9., I ‘ .m L . . 1 i . . 1.334%. VIII .‘uhfau. . T v t! . In ~ Li)‘ . > a . .. .uo‘unnwualu. ! .3 . : 4rd 1,t-.~Luu.‘..hv.u 5.99).... L l. .33}. I.» :51. 35%. 1‘. a. . . .J 3.. _ n! 2.. 3am... . ufl....m.... .3.“ .1... #1 I A. iiéh , ... .fififififi IHIUHIHJUHIHIHUI!llfill!"Hill!!!‘lllllllllll 302074 1207 LEBRARY Michigan State University v“; This is to certify that the thesis entitled Selection Discrimination on the Basis of Sexual Orientation presented by Michael Horvath has been accepted towards fulfillment of the requirements for Master of Arts degree in Psychology Date 5 May 1999 0-7639 MS U is an Affinnative Action/Equal Opportunity Institution PLACE IN REFURN BOX to remove this checkout from your record. TO AVOID FINES return on or before date due. MAY BE RECALLED with earlier due date if requested. DATE DUE DATE DUE DATE DUE SELECTION DISCRIMINATION ON THE BASIS OF SEXUAL ORIENTATION By Michael Horvath A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Psychology 1 999 ABSTRACT SELECTION DISCRIMINATION ON THE BASIS OF SEXUAL ORIENTATION By Michael Horvath Recent trends towards including sexual orientation as a protected characteristic in selection highlight the need for exploration of the antecedents of sexual orientation selection discrimination. A two-part model is proposed and tested. The first part of the model replicates past literature that explored the antecedents of attitudes towards homosexuals, such as religiosity, sex role beliefs, beliefs about controllability of homosexuality, and previous exposure to homosexuals. The secOnd part of the model tests moderators of the relationship between attitudes toward homosexuals and discrimination against homosexuals, as represented with a resume manipulation task. The model was tested using a sample of 236 undergraduates. All hypothesized antecedents significantly predicted attitudes toward homosexuals. Analyses of the attitude — discrimination relationship revealed a higher - order interaction: beliefs in the legality of discrimination, participant / target sex congruence, and target congruence to a masculinity / femininity stereotype interacted in the relationship. Implications and directions for future research are discussed. ACKNOWLEDGMENTS Obviously, many individuals and other entities who have contributed to my success at finally getting my thesis completed. I will attempt here to thank most of the substantial contributors, but I acknowledge that many more helped. First, I would like to thank my advisor, Ann Marie Ryan. She was always ready to help by answering questions and keeping me on track, but she let me run with this project within the safe realm of her guidance, letting me develop and acquire confidence. For that she has my gratitude. Similar thanks go to my other committee members, Rick DeShon and Neal Schmitt. Their comments during my proposal and defense were valuable, but they also helped to provide, outside of a thesis context, the quantitative and analytical skills required for these complex analyses. I am indebted to Cathy Berkman for providing me with a copy of the Previous Exposure to Homosexuals Scale. My support system also deserves substantial credit (for their support, but also for their willingness to be impressed into the project for the purposes of proofreading and pilot testing). Foremost is Anne Papcke, whose continual care and encouragement were invaluable. It’s much easier to work hard and take risks when you know there’s always going to be somebody there no matter what the outcome. Also, I acknowledge Morn'e Mullins, Karen Milner, Melanie Lehman, and Josh Sacco for their friendship and support. iii TABLE OF CONTENTS LIST OF TABLES ................................................................................. vi LIST OF FIGURES .............................................................................. vii INTRODUCTION ................................................................................. 1 Introduction to the Model ......................................................................... 5 Factors Associated with Attitudes Toward Homosexuals ........................... 7 Beliefs About the Controllability of Homosexuality ....................... 7 Previous Exposure to Homosexuals ........................................... 9 Sex Roles ........................................................................ 9 Religion / Religiosity ........................................................... 12 Subject / Target Sex Congruence ............................................. 15 Stereotypical Qualities of Lesbians and Gays .............................. 17 Perception of the Legality of Sexual Orientation Discrimination ........ 20 Other Predictors of Attitudes Toward Homosexuality Not Included in the Model ...................................................................... 22 Age ...................................................................... 22 Education ............................................................... 22 Region ................................................................... 23 Demographic Variables Not Found to Affect Attitudes Toward Homosexuality .................................................................. 24 Summary ......................................................................... 24 METHODS ........................................................................................ 24 Participants ............................................................................... 24 Procedure ................................................................................. 25 Measures .................................................................................. 27 Multidimensional Measure of Religious Involvement ..................... 27 Scale of Beliefs about the Controllability of Homosexuality ............. 28 Measure of Sex Role Beliefs .................................................. 28 Measure of Previous Exposure to Homosexuals ........................... 28 Attitudes Toward Lesbians and Gays (ATLG) Scale ...................... 29 Scale on the Beliefs of Employing Homosexuals .......................... 29 Measure of the Perceptions of the Legality of Sexual Orientation Discrimination ............................................................... 30 Measurement of Discrimination .............................................. 30 RESULTS ......................................................................................... 31 Tests of Antecedents of Beliefs about Employing Homosexuals ................. 32 Tests of a Structural Equations Model ................................................ 34 Tests of the Moderators of the Relationship Between Beliefs about Employing Homosexuals and Ratings of Resumes ............................................... 36 iv DISCUSSION .................................................................................... 49 Perceptions of Legality as a Moderator of the Attitude-Discrimination Relationship ........................................................................... 5 l Stereotypicality of Homosexuals as a Moderator of the Attitude-Discrimination Relationship ............................................. 52 Gender Congruence as a Moderator of the Attitude-Discrimination Relationship ........................................................................... 53 Limitations and Directions for Future Research .................................... 53 Practical Implications ................................................................... 56 Conclusion ................................................................................ 58 REFERENCES .................................................................................... 59 Appendix A: Resumes ........................................................................... 67 Appendix B: Job Description ................................................................... 84 Appendix C: Informed Consent Form ......................................................... 86 Appendix D: Religiosity Measure .............................................................. 88 Appendix E: Controllability Beliefs Measure ................................................ 90 Appendix F: Sex Role Beliefs Measure ....................................................... 92 Appendix G: Measure of Previous Exposure to Homosexuals ............................. 94 Appendix H: Attitudes toward Lesbians and Gay Men Measure .......................... 96 Appendix I: Beliefs about Employing Homosexuals Measure ............................. 98 Appendix J: Perceptions of the Legality of Discrimination Measure ..................... 100 Appendix K: Rating Form ...................................................................... 102 LIST OF TABLES Table 1. Means, Standard Deviations, Internal Consistency Reliabilities, and Intercorrelations among Variables in the Path Model (N=232) ............................ 33 Table 2. Direct and Indirect Effects for the Original Model ............................... 34 Table 3. Direct and Indirect Effects for the Revised Model ............................... 38 Table 4. Characteristics of the Resumes, Mean Ratings, and Standard Deviations 39 Table 5. Breakdowns of Resume Ratings by Gender, Masculinity / Femininity, and Sexual Orientation (N=235) .................................................................... 40 Table 6. Source Table for the Repeated Measures Regression ............................ 43 vi LIST OF FIGURES Figure 1. Theoretical Model of Sexual Orientation Selection Discrimination ........... 6 Figure 2. Tested Model of Sexual Orientation Selection Discrimination ................ 8 Figure 3. Path diagram of the original model ................................................ 