Partnership patterns and homeownership: a cross-country comparison of Germany, the Netherlands and the United Kingdom

Abstract

Using detailed micro-level survey data for three advanced European welfare-state economies (Germany, Netherlands and UK), our analyses suggest a fairly common hierarchy to homeownership, according to partnership status, exists. In all three countries, a variety of interrelated factors appear to encourage greater propensities for homeownership amongst co-residential households (married/cohabiting), as compared to single-person households. However, important macro-contextual differences do appear to play a significant role in mediating the magnitude of difference within this hierarchy. For instance, in Germany the importance of marriage as a predictor of homeownership is found to be particularly strong, with married couples having far higher propensities for homeownership, even when compared to non-married cohabiters. In the Netherlands and UK, where an emphasis on traditional family and marriage is less pronounced, and where homeownership is generally more popular and accessible, the differentiation between married/unmarried partners is greatly reduced. Furthermore, we find no evidence to suggest that living-apart-together partners are more/less likely to own their home than singles.

Keywords:

• Homeownership
• partner relationships
• welfare regimes
• second demographic transition
• comparative analysis
• Germany
• Netherlands
• United Kingdom

Introduction

Partner relationship status and the tenure type of the home are arguably two of the most crucial factors that determine a person’s well-being through the life course. Furthermore, a considerable volume of literature informs us that they are far from independent in their development. For example, a recurrent finding in housing research is that one of the strongest determinants of homeownership (as compared to renting) is partnership status (understood as a combination of marital and cohabitation status), combined with family status. Whilst part of this impact undoubtedly runs via wealth and income, the relationship holds, even after accounting for the availability of these resources (Mulder, 2013; Mulder & Wagner, 1998).

Yet, when we talk of partnership status, we are increasingly referring to what is a far more complex subject than perhaps it was just a few decades ago. After all, the vast majority of the world’s advanced democracies have witnessed a substantial increase in the diversity of living arrangements, family formations and partner relationships, and indeed the norms and attitudes that prescribe them. For some, this long-term, and still ongoing, diversification represents a Second Demographic Transition (SDT), a transition said to be driven, in part, by a weakening of the traditional institutions of ‘the family’ and marriage (Cherlin, 2004), coupled with a corresponding rise in secularisation, individualism and self-articulation (see Lesthaeghe, 2010; Lesthaeghe & van de Kaa, 1986). The emergence and acceptance of same-sex, non-marital, part-time and post-marital cohabitation, as well as the increasingly discussed phenomenon known as living apart together (LAT) – where partners live in separate dwellings – are important markers of diversification in today’s society (Latten & Mulder, 2013). They are also important factors to consider within housing studies, given that, at the micro level, different forms of partnership are correlated with different forms of housing tenure, and that, at the macro level, the increased proliferation of non-traditional partner relationships and family formations may inform contemporary and future demands for different forms of dwelling type and tenure, as well as their relative affordability.

Bearing this in mind, it is certainly surprising that no housing research, at least to our knowledge, has been developed in a way that enables the exploration of this increased diversity. Thus, whilst data limitations preclude attempts to incorporate all possible contemporary partnership configurations, this paper seeks to build on the traditional use of a tripartite partnership model (single, cohabiting or married) by including LAT relationships as well as more detailed ‘never married’ and ‘previously married’ specifications of singlehood. Further to this, we seek to explore potential cross-country variations in the association between different partnership patterns and homeownership. Whilst utility theories are valuable tools for understanding the micro-level mechanisms that lay behind the desire/ability to enter one form of partnership or housing tenure over another, they can be limited in their value if important structural processes are ignored. In most cases, we should expect the sociocultural and institutional context of a country to play an important role in framing and informing micro-level desires and outcomes.

Internationally, comparative research has two major forms. In one form, as many as possible countries are included in the analyses and characteristics of the countries are incorporated into multilevel statistical frameworks, as variables at the country level. We apply a second form, in which a small number of countries are included and hypotheses on differences between the countries are derived from a careful weighing of combinations of country-specific characteristics. Comparative research of the second kind is often fruitful if the countries compared share many similarities, but at the same time reflect important differences. Thus, in this analysis, we compare Germany, the Netherlands and the United Kingdom (UK). All three are advanced economies and welfare states, but with different welfare regimes and sociocultural traditions; there are many similarities in how the housing markets operate in the three countries, but there are also important differences. For our analysis, therefore, we draw on three sources of comparably detailed and nationally representative microdata, namely pairfam (Germany), NKPS (Netherlands) and Understanding Society (UK), and employ binary logistic regression on a sample of individuals aged 25–40 in order to obtain estimates of the predicted probability of homeownership according to different partnership patterns and country-specific contexts.

Conceptual background

From the perspective of rational choice or utility theory, different types of partnerships provide different benefits and costs. These benefits and costs are related to a number of welfare dimensions. Emotional benefits can be realised by any type of partnership – marriage, cohabitation or LAT – though their interpretation may vary across the life course. For example, whilst international comparisons on the topic have been rare to non-existent, LAT relationships are said to be particularly diverse in their reasoning and expected durability. For some, LAT can reflect a particularly short-term relationship such as dating, or a transitional phase that leads to a more concrete co-residential partnership in the future. Indeed, for younger adults, LAT relationships may often be rather involuntary, where a number of factors (e.g. the geographic location of places of work/study and other financial and housing issues) work in combination to delay, or even prevent, the occurrence of a desired co-residential relationship (Reimondos et al., 2011). However, for certain other groups, particularly older and/or divorced people, LAT may more accurately portray a longer term and desired alternative to co-residential partnership – where it can enable the maintenance of independence in a new relationship and, on certain occasions when dependent children are involved, a desired preference to avoid introducing step-parents into the immediate family home (De Jong Gierveld, 2004; Haskey & Lewis, 2006; Liefbroer et al., 2015). Likewise, different forms of co-residential partnership come with their own divergent benefits. We know, for instance, that marriage tends to be a more secure and durable partnership status than non-married cohabitation (Tach & Edin, 2013; Teachman et al., 1991), though the extent to which the durability and security of marriage can be considered beneficial is largely dependent on the partners’ perceived satisfaction and the conditions of their marriage. Moreover, all forms of co-residential partnership should be expected to provide a greater opportunity for more intensive partner interaction, when compared to non-cohabiting, LAT, partnerships.

Yet, beyond the relative emotional costs and benefits of different partnership formations, in the context of housing, and perhaps homeownership in particular, there are a number of additional financial and instrumental factors that must be considered. Certainly, the pooling of resources between co-residential partners can greatly improve the likelihood of being able to meet the necessary financial costs of homeownership (Mulder & Wagner, 1998); costs which range from the initial search and down payment, right through to mortgage interest and repayment as well as everyday maintenance. Singles and those in LAT partnerships therefore fail to capitalise on the advantages of a common household production, while the latter group are additionally confronted with higher transaction costs associated with partner interaction. However, in some cases, previously married singles¹ could still maintain some form of resource pooling (e.g. alimony) linked to the divorce settlement – a factor which could increase their overall propensity for homeownership when compared to singles who have never married.

