1Births outside marriage are considered differently from one country to another. While they are frequent in certain European countries such as France, where almost 60% of the children born in 2016 had unmarried parents, they are still rare in others, where marital childbearing remains the norm. For this reason, the landscape of unmarried cohabiting couples in Europe is highly diverse. In this article, Zuzana Žilinčikova examines the risk of non-marital union dissolution in comparison with marriage. Is the presence of children a stabilizing factor? If so, is the effect stronger for cohabiting unions than for marriages? And does the impact differ according to the prevalence of cohabitation in the country concerned? Using retrospective data from the international Generations and Gender survey, the author compares the risk of dissolution of unions formed since the 1990s in 14 European countries by type of union and the presence or not of children. She reveals points of convergence and divergence in conjugal behaviour between the different countries under study.
2“The substantial rise in unmarried cohabitation… constitutes a hallmark of the ongoing changes in family life in most developed countries” (Sobotka and Toulemon, 2008, p. 97). Indeed, the spread of unmarried cohabitation in western countries is well documented by recent research (Manning, 2013). Cohabitation has become a normative start to a union in many countries and is increasingly becoming an acceptable context for childbearing (Musick, 2007; Sobotka and Toulemon, 2008), as illustrated by the steep rise in childbearing within cohabiting unions in recent decades (Kennedy and Bumpass, 2008; Sobotka and Toulemon, 2008).
3Although recent research deals extensively with the topic of cohabitation and its implications for different areas of life, there are still many unanswered questions. One largely unexplored topic is cohabitation dissolution (Amato, 2000, 2010; Graefe and Lichter, 1999; Kalmijn et al., 2007; Smock, 2000), which is nonetheless crucial, as we know that cohabiting unions are highly unstable and the number of dissolutions is rising (Clarke and Jensen, 2004; Lichter et al., 2006). As yet, no study has provided a satisfactory picture of cohabitation dissolution across European countries, despite its importance for understanding the extent to which union instability affects partnership and family life in Europe.
4While some studies have partly investigated the dissolution of cohabitation, they do not account sufficiently for the heterogeneity of cohabiting unions. Of particular interest are those that include one or more children. Past research has clearly shown that family stability is a key aspect influencing child development (Hao and Xie, 2002; Hill et al., 2001). Moreover, the separation of a couple with a child differs qualitatively from that of a childless couple in its course and consequences (Koo et al., 1984). It is therefore important to ask how stable these unions are, relative to both childless cohabitation and marital unions. International comparison provides a complex picture of cohabitation stability across European countries.
5This article aims to extend current knowledge in two ways: first, to provide cross-national estimates of dissolution rates of cohabiting unions (formed after 1990) with and without a child present; and second, to investigate the stability of cohabiting unions relative to marriage and distinguished by the presence of a child, across European countries. These questions are addressed using data from Generations and Gender Surveys for 14 European countries.
I – What do we know about cohabitation dissolution?
6As already outlined, cohabiting unions are generally unstable and shortlived (Andersson and Philipov, 2002; Lichter et al., 2006; Liefbroer and Dourleijn, 2006). There are several estimates of dissolution rates for cohabiters, but they vary depending on the definition of cohabitation and the country. In the United States, Kamp Dush (2011) finds that 64% of parents who were cohabiting at the birth of a child had separated by the fifth year; however, of these, only 76% of the respective relationships broke up completely, with the rest continuing in an informal non-cohabiting manner. Slightly lower estimates for all cohabiting unions are presented by Lichter, Qian and Mellott (2006), who observed that 46% of cohabiting unions were dissolved within five years.
7Research conducted in the 2000s suggests that dissolution rates for cohabiting unions are generally higher in the United States than in Europe (Andersson, 2003, 2004; Andersson and Philipov, 2002). According to an analysis by Andersson and Philipov (2002), 69% of cohabiting unions (censoring at marriage formation) end in dissolution by the fifth year in the United States, while in European countries the proportions are much lower, ranging between 7% in Poland and 56% in Latvia. Findings are similar concerning the instability of cohabiting couples who have a child together. The share of unions (regardless of subsequent transition to marriage) that have dissolved by the time a child is 6 years old varies between 7% in Poland and 51% in Latvia. In the United States, the figure is 56%.
