INTERDEPENDENCIES BETWEEN MARITAL INSTABILITY AND FERTILITY

Divorce rates increased sharply in most developed countries in the second half of the 20^th century, and had reached historically high levels by the end of the century (Goldstein, 2003). Since the beginning of the 1960s the total divorce rate has increased in almost all European countries. The forerunners in the trend towards marital instability were the countries of northern Europe, like Denmark, Sweden, and the United Kingdom; as well as some republics of the Soviet Union (Estonia, Latvia, Lithuania, Russia, Ukraine). In 2000, a large share of European countries had total divorce rates of over 0.4, and a few countries had rates of 0.5 or higher (Belgium, Finland, Sweden); although some European countries still had rates of 0.25 or lower (Bulgaria, Cyprus, Spain, Italy, Greece, Croatia, Serbia, Poland, Romania, Slovenia) (GGS Contextual Database). In the U.S., levels of marital instability rose steadily through the second half of the 20^th century, and by the beginning of the 21^st century almost half of first marriages in the U.S. ended in divorce (Schoen & Canudas-Romo, 2006).

The rapid and lasting rise in marital instability has attracted the attention of a range of social scientists. The basic challenge is to determine how common divorce actually is (Amato, 2010; Preston & McDonald, 1979; Schoen & Canudas-Romo, 2006). In addition to attempting to improve the measurement of divorce rates, researchers have investigated the factors that contribute to and the consequences of divorce. The macro-level factors that shape marriage dissolution are embedded in legal, economic, and social frameworks. These macro-level factors explain the differences in divorce rates across populations, and underlie the changes in divorce patterns over time. The micro-level factors that influence individuals' decisions about whether to remain in a marriage - i.e., the characteristics of the individual partners and of their unions, and the socio-economic backgrounds of the partners - are studied in order to identify causal dependencies or socio- -demographic differentials in marital instability (Jalovaara, 2002; Lampard, 2013; Zeng, Morgan, Wang, Gu, & Yang, 2012).

The rise in the frequency of divorce has also led researchers to investigate the consequences of divorce. The increasing share of marriages that end in divorce and the rising numbers of people who remarry shape family structures by increasing the numbers of single-parent families, couples in which at least one of the partners was previously married, and reconstituted families. The consequences of divorce on individuals and society are varied and profound. For example, individuals, and especially women, may experience materiał hardship after divorce. In addition, the family relationships and family networks of divorced individuals become more diversified and volatile, which could affect their well-being and security. The main topic in the research on the consequences of divorce has been the impact of divorce on the children of the divorcing couple. The emotional, behavioural, social, health, and educational outcomes of these so-called "children of divorce" have been studied intensively (Booth & Amato, 2001; Fomby & Cherlin, 2007; Huurre, Junkkari, & Aro, 2006; Kim, 2011; Potter, 2010; Sigle-Rushton, Lyngstad, Andersen, & Kravdal, 2014; South, Crowder, & Trent, 1998; Steele, Sigle-Rushton, & Kravdal, 2009; Sun, 2001). The effects of parental divorce might also extend to different aspects of a child's adult life, including to his or her well-being, family life, and social networks (Amato, 2010). However, divorce also affects the divorcees themselves. In particular, researchers have examined whether divorce influences the partners' material circumstances, life satisfaction, mental and physical health, and mortality (Amato, 2000; Lesthaeghe, 2010; Musick, Brand, & Davis, 2012; Musick & Meier, 2012; Rydzewski, 2010; Uunk, 2004).

As I have shown in this brief overview, the existing research on divorce is broad in scope, as many of the factors associated with divorce appear to both contribute to and be the result of divorce. In this work I concentrate on a particular dimension of the on-going research on marital instability: i.e., on its associations with fertility, whereby fertility is considered both as a factor that influences marital instability and as a process that is affected by marital instability. Fertility is the key component in the transformation of the family that has been occurring in developed countries since the 1960s, and shifts in fertility have led to fundamental changes in the reproduction patterns of populations. The decline in fertility to below replacement levels, the tendency to delay or forgo marriage, and the increase in levels of marital instability have led to dramatic changes in the family over the past 50 years. As the decrease in fertility to below-replacement levels has tremendous effects on the ageing processes and the demographic structures of populations, as well as on the economic and social progress of societies, fertility is a central topic in demographic and social research. Because fertility affects population size, the balance between the generations, social structures, and the prospects for sustainable growth, public policy makers have long been concerned about this demographic outcome (European Commission, 2005, 2006; Sobotka, 2008a). Given the extent to which changes in fertility and marital instability affect the current and future demographic regimes, it is surprising that we have relatively little knowledge about the interdependence of these two processes.

In my study I explore the interdependencies between marital instability and fertility by investigating the effects of fertility on marital instability, and, conversely, the effects of marital instability on fertility. To this end I address two research questions:

1. How does having children affect marital instability?

   To answer this question, I will explore the effects of fertility on marital instability on a micro level.

2. To what extent do changes in the risk of marital disruption contribute to changes in completed fertility?

   To answer this research question, I will explore the effects of marital instability on achieved fertility by linking the dependencies observed on a micro level to the outcomes measured on a macro level.

