Stability of fertility preferences and intentions

(1)

Stability of fertility preferences and intentions

A new angle on studying fertility behavior in Germany

Antonia Spath

Department of Sociology, Demography Unit (SUDA) Master’s Thesis 30 HE credits

Subject: Demography

Multidisciplinary Master’s Programme in Demography (120 credits) Spring term 2018

Supervisor: Sunnee Billingsley

(2)

Stability of fertility preferences and intentions

A new angle on studying fertility behavior in Germany

Antonia Spath

Abstract

Prevailing low fertility rates in several European states, such as Germany, have been studied widely in recent years. Findings include discrepancies between fertility preference and actual family size as well as between fertility intentions and fertility behavior; an ‘unmet need’ for children found on the individual and the societal level. Fertility preference is specified as the individual ideal number of children, and fertility intentions as the long- or short-term plans to have a child. Apart from investigating the rates of realization, these measures have been understudied.

The objective of this study is to illuminate a new angle of low fertility rates in Germany by reviewing fertility measures previously considered to be stable predictors of fertility behavior. The aim is to investigate the stability of fertility preferences and of positive short-term fertility intentions of Germans in their reproductive age. According to the Theory of Planned Behavior and the life-course perspective, attitudes and experiences can influence fertility preferences and short-term fertility intentions. In this study, the suspected connection between unstable preferences and intentions and certain attitudes towards and experiences with the career, working life, and childcare situation is examined. These processes are expected to differ between men and women, and between childless individuals, parents with one child and parents with more than one child.

Using data from seven survey waves of the German family panel pairfam, fixed-effects and random- effects regression models are run separately for women and men and for those of different parities.

The results suggest that those with high career importance and those who expect or perceive a negative effect of children on the career are more likely to have unstable positive intentions. Although fertility preferences are shown to be somewhat unstable, no relevant relationships can be found. The differences between the findings on men and women regarding relevant determinants and direction of the relationships are unexpectedly small. Childless individuals are as likely to hold unstable preferences and intentions as parents.

Keywords

Fertility preferences, family size preference, short-term fertility intentions, fertility intentions, Germany, Theory of Planned Behavior, life course perspective, attitudes, working life, female labor force

participation, gender, parity.

(3)

List of Tables

Table 1: Number of respondents per wave, pairfam and DemoDiff data sets ... 18

Table 2: Sample sizes (individuals and observations) of separate models... 19

Table 3: Outcomes of positive short-term fertility intentions by sample group (percentage). 27 Table 4: Women’s fertility preferences (dependent variable: FD ideal number of children) .. 29

Table 5: Men’s fertility preferences (dependent variable: FD ideal number of children) ... 30

Table 6: Women’s fertility intentions (dependent variable: Short-term fertility intentions) ... 33

Table 7: Men’s fertility intentions (dependent variable: Short-term fertility intentions) ... 34

Table A1: Number of under 3-year olds enrolled in daycare since 2005 ... 50

Table A2: Ideal number of children (mean) by cohort groups ... 50

Table A3: Descriptive statistics (Fertility preferences): Explanatory variables ... 51

Table A4: Descriptive statistics (Fertility preferences): Control variables ... 51

Table A5: Descriptive statistics (Fertility intentions): Explanatory variables ... 53

Table A6: Descriptive statistics (Fertility intentions): Control variables ... 53

List of Figures

Figure 1: The Theory of Planned Behavior applied to fertility decisions ... 3

Figure 2: Diagram of filter questions in questionnaire ... 20

Figure 3: Ideal number of children (distr. in % and total numbers) ... 26

Figure 4: Ideal Number of children: Mean absolute deviation from the mean (within id) ... 27

Figure A1: Ideal number of children: Difference between waves, total sample ... 47

Figure A2: Ideal number of children: Difference between waves by sample groups... 48

Figure A3: Ideal number of children: Mean abs. dev. from the mean by sample groups... 49

(5)

1

1. Introduction

Low fertility has become a concern for German policymakers debating demographic sustainability in a society that exhibited a total fertility rate between 1.3 and 1.5 since the 1970s (Statistisches Bundesamt, 2015, p.°27). Although sub-replacement fertility levels have become the norm in all European countries, several states such as Germany experience a more persistent decline. The causes have been of interest to researchers for years. A more in-depth approach concerns the link between individual family size preferences, childbearing intentions, and actual behavior. Family size or fertility preferences are measured as the number of children that would be preferred under ideal circumstances, and childbearing or fertility intentions are long- or short- term plans to have a child. This angle is especially interesting today, as a gap between the preferences, the intentions and the actual childbearing behavior would indicate an ‘unmet need’

for children (Liefbroer, 2008, p.°364). Factors and potential barriers related to this discrepancy, which has been shown in several studies on the societal and individual level, can serve as a reference for policies aimed at enabling individuals to realize their fertility goals and raising fertility levels. As the relevance of studies on this topic is obvious, the concepts used in these studies can certainly be evaluated closer.

Generally, family size preferences have been perceived as relatively stable across Western societies, cohorts, and over the life course. Fertility desires of individuals are conceptualized as

“fundamental and quite general value orientations, or as expressions of family size norms”

(Kuhnt, Kreyenfeld, & Trappe, 2017, p. 237), and thus not expected to vary a lot. Similarly, intentions, once positive, are not expected to change until a child is born. After the conditions are deemed suitable to implement underlying fertility preferences, positive short-term intentions should predict childbearing in the following years. However, the predictive power of fertility intentions has recently been doubted, and the case of positive intentions being changed to negative intentions is an interesting occurrence to be studied. As the life course perspective has gained ground, the notion of values and behavior being modified by experiences and resources has become more established. In addition to the knowledge that has been gained on factors associated with positive fertility intentions and the realization of fertility intentions, it will be interesting to examine aspects that are possibly related to individuals holding unstable positive fertility intentions or even fertility preferences.

(6)

2

This study will contribute to the existing literature on fertility preferences and intentions by expanding the knowledge on the stability of fertility preferences and intentions (see Kuhnt et al., 2017, pp. 238–239 for an overview of previous studies). As well as reviewing established influences on positive fertility intentions and fertility outcomes, the effect of individual experience with the double burden of childrearing and working life, parental leave, and childcare, as well as attitudes towards childcare and career will be studied. The stability of fertility preferences and positive intentions is suspected to be linked to the perceived and actual compatibility of career and childbearing. Data from seven waves of the German Family Panel pairfam enables a detailed analysis of fertility histories and attitudes for individuals born between 1971 and 1993.