8It is thus very important to take account of cohabitation dissolution in studies of instability in families which include children. Indeed, in the United States, accounting for transitions to and from marriage but not to and from cohabitation leads to the omission of 23% of overall instability affecting white children and 53% affecting black children in that country (Raley and Wildsmith, 2004).
II – Existing research on cohabiting unions
9Existing country-specific research that primarily focuses on cohabitation dissolution is scarce and is found outside the European context. Moreover, the sample is often selective for a certain type of cohabitation (Kamp Dush, 2011; Wu and Balakrishman, 1995), or the different types of cohabiting unions are not distinguished (Lichter et al., 2006). Cross-national studies do not focus primarily on the dissolution of cohabitation but present it as part of a different or broader topic. There are two distinct approaches in this area. The first focuses on relationship status – cohabiting, married with prior cohabitation, married without cohabitation at the moment of dissolution (Kiernan, 2002; Liefbroer and Dourleijn, 2006) – but works only with first partnerships. The second investigates the type of union at the beginning of a relationship or at the moment of childbirth, regardless of the situation (separation or marriage) at the moment of dissolution (Andersson, 2003, 2004). Only Andersson and Philipov (2002) use a mixture of these two approaches to investigate the stability of cohabiting unions after childbirth, with censoring at marriage formation. However, again, they do not cover all cohabitation types, especially higher order and childless cohabiting unions. This article aims to provide a more systematic cross-national exploration of the stability of more recent cohabitating unions in Europe. It also extends current knowledge by distinguishing cohabitation with and without a child present and compares it to equivalent marital unions, thus providing a broader perspective on dissolution rates.
1 – Childbearing within cohabitation
10Although childbearing in cohabiting unions in the European context varies greatly across countries, the trend is very similar, with an increasing number of children being born or raised in cohabiting unions (Hiekel, 2014; Perelli-Harris et al., 2010; Sobotka and Toulemon, 2008). Most striking, however, is the fact that the vast majority of children in cohabiting unions will experience family transition at some point in their lives (Andersson, 2004). Recent evidence shows that cohabiting unions with children are highly unstable relative to married families and these findings are constant across a wide range of countries (Andersson, 2004; Andersson and Philipov, 2002; Clarke and Jensen, 2004; Heuveline et al., 2003; Jensen and Clausen, 2003). These findings are usually explained by selection of less dissolution-prone individuals to marriage (Axinn and Thornton, 1992; Clarkberg, Stolzenberg and Waite, 1995), or by the less binding character of cohabitation in comparison to marriage. The selectivity argument suggests that married individuals (and especially those with children) are more committed (Nock, 1995; Stanley et al., 2004), have more satisfying relationship quality (Brown and Booth, 1996; Wiik et al., 2009) and are more family-oriented (Lesthaeghe and Moors, 1995). The second argument is based on the fact that cohabitation is a less institutionalized relationship, and poses fewer barriers to dissolution. Consequently, couples with the same characteristics are potentially able to separate more easily and quickly if they are cohabiting rather than married. We therefore posit that cohabiting unions which involve a child are less stable than marriages with a child and that this should be constant across all national contexts.
11According to the economic theory of marriage, children are a form of joint investment by the partners (Becker et al., 1977). This is, of course, true for both marital and cohabiting unions, although the returns of such investment might be higher in marriage. First, cohabiting parents, and more especially fathers, do not enjoy the same legal rights with respect to their children as married fathers in most European countries (Perelli-Harris and Gassen, 2012); and second, children born to cohabiting partners are more often unplanned than in marriage (Wellings et al., 2013). Both factors may play a role in commitment towards the family, but also in the decisions about union dissolution.
