be due to the different populations from which the samples of respondents were drawn.

Nosworthy, Lea, and Lindsay (1995) asked 192 non-Black Canadian undergraduates to evaluate four AAPs that could increase enrollment of Blacks at their university. Endorsement was strongest for the use of advertising, next for the use of additional fellowships for Blacks, weaker for the use of quotas, and weakest for the use of lower standards for Black students than for White students.

In Summers (1995), 80 male and female continuing education students evaluated three possible AAPs. A special training program received positive evaluations. Quotas received slightly negative evaluations. Differential scoring of selection tests (which would result in strong preferential treatment) received negative evaluations. All differences among the AAPs were statistically significant.

White and Hispanic undergraduates evaluated eight different AAPs targeted at Black employees in Kravitz (1995). Respondent evaluations to two plans, one of which involved the elimination of discrimination and the second of which involved the elimination of all information about race, were quite positive. Respondents gave less positive but still favorable evaluations to a compensatory procedure (training), a diversity plan (recruitment), and a plan that involved proportional hiring based on the number of qualified applicants (which would result in a satisfactory utilization analysis in an OFCCP report). They gave neutral evaluations to a plan that was equivalent to the proportional hiring plan, but used the loaded term "quotas" in the description. Their evaluations of this quota plan were not significantly different from their evaluations of the recruitment, training, or proportional hiring plans. Finally, respondents negatively evaluated weak and strong preferential treatment, and these evaluations were lower than those of the other six plans. These results are generally consistent of those previously reported. They confirmed opinion poll results indicating that compensatory procedures are positively evaluated, but they also showed that compensatory procedures were liked less than the elimination of discrimination.

In the research by Kravitz and his colleagues discussed above (Kravitz & Platania, 1992, 1993; Kravitz et al., 1994), participants evaluated 10 to 25 potential components of an AAP and rated the likelihood that an AAP would actually include the components. Participants also reported their attitudes toward affirmative action, so the relation between beliefs and attitudes could be assessed. Results varied somewhat across the studies. In general, respondents favored equal opportunity, the elimination of discrimination, proportional hiring based on the availability of qualified applicants, training, recruitment, the targeting of organizations with histories of discrimination, and the provision of employment information to the federal government. They opposed the hiring of unqualified applicants, proportional hiring that ignored qualifications, and all versions of preferential treatment. Confirming the underlying premise of the research, likelihood ratings predicted attitudes. Attitudes were most strongly related to beliefs about the selection of unqualified applicants, preferential treatment, and proportional hiring that ignored qualifications.

Bell, McLaughlin, and Harrison (1996) reported results of three studies of attitudes toward AAPs. In the first study they elicited salient (cognitively accessible) attributes that 129 students and managers thought were associated with AAPs. Content analysis produced 14 features that comprised the top 90% of elicited attributes, including AAP purposes designated by the original Executive Order (e.g., "create job opportunities for minorities and women," "remove discrimination") as well as other issues that reflect the current national debate (e.g., "restrict a business's freedom to make decisions," "operate as a quota system," "increase the diversity of the workforce"). In the second study, these salient attributes were used to explain AAP attitudes using Fishbein and Ajzen's (1975) theory of reasoned action. The sample of undergraduate students was diverse in terms of gender (63 male and 61) and ethnicity (66% White, 12% Black, 8% Hispanic, 12% Asian, and 2% Native American backgrounds); 78% were currently employed. As hypothesized, AAP attitudes fit an additive function of the products of (a) the strength with which the respondents associated each salient attribute with AAPs in general, and (b) and how positively or negatively they evaluated that attribute. The third study paralleled the second in both procedures and results, but involved 138 male and 64 female managers (74% White, 11% Black, 7% Hispanic, and 8% Asian backgrounds) sampled from a variety of firms and industries.

In a related study, Bell (1996) examined the relationships among attribute-based beliefs, evaluations, and attitudes toward AAPs. The respondents (N = 610) were diverse in terms of employment (50% were students and 50% were full-time employees in various firms and industries), gender (55% male) and ethnic background (60% White, 10% Black, 12% Hispanic, 11% Asian, and 7% Other). Attitudes were again found to be a multiplicative function of beliefs about AAPs and evaluations of those attributes.

Konrad and Linnehan (1995a) interviewed several HRM professionals who specialize in EEO/AA programs. Based on these interviews, they developed a list of 54 identity-conscious actions and 63 identity-blind actions that organizations can take. They then obtained information from 138 companies about their use of these actions. On average, these firms used 37% of the identity-conscious actions and 58% of the identity-blind actions. In both cases, the number of actions used correlated positively with top management support for EEO/AA. Konrad and Linnehan (1995a) listed all 117 actions, but did not provide frequency data for each individual action.

Konrad and Linnehan (1995b) obtained evaluative ratings from 242 managers in four medium to large firms in the Philadelphia area. They evaluated 26 of the actions listed by Konrad and Linnehan (1995a). A factor analysis revealed two factors: 4 items dealt with identity-blind actions and 18 items involved identity-conscious actions. Identity-blind actions were evaluated positively. Reactions to the identity-conscious actions varied with subject group, being most liked by people of color, next by White women, and least by White men. The mean was neutral for White men, and significantly positive for the other groups.

A similar study was reported in Davis and West's (1984) survey of 403 municipal personnel administrators in 1979. Their study included multi-item scales measuring support for equal opportunity and affirmative action principles. Results revealed greater support for equal opportunity principles than affirmative action principles. Support for specific actions appeared to be highest for special recruitment efforts, followed by weak preferential treatment when minorities are under-represented, merit hiring rather than quotas, and lowest for special training to prepare minorities for recruitment. Support for these four actions varied significantly with the administrator's general support for EEO versus affirmative action.

(4) Conclusions

There is greater support for the principle of equal opportunity than for the principle of affirmative action. Evaluations (i.e., attitudes and fairness judgments) of affirmative action are strongly influenced by actual or presumed AAP structure; they are inversely related to the weighting of demographic status in decision making. Although people support compensatory actions and diversity efforts, they prefer to limit affirmative action to the elimination of discrimination. There is some evidence that there is a great deal of variance in the understanding of what affirmative action entails, and some have argued that this partly underlies some of the public opposition to affirmative action. Given the relation between beliefs and attitudes, and the open question about public understanding, there is a need for additional research on beliefs and evaluations of specific actions that can be incorporated in EEO/AA plans. This research could take advantage of Konrad and Linnehan's (1995a) listing of EEO/AA actions.

c. Possible Mediators of Effect of AAP Structure on Attitudes

Given that attitudes toward AAPs are strongly influenced by details of the AAP, the next question to address is why this influence exists. Theoretical explanations have focused on two possible mediating variables -- fairness and self-interest. We now turn to research dealing with these variables.

(1) Perceived Fairness

(a) Introduction

Although affirmative action is designed to ensure macrojustice (justice between groups), opposition frequently is based on concerns about microjustice (justice for individuals) (Clayton & Tangri, 1989). Typical arguments are that affirmative action results in "reverse discrimination" (e.g., Burstein, 1985), that affirmative action penalizes young White men who were not responsible for discrimination (e.g., Glasser, 1988; Groarke, 1990), and that affirmative action forces organizations to "change the rules in the middle of the game" (e.g., Crosby, 1994). Heilman (1994) pointed out that reactions of nontarget group members will be especially negative if they believe affirmative action has caused them to be denied employment outcomes to which they were entitled. Nacoste (1987) reported results consistent with his hypothesis that one reason for negative reactions to strong AAPs is that they imply the organization is not committed to fairness. In short, much of the debate about affirmative action has centered on the issue of fairness.

Consistent with this emphasis, many researchers have implicitly or explicitly assumed that attitudes toward affirmative action are driven by fairness judgments, and have assessed perceptions of fairness rather than attitudes per se. This work was discussed above, and revealed similar effects involving attitudes and fairness judgments. The following review begins by summarizing three studies that qualitatively explored the importance of fairness perceptions in peoples' understanding of affirmative action. We then turn to assessments of both perceived fairness and attitude.

(b) Qualitative Research on Fairness Perceptions

Three studies used content analyses of interview data determine how people think about the fairness of affirmative action. Edelman, Erlanger, and Lande (1993) interviewed 10 management personnel who handled discrimination complaints in 10 large companies. These administrators construed EEO/AA most often in terms of fairness, primarily procedural fairness, rather than in legal terms. This focus on fairness is especially meaningful because these administrators were responsible for ensuring that their organizations complied with antidiscrimination laws.