The research questions this study seeks to answer are: How stable are fertility preferences and positive fertility intentions across the reproductive lifespan? Which characteristics, attitudes, and experiences are associated with a) unstable reports of the number of children Germans would like to have under ideal circumstances, and b) reporting negative fertility intentions after not realizing their recent positive intentions? Furthermore, the analyses are conducted on separate samples to observe possible differences by gender or parity.

This thesis is structured as follows: After the introduction, the terms and concepts of fertility preferences and fertility intentions will be discussed, and the underlying theories, the Theory of Planned Behavior and the life course perspective, are presented. This section introduces the relevant previous research. Following this, research hypotheses are deduced. The fourth section covers the data and the method. Thereafter the results are presented, and the hypotheses validated. Finally, the findings of the thesis are discussed, and a conclusion is drawn.

(7)

3

2. Theory and previous research

2.1. Fertility preferences and fertility intentions

The concepts of family size or fertility preferences and fertility intentions have been shortly introduced above. The terms fertility ideals, fertility preferences, and fertility desires have been used interchangeably. However, fertility ideals can be understood to reflect perceived social norms and expectations, and fertility preferences or desires measure personal attitudes and values. The theoretical basis for the link between preferences, intentions, and behavior is the Theory of Planned Behavior (Ajzen, 1991). According to Ajzen, intentions and perceived as well as actual behavioral control predict behavioral achievement (Ajzen, 1991, p. 182). Intentions are described as “motivational factors that influence a behavior; […] indications of how hard people are willing to try, of how much of an effort they are planning to exert” (Ajzen, 1991, p. 181).

Intentions can be predicted by the personal attitude towards the behavior, the subjective societal norm, and the degree of perceived behavioral control (Ajzen, 1991, p.°188).

This theory has been widely applied in fertility research. Although unintended pregnancies do occur, it can be assumed that “having a child is the result of a reasoned decision” (Ajzen &

Klobas, 2013, p. 204). In Germany, where contraception is easily available and widely accepted,

“[f]ertility-related behaviors are preceded by corresponding intentions” (Kuhnt & Trappe 2016, p.°17). This underlying intention to have a child follows an evaluation of expected costs and benefits, perceived obstacles, social norms, as well as of the individual, demographic, and societal background factors (Kuhnt & Trappe, 2016, pp. 17–18). Ajzen and Klobas created a diagram detailing factors affecting fertility decisions:

Figure 1: The Theory of Planned Behavior applied to fertility decisions (Ajzen & Klobas, 2013, p. 206)

(8)

4

The Theory of Planned Behavior implies that both individual attitudes and subjective social norms and expectations can predict intentions. Both measures, fertility preferences and intentions, and especially the stability across the years of those in reproductive ages, are interesting to study together, as they tell a similar, but complementary story. Instability of fertility preference relates to a change in the desired number of children, while the instability of short- term fertility intentions relates to a change in the planning concerning the timing.

2.2. Stability of fertility preferences and intentions

2.2.1. Stability of fertility preferences

Although the preferences are generally expected to remain relatively stable over the course of the life, the Theory of Planned Behavior gives an indication as to how preferences might be adjusted. According to the socialization hypothesis, values and general attitudes are formed in childhood and adolescence and are not prone to change afterward (Kuhnt et al., 2017, p. 237), as the individual background, upbringing and early experiences, are important milestones.

However, Ajzen and Klobas (2013) also invoke the demographic and societal background influencing the beliefs about the consequences of and social support for having a child, and about the enabling or interfering factors. Certainly, these beliefs can change with age, income, and perceived societal or cultural changes.

2.2.2. Stability of fertility intentions

As described earlier, fertility intentions are expected to remain negative until the circumstances are sufficiently good. Once an individual intends to have a child, this intention should remain positive and stable until it is realized. After a child is born and the intention is realized, the positive intention is expected to turn negative. A reversal of positive fertility intentions before intentions are realized could be considered counterintuitive and unlikely at least. However, the Theory of Planned Behavior suggests why this might happen: As the intention to have a child is derived from the attitude towards, the subjective norm for, and the perceived control over having a child, a change in these dimensions through a change in underlying beliefs can alter the intention. Also, the perceived control might be influenced by unsuccessful attempts to induce pregnancy or revealed medical issues, by the actual control over having a child.

2.2.3. Attitudes and stability of preferences and intentions

According to the Theory of Planned Behavior, an attitude “represents a person's general feeling of favorableness or unfavorableness toward some stimulus object” (Fishbein & Ajzen, 1975,

(9)

5

p. 216). Individuals hold attitudes towards many objects in their lives. Concerning fertility decisions, attitudes towards having a child are the most relevant for the process. These attitudes

“develop reasonably from the beliefs people hold about the object of the attitude” (Ajzen, 1991, p.°191). Naturally, these beliefs about consequences of having a child concern not just the family life, but also all other aspects of life. Attitudes are thus formed under consideration of consequences for one’s career or working life as a large and important sphere of adult life.

Attitudes towards the career, education, jobs, and the compatibility of family life and working life are continuously formed and altered. As beliefs and attitudes are adjusted, fertility preferences and, indirectly, fertility intentions might be adjusted, too. It is important to mention that these processes can be examined to operate in both directions. Giving birth has been shown to influence Swedish women’s attitudes towards their careers and working lives, although just for a short period of time (Evertsson, 2013, p.°150).

2.2.4. Experiences and stability of preferences and intentions

Both fertility preferences and fertility intentions can arguably be influenced by experiences. “We form beliefs […] by associating it with certain attributes, i.e., with other objects, characteristics, or events” (Ajzen, 1991, p.°191). Beliefs about the consequences of having a child, as well as about the social support for having a child, and about the enabling or interfering factors are thus altered if new experiences prove formerly held beliefs to be wrong. Consequently, attitudes might be adjusted. For example, fertility preferences can be adjusted downward after experiencing difficulties combining work life and family life, either in the workplace or in the home. Similarly, positive or negative experiences of colleagues, acquaintances, or relatives might lead to a person changing their beliefs and thus their attitudes. Changing beliefs can indirectly affect fertility intentions through changing preferences. However, fertility intentions can also change as circumstances change. Adjusting positive intentions without realizing them is likely connected to an unforeseen event or experience changing circumstances to become unfavorable. To understand how experiences or events across the life course can interact with fertility preferences and intentions, the life-course perspective is introduced.

