Does intergenerational educational mobility vary by sexual identity? A comparative analysis of five OECD countries

Abstract

Lesbian, gay and bisexual people are disadvantaged in terms of health and socio-economic status compared with heterosexual people, yet findings pertaining to educational outcomes vary depending on the specific identity and gender considered. This study delves into these unexplained findings by applying a social-stratification lens, thereby providing an account of how intergenerational educational mobility varies by sexual identity. To accomplish this, we use representative data from five OECD countries and a regression-based empirical specification relying on coarsened exact matching. We find that gay and lesbian people have higher educational attainment than heterosexual people in all five countries and that these higher levels of education stem from greater rates of upward educational mobility among gay/lesbian people. There were, however, few differences between heterosexual and bisexual people. Variation across countries emerged when analyses were stratified by gender, with higher rates of upward mobility observed for gay men in Australia, Chile, the United Kingdom, and the United States and lesbian women in Australia and Germany. Overall, our results align with previous claims that education can be a strategy for gay/lesbian people to avoid actual or anticipated discrimination. However, variation in these patterns across groups suggests that other mechanisms may also be at play.

Lesbian, gay and bisexual (LGB) people experience worse health, well-being, and socio-economic outcomes than heterosexual individuals, including disparities in earnings and occupational standing (Duncan and Hatzenbuehler, 2014; Gates, 2015; Russell and Fish, 2016; Perales and Campbell, 2020).¹ Understanding the social determinants of these disparities, and the extent to which they represent a violation of principles of equality of opportunity, is therefore an important endeavour. Evidence regarding educational outcomes, which are central to the above processes, primarily comes from the United States and is surprisingly mixed (see summary table in online Supplementary Appendix A).

With some exceptions (Badgett, 1995), studies using behaviour-based measures of sexual orientation report higher educational attainment among men and women in same-sex unions or with a history of same-sex sexual behaviour (Black et al., 2000, 2003; Verbakel, 2013). However, findings from studies measuring sexual orientation through self-reports vary by gender, sexual orientation category, and educational outcome. Some report lower educational attainment among bisexual men and bisexual/lesbian women than among heterosexual individuals (Mollborn and Everett, 2015; Mittleman, 2022), whereas others report no differences for bisexual men (e.g. high-school graduation rates, Mittleman, 2022) or higher attainment among certain subgroups of bisexual men (e.g. those enrolled in college, Mollborn and Everett, 2015) and lesbian women (e.g. older and white cohorts, Mittleman, 2022). Quantitative studies consistently find that gay men are more highly educated than heterosexual men (Mittleman, 2022), despite qualitative studies showing that gay men struggle more in school than both heterosexual people and lesbian women (Pascoe, 2011). Most of these studies are conducted in the United States. As an exception, Verbakel’s (2013) study documented higher levels of education among men and women in same-sex relationships in the Netherlands, as well as a greater likelihood of having studied a care-related field of education. In sum, the relationship between sexual identity and educational attainment remains poorly understood, with existing findings also being largely parochial to the United States.

In this study, we aim to enhance our understanding of sexual-identity differences in educational outcomes by considering intergenerational educational mobility—that is, the extent to which adult children’s educational outcomes mirror their parents’ outcomes. This approach offers greater insight into processes of (in)equality of opportunity (Li and Heath, 2016). First, intergenerational mobility analyses account explicitly for social background. If group differences in outcomes are produced by differences in social background, outcome differences are only poor indicators of inequality of opportunity (Chetty et al., 2020). Second, intergenerational mobility analyses can identify whether differences in socio-economic outcomes are primarily produced by challenges in attaining education among low-status groups or the ability of high-status groups to maintain advantage (Bernardi, 2014). Besides redirecting attention towards questions of intergenerational mobility, we also expand the geographical scope of this research area by analysing five large-scale representative surveys from Australia, Chile, Germany, the United Kingdom, and the United States. This is an important step forward in the literature, as the experiences of LGB people differ considerably across country contexts. Methodologically, we innovate by deploying an empirical regression-based strategy relying on coarsened exact matching (CEM).

Conceptual framework

Theorizing sexual-identity differences in the intergenerational transmission of education

Educational advantage is transmitted from parents to children through a wide variety of channels, including different forms of economic, cultural, and social capital (Bourdieu, 1986; Coleman, 1988). If sexual identity shapes access to these resources, the connection between parental education and educational attainment could differ between LGB and heterosexual individuals.

On the one hand, LGB people may have lower access to parental resources than heterosexual people. Within heteronormative societies, LGB people risk experiencing parental rejection due to their sexuality (Russell and Fish, 2016), as reflected in lower levels of closeness to parents, more conflict with parents, and less support from parents (Hank and Salzburger, 2015; Perales and Huang, 2020; Fischer and Kalmijn, 2021). These experiences can reduce the intensity of parental efforts to transmit financial, cultural, and social capital to their offspring. Indeed, parental investments are not equally distributed across children, and those with whom parents have the closest relationships receive more and greater transfers (Lennartsson et al., 2010). Overall, this perspective suggests that the transmission of advantage will be weaker among LGB individuals than heterosexual individuals (H1a).

On the other hand, LGB youth are at an increased risk of experiencing school bullying and social exclusion, and they frequently report homophobic behaviours from both peers and teachers (Pascoe, 2007; Pearson and Wilkinson, 2018). Exposure to hostile school environments can negatively affect school performance by increasing absenteeism and disengagement (Robinson and Espelage, 2012). Therefore, parental support may be comparatively important for LGB individuals to overcome these challenges (Russell and Fish, 2016). If parental resources matter more for the educational outcomes of LGB than heterosexual individuals, then a competing hypothesis can be formulated: the transmission of advantage will be stronger among LGB than heterosexual people (H1b).

