Relevant literature

There can be little doubt that the uneven distribution of wealth and assets across individuals and families is a key driver of inequalities in other important socio-economic outcomes such as earnings, qualification attainment, and social class (Davenport et al., Reference Davenport, Levell and Sturrock2021). Hamnett (Reference Hamnett1991), for example, used Inland Revenue data to study the growth of housing inheritance in Britain since the late 1960s. Finding a strong social patterning in inheritance by housing tenure, social class and region, he noted that housing inheritance has not disrupted existing patterns of class-related inequality, but could rather be seen as reflecting and perpetuating these existing inequalities. From a vantage point preceding the ongoing housing boom that began in the 1990s, Hamnett anticipated that divisions between those who inherit housing wealth and those who do not would sharpen; those excluded from homeownership would come increasingly from renting households with lower income and skills. These expectations have largely been borne out, with housing wealth now more unevenly distributed than income (Blanden et al., Reference Blanden, Eyles and Machin2023; Crawford et al., Reference Crawford, Innes and O’Dea2016).

In addition to direct intergenerational wealth transfers through gifts and inheritance, family wealth can also perpetuate inequalities by facilitating access to privilege-generating locales, institutions, and networks. For example, even with comparatively low earnings, wealthy families can invest in private education for their children, tutoring to pass school entrance exams, or they can move to areas with high-performing state schools. These strategies increase the chances of their offspring graduating from a high-status university. In turn, this facilitates entry to well-paid professional occupations and the ability to purchase a home and invest in wealth-producing financial instruments, thus perpetuating the intergenerational cycle of wealth accrual and transmission (Macmillan et al., Reference Macmillan, Tyler and Vignoles2015). Similarly, wealthy families can support investment in housing for their adult children through provision of deposits, which enables them to live in towns and cities with high-productivity labour markets. And there are myriad other ways in which family wealth is deployed to perpetuate material and opportunity advantage from one generation to the next. Wealth is therefore key to understanding social and economic immobility across generations.

Yet wealth is also notoriously difficult to measure in the kinds of surveys that are generally used in studies of social mobility. Partly, this is a result of the complex and multi-faceted nature of wealth, but it is also because many people are unwilling to disclose their wealth in surveys. For these reasons, the role of family wealth in shaping other dimensions of social mobility is currently not well understood. Recent exceptions to this are Gregg and Kanabar (Reference Gregg and Kanabar2021) who find evidence of increasing intergenerational wealth persistence among recent cohorts in Britain using the Wealth and Assets Survey (WAS). These authors also demonstrate that the increase in the intergenerational persistence of wealth is driven mainly by inequalities in home ownership (Gregg & Kanabar, Reference Gregg and Kanabar2022). For example, they found that by age 35 years, the rate of homeownership was three-times higher among adults whose parents were high-educated homeowners compared to those from a low-educated renter background. The high-educated home-owner group also held approximately ten times the level of housing wealth than the low-educated renters. If family wealth is itself generative of disparities in educational, occupational, and income attainment then its increasing concentration amongst the already wealthy means that, ceteris paribus, social mobility in these domains will be inhibited.

To contextualise our later findings, it is necessary to provide a brief account of the origin and distribution of ethnic minority groups in Great Britain. The current composition and spatial patterning of ethnicity in Britain has been shaped by a complex mix of historical migration flows, British colonialism, trade, and globalisation with most of the large-scale migration beginning after the Second World War (Byrne et al., Reference Byrne, Alexander, Khan, Nazroo and Shankley2020). Labour shortages to support post-war reconstruction and staffing the nascent National Health Service led to the government actively encouraging migration from Commonwealth nations. This included people from the Caribbean coming to the UK in search of improved economic prospects from the 1940s, the so-called ‘Windrush’ generations. Similarly, many Indian, Pakistani, and Bangladeshi people migrated to Britain during the 1950s and 1960s, filling vacancies in manufacturing, textiles, and services, notably across towns and cities in northern England. The Chinese population in Britain is diverse, including many migrants from Hong Kong – due to the island’s history as a British colony – and more recent migration from Mainland China for work and education. With the expansion of the European Union from 2004, migration from Eastern European countries has also increased, as people migrated to Britain to fill labour shortages, or for education. This has led to an increase in the ‘White Other’ ethnic category, which also includes migrants from Australia, New Zealand, Canada, and the United States.