   The question of whether having children contributes to marital stability has long been debated by social scientists. Although the literature on this topic is extensive, the results of studies that have examined the effects of having children on marital stability have not been conclusive. Both the theoretical considerations and the empirical findings indicate that having children can either stabilise or destabilise relationships, although the view that the presence of children lowers the risk of divorce seems to predominate. According to some theoretical arguments, having children decreases the risk of marital disruption because children constitute a form of marriage-specific capital that loses its value if the marriage ceases to exist (Becker, Landes, & Michael, 1977). The presence of children also creates additional barriers to dissolving a marriage, as having children increases the material, social, and psychological costs of a marital break-up, while also raising the utility of marriage and of role specialisation within marriage (Becker et al., 1977; Cherlin, 1977; Levinger, 1965; Parsons, 1940). On the other hand, the presence of children can place pressure on a family's resources, and can thus lead to increased strain on the partners' relationship. Moreover, the birth of a child demands that both parents make major adjustments, which can create tensions that are destructive to marriage.

   Empirical studies have often found a negative association between the risk of marital disruption and the presence of children (e.g., Berrington & Diamond, 1999; B. Hoem&Hoem, 1992; Todesco, 2011). However, other studies have shown that the stabilising effect of having children diminishes as the children age (Andersson, 1997; Toulemon, 1995), differs according to the parity of each child (Lillard & Waite, 1993), and becomes destabilising after the endogenous effects of fertility on dissolution risk are taken into account (Fan, 2001; Svarer & Verner, 2008); and that the relationship between marital instability and fertility is positive, even if the endogeneity of fertility is not acknowledged (Boheim & Ermisch, 2001; Chan & Halpin, 2002).

   As the available theoretical reasoning and empirical evidence on the effects of having children on marital stability are not yet conclusive, my study contributes to the discussion in two ways. First, I conduct a detailed analysis of the existing theoretical approaches, and of their implications for empirical investigations. The theories identify the aspects of the presence of children in a marriage that are important for marital duration, and the mechanisms through which children influence marital disruption. These theoretical considerations, together with a review of empirical research, help to form the basis for my empirical study of the effect of children on marital stability in Poland. The second contribution of my study to this research topic is the empirical study itself. I conducted the study using event history analysis, which is a standard statistical technique for analysing the timing of an event; in this case, of marital disruption (Feijten, Boyle, Feng, Gayle, & Graham, 2009). The regression of hazard of marital disruption takes into account that the intensity of marriage disruption changes with time and under the influence of time-constant and time-varying covariates, and that some couples do not experience the event within the observation period, or ever (right censoring) (e.g. Allison, 1982; Blossfeld, Golsch, & Rohwer, 2007; Singer & Willett, 1993). These features of event history analysis enable me to capture the effects of having children on the risk of marital disruption without distortion due to other dependencies on the observed factors and their time-varying nature. Following recommendations from previous research, I also consider the potential endogeneity of fertility in the equation for marital instability due to unobserved time-constant individual characteristics. The failure to take into account the potential endogeneity of fertility could lead to a bias in the estimated effects of fertility on marital instability, because the initial analysis may have captured some unobserved characteristics that influence both the propensity to conceive a child and the duration of the marriage. Moreover, I control for the selection of entry into marriage, which may be another distorting factor. If there are some unmeasured characteristics that jointly influence the propensity to form or dissolve a marriage, and that are correlated with covariates of marital dissolution, then the estimates of those covariates will be partly the product of self-selection. In the analysis, I am able to take into account both the endogeneity of fertility and the selectivity of marriage through the simultaneous modelling of marriage formation, marital disruption, and fertility (Bernardi & Martinez-Pastor, 2011b; Brien, Lillard, & Waite, 1999; Steele, Kallis, Goldstein, & Joshi, 2005; Upchurch, Lillard, & Panis, 2002). There are only a few studies on the association between the presence of children and marriage instability that were based on multi- process multi-level models. To the best of my knowledge, none of them took into account the selectivity of marriage.

   The discussion about the effects of marital instability on childbearing has become increasingly important in recent years. The assumption that marital instability has negative effects on fertility was seldom disputed until a reversal of the correlation between the macro indicators of divorce and fertility was observed. Before 1990, the correlation between those two indicators was negative: i.e., countries with higher levels of marital instability had also lower fertility. But during the 1990s, the correlation between divorce and fertility became positive: i.e., countries with higher levels of marital instability also had higher fertility (Billari & Kohler, 2004; Prskawetz, Mamolo, & Engelhardt, 2010). After noting this reversal in the macro dependencies, researchers started to debate whether parallel changes were occurring on a micro level, and whether "union instability has become an engine of fertility" (Billari, 2005; Thomson, Winkler- Dworak, Spielauer, & Prskawetz, 2012). The idea that union instability may be positively linked to fertility outcomes is based on the observation that re-partnered individuals often have child(ren) with their new partner, even if they already had children from a previous union. Thus, it is possible that over the life course of an individual his or her subsequent unions will bridge the gap in fertility created by the dissolution of his or her first union. The existing research has not confirmed this assumption, as most studies have found that the average fertility outcomes of people who have experienced a separation did not exceed the outcomes of people who remained in a union (Beaujouan & Solaz, 2008; Thomson et al., 2012; Van Bavel, Jansen, &Wijckmans, 2012). At the same time, the results suggest that the fertility outcomes of individuals with discontinuous union histories are highly variable, and that the higher degree of dispersion may be a prelude to the emergence of a positive association between marital instability and fertility on the individual level (Van Bavel et al., 2012).

