Social Disparities in Private Renting Amongst Young Families in England and Wales, 2001-2011

Abstract In Britain, the proportion of young families living in the private rented sector (PRS) has risen sharply in recent years. There is mounting concern that this trend could be particularly pronounced amongst less advantaged young families, who may be disproportionately channelled into relatively costly, insecure and lower quality accommodation in the PRS by growing difficulties accessing other tenures. In consequence, this paper uses the Office for National Statistics Longitudinal Study of England and Wales to compare how family structure and socio-economic characteristics shaped rates of private renting amongst young adults heading families in 2001 and 2011. The results show that social disparities generally increased during this period as private renting expanded most rapidly amongst some types of lone parent and amongst young adults heading couple families with a less advantaged class position. Increasing housing inequalities between young people may thus be as much a feature of “Generation Rent” as deepening divides between generations.


Introduction
British society is being reshaped by a revival of private renting. Between 2003 and 2014, the proportion of English households renting privately rose from 11 to 19% as mortgaged homeownership declined (DCLG 2015: Annex Table 1.1). As these tenure shifts have been particularly pronounced amongst younger adults, public debates often characterize young Britons as "Generation Rent" (Rugg and Quilgars 2015). Although the Generation Rent phenomenon arguably exposes the long-term risks of tying economic prosperity and welfare provision to housing assets, no British Government wishes to preside over falling owner-occupation and policy responses have thus focused on assisting young people into homeownership (Ronald, Kadi, and Lennartz 2015).
Young Britons' deepening reliance on the private rented sector (PRS) 1 is typically thought to be the result of a cocktail of interlinked choices and constraints produced by changing life course trajectories and contextual conditions (McKee 2012). On the one hand, the flexibility and low costs of moving within the PRS make it an attractive tenure as young people defer the transition to settled adulthoodfor example by spending longer in education, "job-shopping" or by postponing family formation. However, young people are also finding it harder to live outside the parental home in other tenures (Clapham et al. 2014). Stock contraction means that access to the social rented sector (SRS) is increasingly restricted to only the most vulnerable, while high house prices, more stringent mortgage lending and the precariousness generated by low incomes, job insecurity and student debts constrain homeownership (Rugg and Quilgars 2015). Although some of these pressures are also affecting young people in other countries, in many parts of Europe there has been a much less marked growth of private renting in young adulthood (Lennartz, Arundel, and Ronald 2015). This indicates that institutions and structural conditions influence how young people's housing transitions and hence pathways to adulthood are being reshaped in the wake of the Global Financial Crisis (GFC).
Public debates about young adults' changing housing careers often highlight two adverse consequences of their growing dependence on private renting. The first fear is that this trend could exacerbate social inequality. While there has been much discussion of how the increasing difficulty of accessing homeownership may be deepening intergenerational wealth inequality (Griffith 2011), many authors worry that housing inequalities between young people are also growing as only those with abundant personal and familial resources can afford to become homeowners (McKee 2012). In the absence of an accessible SRS, this means that less economically advantaged young adults are probably becoming particularly reliant on the PRS (Clapham et al. 2014). Over time, this could deepen social inequality by disproportionately exposing less affluent young people to the relatively high housing costs and poor dwelling conditions prevalent in the PRS (Shelter 2012). 2 The welfare cuts enacted since the Coalition Government came to power in 2010 may have exacerbated this marginalization process by reducing young people's housing benefit entitlements, thereby channelling the less affluent into shared rentals and poor-quality accommodation (Powell 2015). 3 These trends appear to be a part of a broader European pattern, whereby austerity policies and the commodification and financialization of housing systems deepen social divisions by displacing the responsibility for welfare provision onto citizens while making housing less secure and affordable for the poor (Dewilde and De Decker 2016;Kennett, Forrest, and Marsh 2013).
Secondly, many academics, politicians and charities are concerned that the PRS is failing to adequately house the growing number of young families who are renting privately as they cannot access homeownership or the SRS (Kemp 2015;Labour Party 2012;Shelter 2012). Until recently few families lived in the PRS, which primarily housed mobile and transitional population groups such as students, young workers, newly formed households and recent immigrants (Rugg and Rhodes 2008). Although some young families naturally also value the opportunity to move easily (Rugg 2010), the short-term nature of most private tenancies can make it difficult for the majority of families seeking stable long-term accommodation to avoid expensive and disruptive residential moves. Moreover, the absence of long leases with predictable rents can generate ontological insecurity, anxiety, stress and possibly a reluctance to engage with and invest in the local community (Bone 2014; Pennington, Ben-Galim, and Cooke 2012; Scanlon, Fernandez, and Whitehead 2014). Rising rates of private renting amongst young families thus pose a host of challenges spanning policy domains ranging from education to health, welfare and community cohesion.
In the light of the above, this paper examines whether the growth of private renting is deepening housing inequality amongst young families as less advantaged parents are disproportionately channelled into the PRS. The study therefore responds to Cole, Powell, and Sanderson's (2016) plea for research into divisions within Generation Rent, as well as Murie and William's (2015, 672) call to develop richer sociological perspectives on disparities and inequalities within housing systems. To meet these objectives, the next section outlines the background to the study. I then describe the data and methods before presenting the results. The paper concludes with some broader reflections and by outlining a research agenda to extend our understanding of the links between families, housing and social inequality.

