1 Introduction

Since the second half of the twentieth century, family forms have become more diverse everywhere in Europe. The decreasing propensity to marry and have children, the increasing tendency to postpone marriage and childbearing, and the general decline in the centrality of marriage in many societies, have led to the formation of very different types of living arrangements (Kapella et al. 2010). The traditional family model—i.e., a household consisting of a couple and their children—has been replaced by a range of family constellations. Living as a family in Europe today means living longer in smaller (with fewer siblings), often more deinstitutionalized (non-marital) and non-co-resident families (Hantrais 2006), in which the kinship networks have become “tall and lean.”

Despite these pronounced changes in family structures, most studies on family life conducted over the past few decades have focused on the traditional family, and have not considered alternative family structures and the consequences of these non-traditional arrangements for the life satisfaction (LS) levels of individuals (Uhlendorff et al. 2011). Indeed, although a wide range of social, economic, and institutional factors have been suggested as important determinants of LS (e.g., Bjørnskov et al. 2008), the link between family status and LS remains largely unexplored. While there is some evidence on this topic for the US (Spreitzer and Snyder 1974; Campbell et al. 1976; Clemente and Sauer 1976; Glenn and Weaver 1979), only a few studies have examined the LS-family status link in Europe (e.g., Saraceno et al. 2005; Kotowska et al. 2010).

In general, ascribing higher levels of LS to traditional living arrangements seems straightforward. However, explaining differences in life satisfaction by differences in living arrangements is far more complex than simple predictions would suggest. What is not clear is whether the correlation between high LS and living in a traditional family is partly, or even strongly, mediated by the fact that “atypical” families tend to be under greater economic strain. Being a member of a socioeconomically deprived family has been found to be associated with lower levels of life satisfaction (Amato 2000; Breivik and Olweus 2006; DeBell 2008). Thus, the lower levels of LS observed in families led by single mothers or fathers and in extended families could be driven by their lower wealth levels.

Using data from the 2007 European Quality of Life Survey (EQLS), we focus on the link between diverse living arrangements and LS in an enlarged Europe, and explore the possibility that family structures might affect men and women differently. We also seek to verify whether the relationship is mediated by the socioeconomic position of individuals. Although implicit in much of the literature, this previously unexamined perspective provides important insights into the nature of diverse forms of living arrangements, including the role of children within these relationships. From a statistical point of view, we adopt a multilevel approach in order to acknowledge that the LS of an individual can be shaped by the country-specific context in which he or she resides.

After this introduction, we review the theoretical perspectives and the empirical evidence that constitute the premises of the paper. We then present the analytical strategy adopted, followed by a summary of our results. The paper closes with a concluding discussion.

2 Life Satisfaction and Living Arrangements

2.1 Perspectives and Evidence

In addition to increasing life expectancy, low and lowest-low fertility rates, together with changes in partnership dynamics over the past few decades, are altering the shape and structure of families in Europe. In addition to population aging, the key mechanisms that link living arrangements with LS are related to childbearing and partnership dynamics. In a nutshell, different living arrangements are generated by partnership and childbearing dynamics, and the latter processes are in turn increasingly intertwined with the LSFootnote 1 of individuals (Kohler et al. 2005; Billari 2009; Billari and Kohler 2009; Parr 2010; Kohler 2012).

The positive effects of being in partnership on LS are mainly linked to social and emotional support mechanisms. Having a partner can help an individual cope with the strains of life (Johnson and Wu 2002). People in partnerships are likely to feel less lonely and helpless (Blanchflower and Oswald 2004), and they benefit from sexual intimacy (Waite and Joyner 2001). From a health perspective, having a partner may increase the likelihood of early detection of symptoms of illness and of receiving appropriate medical treatment. Being in a partnership may also discourage unhealthy behaviors, such as drinking and smoking (Joung et al. 1997). From an economic perspective, having a partner encourages the sharing of resources (ibid). Mutual financial responsibility provides a good basis for a division of labor and contributes to an increase in the joint utility of the household.