Beyond the immediate benefits of joint household production, it is also possible that marriage, in itself, represents an informal indicator of relative financial durability and security. Oppenheimer (1988) has argued that widespread cultural norms profess an expectation that marriage be delayed to a point where a couple’s relative economic security has been achieved, or at least understood as such. Consequently, to the extent that economic security is a critical prerequisite for mortgage acceptance and home purchase, higher propensities for homeownership amongst married couples can be expected, even after controlling for important characteristics such as age, co-residence, income and human capital. Furthermore, in sociocultural contexts where cohabitation more likely reflects a stage in the marriage ‘process’ as opposed to a viable longer term alternative to marriage (Heuveline & Timberlake, 2004), this selection into marriage should be even stronger, with the differences between married and non-married couples being even greater.

A costs-benefits approach is also useful for explaining differing homeownership patterns, where again, financial and non-financial considerations are thought relevant. From a benefits perspective, homeownership is widely associated with increased security of tenure and protection against eviction, as well as more general additional advantages related to the improved relative location and condition of owner-occupied homes, as compared to rented dwellings. For instance, it is suggested that owner-occupied homes tend to be larger, of better quality and situated in generally more attractive residential environments (Megbolugbe & Linneman, 1993; Mulder et al., 2006; Mulder & Wagner, 1998). From a financial perspective, the purchase of a home can be an attractive and relatively safe long-term investment, particularly in countries that actively promote homeownership with subsidies and tax incentives (Matha et al., 2014). However, as briefly mentioned above, the initial financial costs, including for instance transaction charges and transfer taxes, as well as any ongoing costs associated with the repayment of mortgages and the maintenance of the dwelling, are all burdens to which renters are largely immune. Consequently, the costs of homeownership are often said to be heavily front-loaded, that is, in the initial years following the purchase, homeownership is often considerably more expensive than renting, even when the quality of the dwelling is comparable (Mulder & Wagner, 1998). Yet, as with different partner relationships, the balancing of the various benefits and costs will vary greatly depending on which stage in the life course we choose to focus upon. For instance, private renting, with its increased flexibility, its short-term contracts and low transaction costs, may well be the most desirable tenure type for young adults, and particularly singles, be they at university or at the start of their occupational careers. Likewise, given a decline in resources and the need for less space, recently separated/divorced individuals are also associated with moves into the rental sector (Feijten, 2005), though the motivation for such a move is clearly very different to that of the younger single adults. Conversely, for the reasons of space, quality and security already raised, homeownership will often present itself as the most attractive form of tenure to established couples and families with children (Lauster, 2010; Mulder & Wagner, 1998). Nonetheless, previous studies in the UK have found homeownership to be associated with lower and delayed fertility (Hakim, 2003; Murphy & Sullivan, 1985), where it is thought the costs of homeownership compete with the costs of raising a child (Courgeau & Lelièvre, 1992).

Given the above, we can be fairly confident in expecting individuals who are married, and perhaps more broadly those who are in a co-residential partnership, to reflect certain characteristics that lead to generally higher rates of homeownership, when compared to individuals in one-person households. However, it is less obvious what we should expect for those in LAT relationships as compared to unpartnered singles. For example, where an LAT relationship represents a dating and/or inconvenient temporary phase in the partnership, as it may, particularly for younger adults, there is a strong argument for expecting homeownership propensities to be closely aligned with those of singles. However, where one LAT partner owns their home, and where again LAT is not the desired long-term solution, it should be expected that this form of partnership would very soon give way to cohabitation in the owner-occupier’s home. In this situation, those who remain in LAT partnerships could be less likely to own their home than unpartnered singles – though again this would mostly concern the older and/or divorced LATs in our sample, given the requirement for existing homeownership (discussed later). Yet, as mentioned above, LAT relationships can also be longer term, the result of careful consideration and preference – again particularly amongst older and/or separated/divorced individuals. Thus, where theory suggests that a preference for LAT is motivated by a desire to maintain independence, a pre-existing status as a homeowner may very well be one of the key factors of independence that formed the basis of the LAT relationship in the first place. If this is indeed correct, older LATs could be characterised by relatively higher rates of homeownership. Together, these arguments suggest that the relationship between the different partnership patterns and homeownership may well vary depending on the age group studied, and the approximate stage in the lifecourse. It is also worth noting that, at least in post-divorce scenarios, this relationship could be further complicated by gender variations, given that women tend to be more vulnerable to deteriorating housing careers in the years that follow divorce (Feijten, 2005).

Linking the macro level to the micro: comparing Germany, the Netherlands and the UK

The notion of the welfare regime, incorporating the welfare triad of the family, the market and the state, has proved particularly useful for those attempting to make sense of contextual differences. According to Esping-Andersen’s (1990, 1999) trichotomy of welfare regimes, countries with a more liberal approach are characterised by an increased emphasis on individualism and the market (Arts & Gelissen, 2002). As such, the welfare system is structured so as to provide little more than a ‘safety net’ against poverty (McGinnity, 2004). The UK, despite its National Health Service and universal state pension, is generally seen as an example of a liberal welfare regime.

Aside from the unique historical deviation of the former German Democratic Republic (GDR), Germany is often projected as the quintessential example of the conservative welfare tradition. With influences from Catholic social teaching, the conservative approach is committed to the maintenance of status differentiation, including, for instance, those pertaining to the traditional ‘male-breadwinner’ family structure – wherein men are expected to undertake full-time paid employment within the labour market, whilst women largely operate as caretakers within the unpaid realm of the household, or in part-time paid work (Esping-Andersen, 1999). Indeed, Germany’s adherence to a conservative approach remains evident in government policy. For example, marriage continues to be actively encouraged through the allowance of spouses to jointly declare their taxable incomes; this can result in significantly lower overall tax liabilities for married couples, particularly where the difference in wages between partners is large (Bach et al., 2013). Likewise, the maintenance of status differentiation is evident in the provision of social security, where entitlements are strongly tied to employment status and are provided through an insurance-based system which discriminates according to employment history and contribution (McGinnity, 2004). Whilst this approach encourages status differentiation across broad socio-economic lines, the importance of one’s current and previous labour market status is also key to maintaining traditional gendered family structures, where unpaid caregivers (who are mostly women) are dependent on their employed partner (traditionally men) for access to social security (Schmitt, 2012). With that said, over recent decades, German policy has shifted, albeit slowly, towards greater de-familialisation and increased incentives for women to enter the labour market (Fleckenstein, 2011; Leitner et al., 2007). Furthermore, it is important to keep in mind that despite the 25 years that have passed since reunification, the legacy of the GDR – which emphasised widespread female labour market participation, the reduced role of the Christian church (Wagner, 1997), and strong restrictions to homeownership – should still be important when it comes to explaining differing partnership patterns and homeownership rates within the country.

Whilst it is clear that no one country fits the typology perfectly, the position of the Netherlands is a particularly difficult case to define. Like Germany, it has many characteristics of a conservative employment regime (Visser & Hemerijck, 1997), but also quite a few of a social-democratic regime, for instance with high levels of redistribution and state pension provision (Goodin et al., 1999; Kammer et al., 2012). And to some extent, liberal approaches to family policy are also evident, where restrained state influence is combined with an emphasis on individual responsibility and a variable degree of employer support (Knijn & Ostner, 2008).