Background
In the post-war decades private renting declined across England and Wales as homeownership expanded, council housebuilding decommodified renting and strong regulation of rents and tenure security made the PRS unattractive to investors (Houston and Sissons 2012). However from the 1980s policy-makers became keen to revive the PRS, arguing that the ease and low costs of moving between privately rented dwellings could boost labour mobility and economic performance (Crook and Kemp 2014). In the late 1990s the PRS began to expand and this recovery accelerated with the onset of the GFC in 2007. The groundwork for this renaissance was laid by policy interventions in the 1980s and 1990s to deregulate rents and weaken tenure security (Kemp 2015). Following these reforms, economic factors such as financial innovation through Buy-to-Let mortgages, the prospect of capital gains from inflating house prices and a dearth of alternative investment opportunities helped revive the PRS, predominantly through the conversion of owner-occupied dwellings and some far more limited direct construction (for detailed reviews of PRS dynamics see Crook and Kemp 2014;Kemp 2015;Rugg and Rhodes 2008). Despite some calls to foster institutional investment in private renting, much of the recent PRS growth seems to have been driven by a further influx of the non-professional landlords owning small and often highly localized portfolios who have dominated the sector for many years (Crook and Kemp 2014).
These supply side factors can only partly explain the ongoing expansion of the PRS. Conceptualizing housing tenures sociologically as institutions comprising "socially constructed configurations of property rights and obligations" (Kemp 2015: 602) reveals how changes in life course trajectories and the contexts within which they unfold have also increased demand for private renting (Murie and Williams 2015). Although the PRS has always comprised a heterogeneous mix of niches catering to people in diverse circumstances (Rugg and Rhodes 2008), twentieth-century models of housing careers held that the PRS was particularly important early in the life course (Di Salvo and Ermisch 1997;Murphy 1984). This is because the short tenancies, absence of commitments and low transaction costs offered by the PRS enable people to move easily to adjust their residential circumstances to the life course transitions in education, work and family that often cluster into young adulthood. Conventional models held that as people's lives subsequently settled down and they began to accumulate resources, they would then exit the PRS for the "lifetime" tenures of homeownership or social renting where housing was more secure, cost-effective and perceived to be better for child rearing (Di Salvo and Ermisch 1997;Ineichen 1981;Murphy 1984;Payne and Payne 1977). Ford, Rugg, and Burrows (2002) argue that this orderly life cycle model of housing careers began to break down in the 1980s as structural changes in demography, labour markets, institutions and housing systems came together to reshape young adults' housing options and choices. Since the 1980s, the growing protraction, reversibility and precariousness of transitions to adulthood have increased young people's effective demand for flexible private rental housing, as well as their reliance on safety net accommodation in the parental home (Rugg 2010;Stone, Berrington, and Falkingham 2014). Prolonged participation in education and the burden of student debts, difficulties obtaining secure well-paid work, curtailed welfare support and delayed partnership formation have all combined to reduce young people's ability and inclination to take on the long-term commitments of homeownership or social tenancies (Clapham et al. 2014). Furthermore for some groups, private rentingoften in shared accommodationmay have become an increasingly accepted and valued lifestyle, most notably amongst affluent, highly educated young people wanting to minimize their commitments and live in opportunity-rich urban centres where housing is often costly (Rugg and Quilgars 2015). By contrast, sharing privately rented accommodation may be much less volitional and a far less positive experience for poorer young singles whose housing benefit support is capped at the local cost of renting a room (for further discussion see Beatty et al. 2014).
At the same time, the growing diversity of private tenants, a rise in long-term private renting (Crook and Kemp 2014) and continuing public preferences for owneroccupation (Jessop and Humphrey 2014;Taylor 2011) suggest that difficulties accessing other tenures are also boosting young Britons' reliance on the PRS (Clapham et al. 2014;Pennington, Ben-Galim, and Cooke 2012;Shelter 2012). This is probably especially true for less economically advantaged young people. On one side, their access to relatively secure and affordable housing in the SRS has been reduced by the stock contraction produced by limited construction and privatized commodification through the Right to Buy. Sprigings and Smith (2012) show that a sizeable proportion of council dwellings bought under Right to Buy have subsequently leaked into the PRS, where many are being let at higher rents to people who might formerly have expected to live in social housing. 4 These trends mean that in many parts of the country, the scarce social stock is now basically allocated according to needs assessments and hence is mainly accessible to only the most vulnerable individuals, for example the homeless or those living in very poor-quality accommodation. The lack of political appetite at Westminster for large-scale investment in new social housing suggests that this trend is likely to continue, at least in England (Clapham et al. 2014;Kennett, Forrest, and Marsh 2013).
