![Sample our Medicine, Dentistry, Nursing & Allied Health journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5944/TOC-Subject-Sample-2021-Medicine-Dentistry-Nursing-Allied-Health.jpg"})
Abstract
Although attitudes toward homosexuality have become more liberal, particularly in industrialized Western countries, there is still a great deal of variance in terms of worldwide levels of homonegativity. Using data from the two most recent waves of the World Values Survey (1999–2004, 2005–2009), this article seeks to explain this variance by means of a multilevel analysis of 79 countries. We include characteristics on the individual level, as age or gender, as well as aggregate variables linked to specificities of the nation-states. In particular, we focus on the religious denomination of a person and her religiosity to explain her attitude toward homosexuality. We find clear differences in levels of homonegativity among the followers of the individual religions.
ACKNOWLEDGMENTS
The authors would like to thank Janika Spannagel and Carola Fricke for their work on earlier versions of this article as well as Lena Münke for improving the readability.
Notes
1. In this article, the word homosexual refers likewise to gay and lesbian people. Not only in public debates but also in surveys the difference between gay and lesbian people is not always explained—thus the word homosexual is often equalized with the word gay. We recognize that this may lead to a false interpretation of the attitudes toward lesbian people. Due to constraints of the data—not distinguishing between attitudes toward lesbians and gays— this problem cannot be sorted out. All statements and wording on homosexuality and homosexual people in this article are to be understood with this in mind.
2. We avoid using the term homophobia, as its etymological roots imply debates over the causes of fear of homosexual people. While these debates are legitimate, these debates are not to be mistaken for the structures of prejudice that are the focus of our analysis.
3. Due to its larger range of values (10-point scale), the first operationalization of homonegativity, perceiving it as an objection to a social practice, is better suited for the actual analysis than the neighbor variable—at least from a statistical point of view. Measuring on a 3-point scale and thus containing less information, this second operationalization seems to be predestined for the cross-check. An overall more precise operationalization of homonegativity is thwarted by the lack of data. In contrast to socio-psychological research, where much effort is made to construct homonegativity scales out of a wide range of interview questionnaire items (usually by means of a factor analysis) that best measure homonegativity (Davies, Citation2004; Hudson & Ricketts, Citation1980), the dependent variables used in this article have to be based on the existing data and therefore the questions of the WVS, which inevitably results in decreased accuracy. Many studies that do not conduct their own autonomous fieldwork and therefore must rely on secondhand data encounter this problem (Hooghe, Claes, Harrell, Quintelier, & Dejaeghere, Citation2010; Ohlander, Batalova, & Treas, Citation2005).
4. In addition to the here mentioned explanatory variables, a further set of individual factors is often tested through an implicit association test (IAT), particularly in psychology. These individual factors include, for example, having homosexual people as acquaintances, self-reported gender role characteristics (Black & Stevenson, Citation1984), as well as the essential extent of prejudice against minority groups or “right-wing authoritarianism.” Without having any primary data, these factors can nevertheless only be tested approximately using the data available from the WVS.
5. In his Christmas address to the Roman Curia in 2008, the Pope said that disrespecting the Catholic principle of “human being as man and woman” amounts to a “self-destruction of man himself, and hence the destruction of God’s own work” (Benedict XVI, 2008).
6. According to Jan-Erik Lane (Lane, Citation2008, p. 232), religious fundamentalism constitutes one of the greatest challenges for a global and open society. This is especially true when fundamentalist preferences are officially accepted or even actively supported in a country. In Islam (Roy, Citation2004), in Evangelical Protestant Free Churches (Harris, Citation1998), as well as in Hinduism (Bhatt, Citation2001), particularly fundamentalist trends can be found. Radical Hinduism goes along with a strong nationalist component that advocates, in particular, a distinction with respect to Islam (cf. the Pakistan–Kashmir conflict) and is therefore less concerned with the moral–religious level than Islamic or Christian fundamentalism.