Distinguishing upward from downward mobility

From an inequality of opportunity perspective, distinguishing between downward educational mobility (i.e. being less highly educated than one’s parents) and upward educational mobility (i.e. being more highly educated than one’s parents) signals more directly unequal educational opportunities.

Concerning downward educational mobility, studies on the transmission of socio-economic advantage have shown that well-resourced parents are able to prevent the downward mobility of their children. When their children face barriers to educational success, well-off parents increase their investments to preclude these factors from affecting their educational attainment—a process referred to as ‘compensatory advantage’ (Bernardi, 2014). If LGB people from advantaged backgrounds have less access to their parents’ resources than heterosexual people, this will put them at an increased risk of downward mobility when compared with heterosexual individuals with the same parental background (H2).

Minority stress—or the negative treatment of individuals due to their LGB status (Duncan and Hatzenbuehler, 2014)—can also contribute to downward educational mobility among LGB people. Minority stressors such as discrimination and micro-aggressions may put pressure on their academic performance. Consistent with this, a voluminous body of research has documented how LGB youth experience a range of unique stressors within the school context (Robinson and Espelage, 2012; Pearson and Wilkinson, 2018), in addition to those experienced within families, neighbourhoods, and other social environments.

The mechanisms discussed so far suggest higher rates of downward educational mobility for LGB individuals (H2), but they would also predict lower rates of upward educational mobility (H3a). However, there are also factors that may contribute to higher rates of upward educational mobility among LGB individuals. The sociology of education literature distinguishes between primary and secondary effects on educational attainment. Primary effects are those that affect school performance, whereas secondary effects are those that affect educational choices, such as deciding to continue education or enrol in university (Boudon, 1974). For example, students from socio-economically disadvantaged backgrounds are less likely than students from more advantaged backgrounds to enrol in university, even when they attain the same grades (Jackson et al., 2007). Sexual identity could have secondary effects on educational attainment, chiefly by affecting educational choices and incentives to attain university degrees.

Stigma towards sexual minorities is more palpable and intense in ‘low-education’ and ‘low-income’ environments. For example, individuals with lower education and income hold less supportive views towards LGBTQ+ people (Perales and Campbell, 2018), and sexual minorities experience greater discrimination in occupations requiring lower-level educational credentials (Ueno et al., 2013; Mollborn and Everett, 2015). University degrees can thus be a pathway for LGB people to escape or avoid social environments where sexual stigma is comparatively high, and this may increase their perceived benefits of—and incentives for—higher education. A university degree may also provide LGB individuals with greater financial and social resources to avoid or buffer minority stressors and to overcome economic, social, or legal obstacles towards achieving normative life goals (e.g. marriage or parenthood) (Barrett et al., 2002). Furthermore, research shows that many young gay/bisexual men—especially those who conceal their identity—derive their self-worth from academic success and competition (Pachankis and Hatzenbuehler, 2013). In short, when considering these incentives, we can formulate a competing hypothesis to H3a: LGB individuals will exhibit higher rates of upward mobility than heterosexual individuals (H3b).

Theorizing differences by sexual identity

So far, we have discussed LGB people as a monolithic group. However, there are plausible reasons to expect differences in intergenerational education mobility across different minority identities. Bisexual people face pronounced minority stressors—including biphobia, misrecognition or erasure of their identities, and stereotypes about their sexuality (e.g. around promiscuity or it being ‘a phase’)—which makes them more prone to exhibiting poor health and well-being outcomes (Mize, 2016; Pollitt et al., 2017). These additional stressors could lead to higher rates of downward educational mobility and lower rates of upward educational mobility among bisexual compared to gay/lesbian individuals (H4). In addition, bisexual people are less often ‘out’ to their families, friends, and co-workers than gay/lesbian people (Pew Research Centre, 2019; Doan and Mize, 2020). This could weaken the relationship between having a bisexual identity and direct experiences of general discrimination or parental rejection—although it could also put pressure on educational performance through other pathways (e.g. increased stress due to identity concealment).

Theorizing gender differences

Given previous research revealing substantial disparities in the educational experiences of lesbian women and gay men (Pascoe, 2011), there may also be gender differences in the relationship between sexual identity and intergenerational educational mobility. Mittleman (2022) showed that gay men in the United States are more highly educated than heterosexual men, whereas lesbian and bisexual women are less highly educated than heterosexual women (with the exception of older white cohorts of lesbian women). One explanation for this pattern is the persistence of cultural expectations to ‘do gender’ normatively within schools. This includes the perception of academic effort being incompatible with socially constructed understandings of masculinity and the perception of diligence in school settings as being a feminine trait, a finding that is also central to studies on gender gaps in educational attainment (Legewie and DiPrete, 2012). Lower incentives for gay/bisexual men to perform ‘manhood’ may thus help them avoid the ‘academic cost of masculinity’ (Mittleman, 2022). In contrast, lesbian/bisexual women may on average underinvest in education to avoid being considered feminine, which is consistent with their comparatively high propensity to be subjected to school disciplinary measures (Mittleman, 2018). If these propositions hold, we would anticipate higher upward mobility and lower downward mobility among LGB men when compared with LGB women (H5).