On the other side, house price inflation, job insecurity and the more stringent deposit requirements imposed after the GFC are simultaneously making it harder for less affluent young people to access homeownership without significant financial support from family members (Jessop and Humphrey 2014;NHF 2014;Tatch 2007). This threatens to exacerbate well-documented inequalities in young Britons' housing careers by more deeply stratifying their trajectories and experiences by class position and family background (Coulter 2016). Lennartz and colleagues (2015) show that these issues have international resonance as young people in many European countries are finding it harder to enter homeownership. However, affordability constraints appear to have triggered a particularly strong switch to private renting in Britain, in contrast to the more prominent trend of extended parental co-residence evident in some continental countries. Given the strong British political and ideological commitment to homeownership (Ronald 2008), it is perhaps unsurprising that Westminster's main policy response to this trend has been a succession of interventions to assist younger people into owner-occupation (for example, schemes such as First-Buy, Help to Buy and Starter Homes). So far, these have had relatively modest impacts (Jones 2016) which are probably dwarfed by the negative effects of welfare reform on the housing position and options of less advantaged young tenants (Cole, Powell, and Sanderson 2016).
Public debates about Generation Rent highlight how one group of young private tenantsfamilies with children unable to access homeownership or the SRSare a particular concern for policy-makers and housing charities (Labour Party 2012; Shelter 2012). According to Shelter (2015, 3), the proportion of family households with dependent children living in the PRS rose from around 9 to 24% between 2003/2004 and 2013/2014 (an increase of approximately 1 million households). This trend is problematic because although there have been some changes in PRS institutions since the 1990s (most notably regarding mortgage finance) and some aggregate improvements in dwelling quality (Rugg and Rhodes 2008), these changes have not been matched by significant and widespread adjustment of practices to cater for a changing tenant profile (Kemp 2015). Despite housing an increasingly diverse array of households, the short-term nature of PRS contracts and the broader structure of the sector remain oriented towards providing flexible and transitional housing for mobile population groups (Shelter 2015).
Renting privately can pose extra challenges for families above and beyond the more general issues of constrained affordability and poor dwelling conditions that disproportionately afflict the PRS (NHF 2014;Shelter 2012). Foremost amongst these are the ways in which the short duration of private tenancies can create insecurity (Bone 2014). Although housing insecurity can be problematic for anyone, the expiry of short contracts or brinkmanship by landlords seeking to raise rents between leases may be especially detrimental for families if this makes it difficult for them to avoid making costly and disruptive residential moves (Bone 2014). Some authors suggest that such mobility canin some circumstances and especially if it occurs frequentlyadversely affect children's educational outcomes, socialization and health (Jelleyman and Spencer 2008;Oishi and Talhelm 2012;Scanlon and Devine 2001;Shelter 2012). Even though other studies report more equivocal findings (Gasper, DeLuca, and Estacion 2010), families certainly perceive relocating to be disruptive and typically strive to avoid school changes when they move (Shelter 2012). This desire for educational stability can lead them to accept long commutes, higher rents or a less suitable dwelling in order to keep their children in the same school (Shelter 2012).
As most private tenancies end voluntarily perceived housing insecurity and the feelings of powerlessness and passivity this can generate may be a more potent concern than unwanted residential mobility for many young families in the PRS (Crook and Kemp 2014). The normative duty of providing children with a stable home environment means that uncertainty about tenancy renewals and future rent levels can generate ontological insecurity, stress and anxiety amongst parents keen to stay put (Bone 2014;Scanlon, Fernandez, and Whitehead 2014). This uncertainty may help explain why private tenants tend to engage less with their local community and often report a weaker sense of neighbourhood attachment and belonging than their home-owning peers (Pennington, Ben-Galim, and Cooke 2012).
When taken together, these actual and perceived obstacles to establishing a stable family home in the PRS could deepen social inequality if less economically advantaged young families are becoming increasingly likely to rely on privately rented accommodation vis-à-vis their more prosperous peers. Prior research suggests that two interlinked attributes may be particularly relevant markers of socio-economic (dis)advantage amongst young families. First, family structure is likely to matter as lone mothers (who make up the bulk of single-parent families) are disproportionately likely to be in poverty or have part-time, insecure and lower paid jobs (Berrington 2014;Bone 2014). Second, a long research tradition has shown that labour force participation and occupational class position strongly affect young people's access to resources and hence both their ability to enter owner-occupation and the de facto priority they receive in social housing allocations (Ermisch and Halpin 2004). The rest of this paper therefore concentrates on how these two variables have shaped the housing position of two cohorts of young parents with co-resident children.

Data and Sample
This study uses data from the Office for National Statistics Longitudinal Study of England and Wales (LS). The LS is a relational database containing the linked decennial census records of a 1% sample of the population of England and Wales. The original sample was drawn from the 1971 census by extracting data from all individuals born on one of the four selected dates of the year (Lynch et al. 2015). New babies and immigrants with these birth dates are continuously added to the sample so that it remains representative. Census data from individuals living with Longitudinal Study Members (LSMs) are included in the LS, although these people are not followed through time.
The LS has several advantages for this project. The large sample, high linkage rate and low levels of census non-response allow robust analysis of population subgroups while reducing the problems of selective participation that often bedevil longitudinal surveys (Lynch et al. 2015). The LS also offers rich detail on the long-term development of family and housing trajectories. As the LS is a sample of individuals enriched with relational data, it further allows us to identify young people living in "concealed" families which do not contain the household reference person. This individual-level focus is important because young adults' household formation patterns and housing tenure position can both change over time, for example if tight housing markets make it harder for young people to live outside the parental home. These interlinked changes in tenure and living arrangements are difficult to disentangle using surveys where the sampling frame and target population comprise households rather than individuals (Rugg and Quilgars 2015). Finally, the long period covered by the LS enables studies to examine and control for the ways in which young people's family and housing trajectories are shaped by their childhood circumstances and parental attributes.