7. Our coding of faith traditions is definitely a very crude measure, and we would have loved to improve it by parsing out more theologically conservative types of certain religions from their more liberal counterparts, as one of the anonymous reviewers rightfully observes. However, there is nothing like the RELTRAD measure (Steensland et al., Citation2000) that enables researchers to classify faith traditions in the United States in a very concise way (particularly regarding the nuances between the different Protestant denominations) that we could use for our analysis on a global scale. Neither is it possible based on the available WVS data to create a further religious attitudes measure using, for example, literalism or inerrancy. These factors are just not asked in the WVS. Yet we tried to build our categorization of religions on three criteria that cover a broad array of aspects relevant for the positioning of a religion with respect to the other faith traditions on a liberal–conservative scale. Furthermore, scriptural attitudes such as inerrancy or literalism are important for the question of religiosity, and we assume these two factors to be major determinants for the values of our religiosity measure (s. 2.1.2).
8. This assumption does not hold for every religious subgroup as, for example, in the United States there are certain denominations that have concrete gay and lesbian–friendly policies at work, such as the Evangelical Lutheran Church in America. This denomination opened its ministry in 2009 to “gay and lesbian pastors […] living in committed relationships” (ELCA News Service, Citation2009). We want to thank one of the anonymous reviewers for bringing up this fact. Yet these denominations are still exceptions proving the rule that religions, in general, tend to promote homonegativity.
9. There is difficulty in assigning some of the religions asked in the WVS (e.g., Bahaism, Jainism, Zoroastrianism) to the bigger religions; they were therefore placed in a residual category that will not be further analyzed.
10. Studies written from a Muslim perspective—and thus not being Western-biased—also show that (male) homosexuality is not as accepted in the Muslim world as, for example, in Western countries. These studies nevertheless emphasize that Muslims do not irrationally fear homosexuality, but they disapprove of it, and this disapproval would not be irrational (Halstead & Lewicka, Citation1998; Sarwar, Citation2004).
11. Catholicism and the Protestant Free Churches were assigned the same rank. This is because the religions do not differ in their holy texts; and though the Catholic leadership is more explicit in its rejection of homosexuality, this is offset by the greater influence of the fundamentalist currents of the Protestant Free Churches (Barton, Citation2010). With Orthodox Christianity being comparable to the Catholic Church in terms of its organizational structure as well as its rigidity of religious statements, it is placed under the same category.
12. In comparison to the measures that collect an implicit aversion, self-evaluations are subject to a greater risk of conscious manipulation and bias due to giving socially acceptable answers (Banse, Seise, & Zerbes, Citation2001).
13. Concerning the primarily studied subject of gender equality, Inglehart and Norris refer to works from sociology, social-psychology, and anthropology that view religion in general as one of the “most important agencies of socialization determining social norms and moral values with regard to gender equality” (Inglehart & Norris, Citation2003, p. 50). Furthermore, they see the active attempts of single religions to strengthen social norms such as the subordination of women as an indication of a connection between secularization and a growing support for gender equality.
14. While the composition of the post-materialism index is criticized as insufficient, the basic assertion of the trend of modernization is largely undisputed (Kadishi-Fässler, Citation1993; Klages, Hippler, & Herbert, Citation1992).
15. The intraclass-correlation coefficient (see section 3) shows that the cross-country differences are indeed considerable.
16. The purely economic argument can be linked to the shift from “survival orientation” to “self-expression” as described by Inglehart (Citation1977); it can be applied as a characteristic of value change and its correlation to changed attitudes toward socially controversial topics such as prostitution or gender equality has been verified (Inglehart & Baker, Citation2000; Inglehart & Norris, Citation2003; Inglehart & Welzel, Citation2006).
17. However, the opposite causal effect, where a liberal attitude already manifested within the society is set into law after a certain delay, should not be excluded. But even in such a case, a correlation between the legal configuration and the attitudes of the population should be found.