Yet, regardless of the differences in the welfare approach employed, a clear and common shift towards the increased prevalence of non-traditional partnership patterns and family formations has been observed in the three countries, a shift which is itself understood as a reflection of the deep and long-standing sociocultural changes associated with the SDT (Lesthaeghe, 2010). However, this ‘shift’ towards an increasing occurrence and acceptability of alternative living arrangements need not follow the same path, nor work at the same speed, even across the three countries studied here. After all, differences in sociocultural norms, as well as the extent to which the state plays an active role in family welfare, can be expected to influence the form and speed with which countries transition towards the diversified formations and expectations associated with the SDT. For instance, the enactment of government policies that formalise a conservative deference for marriage and traditional family formation may well act to slow the transition towards partnership diversification and acceptability, whereas countries with more liberal or social-democratic sociopolitical systems may have more neutral, or enabling, policy approaches and discourse. Furthermore, the degree to which a country has transitioned towards greater a diversification in partnership patterns could be important for interpreting the relative significance of marriage as a predictor of homeownership. As Lauster & Fransson (2006) have noted, this increasing diversity of partnership patterns appears to have coincided with changes in the relationship between partnership types and tenure choice. Whilst the events of marriage and homeownership have traditionally proved to be strongly concurrent (Clark & Dieleman, 1996; Mulder & Wagner, 1998), the sociocultural shifts associated with the SDT are thought to be undermining this simultaneity (see Lesthaeghe (2010) for a discussion of the similar disconnect between marriage and procreation), and thus diminishing the relative role of marriage as an important determinant of housing tenure choice. With this in mind, the resilience of the traditional link between marriage, family and homeownership may be greater in the more conservative context of Germany (where transitions have been slower), than perhaps it is in the Netherlands and the UK (where the transition is more advanced and where unmarried cohabitation is a more realistic long-term alternative to marriage).

On the specific topic of housing, Esping-Andersen’s original thesis was rather quiet. However, over more recent years, a small but increasing number of scholars have sought to investigate interrelationships between welfare regimes and housing (Fahey & Norris, 2011; Hoekstra, 2005) and proposed specific categorisations of countries according to housing market characteristics (Mulder & Billari, 2010). Unsurprisingly, given our discussion above, these housing markets differ in the relative role of state regulation as well as the degree to which homeownership is normatively prescribed. In the UK, for instance, the housing market for privately rented accommodation is lightly regulated with short-term (typically 6–12 months) tenancies, while the social housing sector is exclusively for people in need (Norris and Shiels, 2007). Thus to some extent, when compared to the private and social rental sector alternatives, homeownership in the UK represents an almost unchallenged source of tenure security and quality. This compares to Germany and the Netherlands, where the market for rented accommodation is more regulated, offering a comparatively large proportion of rental housing characterised by lower rents, a generally higher quality of dwelling, and improved rights for tenants (Kirchner, 2007; Van der Heijden, 2002). Moreover, most of this rental accommodation is privately owned in Germany, whereas the Netherlands has a particularly large and comparably better-resourced social sector, most of which is owned by not-for-profit housing associations (Boelhouwer, 2005). Additional differences in the fiscal treatment of homeownership also exist. For instance, the Netherlands has had full tax deductibility of mortgage interest for a long time, with only small reductions introduced in the last few years (Haffner, 2002; Mastrogiacomo, 2013). Likewise, the UK had a similar approach to tax deductibility of mortgage interest, though this was gradually phased out through the 1990s and was completely removed by the year 2000 (Ermisch & Halpin, 2000). Germany, on the other hand, has offered few benefits for homeowners in general; though a specific 8-year-limited tax allowance for home buyers (Eigenheimzulage), which provided additional benefits when the household contained children, was in place until January 2006 (Oxley & Haffner, 2010; Voigtländer, 2009). However, with this allowance being granted only once in a householder’s lifetime, the benefits of short-term residential durations were comparatively small. Thus, according to Voigtländer (2009), the scheme may have actually worked to delay homeownership until household members were satisfied that they had accumulated sufficient capital to acquire a home that fulfilled their longer term residential requirements. Indeed, whilst Germany’s comparably stringent mortgage criteria (e.g. high mortgage down payments, rigorous current income requirements, long-term fixed interest rates and early repayment penalties) and widely available low-cost and high-quality rental alternatives are considered critical (Kurz, 2004; Mulder & Wagner, 1998; Scanlon & Whitehead, 2004; Voigtländer, 2009), this additional factor may also underpin the particularly low homeownership rates observed in the younger German cohorts.

By European standards, the national homeownership rate in Germany is low, with just 43 per cent of the housing stock being owner-occupied (Statistisches Bundesamt, 2013, p. 17). This compares to the Netherlands and UK where the national averages are 60 per cent (MBZK, 2013, p. 7) and 64 per cent (Beckett, 2014, p. 4), respectively. Though whilst low levels of homeownership, particularly amongst younger adults, may be a traditional pattern in Germany’s housing market, an unprecedented decline in homeownership rates has been observed for younger adults throughout many other European countries, including in countries like the UK where normative preferences for homeownership are dominant (McKee, 2012). Despite homeownership remaining the most common form of tenure in the UK, a series of interrelated macro-financial/economic trends appear to have combined in making homeownership an increasingly unobtainable goal for young first-time buyers. Such trends are said to include: the rapid rise in house prices; tighter lending requirements; a rise in inflation that exceeds that of pay; and an increase in the average house price to wage ratio (ONS, 2014). Moreover, the continued proliferation of the buy-to-let market should also be acknowledged (DCLG, 2013). Thus, broadly speaking, the costs of homeownership have increased while the ability of households to save sufficient capital, in order to obtain a mortgage and purchase a home, has decreased. The decline in homeownership in the UK is contrasted by rather stable rates in Germany and the Netherlands (Eurostat, 2015a). This unique decline in homeownership in the UK has the potential to be a key social and political issue in the future given other macro-economic transitions that have occurred over recent decades. Indeed, homeownership has been said to be playing an increasingly important role as a key source of asset-based security and welfare. Coinciding with, and perhaps encouraging, the declining (individual and state) investments in collective welfare-state provision, housing equity has been argued to be increasingly valued as a financial reserve that can be used to maintain family/household welfare in cases where traditional (collective) sources of welfare provision have either receded or do not exist (e.g. to cover children’s educational costs or long-term elderly care needs) (Lowe et al., 2011; Smith & Searle, 2010). Clearly, if asset-based welfare is indeed growing into a key source of welfare provision in the UK, the decline in homeownership among its younger adult population is a concern.

In an attempt to synthesise the points raised so far, we derive two key research hypotheses as follows:

(1)	Co-residential married couples will have higher propensities for homeownership than co-residential cohabiting couples and co-residential couples will have higher propensities than individuals who are single (never married or previously married) or living apart together.
(2)	Variation between the countries will exist due to differing welfare traditions and contextual characteristics: Married couples in Germany will have a particularly raised propensity for homeownership when compared to non-married cohabiting couples, while these differences will be relatively small in the Netherlands and UK.

Due to differing, and indeed competing, expectations, the expected difference between LATs and truly single individuals remains vague and we therefore avoid forming a strong hypothesis on this matter.

