Chapters 2 and 3 have laid out the extent and limits of our knowledge on how to reduce homophobia. Contact with LGB people and education about sexuality and prejudice have been shown to reduce homophobia on standardised measures, and participants in such interventions have generally reported a positive, enlightening experience. Most studies, however, have been conducted with participants who are less likely to hold homophobic beliefs in the first place. Thus, samples were predominantly young, female, and educated (see Chapter 2, ‘Study-space analysis’), while levels of homophobia tend to be higher in older, male, and less well educated individuals (see Chapter 1, ‘Homophobia is intertwined with other values’).
Moreover, almost all studies have been performed in North America or Western Europe, where LGB people are more likely to be socially accepted and protected by the law (see Chapter 1, ‘Homophobia varies across space and time’). Participants themselves often criticised the interventions for being inadequate for the contexts where they were performed, aiming at a decrease in homophobia that was perceived to be either too slow or too fast for the respective organisation or community (see Chapter 3). Finally, interventions to reduce homophobia usually occurred on a small scale, with a few dozen people participating and with the results not being monitored on the long term (see Chapter 2, ‘Study-space analysis’).
As explained in Chapter 1, this thesis focuses on homophobia in the UK and Romania, and on the contrasts between Eastern and Western Europe more generally.
Since most research on homophobia and on how to reduce it has been conducted in the US, it would be theoretically and practically useful to extrapolate the results to other regions – especially to Eastern Europe where homophobia is very widespread (see Chapter 1). However, given the different histories of (homo)sexuality in North America and in Western and Eastern Europe, it is far from self-evident whether such an extrapolation is valid.
In the present chapter, I undertake to examine the reduction of homophobia beyond the limits set by the literature reviewed in Chapters 2 and 3. I reanalyse data
from the World Values Survey (World Values Survey Association, 2015) and its sister project, the European Values Study (European Values Study, 2015; henceforth WVS/EVS), in order to study homophobia in a large, cross-cultural sample that spans several decades. Such a dataset allows me to examine the role of participant characteristics on which the corpus of Chapter 2 was too homogeneous (such as age and education) and those potential factors which have not been consistently assessed (such as authoritarianism and religiosity). Given the four-decade history of the WVS/EVS also allows an insight into long-term change not afforded by other methods. Most importantly, a reanalysis of the WVS/EVS data enables me to explore a key question opened up by the systematic reviews: can knowledge about homophobia gained in the West be extrapolated to other societies?
Homophobia and Its Correlates in the World Values Survey
The WVS/EVS systematically collects data on people’s attitudes on dozens of issues, regularly surveying about one thousand participants from the majority of countries in the world. The surveys have been performed in six waves between 1981 and 2014, in a total of 113 countries. The breadth of this research, and the public availability of the data, has allowed both the confirmation of known predictors and the emergence of new explanations for homophobia. Inglehart (1997), the initiator of the WVS/EVS, has argued that ‘[e]conomic, cultural, and political change go together in a coherent pattern that change the world in predictable ways.’ (p. 7).
Consequently, he has sought to identify both a small number of overarching value dimensions (via factor analyses of WVS/EVS data), and large-scale patterns of change.
Inglehart (1997) has not only observed, but also theorised change throughout the duration of the WVS/EVS. In Western societies, according to Inglehart, the impact of industrialisation has reached (and passed) its maximum. While many societies worldwide are undergoing modernisation, i.e., moving from traditional authority and religious values towards economic growth and rationality, the West is going through postmodernisation, i.e., shifting towards an emphasis on personal wellbeing and fairness (Inglehart, 1997). The reduction of homophobia in the West is considered a key aspect of postmodernisation, as it is both illustrative of post-industrialised
societies' focus on diversity and fairness, and one of the most spectacular changes documented by the 30-year history of the WVS.
Widely-recognised predictors19 of homophobia (such as demographic characteristics, authoritarianism, religiosity, racial prejudice, and economic and historical conditions) have been discussed in Chapter 1 (‘Homophobia is intertwined with other values’). I now proceed to briefly examine WVS/EVS research on these predictors. First, as far as demographics are concerned, women, as well as people who are younger, more educated, and wealthier tend to be less homophobic (for a review, see Herek & McLemore, 2013; see also Chapter 1, ‘Homophobia is intertwined with other values’). These associations have generally been supported by studies based on the WVS/EVS (see, e.g., Andersen & Fetner, 2008a, for North America; Hadler, 2012, for Europe). However, some regional variation has been reported in the relationship between demographics and homophobia: e.g., in Sub-Saharan Africa, women tend to be less tolerant than men, while education and age are not related to homophobia (Bangwayo-Skeete & Zikhali, 2011).
Second, WVS/EVS studies have typically confirmed the link between religiosity and homophobia (e.g., Adamczyk & Pitt, 2009; Andersen & Fetner, 2008) which is broadly supported by other lines of research (for a review, see Herek & McLemore, 2013). However, this link may be moderated by other culture-specific values: for example, religiosity may have a stronger association with sexually restrictive values in egalitarian societies (Vauclair & Fischer, 2011). In Europe, the relation between religiosity and homophobia is stronger in the West than in post-communist countries (Jäckle & Wenzelburger, 2011).