The findings in the empirical literature support theoretical predictions regarding the positive link between being in a partnership and life satisfaction. Studies conducted in the US have consistently found that people who live in a partnership report the highest levels of LS. Research has also shown that people who lack a partner—regardless of whether they are single, widowed, or separated—report the lowest levels of LS (Spreitzer and Snyder 1974; Campbell et al. 1976; Clemente and Sauer 1976; Glenn and Weaver 1979). For Europe, a key study on this issue was conducted by Kohler et al. (2005), who investigated the impact of children and partnership status on LS. Using a unique dataset from Denmark which allowed the authors to control for unobserved endowments (such as preferences and capabilities due to genetic dispositions, family background, and so on), the study found that men and women who were currently in a partnership were definitely happier than those who were not.

The other main source of differences in family structures is the childbearing process. The connection between having children and LS has recently received renewed attention in the socio-demographic literature (see Baranowska 2010 for an overview). Fertility theories seem to suggest a positive link between LS and parenthood. One of the most prominent theories that seeks to explain fertility variation is New Home Economics (Becker 1981), which assumes that childbearing is associated with opportunity costs, and that parents tend to specialize in either market or household production in order to maximize utility. According to this economic model of fertility, the utility derived from having children plays a crucial role, although this assumption has been neither tested nor explained in detail (Baranowska 2010). Testing the utility derived from childbearing requires the specification of the innate value of children, a component that is missing from models of this kind. For parents, having children may lead to a strengthening of social ties, provide enjoyable and novel experiences, and contribute to a sense of personal fulfillment (Hoffmann and Hoffmann 1973). In addition, a child can represent a potential source of support in old age. More generally, according to Billari (2009), the concept of “happiness commonality” is the key to understanding fertility differences. This theory is based on the observation that in rich contemporary societies, fertility is positively correlated with happiness, but is mediated by the policy environment. At the micro level, this approach suggests that in countries with low fertility, a certain level of happiness is considered to be a prerequisite for having a child, and that the perception of an increase (or a decrease) in an individual’s own happiness from having a child is a key factor in the decision to have (or not to have) a child. Finally, evolutionary theories suggest that parenthood may have a positive effect on LS because humans have evolved a predisposition to nurture (e.g., Rodgers et al. 2001). This implies that parents may enjoy taking care of children and fostering their intellectual and physical development.

All of these arguments suggest that parents derive satisfaction and economic benefits from having children. However, psychological studies have also shown that there are negative consequences associated with childbearing and childrearing. Parents experience stress related to their increased financial responsibilities after having children (Zimmermann and Easterlin 2006; Stanca 2009). In addition, becoming a parent reduces leisure time (Sanchez and Thomson 1997), affects the quality of the couple’s relationship (Lavee et al. 1996), and exerts pressure that might have negative effects on psychological well-being (McLanahan and Adams 1987).

While the impact of having a partner on well-being has been studied extensively, research on the effects of having children on LS is more scarce (e.g., Stutzer and Frey 2006). Furthermore, the evidence appears to be conflicting. While some studies have found a positive association between parenthood and happiness (Saraceno et al. 2005; Kotowska et al. 2010), other empirical studies that have removed bias resulting from selection into parenthood have shown that having children has either non-significant or negative effects on levels of LS (Clark and Oswald 2002; Clark et al. 2008; Angeles 2009a, b). Taken together, the few available studies appear to indicate that any positive effects of parenthood on LS—if they exist at all—can be found among women who have given birth to their first child (Kohler et al. 2005).

2.2 Mediators: The Role of Socioeconomic Status

One of the most important lines of research on well-being has been dedicated to the role played by socioeconomic factors in shaping people’s levels of satisfaction with their own lives. In particular, individuals’ education and professional status have attracted considerable attention, in part because they are valid surrogates for hard-to-measure concepts like opportunity costs, but also because they represent valid markers of individual autonomy, intellectual abilities, and independence of social norms.

