Comparing regional patterns of homeownership entry across four British birth cohorts

Abstract Difficulties accessing homeownership and reduced rates of owner-occupation among recent birth cohorts are a major concern for Global North policymakers. However, surprisingly little is known about how patterns of entry into homeownership have varied spatially across the early lives of recent birth cohorts. Using life course perspectives and survey data, this study examines how regional disparities in homeownership trajectories and transitions have varied across the life courses of four birth cohorts who entered the British housing system after 1990. The results show a nonlinear pattern of postponed homeownership across cohorts which has not varied greatly across regions. London is the most distinctive area and delayed homeownership transitions have long been a feature of the capital’s housing market. Taken together, the findings illustrate the value of more thoroughly examining how place intersects with biographical and historical time in nuanced ways to shape housing careers.


Introduction
In many countries, buying one's first home is widely viewed as a significant life course milestone that enhances social status, security, housing quality and provides opportunities for wealth accumulation.Yet many governments are now struggling to deliver on their promises of mass homeownership as younger cohorts enter the tenure more slowly than previous generations (Arundel & Ronald, 2021).These issues are especially prominent in Britain, where the owner-occupied sector contracted from the mid-2000s before experiencing a market slump after the 2008 Global Financial Crisis (GFC) followed by a spatially polarised rebound (Hamnett & Reades, 2019).Successive Westminster governments have viewed this volatility and declining rates of young adult owner-occupation with concern and used multiple policy levers to try and bolster homeownership access.These interventions range from the post-2013 Help to Buy package of financial support measures for first-time buyers (including equity loans, mortgage guarantees and savings products paying a government bonus at home purchase) through to wider efforts to increase housing supply and reinvigorate discounted Right to Buy sales of social housing.
Underpinning these interventions is an awareness that the restructuring of housing systems is bound up with changes in the structure of people's lives.This knowledge has inspired two strands of research using life course perspectives.One strand examines how housing careers -defined as the succession of residential states passed through over the adult life course -are intertwined with people's non-linear 'careers' in other domains of life such as education, employment, health, and family (Coulter et al., 2020).Much of this work concentrates on how institutionally mediated shifts in the types and timing of other life course processes have consequences for homeownership entry.Trends towards later partnership and family formation, prolonged debt-funded education, the deterioration of economic security and incomes plus the concentration of highly educated young adults into expensive urban housing markets have all dampened recent cohorts' desire and ability to make early housing purchases (Choi et al., 2019;Clapham et al., 2014;Sissons & Houston, 2019).
A second strand of research focuses on how strengthening housing access and affordability constraints are making it more financially challenging for younger cohorts to enter homeownership.This literature argues that delayed or reduced entry into owner-occupation is increasing intergenerational disparities in housing costs, quality, security, space and wealth (Corlett & Judge, 2017;Green, 2017).In this more constrained context, intergenerational family support -provided in forms ranging from subsidised co-residence and financial transfers through to intangible housing know-how -is becoming a more powerful factor in younger cohorts' housing careers (Suh, 2020).These issues are probably most pronounced in more expensive housing markets, although McKee et al. (2017) note that surprisingly little is known about the geographical patterning of tenure transitions through early adulthood.
This article bridges and develops these two research traditions by examining how homeownership entry in Britain has varied over time and space since the 1990s.More specifically, the paper extends previous literature by moving beyond national-level comparisons of generational housing trajectories to instead focus on regional patterns of homeownership entry across the early lives of four tightly defined birth cohorts.We focus on Britain as this is a classic liberal homeowner society where the tenure is culturally valued, supported by policy, the mortgage market is developed and rental alternatives are either scarce (social renting) or do not provide the security and control people need over much of the life course (private renting).This context means that any difficulties accessing homeownership and delayed entry to the tenure may have damaging impacts on people's quality of life, subjective well-being and may potentially lead to the postponement of other transitions such as family formation (McKee et al., 2017).There are also stark disparities in regional housing markets across Britain which, together with the fragmentation of housing policy across the UK's devolved regions, creates the conditions for substantial subnational geographic patterning of housing careers.