Ayers (1992) interviewed 13 "women of color" who she knew had been involved in affirmative action programs. Virtually all their statements about affirmative action in principle focused on fairness rather than unfairness. Of their statements concerning affirmative action in practice, however, approximately 40% dealt with issues of unfairness. In terms of Leventhal's (1980) justice model, more statements dealt with consistency than with any other justice rule. The distinction between affirmative action in principle and in practice is consistent with the results of numerous surveys indicating public support for the elimination of discrimination, but opposition to most programs designed to attain that end (e.g., Lipset & Schneider, 1978).

In a telephone survey of 68 community residents and a questionnaire study with 86 undergraduates, Kravitz and Van Epps (1995) asked respondents why they considered affirmative action to be fair and/or unfair. Explanations of why affirmative action was fair were most likely to include statements about equal opportunity. Explanations of why affirmative action was unfair were most likely to include statements about reverse discrimination and decision making based on demographics.

In summary, this research revealed that individuals do think about affirmative action in terms of fairness. This, along with the emphasis on procedural fairness, equal opportunity, and reverse discrimination, is consistent with the public debate on affirmative action.

(c) Research on the Relation Between Attitudes and Perceptions of Fairness

As mentioned above, many of the studies on AAP structure examined its relationship with fairness judgments rather than attitudes. This emphasis on fairness may be based on the assumption that perceptions of fairness mediate attitude formation. Surprisingly, we know of only four studies that have included measures of both fairness and attitude. Ozawa, Crosby, and Crosby (1996) and Tougas, Crosby et al. (1995) used both fairness and attitude as dependent variables; Taylor-Carter, Doverspike and Alexander (1995) used attitude as a predictor of fairness ratings; Kravitz (1995) used fairness ratings as a predictor of attitude.

In two studies, Tougas, Crosby et al. (1995) studied reactions of male francophone Canadian managers and professionals to three or four different AAPs. Ratings of attitudes and fairness correlated significantly in both studies (.67 and .68). Ozawa et al. (1996) asked undergraduate students in the U. S. (N = 53) and Japan (N = 65) to evaluate an AAP (minimum standards followed by quota hiring) on goodness (attitude) and fairness scales. The attitude-fairness correlation was .56 in the Japanese sample and .60 in the U. S. sample.

In Taylor-Carter, Doverspike and Alexander (1995), 128 female undergraduates initially read a definition of weak preferential treatment and reported their attitudes toward it. They then read a favorable or unfavorable message and recorded their relevant cognitions. They ended by rating the fairness of preferential treatment as initially defined. Regression analysis revealed that fairness ratings were related as expected to initial attitude, message, number of positive cognitions, and number of negative cognitions. The correlation between fairness ratings and initial attitudes was .71.

The assumption that attitudes toward affirmative action are based on fairness perceptions was tested by Kravitz (1995), who contrasted the fairness model with alternative models based on self-interest, racism, and political ideology. Kravitz manipulated AAP and measured the remaining variables. The target group was Blacks, and respondents were non-Black undergraduates. The manipulation of AAP influenced judgments of self-interest, fairness, and attitudes. The effect of the manipulation on attitudes was completely mediated by the fairness judgments, but not by ratings of personal or collective self-interest. The correlation between fairness ratings and attitude was .86.

In summary, these studies revealed that fairness perceptions and attitudes toward affirmative action are closely related. Given the correlational nature of the observed relations, it is impossible to determine the causal sequence. Finally, in all six samples this relation was assessed at the operational level -- it dealt with specific AAPs rather than with the general policy of affirmative action. There may also be a positive relation between attitudes and perceived fairness of affirmative action in principle, but this relation should be empirically tested.

(d) Discussion

Qualitative research shows that individuals think of affirmative action in terms of fairness. Individuals distinguish between affirmative action in principle and in practice, are sensitive to issues of procedural justice, and focus on implications for equality of opportunity and reverse discrimination. Research on the structure of AAPs revealed that ratings of fairness are inversely related to the weighting of demographic status, and that attention to status at preliminary stages (e.g., recruitment) is considered fairer than attention to status in final selection decisions. Selection decisions are considered fairer when candidates with higher qualifications are selected. Finally, attitudes toward an AAP are positively related to subsequent fairness judgments of the AAP, and effects of an AAP manipulation on attitudes are mediated by fairness judgments.

Some of these results must be qualified. As will be discussed in more detail later, several studies obtained interactions involving affirmative action procedure, respondent gender, and/or respondent race (Heilman et al., 1991; Arthur et al., 1992; Doverspike and Arthur, 1995; Gilliland and Haptonstahl, 1995). Same-gender favoritism in evaluating the fairness of a promotion decision was reported by Saal and Moore (1993), and such in-group favoritism is also likely to occur in evaluations of affirmative action. Singer (1992) obtained evidence of egocentric bias; ratings of fairness depended on whether the respondent acted as an observer (N = 222) or role-played the majority candidate (N = 112). In short, reactions to specific AAPs vary as a function of respondent gender and race/ethnicity. It is apparent that White males respond more negatively than do other types of respondents to AAPs not based purely on merit. These interactions may have been due to the effects of self-interest, which is addressed in the following section. Despite these qualifications, it is clear that fairness judgments and attitudes are closely related, and both are influenced by AAP structure.

Additional qualifications are suggested by research on individual and situational determinants of attention to fairness. Gilliland and Haptonstahl (1995) found that fairness judgments were most closely related to perceptions of equity when equity was clearly satisfied (superior candidate is selected) or clearly violated (inferior candidate is selected). Fairness judgments were most closely related to perceptions of equality when decisions did not clearly satisfy or violate equity (candidates had equal qualifications). Overall fairness ratings were significantly related to perceptions of need only when need was clearly violated -- when the equally qualified minority candidate was not selected. In related research, Rasinski (1987) developed a scale to assess individual differences in the conceptualization of fairness. Whereas some people endorse the principle of proportionality, others endorse egalitarianism. These two studies suggest that perceptions of fairness and the importance of fairness will vary with individual differences and with the situation.

Finally, with few exceptions (Ayers, 1992; Edelman et al., 1993; Kravitz & Van Epps, 1995; Tougas, Crosby et al., 1995), affirmative action studies that have dealt with fairness judgments relied almost entirely on undergraduate respondents. Several studies involved role-play scenarios (Arthur et al., 1992; Doverspike & Arthur, 1995; Nacoste, 1985, 1987; Nacoste & Lehman, 1987), and some asked respondents to play an unfamiliar role. There is a need for additional research on employees in organizations with AAPs.

(2) Implications for Personal and Collective Self-interest

A number of studies have revealed more positive attitudes among individuals who feel their personal interests are being served by affirmative action programs than among those who feel disadvantaged or threatened. For example, significant correlations of attitudes toward AAPs and self-interest were obtained in a 1978 national survey of 1,584 Whites (Jacobson, 1985) and a sample of over 450 employees in a large Canadian firm (Tougas & Beaton, 1993). Similarly, in a study of 133 Hispanic undergraduates, Kravitz and Meyer (1995) found that attitudes toward AAPs targeted at Hispanics correlated significantly with rated implications for individual but not collective self-interest. Attitudes toward affirmative action are clearly affected by implications of affirmative action for the respondent's self-interest.

Other research dealt with more complex aspects of self-interest. Kluegel and Smith (1983) distinguished between simple economic self-interest and cooperative self-interest in their national survey. Cooperative self-interest refers to the belief that AAPs designed to help others will indirectly help oneself. Results showed that both simple and cooperative selfinterest were significantly related to attitudes toward programs to hold a certain number of positions for qualified Blacks in colleges and in business.

Tougas and her colleagues extended the concept of self-interest from the personal to the collective level, where collective self-interest refers to anticipated effects on the respondent's demographic group. Tougas, Beaton, and Veilleux (1991) studied 197 women in a large Canadian firm, and included measures of both personal and collective self-interest. They reported that the relationship between personal self-interest and attitudes was moderated by two intervening variables: collective self-interest and collective relative deprivation (to be discussed shortly). Tougas, Brown, Beaton, and Joly (1995) studied male students and male employees in a Canadian firm with a strong AAP targeted at women. They reported that attitudes toward the organization's AAP and affirmative action in general were related to collective self-interest.

