Abstract

Fertility rates among individuals in their 20s have fallen sharply across Europe over the past 50 years. The implications of delayed first births for fertility levels in modern family regimes remain little understood. Using microsimulation models of childbearing and partnership for the 1970–1979 birth cohorts in Italy, Great Britain, Sweden, and Norway, we implement fictive scenarios that reduce the risk of having a first child before age 30 and examine fertility recovery mechanisms for aggregate fertility indicators (the proportion of women with at least one, two, three, or four children; cohort completed fertility rate). Exposure to a first birth increases systematically in the ages following the simulated reduction in first-birth risks, leading to a structural recovery in childbearing that varies across countries according to their fertility and partnership regimes. Full recovery requires an increase in late first-birth risks, with greater increases in countries where late family formation is uncommon and average family sizes are larger: in scenarios where early fertility declines substantially (a linear decline from 50% at age 15 to 0% at age 30), first-birth risks above age 30 would have to increase by 54% in Great Britain, 40% in Norway and Sweden, and 20% in Italy to keep completed fertility constant.

Introduction

Cohort fertility levels have decreased unevenly across European countries. On the one hand, individuals experience a slower transition into adulthood and first births (Billari and Liefbroer 2010; Clark 2007), contributing to a substantial fertility decline at younger adult ages. On the other hand, few countries have witnessed sufficient fertility increases at later ages to compensate for declining fertility before age 30 (Beaujouan and Toulemon 2021). Some authors have argued that factors such as a high level of development (Myrskylä et al. 2009, 2011) or increased gender equality (Sobotka 2017b) favor fertility recovery. Others, however, have noted that social norms favoring small families, increasing time spent in education systems, and alternative life goals developed earlier in life lower the likelihood of full recovery (Buhr and Huinink 2017; Ciganda and Todd 2019; Gaddy 2021).

Thus, the extent of fertility recovery varies across contexts, and we suggest that studying the demographic mechanisms triggered by the decline in individual first-birth risks at younger adult ages (before age 30) will improve our understanding of this variation. Primarily, if fewer women are transitioning to motherhood at younger ages, those who remain childless will continue to be at risk of having a birth and may recover births at a later age. In addition, partnership patterns shape the potential to have children over the life course: these patterns are likely to be influenced by fertility delays and affect fertility recovery. Despite the wide variation in partnership trajectories across Europe, their importance for aggregate fertility levels and the recovery process remains little researched (Hellstrand et al. 2020; Sassler and Lichter 2020; Winkler-Dworak et al. 2017), with the exception of separation (Thomson et al. 2012; Van Bavel et al. 2012). One challenge is the limited availability of data covering the ongoing change in union types (Bennett 2017). Another challenge is the complexity of the interrelationship between fertility and partnership risks. In this context, the microsimulation approach is well-suited. Modeling a fictitious population based on the observed population risks of transitioning to childbearing and into and out of unions and marriages enables one to follow the evolution of family status over the life course (Aassve et al. 2006) and to calculate population-level estimates of fertility and union status (Billari 2006).

This research uses microsimulation to investigate the effect of a hypothetical decline in individual first-birth risks before age 30 for chosen macro-fertility outcomes (the share of women entering motherhood, the proportion of women with at least two children, and completed cohort fertility). The study is based on simulated fertility and partnership transitions for the 1970–1979 birth cohorts in Italy, Great Britain, and two Scandinavian countries (Sweden and Norway), highlighting cross-country differences. We first examine, in fictitious scenarios of reduced first-birth risks before age 30, the extent to which the structural fertility recovery resulting from the systematic increase in later exposure compensates for the forgone births at younger ages. We use decomposition techniques to explore the mechanisms triggered by the decline in early first births, including the increase in later exposure by partnership situation. In the second step, we use microsimulation to estimate how much the risks of first birth at later ages would have to increase in each country to keep the fertility indicators constant, still considering the underlying partnership dynamics.

Mechanisms of Fertility Delay and Recovery

As Sobotka (2017a:39) noted, the current trend of delayed parenthood “constitutes a unique experience in human history when most of the younger women and men spend their peak reproductive years sexually active, but consistently and effectively avoiding reproduction.” This trend has implications for individuals’ reproductive trajectories and aggregate fertility levels.