Interconnected careers
The life course perspective provides a powerful framework for conceptualising how housing careers are intertwined with the dynamics of people's lives in ways that are shaped by the wider housing, economic and social context.The framework conceptualises lives as comprised of a series of dynamic and non-linear parallel careers in domains such as education, employment, health, family and housing.These careers are interconnected as events in one can affect the course of others.Such events are termed 'turning points' if they alter the trajectory of a person's life by generating new roles and responsibilities (Stone et al., 2014).
Life course perspectives are helpful for understanding how patterns of homeownership entry vary across space, time and social groups.In Anglophone societies, becoming a homeowner has long been institutionalised as a culturally significant housing career milestone and a major life transition that enhances social status (Ronald, 2008).In these contexts, entering homeownership is a life course turning point that involves assuming new roles and responsibilities, for example for managing a large mortgage debt (Smith, 2015).People moving into owner-occupation from the deregulated private rental market also typically gain far greater control of their dwelling's use, décor, upkeep and disposal.
Economic motives help explain why most people in Anglophone nations want to become homeowners (Preece et al., 2020).In Britain and many other countries, decades of real house price inflation coupled with low interest rates since the GFC and favourable fiscal policies (for instance the absence of capital gains tax on owner-occupied property) have fuelled a popular belief that homeownership is economically beneficial and a sound investment (Arundel & Ronald, 2021).This narrative has been reinforced by policy discourse, the property industry and the popular media.Meanwhile, cuts to collective welfare provision have added urgency to homeownership entry as housing equity is increasingly seen as an essential resource that households may need to draw on in old age or to get through hard times (Smith, 2015).
Successive UK governments have further stoked demand for owner-occupation through housing policies.In part, this has involved directly promoting homeownership through initiatives including discounted Right to Buy sales of council housing and, more recently, the Help to Buy scheme.At the same time, rental alternatives have been eroded since the 1980s as the social housing sector has contracted while the deregulation of private renting has created a sector characterised by short insecure contracts, uncontrolled rents, limited use rights and problems of poor housing quality (Kemp, 2015).Thus, for many households neither social nor private renting constitutes an accessible and desirable long-term alternative to owner-occupation (McKee et al., 2017).
Yet despite strong demand for homeownership, the proportion of young Britons who own their dwelling has collapsed since the early 1990s (Andrew, 2012).This is partly because the restructuring of other life course careers has reduced successive cohorts' resources and homeownership preferences during young adulthood.On the economic side, weak income growth, delayed labour force entry due to prolonged debt-funded education, a shift to less stable forms of work and welfare retrenchment have all reduced the proportion of younger adults who have the resources and economic security to enter homeownership (Andrew, 2012;Clapham et al., 2014;Hoolachan & McKee, 2019).Highly educated younger adults have in the last two decades also gravitated towards the central neighbourhoods of large cities where housing is expensive and there are limited opportunities to enter homeownership (Choi et al., 2019).
Meanwhile, the demographic events that traditionally triggered strong preferences for owner-occupied housing are also being postponed.Delayed partnership formation leads to postponed ownership not just because two incomes are increasingly needed to finance housing purchases, but also because partnership and family formation are often synchronised with homeownership as couples seek the security, dwelling quality, space and stability the tenure provides when settling down and preparing to form families (Mulder, 2013).Crucially, Mulder (2013) notes that there is not a one-way relationship between family events and homeownership as housing opportunities also impact on demographic decision-making.In Britain, Tocchioni et al. (2021)'s work indicates that the mounting costs of owner-occupation may be reducing fertility among homeowners.
The impact of life course restructuring on homeownership entries is further complicated by its simultaneous effects on household formation.In many contexts, delayed homebuying has been accompanied by prolonged parental co-residence as economic insecurity and the postponement of family events mean leaving home becomes a more drawn out and reversible process (Stone et al., 2014).These twin trends have a complex bidirectional relationship akin to that connecting homeownership with family formation.On one hand, Clapham et al. (2014) document a large 'stay home to own' housing pathway where young people delay household formation in order to save up to directly transition into homeownership (Suh, 2020).They argue this pathway is becoming more common as house price inflation outstrips wage growth.On the other hand, if young people prioritise independence, then delayed home-leaving could simply push back the age at which households form in the private rented sector and cause a knock-on delay in home purchases.In the US, Myers et al. (2005) report that household formation rates dampen ownership levels among younger adults, while increased numbers of young families in the UK private rented sector suggests that many people are unwilling to put their family lives on hold in order to stay home to save up for homeownership (Tocchioni et al., 2021).