18. There is, without a doubt, a problem of endogeneity with these particular variables: it is also plausible that attitudes toward homosexuality are not a result of legal provisions, but rather that legal punishments for homosexual practices arise from predominant homonegativity within a society. Even if causality will not be clarified here, this article presumes the formative power of laws on society.
19. The separate treatment of the legality of lesbian practices is here knowingly forgone, as the data is worse than that of gay men. The legality of gay contact also poses a harsher test than that of lesbians: when homosexual practices are forbidden between women, they are also forbidden between men. On the contrary, there are cases where homosexual practices between men are forbidden, but they are allowed between women (such as in Ghana or Bangladesh).
20. Practical examples include urban neighborhoods that are known for their high proportion of homosexual residents and that have, over time, developed to generally fashionable neighborhoods (e.g., “Le Village” in Montreal, “The Castro” in San Francisco, or the “Glockenbach” quarter in Munich).
21. Qualitative evidence nevertheless shows that attitudes are changing relatively fast toward a greater acceptance of homosexuality in China nowadays. Since 1997, when the hooligan law was abolished, homosexuality is no longer seen as a crime, and especially in cities such as Shanghai or Hong Kong, an open gay and lesbian scene could emerge largely unhindered by official authorities (Lau, Citation2010).
22. The increasing relevance of converts within terrorist groups such as al Qaeda fits this pattern of argumentation (Roy, Citation2010). Their worldview is often far removed from cultural roots, and it presents a one-sided exaggeration of single aspects of a religious teaching.
23. We used HLM 6.08.
24. Hox (Citation2002, p. 184) gave several rules of thumb when using the ICC (from 0.05 to 0.3), whereby he indicated that an ICC of 0.3 is the “most stringent” criterion (“in those cases where on a priori grounds much higher intraclass correlations appear reasonable”).
25. Despite their significance, the employment dummies will not be further used in the following models. This is due to the high number of iterations (85) that is necessary for the assessment of Model 2, indicating a comparatively poor convergence of the model.
26. This result can also be confirmed provided that the individual religions are added as dummy variables. This type of approach would be preferable to the religion index that was used, as it can more clearly convey the influence of single religions on homonegativity. In any case, the low variance of religious adherence in some countries reduces the sample size on the aggregate level or even circumvents a convergence of the model when more than one of these dummy variables are simultaneously applied to the model. A comparison of these models can become very difficult due to the different number of cases. However, this problem does not exist to the same extent in the analysis of the religion index. Alternatively, the dummy variables for the single religions can also be accommodated in separate models. When using this approach, the theoretically expected ranking for the religion index arises, with one exception. People identifying themselves as Catholics show a less homonegative stance than we would have expected beforehand. This demonstrates a comparatively large discrepancy between the clerical teachings and the actual attitude of the religious.
27. This finding is not altered even if no other variables are tested on the individual level in comparison to Model 1. In the interest of space, these two steps of the model construction were compiled in Model 2.
28. Due to the clearly low number of cases (caused by missing data for the law index in many countries), the results of this model can be compared with the other indices only to a limited extent. It is striking, however, that a number of previously significant factors remain significant even under the markedly reduced data set.
29. It should be mentioned, however, that the models did not converge as well as in the GLS/ML estimation. For this reason, the threshold for the termination of the maximum likelihood iteration process was increased to a value of 0.01. This way, a stable solution can be more quickly found and will be able to hold a higher defectiveness.
30. This interaction effect can also be interpreted as a level effect, as homonegativity in (post-) communist countries is in any way higher than in the comparison group, implying that it can only increase less dramatically.
31. One concrete case concerns the values for the religious affiliation of respondents, which had to be recoded for several countries as the categories “Protestant” and “Evangelical” were not stringently applied.
32. The fact that, despite the data problems, the cross-check with the alternative operationalization of homonegativity (wanting homosexuals as neighbors) confirms the results grosso modo (cf. ) can be seen as an indication for the robustness of the analysis.