People with higher levels of education are generally more satisfied with life (Kotowska et al. 2010). Employment status also influences LS (Clark and Oswald 1994; Clark 2003, 2006), as forms of insecure and unstable employment—which have been increasing all over Europe in recent decades—clearly lead to lower LS (e.g., Scherer 2009). The relationship between income and LS has also been extensively studied, with many scholars finding a positive association (e.g., Diener et al. 1995), as having a higher income can help removing obstacles to a successful life. However, research has also shown that after a certain point income has a smaller marginal effect on LS (e.g., Veenhoven 1984; Clark and Oswald 2002; De Santis et al. 2006).

Socioeconomic resources are not only crucial for defining the opportunities people have for maintaining themselves and for achieving personal fulfillment; they also shape family life. Economic independence achieved through the labor market shapes an individual’s living arrangements, his or her opportunities to start a family and to have children, and the stability of his or her family. Furthermore, there is increasing evidence of socioeconomic differentiation in family life in several European countries: stable partnerships with children are becoming more clearly associated with higher socioeconomic status (Härkönen and Dronkers 2006; Salvini and Vignoli 2011). Having greater socioeconomic resources can also help individuals cope when they are faced with unwanted and unexpected events, such as family dissolution. Finally, as a large body of literature on equivalence scales has shown, the economic disadvantages of single living arrangements relative to larger types of family structures stem from the lack of sharing of fixed costs, which can in turn lead to a lower living standard (e.g., De Santis and Maltagliati 2010).

2.3 Other Mediators

Other potent intervening factors can shape the association between LS and living arrangements. LS is an issue that cuts across different domains of life for every human being. It can vary by gender and age, be filtered by people’s health status and degree of religiosity, and is intrinsically embedded in the society in question.

A key factor is gender. Changing family structures influence women’s and men’s LS levels differently (Evans and Kelley 2006). The size of gender differences in LS varies according to societal gender inequality and cultural norms regarding gender equality (Tesch-Römer et al. 2008). In societies that accept gender inequalities, a woman’s identity as a wife and mother may be in conflict with certain non-traditional roles, such as being a participant in the labor market (Neyer et al. 2013). This can lead to gender gaps in LS, with women reporting lower mean levels of satisfaction (Tesch-Römer et al. 2008). By contrast, in countries that accept, welcome, and encourage gender equality, gender differences in LS seem to disappear (ibid.).

Age is traditionally studied as a factor correlated with LS. In older studies, age was not found to be significantly related to LS (Andrews and Withey 1976; Campbell et al. 1976; Michalos 1985). However, most recent literature in this field has indicated that satisfaction with life is generally U-shaped over the lifespan (e.g., Easterlin 2004). In addition to the purely demographic factors, subjective health status has been shown to be one of the most important predictors of LS (e.g., Mannel and Dupuis 1996).

Similarly, a large number of studies have shown that religiosity has a positive effect on subjective well-being and the ability to cope with setbacks (e.g., Lehrer 2004). For instance, Clark and Lelkes (2005) used the European Social Surveys to analyze the role of religion in “buffering the well-being impact of stressful life events.” They found that people who were religious had a higher level of subjective well-being and were more satisfied with life.

Another factor which may mediate the impact of different living arrangements on LS is the specific context of each country, which produces pros and cons associated with having children. The value of children may vary according to the type of society or the social group. For instance, depending on the cultural conditions, having (more) children may improve the social status of parents in some societies or social groups, while it may have no effect in others (Baranowska 2010). In addition to GDP per capita, researchers have looked at a number of macroeconomic variables (i.e., the overall unemployment rate) in seeking to understand the relationship between economic factors in the societal context and LS. Institutional factors, such as government expenditures (e.g., Bjørnskov et al. 2008), or, more generally, the quality of the government and welfare state (Pirani 2013), have been recognized as playing a role in well-being. Empirical research “unequivocally confirms the hypothesis that the welfare state influences individual well-being” (Pacek and Radcliff 2008: 272).

These considerations guided us in conducting an empirical investigation into the association between living arrangements and LS, stratified by gender. Moreover, all of the potential mediators of the LS-family status link mentioned so far were taken into account in our analytical strategy.

