Housing wealth and welfare

In many advanced economies over forty years there has been a shift to a more individualised, asset-based welfare regime, particularly in the UK (Toussaint and Elsinga, Reference Toussaint and Elsinga2009). As noted by Doling and Ronald (Reference Doling and Ronald2010), while private assets (e.g. pension savings) have long been part of a mixed economy of welfare, the shift over the past decades among many advanced industrialised economies has been to housing asset-based welfare. Variations in welfare regimes have meant that housing assets have been a complex part of household welfare provision, for example as identified by Esping-Andersson in southern Europe in his typology of European welfare regimes (1990) and in countries such as Singapore (Doling and Ronald, Reference Doling and Ronald2010). In the UK, there may not have been a specific policy goal that people would rely on their housing assets in their older age, yet broad demographic, economic, and social changes, combined with a policy landscape which provided a benign environment for home ownership (McKee, Reference McKee2012) means that outright home ownership has increased substantially among the over-fifty-fives in the UK over the past thirty years (Advani et al., Reference Advani, Bangham and Leslie2020).

There have been few examples of systematic research on the experience of LGBTQ+ people’s housing wealth and welfare outcomes. Florida and Mellander (Reference Florida and Mellander2010) argue that sexual minority households are seen to be a central part of the wealthy, high-income ‘creative class’ of Florida’s analysis of urban development, but their analysis relies upon cultural observations which might be biassed or selective. Conversely, there is a developing focus on the systematic exclusion of LGBTQ+ people in housing reflected in experiences of homelessness (McCarthy and Parr, Reference McCarthy and Parr2022), with some evidence LGBTQ+ people do have a higher likelihood of experiencing homelessness, although some methodologies rely upon data from surveys of organisations that only work with LGBTQ+ youth (Matthews et al., Reference Matthews, Poyner and Kjellgren2019).

There is also evidence that broader, historic patterns of exclusion could also have impacted on lesbians and gay men, in particular their ability to accumulate housing wealth. An internationally noted theme in the literature on LGBTQ+ neighbourhoods is the exclusion of single men and women from mortgage finance due to discrimination by mortgage lenders. This led to LGBTQ+ people becoming first-wave gentrifiers, buying cheap property with savings and using ‘sweat-equity’ to improve it (Doan and Higgins, Reference Doan and Higgins2011; Ghaziani, Reference Ghaziani2014). Within the UK, women, and thus lesbians and some bisexual women, were traditionally excluded from accessing mortgages due to the requirement to have a male signatory (Smith and Holt, Reference Smith and Holt2005); for much of the 1990s gay men were excluded from mortgage finance as insurers used the stigma towards HIV/AIDS to refuse life insurance cover to gay men, with intrusive questions being asked of applicants (Draper, Reference Draper and Sorrell2003). ^{Footnote 1} Conversely, data from Great Britain also shows that LGB people are more likely to live in London and the South East, or major urban centres, a recognised pattern globally as LGBT+ people seek to be around communities and facilities such as bars, cafes, and shops (Aksoy et al., Reference Aksoy, Carpenter and Frank2018; Ghaziani, Reference Ghaziani2014). This could have two very different impacts: either further excluding LGB people from home ownership due to higher house prices in these areas, or advantaging those who do enter home ownership due to higher relative house price inflation, particularly in London and the South East (Office for National Statistics, 2023).

The housing circumstances of LGBTQ+ people are of further interest due to what we know of the gendered nature of the sharing of household resources. The sharing of resources in a household is understood as being deeply gendered, with households ‘doing gender’ or ‘doing couple’ through their everyday pooling and negotiating of the allocation of resources coming into the household (Bennett, Reference Bennett2013). Housing is both a resource the household must share and rely on, but also one which often requires the pooling of income (to pay mortgage repayments, or rent) and eventually, for home owners, the pooling of wealth assets through shared ownership. Indeed, within the UK sole home ownership among couples is rare – around 8 per cent of male-female-gender couples have sole ownership of property (Lersch and Vidal, Reference Lersch and Vidal2016).