Three final studies addressed the question of whether the effect of AAP structure on attitudes was mediated by self-interest. Kravitz (1995) found that personal and collective self-interest partly mediated the effect of AAP structure on attitudes. Nosworthy et al. (1995) found that measures of personal and collective self-interest correlated strongly with attitudes toward each of four AAPs. However, when regression equations included measures of racism and fairness along with self-interest, there was only one significant effect of personal self-interest and none of collective self-interest. Both Kravitz (1995) and Nosworthy et al. (1995) found that attitudes were more closely related to judgments of fairness than to judgments of personal and collective self-interest. Summers (1995) found that a significant gender difference in attitudes toward affirmative action in general was fully mediated by anticipated effects of affirmative action on the respondents' careers.

In summary, attitudes toward specific AAPs correlate with implications of the AAPs for the respondent's personal and collective self-interest. Collective self-interest appears to be less important than personal self-interest, and both are less important than perceptions of fairness. The effect of AAP structure on attitudes is at least partly mediated by judgments of self-interest.

d. Conclusions

Evaluations of affirmative action and AAPs are strongly influenced by actual or presumed AAP structure; they are inversely related to the weighting of demographic status in decision making. In addition, people respond more positively to selection situations if the more qualified candidate is selected, regardless of minority/majority status. Evaluations of merit selection are almost universally positive, but evaluations of preferential treatment are moderated by such factors as respondent race, gender, and self-efficacy. Although people support targeted recruitment and training, they prefer to limit affirmative action to the elimination of discrimination. These reactions appear to be mediated by perceptions of AAP fairness and implications of the AAP for the respondent's personal and collective self-interest, with fairness being more important.

3. Attempts to Influence Evaluations

Organizations with AAPs are required by law to communicate that fact. They have considerable freedom, however, in how they justify the existence of the AAP and in how much detail they describe the AAP. This information could influence reactions to the program. More generally, attitudes toward the social policy of affirmative action may be influenced by information provided to the public about affirmative action. That conclusion would follow from the preceding review of research on AAP structure.

Gamson and Modigliani (1987) reviewed public discussions of affirmative action in terms of discourse theory. They found that affirmative action had been conceptualized in three ways: remedial action, delicate balance, and preferential treatment. There were four versions of the preferential treatment conceptualization, with the most common being reverse discrimination. The relative emphasis on these versions is determined by sponsor activities (e.g., administration announcements), media practices, and cultural resonances. While the remedial action conceptualization was most popular in early years, by the mid-1980s it had been overtaken by preferential treatment. Interestingly, although delicate balance probably best represents the public's reactions to affirmative action, it has received little attention in public discourse. This neglect is probably due to the lack of a sponsor. Since publication of this article, we would argue that a fourth conceptualization of affirmative action has been added -- affirmative action as a source of cultural diversity (e.g., Levi & Fried, 1994; Thomas, 1990).

Fine's (1992a) survey results relate to Gamson and Modigliani's (1987) analysis. She found that White respondents' opposition to preferential treatment was stronger when the reason for opposition was phrased as "it discriminates against Whites" than when it was phrased as "it gives Blacks advantages they haven't earned." She reported a nonsignificant trend in the same direction among Black respondents. The difference in statistical significance may have been due to the fact that the sample included 1,790 Whites but only 319 Blacks.

In summary, affirmative action is conceptualized and described in various ways, and this is likely to influence attitudes. Most relevant research has dealt with provision of structural details or justification, and we now turn to that work.

a. Providing Information About AAP Details

Fletcher and Chalmers (1991) sampled attitudes of Canadian citizens and decision makers. The citizen sample included 2,084 people obtained in a 1987 nationwide phone interview project using random digit dialing. The government decision-maker sample included 513 elected legislators, 483 members of the executive branch (charged mainly with enforcing the federal affirmative action laws), and 352 senior lawyers in the judiciary. Fletcher and Chalmers asked their respondents about their attitudes toward AAPs targeted at improving the number of women or the number of French-Canadians (an ethnic minority) in top government positions. They found that providing Canadians with arguments contrary to their initial positions caused most of them to say they would change their positions. After being given an argument for the opposite position (i.e., "quotas mean not hiring the best person" or "lack of quotas means that the target group remains economically unequal"), respondents indicated whether they would change their opinion. Of the citizens who initially opposed quotas, 49% said they would feel differently; of the citizens who initially supported quotas, 64% said they would feel differently. Expressed willingness to change was somewhat lower among the decision makers, being 39% and 43%, respectively.

Bell (1996), in a study discussed previously, provided positive, neutral, or negative pieces of information about AAPs to her 610 respondents. For Whites only, the effect of information on AAP attitudes was in the hypothesized direction. Negative information either strengthened or had no effect on the already positive attitudes of Asians, Hispanics, and Blacks toward AAPs.

Taylor-Carter, Doverspike and Alexander (1995) used cognitive response theory to predict effects of pro- and anti-affirmative action messages on ratings of AAP fairness. After reporting their attitudes toward weak preferential treatment, respondents read the message, recorded their thoughts about the message, predicted the importance of affirmative action for their lives, and rated the fairness of preferential treatment. Message polarity influenced the number of positive thoughts in the predicted direction, but did not affect the number of negative thoughts. There were no effects of involvement, contrary to prediction. Most importantly, regression analysis revealed that fairness ratings were related as expected to initial attitude, message polarity, number of positive thoughts, and number of negative thoughts. This indicates that perceived fairness of preferential treatment (and, presumably, other affirmative action procedures) can be affected by pro- or anti- affirmative action messages and the thoughts they stimulate.

In sum, these studies indicate that attitudes toward AAPs can be changed by providing information about structural details of the AAP. This is consistent with research on attitude change, and with research on effects of structural details on attitudes toward affirmative action.

b. Justifying the AAP

Recent theory and research on justification of affirmative action has considered two justifications. According to the compensation justification, affirmative action exists to make up for previous discrimination against members of the target group. This discrimination could have been societal (weak argument) or organizational (strong argument). This justification is consistent with the remedial action conceptualization mentioned by Gamson and Modigliani (1987), and relates to research on the need for affirmative action. According to the instrumental (cultural diversity) justification, affirmative action enhances organizational effectiveness by increasing cultural diversity within the organization. Levi and Fried (1994) point out that this latter argument also can vary in strength. For example, a weak argument would be that increasing the number of minority employees improves the organization's image. A strong argument would be that only a Hispanic individual could serve as an undercover agent in a barrio.

Mann and Fasolo (1992) obtained 87 undergraduates' reactions to concocted faculty hiring decisions in which the affirmative action justification (compensation versus cultural diversity) was manipulated. Justification did not significantly affect fairness ratings of the decision or of the policy. Similarly, internal analysis revealed that ratings were not correlated with perceptions that the college's rationale was compensation. On the other hand, internal analysis revealed that both types of fairness ratings were positively correlated with respondent beliefs that the college's rationale was cultural diversity.

Murrell, Dietz-Uhler, Dovidio, Gaertner, and Drout (1994) had 135 White male and 202 White female undergraduates evaluate AAP scenarios that varied in terms of the target group, justification, and setting. In two conditions the AAP was framed in different ways but no justification was presented; in the other two conditions the AAP was justified, with the justification involving either compensation or enhancing cultural diversity. There was a significant effect of justification. Reactions were more positive when a justification was presented (with no difference between justifications) than when no justification was presented.

Matheson et al. (1994) asked 19 female police trainees to evaluate five possible reactions to discrimination (AAPs) in a police department. Two reactions involved strong preferential treatment which was justified in terms of reparation or the value of diversity. Evaluation of these two AAPs did not differ significantly, and the similarity of the means suggests that this nonsignificant effect was not simply due to a lack of statistical power.

Bobocel and Farrell (1996; Experiment 2) asked 123 White male students to read two cases, one of which described a male police officer who had filed a lawsuit when an equally qualified woman was promoted instead of him. One independent variable was level of justification: none versus compensation versus instrumental (increasing diversity). Ratings of interactional fairness revealed a main effect of justification; the None condition was evaluated more negatively than the other two conditions, which did not differ. This effect was entirely mediated by the perceived adequacy of the justification.

In Heilman et al. (1996) 162 male undergraduates performing a one-way communication task were told their female partner would be the leader because of her sex. Two variables were manipulated: information about scores on an ability test, and justification for the use of sex (none vs. compensation). Although the interaction was not significant, post hoc tests suggested that justification affected fairness judgments only when the male subjects believed their test performance had been equal to the female's.

