Does Parenthood Stabilize Relationships, after All? Disentangling the Effects of Family, Household and Marriage Formation on the Risk of Separation

Christine Schnor, Vrije Universiteit Brussel

This study examines the risk of separation over relationship duration – including the first non-residential episode – and compares the effect of family formation to the stabilizing effects of household and marriage formation.

Little is known about the variation of relationship stability with household, marriage and family formation when the length of the relationship is measured from its very start. Literature shows that children positively affect the stability of cohabiting and married couples at least in their pre-school years. These studies however neglect that household and marriage formation are often synchronized with first childbearing and thus, the positive effect of first children may be partly attributed to a ‘honeymoon’ effect after household or marriage formation. Furthermore, a pregnancy can have little or even a negative effect on the stability of a non-residential partnership, an aspect that literature has not addressed so far. In sum, the effect of children on overall relationship stability remains still to be evaluated. This study treats cohabitation and marriage as episodes of the same relationship and disentangles the relative effects of household, marriage and family formation on the risk of separation.

We use fertility and partnership histories of women born 1971-1973 and 1981-1983 gathered in the German Family Panel (pairfam) and control for their observed and unobserved characteristics in a piecewise continuous hazard model. The empirical findings reveal that family formation stabilize relationships, but to a lesser extent than other relationship events like household formation and marriage. A closer look on marital status suggests that childbearing increases stability among non-marital relationships, while it does not lead to much further stability increases among marriages.

Introduction & background

Traditionally, the first childbearingwas closely linked to household and marriage formation. In recent decades,these decisions have become more decoupled, which has led to increasingdiversity. Studies have confirmed the stabilizing influence of first childrenin marriages and cohabitations at least in their preschool years. However,studies have overlooked the – often first – partnership period in which thepartners lived in different households. There are at least three reasons why toconsider how a pregnancy affects the separation risks on the partnership level – including couples without a joint household:

First, taking marriage or householdformation date as the partnership start point results in some selection bias,because the risk of separation is found to be strongly duration-dependent. Somecouples move together directly after partnership formation, others spend yearsin separate households and thus, their risk of separation after moving togetheris different. This unobserved selectivity might have biased prior findings.

Second, in traditional family patterns,household, marriage and family formation often were synchronized which made itdifficult to disentangle the effect each partnership event has on relationshipstability. Considering the partnership dimension among women of recent birthcohorts allows to reflect the diversity of living arrangements in modernfamilies and brings in sufficient variation in the timing of partnership eventswhich should help to identify the effect of childbearing on stability.

Third, we know very little about thestability of relationships in which the partners do not live together. Stabilitystudies often assume that persons without a coresiding partner are notpartnered and thus not under the risk of separation. But, these persons couldhave a partner outside the household and form a family in this partnership. InEuropean countries, between 9% (Norway) and 29% (UK) of first conceptions inthe years 1995-2004 took not place within a non-marital cohabitation or amarriage. Detailed partnership data on German women born 1971 to 1973 revealthat about two third of these women were not truly without a partner, but livedin a partnership with separate households. It is not clear whether a pregnancyadds to stabilize these rather fragile relationships, because their degree ofinstitutionalization is rather low and a pregnancy may be unintended in thisstage.

In this study, couples are under therisk of separation starting with the date of partnership formation. Cohabitationand marriage are included as episodes of the same relationship in the analysis.I examine the risk of separation over relationship duration – including thefirst non-residential episode – and compare the effect of family formation tothe stabilizing effects of household and marriage formation.

Data & Methods

The analysis is based on data fromthe German Family Panel (pairfam) which provides full retrospective fertilityand partnership histories of men and women born in 1971-1973, 1981-1983 and1991-1993. The preliminary analyses is based on 7942 partnerships, drawing on aspecial event history data set that includes the first three pairfam waves(2008/2009, 2009/2010, 2010/2011) and the first waves (2009/2010, 2010/2011) ofthe additional oversample of eastern German respondents.  In a next step, thedata set will be extended to information gathered up to wave eight (release8.0).

I used piecewise continuous hazardmodels to estimate the relative risks of separation, controlling for observedand unobserved characteristics. The observation was censored 10 years after relationshipformation. The hazard function consists of the baseline hazard, defined as apiecewise linear spline with knots 1, 2 and 6 years after relationship start. Durationsplines are used to specify the effects of continuously varying effects ofhousehold, family and marriage formation.

Preliminary results

The preliminary results are shown inFigure 1 as an hypothetical example which builds on the median starting pointsof the respective events in the data. The risk of separation is highest at thebeginning of the relationship and remains fairly constant after three years of relationshipduration (the slight increase after six years is not statisticallysignificant). The conception of the first child significantly increases relationshipstability, but this effect is less strong than household formation andmarriage. The risk of separation after family formation seems to increase againwith the age of the first child, but these results are not found to besignificant. Coresidence has a strong, but only temporary positive effect,whereas marriage stabilizes the relationship throughout the observed period. Infurther models (not shown), non-marital and marital family formation aredistinguished. The median non-marital union starts childbearing earlier than themedian marital union. Moving in together as well as marriage formationsignificantly stabilizes the relationship. Childbearing only increases theunion stability among non-marital unions, while it does not lead to muchfurther stability increase among marriages.

Figure 1: Transition to separation,duration splines

Significance of slopes ***p<0.01, ** p<0.05, * p<0.1


Presented in Session 1109: Families and Households