12There is also mixed empirical evidence on the stabilizing effect of children in cohabiting unions. Some researchers have found that childbirth within cohabitation, in contrast to that within marriage, does not stabilize the union in the United States (Manning, 2004) and Great Britain (Berrington, 2001; Boheim and Ermisch, 2001). Conversely, childbirth has been found to have a stabilizing effect on cohabiting unions in Norway (Poortman and Lyngstad, 2007), The Netherlands (Manting, 1994), and in the United States according to more recent research (Guzzo, 2014). These contradictory results suggest that there is considerable variation between countries in the effect of child presence in cohabitation, which might be related to different childbearing rates among cohabiting couples in different national contexts. In countries where non-marital childbearing is less common and less well accepted, it is more likely that cohabiting parents are more selected in characteristics that might increase the risk of partnership break up.
13The other, less prominent, line of research suggests that the difference in the risk of dissolution between married childless women and married mothers diminishes over time (Andersson, 1997). Lyngstad and Jalovaara (2010) suggest that this could be caused by two trends. The first is the increased selection for marriage in the case of childless couples; the second is the weakening norm against divorce for couples with a child.
14Although the theoretical explanations are contradictory, the empirical evidence suggests that cohabiting unions with a child are in general more stable than childless ones, but the difference should nevertheless be smaller than for marriage.
2 – Cross-national differences in cohabitation
15As suggested by previous research, the meaning of cohabitation, union duration and childbearing differ across the countries, so the findings of country-specific research are difficult to generalize. A cross-national comparison provides a complex picture of the stability of unions in Europe and also enables us to grasp differences in the stability of the various cohabitation types, at least at the country level. For example, in Sweden and France, cohabitation is widespread and childbearing in these unions is very common and accepted. Cohabitation in these countries is most commonly considered to be an alternative to, or indistinguishable from, marriage (Heuveline and Timberlake, 2004; Hiekel et al., 2014; Kasearu and Kutsar, 2011). In other western European countries present in the sample (Austria, Belgium and Germany) cohabitation is also very common, although childbearing is still predominantly tied to marriage. Cohabiting unions in these countries are classified by most scholars as a stage in the marital process (Heuveline and Timberlake, 2004; Hiekel, 2014; Kiernan, 2001). Lastly, cohabitation is also becoming a norm at the start of a relationship in central and eastern European countries, although it is less prevalent and shorter in duration than in western European countries. In general, childbearing in such unions is also less frequent, although there are quite extensive national differences in this respect (Hiekel, 2014; Kasearu and Kutsar, 2011). Moreover, cohabiting unions might be selective of individuals who are “too poor to marry” (Hiekel et al., 2014).
16Therefore, the stability of cohabitation relative to marriage is likely to depend on the character of cohabitation in a given country. Cohabitation with a child present should be more stable in countries where it often serves as an alternative to marriage and as an acceptable context for childbearing (especially in Sweden, but also in other western European countries). Moreover, in countries where cohabitation is widespread and non-marital childbearing is widely accepted, cohabitation is more often considered by parents as a sort of marriage-like commitment, whereas this is less the case elsewhere. Therefore, we expect the difference in the stability of unions with and without a child present to be greater for cohabitation than for marriage, in particular in the observed western European countries and in Sweden.
III – Data and sample
17This article uses data from the first wave of GGS (Generations and Gender Survey) conducted in 16 mostly European countries: Austria, Belgium, Bulgaria, Czech Republic, Estonia, France, Georgia, Germany, Hungary, Lithuania, Poland, Romania, Russia and Sweden.  Respondents were 18-79 years old and approximately 10,000 individuals in each country were interviewed in each wave.
18The information about unions was drawn from the respondents’ retrospective partnership histories. In all countries, a maximum of five unions per respondent were recorded, even if some respondents in some countries reported more than five. The retrospective dataset provided limited information on the respondents’ life histories.  However, there was detailed information on the biological children of each respondent, and on their educational history, which could be used as a proxy for socio-economic status. Controlling for other characteristics of each respondent and his/her past partners would have severely reduced the sample size and the number of countries, thus contradicting the objective of this study.