The life-course perspective has been developed and implemented in different disciplines since the 1960s (Elder, 1985). For panel data analysis, it has become a standard approach and an

“indispensable framework for demographic research” (Huinink & Kohli, 2014, p.°1295). It is based on the assumption that an individual’s life consists of a sequence of life events embedded in a distinct career. Life courses are “part and product of a societal and historical multilevel process” (Mayer, 2004, p. 166) and interrelated with others’ life courses. They are

(10)

6

multidimensional as they evolve in “mutually related and mutually influencing life domains”

(Mayer, 2004, p. 166), and they are self-referential, as individuals “act[…] or behave[…] on the basis of prior experiences and resources” (Mayer, 2004, p. 166).

Fertility theories have been argued to be conceived under the terms of life-course dimensions, as individuals “conceive their actions in […] life course terms” (Huinink & Kohli, 2014, p.°1295).

Human fertility is embedded in a spectrum of social, cultural, and personal factors. It interrelates with other domains of life, is affected by experiences from the past life course and impacts the future welfare (Huinink & Kohli, 2014, p.°1295). Fertility and fertility desires and intentions can be seen as dynamic; processes that unfold and can be revised over time (Iacovou & Tavares, 2011, p.°116). “The dynamic nature of fertility may be somewhat less obvious since childbearing itself is irreversible; nevertheless, fertility proceeds one child at a time, and decisions about future births—both number and timing—can be re-evaluated at any point as relevant circumstances change” (Brewster & Rindfuss, 2000, p. 290, Udry 1983, Namboodiri 1972).

“Childbearing is a sequential decision-making process that is very sensitive to changes in individuals’ circumstances, and factors affecting childbearing decisions may vary as the number of children increases” (Chen & Yip, 2017, p. 585, Iacovou & Tavares 2011). Very different processes can be expected to be at play when it comes to the adjustment of fertility preferences or intentions for childless individuals and for parents. Although the transition to parenthood is an irreversible, incomparable long-term commitment, so is having another child. It will be

“closely related to the previous experience of raising the first child” (Chen & Yip, 2017, p. 585).

It follows logically that fertility desires and even the more predictive intentions can be changed, adjusted or revoked through influences such as own experiences, societal pressures and trends, experiences of social contacts, and the perceived appeal and effectivity of policy that influences parenthood experiences.

Variations in the relationship between women’s and men’s life situation or reality and their fertility preferences and intentions can be explained by the life course perspective. It was suggested that changes in fertility preferences are related to changes in “external life circumstances” (Liefbroer, 2008, p.°365), such as cohabitation, marriage, or separation, or in employment status. Progression in age, as well as parity, have been found to affect the ideal number of children individuals state in interviews. Young adults might either have an idealized view of childrearing and adjust it after having contrary experiences, or they have not yet considered timing of childbearing or number of children, and their attitudes evolve once they are older. Also, individuals who feel that their age is too advanced to realize their original fertility

(11)

7

desire might readjust the ideal number of children downwards “to minimize the negative consequence of the failure to realize [it]” (Liefbroer, 2008, p.°366).

As to parity changes, it has been shown that with every additional child increases the fertility preferences, as even after an unintended pregnancy most individuals “rationalize” the additional child as “wanted” (Kuhnt et al., 2017, p.°237). The perception of how children influence their career, how important their work life or education is to them, or to what extent they are willing to cut back in their job to accommodate children’s needs certainly ebbs and flows during their working lives. First- or second-hand experiences they gain along the years, promotions, a new career orientation, a new position or project being less or more satisfying – individuals have dynamic working lives, and with this, the ideal number of children or short-term fertility intentions are expected to change.

2.2.5. Differences by gender and parity

These processes of changing fertility preferences or intentions might differ between men and women and between childless individuals and those who are parents. Philipov (2009) illustrates the relevance of perceived behavioral control, as introduced before, for the formation of fertility intentions. He argues that “[c]onventional policy instruments such as child allowances and parental leave” have an effect on this perceived control, and especially on the intentions to have a second child (Philipov, 2009, p. 359).

In Germany, there is currently no gender equality regarding the distribution of childcare, housework, and the consequences of having children for the working life. Official policies, workplace regulations, daycare provision, and societal expectations are not alligned to offer men and women the same opportunities as working parents. Germany is considered a welfare state with conservative family policies. Western European conservative welfare regimes “support[…]

men’s primacy at the labour market but also provide[…] possibilities for women to combine paid work and family responsibilities”, and while German-speaking countries are included in this group, policies are “less supportive for women’s labour force participation” (Oláh, 2015, p. 2).

The German Federal Ministry for Family Affairs, Senior Citizens, Women and Youth states the government aims at “supporting a good reconciliation of family and work, maintaining economic stability of families, promoting child-wellbeing as well as helping people to fulfil the desire to have children” (“Families”, n.d.).

Since the middle of the 2000s, before the first interviews in the pairfam study were conducted in 2008, several policy changes have been implemented to better support mothers and eventually raise fertility. The parental leave and parental allowance act, Bundeselterngeld- und

(12)

8

Elternzeitgesetz, taking effect on January 1^st, 2007, was a revisal of the previous policy.

Employed mothers profit from a higher parental allowance, but the paid leave was shortened from 24 to maximal 14 months. “During the period from the beginning of their pregnancy to the end of the maternity leave, at least to the end of four months after giving birth, women are also protected against dismissal” (“Families”, n.d.). Fathers are encouraged to take part of the paid parental leave. On average, of the 14 months paid parental leave available if fathers take at least two of these months (otherwise, only 12 months are paid), women took 11,5 months and men 3,7 in 2017 (“Durchschnittliche (vorraussichtliche) Bezugsdauer”, 2018). Since the commencement of the law, the share of fathers taking advantage of it rose from 21% to 34% in 2017 (Bundesministerium für Familie, Senioren, Frauen und Jugend, 2016, p. 6). If fathers apply for parental allowance and parental leave, in 65% of cases, this is at least partially a joint leave with the partner (Egeler, 2012, p. 10).