Singer (1996) presented adults in New Zealand with a report on affirmative action purportedly written by a panel of male or female experts. The panel argued in favor of merit selection, preferential treatment, or diversity, and provided arguments to support the position. The preferential treatment and diversity conditions differed only in the term used to describe the procedure. Respondents reported their attitudes toward merit selection and preferential treatment. Attitudes toward merit selection were affected by the justification and the interaction of Justification X Panel Gender. Attitudes toward merit selection were positive except when a panel of women argued for preferential treatment, in which case they were neutral. Attitudes toward preferential treatment were also affected by the justification, being neutral when the diversity justification was used and negative otherwise.

Levi and Fried (1994) have developed a theoretical model in which justification (referred to as affirmative action framing) plays a central role. Briefly, acceptance of affirmative action by non-target group members is directly affected by perceived justification of affirmative action. Perceived justification, in turn, is directly affected by affirmative action framing (organizational justification), the extent to which affirmative action is perceived as legitimate remediation, and the extent to which ascribed characteristics are perceived as legitimate qualifications. Perceived justification is also affected by the interactions of affirmative action framing with the other two variables. Specifically, if respondents do not believe affirmative action is a legitimate remediation (which will be affected by their ideology and beliefs about past discrimination), they will respond very negatively to a compensatory framing. Similarly, if respondents do not believe ascribed characteristics are legitimate qualifications (which will be affected by their ideology and perceived contributions of ascribed characteristics to the organization), they will respond very negatively to an instrumental framing. This model merits empirical investigation.

In summary, both Murrell et al. (1994) and Bobocel and Farrell (1996) found that providing a justification for affirmative action led to more positive evaluations than did providing no justification. Heilman et al. (1996) found a similar effect only when the male subject was told the female leader's score on an ability test was identical to his own. This is consistent with theory and research on procedural fairness (e.g., Cropanzano & Greenberg, in press). In four studies (Bobocel & Farrell, 1996; Mann & Fasolo, 1992; Matheson et al. 1994; Murrell et al., 1994), reactions to the compensation and instrumental justifications did not differ. On the other hand, internal analysis in Mann and Fasolo (1992) suggested that people who believed affirmative action to be instrumentally justified responded more positively. Also, Singer (1996) found that reactions to preferential treatment were more positive when a panel of experts had justified it in terms of diversity. Finally, the work of Murrell et al. (1994), Heilman et al. (1996), Singer (1996), and Levi and Fried (1994) suggests that reactions to different justifications may be moderated by other factors such as: the source of information about the AAP; the AAP target; the AAP structure; and the respondent's ideology, beliefs about the appropriateness of compensation, and beliefs about whether diversity has any instrumental value.

c. Conclusions

As implied by research on AAP structure, attitudes toward affirmative action can be changed by providing individuals with information about the AAP. Attitudes of ordinary citizens appear to be more flexible than those of government decision makers, presumably because the former have devoted less cognitive energy to the issue. Consistent with research on fairness, attitudes toward AAPs are more positive when the AAP is justified. The specific justification does not strongly affect attitudes, perhaps because such effects are moderated by other variables.

4. Identity of Target Group

Fletcher and Chalmers (1991), discussed above, found that attitudes of both citizens and decision makers were more negative toward AAPs targeted at French Canadians than toward AAPs targeted at women. In an interaction of Target X Respondent, however, the effect of target group was observed only among the majority Canadians; the French-Canadians rated affirmative action toward women and French-Canadians equally.

Sniderman, Piazza, Tetlock, and Kendrick (1991) surveyed 709 White adults in the San Francisco Bay area. They were asked how much the government should do to ensure equal opportunity for women or for Blacks and minorities. The target was manipulated between groups. Both liberals and conservatives were more supportive of government intervention when women were the target group. Further analyses revealed that this target effect was significant among the respondents with a high school degree or less, but was not significant among the college graduates. Among respondents with some college, the target effect was significant among conservatives but not liberals.

Clayton (1992) provides a brief report of a relevant study, in which respondents were told that a fictitious company had either 40% women or 25% minority employees. In addition, respondents either were or were not told that the company was trying to recruit more employees of the same type. Respondents estimated their chances of being treated fairly by the company. These ratings were affected by the interaction of category type by presence/absence of recruitment. Responses were most positive when the company with 40% female employees was trying to recruit more women, and were most negative when the company with 25% minority employees was trying to recruit more minorities.

Kravitz and Platania (1993) randomly assigned respondents to conditions in which they read descriptions of one of three different AAPs. These AAPs differed only with respect to their target: women, minorities, or the handicapped. The sample of undergraduates was diverse in terms of gender and race/ethnicity. Results showed that overall attitude toward AAPs did not differ across the three target groups. Analyses of the individual components of an AAP yielded a few target group differences. Respondents were less opposed to hiring someone who was unqualified when that person was handicapped rather than a minority. In an interaction of Target X Respondent, the possibility of requiring all organizations to have an AAP was seen equally negatively across targets by males, but was seen as more positive by females if the target group was the handicapped rather than minorities.

Kravitz et al. (1994) surveyed employed adults in the Miami area by telephone, asking questions about AAPs targeted at Hispanics or Blacks. Of the 60 respondents, 23 were Black, 24 were Hispanic, and 13 were White. Target group had no main effect on overall attitudes toward the AAP, but it did interact with respondent ethnicity: respondents were more positive about AAPs targeted at their own ethnic group.

Murrell et al. (1994), discussed above, assessed undergraduates' reactions to AAP scenarios that varied in target group, justification, and setting. Respondents had significantly more negative attitudes toward AAPs directed at Blacks than at either the elderly or the handicapped. This effect was heightened by lack of justification for the AAP, in that AAPs described without justification and targeted at Blacks were seen more negatively than any other combination of target and justification (a significant element of a marginally significant interaction of Target X Justification).

Gilliland and Haptonstahl (1995) asked 180 undergraduates to evaluate four scenarios that described a selection decision involving a target and non-target group member. The respondents varied in terms of gender and race/ethnicity. The target group member was a woman, an African-American, or wheelchair-confined. Gilliland and Haptonstahl did not observe a main effect of target, but did find a significant three-way interaction of Respondent Race X Scenario X Target that was consistent with self-interest.

In two studies of undergraduates at different colleges (s = 71 and 119), Clayton (in press) asked her respondents to rate the appropriateness of using information about various attributes (race, sex, religion, sexual orientation, major) for different purposes, one of which was the promotion of diversity in admissions (affirmative action). The interaction of attribute by purpose was significant. In both studies it appears that opposition the use of information for diversity purposes was strongest when the information dealt with religion and sexual orientation, next strongest when it involved race, and weakest when the information concerned college major and gender.

In summary, the target group of an AAP appears to influence AAP attitudes. Programs directed at Blacks or minorities are viewed less positively by Whites than programs directed at women or the handicapped. Reasons for this difference are unclear, though results demonstrates that it may be moderated by details of the AAP and by respondent education, political perspective, gender, and race/ethnicity. Smith and Kluegel (1984) surveyed 1,507 English-speaking residents of the U.S., and found that the public believes women suffer more from discrimination than do Blacks. This could also play a role in the effect of target group on attitudes. Results of several studies suggested effects of ingroup-outgroup differences or self-interest, and research discussed above has shown that self-interest considerations influence attitudes. Eberhardt and Fiske (1994) discuss several intriguing and reasonable hypotheses that might account for the effect of target, as do Taylor-Carter, Doverspike, and Cook (1995). Other portions of this review cover the constructs proposed in these hypotheses and frameworks.

5. Setting

The effect of setting on attitude toward AAPs has been investigated in several studies. In their national survey of Whites, Kluegel and Smith (1983) described two affirmative action programs to all respondents. One program involved college admissions and the other involved hiring employees; both were described as programs in which a "certain number of positions would be set aside for Blacks and other minorities." A reanalysis of their data reveals greater support for set-asides in education than in business.

Bobo and Smith (1994) summarize responses to the 1986, 1988, and 1990 National Election Study (NES) surveys. These surveys included representative samples of American adults. Among White respondents, support for quotas in education (24.1% to 30.9%) was clearly higher than support for preferential hiring and promotions in employment (10.7% to 17.5%). In 1986 and 1988, a higher percent of Black respondents supported quotas in education (74.8% and 82.6%, respectively) than preferential hiring and promotions in employment (61.2% and 65.1%, respectively). In 1990, in contrast, Blacks' support for preferential treatment in employment (74.8%) was slightly higher than support for quotas in education (73.5%). Unfortunately, differences in question phrasing (quotas in education, preferential treatment in employment) make it difficult to interpret these differences.