Among male-female-gender couples, home ownership is closely related to relationship formation, and marriage provided an easily accessible way to formalise joint-ownership of housing (Bayrakdar et al., Reference Bayrakdar, Coulter, Lersch and Vidal2019). Heteronormative housing pathways often have marriage and entering into home ownership as closely tied events, with formal legal recognition being associated with ‘nesting’ together. In the research to date (from the USA and Canada), it has been shown that same-sex couples have lower home-ownership rates than married opposite-sex couples, with same-sex male couples, in particular, having a larger home-ownership gap (Dilmaghani and Dean, Reference Dilmaghani and Dean2020; Leppel, Reference Leppel2007). This is explained by lower relationship commitment from men in all relationships, which could be more positively understood as differing relationship expectations and temporalities (Dilmaghani and Dean, Reference Dilmaghani and Dean2020). Same-gender couples are still restricted from accessing marriage rights in most countries, and in other states access to these rights has been relatively recent. In the UK, same-gender couples could form civil-partnerships from 2005, and marriage equality was granted in 2015. This may have presented a barrier to same-gender couples entering into ownership, either practically (legal agreements would have had to have been drawn up) or symbolically. Analysis from Canada, one of the first countries to equalise marriage rights in 2005, did show a reduction in the home-ownership gap between same-sex and opposite-sex couples after the legal change (Dilmaghani and Dean, Reference Dilmaghani and Dean2020).

These specific inequalities, or differences, experienced by LGBTQ+ people in accumulating housing assets, co-exist with broader changes in housing. While housing assets have become a more important part of how individuals, families, and households maximise their welfare over the life-course, there is growing concern that inter-generational inequality is systematically excluding people from home ownership (McKee et al., Reference McKee, Moore, Soaita and Crawford2017). Analysis by a wide range of research organisations has shown that in many countries in the Global North home-ownership rates are falling, and the age at which people purchase their first home is increasing as house price growth outstrips wage growth. In the UK in particular, this has paralleled a growth in outright ownership as the ‘boomer’ generation has paid off their mortgages (Bourquin et al., Reference Bourquin, Brewer and Wernham2022).

This coincides with substantial reductions in home ownership among younger people, with people under forty being dubbed ‘generation rent’ (McKee, Reference McKee2012; McKee et al., Reference McKee, Moore, Soaita and Crawford2017). Various commercial data sources suggest that the average age that people now make their first house-purchase is the mid-thirties and that this age is rising. Office for National Statistics data in England shows that in 1997, 68 per cent of thirty-five to forty-four year olds had a mortgage, and this had fallen to 50 per cent twenty years later. The proportion of the population in the private rented sector has increased from 11 per cent to 25 per cent over the same period (Office for National Statistics, 2020b).

Importantly for our analysis, this trend towards falling age of first-house purchase could interact with the known age-profile of sexual minority populations which has been impacted by historic discrimination. Until the last decade in Great Britain, lesbian, gay, and bisexual people experienced direct discrimination in the law, with access to marriage rights only being afforded in 2014/2015, and the last act that specifically discriminated against men (the criminalisation of buggery) only removed from the statute book in Scotland in 2013 (Tatchell, Reference Tatchell2017). More importantly for our discussion here is the centuries of discrimination against LGBT+ people that meant that people accepting they had a non-normative sexuality and being open about this was an extremely dangerous act (Weeks and Porter, Reference Weeks and Porter1998), and so people remained ‘heterosexual’. Yet, sociologists have suggested we live in an era of the ‘declining significance of homophobia’ (McCormack, Reference McCormack2012) due to the wider social acceptance of non-normative sexual identities and the lack of acceptance of homophobic attitudes in wider society (Swales and Taylor, Reference Swales and Taylor2017).

The impact of the changing nature of discrimination is reflected in a wide range of survey data. Our own analysis here (data and approach described below) shows that 1 per cent of over-seventies describe themselves as LGB, compared to 6 per cent of under-twenty-fives, with the biggest variation being the growing numbers of young people describing their sexual identity as bisexual. Thus, age interacts in a complex way with sexual identity and housing, as older people are always more likely to own a home, but further the likelihood of owning your own home has declined significantly in recent decades in the UK.

Methodology

Our following results are derived from secondary analysis of the UKHLS, also known as ‘Understanding Society’ (University of Essex Institute for Social and Economic Research, 2023). The UKHLS is a longitudinal household panel survey, which interviews a representative sample of up to 40,000 households across the UK in each wave of fieldwork. In this paper we used UKHLS data waves three (2011/12) through to eleven (2019/20). As the focus of this research project is Great Britain, we did not include responses from Northern Ireland. The unique history of Northern Ireland, and the experience of sexual minorities there due to ethnonationalism, makes inclusion of this data problematic. Put simply, the experiences of LGB people in Northern Ireland are different, and worse, than those in the rest of Great Britain (Hayes and Nagle, Reference Hayes and Nagle2019).

