The Gap Between Lifetime Fertility Intentions and Completed Fertility in Europe and the United States: A Cohort Approach

The fertility rate of earlier cohorts may be taken as a proxy for prevailing cultural childbearing norms during childhood and adolescence. We thus use the fertility level of cohorts born in 1950–1954, i.e., around 15 years earlier, to represent family structure while growing up. Where previous cohorts had higher numbers of children, cultural norms of large families prevail (Testa and Grilli 2006), and if people experienced growing up with many siblings, their fertility intentions tend to be higher too (Axinn et al. 1994; Kotte and Ludwig 2011; Régnier-Loilier 2006). Therefore, in countries with a strong decline of fertility over cohorts, a marked gap between reported intentions and behavior is likely to arise. The fertility decline was most pronounced in the Southern European countries (Greece, Spain, Italy, and Portugal) as well as in Bulgaria (see Table 1). Compared to Southern Europe, the drop in fertility rates in the late-1960s and early 1970s cohorts was generally smaller in the Central and Eastern European countries (Frejka and Calot 2001). With regard to childlessness intentions, we refer to the prevalent social norms (see Table 1). Indeed, previous studies revealed a clear East–West divide in the cultural acceptance of childlessness, with Eastern European populations clearly opposing childlessness (Merz and Liefbroer 2012).

We examine three explanations for variations in the actual number of children: economic situation, work–family combination, and unplanned births. Past studies confirmed that economic conditions are central for men’s and women’s fertility behaviors. A negative correlation between the unemployment rate and the total fertility rate has been documented (D’Addio and Mira d’Ercole 2005) and individual-level research showed that unemployment may lead couples to delay or forgo childbearing (Adserà 2011; Kreyenfeld and Andersson 2014). In 1995–2004, Western Europe and the US were least affected by unemployment, although the unemployment levels were substantial in France, Belgium, and Germany (see Table 1). The unemployment rate was particularly high in some Southern European countries (notably in Spain and Italy) and in most of the East, particularly in Bulgaria and the Baltic states, which was linked to a post-1989 economic depression.

Second, work–family compatibility is another important factor affecting fertility levels. There are several possible indicators to measure the conditions of work–family reconciliation. In terms of family policies, childcare services and childcare leaves are two key instruments (Matysiak and Węziak-Białowolska 2016), and we focus here on childcare services.⁴ In addition, we present the employment rate of mothers with children below age 15 as a measure of the outcome of work–family conditions (Table 1). Well-paid parental leave of moderate length and a well-developed childcare infrastructure strengthen women’s ties to the labor market (Dearing 2016) and are related to higher fertility rates (Luci-Greulich and Thévenon 2013). There is also a positive correlation between the female labor force participation rate and the fertility rate (Ahn and Mira 2002; Engelhardt et al. 2004). As shown in Table 1, mothers’ employment rates are particularly low in the Southern European countries (with the exception of Portugal), which are characterized by weak family support policies, e.g., low childcare participation and low availability of part-time work (Adserà 2004, 2005; Del Boca et al. 2009). Several Central and Eastern European countries report low employment rates for mothers (especially Hungary) and often low enrolment rates for children below age three (the Czech Republic, Bulgaria, and Hungary). In some of these countries—particularly in the Czech Republic and Hungary—mothers take rather long leaves after childbirth (OECD 2017). The Central and Eastern European countries are similar to the South of Europe in their labor market structures (full-time work, high unemployment), but have higher public spending on family benefits (OECD 2014). After the regime change (when the women were surveyed), there was a rapid and dynamic societal transformation which included labor market reforms, social benefits, and family policies (Frejka and Gietel-Basten 2016). In Germany and Austria, mothers’ labor force participation rates are comparatively high while childcare enrolment rates of children below age three are very low: mothers commonly take long employment breaks after the birth of a child and tend to return to the workplace on a part-time basis (Berghammer 2014; Konietzka and Kreyenfeld 2010). Switzerland differs in that early childcare enrolment is frequent, but kindergarten opening hours are very short. In the UK and The Netherlands, women tend to re-enter the labor market rather fast after the birth of a child, mostly on a part-time basis. In the other countries, including France, Belgium, the US, and the Nordic countries, childcare enrolment rates are high and women combine full-time employment with childrearing duties.