Fletcher and Chalmers (1991) examined organizational setting as a possible influence on attitudes toward AAPs and quotas among Canadian citizens and government decision makers. The researchers did a within-subject comparison of attitudes toward gender-based quotas in the Canadian federal government and in large, private companies. There was greater support of AAPs for government positions in all subsamples. The difference was especially pronounced for the decision-maker subsamples (a Role X Setting interaction). Unfortunately, this setting difference was confounded with a small but important difference in question wording. Questions about AAPs in the private sector used the politically charged word "quotas" while questions about AAPs in the public sector did not.

Fine (1992a) analyzed opinion survey data from a 1986 national probability sample of adults in the U. S. (319 Blacks and 1,790 Whites). More Blacks and Whites favored preferential treatment in colleges and universities (74% and 22%, respectively) than in hiring and promotions (68% and 17%, respectively). Unfortunately, the questions were phrased differently, with the former including the loaded term "quotas."

Kinder and Sanders (1990) manipulated organizational setting in a telephone survey using a national probability sample of voting age citizens (N = 380). Using a within-subjects design, they described general AAPs targeted at Blacks in employment (hiring and promotion) decisions and in college admissions decisions. There was no effect of setting on attitudes toward AAPs.

In the study by Murrell et al. (1994) described above, there were no main effects or interactions involving setting. This study included three settings: business, college, social organization. Finally, in Study 1 of Matheson et al. (1994) the social domain (education versus employment) did not affect reactions to the four AAPs.

In summary, the organizational setting of an AAP seems to have limited influence on AAP attitudes, either as a main effect or in conjunction with other variables. Affirmative action may receive more support in education than in employment, but despite large samples this difference was statistically significant in only three of the six relevant studies. Furthermore, in two of these studies the setting was confounded with the description of affirmative action (e.g., quotas versus preferential treatment).

6. Need for Affirmative Action

No studies were found that examined relationships between objective need for affirmative action programs and affirmative action attitudes, if objective need is operationalized in terms of economic indicators and utilization ratios. A number of studies, however, dealt with the organization's reported history of discrimination. If we assume that a history of discrimination implies a need for affirmative action, this research is relevant. Other work has assessed respondents' perceptions of the need for affirmative action.

Nacoste (1985, 1987; Nacoste & Lehman, 1987) asked female undergraduates to play the role of a female professor who received a grant when competing with a male professor. In these studies Nacoste manipulated the organization's reported history of discrimination, along with procedural and qualification variables. These studies included numerous dependent variables, with ratings of fairness of procedure and decision being most relevant to the question of attitudes. In no case were there any significant effects involving the history manipulation.

In Kravitz and Platania (1992, 1993), discussed above, undergraduates evaluated potential components of AAPs. One component was that AAPs would be designed to compensate for previous discrimination. This component received a neutral evaluation in the 1992 study and a positive evaluation in the 1993 study. Kravitz et al. (1994) employed a similar procedure in a telephone survey of employed adults, most respondents being Black or Hispanic. These respondents favored requiring organizations with histories of discrimination to target AAPs at the group (Blacks or Hispanics) that had been victimized. In his analysis of data from a national survey of Whites, Jacobson (1985) found that support for affirmative action was significantly related to the belief that Blacks experienced discrimination in a variety of areas.

Matheson et al. (1994) manipulated the extent of prior discrimination in two scenario studies with female undergraduate respondents. The three levels were: (a) only males would be considered; (b) females would be considered only if no qualified males were available; (c) when qualifications were equivalent, the male would be preferred to the female. The authors hypothesized that respondents would prefer AAPs that precisely matched the level of prior discrimination -- an interaction between the discrimination and AAP manipulations. In both studies the interaction was nonsignificant. Furthermore, the main effect of discrimination also was not significant: Evaluations of the possible reactions (AAPs) were not affected by the level of prior discrimination.

Although results have been mixed, this research suggests that people may support certain kinds of affirmative action if they believe past discrimination warrants it. It follows that opposition to affirmative action could be based on the belief that discrimination does not exist. Kluegel (1985, 1990) reviewed large national surveys completed between 1972 and 1989, and concluded that most White Americans attribute Blacks' economic disadvantages to personal rather than structural factors. That is, they believe the problem is that Blacks lack motivation and skills, not that they lack opportunities. Bobo and Kluegel (1993) reported the results of a national survey of 159 Blacks and 1,150 Whites. They found that perceived discrimination and attributions of the Black-White gap in socio-economic status to motivational causes were two of the strongest predictors of support for policies designed to enhance the opportunities and incomes of Blacks, though none of the policy items referred to affirmative action per se. A study by Witt (1990) is relevant to this issue. She asked 492 university faculty whether minority of female status constituted an advantage or disadvantage for each of 15 career dimensions. Ratings given by White faculty were significantly higher than those given by Black faculty when considering the role of minority status (all 15 dimensions) and female status (12 of 15 dimensions). In the case of minority status, inspection of the means is suggestive. For the key dimensions (initial selection, salary, tenure, promotion) means given by the Whites were above the scale midpoint and means given by Blacks were below the scale midpoint. Minority status has negative effects via discrimination and positive effects via affirmative action. Apparently the White faculty believed the positive effects of affirmative action outweighed the negative effects of discrimination, whereas the Black faculty believed the opposite. Support for affirmative action is likely to be decreased when people believe it overcompensates for discrimination.

In sum, there is no research on the relation between objective measures of need and attitudes toward AAPs. Limited research suggests that respondents are somewhat more likely to support affirmative action if they believe or are told that the target group has suffered discrimination. These results, however, have been inconsistent. Survey research suggests that many White Americans believe discrimination is a problem of the past, and that Blacks have themselves to blame for their economic disadvantages. Other research finds that Whites believe minorities profit more from affirmative action than they are hurt by discrimination. A related finding based on respondent beliefs about discrimination is addressed below where we discuss research on relative deprivation. This emphasis on need is consistent with court decisions, which have emphasized that AAPs should be remedial (Newman, 1989).

7. Conclusions

Attitudes toward affirmative action and AAPs are strongly influenced by structural factors. Details of the AAP (weighting of demographic status) appear to have the strongest effect on attitudes, and this effect is mediated at least partly by implications for fairness and self-interest. Attitudes are more positive when the AAP is justified and when evidence indicates that affirmative action is needed. Reactions tend to be less positive when the target group is a racial minority. Organizational setting has little effect on evaluations.

C. Individual Differences Bases (Respondent Dimensions)

The research described above has focused on structural aspects of the situation and of the AAP. Other research has addressed the role of individual differences. The implicit assumption in much of this work is that individual differences will affect a person's attitude toward affirmative action in general, and this in turn will affect the person's attitudes toward specific AAPs. This research has included the respondent's role, demographic variables, prejudice, relative deprivation, and ideology. When considering this research it is important to realize that it, like all passive observation research, does not permit strong causal conclusions.

1. Respondent Role

Two straightforward characterizations of a respondent's role have received some investigation. The first deals with differences between those who make decisions about or administer AAPs versus those who do not. The second characterization of role deals with differences between those who are members of the target group of an AAP and those who are not. We discuss these two areas in turn.

a. Decision Makers Versus Others

Noble and Winett (1978) surveyed approximately 60 department chairs and 100 hirees for faculty positions at the University of Kentucky between 1974 and 1976. They provided no demographic data for the chairs, but reported that 73% of the hirees were male and 96% were White. They found no differences between the department chairs and the hirees in their support for hiring minority members or women for a faculty position if the candidate was as well qualified as a White male (50% reported positive attitudes). They also found no differences across the two roles in their support for hiring a lesser qualified female or minority member.

In their study of women at a liberal arts college, Goldsmith et al. (1989) compared AAP attitudes of 15 students, 15 faculty, 17 staff members, and 15 administrators. The researchers found no differences in AAP attitudes across the four role-based groups, using either questionnaire items or coded statements from interview transcripts.