3 Empirical Approach

3.1 Data

The data used stem from the EQLS, which was carried out by the European Foundation for the Improvement of Living and Working Conditions from September 2007 to February 2008 on 15 countries that became Member States before 2004, the 12 countries that joined the EU in 2004 and 2007, plus Norway. About 1,000 persons aged 18 years and over were interviewed face-to-face in each country. Due to their larger populations, about 1,500 interviews were completed in France, Italy, Poland, and the UK; and approximately 2,000 were conducted in Germany. The questionnaire, which was developed by a research consortium, covers a broad spectrum of life domains, as well as the quality of a society and subjective well-being (Anderson et al. 2009). Large variations in national response rates were observed (less than 40 % in France, Greece, the Netherlands, and the UK; and more than 80 % in Bulgaria, Ireland, and Romania).Footnote 2 Obviously, low response rates may introduce sample selection bias, but we considered the average response rate of 58 % to be satisfactory.

The subjective assessment of LS was our variable of interest. Our approach acknowledged that each individual has his or her own ideas about what constitutes a “good life” (Naess 1999), and assumed that whether people believe they are living a good life can be measured by asking them how satisfied they are with their lives (Frey and Stutzer 2002). Using the response on a 1-10 scale as the dependent variable, we relied on the following question: “All things considered, how satisfied would you say you are with your life these days?” Our intent was to search for the crucial factors associated with LS, not to compare levels of LS between different areas.

The independent variable of greatest interest in this research was the household structure. An eight-category typology was developed specifically for this analysis: living alone, living with parent(s), living as a couple without children or other household members, living as a couple without children in an extended household (with other household members), living as a couple with children and without other household members, living as a couple with children in an extended household, living as a single parent without other extended household members, and living as a single parent in an extended household. It is important to note that, when developing the household structure, we did not differentiate between relatives and friends. We chose this approach for two reasons: first, a variable containing more typologies would have been unusable for analytical purposes; and, second, we were interested in various types of support. Usually, relatives are considered more obligated to help than friends. However, it can be argued that, because people can choose their friends but not their relatives, friends sometimes may prove to be more supportive (Ball 1983).

Due to data limitations, we could not account for marital status in our analysis. This means that we could not distinguish between married and cohabiting couples; or, when looking at people who did not have a partner, between single, widowed, separated, or divorced individuals. There are, however, well-established theoretical concepts explaining why marriage should improve LS more than cohabitation does (Cherlin 2004; Musick and Bumpass 2006; Evans and Kelley 2006). Thus, this is a limitation of our investigation.

The evaluation of LS varied depending on the living arrangements and the ages of respondents in all of the European countries studied. In order to provide a synthetic overview of these relationships, we illustrate in Fig. 1 the percentage of people who reported high LS levels (7-to-10) by grouping European countries into four relatively homogeneous areas. In each European area, living in a couple seems to have been the family typology associated with the highest levels of LS, regardless of age; whereas being a single parent was associated with low levels of satisfaction. Thus, at least initially, we can detect some degree of association between LS and living arrangements.

Fig. 1
figure 1

Percentage of people reporting a high LS level (7–10) by living arrangement, age group, and macro area of residence. Source Authors’ elaboration on EQLS 2007 data. Weighted distributions

Full size image

3.2 Method

Our data are naturally hierarchically structured (individuals living in different countries), and we believe that context matters in shaping the relationship between LS and living arrangements. Thus, the following analysis was based on multi-level modeling techniques (Goldstein 1987). Since the question relating to the assessment of LS is ordinal in nature, we applied an ordered logit multi-level regression model with random effects. This model allowed for the grouping of observations into countries. Since clustering is not an occasional nuisance, but an intrinsic characteristic of the population, it is explicitly considered in the model (Snijders and Bosker 1999). By adding a country-specific random effect to the predictor, this class of models explicitly introduces a hierarchical structure into the analysis, modeling unobserved heterogeneity and producing valid standard errors.