Third, authoritarianism (Adorno et al. 1950) has been shown to predict prejudice towards various groups, including LGB people (see Chapter 1, ‘Homophobia is intertwined with other values’). However, most standardised measures of authoritarianism (such as the F scale by Adorno et al., 1950) claimed to predict
19 For convenience, these variables will be collectively referred to as ‘the predictors’ throughout the paper, even when the term would not be normally used in the context of a particular analysis (e.g., for correlations).
prejudice towards minority groups while they also contained questions about those same groups, making the authoritarianism-prejudice relationship tautological.
Stenner (2005) proposed the use of the WVS/EVS items pertaining to childrearing values to measure authoritarianism. The respective questions assess the importance placed by participants on such authority-focused values as obedience, and such independence focused values as creativity. Stenner argued that childrearing values allowed for a measure of authoritarianism that was cross-culturally meaningful. Most importantly, such a scale would not overlap with the constructs that authoritarianism was expected to predict. This WVS/EVS-based measure of authoritarianism has been shown to correlate with homophobia, although the strength of this relationship varied across countries (Stenner, 2005).
Fourth, homophobia is related to other forms of prejudice (such as ethnic prejudice, Whitley & Lee, 2000), and to the political ideologies that promote them (see Chapter 1, ‘Homophobia is intertwined with other values’). WVS/EVS data tend to support these associations (e.g., Hadler, 2012), but some regional variation is present: whilst homophobia is typically associated with ethnic prejudice, it is related to tolerance towards other minority groups in Sub-Saharan Africa (Bangwayo-Skeete
& Zikhali, 2011).
Apart from the psychosocial predictors of homophobia discussed above, country-level factors also play an important role in shaping and changing attitudes towards LGB people. Cross-cultural datasets, such as the WVS/EVS, are indispensable to the study of such factors. Most notably, poorer countries have been consistently shown to have higher levels of homophobia (Andersen & Fetner, 2008; Hadler, 2012).
As discussed in Chapter 1 (‘Homophobia varies across time and space’), differences in countries’ histories can also be important: post-socialist countries in Eastern Europe have higher levels of homophobia than their Western counterparts (Štulhofer
& Rimac, 2009; Kuyper et al., 2013).
The Present Study
The central question of this chapter is whether models of homophobia elaborated in research-intensive Western societies can be transferred to countries
with higher levels of homophobia. Based on the research discussed above, the established predictors of homophobia are broadly confirmed by WVS/EVS analyses, but their cross-cultural relevance is variable. Therefore, it is likely that the extant (mostly Western) knowledge on homophobia offers a broad template which could be extrapolated to other cultures with necessary adjustments. It is the aim of this chapter to explore the extent of these adjustments, and thus the possibilities and limits of such generalizations.
The present chapter aims to explore the relationship between homophobia and its known predictors in large cross-national samples. I will examine these relationships both cross-sectionally and longitudinally, comparing explanatory models of homophobia across countries and across time. Therefore, I will conduct four analyses of WVS/EVS data (see Table 1). On the one hand, I assess the value of known predictors of homophobia in the present, relying on the most recent WVS/EVS data (Analyses 1 and 3), but I also compare these recent results with those from the early 1990s in order to study change (Analyses 2 and 4). On the other hand, I analyse individual-level data to gain insight into individual differences in homophobia (Analyses 1 and 2), but I also explore country-level data in order to understand regional differences in attitudes and the role of economic development (Analyses 3 and 4).
Table 1
Plan of the Chapter, by Time Frame and Level of Analysis
Level of Analysis Present Change Across Time
Individual Analysis 1 Analysis 2
Country Analysis 3 Analysis 4
First, I ask whether a similar theoretical model can explain homophobia in the countries where most research has been performed (i.e., the US and the UK) and those where the results of the research are most needed (e.g., Romania) (Analysis 1).
Second, I ask whether the same variables that explain individual differences in homophobia can also explain change across time (Analysis 2). Third, I explore
whether the variables that can explain individual differences in homophobia can also explain differences between European countries (Analysis 3). Finally, I examine whether changes in the extent of homophobia on a national level is related to changes in the predictors of homophobia (Analysis 4).
General Method
Data Sources
WVS/EVS data are freely available online (www.worldvaluessurvey.us/). For Analyses 1 and 2, individual-level data from Romania, the UK and the US were used.
In Analyses 3 and 4, data from 17 Eastern and 20 Western European countries were used, and country-level averages were computed for all variables of interest (see below). The Eastern countries were those that have had socialist regimes before the 1990s: Bosnia and Herzegovina (BH), Bulgaria (BG), Belarus (BY), Croatia (HR), the Czech Republic (CZ), Estonia (EE), Hungary (HU), Latvia (LV), Lithuania (LT), Moldova (MD), Poland (PL), Romania (RO), the Russian Federation (RU), Slovakia (SK), Slovenia (SI), Ukraine (UA) and Macedonia (MK). The Western countries were those that did not have socialist regimes before 1990: Austria (AU), Belgium (BE), Cyprus (CY), Denmark (DK), Finland (FI), France (FR), Germany (DE), Greece (GR), Iceland (IS), Ireland (IE), Italy (IT), Luxemburg (LU), Malta (MT), the Netherlands (NL), Norway (NO), Portugal (PT), Spain (ES), Sweden (SE), Switzerland (SW), and the UK. (Serbia, Montenegro and Kosovo were left out because border changes and disputes made comparisons difficult.)