Incorporating a cross-national angle

In an effort to move the existing academic debates in the field beyond their current US-centric focus, we examine sexual-identity-based educational disparities and processes of intergenerational educational mobility across five OECD countries—Australia, Chile, Germany, the United Kingdom, and the United States. These are, to our knowledge, the only five countries for which sufficiently large datasets with information on sexual identity, respondents’ education, and parental education exist (OECD, 2019). Importantly, there is meaningful variation in terms of discrimination and stigma against sexual minorities across these countries. For example, in the 2005–2009 wave of the World Values Survey, the percentage of respondents who did not want to have ‘homosexuals’ (sic) as neighbours was ~34 per cent in Chile, ~25 per cent in the United States, ~21 per cent in Australia, ~17 per cent in the United Kingdom, and ~15 per cent in Germany (Adamczyk and Liao, 2019). Importantly, the percentage of unsupportive individuals differs by respondents’ education. Specifically, the difference in the share of lowly and highly educated individuals who do not want ‘homosexuals’ (sic) as neighbours lead to a similar country ranking: ~12 percentage points in Germany, ~20 in Chile, ~25 in the United Kingdom, ~27 in Australia, and ~41 in the United States.² These differences support the argument that education may be a more effective resource for LGB people to escape discrimination in certain countries (Pachankis and Hatzenbuehler, 2013; Mittleman, 2022). Regarding the legislation operating during our observation period (2008–2019), Australia (2007–2013), Germany (2006), and the United Kingdom (2007–2010) had all included sexual orientation in anti-discrimination laws, whereas this only applied to certain states in the United States (although workplace discrimination based on sexual orientation or gender identity was ruled illegal by the US Supreme Court, in Bostock vs. Clayton County, in 2020). In Chile, same-sex sexual behaviour was illegal until 1999, but rapid changes have taken place, including new anti-discrimination laws in 2012.

Many of the mechanisms discussed in the theoretical framework relate to sexual stigma, and LGB people’s incentives to attain education to escape discrimination or accumulate resources to cope with it. From this prism, two competing expectations can be formulated regarding cross-national differences in the relationship between sexual identity and educational attainment. On the one hand, because of reduced barriers to education, one could expect LGB people’s average educational attainment to improve as discrimination and stigma decline. This situation would then result in higher rates of upward educational mobility and lower rates of downward educational mobility in contexts with less discrimination and stigma. On the other hand, if reductions in discrimination and stigma make sexual identity less relevant for educational outcomes, rates of upward and downward educational mobility may be more similar across sexual identities in contexts with less discrimination and stigma. Because several studies have reported that educational attainment is higher among LGB individuals than heterosexual individuals, this may imply relatively lower levels of educational attainment among LGB individuals in contexts characterized by low stigma and discrimination.

While stigma and discrimination are salient factors, they may not be the only country characteristics influencing the relationship between sexual identity and educational mobility. For example, factors such as age at educational tracking, standardization of curricula and examinations, and financial returns to education may also play a role (Bol and Van de Werfhorst, 2013). Furthermore, it is possible that differences in measurement instruments across countries may contribute to any observed disparities. We return to these considerations when we discuss our findings.

Data and methods

Datasets and sample selection

Our analyses are based on data from five nationally representative surveys: the Household, Income and Labour Dynamics in Australia (HILDA) Survey; the Chilean National Survey of Socioeconomic Characteristics (CASEN); the German Socioeconomic Panel (SOEP); the UK Household Longitudinal Study (UK-HLS); and the US General Social Survey (GSS). Detailed descriptions of these data sources have been published elsewhere (Buck and McFall, 2011; Watson and Wooden, 2012; Ministerio de Desarollo Social y Familia, 2017; Marsden et al., 2020; Fischer, 2021). Three datasets are annual household panel surveys in which information on respondents’ sexual identity was collected in multiple waves: 2012, 2016, and 2020 in HILDA, 2016 and 2019 in SOEP, and 2011/2012 and 2017/2018 in UK-HLS. For these datasets, we maximize information by pooling observations across years and clustering standard errors by individuals. Robustness checks where we use the last available wave with complete information for each individual yield similar substantive results (online Supplementary Appendix B).³ The GSS and CASEN are cross-sectional surveys. The GSS is conducted every 2 years, and we pool observations across iterations in which sexual-identity information was collected (2008, 2010, 2012, 2014, 2016, and 2018). The CASEN data were collected in 2017.

To arrive at the analytical sample, we first identified all individuals interviewed in waves collecting sexual-identity data. Second, we selected respondents aged 25–85 years to focus on a population that has mostly completed their educational career. This excluded 11 per cent (United States) to 36 per cent (Chile) of the observations. Third, we dropped cases with missing information on self or parental education, sex/gender, ethnicity/migration background, age, or the educational level of both parents. This excluded a further 3 per cent (Australia/United States) to 15 per cent (United Kingdom) of observations.⁴ In Chile, parental education was only asked of the head of household, leading us to drop a further 51 per cent of the sample (people who were not the head of household). The final analytic sample sizes were 25,000 in HILDA; 48,604 in CASEN; 41,167 in SOEP; 53,564 in the United Kingdom; and 9,462 in the United States. In additional analyses, we replicated our main results using imputation techniques to account for missing data on parental and own education. Since the pattern of results remained unchanged, we only present these in online Supplementary Appendix C and retain the more parsimonious analyses here.

Measuring sexual identity

As others before us (see e.g. Mize, 2016; Mittleman, 2022), we identify LGB people based on information on self-reported sexual identity. In HILDA and UK-HLS, this information is collected through the same question: Which of the following options best describes how you think of yourself?, with the following response options: Heterosexual or Straight, Gay or Lesbian, Bisexual, Other, Prefer Not to Say, and Don’t Know (UK-HLS) or Unsure/Don’t Know (HILDA). In Casen and the GSS, the question is similar, but the Other response option is not included in the GSS. In Germany, the response options translate to Heterosexual, Homosexual, Bisexual, None of the above, and No answer/Prefer not to say. In the 2019 SOEP wave, the None of the above option was no longer available.

Self-reports of sexual identity sometimes can vary over time for the same individuals participating in multiple waves of HILDA, SOEP, or UK-HLS (see online Supplementary Appendix D). Because of this, our analyses pool observations across waves—that is, we treat each observation from an individual as a different ‘case’. For example, if an individual participated in two survey waves and was therefore asked the same questions, we would include each wave as a separate observation. To prevent this course of action from resulting in an artificial reduction in the standard errors, we followed the convention in studies using these methods and clustered the standard errors in the regression models by individuals.