Tickamyer, Scollay, Bokemeier, and Wood (1989) explored attitudes about the need for AAPs in university settings. Ninety-one percent of 150 university affirmative action officers thought that some (undefined) steps needed to be taken to increase numbers of underrepresented groups, and 62% were positive about setting specific hiring goals. Seventy-four percent of the 1,900 other university administrators were positive about taking some steps, and only 22% supported specific hiring goals.

Fletcher and Chalmers (1991) made several comparisons of AAP attitudes among Canadian citizens and government decision makers. Citizen and decision-maker attitudes toward gender-based quotas in industry were similar. When the decision-maker sample was split into legislative, executive, and judicial subsamples, however, important differences emerged. Law-makers were more positive than the citizenry, and members of the judicial branch were more negative. Members of the executive branch had intermediate attitudes, matching the level of the general population. Both citizens and decision makers reported they would change their opinions when given an argument contrary to their initial opinion, but citizens were more likely than decision makers to state they would change.

Few general conclusions can be drawn from this research. It seems likely that the strongest support for affirmative action will be observed among those whose jobs involve the maintenance of AAPs (Tickamyer et al., 1989). Furthermore, attitudes of decision makers who have thought the issue through are likely to be less flexible than attitudes of typical citizens (Fletcher & Chalmers, 1991).

b. Target Group Members Versus Non-members

Another characterization of role involves differences between those who are members of the target group of an AAP and those who are not. Classification of respondents into target member versus non-target member simplifies interpretation of some of the two-way interactions to be discussed in the next subsection. Simply put, persons are more positive toward AAPs that are targeted at their own demographic group. This trend is clear and strong for ethnicity, but is somewhat less strong for gender. This effect of target-respondent match can be explained in terms of collective or personal self-interest.

2. Demographic variables

a. Gender and Race/Ethnicity

Numerous studies have examined differences in AAP attitudes associated with respondents' gender or ethnicity. Many of these studies focused on other questions and have been discussed above. In these cases we limit the present discussion to results involving respondent gender or ethnicity. All results mentioned below are significant at the .05 level, although effect sizes varied. If a potential difference is not mentioned, it was not statistically significant. We begin by reviewing studies that assessed only race differences, then turn to studies that assessed only gender differences, and end with studies that dealt with both variables.

We begin with survey research. Lipset and Schneider (1978) summarized some of the earliest data regarding race-based differences in attitudes toward AAPs. Citing a New York Times/CBS News survey carried out in 1977, they noted that Blacks were much more positive than Whites toward four versions of affirmative action. The programs included hiring "a certain number of minority workers," "give special consideration to minority applicants" to college or graduate school, "requiring large companies to set up special training programs for members of minority groups," and a school "reserving a certain number of places for qualified minority applicants." Sigelman and Welch (1991) also analyzed many national surveys of attitudes toward affirmative action. They concluded that support for equal opportunity, affirmative action in general, preferential treatment, and quotas was higher among Blacks than Whites. In the Kinder and Sanders (1990) survey described earlier, AAPs in employment were strongly favored by 64% of the Black respondents and 6% of the White respondents. Affirmative action plans in college admissions were strongly supported by 63% of the Black respondents and 9% of the White respondents. Strong differences were also obtained in Fine's (1992a) nationwide probability survey, discussed above. The proportion of Blacks favoring AAPs in employment and education was more than three times the proportion of Whites favoring these programs. Bobo and Smith (1994) reviewed responses to the 1986, 1988, and 1990 National Election Study (NES) surveys. In 1990, support for quotas in education was reported by 31% of Whites and 74% of Blacks; support for preferential hiring and promotions was reported by 18% of Whites and 75% of Blacks. Large race differences were also apparent in the two previous surveys. Sherman, Smith, and Sherman (1983) reported a study of Maryland undergraduates' reactions to a fictitious court case involving promotion decisions. Of the 175 students involved, 106 were Black and 69 were White. Attitudes toward the use of "racial quotas . . . in promotions and/or hirings" were more positive among Black than White respondents. In summary, these many surveys reveal that affirmative action plans of all types receive more support from Blacks than from Whites.

Arthur et al. (1992) asked volunteer respondents to play the role of a student who had been awarded a prestigious graduate fellowship. In the equally qualified condition, participants read that both candidates were equally qualified for the position. In the less qualified condition, participants read that the recipient of the fellowship was less qualified than the other candidate. Participants read that the fellowship was either awarded (a) primarily on the basis of gender, or (b) on the basis of gender as long as the recipients had the necessary qualifications. Of the respondents, 106 were White, 116 were minority group members (split evenly between Blacks and Hispanics, although analyses did not differentiate the two groups), and 49 were international students. International students reported more positive affect in role-played responses to the scenarios than did Whites. Minority students' affect fell between the two, but did not differ significantly from either. Another dependent variable, agreement with the decision, could reasonably be construed as attitude toward the AAP (the AAP was the basis for the fellowship decision). There was no main effect of ethnicity on agreement, but there was a significant interaction of Ethnicity X Structure on this variable. AAP structure had a larger effect among members of minority groups than among Whites. AAP structure had no effect on attitude of international students.

We turn now to research that dealt with gender differences only. In their study of women at a liberal arts college, discussed above, Goldsmith et al. (1989) found significantly more positive attitudes toward affirmative action among women than men. This difference was observed in respondents' self-reported attitudes and interviewers' ratings of respondents' attitudes. These comparisons were limited to the 47 faculty, staff, and administrator respondents because all student respondents were women. Singer (1993) reported that undergraduate females were more positive than males about an AAP in which the gender ("female-ness") of the candidate was given strong weight. In a study of American and Japanese undergraduates, Ozawa et al. (1996) found that women had more positive attitudes than men in response to two procedures, merit hiring and quota hiring following a minimum standard. The women also responded more negatively to the case of gender discrimination used in the scenario.

Tougas and Beaton (1993) studied AAP attitudes among 185 female and 277 male employees at a large Canadian company with a history of discrimination against women, as well as repeatedly unsuccessful attempts to increase the percentage of women in the workforce. The researchers sent a questionnaire describing the company's history and its attempts to redress discrimination to roughly 1000 employees. Males and females did not differ in their attitudes toward AAPs that had the goal of "eliminating systematic barriers" to women. However, females were more positive than males about AAPs that gave "preferential treatment" to females (i.e., that gave preference to women when they were of equal competence as men). In a somewhat contradictory finding, Murrell et al. (1994) found that women were more negative about specific AAPs than were men. Women were especially negative when the AAPs had no explicit justification (a Gender X Justification interaction).

Summers' (1995) respondents evaluated affirmative action in general and three specific AAPs. They also rated the likely effect of affirmative action in general on their careers (self-interest). Summers found a significant gender difference in attitudes toward affirmative action in general. Men and women did not differ significantly in their evaluations toward a special training program (both positive) or differential scoring of selection tests (both negative), but did differ in their evaluations of a quota system. Women evaluated quotas more positively than did men. In summary, results of these four studies are somewhat mixed. In five studies women had more positive attitudes toward at least some AAPs. In one study women had more negative attitudes. The gender difference appears to be moderated by AAP structure and possibly justification, and may be mediated by self-interest implications.

Finally, many studies have assessed both race and gender differences. In Fletcher and Chalmers' (1991) study, French Canadians had more positive attitudes than did majority Canadians. Gender was not related to attitudes toward AAPs.

Kravitz and Platania (1992) assessed attitudes towards a race-based AAP. The sample of undergraduates contained 66 females and 25 males, as well as two people who did not report their gender. Twelve of the respondents were Hispanic, 19 were Black, and 56 were White. Females were more positive than males about the AAP. Blacks were more positive than Whites, with the mean for Hispanics falling between those two but not differing significantly from either. There were several gender and ethnicity differences in evaluations of the potential structural components, including: quotas (gender & ethnicity differences), weighting of race in employment decisions (ethnicity difference), training programs (gender and ethnicity differences), recruitment (ethnicity difference), and providing workforce composition data to the federal government (ethnicity difference). All gender differences were due to greater support among women than among men. All ethnicity differences were due to greater support among Blacks than Whites, with support among Hispanics generally falling between the other two groups.