These broad ethnic groupings have had quite different experiences in Britain’s education system and labour market (Li & Heath, Reference Li and Heath2008, Reference Li and Heath2016, Reference Li and Heath2020; Lindley, Reference Lindley2005; Modood et al., Reference Modood, Berthoud, Lakey, Nazroo, Smith, Virdee and Beishon1997). In terms of education, the General Certificate of Secondary Education (GSCE) exams taken at age 16 years, are an important gateway to further vocational or academic study in the UK. On average, the Chinese, Indian, Bangladeshi, Black African, White Irish, and White Other groups are more likely to gain a ‘good pass’ (Grade 5 or above) in English and maths compared to the White majority, whereas Black Caribbean, Black Other, and Pakistani are less likely (DfE, 2022). In term of labour market outcomes, several minority ethnic groups – Black African, Black Caribbean, Pakistani, and Bangladeshi – have higher rates of unemployment and lower earnings compared to the White majority (Li & Heath, Reference Li and Heath2020). The same groups have also fared worse during recessions, experiencing higher unemployment, wage penalties, and later re-entry into the labour market (Li & Heath, Reference Li and Heath2008). Indian and Chinese groups, however, have tended to have similar, or better, socio-economic outcomes compared to White British (Li, Reference Li2018; Li & Heath, Reference Li and Heath2020).

With regards to patterns of intergenerational mobility, higher rates of absolute downward social class mobility have been found for some first-generation immigrant groups. For example, Li and Heath (Reference Li and Heath2016) showed that Black African immigrants to Britain were mostly drawn from higher social class groups in their country of origin but were frequently unable to pass these privileged positions across generations in Britain, post-migration. Similarly, Platt (Reference Platt2005) found higher rates of downward mobility among first-generation Black Caribbeans – in contrast to Indian immigrants who were better able to transmit their comparatively privileged origin conditions to their offspring. While there is some evidence of ‘catch-up’ in outcomes for later generations (Li & Heath, Reference Li and Heath2016), mobility scholars have also shown that minority ethnic groups have greater relative mobility compared to the white majority (Li, Reference Li2018), especially the Black African (Li & Heath, Reference Li and Heath2016) and Black Caribbean groups (Platt, Reference Platt2005).

Higher levels of relative mobility can be the result of success in accessing the salariat but, as noted previously, may also indicate a lower capacity to maintain socio-economic advantage from one generation to the next (Platt, Reference Platt2005). Indeed, Pakistani and Bangladeshi, Black Caribbean, and Black African groups, in particular, have been more likely to experience downward mobility in Britain than other ethnic groups (Li, Reference Li2021; Macmillan & McKnight, Reference Macmillan and McKnight2022). A potentially important factor underpinning these ethnic differences in social class mobility is how patterns of home ownership. Homeownership rates in Britain peaked at 69% in 2001, falling to 64% by 2011 – the first fall since 1918 (Office for National Statistics, 2015). In the 2011 Census, the majority of households of Indian, White British, Pakistani, White Irish, and Chinese background were owner-occupiers. Those of Indian ethnicity had the highest rate of home ownership (69%), followed by White British (68%), Pakistani (63%), and White Irish (61%), while rates were lowest for the Black African (24%) and Arab (27%) groups. Between 1991 and 2011, rates of home ownership decreased for all ethnic groups (Finney & Harries, Reference Finney and Harries2013). The overall picture, then, is not one of the white majority faring best on these outcomes but of heterogeneity, with some groups performing better than the majority group and others significantly worse (Commission on Race and Ethnic Disparities, 2021; Li, Reference Li2021).