35 Figure 4. Path diagram of the revised model ................................................. 37 Figure 5. Ratings of Applicants on the Basis of Beliefs About Employing Homosexuals, Gender Congruence, Stereotype Conformity, and Sexual Orientation for Participants who Believe Discrimination is Legal ..................... 47 Figure 6. Ratings of Applicants on the Basis of Beliefs About Employing Homosexuals, Gender Congruence, Stereotype Conformity, and Sexual Orientation for Participants who Believe Discrimination is Illegal ..................... 48 vii INTRODUCTION Sexual orientation issues, including discrimination, have been traditionally studied only by social psychologists, sociologists, and similar professionals. However, discrimination against members of non-traditional sexual orientations in the workplace has important implications for many of the topics traditionally studied in HQ psychology. Such discrimination can have adverse effects on an individual’s physical and psychological well-being and career, as well as on the organization. One result of discrimination for the individual is workplace stress. Stress has been shown to cause poorer performance (e. g. Jackson & Schuler, 1985), lower job satisfaction (e. g. Netemeyer, Johnston, & Burton, 1990), as well as other psychological and physical problems (e. g. Schuler, 1985). While there are myriad factors influencing the amount of stress an individual feels, the pressure to hide one’s sexual orientation and the fear of being discovered contribute to the overall amount of stress one feels in the wOrkplace (Croteau, 1996; Winfield & Spielman, 1995). In addition to stress acquired as a result of hiding one’s orientation, stress could develop if one’s sexual orientation were known and one is consequently discriminated against. Another potential negative effect of hiding one’s sexual orientation is the restriction of promotions and advancement opportunities for the individual (Escoffier, 1975; Gonsiorek, 1993). Homosexuals, in order to conceal their sexual orientation from their peers and supervisors, will refrain from discussing their private lives around the water cooler, and will not participate as much in social activities. Further, in some organizations it is expected that executives entertain with family -- homosexuals, even those involved in committed relationships, will be hard-pressed to host these social obligations in the “expected” way. These behaviors contribute to the informal networking that is crucial to timely career advancement in organizations. Discrimination on the basis of sexual orientation may also have adverse effects for the organization. Although there has been no mention of sexual orientation in federal civil rights or equal employment legislation, local and federal courts have at times ruled against organizations that used sexual orientation as a basis for the firing or refusal to hire individuals (Siegel, 1991). The verdict in Gay Law Students Association v. PT&T (1979), for instance, stated that government employers in California could not discriminate in hiring on the basis of sexual orientation. However, the most recent Supreme Court ruling has made it clear that state and local governments may outlaw homosexual behavior (Bowers v. Hardwick, 1986; cited in Herek, 1990), which would imply that discrimination in selection might also be legally acceptable. Even though federal legislation and litigation have not overwhelmingly upheld the rights of homosexuals, many state and local governments have included homosexuals in their lists of groups protected from employment discrimination (Riccucci & Gossett, 1996; Sega], 1995). As of 1996, 15 states and 142 county and city governments have, by means of legislation or executive order, declared it illegal to discriminate against homosexuals, although survey results indicate that these policies may not be well enforced (Riccucci & Gossett, 1996). Thus, at least in those locations, research on employment discrimination on the basis of sexual orientation is immediately relevant and necessary. While the individual and organizational effects of sexual orientation discrimination are significant, they would be less meaningful if such discrimination did not actually occur. Thus, it is also necessary to establish that discrimination is prevalent in the workplace. Some research has tried to establish the actual rate of sexual orientation discrimination. Levine and Leonard (1984) asked a sample of 203 lesbians about their experiences with and beliefs about employment discrimination. Sixty percent of the sample expected that, if their sexual orientation were discovered, that they would be discriminated against in some fashion. An assessment of self-reported discrimination revealed that 25% of their sample had experienced discrimination. Of these, 29% had been fired or refused employment, and 10% claim to have been denied advancement opportunities on the basis of their sexual orientation. This study has serious limitations ~- it is not clear how accurate these self-report measures can be, and the restriction of the sample to women also confounds the results with sex discrimination. However, the results do suggest that employment discrimination on the basis of sexual orientation may exist. A more recent and more controlled study was conducted by Croteau and vonDestinon (1994). This research assessed 249 homosexual college student affairs professionals. According to this report, 26% reported that they had been victims of discrimination on the basis of their sexual orientation. Croteau and vonDestinon found that the rate of the discrimination reports was significantly higher than for heterosexuals even after controlling for sex and the type of sexual orientation (homosexual versus bisexual). Badgett (1995), using a sample of 1,680 individuals acquired from the National Opinion Research Center’s General Social Survey, reported that homosexual and bisexual males (as determined by self-reports of behavior) made significantly less money than their heterosexual counterparts after controlling for other variables often associated with discrimination (race, age, etc.). The relationship between sexual orientation and income yielded a correlation of .28, and accounted for as much as a 24.4% difference in salary. The difference between lesbians / female bisexuals and heterosexual females was not significant, perhaps because of the relationship of sex with lower salaries regardless of sexual orientation. In summary, research on workplace discrimination on the basis of sexual orientation usually has been limited to homosexuals’ subjective perceptions of discrimination. These perceptions should not be taken lightly, nor should they be accepted as immediate proof that sexual orientation discrimination in the workplace is rampant. This study attempts to add further to this literature - what are the antecedents and mitigating factors of discrimination? Though specific research on this topic has yet to be published, more general studies of sexual orientation attitudes and discrimination have been conducted and are widely available in the traditional social psychology, sex research, and homosexuality journals. Although many of these studies are now quite old, more recent studies have confirmed previous findings. The whole of the literature does not dispute the fact that homosexuals are discriminated against; the foci of the studies have traditionally been on the factors determining an individual’s propensity to discriminate or to have negative attitudes toward homosexuality. Introduction to the Model A model is proposed depicting factors relevant to an individual’s propensity to discriminate in employment selection on the basis of sexual orientation. As justification for the model, a literature review is provided, primarily drawn from research in the domains of social psychology and general sexual orientation discrimination research (not specific to employment discrimination). Throughout the review, hypotheses specific to sexual orientation selection discrimination are proposed on the basis of the findings of previous research. First a general theoretical model is outlined, followed by descriptions of the variables included in this study. Following the introduction of the model, each factor is considered in depth through a literature review, and specific hypotheses are proposed. The theoretical model can be found in Figure 1. An individual’s attitudes towards homosexuals is determined by several factors: religiosity, beliefs about the controllability of homosexuality (i.e. whether or not a homosexual has a choice in sexual orientation), beliefs in specific sex roles, previous exposure to homosexuals, age, level of education, and the regions in which the individual grew up and currently lives. Further, an individual’s attitudes towards homosexuals will influence beliefs about hiring homosexuals. One’s beliefs about hiring homosexuals will, in turn, lead to intent to discriminate against homosexuals, as well as toward actual discrimination. The individual’s perceptions of the legality of such discrimination, the gender congruence between the individual and the target, and the extent to which the target conforms to traditional stereotypes will moderate the relationship between one’s beliefs and one’s intent to discriminate, as well as the relationship between one’s beliefs and actual =23£Et85 scrub—om 5:855 33% Co .032 365285. ._ oSwE 5E3— :ocaua—um .3 :33 ow< no.3— xom 32022:. E bozom mfiaxomeEeg oEEt35 mime—9:“. "2.533.32— =oua£EtOm§ 2 E2:— Soaa mug—om ESE. 8352 $330850: E 5:52.828 2: Sosa .E—om casein—E86 ._o bane. 05 .3 mcoaaoflom .EEBEOO cabs—Ba occuEwcoo x8 6th \ Sufism 5:56.582 €233. discrimination. Finally, one’s intent to discriminate against a