Data and methods

With a key focus of this paper being the evaluation of cross-country variations, large-scale surveys of comparable detail are essential. Thus, for the UK, the analysis draws on the United Kingdom Household Longitudinal Study, also known as Understanding Society (UoS) (McFall, 2013). Launched in 2009, UoS was designed to provide high-quality longitudinal data on topics including work, income, education, family and social life. Through the achievement of a very large sample, UoS holds great potential in allowing researchers to gain greater insights into particular population subgroups, subgroups that have been hard to measure in conventional longitudinal studies with smaller sample sizes (Bryan, 2011). The data we use here are drawn from Wave 3 (2011/2012, individual n = 49 739), the first wave where specific questions are asked that allow for the identification of those in LAT relationships. We exclude all records that were based on proxy interviews because they do not record information on partner relationships. For Germany, we use data from the German Family Panel, pairfam: Panel Analysis of Intimate Relationships and Family Dynamics. Pairfam is a valuable source of information for researchers interested in family relationships and dynamics, with detailed data drawn from anchor persons and their partners, parents and children (Huinink et al., 2011). Pairfam is a longitudinal study currently carried out in five waves. The study is scheduled to run for 14 years, ending in 2022. For temporal consistency, we use data from Wave 3 (2010/2011) which contains information on men and women of the three birth cohorts, 1991–1993, 1981–1983 and 1971–1973, who are aged 16–20 years, 26–30 years and 36–40 years by Wave 3 (n = 9074). Finally, our data for the Netherlands are drawn from the Netherlands Kinship Panel Study (NKPS) (Dykstra et al., 2007). NKPS is a large-scale panel survey of the Netherlands population aged 18–80 (first wave). Again, in order to maintain temporal comparability, we use data from the most recent wave, Wave 3 (2010/2011, n = 4390), which contains a wealth of information on partner relationships and sufficient information on housing (e.g. information on respondent and partner homeownership).

By Wave 3, the youngest respondents in NKPS are aged 25 and thus pairfam’s youngest cohort is too young for comparison. Consequently, in order to make the three samples comparable, we select only men and women aged 25–40, with 40 years being the maximum age for respondents in the oldest pairfam cohort, the unweighted age and sex distributions for each national sample are given in Table 1. Unfortunately, pairfam’s birth-cohort design means that our German sample does not contain individuals between the ages of 31 and 35, whilst this omission could result in issues for analyses focussed on transitions into homeownership, we do not expect it to influence analyses which focus on the probability of currently living in homeownership. Nevertheless, to check the robustness of our findings, we provide replications of our German analyses which draw on respondents aged 25–40 in Wave 1 of the German Generations and Gender Survey (GGS) (Vikat et al., 2007), see Appendix A and B.

Table 1. Unweighted age and sex distributions for national samples.

	Age			Sex
Sample	Min	Mean	Max	Male (%)	Female (%)
Pairfam	26.0	33.8	40.0	43.5	56.5
NKPS	25.0	35.1	40.0	34.3	65.7
UoS	25.0	33.5	40.0	37.6	62.4

Note:

Figures are rounded.

Given the unique nature of UoS as a household survey, only one person from each household is selected for use. Beyond this, we exclude individuals in each sample who were still living in the parental home – removed pairfam n = 425 (9.1 per cent); NKPS n = 40 (4.7 per cent); UoS n = 873 (10.6 per cent). And finally, cases with item nonresponse on the variables selected for analysis are removed, this resulted in the exclusion of 228 (5.1 per cent), 6 (.7 per cent) and 106 (1.4 per cent) cases for Germany, the Netherlands and the UK, respectively. Given these conditions, we are left with three analytical samples containing 4230 (pairfam), 817 (NKPS) and 7389 (UoS) complete cases. These samples were pooled for analytical purposes so as to allow for the simultaneous modelling of comparative effects (pooled n = 12 436).

Our comparative analytical approach involves the calibration of a binomial logistic regression model with interaction terms for country by partner relationship. The dependent variable in our model is binary in nature, where y = 1 indicates that the respondent (or if applicable their partner) owns (either outright or with a mortgage) the home they are currently living in. Whilst we accept that some partnerships are neither fully co-residential, nor fully LAT, for instance commuter partnerships (Van der Klis & Mulder, 2008), the independent variable of interest, the respondent’s partner relationship status, categorises the respondent into one of five partner relationships: married with partner; no partner, never married; no partner, previously married; partner, but living apart together; and unmarried cohabiting with partner. Moreover, beyond these key analytical variables, age, gender, socio-economic status (educational attainment and occupational class) and settlement type (urban/rural area) are included as controls. Educational attainment is a measure differentiating respondents with a degree level education or above, while occupational class is measured using Erikson and Goldthorpe’s internationally comparable seven-category classification (EGP) (Ganzeboom & Treiman, 1996) – which we further aggregate into a set of four broadly ordinal occupational classes: higher managerial/professional; lower managerial/professional; routine/manual; and no occupation. An interaction between occupational class and gender is also included given that the gender division in unpaid domestic work and labour market participation can result in the term ‘unemployment’ holding different connotations for men and women – e.g. women are more likely to take on voluntary family-care responsibilities while being supported by a partner’s income, and so we can expect financially supported ‘stay-at-home’ women to have higher rates of homeownership than otherwise similar unemployed men in the sample. Finally, we include a separate dummy variable for East Germany (former GDR). Unfortunately, due to the cross-national and cross-sectional nature of our analysis, a comparable measure of duration and stability of household incomes is unavailable.² However, it can be argued that occupational class and the level of educational attainment act as reasonable proxies for this characteristic. Additionally, whilst relationship status will to some extent predict the relative opportunity for the pooling of resources within the household, explicit information on whether the household is single or dual-earner is also of interest, given the sociocultural and welfare variations between the countries studied. Thus, whilst we do not include it in our logistic regression model, due to issues of variable confounding, Table 4 provides an accompanying overview of the share of co-residential couples according to earner configuration (single or dual-earner), as well as the relative homeownership rates by said earner configuration. Table 2 presents a descriptive summary of the independent and dependent variables used in our final analytical model. It should be noted that an alternative specification of the pooled model was explored, treating East and West Germany as substantively separate countries (see Appendix C). The results of this more complex model were in agreement with the more parsimonious model we present below (Table 5). Moreover, separate country-specific analyses are presented in Table 6, where, amongst other things, the seemingly trivial differences observed between East and West Germany justifies our final pooled-sample model specification in Table 5. Further to this, an additional sensitivity analysis, involving the calibration of comparable regression models accounting for survey design and nonresponse, suggests that the unadjusted model findings presented below are robust to such issues (this analysis can be provided upon request).

Table 2. Frequencies of variables (unweighted) and relative share of homeowners.

	N	Homeowner (%)
Homeowner
Yes	6280
No	6156
Partner relationship
No partner, never married	1795	23.2
No partner, previously married	538	30.3
Partner, but living apart together	1109	28.9
Unmarried cohabiting with partner	2617	43.9
Married with partner	6377	66.3
Age
Homeowner mean age	35.0
Renter mean age	32.4
Gender
Female	7537	50.9
Male	4899	49.9
Children in household
Yes	8027	55.8
No	4409	40.9
Degree level education
Yes	5006	57.9
No	7430	45.5
Occupational class
Higher man/prof occupations	1727	68.8
Lower man/prof occupations	2667	60.5
Routine/manual occupations	5104	50.2
No occupation	2938	31.2
Settlement type
Urban area	9708	48.5
Rural area	2728	57.7
Country
Germany	4230	35.7
Netherlands	817	78.6
UK	7389	55.9
East Germany	852	27.1

Results

Empirical variations in family composition and co-residential earner configurations

Before we present our main model findings, it is thought important to provide a brief empirical introduction to the similarities and differences in family composition found between the three countries studied. Table 3 provides details on the proportion of people in different partner relationships and according to whether they live in a household with(out) a child(ren) for Germany, the Netherlands and the UK.³

Table 3. Percentage partner relationship status by children in household by country.