Third, we refer to the prevalence of unplanned births due to contraceptive failure and restricted access to abortion. A high prevalence of unplanned and unwanted births is expected to reduce the gap between intentions and realized fertility. While variation in the rate of contraceptive use was modest across countries in the 1990s (when the women in our sample were generally between 20 and 29 years old), there were major differences with regard to the reliability of the contraceptive methods used (Makay 2015). Couples in Western European countries and several more affluent countries of Central and Eastern Europe used highly effective means of contraception (pill, IUD, and condom), while less reliable methods (rhythm and coitus interruptus) were more widespread in less prosperous and more Catholic Central and Eastern European countries (see Table 1). Slovakia, Lithuania, and Bulgaria report the lowest use of effective contraception; in the first two countries, this is related to the Catholic Church’s opposition to artificial methods of contraception (Stloukal 1999) and in Bulgaria to the high costs of access (Vassilev 1999). In Bulgaria and elsewhere across the Central and Eastern European region, the low prevalence and limited supply of effective contraception were partly compensated for by resorting to abortion; rates of legally induced abortion were overall much higher in the East than in the West of Europe (Table 1). In Italy and Greece, though contraception was used, effective contraception was used less frequently than in most other Western European countries.

In conclusion, based on these dimensions, we formulate the following country- and region-specific expectations.

Our study also examines the variation in the fertility gap between women with different levels of education between ages 25 and 29. Due to data constraints, we perform the education-specific analysis for a subset of 11 countries only: Southern Europe (Italy, Spain), Western Europe (Austria, Germany, Belgium, the UK, The Netherlands, Norway, Switzerland) and the US, Central and Eastern Europe (Czech Republic, Hungary). Prior research found no clear education gradient in lifetime fertility intentions (Berrington and Pattaro 2014; Sobotka 2009; Testa 2014), so we base our expectations about the magnitude of the fertility gap by education on the differentiated constraints to actual fertility by education. Besides possible differences in unplanned births, we consider education-specific labor market opportunities linked to the economic situation and opportunities for combining work and family.

Based on these factors, we formulate the following expectations.

In our study, we compare fertility intentions reported by 20- to 24-year-old women born between 1965 and 1979 (with the majority born between 1970 and 1975) to completed fertility at age 40 in the same birth cohorts. For the education-specific analysis, we focus on the 25 to 29 age range because most will have completed their education (see Table 2 for age at completion of full-time education). At this age, women still have a relatively long timeline to fulfill their reproductive plans, although in some subgroups (especially among the less educated) a large proportion of women have initiated childbearing in their early 20s, and could already have reached their completed fertility (Rendall et al. 2010; Rendall and Smallwood 2003).⁵ In a sensitivity analysis, we found that the country ordering for the fertility gaps were consistent for the 20 to 24 and 25 to 29 age groups, so the patterns were not very sensitive to the specific age range. We could not disaggregate by education, given that fertility data by level of education are not yet available for these later birth cohorts.

The substantial sample sizes of the FFS (more than 500 women 25 to 29 years old in most surveys) enabled us to categorize observations into three different educational groups (see Table 4 for sample sizes and data sources). We excluded Lithuania because the education categories could not be reconciled with the ISCED classification. Data for completed fertility and cohort total childlessness by level of education were not available in a few other countries. Data on final parity were either provided in the Cohort Fertility and Education Database (Zeman et al. 2014), or recalculated based on the Generations and Gender Surveys that took place in the early 2000s. We used the common classification of educational levels into low (ISCED 0–2; up to lower secondary), medium (ISCED 3–4; upper secondary and post-secondary non-tertiary), and high (ISCED 5–6; tertiary). Since information on the level of education pertains to the time of the interview, we had to make the assumption that, in the cohorts studied, education was stable over time, from their mid- to late 20s.

Our analytical approach is as follows: we undertake a side-by-side comparison of different countries for which we provide information on the macro-context, without testing contextual variables in macro-level models. This approach is widely used, but mostly—although not exclusively—for a smaller number of countries (Yu 2015). We settled on this approach because of the lack of contextual data and information on the years of giving birth for a time span of around 20 years (from young adulthood to the early 40s) and for 20 countries: too many data points were missing to estimate macro-level models.