homosexual will lead to actual discrimination. This study will test a subset of the factors proposed above. The omission of variables from the present research was due to several reasons, including the probable restricted range of some of the variables in the present sample. See Figure 2 for a depiction of the model to be tested. Factors Associated with Attitudes Toward Homosexuals Beliefs about the Controllability of Homosexuality A key tenet of attribution theory (e.g. Weiner, 1985) is that an individual will be rated more negatively by others if they possess a characteristic that is perceived by others as being both negative and under the individual’s control. Thus, the degree to which a homosexual individual is perceived to have control over sexual orientation should be a factor in determining the strength of an individual’s attitudes toward that homosexual. Whitley (1990) found that the perceptions of controllability of homosexuality were significantly related to attitudes toward homosexuals ([3 = .25; p < .001 ). VanderStoep and Green (1988) also found that belief about the control of one’s homosexuality correlated with attitudes toward homosexuality (R2 = .03). Aguero, Block, and Byme (1984) found that belief about the origin of homosexuality accounted for 25% of the variance in attitudes toward homosexuality. Hypothesis 1: Beliefs about the controllability of homosexuality will be related attitudes toward homosexuals, with subjects believing that homosexuality is not a choice having more favorable attitudes toward homosexuals. :o..m=.Et8.Q 5.82% 5.35:0 =3wa .8 .08.). no.3... .N 2%.... mfisxomoEo: o. Bamoaxm 3038.“. 8.3. xom .3358... c. .o..om ”220M... 25.0825. .050 £35852. £35852. cozacwztuma 3.3.95 .532 Sean m.o._um 33...? co..a:.E.._um.u .8 £33. 2.. .0 22388.. bite—.8 2.50296 3539.8 5% 635% \ 80.35 ”80.89.53. 5.338059. .8 b..£a..o::eu 2.. 59... 3:8. 5.8.33 Previous Exposure to Homosexuals Most research has indicated that an individual who knows homosexuals will have more positive attitudes toward homosexuals (Berkman & Zinberg, 1997; Bowman, 1979; D’Augelli, 1989; D’Augelli & Rose, 1990; Douglas, Kalman, & Kalman, 1985; Gentry, 1987; Glassner & Owen, 1976; Hansen, 1982b; Herek, 1986). Millham, San Miguel, and Kellogg (1976) found that knowing a homosexual significantly reduces one’s personal anxiety about homosexuals, repression of homosexuals, and one’s moral reprobation against homosexuals. According to Herek (1986), several factors influence the effect of exposure to homosexuality. The contact with a homosexual must be intimate and ongoing, not brief and superficial. The individual and the homosexual must share a range of other beliefs. The most influential contact will occur when the individual and the homosexual have shared goals, instead of being in competition. The homosexual must be “out” about their sexuality; individuals do not change their attitudes if they merely suspect an individual’s homosexuality. Hypothesis 2: Having had previous significant exposure to homosexuals will correlate positively with attitudes toward homosexuals. . Sex Roles Sex role beliefs, which are the beliefs that individuals of different sexes should have specific and separate roles in society, have been studied in relation to attitudes toward homosexuality in a plethora of research (Anderson, 1981; Hansen, 1982a; Krulewitz & Nash, 1980; Laner & Laner, 1980; MacDonald & Games, 1974; Macdonald, Huggins, Young, & Swanson, 1973; Marsiglio, 1993; Minnigerode, 1976; Newman, 1989; Smith, Resick, & Kilpatrick, 1980; Stark, 1991; Storms, 1978; Thompson, Grisanti, & Pleck, 1985; Weinberger & Millham, 1979; Whitley, 1987). For instance, Smith et al. (1980) found that profeminists were more likely to be accepting of homosexuals. MacDonald et a1. (1973), in one of the earliest studies of its type, found that belief in sex roles correlated .65 with attitudes toward homosexuality, and Krulewitz and Nash (1980) similarly found a .65 correlation between attitudes toward feminism and attitudes toward homosexuality. Other research has explored this variable further: Hansen (1982a) administered a ten-item scale to measure one’s belief in traditional sex roles to a sample of 206 undergraduates, along with a scale measuring attitudes toward homosexuality. He found that the correlation between sex-role orientation and negative attitudes toward homosexuality to be .48 for men and .55 for females (p < .001 in both cases). Laner and Laner (1980) measured 51] undergraduates’ liking of lesbians who differed in femininity. According to their research, the most liked lesbian was the one who was feminine, then the one who was effeminate, and lastly the masculine, or “butch,” lesbian. These results lend support to the belief that homosexuals are discriminated against because they violate sex role norms. Storms (1978) presented subjects with descriptions of an individual that varied with respect to both sexual orientation as well as feminine characteristics (as portrayed by college major, pastimes, and dress). Storms found a significant interaction between sexual orientation and sex role (F = 26.96; p < .001). Storms concluded that people dislike homosexuals because they violate sex role norms. Millham and Weinberger (1977) found similar results. They administered the Bem Sex Role Inventory to subjects 10 and then assessed their attitudes toward homosexuality. According to their findings, subjects were more likely to discriminate against gay men (by indicating their willingness to continue participation in the experiment with individuals presented as homosexuals) who violated sex role stereotypes. The sex role of lesbians did not significantly relate to the amount of discrimination directed at them. However, there is at least one very interesting incongruous finding in sex role research. While women’s attitudes toward homosexuality were not significantly different than those of men, Weinberger and Millham (1979) found that female subjects were more likely to support traditional sex roles. These results might imply that the predictiveness of sex on attitudes toward homosexuality is moderated by support of traditional sex roles. Another issue related to homosexuals’ violation of sex role norms is that of sex roles in business. A number of studies (e.g. Schein, 1973, 1975) have indicated that subjects’ concepts of good management traits are more similar to their concepts of male, rather than female, traits. With respect to homosexuality, it appears that individuals who hold the stereotype of a gay man as feminine might see an effeminate man as less qualified for a management job. Hypothesis 3: Belief in traditional sex roles will be related to attitudes toward homosexuals such that stronger traditional beliefs will correlate more negatively with attitudes. While the above hypothesis merely considers sex role beliefs as predictors, later sections of this thesis will further develop hypotheses related to the moderating effects of gender congruence (i.e. whether the target is of the same gender of the subject) and sex role 11 stereotype congruence (i.e. whether the target conforms to stereotypes regarding sex roles). Religion / Religiosity Many of the contemporary religious leaders decry homosexuality as a sinful act. Therefore, it would be expected that religion would be found a significant predictor of attitudes toward homosexuality. N yberg and Alston (1976), in a national sample of 30,018 subjects, found that religion was a significant predictor of attitudes toward homosexuality. Protestant and Catholic subjects were less tolerant of homoSexual orientations, while Jewish subjects and subjects with no religious affiliation were more tolerant. Bowman (1979), however, in his study of New Zealanders, found that religious affiliation had no effect on attitudes toward homosexuality, although those results might indicate that minority sects (Mormons, in his case) would have more negative attitudes. One anomaly exists in the literature. Douglas et al. (1985) were not able to demonstrate a correlation between religion and attitudes toward homosexuality. This is a more recent study than most of the others, indicating that the relationship between attitudes toward homosexuality and religion may be changing. In contrast to religion, which has primarily been measured by a one-item categorical measure of denomination, religiosity is a broader measure of an individual’s beliefs. Gentry (1987), Herek (1988), Irwin and Thompson (1977), and Larsen, Reed, and Hoffman (1980) measured religiosity in terms of frequency of church attendance (it is assumed that their studies was primarily targeted toward Christianity), and found that individuals who attended church more frequently tended to have more negative attitudes toward homosexuality (correlations found were around .45). Nyberg and Alston’s (1976) 12 study, however, explored this relationship in more depth, and found that frequency of church attendance interacted with denomination in predicting attitudes toward homosexuality. In some denominations, increased church involvement was correlated positively with attitudes toward homosexuality, while in other cases the correlation was negative. Berkman and Zinberg (1997) reported that their one-item dichotomous measure of religiosity correlated significantly with attitudes toward homosexuals. The most thorough coverage of the predictive nature of religiosity to attitudes toward homosexuality was published by VanderStoep and Green (1988). They used a scale developed by DeJong, Faulkner, and Warland (1976) that broke religiosity down into six factors: Religious Belief (orthodoxy), Religious Practice, Religious Knowledge, Religious Experience, Individual Moral Consequences, and Social Consequences. The authors