The parental leave reform was thought to encourage women to re-enter the labor market quicker, but critics of this policy change claimed that the subsequent care for the children was not available. Although 2005 childcare reforms increased the availability of subsidized daycare for children under three years old, many parents were still unable to find adequate care for their young children. From August 2013, when only 29 of 100 children¹ below three years could be offered a daycare slot (Müller & Wrohlich, 2016, p.°673), parents can legally claim a slot at a subsidized daycare center for every child after its first birthday; before, this was only possible after the child’s third birthday (Heine, 2013). However, the lengthy process of legally challenging the denial of a spot is often futile, since daycare centers are simply not yet built, and there are not enough qualified daycare workers to fill the positions. From March of 2013 to March of 2017, the share of children below three years of age in daycare facilities grew only slightly, from 29 to 33 percent (Statistisches Bundesamt, 2017, see Table A1 in Appendix).

Availability is not the only issue for parents, as some refrain from placing their child into public care over concerns about the quality of the care or because they have only been offered spaces in inconvenient locations. A 2013 government-commissioned report found only 3 percent of daycare centers to be of ‘good’ quality, while a majority was deemed mediocre or lacking (Heine, 2013).

In general, women’s labor force participation has risen in the ten years since the parental leave reform. In the second and third year of their children’s lives, the share of women in full-time and

1There is a significant disparity in the availability of daycare spaces between East and West Germany. Numbers from 2013 show that in East Germany, 52 of 100 children found a spot, while in the West only 24 were able to (Müller and Wrohlich, 2016, p. 673).

(13)

9

part-time employment has increased, and the number of those working less than 15 hours a week has slightly decreased (Bundesministerium für Familie, Senioren, Frauen und Jugend, 2016, p. 4). Although this can be seen as a step towards more support of working mothers, the 2013 introduction of Betreuungsgeld, a childcare allowance that is additionally paid to parents of children aged 15-36 months that decide to forego publicly subsidized childcare, is part of a conflicting policy. This move was criticized as a step back, discouraging women to re-enter the workforce (Müller & Wrohlich, 2016, p. 673).

Women in Germany still carry most of the childcare burden, while men’s working lives are rarely seriously disrupted. Although the last decade saw changes in family policies that allowed women to better combine work and family life, and encouraged men to be more involved fathers, large deficits can still be found. Many women might not feel there is enough support for them, and some necessary policies are still not satisfactorily implemented. If there is no trust in publicly funded childcare for every child, women might still forego childbearing, or further childbearing, according to their expectations or experiences.

2.3. Previous studies

A number of studies have been conducted on the stability of fertility preferences or fertility intentions over the reproductive life course (Kuhnt et al., 2017, pp. 238–239). The main focus of previous research has been on the gap between fertility preferences and realized fertility, and on the realization of short-term fertility intentions. Most existing research utilizes cross-sectional samples, drawing on the premise of stable preferences and intentions (Heiland, Prskawetz, &

Sanderson, 2008, p. 130). The following section will review firstly studies on the stability of fertility preferences, and secondly on the stability of fertility intentions.

2.3.1. Stability of fertility preferences

Studies on the stability of fertility preferences over the life course have been conducted in Germany and in several African countries. Often, only women’s preferences were considered.

The lack of adequate data dictates the use of only two waves in many cases. In West Germany, Heiland et al. (2008) found variation in fertility preferences in men and women over two survey waves, conducted in 1988 and in 1994/95. They were able to show that stability of fertility preferences is more frequently found in individuals being raised in an intact family and with greater financial resources, while young Catholics and those with higher education were more likely to change their desires. However, these early life experiences were less significant later in

(14)

10

life, when events such as childbearing occur and have a greater influence on fertility preferences (Heiland et al., 2008, p. 151).

A more recent study, including six waves of the pairfam study, investigates women’s and men’s fertility desires across time. Kuhnt et al. (2017) found that the ideal number of children “seem to be unaffected by short-term changes in the respondents’ partnership status and employment situation, but they are sensitive to the achievement of long-term goals” (Kuhnt et al., 2017, p. 248). Their fixed-effects model could not confirm results from their pooled OLS-regression showing satisfaction with financial situation positively correlated with ideal number of children for women and having a partner positively correlated for men. However, they found that an increase in number of children was related to an increase in personal fertility preferences, while women’s age was negatively correlated. Yeatman, Sennott, and Culpepper (2013) found that young Malawian women adjusted their fertility preferences upward after the start of a new relationship, and downward after divorce. Their ideal number was adjusted upwards with a new pregnancy and increasing age.

2.3.2. Stability of positive fertility intentions

The stability of positive fertility intentions was examined by Chen and Yip (2017) who studied women in Hong Kong, using cross-sectional data. They point to increasing age being connected to a change in fertility intentions, not just to lower fertility levels because of postponement and subsequent infecundity in high age. “[E]ven if some women have not met their desired parity in their late thirties or forties, they may not intend or plan to have additional children at all.” (Chen

& Yip, 2017, p. 587). Another study by Sennott and Yeatman (2012) generates knowledge about the stability of short-term fertility intentions by investigating the time frames in which young Malawian women intend to have their next child. Over 18 months, 80% of interviewees changed timing preferences. Gaining a partner, moving to a better house, or the partner finding a better job was found to be related to an acceleration of the pace of childbearing intentions. An anticipated move because of work was linked to a postponement of the intention to have a child.