Rodriguez et al. (1984:5) described the reproductive process as an “engine with its own inbuilt momentum” in which early childbearing behavior and individual characteristics (e.g., education, employment, partnership, adherence to social norms) shape the remaining childbearing experience (Balbo et al. 2013). Our approach reflects this view by considering childbearing as a holistic process with linked successive events. Together, reproductive and partnership processes are the engine of fertility (Thomson et al. 2012), with ages at first birth and union suggested to play a substantial role (Rodriguez et al. 1984). Influencing this process are country specificities, such as norms regarding family formation and birth timing, progression to higher order births and spacing, and the acceptance of alternative union forms for childbearing (Billari and Liefbroer 2010; Sassler and Lichter 2020).

Several mechanisms operate when early first-birth rates decline. On the one hand, delaying first births at younger ages reduces the time to have children and achieve higher order births. On the other hand, individuals who do not have a first child at younger ages remain at risk of childbearing, with some of them catching up very soon and others continuing to delay fertility and thereby increasing the share of people at risk of experiencing a first birth at older ages. Thus, late first births increase because of increased exposure, even if fertility risks at later ages are assumed to be unchanged. The increase in late first births can also increase exposure to higher order births, potentially leading to more second-order or higher order births at older ages. We call that overall increase in late births structural recovery because it occurs systematically when early birth risks decline but late risks are constant, as a built-in mechanism of the fertility engine.

Partnership transitions are an important component of the fertility engine (Thomson et al. 2012; Winkler-Dworak et al. 2017, 2021): they depend on current pregnancy and previous births, and births depend on partnership status (Steele et al. 2005). This association is consistent across societies, although fertility risk differences by union type (e.g., cohabitation vs. marriage) vary across countries (Lappegård et al. 2018; Perelli-Harris and Sanchez Gassen 2012). Specific to the early fertility decline, the absence of a stable partnership or a separation in the 20s may further delay parenthood by reducing the time at higher risk of experiencing a birth. Subsequently, births are likely to occur in later first-order or higher order unions, contributing to the structural fertility recovery at higher reproductive ages (Andersson et al. 2022; Fostik et al. 2023; Thomson et al. 2012; Van Bavel et al. 2012).

The degree to which structural recovery alone can compensate for a fertility decline before age 30 depends partly on the magnitude of later birth risks relative to earlier risks. Age-related infertility and normative constraints limit late childbearing, and birth risks decline significantly beginning at age 35 (Leridon 2004; Nicoletti and Tanturri 2008; Wagner et al. 2019). However, in the 1970s birth cohorts, the risks of first and second births are highest around age 30 or in the early 30s, depending on the country (Beaujouan 2023:47; Human Fertility Database 2023). Higher birth risks around age 30 could drive a substantial structural recovery, with further progress depending on an increased risk of childbearing after age 30. Although births occurring at later ages have intensified since the onset of fertility postponement, the magnitude of the increase theoretically necessary for fertility catch-up has not been assessed.

To better assess the possibility of fertility recovery for birth cohorts that experience a substantial decline in childbearing before age 30, we ask two sets of questions. First, in the event of a hypothetical drop in first-birth probabilities at younger adult ages, how much does fertility decrease (when birth risks at older ages are held constant), what mechanisms are involved (i.e., structural recovery and link to partnerships), and how do these dynamics vary across countries with different characteristics and family regimes? Second, how much do first-birth risks need to increase beyond age 30 to achieve full recovery?

Country-Specific Dynamics

The countries included in our analysis exemplify different fertility patterns. Italy illustrates the southern European countries, with one of the lowest fertility rates and the highest average age at birth among the low-fertility countries (Beaujouan 2020; Beaujouan and Toulemon 2021). Great Britain differs from other European low-fertility countries because it has maintained relatively stable fertility rates despite a liberal welfare state model and relatively high labor market uncertainty (Vignoli et al. 2020). Norway and Sweden represent the Scandinavian countries, with family-friendly policies and welfare state models (Esping-Andersen 1990). These two countries are often considered leaders in the second demographic transition (Lesthaeghe 2020) and have long had fertility levels close to replacement level, with delayed entry into parenthood and strong fertility recovery at later ages (Sobotka 2017b).