	Child(ren) in household
	Germany			Netherlands			UK
Partner relationship	Partner relationship %	With child(ren) %		Partner relationship %	With child(ren) %		Partner relationship %	With child(ren) %
	C %	3.0		C %	.8		C %	10.2
No partner, never married	15.9	R %	10.2	10.2	R %	4.6	16.8	R %	37.9
		C %	3.0		C %	.8		C %	10.2
No partner, previously married	3.5	R %	56.3	2.7	R %	27.2	3.7	R %	70.1
		C %	3.6		C %	1.3		C %	4.2
Partner, but living apart together	9.7	R %	19.1	8.2	R %	6.2	9.4	R %	39.5
		C %	3.4		C %	.9		C %	5.9
Unmarried cohabiting with partner	20.2	R %	34.6	29.1	R %	43.2	23.4	R %	56.7
		C %	12.9		C %	22.4		C %	20.7
Married with partner	50.7	R %	82.4	49.8	R %	83.9	46.7	R %	79.4
		C %	77.1		C %	74.6		C %	59.1

Notes:

R = row percentage, C = column percentage. Pairfam n = 4230; NKPS n = 817; UoS n = 7389, samples cover individuals aged 25–40, estimates are weighted to account for survey design and nonresponse. For those who are LAT, the household relates to those who live with the individual respondent and not the partner.

According to Table 3, the shares of people in the various partner relationship statuses are similar across the three countries, though individuals are somewhat less likely to be without a partner and more likely to cohabit in the Netherlands than in Germany and the UK. Perhaps unsurprisingly, in each country the likelihood of having a child (or children) in the household is highest for those who are married to their partner. However, interesting differences do emerge when we observe some of the less-traditional forms of household composition. For instance, in Germany and the Netherlands, a substantial portion of those who live independently (whether LAT or with no partner) live in households that contain no children. Moreover, a greater proportion of unmarried cohabiters are in households without a child. This compares to the UK where the differences in household composition according to partner relationship status are far less distinct. For instance, unlike Germany and the Netherlands, we find a marginal majority (56.7 per cent) of unmarried cohabiters to live in households containing a child. Perhaps more significantly, for those LAT and those with no partner (never married), the proportion of households with and without a child is far more equal. The 2011 UK Census observes similar findings where lone parents with dependent children represented 26 per cent of all families with dependent children (ONS, 2012). Data compiled by Eurostat (2015b) also reveal the uniqueness of the UK in terms of the high percentage of live births outside of marriage. Between 2001 and 2012, the UK averaged 44.1 per cent of live births outside of marriage, which compares to 30.2 per cent in Germany and 37.6 per cent in the Netherlands. Interestingly, over this same period, the Netherlands has witnessed a rapid increase in the proportion, where it now reflects a similar magnitude to that found in the UK – the percentage in Germany has increased at a slower rate (Appendix D). From the SDT perspective, these patterns would appear to offer some support to the notion that the three countries are indeed transitioning towards greater household diversity, though at quite different rates. Whilst the occurrence of children in LAT and single-never-married households is comparatively rare in Germany and the Netherlands, we could expect these rates to increase over the coming decades.

Table 4 presents the national shares of co-residential couples according their single or dual-earner household status, as well as their subsequent homeownership shares. As with Table 3, the results fit closely to our expectations, given the respective sociocultural and welfare traditions of the three countries. For instance, given the long-standing occupational split between men and women in Germany, the share of dual-earner couples is far smaller in Germany than it is in the Netherlands and the UK. Moreover, we observe high homeownership rates regardless of dual- or single-earner status in the Netherlands (N.B. the high rate for no-earner couples is based on a very small subsample and is therefore extremely unreliable). Conversely, Germany again has low homeownership rates across the earner configurations, though when compared to the UK, no-earner couples have higher rates than they do in the UK – this could reflect the relative importance of parental inheritance in Germany’s housing market (Mulder & Wagner, 1998), with comparably low taxation, meaning that homeownership can be more easily obtained outside of considerations of the couple’s earnings. Ignoring the spurious estimate for no-earner couples in the Netherlands, unsurprisingly, given the importance of pooled resources, we see a common pattern to all countries: dual-earners have higher homeownership rates than single-earner couples and no-earner couples. However, for the UK, the gradient is apparently far stronger, with dual-earner households having considerably higher rates of homeownership when compared to single-earner and no-earner co-residential couples. Again, whilst we are unable to explore the duration and stability of household earner configurations over time, it does appear that being part of a duel-earner co-residential partnership is particularly important in the UK, an observation that again supports the particular importance of pooled resources for enabling homeownership in what is a comparatively (and increasingly) expensive housing market.

Table 4. The percentage of co-residential couples according to earner configuration and the percentage of homeownership by earner configuration.

Earner configuration	Germany		Netherlands		UK
	% of co-res couples	Homeowner %	% of co-res couples	Homeowner %	% of co-res couples	Homeowner %
Dual-earner	27.2	48.2	81.8	83.4	69.3	73.0
Single-earner	56.3	40.8	17.7	76.5	24.8	48.5
No earner	16.5	33.8	.6	90.2	5.9	10.2

Notes:

Pairfam n = 3072; NKPS = 660; UoS = 5262, samples cover individuals aged 25–40 who are in a co-residential partnership (married or unmarried) at the point of survey. Estimates are weighted to account for survey design and nonresponse.

To some extent, our initial empirical findings do appear to fit with the discussions on welfare regime and housing market differentiation. In keeping with the conservative tradition, we find that being married with a partner is by far the most common of partner relationships in Germany, a theme further supported by the fact that the vast majority (77.1 per cent) of individuals with a child in the household are also observed to be married. Similarly, Germany has a relatively high percentage of single-earner couples, a pattern we associate with the more traditional male-breadwinner model. This stands in contrast to the UK where the share of married people is the smallest of the three countries, and where a large portion of non-married singles and couples have at least one child in their household. Indeed, the traditional family structure appears to be far less dominant in the UK, which could, to some extent, be a reflection of its more liberal, non-familialistic, welfare approach. Moreover, the blurriness discussed with regards to the Dutch welfare regime is also somewhat evident in Table 3. The overall share of married people is shown to be the largest of all three countries, while the proportion of non-traditional family formations, for instance those with children living in single-parent households, is far larger in the Netherlands than in a more easily defined conservative Germany. Moreover, the Netherlands also has the highest share of dual-earner households. As was noted above, it is important to remember that the design of the pairfam sample means that there are no individuals recorded between the ages of 31 and 35. To check the robustness of our descriptive summaries in Table 3, a replication using respondents aged 25–40 in Wave 1 of the German GGS was undertaken, the results (see Appendix A) appear to further confirm the assertions made here.