combined the Religious Belief, Religious Practice, Religious Knowledge, and Religious Experience factors into a Religiosity variable (with an internal consistency of .90). The resulting path model indicated that religiosity was a significant predictor of attitudes toward homosexuality, as well as a determinant of two other factors that predicted attitudes (Ethical Conservatism and the belief in the controllability of homosexuality). These variables, as well as the sexual orientation of the subject, predicted attitudes toward homosexuality with an R2 of .5. Hypothesis 4: Religiosity will correlate negatively with attitudes toward homosexuals. A facet of most contemporary religious beliefs is that heterosexuality is primarily the natural sexual orientation of all individuals. That is, while some individuals may have some tendencies toward homosexuality, these tendencies should be able to be easily l3 controlled. Thus, it would follow that an individual who embraces this belief would not be inclined to believe that some individuals would possess uncontrollable natural tendencies towards homosexuality. For this reason, it is expected that religiosity should be related to controllability beliefs in the manner previously reported by VanderStoep and Green (1988). Hypothesis 5: Religiosity will relate to beliefs about the controllability of homosexuality, with higher religiosity being correlated with the belief that homosexuality is a choice rather than a trait that cannot be controlled. The above factors include those that I predict will have direct effects on one’s attitudes toward homosexuals. At this point, before I progress to hypothesized moderators, it is appropriate to formally propose the other direct effects I predict. The literature on the relationships of attitudes to behavior (e.g. Ajzen, 1985; in Ajzen, 1988; Fishbein & Ajzen, 1974) theorizes that an attitude toward an object is one of the, but not the only, influence on actual behavior. Thus, an individual’s attitudes toward homosexuality should be antecedent to actual discrimination. Most of the previous research on sexual orientation attitudes and discrimination has focused on one or the other side of this link — they have been either studies of attitude formation, or studies of the prevalence of discriminatory behavior. This research will attempt to study also the link between attitudes and behavior. Following Ajzen’s framework (i.e. the theory of planned behavior (Ajzen, 1985; in Ajzen, 1988), the more general construct of attitudes toward homosexuals will be related to the more specific construct of beliefs about hiring homosexuals. l4 Hypothesis 6: An individual’s attitudes toward homosexuals will be related to the individual’s beliefs about hiring a homosexual, such that more positive attitudes toward homosexuals will lead to more positive beliefs about employing homosexuals. In the theoretical model, it was predicted that beliefs about hiring a homosexual would have a direct effect on discrimination, as well as a mediated effect on discrimination through an individual’s intent to discriminate (as is consistent with Ajzen’s (1985, in Ajzen, 1988) theory of planned behavior). Because of the nature of this experiment, the measures of attitudes and antecedents of attitudes were collected after the subjects have had the opportunity to discriminate or not discriminate, making it impractical to measure an individual’s intent to discriminate. Therefore, the only hypothesized link between beliefs about employing homosexuals and actual behavior will be direct (with moderators as discussed later). Hypothesis 7: Beliefs about employing homosexuals will be related to an individual’s actual discrimination against homosexuals. The final part of the model involves the three factors hypothesized to moderate the relationships between beliefs about employing homosexuals and actual discrimination against homosexuals: Subject / target gender congruence, stereotype conformity of the target, and subjects’ perceptions of the legality of sexual orientation discrimination. Subject / Target Sex Congruence The relationship between sex and an individual’s attitudes toward homosexuals has been inconsistent. Many researchers have failed to find a relationship (Bowman, 1979; Henley & Pincus, 1978; Hudson & Ricketts, 1980; Irwin & Thompson, 1977; 15 Laner & Laner, 1980; Levitt & Klassen, 1974; Newman, 1989; Nyberg & Alston, 1976; Smith, 1971; Storms, 1978; Whitley, 1987). However, contrary evidence does exist. Aguero et al. (1984), Berkman and Zinberg (1997), D’Augelli (1989), D’Augelli and Rose (1990), Glassner and Owen (1976), Hansen (1982b), Kurdek (1988), and Minnigerode (1976) found that female subjects were more tolerant of homosexuality. Another exception to the bulk of the research was reported by Douglas et a1. (1985), who reported that female subjects had more negative attitudes toward homosexuality than the male subjects did. Another exception to this majority was published by Brown and Amoroso (1975), who found that females had more positive attitudes toward homosexuals than their male counterparts. However, since this sample was comprised of a combination of Brazilian and Canadian subjects, it is unclear as to whether sex has been confounded with ethnicity in this case -- the authors did not report male and female attitude scores broken down by nationality. Kite (1984) conducted a meta-analysis of all studies relating gender to attitudes toward homosexuality and found that, when one weighed the observed effects by their respective sample sizes, the main effect size of gender virtually disappeared ( p = .084). However, when the sex of the target homosexual is taken into account, an interaction effect with subject sex appears to exist. Gentry (1987) found that the correlation between frequency of religious participation and attitudes toward homosexuality was moderated by the sex of the homosexual target. When the homosexual being “rated” was female, the correlation between religious participation and attitudes toward homosexuality was .3 for women subjects but only .17 for men (p < .001 and p < .05, respectively). When the target homosexual was male, the correlation was .22 16 for female subjects and .19 for male subjects (p < .01 and p < .05, respectively). VanderStoep and Green (1988) found that male subjects disliked gay men more than lesbians, while female subjects did not differentiate. Herek (1988) reported a subject gender by target gender interaction in attitudes toward homosexuals. Similarly, Smith, et a1. (1980) found that there was no sex difference in attitudes when the target homosexual was male, but that men were more accepting of lesbians than their female counterparts. Kite’s ( 1984) meta-analysis confirms this effect. Two subsequent exceptions to this finding exist: Emulf and Innale (1987), using a sample of 81 Swedish subjects, and Lieblich and Friedman (1985), using Israeli and American subjects, were not able to produce this effect. Given the majority of these data, however, it appears that negative attitudes towards homosexuals are more salient, at least for men, when the target homosexual is of the same sex as the individual. Hypothesis 8: The congruence between the sex of the individual and the sex of the target will moderate the relationship between the individual’s beliefs about employing homosexuals and their actual discrimination against the target homosexual, such that the relationship will be stronger when the sexes are congruent. Stereotypical Qualities of Lesbians and % According to much of the impression-formation literature, prototypes or stereotypes have significant effects on how objects or individuals are perceived, and these effects have important implications in the study of attitudes toward homosexuals. Simmons (1965), while discussing labeling theory, asserted that individuals, when thinking of a certain group. “overestimate within-group similarity and between-group 17 differences” (p. 225). Lord, Lapper, and Mackie (1984), for example, assert that if a member of a minority group fits the prototype an individual has of that minority group, that individual is more likely to behave toward the minority group member in a way more consistent with that individual’s attitudes toward that minority group member. Congruence with stereotypes would therefore be an important moderating variable. Thus, it is crucial that we ascertain how homosexuals are stereotyped, in order to control for the effects of stereotype congruence. Much of the research on attitudes toward homosexuality also assessed opinions about homosexuality. Bowman (1979) found that gay men are seen to have an effeminate attitude, posture, and dress, and are very emotional and sensitive. Lesbians, similarly, are seen to be tough, aggressive, “butch,” and generally masculine in temperament and behavior. Religious affiliation also has been shown to be stereotypically perceived by the general population. Jenks’ (1988) research showed that the general population overestimated the percentage of gay men who had no religious affiliation, as measured by a sample of gay men. Political affiliation was also stereotypically perceived. Jenks (1988) found that straight individuals underestimated the percentage of gay men who were Democrats, and overestimated the percentage of gay men who had no political affiliation. Levitt and Klassen (1974), in a somewhat dated yet frequently referenced article, discuss the differential amount of support individuals show for homosexuals depending on the occupation of the homosexual. According to their research, homosexuals are not as welcome in occupations of authority and responsibility (such as judge, teacher, 18 minister, and doctor), and are more welcome in more stereotypically feminine