19The sample consisted of women who were at least 18 years old. It was also limited to unions that began after 1990. There were two reasons for this restriction. First, this research was more interested in the mapping of more recent trends rather than historical development; second, this limitation deals, at least partially, with the issue of reliability in the reporting of less recent life events. 
20The sample consisted of 22,442 individuals who had experienced a total of 25,458 unions. Almost 62% of these unions began as cohabiting unions, of which 45% became marriages; 38% of unions started as direct marriages. Further, 75% of unions were first unions, and the majority of unions included at least one child at some point in their duration (72%).
IV – Methods and measurements
21Event history analysis was employed to estimate a set of discrete-time logistic models. For this purpose, the original data file was transformed to the person-union format, where one line constituted one union with the same partner. The measurement of duration was changed from months to half-years, considering the long average union duration. The mean duration of all unions was 11.5 years. The half-years were then used to expand the dataset in order to create a union-period format which consisted of 414,150 union-periods.
22The main dependent variable was the event of union dissolution. The partnership was observed from the beginning of cohabitation to dissolution as defined by the respondent (or until the death of a partner or the interview in case of censoring).
23The main independent variable was child presence, a time-varying variable that indicates whether there was at least one biological child living in the household at a given time point. The information on children was restricted to the respondent’s biological children. Respondents with adopted and foster children were omitted, as these types of parents constituted a specific group. Unfortunately, there were no full reports of past partners’ children in the GGS data, so the analysis does not account for stepchildren. However, insofar as we tracked only women, they were much more likely to be the ones who brought their own children to their new partnership (Bernhardt and Goldscheider, 2002; Goldscheider and Sassler, 2006).
24All models control for age at the start of the union, age at the start of the union squared, cohort, and country. Further, a dummy variable indicating parental status before the union was introduced to the models to account for the respondent’s child or children born before the particular union formation. The variable might refer either to a child with a current partner who was born outside of the co-resident partnership, or to a child with a previous partner. Unions were further distinguished by their order. A union was classified as first if it was not preceded by cohabitation or marriage with a different partner. All other unions were considered as second or higher order unions. Marriage that followed cohabitation with the same partner was considered to be the same union. Two time-varying covariates were also included: union type (marriage or cohabitation) and education. Union type was a dummy variable indicating whether a couple was married (1) or cohabiting (0). Education was measured in years up to completion to the highest level (for more information see Dourleijn et al., 2002). Unions with missing information on the dependent (1.3%) and independent variables (7.6%) were dropped from the analysis.
25The first step was to estimate a baseline model. The best fit for the dissolution hazard was achieved by introducing categorical specifications of duration (with seven categories). This duration specification was included in all the presented models. Altogether, three models were estimated. Model M1 included all independent variables as presented above; Model M2 further included a new variable, generated as the combination of child presence and union type; and Model M3 added the interaction between this newly generated variable and the country.
26The results section is organized as follows. First, we present the proportion of different union types at the moment of union formation and their dissolution rates by the fifth year since union formation. Descriptive analysis for dissolution rates is based on a slightly different number of unions (19,193), as only unions formed at least five years before the interview are taken into account. Second, we present results from the discrete-time logistic models, M1, M2 and M3.