Differences in home ownership between ethnic groups are shaped by a range of factors, including concentrations of earlier migrant flows, location preferences, local house prices, labour market conditions, and differential access to mortgage financing. Discrimination has also shaped experiences of ethnic minorities across all sectors of the housing market (Phillips, Reference Phillips2003; Rex & Moore, Reference Rex and Moore1969), including residency requirements acting as a barrier to accessing social housing, and discrimination by landlords in the private rental sector (Lukes et al., Reference Lukes, de Noronha and Finney2019). The implications of reduced access to home ownership for wealth accumulation are significant; as housing affordability declines, ownership increasingly depends on intergenerational transfers, the so-called ‘bank of mum and dad’. Sanderson and Udagawa (Reference Sanderson and Udagawa2017), for example, found 35% of first-time buyers in England in 2017 supported their house purchase with a parental gift or loan, up from 22% in 2000. Blanden et al. (Reference Blanden, Eyles and Machin2023) also found that the probability of home ownership has increasingly come to depend on buyers’ parents being home owners. Gregg and Kanabar (Reference Gregg and Kanabar2022) also found that parental wealth has become increasingly predictive of housing wealth in the next generation. Even among current homeowners, those from the wealthiest backgrounds reported ten times more housing wealth than individuals from the most disadvantaged backgrounds.

If home ownership is increasingly the preserve of the offspring of home-owning parents, this will likely exacerbate existing inequalities between ethnic groups. In 2011, the home ownership rate among Black households was 24%, compared to 68% among White British households (Finney & Harries, Reference Finney and Harries2013), and the median housing wealth among Black African and Black Caribbean households in Britain in 2018 was £0, compared with £115,00 among White British households (Office for National Statistics, 2020b). That stark differences in homeownership and wealth exist across ethnic groups is now well known. Less well understood, however, is how ethnic groups differ in how this housing wealth is passed between generations. It is to this question that we now turn in our empirical analysis.

Data and measures

The Office for National Statistics Longitudinal Study (LS) is a 1% sample of the population of England and Wales, linking census and administrative data on births, deaths, and cancer registrations, since the 1971 Census (Shelton et al., Reference Shelton, Marshall, Stuchbury, Grundy, Dennett, Tomlinson, Duke-Williams and Xun2019). The original LS sample was selected from the 1971 Census by identifying records for all individuals born on four (undisclosed) dates in the year. It is one of the largest nationally representative longitudinal studies in the UK with a sample size of over 500,000 in each Census year. The advantage of the LS for social mobility research is that the entire household is observed at each Census. Therefore, parental demographic characteristics, including occupation and housing tenure, are measured when the study member was a child. Such parental characteristics can be used to measure ‘origin’ status on a range of measures and compared to study members’ outcomes on these same variables in adulthood.

Here, we restrict the sample to study members who were aged 8 to 17 years in the 1971, 1981, and 1991 censuses, making this the age of origin for housing tenure for three consecutive cohorts. The analysis sample is composed of study members born in England or Wales (97% of the sample), or who arrived as immigrants during their formative years (3% of the sample). Following Li (Reference Li2018), our ethnic minority sample members can be categorised as second or higher generation migrants on the basis that they would have received most (or all) of their education in Britain, and would acquire the same kind of human capital and have as fluent English as the British White group. We link study members’ records to their data in the Census 20-years later, when they are aged 28–37 years, and identify their housing tenure at this point. This 28–37 years window includes the average age of first home purchase in England and Wales; between 2015 and 2017, the average age of first home buyer in the UK was 31 years and 11 months among White British, and 32 years and 5 months among all other ethnic groups (Office for National Statistics, 2021). We then estimate the association between origin and destination housing tenure status by ethnic group to assess the extent to which an individual’s probability of home ownership depends on whether or not their parents owned their own home when sample members were children. Our novel contribution is to also examine whether this association differs between ethnic groups and how, if at all, it has changed over time.