occupations, such as beautician, artist, musician, and florist. These attitudes, as measured by percentages of subjects approving of homosexuals in these occupations, ranged from 22.8% approval for homosexuals as judges to 86.8% approval of homosexuals as florists. These results, while seeming to indicate a distinct sex role norm for homosexuals, may merely be an outdated stereotype that has fallen apart in the more than two decades since the study’s publication. J enks (1988), in research on labeling theory, found that perceptions of homosexuals were subject to what he called generalizability -- members of groups perceived as deviant in one way would be perceived as deviant in other ways. Jenks measured the perceptions of 146 subjects and found that, as opposed the mainstream groups of Catholics and Republicans, homosexuals were seen to be of a lower socioeconomic Status, used drugs more, and were more liberal. According to the perceptions of the subjects, 92% of homosexuals were liberal politically (Bell and Weinberg (1978) found that only 57% of homosexuals defined themselves as liberal). These findings pose an interesting question: sex role research has shown that homosexuals may be disliked more because they violate traditional sex role norms, but to what extent can a violation of homosexual stereotypes alleviate negative attitudes on the part of the observer? Glick, Zion, and Nelson (1988) showed that women who break the traditional stereotype of women were less discriminated against for a traditionally masculine job than were stereotypical women. Other examples of such results can be found in Heilman’s (1995) review of sex discrimination. Perhaps this phenomenon can also be applied to attitudes toward homosexuals. That is, a homosexual is already 19 targeted as aberrant because of the violation of sexual orientation, and is already given attributions as to personality, occupation, ideology, etc. What effect will a violation of the stereotype of a homosexual have on attitudes toward the homosexual who violates them? Hypothesis 9: The extent to which a target homosexual conforms to traditional stereotypes of homosexuals will moderate the relationship between an individual’s beliefs about employing homosexuals and actual discrimination against that target homosexual, so that stronger relationships between attitudes toward homosexuals and actual discrimination will be found for conforming target individuals. For the purposes of this study, the stereotypes of homosexuals on which I will focus will be the stereotypes of masculinity and femininity (i.e. a feminine gay man, a masculine lesbian). In contrast to the belief in traditional sex role beliefs construct described earlier, stereotype conformity is a characteristic of the target, not the subject. It is possible that these two moderators may interact with each other (i.e. stereotype conformity might be more relevant when the target is of the same gender as the subject), but no specific hypotheses were proposed. Perception of the Legality of Sexual Orientation Discrimination An individual’s behavior is certainly not guided solely by attitudes and beliefs. Obviously, there are other factors that contribute to the formation of the individual’s behavior. For instance, an individual's desire to perform a specific action (e.g. discriminate) can be overruled by another desire (e.g. to not be cited or sued). In terms of sexual orientation discrimination, the individual’s perceptions of the consequences of 20 discrimination should be a factor in determining the individual’s intent to discriminate against homosexuals, as well as actual discrimination. In terms of most laws (for instance, speeding), knowledge of the law could be argued to be a constant; most individuals are aware of the law — they know the posted speed limit and the consequences for exceeding that limit. However, the legal guidelines on sexual orientation discrimination are not as clear — they vary from municipality to municipality, and state to state. Additionally, the consequences for violating those assumptions vary across government lines. It is probable that such increased variability is reflected in individual differences in perceptions of the legality of sexual orientation discrimination, and would likely influence the extent to which attitudes are linked to actual discrimination. The analogous literature in the sex and race discrimination domain has not resulted in promising findings on this variable, however. Rosen and Mericle (1979) did not find any difference in sex discrimination on the basis of whether the anti- discrimination policy was strong or weak (although they did find a significant effect on recommended starting salaries). However, perhaps merely having a policy at all was enough to bias the results; a manipulation of policy / no policy might have produced an effect. Arvey (1988) suggested that knowledge of Equal Employment Opportunity and other similar issues might account for interviewers’ higher ratings of African-American candidates than of white candidates. With an issue as confusing and variable as sexual orientation anti-discrimination policies, an effect is more likely. Hypothesis 10: An individual’s perception of the legality of sexual orientation discrimination will moderate the relationship between beliefs about employing homosexuals and actual discrimination against homosexuals, so that individuals 21 who have negative beliefs about employing homosexuals will be less likely to discriminate against homosexuals if they believe such discrimination to be illegal. Other Predictors of Attitudes toward Homosexuality Not Included in the Tested Model Agg. Some of the earlier studies in sexual orientation research (Levitt & Klassen, 1974; Smith, 1971) were not able to find a correlation between age and attitudes toward sexual orientation. However, subsequent research has contended the original findings. Several studies have found age to be a significant predictor of attitude (Berkman & Zinberg, 1997; Bowman, 1979; Irwin & Thompson, 1977; Nyberg & Alston, 1976). According to these findings, older individuals are more likely to have negative attitudes toward homosexuals, probably attributable to longitudinal changes in our society. Correlations between age and attitudes toward homosexuals have been reported at around .35. The opposite of these findings was found by Hudson and Ricketts (1980) and Kurdek (1988), but they may have used samples that had restricted ranges in terms of the age of subjects (i.e. younger college students have more positive attitudes toward homosexuals than older college students). It is possible, therefore, that a curvilinear relationship exists, but the literature is not yet strong enough for a definitive answer on this question. It was decided to drop this variable in the present study for methodological reasons. Subjects for this experiment were drawn from an undergraduate subject pool. Thus, such a restricted range in age would have resulted in a fruitless investigation. Education. Most studies have indicated that individuals with higher levels of education have more positive attitudes toward homosexuals (Hudson & Ricketts, 1980; Irwin & Thompson, 1977; Nyberg & Alston, 1976; Simmons, 1965). Irwin and Thompson’s study regressed attitudes toward homosexuals on education and found that education had a standardized regression coefficient of .33 (R2 = .15). However, Berkman and Zinberg (1997) and Bowman (1979) found no education effect. Similarly to the age variable, level of education has been dropped from the current study due to restricted range. m. Research that has included measures of region (Hansen, 1982b; Herek, 1988; Levitt & Klassen, 1974; N yberg & Alston, 1976) has found that subjects in a rural environment are less likely to tolerate individuals with homosexual lifestyles. Additionally, research that has compared what kind of environment in which an individual currently lives to where they were raised has found some support for an individual’s location helping to determine attitudes toward homosexuality. Stephan and McMullin ( 1982), using an national sample, found that the size of the individual’s current city correlated with attitudes toward homosexuality at .63, and further that the size of the city an individual lived in at age 16 correlated with attitudes toward homosexuals at .8, indicating that, whatever else may contribute to the cause of the attitudes, it may be formed and solidified before adolescence. While the students (and thus potential subjects) at Michigan State University might be more heterogeneous in terms of region than other nearby educational institutions, a restricted range problem still likely exists. In addition to being comprised mainly of Midwestern students, the subject pool also currently lives in the same area (East Lansing and vicinity). Therefore, this variable was not included in the present experiment. 