Partnership patterns and propensities for homeownership

Our empirical results provide evidence of cross-country variations in the pattern and propensity of different partner relationships and of homeownership. With this in mind, we turn to the model-based analysis of the pooled analytical sample (Table 5). Figure 1 provides the predicted probabilities for homeownership according to partner relationship status in each country. Aside from the independent country effects, where the Netherlands and the UK reflect considerably higher total rates of homeownership than Germany, the largest differentials in the predicted probability for homeownership are located between the different partner relationship configurations. Broadly speaking, the same ownership hierarchy exists in all three countries, where married couples are the most likely to own their home followed by cohabiting couples, and finally a broad group of one-person households who, regardless of type, are substantively equivalent in terms of their lower propensity for homeownership. This shared relationship, which may more accurately be portrayed as the difference between those in a co-residential partnership and those living alone (or separately), certainly fits well with the arguments we put forward earlier.

Table 5. Binomial logistic regression: predicting the likelihood of homeownership.

Predictor	B	SE	95% CI	Odds ratio
(Intercept)	1.504	.096	1.317: 1.692
Age (centred at 34)	.101	.005	.092: .111	1.107
Gender (ref: female)
Male	−.106	.095	−.292: .081	.900
Degree level education (ref: no)
Yes	.255	.048	.161: .349	1.291
Occupational class (ref: lower man/prof occupation)
Higher man/prof occupation	.111	.106	−.096: .320	1.118
Routine/manual occupation	−.483	.076	−.632: −.335	.617
No occupation	−1.109	.081	−1.268: −.952	.330
Settlement type (ref: urban area)
Rural area	.408	.053	.305: .512	1.504
Children in household (ref: no)
Yes	−.356	.065	−.484: −.229	.700
Partner relationship (ref: married with partner)
No partner, never married	−1.801	.086	−1.971: −1.632	.165
No partner, previously married	−1.471	.128	−1.724: −1.221	.230
Partner, but living apart together	−1.429	.098	−1.622: −1.238	.240
Unmarried cohabiting with partner	−.715	.069	−.851: −.580	.489
Country (ref: UK)
Germany	−1.578	.113	−1.800: −1.356	.206
Netherlands	.320	.264	−.189: .848	1.377
East Germany (ref: no)
Yes	−.382	.099	−.576: −.189	.683
Partner relationship × country
No partner, never married: Germany	.144	.178	−.209: .488	1.154
No partner, previously married: Germany	−.316	.247	−.812: .158	.729
Partner, but living apart together: Germany	−.259	.203	−.667: .132	.771
Unmarried cohabiting with partner: Germany	−.323	.130	−.579: −.069	.724
No partner, never married: Netherlands	−.246	.344	−.928: .424	.782
No partner, previously married: Netherlands	−.682	.521	−1.722: .340	.506
Partner, but living apart together: Netherlands	−.225	.389	−.990: .537	.799
Unmarried cohabiting with partner: Netherlands	.341	.269	−.183: .872	1.406
Gender × occupational class
Male: Higher man/prof occupation	.038	.152	−.260: .335	1.039
Male: Routine/manual occupation	.122	.114	−.102: .347	1.130
Male: No occupation	−.636	.155	−.941: −.334	.530
Children in household × country
Yes: Germany	.845	.113	.624: 1.067	2.329
Yes: Netherlands	.775	.253	.276: 1.271	2.171

Notes:

n = 12 436 (pairfam n = 4230; NKPS n = 817; UoS n = 7389).

Figure 1. Predicted probability (95% C.I.) of homeownership for partner relationship status by country.

Note: These are marginal effects at the sample means.

However, given our discussions on the potential relevance of the life course in altering residential and partnership preferences, one could have expected there to be a significant mediating effect of age on the partnership–homeownership relationship. For instance, where the results in Figure 1 propose LATs and singles to be similar in their propensity for homeownership, there is some rationale for predicting significant differences to emerge as age increases – where LAT may become less a phase and more a long-term choice. If this is indeed correct, we ought to see an increased divergence between LATs and singles in the homeownership propensities for the later ages. Yet, exploratory analyses on the importance of age as a mediator in this relationship failed to provide any substantively important differences (these results can be provided upon request). In fact, with the exception of previously married singles in the Netherlands (where a very small negative relationship between homeownership and age was found), homeownership is shown to increase in a similar way with age regardless of partner relationship status and country – this finding presumably reflects the universal importance of time in allowing for the accumulation of sufficient capital to enable homeownership.

Yet, regardless of the similarities in the overall pattern to homeownership, where a co-residential vs. one-person household hierarchy is clearly shared, the magnitude of the differences between the partner relationships is found to vary greatly depending on the national context studied. A glance at Figure 1 quickly reveals Germany’s uniqueness, not only in its lower propensity for homeownership overall, but also in terms of the significant difference that exists between, on the one hand, those who are married with partner, and on the other, the rest (N.B. these differences are reduced in the regions that were previously a part of the GDR (see Table 6 and Appendix C), where again marriage appears to be far less important than it is in West Germany). In the UK, the difference between being married and cohabiting with a partner is far smaller, while in the Netherlands, the estimates indeed overlap. Whilst it is not possible to claim a direct causal link, it is possible that a number of factors relating to the differing sociocultural and institutional traditions of the countries underpin these variations. In countries where mortgages are more easily obtained and where the legacy of familialism and state involvement is less pronounced, the significance of the distinction between cohabitation and marriage appears reduced. These results have been replicated using alternative German and Dutch data, Wave 1 of the German GGS and Wave 1 of the Dutch NKPS (see Appendix B).

Table 6. Separate country-specific regression models.

	East Germany		West Germany			Germany			Netherlands		UK
	B	SE	B		SE	B		SE	B	SE	B		SE
(Intercept)	−.439	.344	−.162		.158	−.162		.142	1.565	.325	1.583		.115
Age (centred at 34)	.138	.021	.132		.009	.132		.008	.106	.025	.087		.006
Gender (ref: female)
Male	−.311	.382	−.073	.162		−.089	.148		.354	.350	−.145	.137
Degree level education (ref: no)
Yes	−.096	.223	.185	.096		.152	.088		−.094	.274	.328	.059
Occupational class (ref: lower man/prof occupations)
Higher man/prof occupations	−.448	.542	−.175	.244		−.240	.221		.790	.405	.102	.134
Routine/manual occupations	−.348	.306	−.353	.144		−.336	.130		−.376	.287	−.622	.101
No occupation	−1.205	.367	−.396	.154		−.498	.141		−.838	.351	−1.431	.105
Settlement status (ref: urban area)
Rural area	.649	.183	.618	.094		.581	.082		.498	.248	.234	.071
Children in household (ref: No)
Yes	.260	.234	.275	.114		.231	.100		.437	.263	−.251	.067
Partner relationship (ref: married with partner)
No partner, never married	−1.722	.378	−1.677	.173		−1.734	.156		−2.097	.342	−1.808	.087
No partner, previously married	−.821	.418	−2.198	.253		−1.910	.212		−2.202	.506	−1.442	.130
Partner, but living apart together	−1.390	.368	−1.760	.207		−1.741	.179		−1.647	.381	−1.436	.099
Unmarried cohabiting with partner	−.744	.221	−1.051	.130		−1.048	.110		−.382	.263	−.734	.070
Gender × occupational class
Male: Higher man/prof occupations	.630	.697	.106	.320		.166	.288		−.537	.599	.077	.200
Male: Routine/manual occupations	.310	.455	.226	.204		.211	.185		−.179	.503	.092	.158
Male: No occupation	−.271	.652	−1.086	.342		−1.054	.298		.241	.678	−.517	.199

Notes:

East Germany n = 852; West Germany n = 3378; Germany n = 4230; Netherlands n = 817; UK n = 7389.