Our research design has several limitations. First, an aggregate approach naturally precludes any statements about the realization of intentions on an individual level. Because we do not follow individuals over time, we cannot know the extent to which women over- or under-achieve their fertility intentions and how this differs across countries. Second, it may be considered a drawback that the cohort fertility gap is only measured after a cohort has completed its fertility (that is, after an approximately 20-year period of childbearing), incurring a long lead time. This is different from period measures which are available for recent periods. Third, some researchers have critically noted that lifetime fertility intentions are difficult to measure. Respondents do not necessarily have an attitude formed at the time of the interview (Bachrach and Morgan 2013) and possibly make up whether they want children or not, as well as how many, because of the coercive nature of the interview situation and possibly influenced by social norms and desirability (Ní Bhrolcháin and Beaujouan 2019). Fourth, it is a general concern in cross-national research that deviations in question wording, differences in survey protocols (e.g., representativeness of the samples), and in survey quality (e.g., high non-response rates) may affect the results (Beaujouan 2013).

Our exploration of the aggregate gap between fertility and childlessness intentions in young adulthood and completed family size and final childlessness in the same birth cohorts brings new insights to the discussion on the “fertility gap.” Inspired by Bachrach and Morgan’s approach (2013), we took the perspective that intentions at young ages were defined by societal fertility norms and family context, and that institutional conditions in peak childbearing years have a large impact on the capacity to meet fertility goals. Our comparative approach allowed us to suggest explanations to why the size of the fertility gap differs across countries and educational groups.

Our results reveal quite distinctive regional patterns for women born (mostly) in the early 1970s. We observe the largest fertility gaps in the Southern European countries (more than 0.6 children per woman in Italy, Greece, and Spain). This is in line with our expectations: we anticipated low completed fertility in the South based on unstable labor market conditions and little support for reconciling work and family life. In these countries, the fertility gap was boosted by the relatively high fertility intentions in young adulthood, certainly driven by traditionally large family sizes. For Central and Eastern European countries, we identified countervailing forces that might influence the fertility gap: poor economic situations and difficulty combining work and family (both being conducive to a small family size), partly counterbalanced by a high prevalence of unplanned births. The intended family size was rather low in that region (around two children per woman in most countries), corresponding to relatively low fertility achieved in the parental cohort. Although we found rather small gaps in most Central and Eastern European countries, in line with our expectations, higher fertility gaps of 0.3–0.4 children per woman were displayed in the region’s weaker economies (particularly the Baltic States). Among the Western countries, in the UK, the US, and France, rather high fertility together with higher fertility intentions (in line with previous cohorts’ fertility levels) resulted in a small to medium gap. It may seem surprising that both France and the US have the smallest gaps between mean intended and total fertility, because France has more developed and generous family policies than the US (Crittenden 2001). However, the high share of “overachievers” in the US (e.g., teenage mothers) may be responsible for this finding (Berrington and Pattaro 2014; Morgan and Rackin 2010; Quesnel-Vallée and Morgan 2003). In the German-speaking countries, we would have expected larger gaps because the support for work–family reconciliation is weak, but we found that this larger gap only applies to Switzerland. Indeed, Germany and Austria are among the countries with the lowest mean intended family sizes, in line with their long history of low fertility, which resulted in moderate gaps. Switzerland, however, had persistently high fertility intentions, despite its long-term low fertility.

How do these results compare to previous research? Studies on the UK and the US have indicated fertility gaps of a magnitude of 0.2–0.3 children per woman for slightly older cohorts than studied here (Morgan and Rackin 2010; Smallwood and Jefferies 2003). Comparing all the countries studied, these gaps are relatively moderate and, depending on the region, they may be significantly higher: fertility gaps are, for instance, about twice as large in the Southern European countries. Based on the intended number of children and the tempo-adjusted total fertility rate, gaps of around 0.3–0.4 children per woman had been reported as a European average which closely correspond to the mean of all the countries we included (Sobotka and Lutz 2010). However, the regional pattern obtained with the tempo-adjusted method clearly diverged from ours as moderate gaps were noted for Southern Europe and large gaps for Central and Eastern Europe (Sobotka and Lutz 2010).