Analysis

Table 1 provides summary statistics for the LS analysis sample described above. The first column reports the sample size for each ethnic group by cohort, the average age of the cohort members, and the average age of sample members’ parents. The second panel reports home ownership rates and the proportion who are in the National Statistics Socio-economic Classification (NS-SEC) groups 1 or 2 (managers and professional occupations). This shows the steady increase in this group for the population and across ethnic groups that has been well documented elsewhere (Buscha and Sturgis, Reference Buscha and Sturgis2018; Platt, Reference Platt2005). Black, Pakistani, Bangladeshi, Indian, and Asian groups have a relatively lower share of the parental generation in managerial or professional roles, compared with White British, White Other, and Mixed and Other groups. The mean age of study members at first follow-up is 32 years (range = 28–37 years) and the mean parental age at origin was 41 years for most ethnic groups and cohorts. This is 10 years later than the typical age of first home purchase (which was, on average, 31 years for White British and 32 years for other ethnic groups between 2015 and 2017) (Office for National Statistics, 2021).

Table 1. Descriptive statistics and sample sizes

Note: Data source is the ONS-LS. N is the sample size; Age is age of the study member at follow-up (‘destination’); parental age is age of the parent at ‘origin’ – when the main study member was a child; NS-SEC 1/2 is the proportion of main study members in a professional or managerial role at ‘destination’; origin owner indicates whether the study member’s parents owned their own home; destination owner indicates whether the study member owned their own home in adulthood.

Figure 1 plots the association between age and homeownership for the three cohorts, separately by housing tenure origin status. As would be expected, for all three cohorts, the probability of homeownership increased with age. However, across successive cohorts, the probability of homeownership declined, with a substantial decrease between cohorts 2 (born 1964–1973) and 3 (born 1974–1983), a trend which has also been reported by Blanden et al. (Reference Blanden, Eyles and Machin2021). The size of the decline in home ownership over the three cohorts is notably larger among those whose parents rented than it is amongst those whose parents owned their home. Among those born between 1954 and 1963 (cohort 1) who grew up in rented accommodation, 71% had become homeowners by age 28–37 years, while for those born between 1974 and 1983 (cohort 3) who grew up in rented accommodation, just 47% had become homeowners by the age of 28–37 years. In contrast, for those of owner origin, the decline in homeownership was much less pronounced. For cohort 1, 86% of those with home-owning parents themselves became homeowners, falling to 75% for the third cohort.

Figure 1. Ownership proportions by cohort, age, and housing tenure. Note: Data source is the ONS-LS.

Table 2 presents transitions between origin and destination housing tenure by ethnicity and cohort. Conditional on being of renter origin, the probability of upward housing mobility (moving from renter origin to owner destination status) was 65% for the White British group, pooled across cohorts. Conversely, conditional on growing up in an owner-occupied home, the probability of downward housing mobility (moving from owner origin to renter destination) was just 19%. These estimates are quite different for minority ethnic groups. Individuals of Indian ethnicity experienced a significantly higher rate of upward housing mobility, at 81%. In other words, the relative risk of Indian individuals who grew up in rental accommodation transitioning to home ownership is 25% greater than for White British. For the Pakistani/Bangladeshi, Mixed and Other, and Black and Other Black groups, the rates of upward mobility were lower compared to White British, with relative risks of .85, .82 and .68, respectively. Across all three ethnic groups, only around half of those from renter origins transitioned to home ownership as adults.

Table 2. Transitions between housing tenure of parents and housing tenure in adulthood

Note: Data source is the ONS-LS. Figures for Chinese & Other Asian are excluded from cohort 1 due to lower cell sizes.

Also noteworthy in Table 2 is the large difference in the risk of downward housing mobility between the Indian and Black ethnic groups. Across all three cohorts, Indians experienced the lowest rate of downward mobility, at just 10%, while for Black individuals this was over three times higher, at 34%. The difference between the Indian and Black groups is also reflected in the higher rates of ownership stability between generations, with the Black and Other Black and Mixed and Other groups considerably less likely to maintain home ownership status and more likely to remain renters from one generation to the next. The cohort-specific tables show that the chances of upward housing mobility have decreased, while downward housing mobility has increased for all ethnic groups. The increase in downward mobility was lowest for Black individuals but this was primarily due to the already high rates of this trajectory in cohort 1.