Opportunities and constraints
Life course perspectives highlight how housing careers are also influenced by contextual opportunities and constraints.Shared patterns of cohort experiences thus emerge as people born in a given period experience shock events and distinct constellations of contextual circumstances at particular ages (Feijten, 2005).In recent years, research in this area has focused on comparing the 'lucky' Baby Boomers born in the two post-war decades with the 'unlucky' millennials born after 1980.This literature has shown that millennials are moving more slowly into homeownership with age and that this delay may be having adverse impacts on their prosperity, housing quality, subjective well-being and sense of self-worth (Choi et al., 2019;Clark, 2019;Corlett & Judge, 2017;Green, 2017).Entry into homeownership among older millennials was also significantly disrupted by the GFC, which disproportionately hit the incomes and economic security of younger adults and housing market 'outsiders' (Clark, 2019;Lennartz et al., 2016).
Empirical evidence suggests that compositional changes in the demographic and socio-economic characteristics of successive cohorts can only partly explain why millennials are moving slowly into owner-occupation (Choi et al., 2019;Corlett & Judge, 2017).Another important explanation is that the contextual constraints impeding first-time buyers have strengthened since the 1990s (Sissons & Houston, 2019).House prices have inflated considerably over the last thirty years and risen faster than wages, but before the GFC easy access to credit and an increased use of family support allowed younger adults to overcome affordability barriers.This came to a halt in the aftermath of the GFC as mortgage lenders required much larger deposits.Over recent years this downpayment constraint has fallen back again as lending has relaxed and the government's Help to Buy scheme has enabled access to ever larger mortgage borrowing.
Contextual opportunities and constraints for homeownership entry vary across space as well as over time (Clark, 2019).Recent house price inflation in Britain has been spatially polarised with prices in London and the southern English regions increasing more rapidly than those in other areas (Hamnett & Reades, 2019; Figure A1).In London, median house prices increased from 6.9 times median earnings in 2002 to 12.8 times in 2019 while median prices in the South East and South West went from around 6 times median salaries in 2002 to 8-9 times by 2019 (Office for National Statistics (ONS), 2020).While housing prices have been consistently lower in more peripheral regions, crude affordability here has also worsened with median price to earnings ratios rising from 3-4 to over 5 between 2002 and 2019 across Wales and northern England (Office for National Statistics (ONS), 2020).
In addition to regional house prices, differences in the size of the owner-occupied sector and the number of vacancies create spatial patterns of opportunity to enter homeownership.London stands out here as the capital has a much smaller proportion of owner-occupied stock (under 50% of households in the 2011 census) than other regions, where owner-occupation rates typically exceed 60%.With this context in mind, we now turn to examine how regional disparities in homeownership transitions have varied across the early lives of four birth cohorts who entered the UK housing system at different points after 1990.

Datasets and cohort definition
Two nationally representative surveys were used in this study.First, Labour Force Survey (LFS) data were used to track the homeownership rates of successive birth cohorts as they aged.The LFS collects data from around 35,000 households each quarter and we used individual data from autumn sweeps conducted after 1996 when the questionnaire asked which household members owned or rented the dwelling.All LFS analyses were weighted using supplied variables.
While repeated cross-sectional data are ideal for tracking cohorts' aggregate homeownership trajectories, such data cannot tell us about the individual-level transitions and processes that are driving these patterns.For this we turned to the 28 waves of longitudinal data collected by the British Household Panel Survey (BHPS) and its successor, Understanding Society.The BHPS began in 1991 when interviews were conducted with around 10,000 adults living in 5,500 households across Great Britain.These respondents were then tracked and re-interviewed annually along with any other co-resident adults.The BHPS sample was boosted in 1999 when additional households from Scotland and Wales were recruited into the panel.
BHPS was replaced in 2009 by Understanding Society.This has a very similar longitudinal design but interviewed a fresh sample of over 50,000 adults in its first wave (Buck & McFall, 2012).Unlike BHPS, data collection for every Understanding Society wave spans two calendar years, although interviewers still aim to return to each respondent every twelve months.Wave one interviews were thus divided into monthly samples spread across 2009-10 before interviewers then returned to households for their wave two interview twelve months later in 2010 (for those originally contacted in 2009) or 2011.Continuing BHPS participants were invited to join Understanding Society from wave two in 2010.For BHPS respondents who took part in Understanding Society, we discarded data covering the period between their last BHPS interview (mostly autumn 2008) and their first entry into Understanding Society in 2010.This is because the time gap separating these observations is larger than that separating all other consecutive wave pairs 11 .
Academic and policy debates often focus on how the housing careers of recent generations are deviating from the normative historical benchmark provided by the Baby Boomers (Hoolachan & McKee, 2019).We extend this by comparing the housing careers of four recent and tightly defined seven-year birth cohorts: those born (1) 1967-73; (2) 1974-80; (3) 1981-87 and (4) 1988-94 (Table 1).These cohorts entered the UK housing system after 1990 (the 1967-73 cohort was around 18-24 when BHPS began in 1991) and have experienced very different housing market and contextual conditions as they aged.Table 1 shows the age at which each cohort experienced two potentially important period effects: the onset of the GFC in 2008 and the 2013 introduction of UK government Help to Buy support for first-time buyers.Our analysis focuses solely on UK-born adults as immigration significantly altered the composition of younger cohorts during this period.