In addition to the mean number of children, we presented evidence on childlessness. The assumption that both childlessness intentions and actual childlessness would be particularly low in the Central and Eastern European countries (resulting in a small gap) was confirmed by our results. Indeed, the cohorts under study grew up under state socialism (until their late teenage years), when there were very strong norms against voluntary childlessness and policies supporting early and almost universal entry into parenthood. In this context, women would have one child rather than none (Frejka and Gietel-Basten 2016; Merz and Liefbroer 2012; Sobotka 2011).⁸ Very low childlessness levels continued long after the fall of communism (Beaujouan et al. 2016). Our results suggest that in the East, the strong negative attitudes towards childlessness were more important for defining intentions and actual fertility than the economic situation or family policies. In addition, we identified two distinct country clusters that featured the largest excess childlessness of around 20%: the Southern European countries (Italy, Greece, and Spain) and the German-speaking countries (particularly Germany and Austria). Excess childlessness was lower (above 10 percentage points) in the other Western countries, where it may have been mitigated by the well-established work–family policies.

The education-specific analyses refined some of our general results. In accordance with prior research, we observed small differences in fertility and childlessness intentions by education, without a clear-cut pattern (Testa 2014). However, our analysis confirmed that highly educated women achieve a lower mean number of children (except for in Belgium and Norway) and a higher level of childlessness (except for in the Czech Republic and Norway), which leads to a larger gap between intentions and final parity than for their lower educated counterparts. Overall, the size of the educational gradient of the gap in mean family size did not seem predicted by economic criteria or welfare regimes. For instance, we had expected moderate education-based gradients in Italy and Spain because childbearing of the low educated is curtailed by high economic insecurity, while childbearing of the highly educated is constrained especially by limited support for combining employment and family. This seemed to apply only to Italy, whereas in Spain, the fertility gap was much larger for highly educated women. In the UK, we observed almost no fertility gap among less educated women, but a marked gap among their highly educated peers, which was in line with previous findings in the UK and the US (Berrington and Pattaro 2014; Morgan and Rackin 2010; Quesnel-Vallée and Morgan 2003). Unintended fertility among the less educated women, linked to low contraceptive use and high rates of teenage pregnancies, could partly explain this result. We did not find that result for women in the US, but the differences with previous research in age group, birth cohort, and measurement may be behind this discordance.

Unlike for the mean number of children, the educational gradient for excess childlessness displayed a clear regional pattern: the gradient was the largest in countries where the combination of work and family is most difficult, namely the German-speaking countries and Southern Europe, as well as in The Netherlands and the UK where the majority of mothers work part-time. We had expected the smallest gradient in Norway given the well-established work–family policies and similar labor market outcomes between educational groups. This is confirmed by our data: the educational gradient in final childlessness is the lowest of all the countries studied together with Belgium. In contrast, in Hungary and the Czech Republic, childlessness levels were close to initial childlessness intentions, and particularly so among the medium educated.

What do our findings suggest in terms of policies? One of the clearest results is that highly educated women have the highest level of excess childlessness (18–26 percentage points) in the German-speaking (Austria, Germany, Switzerland) and the Southern European countries (Italy, Spain), where the obstacles for work–family reconciliation are highest. We could show that, despite highly educated women having moderately stronger preferences for work than their lower educated peers (Hakim 2002), they are not more likely to want to stay childless or to want fewer children than lower educated women. Still, fertility gap and excess childlessness are highest for them in almost all countries. The larger gaps among more highly educated women suggest that focusing policies on the needs of higher educated women to reconcile work and family demands are likely to have the biggest influence on birth rates. A series of measures have been suggested that aim to provide highly educated women with good conditions for having the number of children they want to have and a career simultaneously. This would include a well-developed childcare system when it comes to opening times and quality, a short to medium period of parental leave of up to 1 year with income-based payments available for all types of job, e.g., self-employed (Dearing 2016), and flexibility in terms of time and place to work. In addition, Esping-Andersen has argued that “a return to fertility levels that are more aligned with people’s preferences will require the consolidation of a new, ‘gender egalitarian’ family equilibrium” (Esping-Andersen 2017, p. 56). Highly educated couples generally display more gender egalitarian attitudes and are hence particularly inclined to respond to policies aimed at gender equality. Examples of such policies are reserving part of parental leave for fathers (Dearing 2016) and taking into account fathers’ care responsibilities in employment policies (e.g., flexibility, schedules that can be planned ahead, reducing overtime). In conclusion, if societies want to raise their birth rates, one pathway would be to enable highly educated women to have the children they intend to by fostering the combination of work–family for both mothers and fathers. An increase in fertility in gender egalitarian societies could thus be spearheaded by highly educated women (Esping-Andersen and Billari 2015).