Harmonizing these data, we arrived at the following sexual-identity categories: (i) heterosexual, (ii) gay/lesbian, (iii) bisexual, (iv) other, and (v) unsure/refused/prefer not to say. Because the Other group is complex to interpret, our main analyses focus on the first three groups of individuals. Online Supplementary Appendix D contains a detailed discussion of measurement issues related to sexual identity, whereas online Supplementary Appendix E presents the results for respondents who answered Other or refused to answer. The unweighted number (weighted percentage) of observations in which respondents in our sample identified as LGB was 1,137 (2.9 per cent) in HILDA, 654 (1.5 per cent) in CASEN, 904 (1.8 per cent) in SOEP, 832 (2.5 per cent) in UK-HLS, and 389 (3.4 per cent) in the GSS (Table 1).

Table 1.

Descriptive statistics

	Australia	Chile	Germany	United Kingdom	United States
	%	%	%	%	%
Sexual identity
Gay/lesbian	1.5	1.3	1.1	1.5	1.6
Bisexual	1.4	0.2	0.7	1.0	1.8
Other	0.9	0.1	3.6	0.7
Refused/prefer not to say	5.3	0.2	6.3	2.9	1.8
Gender
Woman	51.8	43.8	51.2	53.6	54.6
Own education
ISCED 1–2 (lower secondary or less)	20.5	23.1	10.5	26.0	6.0
ISCED 3–4 (higher secondary)	35.2	47.5	56.7	34.5	59.1
ISCED 5–6 (tertiary education)	44.3	29.4	32.8	39.5	34.9
Parental education
ISCED 1–2 (lower secondary or less)	31.5	11.6	14.6	13.0	18.4
ISCED 3–4 (higher secondary)	41.5	26.8	68.3	53.4	57.3
ISCED 5–6 (tertiary education)	27.1	11.6	17.1	13.0	24.2
Age group
25-34 years	21.9	14.5	17.2	16.6	22.1
35-44 years	21.0	17.4	17.0	19.8	21.7
45-54 years	20.2	20.7	21.7	21.5	21.0
55-64 years	17.6	21.1	19.1	19.2	19.2
65-74 years	12.9	17.0	14.1	15.3	11.1
75-84 years	6.4	9.2	10.9	7.7	4.9
Migration background
Native born		94.8	84.9	89.1	86.1
Foreign born		5.2	15.1	10.9	13.9
Ethno-migrant background
Native born, Non-Indigenous	66.1
Native born, Indigenous	1.8
Foreign born, other country	20.3
Foreign born, major English-speaking country	11.8
Race
Black					13.5
Other					10.2
White					76.3
n (observations)	35,840	48,347	41,702	37,845	9,508

Notes: Weighted descriptive statistics. Data from 2012, 2016, and 2020 for HILDA (Australia), 2017 for CASEN (Chile), 2016 and 2019 for SOEP (Germany), 2011/2012 and 2017/2018 for UK-HLS (UK), and 2008–2018 for the GSS (United States).

Measuring education and educational mobility

The dependent variables of interest are measures of intergenerational educational mobility, defined as the difference between respondents’ educational attainment at the time of the survey and their parents’ highest level of education. In the main analyses, the highest level of education is divided into three categories: (i) lower secondary education or less (ISCED 1–2), (ii) upper secondary education or vocational qualifications (ISCED 3–4), and (iii) university or other tertiary degrees of at least 2 years of duration (ISCED 5–6). More nuanced classification schemes were not available for all countries (DiPrete et al., 2017). However, given the importance of vocational qualifications in Germany, we also ran robustness checks using a five-category education variable for that country (using the CASMIN classification; see online Supplementary Appendix F). For parental education, we followed earlier studies (see e.g. Bernardi, 2014; Chetty et al., 2020) and used the same categories and considered the highest educational qualification obtained by either of the parents. When information for one parent was missing, we coded this variable based on the information for the other parent only.

Individuals are treated as being intergenerationally mobile if their education differs from the highest level of education of either of their parents, downwardly mobile if their education is lower than the highest level of education of either of their parents, and upwardly mobile if their education is higher than the highest level of education of either of their parents. Importantly, the outcome variables capturing upward and downward intergenerational educational mobility are only derived for individuals at risk of such mobility. Specifically, individuals with tertiary-educated parents are not part of the upward-mobility analyses and individuals whose parents both have lower secondary education (or less) are not part of the downward-mobility analyses. Note that because this classification depends on the highest level of parental education, individuals can have the same level of education as one of their parents (i.e. the lower educated parent) and still be coded as educationally mobile. In online Supplementary Appendix F, we present the results of more complex specifications using an education measure that simultaneously considers both parents’ education. Table 1 describes the samples used in the analysis (for descriptive statistics stratified by sexual identity, see Supplementary Table G1 in online Supplementary Appendix).

Analytic approach

Our analytical approach encompasses two distinct sets of analyses. First, we represent patterns of intergenerational educational mobility using flowcharts, as proposed by Laurison et al. (2020). This approach enables us to visualize the distributions of parental and own education, as well as the mobility rates between these.

Second, we fit country-specific sets of regression models to estimate sexual-identity differences in intergenerational educational mobility. In this analytic component, we provide an overall estimate of sexual-identity differences in such mobility using CEM (Iacus, King and Porro, 2012). The advantage of this method over traditional regression models is that it more effectively controls for observed compositional differences between groups. This includes any possible interactions between the control variables and the focal explanatory variable capturing respondents’ sexual identity. For example, the LGB samples are much younger than the heterosexual samples; if there are any cohort changes in the effect of parental education (or any other variable) on educational outcomes, the inclusion of a simple control for age would be insufficient to account for compositional differences across sexual-identity groups.