The sample for this study comprised all LSMs who were (jointly) heading a family as either (a) a lone parent or (b) as a member of a couple with their identified child(ren) in the 2001 or 2011 census. Families are coded during census processing using information from the household relationship grid included on the census form. In 2001, a family was defined as "a group of people consisting of a married or cohabiting couple with or without child(ren), or a lone parent with child(ren)" (Office for National Statistics, General Register Office for Scotland, Northern Ireland Statistics and Research Agency 2004, 33). Children could be of any age and need not belong to both partners in couples. Although each family can only contain two generations, families could also comprise "a married or cohabiting couple with their grandchild(ren) or a lone grandparent with his or her grandchild(ren) where there are no children in the intervening generation in the household" (Office for National Statistics, General Register Office for Scotland, Northern Ireland Statistics and Research Agency 2004, 33). 5 This definition of a family remained essentially unchanged for the 2011 census, barring the addition of same-sex civil partnerships (legalized in 2005) to the typology of couple families (ONS 2014). Crucially, the census definition of families allows the identification of LSMs living with their children in complex or multigenerational households.
The sample was then restricted to LSMs aged 25-34 in order to provide a common frame of reference to compare patterns in 2001 with 2011. 6 Most people have left full-time education and entered the labour force by 25 and these are the prime years for family formation and entering homeownership. Moreover, those in their late 20s and early 30s are the group perhaps most closely associated with popular narratives of Generation Rent (Cole, Powell, and Sanderson 2016). Although the study concentrates primarily on those aged 25-34, in places comparisons are drawn with young adults aged 16-24 or 35-44 who were heading families in 2001 and 2011. In total the sample contains 59690 parents aged 25-34 (30913 in 2001 and 28777 in 2011). Women are over-represented because few young men are lone parents and because women tend to form partnerships and have children with slightly older men.

Measures and Methods
The dependent variable is a categorical indicator of housing circumstances in 2001 or 2011 when LSMs were aged 25-34. This variable uses information on intrahousehold relationships and housing tenure to code sample members as either a homeowner, social tenant, private tenant or living in the parental home. Although the census does not allow us to unequivocally tell "who lives with who" in a multigenerational household, it seems reasonable to assume that the majority of individuals aged 25-34 live with their parent(s).
The main independent variable of interest classifies sample members' socioeconomic position into 11 categories by combining information on their family type, labour force participation and occupational status. Family type divides lone parents from individuals jointly heading couple families with children. Employment participation categorizes individuals and their partner (if applicable) as either employed or not employed, with full-time students coded as not in employment. This makes it possible to classify lone parents by whether or not they are in paid employment and allows couples to be subdivided into dual-and single-earner partnerships.
For employed individuals and partners (if applicable), the National Statistics Socio-Economic Classification (NS-SEC) of their job was then used to code occupational status. NS-SEC classifies occupations according to employment relations and conditions. To avoid small cell counts the three-category version of NS-SEC was used. This disaggregates those working in higher managerial, administrative and professional roles (NS-SEC 1-2) from those with intermediate (NS-SEC 3-4) or routine and manual occupations (NS-SEC 5-7). Following standard practice, the "higher" status job was used to code the occupational position of dual-earner couples.
The analysis begins by comparing the housing circumstances of young parents with particular demographic and socio-economic characteristics in 2001 with their peers in 2011. A multinomial logistic regression model is then fitted to examine whether these temporal trends in housing circumstances persist after controlling for a range of other factors known to shape housing careers such as ethnicity, health, educational qualifications and region (Long and Freese 2006). Descriptive statistics for all variables in the model are provided in Appendix Table A1.
The sample for the multinomial model is restricted to only those 38582 (71%) sample members enumerated at the preceding two censuses who were also living with a parent when aged 5-14 at their t − 2 census (1981 and 1991 for those aged 25-34 in 2001 and 2011 respectively). This restriction allows controls to be included for spatial mobility between censuses and parental attributes measured when LSMs were aged 5-14. Prior LS research shows that these are important factors in young adults' housing careers as migrants often rely on the PRS, while parental occupational advantage and homeownership increase the odds that young people enter owner-occupation (Coulter 2016). As a result of this additional restriction, the population of interest shifts subtly for the modelling work to encompass only those young adults in families in 2001 and 2011 who have also been residents in England and Wales for three consecutive censuses. Recent immigrants are thus included in the descriptive analysis but excluded from the modelling sample.