Let us consider individual i with i = 1,…,n j living in area j, with j = 1,2,…,J. A set of H individual variables Xij is collected for each individual, as is a set of M contextual variables Zj for each area. In the ordered logit model, there is an observed ordinal variable Y—i.e., LS evaluation in our analysis—which is in turn a function of an unobserved continuous response variable, Y*. The continuous latent variable Y* has various threshold points, and the value on the observed variable Y depends on whether or not Y* has crossed a particular threshold. For subject i, a threshold model determines the observed response:

$$ y_{i} = \left\{ {\begin{array}{*{20}c} 1 & {if} & {y_{i}^{ * } \le \alpha_{1} } \\ 2 & {if} & {\alpha_{1} < y_{i}^{ * } \le \alpha_{2} } \\ {} & \ldots & {} \\ C & {if} & {\alpha_{C - 1} < y_{i}^{ * } } \\ \end{array} } \right. $$

When we consider the logit transformation of the cumulative probability up to the c-th category for person i living in country j, the complete multi-level full model becomes:

$$ logit\{ P(Y_{ij} \le y_{c} )\} = \alpha_{c} - (\beta_{1} {\text{X}}_{ij} + \beta_{2} {\text{Z}}_{j} + u_{j} ){\text{ with c}} = 1,2, \ldots , \, C - 1 $$

where β1 and β2 represent, respectively, the slope parameters for the individual and the contextual covariates; while the parameters αc are the thresholds points. The uj are the random effects representing unobserved factors at the country level, “net” of the explanatory covariates introduced into the model. If the overall intercept is fixed to zero,Footnote 3 we can view uj as a random shift of the thresholds, so that the set of thresholds of country j isα c − u j with c = 1,2,…, C − 1 (Grilli and Rampichini 2012).

In order to test whether our data did or did not require a multilevel analysis, the so-called null model was estimated and yielded a positive answer. We then estimated two models with only first-(i.e., individual-) level variables, with the goal of finding out in what sense, and to what degree, they acted. Finally, we estimated the complete model, including both the first-level and the second-level variables so that we could interpret the variability associated with the context. A comparison between the different model specifications was made through the intra-class correlation coefficient (ICC), which expresses how much of the total unexplained variation in LS is due to the national context. The ICC was computed following standard practice (Snijders and Bosker 1999).

The analysis included 13,558 male and 17,928 female respondents nested in 28 countries. The minimum cluster sizes were 377 for men (Lithuania) and 490 for women (Romania), while the maximum cluster sizes were 916 for men and 1,091 for women (Germany). The unbalanced structure of the sample was efficiently handled by maximum likelihood methods (Snijders and Bosker 1999).

3.3 Model Specification

As was suggested by the theory, the analysis was stratified by gender. Apart from the null model, in this section three multi-level models of LS are illustrated. The overall composition of the sample is reported in the “Appendix”.

In the first model, only the living arrangements, the demographic variables, and the other variables described in the literature were included. In line with the literature, the demographics considered were the respondent’s age, the presence of children, and the age of the youngest child. Our age groups were 18–34, 35–49, 50–64, and 65+. The literature-driven control covariates were subjective health status and religiosity. Specifically, we included individuals’ perceptions of their own health status (good, fair, poor) and the frequency of their attendance at religious services (often, sometimes, rarely, never).

It is important to note that living arrangements can mean very different things during different life stages (McLanahan and Adams 1987). For instance, young parents tend to report higher levels of distress than young non-parents (Cleary and Mechanic 1983; Gore and Mangione 1983; Lovell-Troy 1983; Margolis and Myrskylä 2011; McLanahan and Adams 1987). Studies have found that, at older ages, there is either no difference between childless adults and parents in terms of LS (Connidis and McMullin 1993; Koropeckyj-Cox et al. 2007; Rempel 1985), or that parents display higher levels of well-being than their childless counterparts (Margolis and Myrskylä 2011). Given the general objective of our study, however, we did not confine the analysis to a specific age group. Separate models were run according to age groups (the results are available from the authors upon request). The low number of observations for certain cells did not allow us to present results in a disaggregated way, but they were helpful in explaining and interpreting findings for the overall population. Importantly, our findings remained virtually unchanged if stratified by age.