The CEM procedure relies on selecting samples from two groups that have the exact same observed characteristics. In our case, this method is powerful because we are able to draw, for virtually every LGB individual, at least one exact match from the comparatively large pool of heterosexual individuals. We match on a small set of observable characteristics that can be considered exogenous to sexual identity, differ considerably by sexual identity, and/or are key to our analysis. These characteristics include respondents’ age (in 5-year intervals), gender (for analyses pooling men and women), parental education, and country-specific measures of ethnicity (see Table 1). If more than one heterosexual match is available for an LGB individual, all cases are included and weights are created to adjust for their number. Matches are created separately by country, gender (for analyses pooling men and women), and sexual-identity group.

Applying these procedures, 4,026 of 4,029 LGB individuals had at least one exact heterosexual match in the data. Using these matched samples, we fitted linear probability models of the different mobility outcomes. In robustness checks, we also used logistic regression (see online Supplementary Appendix C) and alternative estimation methods, including ‘unidiff’ and log-linear models (Erikson and Goldthorpe, 1992; Jann and Seiler, 2020) (see online Supplementary Appendix H). Reassuringly, all of these supplementary analyses yielded similar results to those presented in the main body of the paper, adding confidence to our findings.

Visualizing intergenerational educational mobility across sexual-identity groups

Figures 1a–e portray intergenerational mobility flows for the five countries considered in this study. The charts show both the distribution of parental educational attainment (left side) and the distribution of own educational attainment (right side) for each group considered. Accompanying tables containing the numerical values for the transition rates are presented in online Supplementary Appendix F.

Figure 1.

Share of respondents attaining low, middle, and high levels of education by parental education in Australia (a), Chile (b), Germany (c), United Kingdom (d), United States (d)

Notes: Flowcharts based on weighted descriptive statistics (Table 1). Data from 2012, 2016, and 2020 for HILDA (Australia), 2017 for CASEN (Chile), 2016 and 2019 for SOEP (Germany), 2011/2012 and 2017/2018 for UK-HLS (United Kingdom), and 2008–2018 for the GSS (United States).

The flowcharts clearly demonstrate that gay/lesbian individuals have higher levels of education than heterosexual individuals in all five countries. This difference ranges from 6 percentage points in Germany to 17 percentage points in the United Kingdom. For bisexual people, however, the pattern is less clear: they fall in-between heterosexual and gay/lesbian people in the United Kingdom and the United States; they are the highest educated group in Chile; and they have education rates similar to heterosexual people in Australia and Germany.

The charts also reveal differences in parental education by sexual identity, particularly when examining individuals with lower levels of parental education. Compared to differences in own education, differences in parental education between heterosexual and gay/lesbian individuals are relatively small. As an exception, 24 per cent of gay/lesbian individuals in Chile have a tertiary-educated parent, compared with 11 per cent of heterosexual individuals. In the case of bisexual individuals, we observe considerably higher levels of parental education in Chile, the United Kingdom, and the United States. These results suggest that parental education differences may explain a greater portion of the educational advantages observed for bisexual individuals in some countries, whereas differences in mobility rates may be primarily responsible for the higher education levels observed among gay/lesbian individuals.

Inspecting the flows from left to right offers an indication of the level of intergenerational educational mobility. The clearest difference across sexual-identity groups relates to upward mobility, which was greater for gay/lesbian individuals than heterosexual individuals across countries. Higher levels of upward mobility are particularly visible for gay/lesbian persons whose parents have upper secondary (ISCED 3–4) education. To illustrate this point, in Australia, 59 per cent of gay/lesbian individuals whose parents have upper secondary education (ISCED 3-4) attained tertiary education, compared with just 40 per cent of heterosexual individuals with the same level of parental education. The smallest difference was observed in Germany, where these percentages were 39 per cent and 30 per cent, respectively. In Chile, it was bisexual people who displayed the highest rates of upward mobility. For example, among individuals whose parents had ISCED 3–4 education, 67 per cent of bisexual people compared with 46 per cent of gay/lesbian people and 35 per cent of heterosexual people were upwardly mobile.

For downward mobility, few differences can be observed by sexual identity. The exception here is Australia, where gay/lesbian individuals were less likely to be downwardly mobile than heterosexual individuals, with bisexual individuals being the most likely. For example, downward-mobility rates for individuals with tertiary-educated parents were 25 per cent for gay/lesbian people, 32 per cent for heterosexual people, and 39 per cent for bisexual people. In the next section, we provide formal statistical tests of differences in mobility rates by sexual-identity groups while accounting for compositional differences.

Modelling intergenerational educational mobility

Mobility among gay and lesbian people

The key results of our regression models based on CEM are summarized graphically in Figures 2 and 3. Figure 2 compares rates of tertiary-education attainment and different measures of intergenerational educational mobility (overall, downward, and upward) for heterosexual and gay/lesbian individuals matched on their age, parental education, foreign-born status, and ethnicity.⁵ As in the descriptive charts discussed before, we observe higher levels of tertiary-education attainment among gay/lesbian individuals than heterosexual individuals across all countries. The largest difference is observed in Australia, where the share of gay/lesbian individuals with tertiary education is 16 percentage points higher than for heterosexual individuals. Stratifying the analyses by gender adds nuance to these results. While both lesbian women and gay men tend to be overrepresented among tertiary-education graduates, the differences are small and insignificant for women in Chile and the United States and for men in Germany.

Figure 2.

Coefficients of linear probability models based on matched samples explaining mobility among gay/lesbian individuals (reference category: ‘heterosexual’)

Notes: Coefficients from linear probability models (95 per cent confidence intervals) based on CEM. Matching variables: age, ethno-migrant or racial background, parental education, and gender (in the pooled models). Each coefficient comes from a separate model for a specific country and mobility type; for the estimates for men and women, separate models by gender are run.

Figure 3.

Coefficients of linear probability models based on matched samples explaining mobility among bisexual individuals (reference category: ‘heterosexual’)

Notes: Coefficients from linear probability models (95 per cent confidence intervals) based on CEM. Matching variables: age, ethno-migrant or racial background, parental education, and gender (in the pooled models). Each coefficient comes from a separate model for a specific country and mobility type; for the estimates for men and women, separate models by gender are run.