Kravitz and Platania (1993; N = 349) found more positive attitudes toward AAPs in general among female than male undergraduates. Black and Hispanic respondents were more positive than White respondents towards AAPs. There were no interactions of respondent and target group characteristics for AAP attitude. However, there were important gender and ethnicity differences involving evaluations of the structural components of potential AAPs. Significant differences included: quotas (gender and ethnicity differences), weighting of demographic status (ethnicity difference), training (gender and ethnicity differences), recruitment (gender and ethnicity differences), requiring AAPs of all organizations (gender and ethnicity differences), using AAPs to eliminate discrimination (gender difference), providing workforce composition data in an AAP to the federal government (gender difference), and using AAPs to compensate groups for past discrimination (gender and ethnicity differences). Once again, all gender differences were due to greater support among women than among men; all ethnicity differences were due to greater support among Blacks than Whites, with support among Hispanics generally being between the other two groups.

Kravitz et al. (1994) surveyed 60 Miami-area residents about their attitudes toward affirmative action targeted at Blacks or Hispanics. They found that males were more positive than females. Blacks and Hispanics each evaluated AAPs more positively if they were targeted at their own ethnic group rather than the other group.

Doverspike and Arthur (1995) explored gender and ethnicity (Black versus White) differences in a role-playing study in a data set drawn in part from Arthur et al. (1992). The 235 undergraduate respondents role-played a same-sex Black applicant who was awarded a fellowship. The applicant's qualifications were said to be equal or inferior to those of the other finalist, and the AAP was said to involve reverse discrimination or preferential treatment. Respondents rated the fairness of the procedure and decision. Fairness ratings varied with respondent gender and race, and each of these main effects was qualified by an interaction with applicant qualification. Females were more affected by the qualification manipulation than were males. Blacks were less affected by the qualification manipulation than were Whites, but this difference was much larger in the preferential treatment AAP than in the reverse discrimination AAP (a Race X Qualifications X AAP interaction). Finally, there was a Sex X Race X Qualifications interaction: The race difference was smaller in the condition involving male respondents and equal qualifications than in the other three conditions.

Gilliland and Haptonstahl (1995) found that fairness ratings of four selection scenarios varied as a function of respondent gender and minority status. Males provided higher fairness ratings than did females for the scenario in which the majority candidate was selected instead of the equally-qualified minority candidate. Nonminorities provided higher fairness ratings than did minorities for the scenario in which the more-qualified majority candidate was selected.

Kravitz (1995) reported a study of respondent reactions to eight AAPs targeted at Blacks. Undergraduate respondents varied in terms of gender (47 male and 128 female) and race/ethnicity (84 Hispanic, 75 non-Hispanic Whites, 16 other). Blacks were excluded for both methodological and substantive reasons. There were no significant effects of gender or ethnicity (Hispanic versus White) on overall attitude toward AAPs.

Konrad and Linnehan (1995b) studied managers' attitudes toward identity-conscious and identity-blind policies that a firm might adopt. The scales, therefore, can be seen as representing AAPs that either did or did not assign a weight to demographic status. They compared responses of White men (n = 81), White women (n = 76), People of Color (racial minorities; n = 85), and African Americans (n = 50, a subset of People of Color). They obtained no differences in attitudes toward identity-blind procedures. However, Blacks (and People of Color) had more positive attitudes than Whites toward the identity-conscious procedures. Among Whites, women responded more positively than men to such procedures. These differences remained even after controlling for the type of organization from which the respondents were drawn.

Bell et al. (1996) studied 326 undergraduates and managers, and found that attitudes toward AAPs were more positive among women than among men, There was also an effect of race, with Blacks being most positive, Asians and Hispanics being next most positive, and Whites being least positive. Additional analyses, however, revealed that the significant effects of gender and race were fully mediated by the respondents' beliefs about affirmative action and evaluations thereof. In a subsequent study involving 610 students and employees, Bell (1996) obtained similar effects of respondent gender and race, but these effects were only partly (80%) mediated by the respondents beliefs and evaluations of those beliefs.

In summary, Blacks clearly feel more positively about AAPs in general than do Whites, though the size of this difference depends on details of the AAP. Attitudes of Hispanics fall somewhere between Blacks and Whites. It should also be noted that Hispanics are not a monolithic ethnic group. Cuban-Americans may have notably different attitudes than Mexican-Americans, or those from Puerto Rico. Moreover, apart from the Bell et al. (1995) and Bell (1996 studies, there are no data in the literature reviewed on the attitudes of Asian Americans or Native Americans. Similar caveats about assuming homogeneous attitudes within those ethnic groups apply as well. The research by Bell et al. suggests that the race differences are largely due to differences in beliefs about affirmative action and evaluations of those beliefs.

Finally, there are also gender differences in attitudes toward AAPs. In studies in which respondents (especially White respondents) are asked about AAPs in general, women generally have more positive attitudes than men. Those gender effects are also likely to be moderated by the structural properties (details of the AAP and justification for the AAP) of the AAP being studied; women tend to be more responsive to these factors than do men. Of greater interest is the question of why gender differences exist. Once again, the research by Bell et al. suggests that the gender differences are largely due to differences in beliefs about affirmative action and evaluations of those beliefs. Research by Ozawa et al. (1996) suggests that perceptions of the severity of discrimination could serve as the mediator, and research by Summers (1995) suggests that the mediator could be self-interest. Clearly more work is needed on the boundary conditions and reasons for race and gender differences.

b. Other Demographic Variables

A number of studies have included demographic variables such as age, education, and income as predictors of affirmative action attitudes. Demographic variables have been of interest because they are thought to reflect underlying factors such as socialization (in the case of age, education, region of the country, and city size), social status (education and income), selfinterest (income), potential for having personal experience with unequal treatment (race, gender, men married to working women), and even cognitive complexity (education). In some cases, researchers state specific hypotheses regarding demographic variables (e.g., Fine, 1992b). More commonly, demographic variables are used in an exploratory fashion or as control variables, with no specific hypotheses being stated.

Research on demographic variables can be categorized along two methodological dimensions. One dimension is the type of respondents surveyed (general public versus specific population). The second dimension is the type of statistical analysis performed (multiple regression versus oneway analysis of variance). Studies of the general public have employed largescale national opinion surveys, while studies of specific populations have been done with groups as diverse as public personnel administrators (Davis & West, 1984), university administrators and affirmative action officers (Tickamyer et al., 1989), and social workers in Texas (Stout & Buffum, 1993). Studies that have assessed attitudes of the general public have used multiple regression analyses and have assessed the relative contribution of both demographic and attitudinal variables to the prediction of affirmative action attitudes. In contrast, most studies of attitudes within specific populations have used oneway analysis of variance. The exception was Davis and West (1984) who used a measure of association called gamma. The methodology used has implications for the ability to draw conclusions about the statistical significance of effects, about the size of the effects, and about generalizability of the effects.

(1) Surveys of the General Public

Three studies using general opinion surveys included demographic characteristics among the predictor variables. Kluegel and Smith (1983) assessed attitudes of 1,596 Whites, selected from a 1980 opinion poll, toward affirmative action directed at Blacks and minorities. They presented separate analyses for each of their four attitude questions. The effect of respondent age was nonsignificant in all cases. There were, however, significant interactions of age with employment variables such as the extent of segregation in the respondent's industry. In two analyses there were significant regression weights of education. More educated respondents were less likely to report that preferential treatment was fair and that affirmative action programs should be supported. In the other two analyses, regression weights of income were significant. Respondents with higher incomes were less likely to agree that colleges and employers should set aside positions for qualified Blacks and minorities. [Kluegel and Smith's verbal descriptions of the income results appear inconsistent with the regression weights, so the directions of the effects are not entirely clear. The descriptions in this paragraph are taken from Kluegel and Smith's verbal descriptions.]

Jacobson (1984) reports results from a 1978 survey of 1,584 White respondents. Attitudes toward affirmative action were not related to income. Attitudes were related to occupation, age, and education. The relation of attitudes with age was complex. Attitudes appeared most positive among college graduates and least positive among high school graduates; the other two categories were those with less than a high school degree and those with some college.

Fine (1992b), drew a sample of AfricanAmericans from a 1986 national postelection survey. None of her four dependent variables explicitly dealt with support for affirmative action, but two of them seem relevant. Respondents were asked to indicate the extent of their agreement with statements that: (1) it is not the government's job to guarantee equal opportunity for Blacks and Whites (N = 313), and (2) Blacks should overcome prejudice without any special favors (N = 156). When responses to these items were regressed on demographic and core value variables, regression weights of age, education, and income were all nonsignificant.