V – Results
1 – Proportions of different union types and of dissolutions in European countries
27Four broad categories of countries were distinguished. Table 1 presents the proportion of union types at the first measured time interval (i.e. after six months)  and mean age at union formation. The first group (Belgium, France, Sweden) and second group (Austria and Germany) of countries are characterized by a high prevalence of childless cohabiting unions (63-76%), but also a relatively high prevalence of cohabiting unions with a child present (10-19%). The proportion of unions starting as marriage or which transitioned to marriage within the first six months of shared living is quite low, and the mean age at entry into union varies between 24 and 28 years. The third group, consisting of central and eastern European countries (Czech Republic, Estonia, Hungary and Russia), is quite heterogeneous in terms of the prevalence of one or other type of partnership. In all of these countries we observe moderate prevalence of childless cohabiting unions (37-59%) relative to other countries and a relatively high prevalence of cohabiting unions with a child (13-27%). In this respect, this group of countries is closer to western European countries than to other central and eastern European countries in the last group. The mean age at entry into union varies between 25 and 26 years. The countries in the last group (Bulgaria, Georgia, Lithuania, Poland and Romania) are the most traditional, with the vast majority of unions starting as childless cohabiting unions (25-56%) or childless marriages (39-62%), and the individuals are the youngest (23-25 years old) at the moment of union formation.
Prevalence of union types and mean age observed at the first time interval (half year) after union formation (N=25,458)
|Country||Childless cohabitation||Cohabitation with a child||Childless marriage||Marriage with a child||Mean age at union formation (SD)|
|Czech Republic||41.2||17.9||33.7||7.2||25.7 (8.2)|
Prevalence of union types and mean age observed at the first time interval (half year) after union formation (N=25,458)
28Table 2 presents the percentages of dissolved unions among cohabiting unions, marriages and direct marriages with or without a child, within five years of union formation.  Across all countries, we observe a similar trend – although there is substantial cross-national variation of dissolution rates within the union types. Cohabiting unions are less stable than marriages, and childless unions are less stable than unions with a child (with the exception of Germany for marriages following cohabitation). The first group of countries (Belgium, France and Sweden) has a high rate of dissolution for childless cohabiting unions (41-54%) but relatively stable cohabiting unions with a child present (12-18%). The proportion of dissolved childless cohabiting unions in the second group of countries, Austria and Germany, is somewhat smaller (38-40%), but the stability of other union types is quite similar to that in the first group. The third group of countries, consisting of less traditional central and eastern European countries, is characterized by high dissolution rates of most of the union types relative to other countries, although the frequency of dissolution of childless cohabiting unions is not much greater than in the first group of countries (33-52%). On the other hand, the last group including traditional central and eastern European countries is characterized in most cases by high stability of all union types relative to all other countries.
Percentage of unions by type and child presence that were dissolved by the fifth year after union formation (N=19,193)
Percentage of unions by type and child presence that were dissolved by the fifth year after union formation (N=19,193)* Category with fewer than 20 observations.
Note: Only unions formed at least five years before an interview are included in the calculations.
2 – Discrete time models
29We started by estimating the basic model M1 (Table 3), i.e. one without any interaction terms, in order to observe the main effects of the key independent variables and controls. The presence of a child reduces the odds of dissolution by almost 40% and marriage reduces the risk even more strongly (66%). On the other hand, being in a second (or higher order) union slightly increases the chances of dissolution. More years spent in education increases the odds of dissolution, and more recent cohorts were also more likely to dissolve. From the duration specification, it is clear that the lowest odds of dissolution are in the first half-year of union duration. They then increase dramatically for the first ten years before decreasing steadily over the following periods. Having a child prior to union entrance is also associated with a higher risk of union dissolution. The effect of age at the start of cohabitation is non-linear as both linear and quadratic specifications are significant. The linear effect reveals that union stability also increases with increasing age at entry into co-resident partnership. The quadratic effect, however, indicates that the magnitude of this stabilizing age-effect diminishes in higher age groups. The effects of individual countries show, in accordance with the descriptive results, that the risk of dissolution is highest in the Czech Republic, Hungary and Russia, and lowest in Bulgaria, Georgia, Poland and Romania. The risk of dissolution is also relatively high in western European countries, namely Belgium and France, and somewhat lower in Austria and Germany and Sweden.