Measures
In both surveys, information on individuals' legal tenure status and that of their partner (if applicable) was used to code a dummy dependent variable measuring whether a person and/or their partner was a homeowner.One drawback of this is that treating homeownership as a dichotomous state overlooks variation within the tenure: for instance by dwelling size, location or mortgage gearing.Investigating these nuances -as well as the housing careers of migrants -is left to follow-up research.
Region is our main independent variable.To ensure adequate sample size, we took the former Government Office Regions of Great Britain and collapsed the English regions outside London into three categories with broad commonalities in housing market conditions (Figure A1): the South (comprising the South West, South East and East of England), the Midlands (East and West Midlands) and the North (Yorkshire and Humberside, North West and North East).Scotland and Wales are kept separate due to differences in their housing markets, tenure profiles and post-devolution housing policy.
A number of additional predictors of homeownership were defined for the panel analyses (Table 2).An ordinal age variable was included as ownership transitions are likely to broadly increase with age as people accumulate savings and potentially strive to own in order to reduce housing costs after retirement.The transaction costs of buying and selling homes may also become a less relevant deterrent as people age and typically become less likely to expect moving.Additional dummies were included for sex and ethnicity.
Two dynamic variables were included to capture the synchronisation of family events with ownership transitions.First, a categorical partnership indicator was defined using information on (1) partnership status at waves t and t + 1 (where t indicates the baseline wave and t + 1 when transitions into homeownership are measured) as well as (2) relationship type and elapsed duration.We used this information to distinguish people who were consistently single from those who were cohabiting or married at both t and t + 1 and in an established relationship (defined as partnerships that had lasted at least 2 years by t + 1).Additional categories were defined for those in recent cohabitations or marriages at t + 1 (defined as having lasted under 2 years) and for a small number of people splitting from a partner between waves.A two-year threshold was used to define recent relationships as partnership and housing transitions are often imperfectly synchronised (Bayrakdar et al., 2019).
A second family indicator was coded using information on whether the respondent was living with their own and/or their partner's own children at t and t + 1.This indicator distinguishes people never observed to be living with children from those consistently living with children and from those transitioning from not living to living with children.While having children consumes income and may dampen homeownership transitions (a negative resource crowding effect), parents may be driven to enter homeownership in order to obtain greater dwelling control and display their adherence to cultural norms tying good parenthood to owner-occupation (a positive normative effect).
Several variables were used to capture the effect of resources and socio-economic position on homeownership entry.These include a dummy for degree attainment, an indicator of employment status and job type, and a family income variable measured as the total gross personal income of the individual and their partner (if applicable) in 2015 pounds.As discussed by Sissons & Houston (2019), we could not include variables measuring savings, intergenerational transfers or parental homeownership as this information has not been routinely gathered in the surveys.To alleviate potential confounding we included a dummy variable for highest parental occupational status.A categorical variable measured respondents' housing circumstances at wave t.This distinguishes those living in the parental home 22 from those in the private and social rented sectors 33 .In general, we expect renting to consume resources and delay entry to homeownership.Selection effects may exacerbate this as people prioritising homeownership remain in the parental home to save up (Clapham et al., 2014;Suh, 2020).Social renting is expected to particularly dampen homeownership transitions as the tenure provides an affordable and secure long-term housing alternative to owner-occupation.Finally, a regional migration dummy variable was defined to control for changes of region between t and t + 1.

Approach
The analysis began by using the LFS to compare cohort homeownership trajectories.We then used panel data to analyse homeownership transitions and the way these are shaped by life course processes and vary over time and place.To do this, we first descriptively compared national age-graded rates of homeownership transition across the four focal cohorts.Logistic regressions were then used to model the probability that non-owners observed at each wave t had become owners at t + 1. Respondents were removed from the risk sample after they first entered homeownership and to avoid double-counting couples we randomly selected one individuals' record when both partners had been fully interviewed.These procedures left a final sample of 50,465 records provided by 13,030 people (mean = 3.9 records/person).
The logistic regressions were built up in stages.We first entered our principal predictors (age, cohort and region) before adding life course controls.As our main focus lies in examining regional patterns of homeownership entry across cohorts, interactions of region*cohort and region*age were then tested and examined using marginal effects and predicted probabilities (Mize, 2019).All models were fitted with standard errors adjusted for the clustering of records within individuals.