Spéder and Kapitány (2009) investigated outcomes of short-term fertility intentions in Hungary, using two survey waves connected to the Generation and Gender program. They looked at factors related to the possible outcomes of positive intentions (realization, postponement, or abandonment), and of negative intentions (unintended births and consistently opposed) (Spéder

& Kapitány, 2009, p. 507). Regarding the outcomes of the positive intentions, which will be the focus of this thesis, Spéder and Kapitány found that those who fail to realize their intentions were older, and that those with more children were more likely to abandon their intentions as

(15)

11

compared to realizing them. Individuals without a partner were also less likely to realize their positive intentions (Spéder & Kapitány, 2009, p. 514). Those with a lower educational level were more likely abandoners, and non-religious individuals were more likely abandoners and postponers. Individuals who have a higher life satisfaction were more likely to be abandoners than able to realize their intentions (Spéder & Kapitány, 2009, p. 516). The authors showed differences in the processes for men and women. Partnership had a stronger effect for men, and being unemployed was related to a higher likelihood of abandoning their intentions (Spéder

& Kapitány, 2009, p. 517). Among women, those with no religious affiliation were more likely to be abandoners. Among men, those with lower life satisfaction were more likely to be abandoners. Age was significant for both men and women, implying that social pressure was felt by both (Spéder & Kapitány, 2009, p. 519).

In 2012, Kapitány and Spéder followed up with a similar study on the outcomes of positive fertility intentions in four European countries, including data from the Netherlands, Switzerland, and Bulgaria as well as from Hungary. They used two waves from the Generation and Gender survey and two comparable longitudinal survey (Kapitány & Spéder, 2012, p. 600). Kapitány and Spéder found that if the woman in the couple was older, individuals were more likely abandon their intentions than realize them. In the formerly communist countries, if the women was younger, they were less likely to abandon and postpone their intentions (Kapitány & Spéder, 2012, p. 618). In general, postponers were those couples with older women. Parity was shown to be significant in the analysis as well. Having one child was related to a more likely realization, being childless to being postponers and having more than one child to being abandoners.

However, differences between the four countries were pointed out (Kapitány & Spéder, 2012, pp. 618–620). Those who dissolved their partnership were more likely to abandon their intentions than realizing them (Kapitány & Spéder, 2012, p. 620). Educational attainment and religious affiliation showed contradictory results (Kapitány & Spéder, 2012, pp. 620–622).

Regarding employment status, the authors were able to show that Hungarian women with stable employment were more likely in the group of abandoners, while among men, stable employment was related to lower likelihood of abandonment (Kapitány & Spéder, 2012, p. 622).

2.3.3. Stability of related fertility measures

Recent studies on the stability of fertility attitudes also use other measures related to fertility preferences and intentions, namely realistically expected family size and intended family size.

Although the realistically expected family size is distinct from the ideal (preferred) family size, and the intended family size is only loosely connected to the short-term fertility intentions

(16)

12

considered in this study, knowledge about the stability of these two measures provides additional insight.

Buhr and Kuhnt (2012) examined the influence of changes in relationship and employment status on the stability of fertility expectations in Germany, finding that stable relationships are connected to stable expectations. A change from employment to unemployment was not found to be correlated with a reduction in fertility expectations, however, in the opposite case, the expectations were increased (Buhr & Kuhnt, 2012, p. 295).

Iacovou and Tavares (2011) conducted research using data from 17 survey waves in Great Britain, analyzing changes in fertility expectations of women and men. They found that new relationships rather than stable relationships were related to an upward adjustment of expectations, and that experience with having a first child was related to an upward revision if it was positive, and with a downward revision if it was negative. Hayford (2009) analyzed changes in US women’s family size expectations over the span of their childbearing years, showing that the biggest group have normative and stable expectations and are more likely to marry, while those with high and increasing expectations were often from disadvantaged backgrounds, and those with low and decreasing expectations had higher education. Overall, she discovered

“nonmarriage as a major determinant of declining fertility intentions” (Hayford, 2009, p.°779).

As to intended family size, Ní Bhrolcháin et al. (2010) investigated women in Great Britain.

They found that in recent cohorts average intended family size declined with age, and that women over 30 increasingly indicated an expected family size of fewer than two children (Ní Bhrolcháin, Beaujouan, & Berrington, 2010, p. 32). In the Netherlands, Liefbroer (2009) studied the influence of age on the intended family size, using data from six survey waves over the course of 18 years. Family size intentions were found to generally be adjusted downwards with age, but as respondents grew older, the variation between intentions increased. Some were also found to increase their intended family size and others to hold them stable. Respondents without a partner were most likely to report decreased intentions. Educational and occupational careers showed weaker effects on family size intentions. Being full-time employed was correlated with downward adjustment, as was working more hours for both men and women in their thirties.

Women with university degrees showed a steeper decline between age 25 and 35 (Liefbroer, 2008, p. 384).

In summary, especially age has been indicated to influence the stability of fertility preferences.

Fertility preferences and family size intentions decline at higher ages (Heiland et al. 2008;

Hayford 2009; Ní Bhrolcháin et al. 2010; Gray, Evans, & Reimondos, 2013), which is assumed

(17)

13

to be connected to actual or perceived difficulty to conceive. Having a child has been found to be connected to higher fertility preferences (Heiland et al., 2008, p. 150). Parity was also related to the probability to realizing, abandoning, or postponing fertility intentions (Kapitány & Spéder, 2012, pp. 618–620). Changes in relationship status lead to adjustments in preferences and intentions as well (Buhr and Kuhnt, 2012, p. 288); single women have been shown to have less stable expectations than those who are married (Hayford, 2009, p. 777). Being single or separated has been related to a downward adjustment of expectations (Iacovou & Tavares, p. 119) and a lower probability of realizing fertility intentions (Spéder & Kapitány, 2009, p. 514, Kapitány

& Spéder, 2012, p. 620). There were no conclusive findings on the effect of changes in employment status on fertility preferences and intentions. Some results were not statistically significant (Heiland et al., 2008, p. 147; Buhr and Kuhnt, 2012, p. 290), while Iacovou and Tavares (2011) found only a significant association between women’s high income and a downward adjustment of fertility expectations (Iacovou and Tavares, 2011, p. 119). In Hungary, women with stable employment were more likely to have unstable intentions, and for men, stable employment was related to more stable intentions (Kapitány & Spéder, 2012, p. 622). Thus, partnership status seems to have a more straightforward influence than employment status and economic factors (Kuhnt et al., 2017, p.°239).