30The second estimated model M2 (Table 3) includes a variable generated as a combination of two time-varying variables: type of union (marriage/cohabitation) and presence of a child (not present/present), resulting in a new variable with four categories: childless cohabitation (reference category), cohabitation with a child present, childless marriage, and marriage with a child present. Model 2 then shows that childless cohabitation carries the highest risk of dissolution, followed by cohabitation that includes a child, childless marriage, and marriage that includes a child. Marriage with a child present is significantly more stable (odds ratio = 0.742, p = 0.000) relative to childless marriage, although the difference is less pronounced than in the case of cohabitation (odds ratio = 0.544, p = 0.000).  Model M3 (Table 3) adds the interaction between the newly generated variable in M2 and the country identifier. Table 3 shows only the significant interactions; all the results are presented in graphic form in Figures 1 and 2. Figure 1 presents the odds of dissolution for different types of unions distinguished by the presence of a child. Sweden and childless cohabitation are the reference categories. The graph indicates that the pattern of stability is more or less the same across all countries, with childless cohabiting unions being the least stable unions and marriages with children the most stable. Cohabiting unions with a child present are significantly more stable than childless cohabiting unions in ten out of 14 countries (with the exception of the Czech Republic, Georgia, Lithuania and Romania, where the difference was not significant). Marriages with a child were significantly more stable than childless marriages in only six countries (Austria, Belgium, Bulgaria, Hungary, Russia and Romania).
31Figure 2 explores the difference according to child presence within the union types, and thus makes it possible to compare the role of children in cohabitation and marriage more directly. This type of graphic presentation also provides better comparison between the countries. The difference in the stability of cohabiting unions with and without a child was most pronounced in Sweden (reduction of almost 70%). A large difference was also observed in Belgium (55%), Bulgaria (53%) and Austria (42%). For all other countries (with significant results), the difference in the odds of dissolution between cohabiting unions with a child present and childless cohabiting unions ranged between 34% and 24%. In the Czech Republic, Lithuania and Romania no significant differences were found, although the trend is in the same direction as for the other countries. In Georgia the difference is non-significant and in the opposite direction.
32The difference for marital unions with and without a child present are distinctly less pronounced and, in the majority of the countries, non-significant. However, the general trend also points to the greater stability of marriages with a child present. Consequently, in some of the countries, the difference in the stability of unions with a child present relative to childless ones was greater for cohabiting unions than for marriages. In this respect, the biggest difference between cohabiting unions and marriages was observed in Sweden. Similar, although smaller, differences were observed in Belgium, Germany and Poland. In contrast to the large difference in stability of cohabiting unions between childless couples and parents, there was essentially no difference in the stability linked to the presence of children in marital unions. By contrast, the relationship was in the opposite direction in Romania, as the difference in the odds of dissolution was higher for marriages than for cohabiting unions.
Odds ratios of union dissolution estimated from discrete-time models
|Child present (ref. = no child present)||0.606***|
|Marriage (ref. = cohabitation)||0.340***|
|Education (in years)||1.016***||1.015***||1.017***|
|Cohort 1951-60 (ref. = 1921-50)||2.057***||2.057***||2.097***|
|Categorical duration specification 2-5 semesters (ref. = 1)||4.886***||4.896***||4.836***|
|Birth of a child before union entry (ref. = childless before union entry)||1.330***||1.374***||1.342***|
|Age at start of the union in years||0.897***||0.900***||0.900***|
|Age at start of the union in years squared||1.001***||1.001***||1.001***|
|Second and higher order union (ref.= first)||1.289***||1.293***||1.276***|
|Belgium (ref. = Sweden)||1.372***||1.367***||1.293***|
|Cohabitation with a child present (ref. = childless cohabitation)||0.544***||0.313***||0.313***|
|Marriage with a child present||0.210***||0.177***|
|Belgium*Cohabitation with a child present||1.405*|
|France*Cohabitation with a child present||2.384***|
|Austria*Cohabitation with a child present||1.842***|
|Germany*Cohabitation with a child present||2.198***|
|Germany*Marriage with a child present||1.500*|
|Czech Republic*Cohabitation with a child present||2.457***|
|Czech Republic*Childless marriage||2.399**|
|Czech Republic*Marriage with a child present||1.787**|
|Estonia*Cohabitation with a child present||2.098***|
|Estonia*Marriage with a child present||2.565***|
|Hungary*Cohabitation with a child present||2.119***|
|Russia*Cohabitation with a child present||2.120***|
|Russia*Marriage with a child present||1.558**|
|Georgia*Cohabitation with a child present||3.622**|
|Lithuania*Cohabitation with a child present||2.129**|
|Poland*Cohabitation with a child present||2.080***|
|Romania*Cohabitation with a child present||2.511**|
Odds ratios of union dissolution estimated from discrete-time modelsSignificance levels: †p < 0.10, *p < 0.05, **p < 0.01, ***p < 0.001.