In Analyses 1 and 3, I used the most recent WVS/EVS data for the countries of interest (Wave 6 of data collection). For Analyses 2 and 4, data from Wave 2 (1990-1994) and Wave 6 (2005 onward) were compared. The WVS/EVS does not provide individual-level longitudinal data, but a time-series of cross-sectional surveys. Data collection for Wave 2 occurred immediately after the fall of socialist regimes in Eastern Europe (1989-1993); data were collected by EVS in 1990 for the UK and the US, and in 1993 for Romania. Therefore, a comparison of these two waves allows an exploration of the changes that occurred in the first two post-socialist decades.
Apart from WVS/EVS data, I also used information on countries’ economic development in Analyses 3 and 4. In this analysis, gross domestic product per capita (adjusted for purchasing power parity; GDP) is used to measure each country’s economic development. Specifically, I retrieved the IMF-reported GDP (International Monetary Fund, 2015) of each of the 37 countries, and I averaged it for the years of Wave 2 (1990-1994) and Wave 6 (2010-2014).
WVS/EVS Measures
Participants’ gender (x001) 20, age (x003), highest educational level (x025) and income level (x047) were used as demographic variables. Gender was dichotomised as male (1) versus female (2). Education was quantified on a scale ranging from 1 (Inadequately completed elementary education) to 8 (University with degree). For assessing income, WVS/EVS researchers in each country determined the deciles; the resulting income brackets were coded from 1 (lowest step) to 10 (highest step). For the UK, only a similar twelve-step income measure (x047c) was available; this was recoded into ten steps in order to match data from Romania and the US.
Apart from demographic variables, four predictors of homophobia were used.
In the WVS, religiosity was assessed through a question (f034) allowing people to categorise themselves as either religious, non-religious or atheists, or to give a different answer. I dichotomised this variable into ‘religious’ (1) versus ‘non-religious’
(0).
Postmaterialism was conceived by Inglehart (1997) as a measure of the postmodernisation of individual values. It is assessed by asking participants to prioritise two out of four societal goals. Participants receive a high score (3) if they select the two postmaterialistic goals (democratic decision making and freedom of speech), a low score (1) if they select the two materialistic goals (public order and price control), and an intermediary score (2) if they select any combination of the two types of goals. The resulting variable (y002) approximated a normal distribution (skewness and kurtosis values in all three countries < 2).
20 Following the conventions of previous WVS-based reports, I use codes to identify variables in the dataset.
National pride was used as a proxy measure of ethnic prejudice. All other measures were deemed culture-specific. For example, attitudes towards immigrants would only be relevant in countries with a high level of immigration, such as the US (where 14.3% of the population is foreign born; United Nations, 2013) and the UK (12.4%). Romania, on the contrary, has low levels of immigration (0.9%), while the marginalisation of Gypsies is a major social issue (INSOMAR, 2009; Marcu et al., 2007;
Tileagă, 2005). Pride of one’s nationality was assessed on a scale (g006) ranging from 1 (very proud) to 4 (not at all proud). Due to the asymmetry of the distribution, I dichotomised this variable into ‘very proud’ (1) versus ‘not very proud’ (0).
Authoritarianism was computed based on the WVS items on childrearing values (a027 – a034; see Stenner, 2005). All of these variables were dichotomous;
participants were asked whether they value specific characteristics in a child. In this study, I summed the scores for obedience (a042), tolerance (reverse scored, a035) and independence (reverse scores, a029) to obtain a measure of authoritarianism that ranged from 0 to 3 and approximated a normal distribution (skewness and kurtosis values in all three countries < 2)21. As anticipated (Stenner, 2005), the measure of authoritarianism computed from childrearing values had low internal consistency, particularly in Eastern European samples (for Wave 5, Cronbach’s α = .231 in the US, .256 in the UK, .142 in Romania).
Two measures of homophobia were used. A dichotomous survey item (a124_09) indicated whether participants chose ‘homosexuals’ from a list of potentially undesirable neighbours (some other options being people who have AIDS, speak another language, or drink heavily). I will call this measure social distance.
Participants also assessed the morality of homosexuality (f118) on a scale ranging from ‘never justifiable’ (1) to ‘always justifiable’ (10). Some other issues raised in the same set of questions were divorce, suicide, and stealing. In many countries (including Romania and the US), (1) is the most frequent answer. Therefore, these variable have often been dichotomised (e.g., Inglehart, 1997) to contrast those for
21 Unlike Stenner (2005), I omitted good manners and imagination, which are conceptually less linked to authoritarianism (see also Singh & Dunn, 2013).
whom homosexuality was never justifiable (1) with all other respondents (0). I adopted the same approach in this chapter, and called this measure moral rejection.