2.3.4. Realization of fertility intentions and independent variables

A large amount of studies has been conducted on the realization of fertility intentions and on the association between different aspects of female labor market participation and fertility intentions on the individual and aggregate level. Begall and Mills (2011) used data from 23 European countries to found that women with higher levels of work control² were more likely to have positive intentions to have a second child, while higher levels of time pressure in the job were correlated with significantly lower intentions for mothers in societies with low childcare availability. However, intentions to have a second child were higher among mothers who reported a work-family conflict, which was explained by these women being highly invested in both areas of life (Begall & Mills, 2011, p. 433). Bernardi et al. (2015) conducted qualitative research in four European countries and found that uncertain fertility intentions were most often rationalized by “missing elements of behavioural [sic] control […] linked to the labour [sic]

market” (Bernardi, Mynarska, & Rossier, 2015, p.°123), such as unstable employment or unemployment, financial insecurity, and inadequate housing. They also cited difficulties in

2 Work control is a measure of work quality. It concerns an individual’s ability to make decisions over their work tasks, pace, or similar issues in their workplace.

(18)

14

combining their job with childrearing. Financial situation has also been shown to be a relevant factor for Australian women: an increase in satisfaction with finances was associated with an increase in long-term fertility intentions (Gray et al., 2013, p.°147). Shreffler and Johnson (2013) found that the importance of career considerations mediates the relationship between working hours and fertility while studying US-American women. In their study, higher career importance could be associated with decreasing birth odds. They were able to show that “career importance is more important than work hours in driving the negative employment-fertility relationship”

(Shreffler & Johnson, 2013, p. 293). They emphasized the relevance of including preferences and values when investigating fertility behavior. Shreffler, Pirretti and Drago (2010) showed that in US dual-earner couples, neither male nor female respondents’ own subjective work-family conflict predicts fertility intentions, but that men’s fertility intentions were significantly lower if they perceive such a conflict in their partners’ lives. Men with more conservative gender roles were more likely to intend to have a child (Shreffler et al., 2010, p. 235).

On a societal level, German women were found to be less likely to have positive short-term fertility intentions compared to women in Nordic and Central Eastern European countries, being analyzed “in relation to a policy index including FTE of paid maternal, paternal and parental leave, and the availability and cost of childcare” (Fahlén, 2013, p. 655). In Portugal, subjective childcare load was shown to indirectly influence childbearing intentions via the perceived justice of the division of childcare on a sample of 82 women (Andrade & Bould, 2012, p. 33). Spéder and Kapitány (2014) investigated the realization and failure to realize fertility intentions in Western and Eastern European countries and were able to show an East-West difference (Spéder

& Kapitány, 2014, p. 399). Individuals in France and Germany who planned to have a child in the short-term were able to realize this intention at a higher degree than those in the post- communist countries Hungary, Georgia, and Bulgaria (Spéder & Kapitány, 2014, p. 407). If the female member of the couple was in employment, individuals were less likely to realize their intentions (Spéder & Kapitány, 2014, p. 407).

In this thesis, the stability of fertility preferences and of positive short-term intentions will be analyzed. As summarized above, existing research includes some similar studies, which however either used different measures or included only one or two waves. Also, the focus was often on age, relationship status, or employment status as explanatory variables, and these were not consistently found to be significant. I argue that individual experience with working and family life have been overlooked as a likely influence on both fertility preferences and intentions.

Women with children are faced with combining their occupational and childbearing career and are likely to revise attitudes after experiencing difficulties and becoming “more aware of the

(19)

15

conflict” (Brewster & Rindfuss, 2000, p. 282). In countries with weaker work-family policies such as Germany, women who report high work importance have more to sacrifice once they have a(nother) child. Those with lower work importance would be more willing to have more children and disrupt their careers. Their labor force attachment is lower (Fahlén, 2013, p. 657).

Experiences or expectations with childcare, parental leave, and re-entering the labor market, are likely to influence future intentions and even preferences. Importance of work and career, satisfaction with childcare, and the perceived effect of children on the career are thus introduced as possible factors. These experiences and attitudes have not yet been explored by other researchers in relation to stable preferences and positive intentions.

(20)

16

3. Research hypotheses

This section introduced six hypotheses that were derived from the theory and previous research discussed above.

3.1. Fertility preferences

Hypothesis 1: Fertility preferences are expected to be mostly stable.

Hypothesis 2: Women and parents are more likely to hold unstable preferences.

Hypothesis 3: Individuals who attach high importance to work and career, are less satisfied with childcare arrangements, and perceive more negative effects of children on the career, are more likely to hold unstable preferences.

3.2. Fertility intentions

Hypothesis 4: Positive short-term fertility intentions are expected to be less stable than preferences.

Hypothesis 5: Individuals who attach high importance to work and career, are less satisfied with childcare arrangements, and perceive more negative effects of children on the career are more likely to hold unstable short-term fertility intentions.

Hypothesis 6: Individuals who are unemployed, not in employment, hold temporary work contracts, and are less satisfied with their financial situation are more likely to hold unstable short-term fertility intentions.

(21)

17

4. Data and method

4.1. Data set

The German Family Panel pairfam (“Panel Analysis of Intimate Relationships and Family Dynamics”) is a multidisciplinary longitudinal study. The planning, collection, and preparation of the data is a cooperative effort linking five German Universities in a collaboration with the Max Planck Institute for Demographic Research. The project is overseen by an international board and funded by the German Research Foundation (DFG)³. The aim is to investigate couple and family dynamics, with a special focus on partnerships, childbearing, attitudes, and values.

The questionnaire includes retrospective questions on family history, education and employment, and a self-assessment of personality, health, and satisfaction. Since the initial survey in 2008, annual follow-up studies have been conducted. To date, eight waves are available, providing information on more than 12,000 randomly chosen persons of the birth cohorts 1971-73, 1981-83, 1991-93 and their partners, parents, and children (Brüderl et al.

2017a).

The unique richness of data on attitudes, values, and partnerships, family, and employment status, is of great use for this research topic. This accumulation of detailed information on the respondents’ life events is rare and has not been exploited for many life course studies. The interviewed cohorts are sufficiently young to analyze the change in fertility preferences and intentions in childless individuals, young parents, and even those farther into their reproductive careers and the gaps of seven years allow for a comparison between the cohorts. Especially the middle cohort is followed closely during the years of the first childbearing⁴, while the oldest cohort will have mostly finished their fertility career, having reached an average age of 44 in the year of the last interview.