There are three challenges with the data we are using that must be acknowledged: the quality of the survey measure of sexual identity; sample size; and data on estimated house values.

There are long-standing and well-known issues with asking survey participants about their sexual identity. These issues range from the question of whether we are asking about sexual behaviour, romantic attraction, or identity, to more profound questions as to whether it is correct to use categories developed as a form of oppression within research about sexual minorities (Browne, Reference Browne, Browne and Nash2010; Guyan, Reference Guyan2022).

Sexual identity (referred to as sexual orientation in the dataset) was asked in waves three, five, seven, nine, and eleven of UKHLS. Following the approach of previous analysis, we selected participants’ most recent answer as their identity (Mann et al., Reference Mann, Blackaby and O’Leary2019; Powdthavee and Wooden, Reference Powdthavee and Wooden2015). The sexual identity question had five possible answers: straight/heterosexual; lesbian, gay/homosexual; bisexual; other; prefer not to say. We treated those who selected ‘other’ (0.84 per cent of the weighted sample) and ‘prefer not to say’ (2.83 per cent) as missing as we simply cannot infer anything about their sexual identity. This leaves us with around 97 per cent of respondents identifying as heterosexual and 3 per cent as LGB. The small sub-sample of non-heterosexual respondents can raise analytical challenges, for instance leading to non-significant results. The UKHLS is already a relatively large sample, but to boost the sample we chose to include multiple records per person for every year in which they have a UKHLS response, e.g. if a respondent appears in waves three, four, and five but drops out in wave six, they will have three records in our sample. We then used robust standard errors to account for this data structure. Increasing the sample size in this way allowed for more robust modelling, including a greater number of controls. Most analyses of UKHLS data also use sampling weights to adjust for known bias in the sample. UKHLS standard weights take account of other demographic variables but do not currently use sexual orientation data (comparable data was collected in the UK censuses for the first time in 2021/22, so future releases might include weights that also use this measure). Following standard advice we used the available individual weighting variable to correct for response bias (Institute for Social and Economic Research, 2022). However, sensitivity analysis was carried out comparing the weighted and unweighted data which showed negligible differences in the results.

Another challenge with the data is that participants in UKHLS are simply asked to estimate the value of their home, compared to other surveys which may rely on external surveyor’s valuations. There is a possibility this may not produce accurate data as home owners may have poor knowledge of relevant issues such as local housing market conditions. Nevertheless, this UKHLS data has been used in other analyses on inequalities and housing (Nutz and Gritti, Reference Nutz and Gritti2022; Smith et al., Reference Smith, Clark, Ong ViforJ, Wood, Lisowski and Truong2022). We analysed the house value data through several alternative functional forms but found comparable results across different operationalisations. Our results below often use a binary variable for if people owned a home worth over £200,000. The average UK house price in 2011 was £169,000 rising to £250,000 in 2020. As we are using data from across this period, £200,000 represents a proxy for owning a home above or below the national average. Our analyses also used housing tenure categories based on a direct question that is widely used in survey studies, asking people if they: own their home outright; own it with a mortgage; rent from a local authority or housing association; or rent from a private landlord.

Once we had cleaned the data to remove non-response and combined the waves of data, we were left with a sample as outlined in Table 1.

Table 1.

Records from waves 3–11 (years 2011–2020) of the special licence access version of UKHLS. UK Data Service. SN: 6931. Includes multiple records per person, five records per person on average. Respondents from NI excluded. “prefer not to say” and “other” sexual identities excluded. Weighted indinub_xw

For our analysis, we first of all provide descriptive statistics with housing tenure and the binary housing value variable. We then carried out regression modelling, with these as our dependent variables, LGB status as a key explanatory variable, and numerous other controls for variables which would likely have an impact on housing outcomes (described in more detail below). Gender presents one complication to modelling the outcomes of LGB respondents, since we could consider modelling women and men separately to allow for the many different ways that social processes often work differently for men and women. However, this has the challenge of reducing the LGB sample size further. We ran numerous sensitivity analyses which applied models separately for males and females, however in practice these did not normally lead to important differences in conclusions, and for the results presented below we usually combine the whole population and only include an interaction term between LGB status and gender to account for possible variation in the effect of LGB status on housing outcomes by gender.