In Italy, births below age 30 declined for female birth cohorts between the 1950s and 1970s, and increases in fertility did not match the decline, resulting in rapidly decreasing fertility (Beaujouan and Toulemon 2021; Zeman et al. 2018). Parity progression ratios to first births declined rapidly among the 1930–1966 birth cohorts, reducing the number of women at risk of experiencing a transition to second births (Frejka 2008). Parity progression ratios to second and third births also declined, with the latter drop being particularly steep (Frejka 2008). Despite the prevalence of traditional family norms in Italy, the diffusion of unmarried cohabitation has increased in recent decades (Castiglioni and Dalla Zuanna 2009); nonetheless, significant differences across geographic regions and educational levels persist (Gabrielli and Vignoli 2013; Kertzer et al. 2009). Although the share of nonmarital births has grown, marriage remains closely linked to childbearing (García Pereiro et al. 2014), with marriage and fertility being postponed in parallel (Kertzer et al. 2009). Therefore, in Italy marriage presumably remains an important engine of fertility recovery in a context of delayed marriage.

Fertility levels are higher in Great Britain than in other European countries, but behaviors are more heterogeneous. Among women born between 1959 and 1976, adolescent childbearing remained frequent and stable, with 7.5% of mothers entering motherhood as adolescents (Tomkinson 2019). Thus, before its more recent steep decrease (Heap et al. 2020), teenage childbearing contributed to higher fertility despite an overall decline in births before age 30 (Beaujouan and Toulemon 2021). On the other hand, birth risks are comparatively low among women who are childless at older ages, who are selected, for instance, among the highly educated (Berrington et al. 2015). Childlessness also appears to be associated with first-birth postponement, and one third of women intending to have a child remain childless at age 42 (Berrington 2017). For higher order births, the decline with age in the progression to a second birth persisted despite delays in first births across the 1959–1976 birth cohorts (Tomkinson 2019). Hence, a declining proportion of women experienced a transition to a second child. Finally, unmarried births, cohabitation, and divorce are common and socially accepted in Great Britain (Thomson et al. 2018). Given that union dissolution and repartnering are common, their effect on fertility levels appears less pronounced than in traditional settings, such as Italy (Winkler-Dworak et al. 2017), and second-order or higher order unions may fuel structural fertility recovery.

Birth rates in the Nordic countries have been relatively stable over time and have been higher than in other European countries (Hellstrand et al. 2020). However, recent publications noted that late fertility in the Nordic countries declined in the last decade (e.g., Hellstrand et al. 2020). Comolli et al. (2021) found that between 2008 and 2017, the relative risk of experiencing a first birth declined by 33% and 15% in Norway and Sweden, respectively. Thus, although the relative risk of second birth declined only marginally in both countries, the overall number of women at risk of experiencing a second birth decreased, driving a slight fertility decline (Comolli et al. 2021). In Norway and Sweden, cohabitation is a widely accepted alternative to marriage, and nonmarital childbearing is common (Lesthaeghe 2020). Thus, the exposure to cohabitation and the risk of experiencing a first birth during cohabitation is higher than in other countries in this analysis. Delayed marriage in Sweden and Norway is probably not closely linked to fertility delay, given that individuals often experience first births in cohabiting unions before potentially entering marriage later on.

