In this book, Jonas Wood offers a comprehensive analysis of socio-economic differentiation in reproductive behaviour. Extensive geographical coverage also enables him to take into account differences in economic context and social policy across European countries.
The book opens with a well-written introductory chapter summarizing the relevant theoretical ideas; the author presents not only the standard economic arguments (new home economics) but also cultural theories (second demographic transition) and the institutional aspects (welfare state, norms, gender equality) of differential fertility.
The four main chapters present original empirical analysis; an additional chapter focuses on data quality. Each chapter is structured as an independent research article as the book is derived from Wood’s doctoral research at the University of Antwerp under the supervision of Karel Neels. Jonas Wood holds a PhD in sociology but his work is strongly interdisciplinary, covering cultural, economic, demographic and institutional aspects. Wood also holds a Master’s degree in statistics and uses sound statistical and econometric methods.
The question of how varying configurations of economic and institutional characteristics across European countries coincide with different demographic behaviours is of strong policy relevance in a context where below-replacement fertility levels represent a serious challenge to the financial sustainability of pension systems and public health care. The macro-economic contexts Wood addresses vary from education expansion to economic fluctuations and family policy settings. The combination of individual-level survey data, primarily from the Generations and Gender Survey (GGS) and the European Social Survey (ESS), and aggregate-level data (OECD, World Bank) make a cross-national comparative approach possible. The surveys focus on fertility, education and employment viewed retrospectively at the time of the interview (i.e., information is available on length of education, time elapsed since first cohabitation, since first job, etc.).
Throughout the work, the issue of self-selection bias for higher order childbearing behaviour is accurately taken into account. Women at risk for a second or third birth may constitute a selective group as they already have one child. Self-selection into the group “at risk” for having a second or third child could be linked to socio-economic characteristics such as education. The opportunity costs of having a first child are heaviest for highly educated women, and those who decide to do so despite those costs might have particular characteristics – e.g., being more family-oriented – that are potentially related to the probability of their having more than one child. This selectivity may affect the impact of education on the transition to second or third childbirth as well as sensitivity to economic and institutional context. To control for this, Wood proposes a random effects discrete-time hazard model, in which the random effect is included at the individual level (shared frailty). This allows for controlling for time-constant unobserved individual-level characteristics and therefore captures selectivity in connection with the transition to parenthood. The author finds that women-specific characteristics connected to timing of entry into motherhood have little impact on the educational gradient in progression to second and third births. Selection in terms of timing and occurrence of first births thus does not affect educational patterns in progression to second and third births. This is an important and helpful insight for research based on datasets where the panel is too short to apply shared frailty models.
The first research chapter or essay investigates the educational gradient in completed fertility. Cohort parity progression ratios to the first, second and third birth up to age 39 are estimated as a function of education for women born between 1940 and 1961 on the basis of data from 13 European countries (GGS). The regression-based proportions are then illustrated by cohort and education level. No causal relationship was found, since not only is the number of children a woman has at the end of her childbearing life likely to be influenced by education level but fertility level may also determine education level. For entry into parenthood, Jonas Wood finds a significant negative correlation: the lower the education, the higher the cohort progression ratio to first birth. The negative educational gradient is weak in Norway, Belgium, the UK and CEE countries and strong in Italy, Spain and the Netherlands. In all countries, the negative coefficient decreases between cohorts 1940/1944 and cohorts 1956/1961. For second births, country results are more heterogeneous. The educational gradient remains negative in Italy, but is also negative in Bulgaria, Estonia, Romania and Russia. A U-shaped curve (i.e., women with mid-level education showing the lowest progression ratio to second births) is found for France and Hungary. The pattern is flat in Australia, the UK and Spain, at least for the youngest observed cohort (1956/1961), while a positive pattern is found for Norway, the Netherlands and Belgium. For third births, a negative association is confirmed for the UK, Spain, Bulgaria, Estonia, Romania and Russia, while the pattern is U-shaped for Norway, Australia, Belgium and France. For all birth orders, the strongest variation in between-country progression ratios is found for highly educated women (tertiary education).
The strong country heterogeneity points to the importance of country context. Institutions seem to play a crucial role, especially for highly educated women, when it comes to deciding whether and when to start or enlarge a family. The finding of consistent negative educational gradients in first births across Europe is of high policy relevance. Jonas Wood mentions work-family incompatibility as a possible explanation for countries with a negative gradient, but between-country differences in institutional settings are not explicitly modelled in this chapter. Multilevel modelling is provided in the other chapters. However, to reinforce the hypothesis of the importance of institutional context in this chapter, it might have been helpful to perform regressions by country groups to distinguish those with high versus those with low work-life balance support. Another shortcoming, which the author acknowledges, is that due to missing data in the GGS, partners’ characteristics are not controlled for and male fertility decisions not modelled. Controlling only for female education means a risk of capturing male characteristics but interpreting them as female ones. For example, assortative mating can lead to progression ratios that are lower for highly educated women if their likewise highly educated partners tend to have lower fertility preferences. Another possible interpretation is that women’s progression ratios are higher in households less subject to budget constraints due to those women’s high-income partners. These claims are purely hypothetical, but they illustrate the importance of taking into account interactions between women and their partners. Without information on male education, we do not know the extent to which educational gradients vary by gender or how interdependent they are. The results must therefore be interpreted with caution–a challenge that the author meets very well.