23 Demographic Variables Not Found to Affect Attitudes Toward Homosexuality The accumulated research on attitudes toward homosexuality has revealed several demographic variables that do not seem to be related to such attitudes. Levitt and Klassen (1974) and Irwin and Thompson (1977) did not find socioeconomic status to be related to attitudes toward homosexuality, although Irwin and Thompson did find evidence to suggest that family income was weakly related. Other factors not found to be correlated with attitudes toward homosexuality are marital status (Berkman & Zinberg, 1997; Bowman, 1979; Douglas et al., 1985; Levitt & Klassen, 1974) and race (Irwin & Thompson, 1977; Marsiglio, 1993). Summary An individual’s attitudes towards homosexuals are proposed to be derived from many factors. In this study, religiosity, beliefs about the controllability of homosexuality, beliefs in traditional sex roles, and previous exposure to homosexuals will be researched as likely predictors. It is hypothesized that these attitudes subsequently influence actual discrimination behaviors, with the strength of this relationship moderated by the congruence of subject / target gender, the target’s conformance to stereotype, and beliefs in the legality of discrimination on the basis of sexual orientation. Methods Participants Participants were recruited from the psychology department subject pool from April to October of 1998. Pilot testing was conducted from April to August, and the actual experiment was conducted primarily for two weeks in mid-September. After examining the preliminary data, it was decided to reopen the experiment to males in order 24 to increase male representation in the sample. A total of 236 individual participated, although 10 individuals were dropped due to incomplete or unlikely responses (e. g. respondents claiming that parents, siblings, children, spouses, ex-spouses, and individuals in all other categories were all homosexuals). The remaining participants were 77% female. Procedure This experiment used a 2 x 2 x 2 within-subjects design - gender of the target, the sexual orientation of the target, and the degree to which the target conforms to the stereotype of his or her sexual orientation were manipulated. It was decided to use a within instead of a between-subjects design because of the nature of the experiment — participant ratings of resumes are the dependent variables, to be evaluated in light of several attitude measures. In a between subjects design, resume ratings would have been confounded with attitudes. Participants were presented with a variety of resumes representing the above fully—crossed design, along with eight other lower-quality resumes added to the packet in order to blind the participant as to the nature of the experiment (resumes can be found in Appendix A). This procedure is in alignment with studies on sex and race discrimination that utilized resumes and job applications (e.g. Cash, Gillen, & Burns, 1977; Dipboye, Fromkin, & Wiback, 1975; Heilman & Saruwatari, 1979; Click et al., 1988; Marlowe, Schneider, & Nelson, 1996; McIntyre, Moberg, & Posner, 1980; McRae, 1991). It should be noted that by 1979, authors had already begun criticizing the “overreliance” on this technique, calling for other methods of examining discrimination (Arvey, 1979, p. 760). However, the research on sexual orientation discrimination is not as mature as research on sex and race discrimination. In order to relate this research to 25 the research on other types of discrimination, it was useful to use a resume manipulation approach to examine sexual orientation discrimination. In order to control for the masculinity / femininity of the job, the job for which participants were ostensibly hiring was chosen because it was gender neutral. Four potential occupations were originally chosen because of their equal gender balance according to the US. Department of Labor’s Bureau of Labor Statistics (1998). A pilot testing of job descriptions for these jobs showed Technical Writer to be perceived as the most gender neutral (see Appendix B for the job description used). The eight resumes to be used in analysis were originally created without reference to gender, masculinity / femininity, or sexual orientation. Iterative pilot tests were conducted in order to adjust the resumes so that they were perceived to be as equal in quality as possible. Next, masculinity / femininity was added to the resumes in the form of stereotypical hobbies and organizational affiliations, and iterative pilot tests were again conducted to establish the perceptions of the resumes as equally feminine or equally masculine. Finally, gender and sexual orientation were added (the former by the name on the resume, and the latter by membership in sexual orientation-related organizations). A manipulation check was conducted in order to determine 1) whether participants noticed sexual orientation, and 2) whether participants could guess the purpose of the study. Participants were shown to have noticed the sexual orientation of the applicants. While approximately 30% of pilot study participants had some inkling as to the nature of the experiment, it was decided to proceed with the experiment. Participants were told that they would be asked to review several applications for a Technical Writer position and would rate the applicants, after which they would fill out 26 several questionnaires. To minimize any order effects in the presentation of resumes, the order of the resumes varied (i.e. half the subjects received the resumes in the reverse order. Analyses revealed no significant order effect in the ratings of applicants; the resumes in Appendix A are presented in one of the orders). Though it was hypothesized that the variables measured in the surveys are antecedent to discrimination, it was necessary to measure them after the resume evaluations because the content of the measures would almost certainly inform the subjects of the purpose of this study. Additionally, in order to minimize any effect the nature of the resumes had on responses to the measures, clear instructions were given to subjects asking for honesty in their responses, and again assuring them of the confidentiality of their responses (a c0py of the informed consent form is included as Appendix C). Measures Multidimensional Measure of Religious Involvement. This scale, included as Appendix D, was included as part of the National Survey of Black Americans (NSBA) in 1979 and 1980 (see Jackson, 1991). Levin, Taylor, and Chatters (1995) validated a three- dimension model of the scale with the NSBA sample of 2,107 subjects. Their model, analyzed with structural equations modeling, demonstrated excellent model fit (i.e. traditional fit statistics were greater than .95) as well as good construct and predictive validity. This 12-item scale was shown to measure three dimensions of religiosity - organizational, nonorganizational, and subjective religiosity. This scale is of particular interest because, contrary to the majority of religiosity scales depicted in the literature, it attempts to be religion-neutral in its assessment (i.e. it does not assume a Christian background). Though the scale was developed with African-Americans in mind, Levin et 27 al. hoped that it would be subsequently validated on other populations, and a subjective evaluation of the items by the author of this proposal did not result in any reason why this scale would function any differently in another population. This scale, as included in Levin et al.’s (1995) article, did not include the scale anchors. Therefore, anchors were created (see Appendix D for this study’s anchors to their scale), and this scale was included in the pilot test of this experiment in order to test both the psychometric properties of the scale with the added anchors and the dimensionality of the measure (this test was given to all pilot test participants, even as the nature of the resumes was changing). Me of Beliefaabout the Controllability of Homosexua_li_t_y. The only measure found regarding the controllability of homosexuality was developed and used by Levitt and Klassen (1974). The authors did not provide much information about the psychometrics of this scale, and indeed only reported percentages of each response option for each of the questions. This scale was reviewed as part of the pilot study for this experiment, and was found to possess low internal consistency reliability. Some items were dropped and new items added to form a more reliable measure (see Appendix E for the final measure). Measure of Sex Role Beliefs. The Attitudes Toward Women Scale (AWS, short version; Spence & Helmreich, 1978) is one of the most often used measures of sex role beliefs. It has been mentioned in the literature that the scale may soon become outdated (McHugh & Frieze, 1997; Spence & Hahn, 1997). The first author of the scale has reported recently that restriction of range on some of the items may be beginning to occur (Spence & Hahn, 1997). However, talk of its imminent demise may yet be unwarranted. 28 The most recent evaluation of the scale concluded that there is still “sufficient within- group variability for the scale to retain its utility” (Spence & Hahn, 1997, p. 31). Previous reports of internal consistency have been over .8 (Spence & Hahn, 1997). This scale can be found in Appendix F, and was coded such that higher values indicate beliefs in more prescribed sex roles. Measure of Previous Exposure to Homosexuals. This scale was recently developed by Berkman and Zinberg (1997). Berkman and Zinberg found adequate validity for the measure in their research, and it appears to cover a wide range of types of individuals with whom subjects may come into contact. This measure is presented in Appendix G. Attitudes Toward Lesbiarmmd Gays (ATLG) Scale. This 20-item scale was developed and validated by Herek (1988), and has been used in many of the studies conducted since then. The scale is composed of two lO-item subscales (one regarding male and the other regarding female homosexuals). Each subscale has been shown to be useful on its own, as well as a valid measure when combined. Herek’s initial validation study reported a coefficient alpha of .90 for the entire 20-item scale, .89 for the lO—item Attitudes Toward Gay Men (ATG) subscale, and .77 for the Attitudes Toward Lesbians (ATL) subscale. Even though the scale was produced over a decade ago, a cursory review of scale content showed that it still seemed appropriate as a measure in current society. Higher values indicate more positive attitudes toward homosexuals. See Appendix H