Patterns and Trends
To contextualize the main analyses, Table 1 shows how the percentage of sample members aged 25-34 in 2011 living in each housing situation varied with the demographic and economic attributes of their family (henceforth "family type"). The table shows that over one-third of lone parents and one-quarter of individuals heading couple families rented privately in 2011. On average, lone parents were much more likely to be social tenants and far less likely to be homeowners than individuals heading couple families. While around 9% of lone parents lived with a parent, this arrangement was unusual for individuals living with a partner and child(ren). Table 1 shows that the percentage of lone parents aged 25-34 who were living in the PRS in 2011 did not vary greatly by labour force participation or occupational class position. However, lone parents working in higher managerial, administrative and professional occupations (NS-SEC 1-2) were considerably more likely to be homeowners and less likely to be social tenants than their peers with routine and manual jobs (NS-SEC 5-7) or without paid employment. By contrast, the percentage of young adults heading couple families who were living in the PRS was much more stratified by economic position, largely due to differential rates of homeownership. Private tenancies were least common amongst individuals in dual-earner couples with higher (NS-SEC 1-2) occupations (17.4%) and most common amongst those living in couples where neither partner had paid employment (36.9%). Interestingly, Table 1 shows a clear occupational class gradient in the rate of private renting  amongst individuals heading dual-but not single-earning couples. This appears to be because the class gradation of homeownership (social tenancies) is less (more) pronounced amongst individuals in single-earning couples as compared with those in dual-earning partnerships. 7 Table 2 Table 2 shows that there is much truth in Generation Rent narratives as young people heading all types of families in 2011 were considerably more likely to live in the PRS than their peers in 2001. Amongst lone parents, this trend offset a large absolute decline in social renting (−11.8 percentage points) and a smaller absolute reduction in owner-occupation (−5.7 percentage points). By contrast, between 2001 and 2011 there was little change in the proportion of individuals heading couple families living in the SRS. For this group the growth of private renting was almost entirely counterbalanced by falling owner-occupation (−16.8 percentage points). Despite strong relative growth from a low base, in absolute terms there was only a small increase in parental co-residence between 2001 and 2011, most notably amongst lone parents. Table 2 shows that between 2001 and 2011, private renting expanded particularly dramatically for some types of young family heads. For lone parents, the most rapid absolute and relative growth in private renting was amongst those with more advantaged jobs, who were much less likely to be homeowners in 2011 than 2001. By contrast, the fall in social renting amongst less economically advantaged lone parents has been slightly slower and made for a somewhat more muted increase in private renting. Amongst young adults heading couple families, the largest absolute and relative percentage increases in private renting between 2001 and 2011 were for those employed in routine and manual occupations (NS-SEC 5-7). For example, the percentage of individuals in higher managerial, administrative and professional (NS-SEC 1-2) dual-earner couples who were living in the PRS rose by 11.6 percentage points (nearly 200%) between 2001 and 2011. However, this was dwarfed by the 26.0 percentage point (435%) increase in private renting amongst dual-earner couples where both partners were employed in routine and manual jobs. This deepening class stratification of private renting is linked to trends in the owner-occupied sector where homeownership has declined most rapidly amongst individuals heading single-earner couple families and families where the partners have less advantaged occupations. 9 Overall, these patterns suggest that the growing difficulty of accessing homeownership is combining with persistently constrained access to social housing to deepen class disparities in which young adults raise children in the PRS.

Modelling Results
These bivariate results provide only preliminary insights because unobserved differences between young adults heading families in the two census years may also influence their housing careers. In consequence, Table 3 presents the results of a well-fitting multinomial logistic regression model which predicts the housing position of sample members while controlling for a range of additional factors known to be associated with family and housing trajectories (Appendix Table A1 for details). To address the main research question, the model includes a dummy for 2011 (reference category = 2001), a categorical indicator of family type and a series of   Table  A2). Robust standard errors. * = p < 0.05; ** = p < 0.01; *** = p < 0.001.
terms to capture interactions between these two variables. For brevity, only the effects of these three independent variables are presented for interpretation in Table 3. All control variable parameters are however broadly in line with expectations and are shown in Appendix Table A2. The reference category for the multinomial model is private tenancy and all results are reported as relative risk ratios (RRR). RRRs greater than 1 indicate that a one unit increase on that variable raises the risk of a specified outcome relative to the risk of private tenancy, whereas RRRs less than 1 indicate the reverse. This means that the results in Table 3 should be interpreted as multiplicative effects (Buis 2010). Table 3 validates Generation Rent narratives by showing that the relative risk that young family heads lived in owner-occupation, social tenancies or the parental home were all lower in 2011 than in 2001. As the model interacts census year with family type, these parameters refer specifically to the reference group of individuals in dualearner couples with higher (NS-SEC 1-2) occupations (Buis 2010). Turning to the RRRs for owner-occupation, we see that the relative risk of being in homeownership as opposed to a private tenancy was substantially and significantly lower for individuals in all family types as compared with those in dual-earner NS-SEC 1-2 partnerships. As expected from Tables 1 and 2, lone parents and those heading workless families have particularly low relative risks of homeownership. There is also a significant negative interaction effect for individuals living in both dual-and singleearner NS-SEC 5-7 couples. This suggests that the growing financial difficulty of accessing homeownership between 2001 and 2011 may have disproportionately impacted upon less occupationally advantaged young couples and increasingly channelled them into the PRS.