The second panel of Figure 2 presents differences in the share of individuals who are educationally mobile (overall). We observe generally higher rates of educational mobility among gay/lesbian individuals compared with heterosexual individuals. Differences are, however, less striking than for educational attainment, especially in Australia.

In the third and fourth panels, intergenerational educational mobility is separated into downward and upward mobility, and the analytic samples are restricted to individuals at risk of experiencing those types of mobility (i.e. those with a tertiary-educated parent are excluded from upward-mobility analyses and those with lowly educated parents are excluded from downward-mobility analyses). Overall, there are no notable or statistically significant differences in downward mobility rates between gay/lesbian and heterosexual people, with the exception of Australia (where downward mobility is lower among gay/lesbian individuals).

However, and importantly, we do observe significantly higher rates of upward mobility among gay/lesbian individuals in all five countries. Furthermore, the magnitude of the estimated effects is sizeable, ranging from a 7-percentage-point difference in Germany to a 19-percentage-point difference in Australia. To illustrate this point, 50 per cent of heterosexual individuals with lower or middle parental education in Australia are upwardly mobile, compared with 70 per cent of gay/lesbian individuals. Altogether, these results suggest that it is primarily the attainment of LGB individuals of lowly educated backgrounds that is responsible for differences in educational attainment and educational mobility between gay/lesbian and heterosexual people. Additional analyses (shown in online Supplementary Appendix H) confirm that differences in educational attainment between gay/lesbian and heterosexual individuals are smaller and, in many cases, inexistent among those with higher-educated parents.

Estimating the models separately for men and women reveals some interesting differences. For example, the results reveal small and insignificant differences in upward intergenerational educational mobility among lesbian women in Chile, the United Kingdom, and the United States. In Germany, however, it is gay men who do not appear to attain greater mobility than their heterosexual counterpars.⁶ In all other contexts and gender-based groups, gay/lesbian people outperform all other sexual-identity groups in terms of their rates of upward intergenerational educational mobility.

Mobility among bisexual individuals

Figure 3 displays the results of models comparing matched samples of bisexual and heterosexual individuals. The results show that the educational advantages among bisexual individuals in the United Kingdom and the United States observed in the descriptive flowcharts (Figure 1a–e) disappear when respondents are matched on their observable characteristics. This pattern of results indicates that compositional differences are responsible for the higher levels of education observed among bisexual individuals in those countries. In Chile, however, bisexual people remain more highly educated than their heterosexual matches. In Australia, significantly lower levels of attainment for bisexual individuals are still observed when using matched samples.

The estimates for intergenerational educational mobility show that all coefficients for overall mobility are close to zero and not statistically significant. A few statistically significant differences emerge once distinguishing between downward and upward mobility. In Australia, we observe a particularly disadvantageous pattern for bisexual people. Compared with heterosexual individuals, the share of bisexual individuals that is downwardly mobile is 10 percentage points higher than among heterosexual individuals. At the same time, the share of bisexual individuals that is upwardly mobile is 7 percentage points lower. Some differences also emerge when distinguishing between men and women. In Chile, for instance, the share of bisexual men who are downwardly mobile is 17 percentage points lower than for heterosexual men, although small cell sizes cast some doubts over this estimate (See footnote 5).

Additional analyses

To ascertain the validity and reliability of our findings, we conducted several sets of sensitivity analyses and robustness checks. These included (i) using more traditional methods to estimate mobility differences (‘unidiff’ and log-linear models), (ii) considering both parents’ educational attainment, and (iii) estimating our main results using alternative model specifications. In this section, we briefly summarize the results of these additional analyses, with the results being presented in the online Supplementary Appendix.

First, we replicated our analysis of overall intergenerational educational mobility using ‘unidiff’ models. These models estimate the extent to which the association between parental and own characteristics differs across groups or contexts (Erikson and Goldthorpe, 1992). We also used log-linear models to examine whether upward and downward intergenerational educational mobility are more commonly observed among gay/lesbian and bisexual individuals than heterosexual individuals. These models rely on Poisson regressions to predict the frequency of combinations of parental and own education, including full ‘sexual identity × parental education’ and ‘sexual identity × mobility category’ interactions. The results, detailed in online Supplementary Appendix H, are highly consistent with those presented in the main text, albeit more often statistically significant (quite likely, because these models do not match individuals on observable characteristics). The only substantive difference was observed for gay men in the United States, for whom log-linear models yielded a significantly greater likelihood of being downwardly mobile, compared with heterosexual individuals.

Second, we expanded our descriptive analyses by simultaneously considering the education levels of both parents (see online Supplementary Appendix F). When doing so, we observed little difference in educational outcomes by sexual identity for individuals with one or two highly educated parents. This pattern is consistent with our results for downward mobility reported earlier. In addition, these additional analyses indicated that gay/lesbian individuals are more likely to have tertiary education than heterosexual individuals across all other groups, which aligns with our findings on upward mobility. However, there was one exception to this general trend, namely the absence of clear sexual-identity differences in educational outcomes among individuals who have two lowly educated (ISCED 1–2) parents, except in the United Kingdom.

Third, we replicated our analysis for Germany using the CASMIN (instead of ISCED) classification to capture both own and parental education. This classification enables a more nuanced partitioning of vocational education, which is an important education pathway within the German context. These results were substantively similar to those reported in the main text above, although the disadvantageous patterns of intergenerational educational mobility previously reported for bisexual men (i.e. higher downward mobility and lower upward mobility) became statistically significant (see online Supplementary Appendix F).

Finally, alternative specifications of our matching analysis using logistic regression and multiple imputation rendered practically identical results to those reported in Figures 2 and 3, as can be appreciated from the results presented in online Supplementary Appendix C.