Note: Only significant interactions are presented. Number of individuals = 22,442; number of unions = 25,458; number of union-periods = 414,150.
Odds ratio of union dissolution calculated from Model M3 which includes interaction between type of union, child presence, and country
Odds ratio of union dissolution calculated from Model M3 which includes interaction between type of union, child presence, and countryInterpretation: Example of a calculation of the odds ratios (France, cohabitation with a child present): 0.949 x 0.313 x 2.384= 0.709.
Difference in the odds of dissolution according to child presence for cohabitation and marriage calculated from Model M3 which includes interaction between type of union, child presence, and country
Difference in the odds of dissolution according to child presence for cohabitation and marriage calculated from Model M3 which includes interaction between type of union, child presence, and countryInterpretation: Example of a calculation of the odds ratios (France, difference in the dissolution odds between cohabitation with and without a child present): 0.949/(0.949 x 0.313 x 2.384)=0.747.
33This paper aimed to map dissolution rates across Europe, distinguishing childless cohabiting unions and those with a child present. It compared the dissolution rates of these different cohabitation types across countries and contrasted them to marriage in order to obtain a broader picture of their stability. Indeed, the results show that the stability of unions varies across countries, with Russia and Hungary having the highest dissolution rates and Georgia, Bulgaria and Romania the lowest. Dissolution rates in all other European countries are roughly similar.
34The results confirm that cohabiting unions are less stable than marriages, and unions including a child are more stable than those without a child present. Moreover, the comparison of different types of unions, with or without a child present, confirms results from previous studies, i.e. that in all countries cohabiting unions with a child present are less stable than marriages with a child present (Andersson, 2004; Andersson and Philipov, 2002; Clarke and Jensen, 2004; Heuvelineet al., 2003; Jensen and Clausen, 2003; Wu and Musick, 2008), but also less stable than childless marriages. This proved to be constant in the wide range of European countries studied.
35Nevertheless, the comparison of unions with and without the presence of a child within cohabitation and marriage provides interesting insight into the nuances of the pattern. Cohabiting unions with a child present were found to be more stable than childless cohabiting unions in ten of the 14 countries studied, in line with the more recent studies (Guzzo, 2014; Poortman and Lyngstad, 2007), thus confirming the stabilizing effect of child presence, in contrast with studies that found no effect (Berrington, 2001; Boheim and Ermisch, 2001; Manning, 2004).
36Comparing the difference in union stability according to child presence showed that in four countries (Belgium, Germany, Poland and Sweden) the difference in stability is greater for cohabitation than marriage, and the effect is reversed in only one country (Romania). These findings, therefore, provide limited evidence for the expectation that child presence matters more for the stability of marriage. In contrast, the results tend to support the findings of Andersson (1997), suggesting that the difference between childless women and mothers is narrowing in marriage but widening for cohabiters. This reflects the fact that cohabiting unions may be stable and long-term unions, for parents especially.