Unsurprisingly, the RRRs for social tenancy are largely the inverse of those for owner-occupation. Reflecting the SRS's "safety net" function, the risk of a social relative to private tenancy is significantly higher for lone parents and those individuals in couples with less advantaged jobs than for those living in dual-earner NS-SEC 1-2 families. The interaction terms indicate that this pattern seems to have strengthened between 2001 and 2011 for young adults heading workless families and some types of single-earner couples. In contrast, the results in the final columns of Table 3 indicate that lone parents had a considerably higher relative risk of living in the parental home than individuals in dual-earner NS-SEC 1-2 couples. Mirroring the results in Table 2, the interaction terms suggest that less economically advantaged lone parents may also have become relatively more reliant on the parental home over time. Overall, the model estimates suggest that the growing stratification of young families' chances of living in the homeownership and private rental sectors cannot be completely explained by observable differences between young people in different family types.
To glean a richer understanding of these estimates, it is useful to also consider the modelling results in terms of probability using marginal effects. According to Williams (2012, 323), marginal effects for a categorical variable in a non-linear model show "how P(Y = 1) changes as the categorical variable changes from 0 to 1, after controlling in some way for the other variables in the model". As there are four possible housing outcomes in the multinomial model, each variable has four marginal effects (one per outcome). Table 4 estimates the marginal effect of the 2011 dummy on the probability of renting privately for sample members in different family types, holding all other independent variables at their observed values. 10 Comparing these adjusted results with the bivariate percentage point changes in Table 2 suggests that only a relatively small proportion of the percentage point inter-censual trends are probably due to the attributes of young people living in each family type. After adjusting for compositional factors, Table 4 shows that the probability that young individuals heading all types of family lived in the PRS was between 10 and 18 percentage points higher in 2011 as compared with 2001. Amongst lone parentsthe strongest marginal effect is for those working in higher (NS-SEC 1-2) occupations (0.180), while there is a slightly weaker period effect for lone parents without paid employment (0.122) and those working in routine and manual (NS-SEC 5-7) occupations (0.142).
Concerns that private renting has increased most dramatically amongst less advantaged young couples are borne out by Table 4 as the marginal effect of the 2011 dummy is weakest for individuals heading dual-earner families with NS-SEC 1-2 occupations (0.102). By contrast but in line with Table 2, the marginal effect is stronger for individuals living in single-earner couples or couples with NS-SEC 5-7 occupations. Crucially, the 95% confidence intervals for individuals heading NS-SEC 5-7 couple families do not overlap with the confidence interval for dualearner NS-SEC 1-2 couples. This indicates that couples with a less advantaged occupational class position became significantly more dependent on the PRS between 2001 and 2011 than their peers with more advantaged jobs. Examining the marginal effect of the 2011 dummy on the other categories of the dependent variable (not presented) shows that this pattern is counterbalanced by a particularly strong fall in the probability of homeownership amongst young adults heading couples with less advantaged occupations. Figure 1 builds on these results by plotting the predicted probability that a young adult heading different types of family in 2001 and 2011 was living in each housing position. These predictions are derived for a hypothetical sample member of mean age with modal values on all categorical attributes (Appendix Table A1 for details). The lower panel shows that the predicted probability of private tenancy increased significantly across all family types between 2001 and 2011. Amongst lone parents this led to socio-economic convergence as growth in the probability of private renting was particularly pronounced amongst those lone parents with more advantaged occupations. By contrast, for young adults heading couple families the probability of private renting was considerably more stratified by labour force and  Figure 1 reiterates that these trends are linked to changes in access to homeownership rather than the social rental sector. Across all family types, the predicted probability of social tenancy does not vary a lot between 2001 and 2011. Indeed, for individuals heading less occupationally advantaged couple families, the predicted probability of renting socially is actually higher in 2011 than in 2001. By contrast, the predicted probability of homeownership fell most dramatically across time amongst young single-earner couple families and those couple families with a routine and manual occupational class (NS-SEC 5-7). Amongst lone parents the biggest decline was for those working in more advantaged occupations, largely because few lone parents without a job or working in NS-SEC 5-7 were homeowners in 2001. Overall, these results indicate that the social stratification of young families in the PRS has in general increased over time. This appears to be primarily because homeownership amongst young adult families is becoming less socially inclusive as it is increasingly confined to only those in dual-earning couples with advantaged jobs.

Discussion and Conclusions
In many Western societies the increased diversity, fluidity and uncertainty of housing careers is a key reason why transitions to adulthood have become more complex, protracted and precarious in recent decades. Following the GFC, fewer young Europeans are entering homeownership and more are either renting or living in the parental home (Lennartz, Arundel, and Ronald 2015). In Britain, the shift to private renting has been especially pronounced and policy-makers have responded with measures to assist young people into owner-occupation. These interventions are at least partly motivated by fears that young adults' deepening reliance on the PRS is helping polarize society as older generations stockpile housing and wealth, while only an increasingly advantaged tranche of young people are able to become homeowners (Pennington, Ben-Galim, and Cooke 2012;Ronald, Kadi, and Lennartz 2015). Moreover, growing dependence on private rental housing is thought to present a particular threat to the current and future welfare of young families as the PRS caters relatively poorly for longer term tenants seeking control, security and stability (Shelter 2012). In the light of these concerns, this paper has examined whether young adults heading less advantaged young families were disproportionately channelled into the expanding PRS between 2001 and 2011.