Discussion

In this study, we have offered first-time insights into how sexual identity relates to intergenerational educational mobility in five OECD countries: Australia, Chile, Germany, the United Kingdom, and the United States. To accomplish this, we leveraged nationally representative datasets and a CEM estimation approach. A key finding from our analyses was that individuals identifying as gay/lesbian are more highly educated than heterosexual people in all five countries and that these higher levels of education stem from greater rates of upward educational mobility among gay/lesbian people with lower- or middle-educated parents. These disparities were not only statistically significant, but often sizeable in magnitude.

Although our analyses are not designed to identify the specific mechanisms underpinning these processes, higher upward mobility among gay/lesbian people is consistent with theoretical perspectives emphasizing higher incentives to, and benefits from, educational effort and attainment for this group compared with heterosexual people (H3a). Previous research has suggested that pro-school attitudes among LGB youth can be a deliberate strategy to prioritize achievements that are less dependent on specific individuals who may disapprove of their identity, including peers and family members (Pachankis and Hatzenbuehler, 2013; Mittleman, 2022). A similar, broader argument can be found in early social-stratification research, positing that social mobility is driven by the experience of psychological distress and the desire to escape ‘toxic’ childhood environments (Houle and Martin, 2011). If the challenges stemming from holding non-heterosexual identities are particularly pronounced in lower socio-economic environments, this would incentivize LGB youth from these backgrounds to invest in academic success.

Comparisons between bisexual and heterosexual people, however, revealed similar rates of overall, upward, and downward intergenerational educational mobility between these two groups. At first sight, this finding is inconsistent with the incentive-based proposition described above. The similarities in the intergenerational educational pathways of bisexual and heterosexual people may stem from the observation that many bisexual individuals eventually settle for heteronormative lifestyles. For example, bisexual people may feel attraction for, and have a history of sexual relationships with, both men and women; yet, most are in long-term, committed relationships with different-sex partners (e.g. 84 per cent in a study by the Pew Research Center, 2019, see also Gates, 2015). Furthermore, compared with gay/lesbian people, many bisexual people are not out to colleagues, family, or friends, and, as a result, they may be less likely to be the direct target of discriminatory behaviours (Pew Research Centre, 2019; Doan and Mize, 2020). Hence, normative pressures could lead bisexual people from lower socio-economic backgrounds to settle for heteronormative lifestyles instead of pursuing upward mobility to escape stigmatizing social environments. We nevertheless found some support for theoretical predictions based on heightened levels of discrimination and stigma towards bisexuality (H4). In Australia, bisexual individuals were more likely to be downwardly mobile and less likely to be upwardly mobile when compared with heterosexual individuals.

Gender differences

Several differences and nuances emerged when stratifying the analyses by gender. In Chile, the United Kingdom, and the United States, differences in upward-mobility rates between lesbian and heterosexual women were relatively small and statistically insignificant (which supports H5). In Germany, however, higher rates of upward mobility were observed for lesbian women compared with heterosexual women, but not for gay men compared with heterosexual men (contradicting H5). Similar results applied to differences in educational attainment.

Our results for the United States are similar to those of Mittleman (2022), who documented higher educational attainment among gay men, but lower educational attainment for (younger cohorts of) lesbian women, when compared with their heterosexual counterparts. Mittleman suggested that gender norms about masculinity may explain gay men’s educational premium. Dominant conceptions of masculinity position academic effort as ‘feminine’, and by challenging hegemonic masculinity, young gay men may be able to escape the reach of these norms and bolster their educational success (Mittleman, 2022). Yet the reverse may hold true for young lesbian women, for whom challenging gender-typical norms could result in lesser educational efforts (Mittleman, 2018). This proposition resonates with research findings on the reversal of the gender gap in educational attainment (Legewie and DiPrete, 2012; DiPrete and Buchmann, 2013)—especially in combination with our finding that educational mobility differences are limited to gay men from lower educational backgrounds. The contemporary gender gap in educational attainment favouring women is higher in lower socio-economic environments, where boys are particularly susceptible to adopting masculine identities involving anti-school attitudes (Legewie and DiPrete, 2012). If anti-school attitudes are more central to the masculine identities of heterosexual young men from lower socio-economic strata, challenging hegemonic norms about masculinity can lead to more pronounced differences in academic achievement between heterosexual and gay men from these backgrounds.

Results for Germany, however, cast some doubts over the universality of these claims. In Germany, higher levels of educational attainment and rates of upward mobility were restricted to lesbian women and not observed for gay men. This finding suggests that country context may moderate the relationship between gender, sexual identity, and educational outcomes and that results for the United States—where most research has been conducted—cannot be readily extrapolated to other countries. A notable feature of the German context that may explain the difference in findings is that the gender gap in educational attainment still favoured men within the sample studied. In fact, heterosexual women in Germany lagged behind other groups, with 35 per cent of them having completed higher education, compared with 45 per cent of heterosexual men, 49 per cent of lesbian women, and 46 per cent of gay men. As a point of comparison, in the United Kingdom (similar to Australia and the United States), it was heterosexual men who had the lowest rates of higher-education attainment, at 41 per cent (compared with 50 per cent, 59 per cent, and 55 per cent for heterosexual women, lesbian women, and gay men, respectively). In other words, gay men and lesbian women are generally highly educated, but whether they are significantly more educated than heterosexual men and women, respectively, depends on the educational attainment of the latter two groups, which seems to vary more across contexts. This explanation is consistent with the notion that incentives to pursue higher-education options are high for gay/lesbian individuals regardless of country context (due to the pervasiveness of sexual stigma across institutional contexts), yet these incentives vary more for heterosexual individuals (e.g. due to country-level gender or masculinity norms that disincentivize educational attainment).