Moving towards the wider context of the family, we uncover further comparative differences in homeownership. Owner-occupied dwellings tend to be larger, of better quality, situated in better environments and associated with enhanced security of tenure (Mulder & Wagner, 1998), thus the raised likelihood of homeownership for households with children in Germany is as we might expect. The directional relationship found in the Netherlands is the same as in Germany but, for reasons likely related to limitations in sample size, the estimates come with very wide and overlapping confidence intervals. Indeed, the larger Wave 1 NKPS sample used in the replication model confirms a positive relationship for the Netherlands (see Appendix B). Interestingly, for the UK, the opposite directional association is found – the predicted probability of homeownership is greater when there are no children in the household. As such, our findings fall in line with previous research in the UK (Hakim, 2003; Murphy & Sullivan, 1985) and lend some support to the continued relevance of the competing costs hypothesis for the UK context (Courgeau & Lelièvre, 1992), a process that can lead to a delay in childbearing, given the high costs associated with home purchase in the UK.

With regards to the remaining covariates, homeownership is found to increase with age, those living in rural areas are more likely to be owner-occupiers than those living in urban environments (compare Mulder & Wagner, 1998), while the control for East Germany reveals an anticipated negative association with homeownership. In terms of the socio-economic characteristics included in our model, a raised level of human capital, measured here as a degree-level education, is associated with greater levels of homeownership. Likewise, individuals from both the lower and higher managerial and professional occupations are more likely to be homeowners than those with no occupation as well as those from the routine and manual occupations. With that said, the negative impact of having no occupation is significantly greater for men than it is for women. Of course, given the continued likelihood for women to be allocated care and domestic responsibilities (Arts & Gelissen, 2002; Lewis, 1992), non-participation in the labour market for men may in general reflect a more precarious financial state than is the case for otherwise similar women – that is, the likelihood of non-employed women to be in a financially supported ‘stay-at-home’ role, as opposed to being unemployed but seeking work, is greater – particularly in the family forming 25–40 age range studied here. Again, the separate country-specific models (Table 6) suggest this gendered relationship to be particularly strong in the more conservative West Germany, while there appears to be no such relationship for the Netherlands.

Conclusion

Whether it is a question related to the suitability of the current housing stock or one focused on the relative affordability of homes, the increased proliferation of non-traditional partner relationships and family formations is a critical issue for those interested in housing. Consequently, in an attempt to address a small part of what is a very considerable dearth in the literature, this paper has presented an analysis designed with the purpose of uncovering the differing propensities for homeownership according to partner relationship status in three advanced European welfare-state economies.

Our findings suggest that a rather universal hierarchy to homeownership exists. For a variety of financial and non-financial reasons, the probability of homeownership is found to be higher for those in co-residential partnerships as compared to those in LAT relationships and singles (both ‘never married’ and ‘previously married’). Whilst there are plausible reasons for expecting LATs to have different homeownership propensities to singles, for the countries and ages studied here, we find no substantive difference. As we have argued, each form of partnership can be interpreted from the perspective that it provides a unique series of relative benefits and costs. For example, when compared to those in LAT relationships, co-residential partnerships offer the raised opportunity for cheaper and more intensive inter-partner interaction as well as the benefits of a common household production. Many of these same benefits and costs can also be assumed to be inextricably tied to the individual and/or joint decision and ability to enter homeownership. For instance, the benefits of a shared household production, the pooling of resources and the need for greater space are all important factors that make homeownership a more desirable and achievable goal for co-residential partners and families. Yet these same individual and/or joint decisions and abilities are also framed and influenced by wider social and institutional factors. Differences in normative expectations, housing market configurations, and approaches to taxation and state regulation undoubtedly impact on micro-level behaviours and outcomes regarding homeownership.

Summary statistics show there to be clear differences between the countries in terms of their headline homeownership rates as well as their relative shares of different household formations. And these differences do fit closely to what would be expected, given their respective welfare approaches and sociocultural traditions. However, with greater relevance to the immediate interests of this paper, the macro context would also appear to be critical when it comes to mediating the magnitude of difference in homeownership propensities between the different partnership types. Despite the common cross-country hierarchy to homeownership, Germany’s conservative approach to the family and the (housing) market appears to have resulted in a comparatively low total homeownership rate as well as particularly pronounced micro-level differences according to marital status in the probability of owning a home. Indeed, with a legacy of normative and policy-implemented preferences for marriage and traditional household formation, a reserved approach to lending, and a limited stock of owner-occupied housing in the market, homeownership appears to be largely restricted to married couples. Conversely, for the Netherlands and the UK, where the emphasis on traditional family formation and marriage is less pronounced, and where homeownership is a far more accessible and popular option, the differentiation in homeownership between married and unmarried co-residency is found to be significantly reduced, at least for the 25–40 age group studied here.

In terms of its relevance for housing researchers and practitioners, the findings in this paper suggest that greater attention should be paid to the long-term demographic shifts that have occurred across Europe, and much of the developed world. The transition towards greater partnership and family diversity has changed the composition of households in profound ways, though the rate at which this diversification has occurred appears to be strongly influenced by the sociocultural and welfare traditions of the country studied. Whilst Germany appears to be somewhat behind the Netherlands and the UK in terms of its transition towards greater family/partnership diversity, it may well be the best placed to deal with the potential issues that this increasing diversity will raise with regards to future housing needs and provision. Indeed, the non-traditional partnership formations appear, all else equal, to be associated with lower rates of homeownership as compared to (married) co-residential couples. Thus, in a country such as Germany that has a less significant tradition of homeownership, driven in part by high-quality and affordable alternative dwellings in its private rental sector, the transition towards greater diversity and delayed homeownership needn’t be such an issue, so long as the alternative is maintained. However, at the other end of the scale, a country such as the UK, where homeownership is normatively prescribed and represents an unchallenged source of housing security and quality – given the lack of alternatives in the private and social rental sectors – the transition towards greater family and partnership diversity, on top of already declining homeownership rates among the youngest cohorts, should be of real concern for policy-makers and housing practitioners; and particularly if asset-based welfare is to be relied upon as an alternative to traditional collectivist provision in certain welfare scenarios.

Finally, some thought should be given to the limitations of this study. As with most analyses involving family and partner relationship variables, we are restricted to sample surveys which contain detailed attribute data, but relatively small samples. When such surveys are used in comparative studies, the requirement for comparability often dictates the variables that can be used and, as was the case here, the specific sample subgroups that can be drawn on. In our case, the requirements for temporal comparability meant the restriction of the analysis to sample members aged between 25 and 40 years. The implications of this restriction include the relatively small final sample for the Netherlands (n = 817) as well as the analytical limitation of not being able to explore the potentially interesting differences for individuals in older age groups, particularly LATs. Furthermore, due to a lack of survey waves with sufficiently detailed partner relationship status indicators for each country, we were restricted to a cross-sectional approach. This is unfortunate given that alternative studies of homeownership have been successful in shifting towards more informative longitudinal analyses, for instance where transitions into homeownership have been explored. If and when a sufficient number of waves containing detailed partnership data do become available, future research in this area should build on this analysis by employing a more detailed longitudinal analytical approach. Further to this, future research incorporating different countries with differing homeownership rates and differing welfare regimes will be important for further testing and developing our understanding of the macro–micro dynamics discussed here.