For this study, seven of the eight waves of the pairfam study and its supplement DemoDiff⁵, release 8.0, are used. The DemoDiff study on Eastern German individuals closely follows the design of pairfam and was integrated into the pairfam sample after wave 4. The first DemoDiff wave was conducted parallel to the second pairfam wave (Brüderl et al., 2017b, p. 108). The first wave is excluded, as one of the independent variables, satisfaction with the childcare situation, has only been measured from the second wave forward.

3 A detailed description of the study can be found in Huinink et al. (2011).

4 They are aged 26 to 33 on average.

5 DemoDiff stands for “Demographic Differences in Life Course Dynamics in Eastern and Western Germany”.

(22)

18

4.2. Sample

The pairfam study has relatively good response rates, but panel attrition is still a concern. The original sample in the first wave, which will not be included in the analysis, had 12,402 respondents. The following samples sizes decreased with every wave, as previous respondents declined to participate further or could not be reached, which can have several reasons. The overall attrition from wave 1 to wave 8 was 46%, which is slightly lower than in the expected range of lengthy panel studies in Germany (Müller & Castiglioni, 2015a, p. 386). A suspected significant nonresponse bias could not be confirmed (Müller & Castiglioni, 2015a, p. 403), although for example the youngest cohort and migrants were found to have higher attrition rates (Müller & Castiglioni, 2015a, p. 400), and an association between partnership dissolution and later attrition was detected (Müller & Castiglioni, 2015b, p. 214). The respondents from the supplement DemoDiff were included in the pairfam sample from wave 5. No new individuals were recruited to the main data set used for this analysis.

Table 1: Number of respondents per wave, pairfam and DemoDiff data sets. Own illustration

This study does not include all individuals and all observations for its models. 89 observations and 8 individuals were dropped because of inconsistencies in the number or type of children they reported, reducing the general sample to 65,178 observations and 13,874 individuals. The first wave was dropped. From this point, two different samples were created from the same basis sample to fit the two different dependent variables. Dropping observations or individuals because of missing responses in one of the two dependent variables (with no missing response in the other dependent variable) would undesirably reduce the sample size.

For the sample for the analysis of fertility preferences, four individuals were dropped as they had implausible responses, and 41 observations were deleted as the answers, seen as compared to the answers in the other waves, were seemingly mistakes on the interviewer’s or interviewee’s side.

At the same time, 100 individuals (1,265 observations) who did not wish to answer the question or stated, “do not know” were deleted from the data set. At this stage, 10,960 individuals and 51,498 observations were left. This was reduced by those who had missing data in independent variables. Only individuals with more than one observation were included in the models. The final sample of all individuals and observations included in one of the samples on fertility preferences consists of 8,402 individuals and 33,381 observations. 1,098 individuals appear in two samples as the number of children changed throughout the survey years.

Wave 1 Wave 2 Wave 3 Wave 4 Wave 5 Wave 6 Wave 7 Wave 8

pairfam 12,402 9,069 7,901 6,999

DemoDiff - 1,489 1,173 1,074

Total sample 12,402 10,558 9,074 8,073 7,330 6,710 6,109 5,701

7,330 6,710 6,109 5,701

(23)

19

For the fertility intentions, observations and individuals are excluded who report to be infertile or have an infertile partner, as they are not asked about their fertility intentions. This reduced the sample by 403 individuals and 4,686 observations. As many individuals never or only in a few observations report positive fertility intentions, many observations were dropped from the data set. This results in a sample that includes 2,784 individuals and 5,898 observations. Those observations that have missing information for relevant variables are excluded again. The final data set for the analysis of fertility intentions includes 2,496 individuals and 5,178 observations.

Table 2: Sample sizes (individuals and observations) of separate models to be used in forthcoming analyses. Own illustration

4.3. Dependent variables

The dependent variable for the first part of the question, the change in fertility preferences, is the change in ideal number of children. Studying the individual ideal number of children indicated in surveys requires a discussion about possible issues with this measure. First, it is crucial to understand what respondents address: perceived societal norms regarding family size or their own family size ideal (Chen & Yip, 2017, p.°586). Here, the question directly addresses the individual’s own preferences: “Assuming ideal circumstances: How many children would you like to have altogether?”. In the original German, this question reads: “Wenn Sie einmal alle Hindernisse außer Acht lassen: Wie viele Kinder würden Sie im Idealfall insgesamt gerne haben?”, which literally translates to: “If you could put all obstacles aside for once: How many children would you like to have in the ideal case altogether?”. It is an open question to be answered with any number. However, even though the individual ideal number is inquired, it is absurd to assume that general societal norms do not form or manipulate the respondents’

attitudes. Also, some might adjust their answer to fall inside the norm if they do not want to be judged by the interviewer. Moreover, respondents might interpret “ideal circumstances” as actually unattainable, such as “if I were younger”, and thus not give realistic answers. The ideal family size can be understood as “wishes and emotional feelings without containing any commitment to act” (Testa, 2012, p. 7). These considerations have however little bearing for this study, as it is not concerned with the gap between desired and actual fertility, but with the change

individuals observations individuals observations

Women, no child 2,450 9,484 788 1,664

Women, 1 child 998 2,917 505 841

Women, 2 + children 1,428 5,146 182 252

Men, no child 2,927 11,002 875 1,672

Men, 1 child 672 1,844 360 566

Men, 2 + children 857 2,988 133 183

Fertility preferences Fertility intentions

(24)

20

in an individual’s ideal number over their life course. Essentially, the fact that there is such variance in answers over multiple waves, partially overcomes concerns with the measure, as societal norms would not be reported differently every wave. Also, if ideal circumstances would be interpreted as an unattainable, unrealistic situation, the number of children would similarly be expected to be stable. Of the 8,402 individuals in the cleaned data set for the analysis, 55% have reported at least two different values for ideal number of children. The question was posed to all respondents and in all conducted waves. Those observations that had an unreasonably large reported ideal number of children were examined and excluded when they were larger than 10.

The answers thus range from 0 to 10. As the stability of the ideal number of children is of interest, the dependent variable is created as a first difference variable; the difference between number reported in the previous and current wave. It ranges from -8 to 8.

The second part of the question addresses changes from positive short-term fertility intentions.