Institutional context

Our findings invite additional observations regarding the possible influence of institutional context on the educational outcomes of LGB people. In the theoretical section, we discussed mechanisms leading to competing expectations, namely higher or lower educational attainment among LGB people in less stigmatizing contexts. Although higher levels of education and higher rates of upward mobility were observed for gay/lesbian people across countries, effect sizes differed. More specifically, effects were larger in Australia and the United States and weaker in Chile and Germany. It is hard to reconcile these results with these countries’ relative levels of discrimination, as the two contexts that would rank highest in stigma based on attitudinal data—Chile and the United States—are on opposite ends of the spectrum. Large confidence intervals and variation in measurement instruments can complicate the comparison of effect sizes across countries using these data, which calls for future research that (i) exploits variation within the same countries over time or (ii) deploys purposively collected cross-national datasets suitable for multilevel analytic techniques.

A final noticeable difference across countries was the diversity of outcomes for bisexual people. A disadvantageous pattern of intergenerational educational mobility was observed for bisexual men and women in Australia and bisexual men in Germany. This pattern of results was consistent with hypothesis H4, which underscored the additional stigma and stressors faced by bisexual individuals. However, the outcomes of bisexual people differed little from those of heterosexual individuals in other countries. It is possible that cross-national differences in attitudes towards bisexuality can explain this divergence, which constitutes a question that could be addressed in future research.

As noted earlier, an important area for future inquiry involves fully theorizing and empirically examining the role of additional contextual factors in moderating the relationships between sexual identity and intergenerational educational mobility. For instance, educational systems differ across countries, and features such as the age of selection into educational tracks might influence the association between sexuality and educational outcomes (Bol and Van de Werfhorst, 2013).

Study limitations

Some study limitations should be borne in mind when interpreting our results. First, our study shares limitations in measuring and operationalizing sexual orientation with other studies in the field based on general social surveys (Mize, 2016; Mittleman, 2022). These include the inability to capture fluidity in people’s sexual orientation (Diamond, 2008), particularly during educational careers; overly coarse sexual-identity measures that miss smaller identity groups (Goldberg et al., 2020); non-negligible shares of non-informative responses (Elliott et al., 2019); and reliance on a single dimension of sexual orientation (Mize, 2016). Although an array of sensitivity analyses offered reassurance regarding the validity and reliability of our key findings, future studies should aim to replicate our analyses using more nuanced, longitudinal measures of sexual orientation that go beyond individuals’ identity (e.g. encompassing sexual attractions and behaviour) and capture sexual identity as individuals navigate their educational careers.

Concluding remarks and implications

We hope that the present study provides a conceptual map for social scientists to further explore the intersections between sexual identity, educational attainment, and social mobility. Our findings have direct implications for an emerging literature aimed at documenting and explaining differences in educational attainment between gay/lesbian, bisexual, and heterosexual people (Black et al., 2000; Verbakel, 2013; Mollborn and Everett, 2015; Mittleman, 2022). We found that gay/lesbian people’s higher educational attainment is primarily produced by higher upward mobility among those from lower educational backgrounds. Such results suggest that explanations for educational differences by sexual identity should focus on the contexts within which individuals from lower educational backgrounds operate.

Our findings also bear implications for the principles of equality of opportunity. Overall, they suggest that academic attainment may be a way for LGB youth to invest in self-worth and increase their chances to independently construct their adult lives (Pachankis and Hatzenbuehler, 2013). If this is the case, gay/lesbian people’s higher upward mobility rates may be less an indicator of equal opportunities and more of an adjustment to experiences of early disadvantage. The educational advantages observed in this and other studies (Mittleman, 2022) contrast with studies on the labour market documenting ‘pay gaps’ that are especially unfavourable for gay men (Mize, 2016; Valfort, 2017). In this regard, there are clear parallels to research on gender inequality: women are now more highly educated than men, but a considerable gender gap in earnings and hourly wages persists (England, 2010). In short, gay/lesbian people are found to be more highly educated than their heterosexual peers, but this could be seen as a response to unequal environments rather than a sign of equality of opportunity.

More broadly, this study has showcased the need for social-stratification research to engage more seriously with issues of sexual orientation—a socio-demographic trait that has been largely absent from this literature. Sexual orientation is not only a salient socio-demographic feature, but, as our results underscore, a trait that can meaningfully alter processes of intergenerational mobility that are core to the field. Failure to recognize that background shapes the status-attainment process in different ways for heterosexual and LGB people can result in an incomplete picture regarding the population groups that are disadvantaged and the processes underpinning their disadvantage.

Supplementary Data

Supplementary data are available at ESR online.

Funding

This work was supported by the Spanish Ministry of Science, Innovation and Universities [RYC2019-026510-I] [PID2021-124267OB-I00], the European Research Council [ERC-2020-STG-948557-MINEQ] and the Australian Research Council Centre of Excellence for Children and Families over the Life Course (project number CE200100025). L.P. acknowledges assistance provided by the Population Research Institute at Penn State University, which is supported by an infrastructure grant by the Eunice Kennedy Shriver National Institute of Child Health and Human Development (P2CHD041025).

Diederik Boertien is a researcher at the Center for Demographic Studies (CED) in Barcelona. His research addresses questions about relationships and social inequalities. His work has been published in journals such as Demography, European Sociological Review, and Social Forces.

Francisco Perales is an Associate Professor of Sociology at The University of Queensland. His research uses longitudinal and life-course approaches to enhance our understanding of socio-economic inequalities in contemporary societies. His work has been published in outlets such as Social Forces, European Sociological Review, Demography, and Journal of Marriage and Family.

Léa Pessin is an Assistant Professor of Sociology and Demography at The Center for Research in Economics and Statistics (CREST) in ENSAE Paris and an external research affiliate at the Population Research Institute at the Pennsylvania State University. Her research focuses on the unequal consequences of the gender revolution on women’s work-family outcomes across class, race–ethnicity, and geographic contexts. Her recent work has been published in Demography, the Journal of Marriage and Family, and Social Forces.