Findings

Table 2 shows the proportion of each group living in different housing tenures and parallels previous findings from analysis of data relating to sexual identity and housing in the UK. This shows a higher proportion of LGB people living in the private-rented sector, a lower proportion in owner occupation, and a higher proportion in socially rented housing. As mentioned above, this may be related to the lower age profile of the LGB group.

Table 2.

Records from waves 3–11 (years 2011–2020) of the special licence access version of UKHLS. UK Data Service. SN: 6931. Includes multiple records per person. Respondents from NI excluded. “prefer not to say” and “other” sexual identities excluded. Owner occupation includes owned with a mortgage. Socially-rented includes housing rented from a local council, housing association or housing cooperative. Weighted indinub_xw

Figure 1 illustrates the contrasting experiences of housing advantage and disadvantage among LGB men and women looking at housing wealth. The figure shows the proportion of people in the broad categories of: no property owned; housing assets of £100,000–£200,000; housing assets of £200,000–£350,000; and housing assets over £350,000. The grey bars show the proportion of each gender (men and women) within each housing category in the data. Without any adjustment, the blue bars represent the change to that proportion for the relevant groups, gay men, bisexual men, lesbians, bisexual women, compared to the non-LGB population. This shows that a larger proportion of all four groups are located in the ‘no property owned’ category than the national average. The most extreme inequality is for bisexual women; whereas about one third of all women fall into the ‘no property owned’ category, this increases to nearly one half of all bisexual women.

Figure 1.

Proportion of the UK population with given housing circumstance, and changes to that proportion associated with selected gender-sexuality groups. Data: records from waves 3–11 (years 2011–2020) of the special licence access version of UKHLS. UK Data Service. SN: 6931, with direct standardisation weights for age adjustments.

As noted, these differences in home ownership may just reflect the far younger age profile of the LGB population compared to the non-LGB population. To account for this, we also use ‘direct standardisation’ to weight the data for LGB people as though they had the same age distribution as non-LGB people. These adjustments are shown by the orange bars of the figures, with the orange line showing the described trends more clearly. This age adjustment still shows a similar patterns of inequality although it does reduce the differences slightly, reinforcing the evidence of inequalities for LGB people. For gay and bisexual men, and for bisexual women, there is still an overrepresentation in non-home-owning. Within home owners, there is underrepresentation for these three groups in those who own property valued between £100,000 and £350,000. However, there is little difference in the proportion of these groups owning property worth over £350,000 when we adjust for age. Lesbians have a very different ‘n’-shaped profile, being underrepresented in the extreme categories, but overrepresented among those who own housing worth £100,000–£200,000.

Regression modelling

Our descriptive analysis suggests there is inequality in home ownership and housing assets owned among LGB people in Great Britain, even when we adjust for the different age profile of this group – a key demographic factor associated with home ownership. We now turn to our regression modelling to add further controls associated with home ownership, with a simplified model presented in Table 3. Detailed results are available in the online supplementary materials. These continue to show the same patterns of inequality as the descriptive analysis, namely LGB people are more likely to be non-homeowners and rent their homes compared to their heterosexual counterparts.

Table 3.

Results from series of regression models with different controls for 4 outcome variables

Records from waves 3–11 (years 2011–2020) of the special licence access version of UKHLS. UK Data Service. SN: 6931. Multiple records per person with robust standard errors. Sample restricted to those aged 18–64.

ns = not significant, + = positive association, – = negative association.

* P < 0.05.

** P < 0.01.

*** P < 0.001

Table 3 shows results from a series of regression models with different controls for four outcome variables. Firstly, logistic regressions for home ownership measured as a binary variable (0=’not an owner’ 1=’owner’). Secondly, OLS regressions for logged scale of home value, placing those who do not own at the bottom of scale. Thirdly, OLS regressions for logged scale of home value restricted to owners only. And finally, logistic regressions for home value measured as binary variable where zero was less than £200,000 and ome was more than £200,000.