The second model specification included the current and perceived affluence of the family in order to evaluate the role of individual and contextual socioeconomic factors in the relationship between family structure and LS. The adopted indicators of individual socioeconomic status were the highest level of education attained by the respondent (lower-secondary education, upper- and post-secondary education, tertiary or higher education), the respondent’s (and, where applicable, the partner’s) employment status (unemployed, inactive or housewife, employed), and the respondent’s perception of his or her financial situation (good, medium, bad).

Finally, in the third model specification, the country context was accounted for based on secondary national-level indicators referring to the year 2007 and collected from Eurostat (Eurostat 2011). First, we considered each country’s net social protection benefit in the family and children function. This indicator was derived from gross social benefits, less the levies (taxes and social contributions) paid on cash social benefits by the recipients, plus the (where relevant) residual fiscal benefits. Second, each country’s full unemployment benefit was adopted as an indicator of the labor market; that is, the benefits that compensate for a loss of earnings.Footnote 4

4 Results

Looking at model 1 in Tables 1 and 2, it is easy to see that family structure is linked with LS. The tables show the odd ratios (ORs) of coefficients estimated for the hierarchical models for life satisfaction assessment, with the 28 European countries representing the second-level units. The OR, which simply corresponds to exp(β), expresses whether a higher assessment of life satisfaction is likely to occur in a certain group with respect to the reference group.

Table 1 Correlates of life satisfaction: outcomes from step-wise multilevel ordered logistic models estimated for men
Full size table
Table 2 Correlates of life satisfaction: outcomes from step-wise multilevel ordered logistic models estimated for women
Full size table

The less traditional families—i.e., families that were not made up exclusively of a couple with their children—displayed lower levels of LS. Single mothers, single fathers, and members of extended families seemed to be the least satisfied with their lives. In this specification, men who were living in couple without children were more satisfied with their lives than men living in couple with children (at a level of 10 %, Table 1).Footnote 5 The LS levels of neither men nor women seemed to be related to the presence of children under age 13 in the household. However, while the number of children did not appear to matter, the age of the youngest child was found to be relevant, but in different ways for men and for women. Men showed an increase in LS in the presence of a newborn (two years old or under), while women reported lower LS in the presence of a child aged three or more.

In this model specification, an age differentiation was also found, with LS decreasing as individuals went through different phases of life. The other individual variables that have been suggested as relevant in the literature helped us paint a more complete picture of the personal characteristics associated with high LS. There also appears to be a religiosity gradient, with LS increasing along with a greater frequency of attendance at religious services. Finally, as anticipated, a report of poor health was among the most relevant predictors of lower levels of LS.

After the various facets of socioeconomic status were controlled for, the picture changed (model 2 in Tables 1, 2), with almost all of the family typologies considered losing their significance in explaining differences in LS. In this model specification, the presence in the household of individuals other than the respondent’s partner and children was found to be associated with a reduction in LS among men only, all other things being equal.Footnote 6 By contrast, being a single parent with no support (within the household) from other members of the family was found to be associated with lower LS levels among women: single mothers reported being 33 % less satisfied than partnered mothers. Considering the reduction in the OR of single mothers after their socioeconomic situation had been accounted for, a large share of the LS disadvantages associated with this family form can be explained by lower levels of family affluence.

In addition, outcomes of model 2 suggest that, among women, the presence of three or more children in the household increased LS by about 26 % (at 10 % significance level); while among men, the presence of a newborn was linked to an increase in LS of about 50 %. It should also be noted that the association with age changed after we controlled the models for a complete set of individual socioeconomic characteristics (model 2). Young adult men were no longer in the most satisfied category, and the elderly were classified as more satisfied than other age groups. For women, this result had already emerged in the simpler model (model 1).

When we look at the socioeconomic variables, we can see that, while LS levels were highest among young people in schooling, LS generally increased with higher levels of educational attainment. As expected, unemployed people reported feeling less satisfied than employed people. Accordingly, a higher rating of one’s own economic and financial situation was found to be positively associated with LS, and vice versa. The rather gender-equal pattern found for the importance of one’s own employment for LS was not replicated regarding the importance of partners’ employment status for men and women. Interestingly, a gendered effect of the partner’s labor market status on LS clearly emerged: women reported lower LS when their male partner was unemployed, while the occupational status of the female partner was not significant for men.