(2) Surveys of Specific Populations

Davis and West (1984) surveyed a national sample of public personnel managers during 1979. They measured several demographic characteristics and environmental characteristics (e.g., city size). Dependent variables were commitment to affirmative action and to equal employment opportunity. Neither age nor education was significantly associated with commitment to affirmative action. Age, but not education, was significantly associated with commitment to equal opportunity, with the youngest managers displaying the lowest levels of commitment.

Tickamyer et al. (1989) surveyed 1,894 university administrators and 148 university affirmative action officers. The survey included four questions about the need for EO/AAPs at their colleges and universities. Among the administrators, age (four groups) was significantly associated with responses to all four need questions. For two questions, perceived need decreased with age; for the other two questions the relation was nonlinear. Among the affirmative action officers, there were no significant relations between age and perceived need.

Stout and Buffum (1993) measured the commitment to affirmative action of 193 Texas social workers. They found significant effects for job position and membership classification. Because statistical tests of mean differences were not reported, it is not possible to report which groups were different from one another. Commitment was higher among those who identified themselves as members of a "sexual minority," but was not associated with level of highest social work degree.

(3) Summary

Researchers have used a variety of individual characteristics in attempting to account for attitudes related to affirmative action. Age, education, and income have been studied most consistently, but many other characteristics have been investigated on occasion. These studies show that, in the majority of cases, demographic factors other than gender and race/ethnicity contribute little or nothing to predictions of attitudes related to affirmative action. In the few cases in which the effects associated with these variables reached statistical significance, the effects were quite small and sometimes nonlinear.

3. Self-Efficacy

Two studies assessed the relations among AAP, self-efficacy, and fairness perceptions. In both cases, undergraduates were told they had been selected either on the basis of merit or on the basis of gender.

Hattrup (1994) found that fairness judgments were affected by the interaction of selection procedure and self-efficacy. Fairness ratings of preferential treatment were minimally related to self-efficacy; virtually everyone considered it unfair. Fairness ratings of merit selection were positively related to respondent self-efficacy; more confident respondents considered their selection fairer. In Brutus and Ryan (1994), ratings of fairness correlated positively with self-efficacy measured before and after selection, and with self-evaluation of performance. Fairness ratings did not correlate with performance in either study.

In summary, AAP structure, perceived fairness, and self-efficacy are complexly interrelated. We are not aware of research involving more traditional dispositional variables. Given the relation between prejudice (discussed below) and dogmatism, it is possible that research on dogmatism would yield interesting results.

4. Opinion Variables

a. Prejudice (Racism and Sexism)

Fernandez (1981) defined [anti-Black] racism "as basic cultural ideologies that state that Whites . . . are inherently superior to minorities . . . solely on the basis of race . . . and that Whites have the power over social institutions to develop, evolve, nurture, spread, impose, and enforce the very myths and stereotypes that are basic to the foundation of racism" (p. 303). Sidanius, Pratto, and Bobo (1996) refer to this assumption of Black inferiority as classic racism, and distinguish it from anti-Black affect. Other theorists have suggested that such forms of racism have been supplanted by modern forms. For example, aversive racism is said to be a combination of racial prejudice and an egalitarian ideology (Dovidio, Mann, & Gaertner, 1989; Gaertner & Dovidio, 1986). Symbolic racism, in contrast, is said to be a combination of racial prejudice and a conservative ideology (Kinder, 1986; McConahay, 1982; McConahay & Hough, 1976; Sears, 1988). The concept of symbolic racism has been critiqued by Sniderman and Tetlock (1986b; see also Bobo & Kluegel, 1993; Kinder, 1986; Sniderman et al., 1991; Sniderman & Tetlock, 1986a), and some of their criticisms apply to aversive racism. One problem with the concepts of modern racism particularly relevant to the present review is that opposition to affirmative action is taken as an indicant of modern racism; as such, it is meaningless to ask whether modern racism leads to opposition. Research examining symbolic and aversive racism does not examine the conditions under which these values do or do not constitute a convenient racist surrogate. Some evidence (Kluegel, 1990) suggests that the fundamental attribution error may augment or enhance symbolic racism. In addition, Murrell et al. (1994) have suggested that it is not possible to assess aversive racism through self-report measures. In summary, there is considerably disagreement about how racism should be conceptualized, and an equal level of inconsistency in how it is operationalized.

Despite this disagreement, research has consistently found that self-report measures of racism are positively associated with opposition to affirmative action targeted at racial minorities. Using data from the national Harris opinion survey of 1978, Jacobson (1985) found that symbolic and overt racism predicted attitudes toward AAPs. In their national survey of Whites, Kluegel and Smith (1983) found that a two-item scale of symbolic racism was associated with attitudes toward affirmative action in education and business. Kravitz (1995) found that racism was associated with opposition to affirmative action in general and to specific AAPs.

Nosworthy et al. (1995) argued that researchers should separately measure the two components of symbolic racism -- racial affect and justice beliefs. Thus, they included measures of symbolic racism, racial affect, and justice beliefs in their study of reactions to AAPs in education. When endorsement of four AAPs were regressed on the three measures, the alternative approach (racial affect plus justice beliefs) explained variance beyond symbolic racism in two cases, but the reverse was not true. When racial affect and justice beliefs were contrasted, they were found to be approximately equivalent predictors. Justice beliefs explained variance beyond racial affect in predicting reactions to recruitment, but neither approach explained independent variance in predicting reactions to the other three AAPs.

Sidanius, Devereux, and Pratto (1992) contrasted a symbolic racism explanation with a social dominance theory of "racial policy attitudes." Social dominance theory as developed by Sidanius et al. (1992) suggests that all social systems consist of hierarchical caste systems with various dimensions of social status and rules/influences on movement within and between castes. These caste systems are maintained over time through institutional discrimination, individual discrimination, and "behavioral asymmetry," or evolution of different behavioral repertoires across castes that perpetuate and reinforce caste differences. Sidanius et al. (1992) analyzed data from a sample of 234 White respondents, and found substantial evidence for social dominance as an antecedent of attitudes toward racial policy. Close inspection of the symbolic racism and social dominance "models" indicates that they contain virtually the same antecedent variables hypothesized to operate in different causal orders. In Studies 1 (N = 3,861) and 2 (N approximately = 614), Sidanius et al. (1996) found that opposition to affirmative action was associated with classical racism, and that this relation increased with respondent education. Anti-Black affect was not associated with opposition to affirmative action.

Reid and Clayton (1992) warned that racism and sexism differ in many ways, and that it is not safe to assume that the two dispositions will relate identically to attitudes toward affirmative action. In comparison to racism, sexism has been neglected in the affirmative action research, but there is some relevant work. Similar to modern racism, neosexism is said to be "a manifestation of a conflict between egalitarian values and residual negative feelings toward women" (Tougas, Brown et al., 1995, p. 843). In a study of 130 male Canadian students, Tougas, Brown et al. (1995) used path analysis to demonstrate that neosexism predicted opposition to affirmative action. Traditional sexism affected reactions to affirmative action only through its influence on neosexism. In a second study of 149 male managers, office workers, and blue-collar workers, it was found that attitudes toward affirmative action in general and toward the program in place were significantly related to neosexism. Joly et al. (1993) studied 96 male managers and professionals. They found that neosexism predicted opposition to affirmative action in general and a specific AAP targeted at women. Tougas, Crosby et al. (1995) report the results of two studies on male francophone Canadian managers and professionals. In both cases, attitudes toward AAPs were significantly related to neosexism scores.

In summary, prejudice (racism and sexism) is associated with opposition to AAPs targeted at the relevant group. Much of the research using measures of modern racism or sexism is flawed because the measure of prejudice is confounded with fairness considerations. A few studies have compared the explanatory power of racism to other variables. Kravitz (1995) found that perceived fairness of AAP structure was much more important, and personal self-interest was somewhat more important than racism. Nosworthy et al. (1995) found that perceived fairness explained unique variance beyond measures of racism, personal self-interest, and collective self-interest in evaluations of four AAPs. Attitudes were more closely related to self-interest than to racism.

b. Relative Deprivation

Relative deprivation is "the emotion one feels when making negatively discrepant comparisons" (Crosby, 1976, p. 88). Building on earlier work, Crosby (1984) distinguishes four types of resentment individuals feel about the distribution of an outcome. Personal deprivation exists when the individual receives less than he or she desires and deserves. Surprisingly, this concept has been neglected in the affirmative action area. In-group deprivation exists when the individual's group receives less than the individual desires and believes it deserves. Within the affirmative action area this has been called collective relative deprivation, and it is disc