Cohort homeownership trajectories
Figure 1 uses four LFS sweeps conducted at seven-year intervals (1998, 2005, 2012 and 2019) to track the national homeownership trajectories of each birth cohort as its members aged.The y-axis shows each cohort's estimated owner-occupation rate when its members were the ages shown on the x-axis.No 18-24 data point is included for the 1967-73 cohort as LFS data from 1991 are not comparable with the later time-series.
Several key points emerge from Figure 1.First, the speed of entry to owner-occupation has declined substantially across successive cohorts.While over 50% of the 1967-73 cohort were homeowners when aged 25-31, fewer than 40% of those born after 1980 owned their home by the same age.Although this trend is well-known, what is most striking from Figure 1 is just how quickly such significant changes in owner-occupation trajectories have occurred in Great Britain.These rapid shifts are missed when comparing the housing careers of more broadly defined generations.
Figure 1 further indicates that declining entry to owner-occupation has been an uneven process.Although homeownership rates at ages 18-24 fell slightly across the three observed cohorts, inter-cohort divergence is far greater at ages 25-31 and 32-38.The most precipitous gap in the age-graded homeownership rate separates the 1974-80 and 1981-87 cohorts at ages 25-31, which they reached in 2005 and 2012 respectively.This fall occurred during a particularly difficult period for first-time buyers as stretched affordability prior to the GFC was followed by economic uncertainty, borrowing constraints and reduced housing market activity post-2008.
Figure 1 provides further evidence that cohort homeownership trajectories changed during the GFC period as from 2005 to 2012 a gap opened up between the homeownership rates of the 1967-73 and 1974-80 cohorts.Aged 25-31, the 1974-80 cohort's ownership rate was tracking that of the 1967-73 cohort, but the former subsequently made much less progress than the latter as they aged through to 32-38.By 30 many Britons have finished education, left home, entered the workforce and formed partnerships and so the postponement of these life course transitions is unlikely to fully explain this divergence.Instead, it seems more likely that GFC-related economic turmoil was especially detrimental for the ownership trajectories of those cohorts who experienced this shock during the peak age phase of homeownership entry spanning the mid-twenties to late-thirties.Finally, the cohort trajectories in Figure 1 do not support more pessimistic narratives that younger cohorts are a Generation Rent permanently locked out of owner-occupation.By the time they were aged 39-45 in 2019, the 1974-80 cohort had attained a similarly high level of homeownership to the 1967-73 cohort at the same age.Meanwhile, owner-occupation among those born 1981-87 rose 20 percentage points from 2012 to 2019.A similar pattern of continuing entry is observed for the 1988-94 cohort whose homeownership rate at age 25-31 (just under 40%) is tracking that of the preceding cohort.Overall, the evidence in Figure 1 suggests that more recent cohorts remain willing and to some extent able to enter homeownership, but this transition is being postponed to higher ages.
Figure 2 disaggregates each cohort's age-graded homeownership rate by region.Overall Figure 2 suggests that the regional patterning of all four cohorts' trajectories is relatively muted.Moreover, what regional patterns in age-graded homeownership rates there are tend to be fairly similar across cohorts.This indicates that similar types of shifts in patterns of homeownership entry through early adulthood have occurred across the entire country since the 1990s.
However, closer inspection of Figure 2 nuances this general picture in several ways.First, across the cohorts it appears that minor regional variations in homeownership at ages 18-24 existed in the late 1990s but have since disappeared.Among the 1974-80 cohort, rates of owner-occupation at ages 18-24 were significantly higher in the South than in London or Scotland.These regional patterns had weakened among the 1981-87 cohort and disappeared entirely by 2012 when those born 1988-94 were 18-24.Very few people now buy homes before the age of 25 and this means spatial variation in ownership entry during the early phase of young adulthood has declined over time.
Second, in the oldest cohort levels of homeownership before the age of 40 were lower in Scotland than anywhere else outside of London before Scottish patterns converged with the other regions at higher ages.No such Scottish pattern appears in the later cohorts.Finally, London is the only region that appears consistently distinctive in Figure 2. Across all cohorts, younger adults living in the capital have experienced slower growth in homeownership with age than their peers living elsewhere.Within each cohort, homeownership rates were particularly reduced in London at ages 25-31 and the much lower homeownership rate of the 1988-94 cohort at ages 25-31 in London (just over 20% as compared to 40% elsewhere) provides strong evidence that entry to the tenure continues to be strongly delayed in the capital.