Two principal conclusions can be drawn from the results. Although young adults heading all types of families became considerably more reliant on the PRS between 2001 and 2011, this trend was more pronounced for lone parents and couples with routine and manual occupations than for dual-earning couples with "salariat" jobs. It is likely that this is due to strengthening constraints rather than changing preferences as relatively few families choose to rent privately and the ease of moving within the PRS is likely to be most useful for mobile young professionals operating in large labour markets (Pennington, Ben-Galim, and Cooke 2012;Shelter 2012). Overall, the growing social stratification of private renting amongst young adults heading families indicates that housing debates need to move beyond Generation Rent narratives of intergenerational conflict to also consider housing inequalities between young people (McKee 2012). Doing this by extending this paper's focus on class disparities is crucial if we are to understand and explain cross-national patterns of housing inequality (Dewilde and De Decker 2016;Murie and Williams 2015).
A second conclusion is that changing patterns of private renting amongst young families are primarily linked to trends in homeownership. Between 2001 and 2011, there was only a small rise in the proportion of young family heads living with a parent and some fairly minor shifts in rates of social renting. By contrast, homeownership amongst family heads aged 25-34 became considerably more stratified by family type, labour force participation and most notably class between 2001 and 2011. Young adults heading dual-earner families with advantaged jobs experienced the smallest fall in homeownership, while there was a much more precipitous decline amongst lone parents and couples with routine and manual occupations.
It seems probable that this is because affordability problems and stringent credit requirements are increasingly excluding young families with fewer resources from entering more marginal forms of homeownership early in life. This trend suggests that the commodification and financialization of housing systemsa field in which the UK excelsnot only generate unequal exposure to housing risk (Kennett, Forrest, and Marsh 2013), but can also qualitatively change the types of risks facing poorer households as they respond to economic constraints by adapting their tenure and living arrangements. However, as yet, we cannot disentangle whether the observed trends are due to shifts in the timing and/or the eventual occurrence of homeownership transitions. On the one hand it may be that more constrained young families are simply taking an increasingly long time to become homeowners relative to their more advantaged peers. On the other, constrained housing supply and the transfer of owned stock into an increasingly competitive PRS may impede less affluent young people from ever accumulating sufficient resources or having the opportunity to buy a dwelling. Using panel data to probe this and test whether social inequalities in housing are changing differently in countries where young people have become increasingly reliant on the parental home rather than the PRS are key future research priorities. Such work could yield insights about which aspects of the institutional and welfare context provide the most favourable conditions for young people from across the social spectrum to leave home and enter the housing system.
Deepening disparity in which young families rely on the PRS has implications for policy and social justice, regardless of whether or not these patterns "wear off" as people age. In the short term, the pressures generated by problems of dwelling quality, affordability and insecurity within the British PRS are likely to disproportionately fall on those with the fewest resources living towards the lower end of the market. Crucially, the marginalization of less affluent private tenants is likely to have intensified since 2011 as stagnant wages and reduced welfare support channel them into less desirable dwellings and locations (Powell 2015). Landlords may also have responded to housing benefit cuts by adopting harsher lettings practices (for example, by taking a more robust stance on arrears or by refusing to let to young people or benefit claimants), or by reducing maintenance expenditure (Beatty et al. 2014;Cole, Powell, and Sanderson 2016). The lack of central government interest in expanding social rental or reforming the PRS means that these processes are likely to negatively affect the housing options and welfare of less advantaged young families for the foreseeable future (Clapham et al. 2014).
In the longer term, less advantaged young parents may also find it increasingly difficult to meet parenting norms stressing the importance of a stable, controllable home environment and the detrimental impacts of residential mobility on child development (Shelter 2012). Parenting difficulties are likely to be especially acute for young non-resident parents constrained to share PRS accommodation by the Shared Accommodation Rate (Cole, Powell, and Sanderson 2016). Moreover, increasingly unequal access to homeownership early in the life course is likely to exacerbate future inequalities as the growing influence of familial transfers and class position on transitions to owner-occupation cements the advantages that children from more privileged backgrounds already enjoy in education and the labour market. This raises serious questions about the wisdom and long-term viability of building housing wealth into welfare strategies (Ronald, Kadi, and Lennartz 2015).
Finally, the results signpost new conceptual and analytic directions for researchers concerned about housing inequality. Although the housing pathways metaphor is enriching scholarship (Clapham et al. 2014), it arguably lacks the theoretical tools to explain how and why housing trajectories are structured in particular kinds of ways which are interlinked with other life domains and the lives of other people. This makes it difficult to understand the vectors and mechanisms through which policies, institutions and structural factors selectively affect housing experiences, for example via the interactions between landlords and tenants. By contrast, the life course perspective guides us to think of human development in terms of multiple intersecting careers which are shaped by relationally "linked lives" and broader structural forces that vary across time and space (Coulter, van Ham, and Findlay 2016). In this perspective, housing decisions and transitions are not unbounded choices but are shaped by micro-level resources and restrictions, as well as macro-scale opportunities and constraints. As McKee (2012) has observed, these insights mean that the life course framework holds much potential for linking analysis of housing experiences to patterns of social and institutional change.