for this scale. Scale on the Beliefs of Employing Homosexua1_8_. O’Brien and Vest (1988) developed a 13-item scale about the consequences of employing homosexuals (see 29 Appendix 1). Higher scores on this measure represent more negative beliefs about employing homosexuals. The potential consequences were developed from focus groups of managers. Each item was presented, along with a social desirability scale and two measures of discrimination for validity estimates, to 182 managers in manufacturing, government, and service organizations. Test-retest reliability was shown to be .9, and Cronbach’s alpha for the scale was .95. To assess predictive validity for the scale, a sum of the scale items was correlated with two one-item measures of discrimination -- one asking the subject whether they would hire a homosexual, and another asking the subject whether they would fire a homosexual on the basis of their sexual orientation. The first measure correlated with the scale at .55, and the second at .63. O’Brien and Vest also reported that the correlation between the two one-item discrimination measures was .69. Such positive results suggest that the scale merits study as a measure of the beliefs of employing homosexuals. Measure of Percepfitions of the Leggity of Sexual Orientation Discrimination. I was not able to find such a measure in the literature. Therefore, a new scale was developed through pilot-testing for the current study. See Appendix J for the final scale used. Measurement of Discrimination. Discrimination research that has used a resume methodology has measured discrimination in several ways. In addition to a mere dichotomous variable (hired or not hired), many studies that used a within-subjects design have also measured subjects’ ratings and rankings of the “applicants.” Both of these techniques have met with some success. Dipboye et al. (1975) found that, although both ratings and rankings illustrated gender discrimination, the technique of ranking, 30 especially subjects’ first choice, showed the most discrimination. Pazy (1992; cited in Marlowe et al., 1996) found gender bias only in rankings. Arvey’s (1979) review contended that a ranking strategy, because it employs a forced-choice strategy, is better able to detect subtle bias. Additionally, ranking of applicants may be more of a reflection of reality than ratings — in a selection situation, one wishes to select the most qualified individual for the job. In short, it appears that rankings would be a more powerful measure of discrimination. However, Marlowe et al. (1996) nevertheless recommend that, because both techniques are used in actual selection practice, both should be measured in research on discrimination. In this experiment, subjects were asked both to rate each applicant on a 100-point scale of hirability with five anchors, as well as by ranking the top four applicants they have seen. A description of the instructions and measures for this task can be found in Appendix K. Results The Attitudes Toward Women (Sex Role Beliefs), Controllability, Attitudes Toward Lesbian and Gay Men, Beliefs About Employing Homosexuals, and Legality of Discrimination scales were computed by summing the items in their particular scales. Means, standard deviations, internal consistency reliabilities, and scale intercorrelations can be found in Table 1. All scales displayed sufficient univariate normality and internal consistency reliability, although the mean on the Attitudes Toward Lesbians and Gay Men was toward the more egalitarian end of the scale, and the mean of the Legality Beliefs scale indicated that most individuals thought it was illegal to discriminate on the basis of sexual orientation. The Previous Exposure score was computed by summing the 31 amount of contact the participant had with each known or suspected homosexual individual. The mean and standard deviation for this variable can also be found in Table 1. In order to determine how many religiosity factors to use, a factor analysis was performed on the 12 Religiosity items (squared multiple correlations were used as prior estimates of the communalities among the variables). One factor was retained according to Kaiser’s criterion. Therefore, a single Religiosity scale was formed by summing the 12 items. The mean, standard deviation, and internal consistency reliability of this scale can be found in Table l. Tests of Antecedents of Beliefs about Employing Homosexuals Verification of Hypotheses 1-4 (that Controllability, Previous Exposure, Sex Role Beliefs, and Religiosity were related to Attitudes Toward Homosexuality) were conducted using correlations. Results of these tests can be found in Table 1. All four hypotheses were confirmed - Controllability Beliefs (32 = .27), Previous Exposure R2 = .03), Sex Role Beliefs (I;2 = .24), and Religiosity (E = .19) each were related to attitudes toward homosexuality. Similarly, Hypotheses 5 (Religiosity correlates with Controllability Beliefs) and 6 (that Attitudes toward Homosexuals would be correlated with Beliefs about Hiring Homosexuals) were also tested through correlations which can be found in Table l. Religiosity was correlated with Beliefs about the Controllability of Homosexuality (BE: .10), and Attitudes toward Homosexuals was correlated with Beliefs about Hiring Homosexuals (R_2 = .41). All correlations were in the hypothesized directions. 32 .0..m..0.m 80.000500 :0 .0: 0-03 >......0..0.. 55.0.0000 .0E0.=. .0... £000 00 005000... 0.03 0.00.0; 050091". 000305 0:... ..0>0. no. 0:. .0 50055.0 000 m. .- 00:. $0. .0 m.. 00:. 00.00% 20.0.0000 .w.0=0w0..0 0... c. 0.0 8.2.32.0. 000.2300 .050.0. ”0.02 ..w. R.- 8- 3.- E. 8. 0.. m. m :0 m... 802.55....033003 2.2.0200 s a. .0. 2. 00- on- 00.. mm _ om... 02 2258.8: 053.050.802.200 .0 00. 2. an- 0...- 3..- 8. .m 3.0. 0.00 massage: 203900202 .0 02 we. 8.. 3.- cm 0 2.0 3 23.8.0323... .4 00. 5. mm. 8 4 3... 0: 0:32.828 50200 ..m o... a. z. _ 000 0.2 02.00 23. am .0 3. mm N. .2 0.... £80.00. A 0 m 4 m N _ 52:52 52552 00 as: 2.2.; damn. .0003. _...0A. 0... :. m0.00..0> w00E0 20:00:00.0...— 000 00.2.5023. 3:020:00 .0E0E. .m:0..0.>0Q 0.00:0.m .2002 .. 050-.- 33 Tests of a Structural Eguations Model Antecedents of beliefs about employing homosexuals (i.e. the variables and relationships proposed in Hypotheses 1-6) were analyzed simultaneously through structural equations modeling. The analysis was conducted using AMOS 3.61. Internal consistency reliabilities were included in the model as estimates of error inherent in the manifest variables, in a procedure described by Hayduk (1987). A simplified path model showing only the latent variables and standardized causal paths can be found in Figure 3. Listings of direct and indirect effects for each variable in this model can be found in Table 2. Although many of the fit indices for this model suggested excellent fit (GFI=.94; NFI=.92; CFI=.93), the chi-square test was significant (x2(6, E = 228)=34.72; 9 =0.00), and the RMSEA was high (RMSEA=.14). All paths currently in the model were statistically significant. Religiosity and Sex Role Beliefs appeared to be strong mile 2. Direcflnd Indirect Effects for the Original Model Dependent Variable Independent Variable Direct Indirect Hypothesis Effect Effect Tested Controllability of Religiosity 5 Homosexuality .35 -- Attitude Toward 4 Homosexuals Religiosity -.62 -.09 Sex Role Beliefs -.7l -- 3 Controllability Beliefs —.27 -- 1 Previous Exposure to 2 Homosexuals . 16 -- Beliefs about Employing Homosexuals Religiosity -- .48 Sex Role Beliefs -- .48 Controllability Beliefs -- .18 Previous Exposure to Homosexuals -— -.l 1 Attitudes toward 6 Homosexuals -.67 -- 34 0.9.00 .coE>o_0:-.m km... .000... .0...m..0 0... ..0 E0.w0..0 50. .m 0.5m... 3:038on01 0.02.0... 09.5.36. N0.- 0.960 0:520:50 w... ..h. 050on 03030.... w.m..mm 0.0m xom 0.8.0.6.. 35 predictors of Attitude Toward Homosexuality, which was a strong predictor of Beliefs about Employing Homosexuals. Beliefs about the Controllability of Homosexuality appeared to partially mediate the relationship between Religiosity and Attitude toward Homosexuality. The Previous Exposure to Homosexuals variable was a weak but significant predictor of Attitudes Toward Homosexuality as well. An analysis of the modification indices suggested that a path be added from Sex Role Beliefs to Controllability Beliefs. This path was added, and the model was reanalyzed- A simplified version of the test of the revised model showing only latent variables and causal paths can be found in Figure 4. This model was very similar to the previous model, except that the path from Sex Role Beliefs to Controllability Beliefs was also significant. The addition of this path to the model improved the fit of the model (GFI=.99; NFI=.98; CFI=.99; X715, 3 = 228)=7.82; p =.17; RMSEA=.05). All paths in the model were significant, and no modification indices existed that would have incrementally improved the fit of this model at a .05 level. This model was thus accepted as the model that best fit the available data. See Table 3 for a listing of direct and indirect effects for variables in this model. Tests of the Moderators of the Relationship Between Beliefs about Employing Homosexualsjand Ratings of Resumes The characteristics of each resume, their mean rating, and the standard deviations of the ratings can be found in Table 4. Breakdowns of mean ratings for females vs. males, masculine vs. feminine individuals, and heterosexuals vs. homosexuals for the eight used resumes can be found in Table 5. There were no significant differences 36 .000... 000.5. 0... .0 5030.0 .000 .0 0.03.. 0.00093 00030.... m... 0.0..0m 0.0m xom 2.3300080... 0.02.0... 00.- 83...... 0.0..0m .00..:>0.0Em mm... m v.- 0.8.0.0.. 