Transitions into homeownership
Each cohort's age-graded homeownership trajectory is produced by people transitioning in and out of the tenure as their life courses unfold.While entry and exit rates both matter for ownership levels, existing evidence indicates that entry patterns are the driving force as the risk of falling out of owner-occupation was low during much of this period (Wood et al., 2017).To better understand entries into homeownership, Figure 3 uses the panel data to plot the proportion of non-owners in each cohort of a given age at t who transitioned into owner-occupation by t + 1.A higher proportion indicates a higher risk of entering homeownership at a given age conditional on not already being an owner.As case numbers are low for single ages, the transition rates were smoothed by calculating them over three-year moving windows centred on the focal age 44 .To test the validity of our transition rate estimates we applied them to predict cohort aggregate ownership trajectories.The predicted trajectories are show in Figure A2 and are similar to those estimated from LFS data (Figure 1) 55 .Figure 3 shows that a nonlinear pattern of declining rates of entry to owner-occupation through the early lives of each successive cohort has produced the trajectories in Figures 1 and 2. The 1967-73 cohort's rate of transition rose sharply through the early twenties, peaked at ages 25-27 and then gradually declined.While the 1974-80 cohort's pattern was similar until their mid-twenties, this cohort's transition rate then slowed significantly with lower levels of entry through the late twenties to late thirties.This caused the two cohorts' rates to diverge as they entered their thirties (Figures 1 and A2).
In line with Figure 1, a major change in the pattern of homeownership transitions occurred between the 1974-80 and 1981-87 cohorts.Rates of homeownership entry for those born 1981-87 rose slowly with age through the twenties and peaked later (around 30) and at a much lower level than was the case for the earlier cohorts.These twin trends have flattened this cohort's homeownership trajectory significantly compared with its predecessors.Transition rates among those born 1988-94 initially started at even lower levels but as this cohort has aged its members have begun to enter owner-occupation more quickly than the 1981-87 cohort.Further analysis of the housing origins of new homeowners (Figure A3) demonstrates a stable pattern as transitions from the parental home decline with age while private rental entries rise 66 .

Modelling homeownership entry
We now examine how homeownership transitions vary across time and space after controlling for observed differences in life course careers.To do this we built up a sequence of logistic regressions which used panel data to model the probability of ownership entry between t and t + 1. Table 3 presents the models with estimates shown as Average Marginal Effects (AMEs).AMEs show the change in the probability of entering homeownership produced by changing a given predictor across all cases in the sample while holding all other variables at their observed values (Mize, 2019).
Model 1 in Table 3 contains only the age, cohort and regional variables of primary interest.In line with Figure 3, the age AMEs in Model 1 confirm that the probability of entering owner-occupation is low prior to 21 before rising to peak at 25-29 and then falling back through the late thirties.Compared with those born 1967-73, the probability of ownership entry is, on average, 1.4 percentage points lower for those born 1974-80, 3.8 percentage points lower for those born 1981-87 and 4.8 percentage points lower for those born 1988-94.The probability of entering ownership is meanwhile 1.8 percentage points lower in London than in the South with smaller negative for the Midlands and Wales.Model 2 adds control variables to test whether this changes the age, cohort and regional AMEs.The age effects persist in Model 2 but are heavily attenuated which indicates that the age-grading of homeownership transitions is partly due to the biographical timing of other processes.By contrast, the AMEs for the cohort dummies remain fairly stable when controls are added.This suggests that the changing structure and tempo of cohorts' life course careers cannot fully explain why they have been delaying entry into homeownership.Comparing the cohort AMEs in Model 2 against one another as well as against the 1967-73 reference category reinforces that the cohort trend is not linear.While the probability of entry into homeownership fell sharply across the three older cohorts, this decline subsequently levelled off and the two post-1980s cohorts have rather similar AMEs.
Finally, a new pattern of regional effects emerges in Model 2 with homeownership entry slightly faster in the North and in Scotland as compared with the South.London, meanwhile, retains its negative AME. Comparing each region against each of the others by varying the reference category in Model 2 (detailed results not shown) indicates that the probability of entering ownership is significantly lower in London and slightly higher in Scotland than in any other region.Contrasts among the South, Midlands, North and Wales are very small and with the exception of the aforementioned South-North comparison these never approach statistical significance.
The control estimates in Model 2 of Table 3 largely match our expectations.Ethnic minorities have slower transitions to owner-occupation while having a degree, working in a managerial or professional job 77 , higher incomes and having more socio-economically advantaged parent(s) are associated with a higher probability of entry.The small negative male effect is less straightforward but may be linked to younger men being more likely than young women to live in the parental home to higher ages.Renting -especially from social landlords -is associated with slower entry to homeownership than remaining in the parental home, while inter-regional migration is associated with synchronous transitions into owner-occupation.Further analysis (not shown) showed that this outmigration effect is especially pronounced in London, hinting that people selectively leave the capital in order to become homeowners.
Model 2 provides evidence that some events in the family life course are powerfully synchronised with homeownership entry.On one hand, going from having no co-resident children at t to having co-resident children at t + 1 is insignificantly associated with ownership entry.Living with children at both t and t + 1 does, however, dampen the entry probability in line with the resource crowding hypothesis.On the other hand, recent partnership formation is a potent predictor of entering homeownership.Being in an established partnership is, meanwhile, associated with a higher probability of entering ownership (as compared with staying single), while those splitting from a partner have a slightly higher tendency to enter ownership.The latter finding is counterintuitive but eyeballing the data indicates that the parameter is based on a very small number of cases who tend to have been in short cohabitations (median = 2 years) and do not always sustain ownership long.It seems plausible that splitting from a short relationship may have triggered some individuals already saving up for homeownership to enter the riskier fringes of owner-occupation.