The study also indicates several avenues for future empirical work. First, multilevel analysis of how young adults' housing transitions vary across space and time could help disentangle which contextual factors are most potent for which groups of young people (for example trends in house prices, incomes or dwelling stock composition). Longitudinal studies could also yield new insights about changes in the relative timing of family and housing transitions, changing subjective experiences of renting and owning and reverse causalities in the association between young people's labour force and housing positions. It would also be valuable to decentre tenure by exploring changing patterns of housing affordability, conditions, security and dwelling suitability within as well as between tenure groups. Overall, there is much to do if we are to better understand and address housing inequalities across the life course.

Acknowledgements
The permission of the Office for National Statistics to use the LS is gratefully acknowledged, as is the invaluable help provided by staff of the Centre for Longitudinal Study Information and User Support (CeLSIUS). CeLSIUS is supported by the ESRC Census of Population Programme [ES/K000365/1]. The author bears sole responsibility for all analyses and interpretations of the data. This work contains statistical data from ONS which is Crown Copyright. The use of the ONS statistical data in this work does not imply endorsement of the ONS in relation to the interpretation or analysis of the statistical data. This work uses research data-sets which may not exactly reproduce National Statistics aggregates. The derivation of NS-SEC classes is provided in Bukodi and Neuburger (2009) "Data Note. Job and occupational histories for the NSHD 1946 Birth Cohort" as part of the ESRC Gender Network Grant, Project 1 'Changing occupational careers of men and women' (RES-225-25-2001

1.
The British rental system is usually portrayed as a dualist model split into distinct social and private sectors which do not directly compete with one another. Housing in the private rental sector is predominantly allocated by the market, although a minority of tenants live rent free or in accommodation tied to employment. Leases can be any length but most landlords offer initial (but renewable) 6-or 12-month fixed term contracts after which the tenancy can be terminated with little notice by either party. By contrast, social rental housing is bureaucratically allocated and leased at below market rents from regulated providers. As demand for social housing generally outstrips supply, local governments operate waiting lists, with priority usually assigned to more vulnerable applicants.

2.
The 2013/2014 English Housing Survey (DCLG 2015, 34-36) shows that private tenants have higher average weekly housing costs (£176) than mortgage holders (£153) or social tenants (<£100). After taking housing benefit support into account, private rents on average also consume a larger proportion of gross household incomes (34%) than mortgages (18%) or social rents (29%). In spite of these cost differentials, the proportion of dwellings failing to meet basic quality standards is highest in the PRS (Rugg and Rhodes 2008;Shelter 2012). 3.
One particularly important reform has involved changing the Local Housing Allowances payable to poorer private tenants (Beatty et al. 2014). Young people have been especially affected by the decision to raise the age threshold below which benefit support is limited to the local costs of renting a room (the Shared Accommodation Rate, which applies to single young adults without co-resident dependent children). This threshold was lifted from 25 to 35 in 2012 (Beatty et al. 2014). As welfare reform was partly designed to reduce public expenditure, it is unsurprising that these changes have resulted in a marked but geographically varied reduction in tenants' benefit entitlements and purchasing power (Powell 2015). Work by Cole, Powell, and Sanderson (2016) indicates that this may be exacerbating class and generational fractures within the PRS as younger and less affluent tenants are increasingly marginalized, excluded and stigmatized by older and better off landlords. 4. Sprigings and Smith (2012) argue that this process has greatly inflated the long-term revenue costs of Right to Buy. They suggest that any capital gains made from stock sales may be outweighed in the long run by the higher welfare cost of supporting tenants in private as opposed to social rental accommodation. 5.
These types of families are of little relevance in this paper as sample members had to be enumerated as the parent of at least one co-resident child. 6.
Focusing on those aged 25-34 also allows variables capturing parental attributes when sample members were aged 5-14 and living in the parental home to be included in models of housing outcomes (see measures and methods section). 7.
Comparing these patterns with those for individuals aged 16-24 and 35-44 yields additional insights (results not shown). In 2011, rates of private renting were higher amongst the youngest group and there was considerably less variation by family type (range 36.7-47.8%). Unsurprisingly, homeownership was comparatively rare for all individuals aged 16-24, while far more of the youngest lone parents lived with a parent (21.0%). By contrast, the socio-economic gradation in the rate of private renting shown in Table 1 is visible but somewhat muted for those aged 35-44. Although interesting, it is risky to use these results to speculate about future trends or life course patterns of housing inequality as it is not possible to disentangle age, period and cohort effects. 8.
Absolute percentage point change is the row percentage of a cell in 2011 minus its 2001 row percentage. Relative change is this absolute change expressed as a percentage of the original 2001 cell value. While absolute change captures the overall magnitude of inter-censual trends, relative change provides a way to compare the importance of absolute percentage point shifts across groups with different baseline values. By way of an example, the total percentage of sample members in private tenancies was 11.1 in 2001 and 27.6 in 2011 (compare Tables 1 and 2). This translates into an absolute percentage point change of +16.5 (27.6-11.1). Making allowance for rounding this absolute change represents a relative percentage shift of +149.5% ((27.6-11.1)/11.1) *100. 9.
Further analysis (not shown) reveals that these broad patterns were also evident amongst individuals aged 35-44. Amongst the youngest age group (16-24), the increase in private renting between 2001 and 2011 was a more secular trend. 10.
Fixing covariates at sample means yields very similar estimates.