   The relatively small number of studies on the effects of marital instability on completed fertility were unable to provide a final answer to the question of whether divorce can be a pro-natalist force. Before this question can be settled, research using different methodological approaches and pertaining to different contextual factors is needed. I respond to this demand by conducting a study on the relationship between marital disruption and lifelong fertility. Using micro-simulation-based decomposition, (i) I check whether the increasing risk of union disruption contributed negatively or positively to the number of children born to a woman during her reproductive career, and (ii) I evaluate the relative contributions of changes in union dissolution behaviour and changes in union formation and fertility behaviour. To this end, I first build a micro- simulation model of family-related behaviours; i.e., of fertility, union formation, and union disruption. Using this model, I simulate the individual family careers of 100,000 women. I estimate the parameters of the model using event history models of basic family-related behaviours: union formation, union dissolution, and fertility. The evaluated outcome of the model is the mean number of children born to each simulated woman at the end of her reproductive career. Changes in the underlying intensities of experiencing the events of different types result in changes in the model outcome. By changing only one behaviour at a time while keeping the remaining behaviours constant, I am able to estimate the effect of the change in this particular behaviour on completed fertility. Comparing the effect of the change in a single behaviour to the effects of simultaneous changes in all behaviours also allows me to calculate the contribution of this behaviour to the total change in completed fertility. The main advantage of using the micro-simulation model is that it allows me to follow the relationship between changes in individual behaviours and societal or lifelong outcomes. Because there are numerous dependencies between the transitions that occur over the life course, this linkage is too complex to be derived analytically. The numerical solution to this problem offered by micro-simulation can be used to bridge the micro-macro gap (Billari, 2006; Matysiak &Vignoli, 2009). Thus, the main methodological advantage of the study is that I link existing knowledge on the interdependencies between processes on an individual level with the outcomes observed on a macro level, which is one of the up-to-date postulates for research in the social sciences (Billari, 2006; Matysiak & Vignoli, 2009; Willekens, 1999).

   The empirical analyses described have been conducted with the use of Polish data. The choice of Poland has important implications for the expected results, because this country has more traditional patterns of family-related behaviours than most other European countries (Kotowska, Jóźwiak, Matysiak, & Baranowska, 2008). Although those behavioural patterns have changed considerably over the past 30 years, marriage is still the dominant form of living together as a couple, and a marital union is considered to be the most appropriate arrangement for childbearing (Kowalska, 1999; Mynarska, 2009). In addition, compared to couples elsewhere in Europe, couples in Poland tend to marry relatively early, and are less likely to divorce (Styrc, 2010; Szukalski, 2013; Wieczorek, 1999). It is also important to note that Polish society is still heavily influenced by the Catholic Church, an institution that rejects the dissolution of religious marriages and extra-marital fertility. Based on the contextual knowledge of Poland and my review of the relevant theoretical and empirical research, I have formulated three specific hypotheses. The first two hypotheses refer to the micro level, and the third hypothesis refers to the macro level.

    

Acknowledgments

Introduction

Chapter 1. Family change in Europe

1.1. Union formation

1.2. Union disruption

1.3. Fertility

1.4. Cross-country correlations between fertility and divorce

1.5. Summary

Chapter 2. Marital instability in Poland

2.1. Legal framework of marital instability in Poland

2.2. Statistical picture of marital instability in Poland

2.3. The cultural settings for marital stability in Poland

2.4. Summary

Chapter 3. Effects of children on marital stability: a theoretical framework and a review of the empirical research

3.1. Having children as a factor that stabilises marriage

3.1.1. Marriage capital

3.1.2. Role-specialisation in the marriage

3.1.3. Barriers and costs of dissolving a marriage

3.2. Having children as a destabilising factor in a marriage

3.2.1. Adjustment, stress, and marital conflict

3.2.2. Children conceived before marriage

3.2.3. Children born before marriage

3.3. The effect of children on marital stability - a summary of the theoretical arguments

3.4. Empirical studies on the effect of the presence of children on marital stability

3.5. Summary and conclusions

Chapter 4. Effect of children on marital instability

4.1. Data

4.2. Method and model specification

4.3. Results

4.4. Summary and discussion

Chapter 5. Effects of rising union instability on aggregated fertility

5.1. Background

5.2. Data

5.3. Methods

5.3.1. Event history models

5.3.2. Micro-simulation model

5.4. Results

5.4.1. Event history analysis

5.4.2. Micro-simulation analysis

5.5. Summary and discussion

Conclusions

Appendices

Appendix 1. The procedure used to compile the data series for chapter 1

Appendix 2. Estimates of the multi-process model from chapter 4

Appendix 3. Event history regression estimates for chapter 5

References