Model B includes controls for: gender, age, age squared, having a degree, having a disability, having children, being married or cohabiting, living in an urban area, living in the South East and London, being white British, not being born in the UK, and wave of UKHLS. In choosing controls for this model, we carefully considered how our variables might relate to LGB status in choosing whether to include or exclude them. We control for disability and health status to recognise the known far-lower rates on home ownership among disabled people (Satsangi et al., Reference Satsangi, Theakstone, Matthews, Lawrence, Rummery, Mackintosh and Boniface2018). We included a quadratic age function to adjust for the different age profiles of the populations, as discussed. Other controls have similar complex relationships. Having a degree is associated with higher rates of home ownership, and LGB people, particularly lesbians and gays, are more likely to be degree educated than their heterosexual counterparts (Scottish Government, 2017). In terms of geography we recognise the wide regional variations in house prices across Great Britain, along with the known patterns of LGB residential choice, described above. To account for these we control for living in an urban area (based on the Office for National Statistics urban-rural classification) and living in the South East and London. Within the UK, ethnicity and immigration status are complexly interwoven with home-ownership, so we also control for these (Office for National Statistics, 2020a).

Model C include the same controls and adds an LGB*Female interaction term to account for potential differences by gender in the effect of LGB status on housing outcomes. Model D is the same as model C but with an additional LGB*Children interaction term; and model E is the same as model C with an additional LGB*Married/cohabiting interaction term. We include being married or cohabiting, and having children, as among heterosexual couples these are widely recognised as key life-stages when people enter into home ownership (Tocchioni et al., Reference Tocchioni, Berrington, Vignoli and Vitali2021).

Going through the model in Table 3 in detail, beginning with home ownership in the bivariate linear regression being LGB has a significant negative association with home ownership in the initial Model A. This effect remains when the controls for other demographics are added in Model B. However, the main effect of being LGB drops out when an interaction between gender, and being LGB is added in Model C, although this interaction is also not significant. This suggests there is a modest pattern that connects LGB status with home ownership probabilities, which may be different for men and women; however, there is not enough power in our dataset to confirm this as statistically significant. Interactions between being LGB and having children (Model D) and being LGB and having a partner (Model E) are significant and negative. This suggests that the negative effect of having children on home ownership is stronger for LGB people, and the effect of having a partner on home ownership is not as strong for LGB people as it is for heterosexual people.

Moving on to look at home value, we have considered models both that include non-owners (value assigned as zero), and that are restricted to the homeowner population. When non-owners are included in our model we find a similar pattern of the main negative effect of being LGB dropping to non-significant when a gender interaction term is added in Model C. Again, we see that the negative effect of having children is stronger for LGB people, and the effect of having a partner on home ownership is not as strong for LGB people. When we go on to look at home value restricted to owners only, in Model A, LGB status is non-significant, but becomes negatively significant when the controls in Model B are added. When an interaction with gender is fitted in Model C, the effect again becomes non-significant. Again, this suggests a similar pattern to the model for home ownership, with some negative association between home value and LGB status. The interaction effect between being LGB and having children (Model D) has a similar pattern to the above models; however, this becomes non-significant, likely due to the smaller sample size. The interaction between being LGB and having a partner (Model E) is significant, but the direction of the effect is positive so this suggests that having a partner has a stronger positive effect on house value among home owners for LGB people than it does for heterosexual people.

When home value is measured as a binary outcome of owning a home worth over £200,000 the same broad patterns are found. The main negative effect of being LGB drops to non-significant when a gender interaction term is added in Model C; and the negative effect of having children is stronger for LGB people (Model D), and the effect of having a partner is less positive, but also not significant in Model E.

As discussed in the methodology, the challenges of analysis of the small sub-population are apparent in the modelling, with results dropping to be not significant with the data we are using, even when we are using a large total sample size. While the modelling results are complex, they do suggest that, controlling for a wide range of factors that may increase the likelihood of home ownership, or owning a home of higher value, LGB people in Great Britain are disadvantaged. However, when we add the interaction terms a lot of these results become more complex, suggesting systematic inequalities impact only some LGB people, conditional upon gender, marital, and family status. Amongst the interaction terms, we highlight firstly that LGB people with children are statistically significantly less likely to own a home, and secondly the pattern that LGB people who are married or are co-habiting are less likely own a home, and less likely to own a home worth over £200,000, although among homeowners their homes are likely to be worth more on average. Overall, this does suggest that LGB people do not benefit from life events such as partnering and having children for home ownership in the same way that heterosexuals do.

Overall, our descriptive analysis and statistical modelling therefore suggest that LGB people do experience relative housing asset disadvantage in the UK, with implications for accumulating housing assets to support their individual welfare.