The interviews included several questions on fertility intentions. A diagram of the structure of these questions is included for visualization:

Figure 2: Diagram of filter questions in questionnaire. Cf. Buhr & Kuhnt (2012)

First, general or long-term intentions were inquired: “When you think realistically about having [additional]⁶ children: How many [more] children do you think you will have? Here we mean children in addition to the ones you already have, or if you or your partner is pregnant, in addition to the child you are expecting.” The answer options were: 0“No (additional) children”, 1“One (additional) child”, 2“Two (additional) children”, 3“Three (additional) children”, 4“Four or more (additional) children”, 5“I'm not sure” and 6“I haven’t thought about that”. These long- term intentions were not considered in this study, as they are less predictive than the short-term intentions.

The following short-term fertility intentions question was only posed to those who had replied that they are considering having (more) children or who are unsure (answer options 1-5), and who are not infertile and whose partner is not infertile. These items were inquired earlier in the questionnaire of all respondents. The question reads: “Do you intend to have a [another] child within the next two years?”. The answer options are as follows: “Yes, definitely”, “Yes, perhaps”, “No, probably not”, “No, definitely not”, “I haven't thought about that”. To simplify

6 The words ‘additional’ and ‘more’ are added if the respondent has at least one biological, adopted, or step- child, is pregnant, or has a has a partner who is pregnant.

(25)

21

the model, and as routinely done in similar studies, the answers have been summed up as “Yes”

or “No/Haven’t thought about that”⁷. The dependent variable is set up in such a way that the three possible outcomes of positive short-term fertility intentions in the previous wave can be compared. It is coded 0 if the fertility intentions remain positive from one wave to the next, 1 if a new biological child has been conceived and thus the intentions realized, and 2 if the short- term fertility intentions are no longer positive, although they have been positive in the previous wave and were not realized since then. This question is asking the respondent to take into account their current life situation and that of the foreseeable future. Generally, it is expected that actual and anticipated financial situation, employment and relationship status would be considered in answering this question (Bernardi et al., 2015, p. 123), while they would not be of relevance in answering the question on fertility preferences.

Due to the manner in which both dependent variables are coded and the focus of this study on the stability of the measures, all individuals’ first survey wave has been dropped from all models.

4.4. Independent variables

Three categories of independent variables are introduced in the models to investigate the change in fertility preferences and the change from positive fertility intentions: The main explanatory variables, the control variables on employment status and relationship status, and the control variables measuring socio-demographic background and situation. As the sample sizes are small for some of the models for fertility intentions, and the variance in some variables was too small or inexistent⁸, it was impossible to calculate models including all variables that could be estimated for the models on fertility preferences. Thus, the variables whose categories were critically small were excluded. This concerns the partner-variables which proved to be problematic in some of the models, as the multinomial logistic regression could not converge.

Regrettably, explanatory power was lost with the exclusion of these variables, and the decision was made to at least include the partner-variables for the analysis of fertility preferences to capture the effects. The comparability is limited, but the knowledge that parents with positive fertility intentions in these samples are almost exclusively partnered might excuse this.

7 Here, “Haven’t thought about it” is seen as a non-positive response, considering that in the previous wave (~12 months ago), the answer was “Yes, definitely” or “Yes, perhaps”. Surely, the respondent had “thought about it” in the past, but now feels too unsure to have positive intentions.

8 This concerns the models for women and men with one or two and more children. Only those with positive fertility intentions at one point in the survey were included in the model, and in these four models, only very few or no individuals have no partner.

(26)

22 4.4.1. Explanatory variables

The main explanatory variables concern the respondents’ attitudes towards and experiences with work, childcare, and the compatibility of career and childrearing. Regarding attitudes, the variable career importance measures the importance the respondent ascribes to the life goal and life domain “pursuing my education or career interests” as compared to four other “goals and domains”, namely “pursuing my hobbies and interests”, “keeping in touch with friends”, “living in a partnership” and “having a(nother) child”. Among these five life domains or goals, 15

“importance-points” had to be distributed. Assigning the value 0 to one or several domains was allowed. The variable career importance ranges from 0 to 15. For parents, experiences are measured as well. Satisfaction with the childcare situation of a child (the younger child for parents of two or more children) is captured as satisfaction with childcare on a scale from 0 to 10. This includes formal and informal care and is reported for all parents. The third independent variable measures the attitude of non-parents and the experience of parents. It shows respondents’ assessment of the effect of being a parent on the life domain of “school, education, career”. Parents and childless individuals answer the questions: “How does being a parent affect other areas of your life at the moment?” and “How do you think being a parent will affect other areas of your life?”, respectively. The answers to the variable effects children on career range from 1 “very positively” to 5 “very negatively”. A fourth explanatory variable concerns the experience with taking parental leave. Parental leave is introduced to measure the effects of having taken parental leave.

4.4.2. Control variables

The first set of control variables concerns the employment status and conditions of the respondent and his or her partner. To ensure that the explanatory variables actually capture the attitudes towards work and the combinability of career and childrearing and are not confounded with the realities of respondents’ employment and relationship, these have to be controlled for.

The variable employment status is set up to control for the following states: being employed, not employed, and unemployed. The category employed includes those being employed full-time, part-time or self-employed, those in military or civilian service, and those holding down another type of job. Respondents were coded as not employed if they are currently enrolled, report being an intern, a trainee, on work experience, or on leave. This also includes housewives and - husbands and those in (early) retirement or with an occupational disability. As argued in previous studies using the pairfam data, those in marginal employment or in occasional or irregular employment as well as those who are unemployed are coded as unemployed (Kuhnt & Trappe, 2016, p. 21).

Stability of fertility preferences and intentions

Stability of fertility preferences and intentions

A new angle on studying fertility behavior in Germany

Antonia Spath

Stability of fertility preferences and intentions

Abstract

Table of Contents

List of Tables

List of Figures

1. Introduction

2. Theory and previous research

2.1. Fertility preferences and fertility intentions

2.2. Stability of fertility preferences and intentions

2.3. Previous studies

3. Research hypotheses

3.1. Fertility preferences

3.2. Fertility intentions

4. Data and method

4.1. Data set

4.2. Sample

4.3. Dependent variables

4.4. Independent variables