The second specification also controlled for two country-level variables. Although we recognize that the strength of the association between family status and LS changes when the socioeconomic position of the family is taken into account, we believe that the role of the broader socioeconomic context (i.e., the country in which the respondents lived), which does not change the estimation of individual correlates, is also important, from both a conceptual and a methodological point of view. The model results showed that higher social protection benefit levels—in the areas of family and children and of unemployment—were associated with higher levels of LS of individuals.

We carried out a multilevel analysis, explicitly accounting for the fact that individuals lived in different countries. The presence of a significantly high estimated country-level variance τ 2 in the multilevel null model (model 0 in Table 3) proved that multilevel modeling performs better than single-level modeling. The estimated ICC—i.e., the percentage of total unexplained variation in LS assessment among European respondents based on their country of residence—was about 16 % for the null model (similar values are estimated for both genders). Passing from the two-level null model to the two-level model adjusted for individual level covariates, the area-level variance τ 2 fell to about 10 % in the complete men and women models (model 2). This decrease shows that a certain part of the geographical variability was actually due to the differing social profiles of the resident populations. After the two country-specific covariates were added, the total unexplained variation in LS attributable to the context further declined to about 5 % for both genders.

Table 3 Territorial variability and ICC: model comparison
Full size table

5 Conclusions

Existing research on family and well-being has generally focused on the traditional family, and has largely ignored the increasing diversity in family forms and relations (Uhlendorff et al. 2011). However, in recent decades, the family life course has become less and less standardized all over Europe, resulting in an increase in previously rare forms of family life, such as lone parenthood and extended families. Our paper contributes to this ongoing debate, investigating the extent of the relationship between diverse patterns of living arrangements and LS in contemporary Europe.

Our findings suggest that levels of LS among people in families consisting of couples with children are significantly higher than among people in other family arrangements; i.e., families headed by single mothers or fathers, or extended families. However, after the socioeconomic situation of the family has been taken into account, the association between family status and LS vanishes almost completely. The positive gradient found between family socioeconomic status and LS is consistent with prior research (Amato 2000; Breivik and Olweus 2006; DeBell 2008). In addition, our results indicate that the lower levels of life satisfaction experienced by people living in atypical families can be essentially attributed to their weaker socioeconomic position. This finding suggests that living in a non-traditional family is a symptom of lower socioeconomic status, and is not a factor necessarily associated with low LS.

Another interesting finding of our large-scale cross-national analysis is that LS appears to be significantly influenced by the specific living context. Nevertheless, compositional effects account for a large share of the variation in LS. In other words, our study suggests that, after individuals’ living arrangements and demographic characteristics have been controlled for and their socioeconomic situations have been taken into account, there is considerably less room left for macro explanations regarding variation in LS across Europe. Only about 5 % of the total variation in LS across Europe could be attributed to the specific country of residence, after the individual-level predictors were aligned with the country-level covariates.

Although this study provides important insights into the relationship between LS and family status, it has limited power to provide us with information about causal relationships. It is important to note that none of the presented analyses account for the fact that individuals who have an innate predisposition to report a higher level of LS may also systematically vary in their propensity to form unions. We cannot exclude the possibility that unobserved universal traits which drive selection into partnership and simultaneously improve LS may have partly affected our outcomes. However, our goal in this paper was to describe the degree of association between family forms and LS across Europe. A descriptive study is a necessary first step; future efforts should be directed at verifying the associations evoked in this cross-sectional research through the use of panel data and the adoption of causal approaches.

Importantly, our findings suggest that being in an atypical living arrangement is not a condition linked per se to a lower level of LS; instead, it appears likely that the (often) more fragile socioeconomic position of people in non-traditional families has the greatest effect on their levels of satisfaction with life. Thus, we believe this analysis raises important questions about family change and individuals’ well-being in post-industrial societies, because even if the trend toward “new family forms” comes to a halt, a return to a traditional family model is unlikely.