The second research chapter investigates the relation between economic conditions and the timing of first births for 22 European countries (including CEE countries) from 1970 to 2005 (ESS). Differences between regions, age groups, education levels and institutional context are taken into account by applying multilevel discrete-time event-history models of first birth hazards. The macro-level variables used are the consumer-price index and unemployment rates, which are interacted with country cluster dummies (Northern, liberal, Western, CEE, and Southern). Wood finds that for all countries combined, labour force entry is an important precursor to motherhood, especially for highly educated women. Aggregate unemployment is found to be negatively associated with first birth hazards, especially in Southern European countries. In Nordic and Western European countries, highly educated women are more sensitive to aggregate-level economic context than in CEE and Southern countries.
As in the previous chapter, the survey data does not allow for controlling for partner characteristics, which is why potential interactions between women and their partners are not discussed. However, the results indirectly provide some indications on those interactions. For example, Wood finds that in Southern Europe, women’s labour market status and job duration is not significant for entry into motherhood whereas aggregate unemployment is. This strongly suggests that it is the partner’s employment status that is crucial in the family’s decisions on whether and when to have a first child. In countries with a dominant male breadwinner model, having a partner who is unemployed or at risk of losing his job represents a major barrier to starting a family. In more gender-egalitarian countries, where women as well as their partners generate family income, it is women’s own successful labour market integration that emerges as an important determinant of motherhood. Thus, despite the lack of information on partner characteristics, multilevel modelling does allow for discussing some potential partner interactions.
Multilevel models are also applied in the third research chapter, where Wood investigates the impact of family policy on childbearing for seven European countries (Belgium, Germany, France, Norway, the Netherlands, Spain and the UK). Availability of formal childcare for children aged 0-2 and family allowance amounts are modelled as second birth determinants, differentiated by age, country and education. Survey data (mainly from GGS) is combined with aggregate-level data from the OECD Family Data Base (1980-2002). Policy measures are found to encourage the transition to a second child for all educational groups. However, family allowances are found to have a stronger positive impact on propensity of second childbirth for women with little education, while childcare availability has a more positive impact for highly educated women. Significant positive impact is mainly driven by between-country variation, whereas within-country variations in second childbirths over time are much less related to variations in family policy settings. The weak results for models with country-fixed effects are probably due to the fact that the time period only runs to 2002. In European countries with relatively high fertility levels, childcare coverage increased during the 2000s, a change that went together with an increase in total fertility rates back to replacement level.
The limited time period is also problematic in the last research chapter, where Wood analyses the impact of parental leave (uptake after birth of first child) on parity progression in Belgium, France and Germany. For Germany and France, the time period studied runs only to 2004. During that period in Germany, parental leave was paid as a lump-sum transfer for a maximum period of 24 months. This form of parental leave was criticized for its polarizing effects: it detached many women from the labour market, particularly those with little education, whereas others, particularly highly educated women, often did not benefit from leave because the income compensation was too low. In line with this observation, Wood finds no significant effect of women’s parental leave uptake on transition to higher order births in Germany except for women with little education, and a significantly negative effect for men. The effect of lump-sum parental leave is found to be positive in France and Belgium, but it is important to keep in mind that in those countries leave is paid for a much shorter time period for the first child.
Germany reformed its parental leave system in 2007, following the Swedish model by introducing a wage substitution for a shorter time period (67% of former net wage for a maximum of 12 months). The purpose of the reform was to encourage female employment before and after childbirth. The income effect is supposed to facilitate transition to higher order births. However, the expected positive effects are slow in coming, as childcare coverage is still low in Germany and many women have difficulty returning to a full time job after parental leave. As Wood explains, this shows that the impact of parental leave depends not only on its design features but also on family policy context. The author concludes that, given the fact that different educational groups respond differently to institutions, comprehensive family policy packages are needed rather than singular measures.
Overall, by bringing together the different research chapters into one book, Wood succeeds exemplarily in highlighting the interplay between individual characteristics and institutional context and the importance of the latter for fertility decisions. He illustrates clearly and intelligibly that for highly educated women work-family compatibility is of major importance, while for women with relatively little education, exclusion from the labour market represents the major barrier to starting and enlarging a family.
The author reaches this conclusion based on solid empirical modelling that carefully and correctly takes into account selection effects. The control for differential selectivity on the risk set for higher order births is applied by mobilizing retrospective survey data. The longitudinal design makes shared frailty models possible, which model births as repeated events with a random term at the individual level. The pitfall of using GGS and ESS data is lack of data on partners and of more detailed information on employment characteristics (weekly working hours, for example). Other European survey data, such as the European Survey of Income and Living Conditions (EU-SILC), include partner information and more specific employment features and thus could be used as alternative datasets. However, the short follow-up in SILC does not allow for applying shared frailty models and the quality of fertility measures is not perfect either. It therefore seems that analyses of socio-economic differentiation in European fertility based on GGS, ESS and SILC data are complementary approaches. Combining and comparing the different results seems a fruitful direction for future research.