0.0:0m 3230:9500 37 Table 3. Direct and mdirect Effects for the Revised Model Dependent Variable Independent Variable Direct Indirect Hypothesis Effect Effect Tested Controllability of Religiosity 5 Homosexuality .49 -- Sex Role Beliefs .41 -- Attitude Toward Homosexuals Religiosity -.63 -.09 4 Sex Role Beliefs -.71 -.O7 3 Controllability Beliefs -. 18 -- 1 Previous Exposure to 2 Homosexuals . 15 -- Beliefs about Employing Homosexuals Religiosity -- .49 Sex Role Beliefs -- .53 Controllability Beliefs -- . 18 Previous Exposure to Homosexuals -- -. 15 Attitudes toward 6 Homosexuals -.68 —- between levels of these breakdowns - that is, widespread discrimination did not occur on the basis of applicant gender or sexual orientation (but see below for how applicant and subject characteristics interacted in producing discrimination). Repeated measures analyses using a high quality / low quality contrast also revealed significant differences in ratings on the basis of the intended quality of the resumes (i.e. the high quality applicants were indeed rated more favorably; E = 29.0; p < .01). A factor analysis performed on the 16 applicant ratings revealed that the low-quality applicants formed the first factor (homosexuals and heterosexuals comprised the other two factors). Tests of Hypotheses 7-10 were performed using repeated measures regression (via SAS’ PROC GLM). In this repeated measures regression, applicant ratings served as the 38 Table 4. Characteristics of the ResumesI Mean Ratings, and Standard Deviations Name Quality Gender Masculinity / - Sexual Mean SD Femininity Orientation Rating Sara Franks High Female Masculine Heterosexual 74.1 20.4 Christopher Grano Low Male 37.0 23.1 Brent Hawks Low Male 30.2 22.2 Mary Hirsch High Female Feminine Heterosexual 78.2 16.3 Nicholas Hoffman -Low Male 32.1 22.4 Wayne Jameson High Male Feminine Heterosexual 87.4 13.9 Douglas Kirgis High Male Feminine Homosexual 85.1 12.9 Janet McClellan Low Female 29.1 22.9 Laura Mercier Low Female 35.5 22.7 Victoria Niles Low Female 30.0 22.2 Nancy Pitcher High Female Feminine Homosexual 83.6 17.5 Jonathan Reyner Low Male 30.9 20.4 Abigail Simmons High Female Masculine Homosexual 77.9 18.7 Michael Smythe High Male Masculine Homosexual 75.7 22.9 Richard Snead High Male Masculine Heterosexual 82.3 14.9 Ellen VanBuren Low Female 40.3 22.0 39 Table 5. Breakdowns of Resume Ratings by Gender, Masculinity / Femininity, and Sexual Orientation (N=235) Comparison Level Mean _S_I_)_ Gender Males 82.6 10.4 Females 78.5 12.3 Masculinity / Femininity Masculine 77.5 13.1 Feminine 83.6 9.4 Sexual Orientation Heterosexual 80.5 10.7 Homosexual 80.6 13.8 (repeated) dependent variables. Beliefs about Employing Homosexuals, participant gender, and Beliefs in the Legality of Discrimination were independent variables. Some brief words are in order regarding the relationship between the repeated measures regression and the initial hypotheses. For the purposes of accuracy in reporting the results, care was taken to remain within the repeated measures regression framework and the variables as used therein. However, it may be noted that the variables in the repeated measures regression are not directly those in Hypotheses 7-10. This discrepancy is because the factors in Hypotheses 7-9 (i.e. sex congruence) are actually interaction terms of some of the variables and contrasts included in the repeated measures regression. First, consider Hypothesis 7 (that Beliefs about Employing Homosexuals would predict discrimination on the basis of sexual orientation). Discrimination on the basis of sexual orientation would be evident if the sexual orientation contrast were significant. Thus, a Beliefs about Employing Homosexuals x Sexual Orientation effect is appropriate to test 40 this hypothesis. Hypotheses 8-10 all predict moderators of this relationship. Thus, the tests of these hypotheses must include the Beliefs about Employing Homosexuals x Sexual Orientation effect. For Hypothesis 8 (sex congruence as a moderator), sex congruence is represented as a participant gender x applicant gender interaction. Thus, a 4-way interaction (Beliefs about Employing Homosexuals X Sexual Orientation x participant gender X applicant gender) is necessary to test this hypothesis (for a similar approach using gender, see Riordan and Shore (1997)). For Hypothesis 9 (stereotype conformity as a moderator), the effects are of a similar order. For an individual to conform to a sexual orientation masculinity / femininity stereotype, the sexual orientation, gender, and masculinity / femininity of the applicant all must be taken into consideration. Therefore, the Beliefs about Employing Homosexuals x Sexual Orientation >< Applicant Gender >< Masculinity / Femininity interaction is the effect to test for this hypothesis. Finally, for Hypothesis 10, Beliefs about the Legality of Discrimination is included in the repeated measures regression as its own variable. Thus, the Beliefs about Employing Homosexuals x Sexual Orientation X Legality Beliefs interaction is all that is necessary to test this effect. The characteristics of the applicants (gender, masculinity / femininity, and sexual orientation) were represented as orthogonal repeated measures contrasts. Because no repeated measures contrast had more than two levels, tests of sphericity did not need to be performed (6 is inherently 1 in this type of design). The assumption of the equality of the variance / covariance matrices of the dependent measures across levels of the independent variables (i.e. rater gender), however, was violated. Originally, I had thought that my hypothesized interactions would have produced such heterogeneity. 41 However, after performing all tests, the intercorrelation matrix of the residuals for each repeated measure still possessed heterogeneity on the basis of gender. Bartlett’s test of homogeneity of variances confirmed that variances of the ratings differed on the basis of gender primarily for homosexual applicants, with males having larger variances. Since the number of male participants was less than the number of women, an indirect pattern existed in the data, a phenomenon known to inflate the probability of obtaining Type 1 errors. An attempt was made to correct for the violation of this assumption by using a SAS procedure designed by Lix and Keselman (1995), but the necessary mathematical ability was beyond the grasp of this author. Therefore, it was decided to violate the assumption and interpret the repeated measures regression with the caution that the Type 1 error rate was probably high. The results of the repeated measures regression can be found in Table 6. The Sexual Orientation x Beliefs About Employing Homosexuals x Legality of Discrimination x Applicant Gender x Participant Gender x Masculinity / Femininity of Applicant effect was significant (F = 4.79; p < .05). This interaction was not hypothesized in this study, but the interaction does include all of the hypothesized interactions. Thus, none of Hypotheses 7-10 were directly supported, as the predicted direct and moderating effects depend on other variables (again, it should be noted that the inflated Type I error rate may have contributed to the significance of this effect). However, all is not lost - as will be explained below, many of the hypotheses were supported within at least one level of another variable. Without supporting theory, a 6- way interaction would normally be considered impossible to interpret. However, the 42 . . .N 0m.0 .0 . 0&0. 0 MmDZmquxxqume 00. N0. . N0. meZMOxroqmzmme 00. 00. . 00. 1....ngme 0m.m 00.000 . 00.000 A0402mxm2 N . . m0.0~ . 8.00 1.0.qume mm. . ..00 . . ..00 roamimxms. .00.... 0.00.0. . 0.00.0. ".2 0m.0.~ w . N 00.00000 2.00003 .0..m. m. .N N. .0m0 . m . .00 0 mmDZmOXAAQquwaAAEmXZmOAmxx 00. 0.0.0. . 0N.0. mmozm05<0qu2m000< 00. >0. . 00. meZme>0402mem000< 00. 0. .0. . 0. .0. memevmeOnE< 00. 0. . . 0. . 40402mx2m000< m . .N 3.00 . mm.~00 40402memOAE< .mm.0m 00.0000 . 00.0000 Zm00m< 08.3% 55.3 0&wa w.~ 00.0.0.0. .0..m 00.~ 00. m . 0m . 00.0 . 0N meZmOXAOAmEm .mm0 00009“ . 02.0mm amozm05<0mq mm. 00.0mm. . 00.0mm. meZmOx>0402m 0N. 0N.0.N . 00.0.N meZmD .m. 00:03 . $.03 A0402m w . . 00.00. . 00.00. 4’0qu 00. m... .00 . mu. . 00 >an..>.m. 0.00.3.5 0003.3. m m2 .0 Wm 00.3.0003. 005000.). 000.03% 0... .0. 0.00... 00.:0m .0 0.00... 43 .528. 0 2.5 .000 N0. . N0 . N0. . N0 meZmOmeOnEOAm$.me0 . 0. 00 .N0 . . 00 .N0 . MmDZmOXZmOnEOx52mem A020 00. 0. .0. . 0. .0. ZmOmmOAmzmem 0N0 N06 . 0 . N0 8.0 éOBXZmOmm/xx00 00. 0 0.00. . 00.00. ZmOaEOxazmem .0000 00.0000 . 00.0000 ZmOmm0402mxm2xzmmumm< N0. N00 0 . . N50 0 . MMQZmOxAOAmzmxmzmeOmm< 00. m . .00. . m . .00. MmQZmOmemeDmm< N0. 0N0: . 0N0... A0402mxm2meOmm< 0N. . 0N.00N . 0N.00 N AOAmzmx02meOmm< 00.0 No.03. . No.09 mEmeOnE< .5000 0 . N 0 0.00000 A00..0..m 00. 00.0N . 00.0N mmDZmOXAOAAEmXOm 00. 00. .0N . 0m. .0N mmDZmOXAOaEmeOm 00. . 00.000 . 00.000 meZmOxOm .0. No.0 . No.0 40402mx00 .0 . :0 00. 00.00. . 00.00. A0. . 00. >Ox.n..>.mx00 2.... N0. 0. .0 . 0. .0 00 00.00. 0.N 00.00.00 £06.85 00.0 00 .000 . 00.000 meZmOXAOAAmemE 44 $3.300. ..o 508:0..0 .0380 u 00 358.300. 00 5.5580". 1......0822 n n..>. 3:00:95 .0 50:00 u ZmOmnz 3506.50 00 .5050 u meZmO Eo..a:.Etom.Q 503505 .0380 00 00.003 n éOm‘. $338080... 0500.95. SO00. £2.00 N >0402m. .mlm . 2...... so 0800 80 38. m 0.5 £08.00 :32 £85.00 .0 2.50 “2'07... 45 00. v m . 0 . .00. 0 . N 00.00. 00 E2x2m0000402mx00 00. 00.00. . 00.00. amazm0x4<0qu2m000O..n..>.m.x00 00. .0. . .0. mzxmmDZmOmeOmmOAm20700 . ..N 00.000 . 00.000 mEmeOmmqu2mx00 00.. 00. . 0N . 00. . 0N mEme000quzmx00 .02. 00. 00.00. . 00.00. mEmeOmm0402mx00 00. 0. .N0. . 0. .N0. meZmOxm2xq0402mx00 .000 00.000 . 00.000 meZmOx02x00 00. 00.0. . 00.0. 02x4<0m4x>04020x00 0... 00. . 0N . 00.0. N mExq.x>ou.n..>.m.x00 ..N. .0 00.000. . 00.000. 02x00 00.00. 0 . N 00. . 0000 .zm000