Funding

The research for this paper is part of the project ‘Partner relationships, residential relocations and housing in the life course’ (PartnerLife). Principal investigators: Clara H. Mulder (University of Groningen), Michael Wagner (University of Cologne) and Hill Kulu (University of Liverpool). PartnerLife is supported by a grant from the Netherlands Organisation for Scientific Research [NOW, grant number 464-13-148], the Deutsche ForschungsGemeinschaft [DFG, grant number WA1502/6-1] and the Economic and Social Research Council [ESRC, grant number ES/L0166X/1] in the Open Research Area Plus scheme.

Disclosure statement

No potential conflict of interest was reported by the authors.

Acknowledgements

We are grateful to Sandra Krapf, Michael Wagner (University of Cologne), Hill Kulu and Julia Mikolai (University of Liverpool) for their helpful comments and insights on an earlier version of this paper. We also thank the three anonymous reviewers for their useful comments.

Notes

1 Previously married singles include widowers as well as divorcees, though in our analytical sample the number of widowers is very small, given the 25–40 age range.

2 Once the panel studies mature this may be possible in the future.

3 For the German and UK samples, children are defined as under 16 years of age, for NKPS children can be of any age so long as they live in the household with their parent(s).

Appendix A.

German GGS replication of percentage partner relationship status by children in household in Germany.

Table

	Child(ren) in household
Partner relationship	Partner relationship %	With child(ren) %
No partner, never married	15.3	R %	12.8
		C %	4.9
No partner, previously married	7.2	R %	21.1
		C %	3.9
Partner, but living apart together	6.2	R %	13.2
		C %	2.1
Unmarried cohabiting with partner	8.7	R %	32.0
		C %	7.1
Married with partner	62.5	R %	52.0
		C %	82.1

Notes: R = row percentage, C = column percentage. GGS n = 2462, sample covers ages 25–40, estimates are weighted to account for survey design and nonresponse. For those who are LAT, the household relates to those who live with the individual respondent, and not the partner.

Appendix B.

Replication of pooled model using alternative Wave 1 GGS data for Germany and Wave 1 NKPS data for the Netherlands.

Table

Predictor	B	S.E.	95% C.I.	Odds ratio
(Intercept)	.918	.075	.772: 1.065
Age (centred at 34)	.096	.005	.086: .106	1.101
Gender (ref: female)
Male	−.024	.046	−.113: .065	.976
Degree level education (ref: no)
Yes	.605	.046	.515: .697	1.832
Settlement type (ref: urban area)
Rural area	.391	.057	.279: .503	1.479
Children in household (ref: no)
Yes	−.446	.062	−.567: −.325	.640
Partner relationship (ref: married with partner)
No partner, never married	−1.893	.084	−2.059: −1.730	.151
No partner, previously married	−1.529	.124	−1.775: −1.287	.217
Partner, but living apart together	−1.464	.095	−1.652: −1.278	.231
Unmarried cohabiting with partner	−.717	.067	−.849: −.586	.488
Country (ref: UK)
Germany	−1.534	.139	−1.806: −1.263	.216
Netherlands	.410	.180	.062: .766	1.507
East Germany (ref: no)
Yes	−.349	.129	−.605: −.098	.705
Partner relationship × country
No partner, never married: Germany	.156	.202	−.248: .545	1.169
No partner, previously married: Germany	−.341	.387	−1.159: .372	.711
Partner, but living apart together: Germany	−.180	.241	−.665: .281	.835
Unmarried cohabiting with partner: Germany	−.238	.176	−.588: .103	.788
No partner, never married: Netherlands	.081	.234	−.381: .537	1.084
No partner, previously married: Netherlands	−.138	.343	−.812: .534	.871
Partner, but living apart together: Netherlands	−.259	.282	−.815: .293	.772
Unmarried cohabiting with partner: Netherlands	.226	.184	−.134: .588	1.254
Children in household × country
Yes: Germany	.767	.139	.495: 1.041	2.154
Yes: Netherlands	.986	.176	.640: 1.331	2.680

Notes: n = 11 454 (German GGS (Wave 1, 2005) n = 2462; Netherlands NKPS (Wave 1, 2002–04) n = 1603; UoS (same as above) n = 7389). Age range 25–40.

Predicted probability (95% C.I.) of homeownership for partner relationship status by country.

Note: These are marginal effects at the sample means.

Appendix C.

Alternative pooled model: East Germany, West Germany, the Netherlands and the UK.

Table

Predictor	B	S.E.	95% C.I.
(Intercept)	1.505	.096	1.319: 1.693
Age (centred at 34)	.101	.005	.092: .111
Gender (ref: Female)
Male	−.110	.095	−.296: .077
Degree level education (ref: No)
Yes	.255	.048	.161: .349
Occupational class (ref: Lower man/prof occupations)
Higher man/prof occupations	.111	.106	−.096: .320
Routine/manual occupations	−.485	.076	−.634: −.337
No occupation	−1.111	.081	−1.270: −.954
Settlement status (ref: Urban area)
Rural area	.409	.053	.306: .513
Children in household (ref: No)
Yes	−.356	.065	−.484: −.229
Partner relationship (ref: Married with partner)
No partner, never married	−1.801	.086	−1.971: −1.633
No partner, previously married	−1.471	.128	−1.724: −1.221
Partner, but living apart together	−1.429	.098	−1.622: −1.238
Unmarried cohabiting with partner	−.715	.069	−.851: −.580
Country (ref: UK)
East Germany	−2.011	.237	−2.482: −1.551
West Germany	−1.563	.122	−1.802: −1.325
Netherlands	.320	.264	−.189: .848
Partner relationship × Country
No partner, never married: East Germany	.072	.384	−.731: .787
No partner, previously married: East Germany	.811	.428	−.062: 1.630
Partner, but living apart together: East Germany	.094	.374	−.682: .795
Unmarried cohabiting with partner: East Germany	−.100	.225	−.546: .336
No partner, never married: West Germany	.160	.192	−.222: .531
No partner, previously married: West Germany	−.635	.283	−1.211: −.098
Partner, but living apart together: West Germany	−.354	.227	−.812: .081
Unmarried cohabiting with partner: West Germany	−.383	.146	−.672: −.098
No partner, never married: Netherlands	−.246	.345	−.928: .424
No partner, previously married: Netherlands	−.682	.521	−1.722: .340
Partner, but living apart together: Netherlands	−.225	.389	−.990: .537
Unmarried cohabiting with partner: Netherlands	.341	.269	−.183: .872
Gender × Occupational class
Male: Higher man/prof occupations	.043	.152	−.255: .340
Male: Routine/manual occupations	.125	.115	−.099: .350
Male: No occupation	−.630	.155	−.935: −.328
Children in household × country
Yes: East Germany	.759	.228	.314: 1.211
Yes: West Germany	.856	.123	.616: 1.097
Yes: Netherlands	.775	.253	.276: 1.270

Notes: n = 12 436 (pairfam n = 4230; NKPS n = 817; UoS n = 7389).

Predicted probability (95% C.I.) of homeownership for partner relationship status by country.

Note: These are marginal effects at the sample means.

Appendix D.

National trends in the proportion of live births outside of marriage.

Source: Eurostat (2015b), Proportion of live births outside of marriage. Luxembourg: Eurostat, the statistical office of the European Union.