37We also posited that the difference in stability between cohabiting unions with and without a child would be greater in countries where cohabitation often serves as an alternative to marriage and is an acceptable context for childbearing. This paper indeed finds that the difference is on average greater and more pronounced for cohabitation than marriage in western European countries in comparison to those of central and eastern Europe. A particularly large difference in stability between cohabiting unions with a child present and childless cohabiting unions (and in comparison to marriage) was found in Sweden. A similar trend was observed in Belgium and to a certain extent also in Germany and Austria. Contrary to expectations, this difference in union stability with and without a child present is not more pronounced for cohabiting unions than for marriages in France, which is also a country with a long history of cohabitation and wide acceptance of childbearing within these unions (Heuveline and Timberlake, 2004). This invites at least two possible explanations. First, Sweden is usually marked as a forerunner of the changes in partnership behaviour (Lesthaeghe, 2010), and this might also be the case for the stability of cohabiting unions with a child present. It would suggest that the difference in stability of cohabitation with and without children will become more notable in the future. On the other hand, cohabiting unions with a child present, or at least a large share of such unions, might represent a special case in France. As suggested by the results of Heuveline and Timberlake (2004), relatively high instability of cohabiting unions with a child present in France may reflect the fact that cohabiting parents remain in this situation for very long periods and only about a quarter of them go on to marry.
38The difference between the stability of cohabiting unions with and without a child present is smaller and more often non-significant in central and eastern European countries, indicating that cohabiting women with children do not live in more stable relationships than their childless counterparts and that the unions are more similar in terms of stability. However, in the context of overall stability, this has a different meaning for the traditional countries (Georgia, Bulgaria and Romania) where cohabiting unions are relatively stable, and in countries such as Hungary and Russia where the odds of dissolution are very high for cohabitation both with or without a child.
39Overall, the results seem to indicate that cohabiting unions in western European countries are more diverse in terms of commitment – with childless cohabiting unions being highly unstable, but cohabiting unions with children comparatively less dissolution prone, whereas in central and eastern European countries they seem to be more alike, both with or without children present. This conclusion is in line with the findings of previous studies that have observed higher prevalence of marriage-like cohabiting unions in western European countries than in central and eastern European countries (Hiekel, 2014).
40A few limitations of this article need to be addressed. The research employed retrospective partnership data which are always affected by several biases. The first is the reliability of individual reports about past events. This drawback was tackled by using only the more recent partnership histories to minimize possible misreports. Second, the number of variables for which we have retrospective information is limited. For example, information on step-children or employment histories was not recorded in a number of the national questionnaires. Finally, the article is not able to make inferences about the effect of child presence, but instead only about the differences in the stability of different union types. Therefore, the sources of stability in unions with a child present are not only the children themselves, but also individual characteristics such as commitment, relationship quality or family-oriented values which are not observed in this study, but which prove to be positively related to both fertility decisions and union stability. For example, couples with children are probably selected in terms of better relationship quality or stronger commitment to the partnership. This calls for the use of panel data in future research. Lastly, although the GGS data are representative, some deviations from the population statistics were also observed for the most recent periods, so the descriptive estimates of the partnership stability should be viewed with caution. The estimates of relative stability, on which the present study relies, should be unbiased, however. Despite these limitations, this article provides a systematic overview of cohabitation stability across European countries and brings the presence of a child in a relationship into the spotlight.
Masaryk University, Brno, Czech Republic.
Correspondence: Zuzana Žilinčíková, Joštova 10, 60200 Brno, Czech Republic, email: firstname.lastname@example.org
Italy and The Netherlands had to be excluded due to variations in their national questionnaires.
Even though there was more detailed retrospective information in some national questionnaires, this article employs only those which were available for most of the countries.
Several papers have suggested that retrospective information for the oldest periods might underestimate fertility and nuptiality. Period estimators from the 1970s onwards are fairly accurate, although in a few countries the nuptiality and/or fertility estimates might be overestimated (Kreyenfeld et al., 2013; Vergauwen et al., 2015).
The presence of a child in the first measured period might indicate that the mother brings her own child to a new relationship, or that the child was conceived prior to the start of a cohabiting union.
Marital status and the presence of a child are measured at the fifth year after union formation or at the moment of dissolution in cases where it occurs.
Results of further tests not shown.