Temporal variation in regional effects
Finally, we tested whether the regional patterns of ownership entry varied across cohorts and with age.Sample size limitations meant that a three-way interaction between all focal variables was not possible and so we fitted two separate models with interactions of cohort*region and then age*region.Only the former is presented as both models yielded very similar substantive results and interacting cohort with region best addresses our research question.As it is not straightforward to interpret the coefficients from interacted variables in logistic regressions we follow Mize (2019) and present the results using plots of marginal effects and predicted probabilities.
Figure 4 shows how the regional AMEs vary across the cohorts.The figure shows that in all four cohorts the probability of entering homeownership across early adulthood was similar in the South, Midlands and Wales.A slight positive AME is observed for the North in all cohorts, but this only reaches conventional levels of significance among those born 1988-94.In line with Model 2, Scotland has a positive AME across all cohorts, although this only attains significance for those born 1967-73 and 1981-87.London has a significant negative effect for the latter three cohorts and further analysis of additional models interacting age with region (results not shown) confirms the picture from Figure 2 that this negative London effect was strongest across all cohorts at ages 21-29.
Figure 5 shows how the cohort and regional variables together shape the predicted probability of entering owner-occupation for a hypothetical person whose age is allowed to vary and whose other characteristics are fixed to favour homeownership.The figure shows regional differences in ownership entry were minimal among the 1967-73 cohort who had higher entry probabilities at all ages than each successive cohort.The negative effect of living in London exists for all subsequent cohorts while few differences between the South, Midlands, North and Wales are observed in any cohort.Similar patterns of predictions emerge when interacting age with region (results not shown) although this model provides somewhat stronger evidence for lower entry probabilities in London through the twenties of those born 1967-73.