Figure 2 Footnote ² shows home ownership rates by ethnic group for individuals from renter (top part of Figure 2) and owner (bottom part of Figure 2) origins, which are again differentiated by ethnicity. Table 1 showed that, across all three cohorts, those of Indian ethnicity had the highest rate of home ownership in the parental generations, while those of Black ethnicity had the lowest. This is important because the combination of a low home ownership at origin paired with a higher intergenerational persistence works to perpetuate inequalities in homeownership and wealth. Across all ethnic groups and cohorts, those of ownership origin had a higher ownership probability compared to those whose parents rented. The general decline in home ownership over the period is predominantly concentrated amongst people whose parents rented, although this difference between origin renters and origin owners is smaller for the Indian and Chinese/Other Asian groups. 91% of cohort 1 Indians with home-owning parents became homeowners themselves, falling by just two percentage points to 89% for the third cohort. For those or renter origin, the corresponding figures were 84% and 75%. In contrast, 88% of Pakistani/Bangladeshis with home-owning parents in cohort 1 went on to become homeowners themselves, with this figure declining to 84% for the third cohort. For Pakistani/Bangladeshis with renter parents, these estimates were 69% and 49%, respectively. This represents a substantial change from 1.28 to 1.71 in the relative risk of home ownership for owner over renter origin individuals in this ethnic group. For those in the Black group, homeownership rates were low across both tenure origin statuses, though particularly so among those with renting parents, where the homeownership rate was 49% for cohort 1, dropping to just 32% for cohort 3.

Figure 2. Ownership proportions by origin tenure, ethnicity, and cohort. Note: Data source is the ONS-LS.

It is likely that some of the differences between ethnic groups in housing tenure and its persistence across generations is a result of the different kinds of areas that ethnic groups tend to live in, particularly the greater tendency of ethnic minorities to live in metropolitan areas, where the cost and availability of housing can differ substantially compared to more rural areas. To control for such factors, and to facilitate summary of estimates in a more succinct and interpretable manner, we use a regression framework for the remaining analyses. We estimate the parameters of linear models, of the form described in Equation 1, using ordinary least squares (OLS):

(1) $${D_i} = \alpha + \beta {O_i} + \gamma {E_i} + \delta {O_i}.{E_i} + {X_i} + {\varepsilon _i}$$

${D_i}$ , is a binary variable denoting housing destination status (renter = 0, owner = 1), ${O_i}$ measures housing tenure at origin for individual i, and ${E_i}$ is a categorical variable indicating ethnic group. The ${O_i}.{E_i}$ term is an interaction between origin housing tenure and ethnic group. ${X_i}$ is a vector of covariates which, in the full specification, includes sex, age and age-squared, dummy indicators for government office region, Local Authority, within-cohort deciles of Local Authority median house prices, and cohort. The cohort indicators are included to control for shocks that affect the entire cohort, such as macroeconomic impacts. $\beta $ , $\gamma $ and $\delta $ are regression coefficients to be estimated and ${\varepsilon _i}$ is an idiosyncratic error term. Results from the simplest version of the model in Equation 1, including only sex, age, and age-squared as controls, are presented in Table 3.

Table 3. Linear probability model of home ownership in adulthood

Note: Data source is the ONS-LS. Control variables included are: sex, age, and age-squared. Robust standard errors in parenthesis.

The main effects for parental homeownership status show the strengthening over cohorts in intergenerational persistence of homeownership for the baseline group, White British, that was also seen in Table 1. For this group, having homeowner parents increased the probability of homeownership – compared to having renting parents – by 16 percentage points for cohort 1, 19 percentage points for cohort 2, and 27 percentage points for cohort 3. Turning to the main effects for ethnicity, there are substantial differences in home ownership by ethnic group (among those of renter origin). Indians are notably more likely to be homeowners by age 37 years compared to all other ethnic groups. Compared to White British, Indians of renter origin were 12 (cohort 1) to 29 (cohort 3) percentage points more likely to be homeowners. In contrast, those of Black ethnicity with renter parents were less likely to own a home compared to White British, by 22–14 percentage points for cohorts 1 and 3, respectively. Other ethnic groups of renter origin did not experience significantly different levels of home ownership compared to White British.

The coefficients of primary interest in Table 3 are the interactions between origin ownership status and ethnicity. These measure how the association between parent and child housing tenure varies by ethnic group (relative to the reference category, White British). The Indian group had a lower association between parent and child housing tenure across cohorts, with coefficients ranging from −0.08 to −0.14 from cohort 1 to cohort 3. This means that for Indian individuals, being of owner origin was less important for becoming a homeowner compared to the White British group. In cohort 3, the probability of being a homeowner if your parents were homeowners was 14 percentage points lower for Indians compared to the White majority. A coefficient of similar magnitude is estimated for the White Other group, who also had lower intergenerational persistence in housing tenure compared to White British, an effect which is mainly driven by cohorts 1 and 2.

The Pakistani and Bangladeshi group had a high intergenerational homeownership association in the third cohort, and for all cohorts pooled. A coefficient of 0.085 in cohort 3 represents an 8.5 percentage point higher intergenerational homeownership association compared to White British. Similarly, the Black group did not differ in their intergenerational homeownership association across the three cohorts compared to White British, but when pooled across cohorts, had a significantly higher intergenerational homeownership association. This reflects the fact that, as we saw in Table 2, the Black group is more likely to be stable renters across generations. The Chinese and other Asian group had a low intergenerational housing association in the first cohort, but this became statistically indistinguishable from the White majority by cohorts 2 and 3, by which time the ownership rate for this group had increased substantially.

Figure 3 presents marginal effect estimates from the pooled model specification in Table 3 in graphical form; the difference in the predicted probability of homeownership for individuals of homeowner origin compared to renter origin is plotted for each ethnicity by cohort. Recall also that parental homeownership rates are important for interpreting the intergenerational housing associations because whether a high association is normatively desirable depends on whether an ethnic group has high or low initial rates of home ownership. For instance, the Black group had low parental homeownership rates, as well as strong persistence of housing tenure across generations, whereas the Indian group had high parental homeownership, with low persistence of tenure across generations – parent ownership was less important for Indian individuals in gaining access to the housing ladder for Indians.

Figure 3. Marginal effect of parents owning a home on main study member owning a home (intergenerational housing mobility) for each ethnicity. Note: Data source is the ONS-LS.

Finally, Table 4 adjusts for the additional covariates. Model M0 in Table 3 reproduces the pooled estimates from Table 2 for ease of comparison. Model M1 adds indicators for Government Office Region (there are nine regions), model M2 replaces region with Local Authority fixed effects, model (M3) adds median house price deciles (which vary by cohort and Local Authority), and model M4 adds cohort indicators to account for secular cohort-specific changes (to, e.g., the macroeconomic environment) over time. The purpose of taking these factors into account is not to ‘explain away’ ethnic group differences in homeownership mobility, but to identify possible underlying mechanisms leading to between group differences in home ownership and its persistence across generations.

Table 4. Pooled linear probability regressions with controls

Notes: Data source is the ONS-LS. Robust standard errors in parenthesis.

Adjusting for government office region, in Model 1, the patterns seen in Table 3 persist; the Pakistani, Bangladeshi, and Black groups have a stronger link between origin and destination housing tenure, whereas the White Other and Indian groups have a weaker link. Adjusting for Local Authority, a more granular geography, in Model 2, the coefficients for Pakistani and Bangladeshi and Black groups become smaller. Both interaction coefficients decrease by approximately a half, indicating that intergenerational differences between the Pakistani and Bangladeshi and Black groups compared to the White British group is in part explained by local area characteristics. The findings also remain largely unaltered after adding time-varying controls for house prices. Finally, adding an indicator for cohort reduces the size of the interaction term for Pakistani, Bangladeshi, and Mixed and Other groups to close to zero. One potential reason for this is that nationwide economic conditions that affect all groups equally – such as the stagnation in real wages and increases in house prices – are an important contributor to differences in intergenerational housing mobility for these groups. In other words, the difference in intergenerational housing persistence may be explained by the disproportionate impact of macro-economic conditions on the Pakistani, Bangladeshi, and Mixed and Other Black groups. However, we also cannot rule out alternative explanations, such as the effects of unobserved confounders which change over time, or differences in statistical power across cohorts.