Conclusions
The causes and consequences of declining access to owner-occupation are a major theme of contemporary housing debates.The current UK government has, like many others, viewed this trend as a problem and responded by pledging to affordability constraints by increasing housing supply and providing demand-side assistance.However, there is growing academic concern that the prevailing narrative of thwarted transitions provides only a partial understanding of changing patterns of entry to homeownership (Coulter et al., 2020).Comparing the housing attainments of broadly defined generations misses more rapid changes occurring over shorter timescales, while sub-national geographic variation in housing careers remains surprisingly poorly understood.To address these issues, this study examined how patterns of transition into homeownership have varied regionally for four tightly defined cohorts of younger adults entering the UK housing system after 1990.
Our results confirm the popular narrative of declining rates of entry to homeownership.However, comparing tightly defined birth cohorts shows that this restructuring has not just happened over a generational timescale, but rather has also occurred rapidly and in a non-linear fashion through the last few decades.Overall, the main pattern is of a shift towards lower, slower and later homeownership with a particularly pronounced inter-cohort decline in entries during the mid-to-late-twenties across the three older cohorts.The fact that significant cohort differences persist after controlling for personal attributes, family dynamics and resources indicates that changes elsewhere in the life course cannot fully explain delayed entries (Choi e al., 2019;Corlett & Judge, 2017).It is highly likely that increased affordability constraints are an important factor as cohorts have thus far partially caught up with their predecessors' homeownership levels as they aged beyond 30.
Cohort trends in entry to owner-occupation have also varied by age but these patterns diminish when changes in cohorts' demographic and socio-economic composition are controlled in regression models.While ownership entry among very young adults has been consistently low since the 1990s, during their twenties the 1981-87 cohort moved much more slowly into owner-occupation than the 1974-80 cohort did at the same age.The fact the 1974-80 cohort only began to lag the preceding cohort at the same historical moment -the period spanning roughly 2005 to the mid-2010s -hints that GFC-related life course disruption may have had a particularly adverse impact on the housing careers of adults in the prime phase of ownership entry at the time.Furthermore, disaggregating the millennials into an early (1981-87) and late cohort (1988-94) reveals that these have experienced different early homeownership trajectories, with the latter's homeownership rate very low in their early twenties before starting to surpass that of the former.This could be due to those born 1988-94 benefiting from a more favourable context in their twenties as state Help to Buy support coupled with easier access to credit eases entry to owner-occupation.From a life course perspective, this highlights how commonalities in cohorts' housing careers are produced by the age at which people experience particular period effects (Myers et al., 2005).
Adding region into the analysis nuances the national picture but less strongly than expected.London emerges as consistently distinctive as living in the capital has dampened early entries to owner-occupation (particularly among those in their twenties) across all four cohorts.This is probably due to particularly high and rapid inflation of housing prices in the capital as well as the lower proportion of owner-occupied stock which together drive delayed homeownership transitions as well as selective outmigration to buy homes (Hamnett & Reades, 2019).Regional differences elsewhere are muted and there seems to have been only minor changes in regional patterns of ownership entry across the four cohorts.Overall, it seems shifts in patterns of homeownership entry across cohorts have for the most part occurred across the entire country rather than in specific regions (Corlett & Judge, 2017).
Taken together, the findings have implications for research and policy in Britain and beyond.The fact that declining rates of homeownership across cohorts seems to reflect postponement rather than permanent lock-out or reduced ownership preferences implies that aspirations to own remain strong but that these are taking longer to realise (Preece et al., 2020).Given that many countries' policies and cultural contexts continue to strongly favour owner-occupation, the fact that homebuying is still being postponed implies research needs to focus on the implications this delay is having for recent cohorts' quality of life, prosperity, well-being and sense of self-worth (McKee et al., 2017).We also need to know more about how deferred homeownership plans fit in with aspirations and decisions across other life course domains, for example relating to leaving the parental home, having children or employment.Slower and later entry to owner-occupation among recent cohorts further hints that entry is becoming more unequal as family support and larger mortgages are required to overcome affordability barriers (Suh, 2020).
Future research using more granular geographical data or qualitative methods could build on these findings by exploring how people striving to enter homeownership make trade-offs in dwelling types, neighbourhoods and city or regional locations.While Sissons & Houston (2019) reported a decline in the propensity to leave cities to buy housing between 1994 and 2008, anecdotal evidence suggests this trend has recently reversed as the housing market rebounds, COVID-19 has boosted working-from-home and dwelling space demands, and as cheaper electric and autonomous vehicles promise to revolutionise private transportation.In this context there is a clear need for more work on how geographies of homeownership entry fit into the broader context of changing residential decision-making over the life course.

Notes
1.The BHPS to Understanding Society transition means that no annual panel data span the 2008-2009 period.2. Couples are defined as living in the parental home if living with either partners' parent(s), although this occurs rarely.3. Cases where people lived in owned dwellings but incomplete information on ownership was provided were removed.4.This means that, for example, the rate for 19-year-olds in a cohort was calculated as the proportion of at risk 18, 19 and 20-year-olds who transitioned into owner-occupation from t to t + 1. 5.Although the age-graded trajectories are similar, the actual rates do vary from those estimated from the LFS in Figure 2.This is to be expected as inter alia the panel predictions do not take account of exits.6. Transitions from social renting are very low for all but the 1967-73 cohort.7. Gender interactions with employment status were examined but discarded as these showed little of interest.

Figure 2 .
Figure 2. Cohort homeownership rates by age and region.

Figure 4 .
Figure 4. Marginal effects of region across cohorts.

Table 3 .
Logistic regression models of homeownership entry.