Overall, the theoretical perspective of recovery for first and all births in the 1970–1979 birth cohorts appears heterogeneous across these three sets of countries in the event of first-birth risk declines in the 20s.1 In Italy, significant structural recovery of first births should occur because late fertility is already particularly prevalent, although the exceptionally late age at first birth might reduce the biological capacity to progress to a second child. Accordingly, the increase in first-birth risks after age 30 might have to be less pronounced than in the other countries to maintain equal transition levels to first births but still large to keep the cohort fertility level constant. In contrast, in Great Britain, to maintain fertility levels equal to those achieved in the absence of fertility delay, first-birth risks would have to increase significantly at higher ages. Indeed, childbearing is rather concentrated around ages 25–29, making structural fertility recovery at higher ages less likely. Yet, we expect partnership dynamics to be particularly relevant in helping to compensate for delayed fertility because of the large prevalence of separation. Finally, in Norway and Sweden, we expect a steep recovery of births beginning in the late 20s because birth rates appear comparatively large at higher ages. Nevertheless, as in the other countries featured, fertility recovery will certainly not be sufficient to offset the decline in the risk of early first births without an increase in the risk of later births, which occurs only if a larger proportion of women than in previous birth cohorts decide to have children at later ages and succeed in doing so.

Data and Methods

We base our analyses on a microsimulation of family life courses, fully described in Winkler-Dworak et al. (2021). This microsimulation model explicitly considers the complex interrelationship between individual childbearing and partnership dynamics and has been parameterized for Italy, Great Britain, Sweden, and Norway. The underlying survey data are the Family and Social Subjects for Italy, the General Household Survey and Understanding Society Survey for Great Britain, and the Generations and Gender Surveys for Norway and Sweden for women born in 1940 or later. From the original data, we estimated hazard regression models for each live birth transition (first to fourth) and for transitions into and out of union (up to third union) for women up to age 50. For each first-order and higher order union event, we estimated competing risks of cohabitation and direct marriage, separation of unmarried cohabitation and marriage, and divorce.2

The synthetic population of women is simulated by running the microsimulation models by 10-year birth cohorts up to age 50 for women born from 1940 onward (Winkler-Dworak et al. 2021). Here, we display simulated family life courses only for the 1970–1979 birth cohorts, the most recent cohort group for which we could still observe and validate transitions up to around age 40 in the different countries. Thus, the population and behavior are almost entirely empirically validated, and the trajectories reflect recent birth and partnership dynamics. This very large synthetic cohort provides the basis for experimentations on the interrelationship between micro-level changes and macro-level indicators, particularly linking family behavior over the life course and population-level outcomes.

For the scenarios, we held all transition risks constant except the age-specific baseline hazard rate of conception leading to a first live birth3 (piecewise splines coefficients retrieved from Winkler-Dworak et al. 2021: supplement, table A-1). To implement the exogenous change in rates, we added two multiplicative splines to the age-specific baseline of first-birth conception (Figure A1, online appendix). We assumed that there is an age gradient in the fertility delay until age 30, such that the younger the woman, the larger the decline in first-birth risk (e.g., see Rendall et al. 2005). We did not postulate a specific age pattern for fertility recovery and assumed uniformly higher first-birth risks after age 30. In that set of scenarios, the spline varies with a slope α before age 30 (resulting in the green-bluish lines for the risk) and is flat beyond age 30; that is, β is a proportional factor, and the implemented change in risk is uniform at all ages beyond 30 (resulting in the orange-reddish lines in the figure). Note that the values for the age-specific baseline hazard rate of a first-birth conception are those for women in the reference category of the partnership status variable (i.e., women in their first marriage with a union duration of less than a year). Varying the baseline risks will thus proportionally vary the first-birth risks contingent on the partnership situation.

The hazard rates constructed for the scenarios were subsequently converted into probabilities of first conception leading to a live birth and used in the microsimulation for the Monte Carlo experiments. Varying the baseline first-birth risks directly affects the simulated first-birth events and indirectly affects the exposure, as explained earlier. Moreover, because partnership transition risks depend on past and current births, a change in baseline first-birth risks might affect the partnership trajectories of women whose first birth is delayed because of the reduced first-birth risks but who otherwise would have had a first birth at an earlier age. Any change in early family life events thus has repercussions on later family events because of the complex interrelationship between fertility and partnership processes.

By varying α and β, we directly manipulate first-birth risk at the individual level. From the simulated family life courses, we calculate aggregate fertility measures depending on the variation imposed on individual fertility before and after age 30. These aggregate measures include the proportion of women having at least one, two, three, or four children and the sum of all these parity progression indicators as a proxy for the completed cohort fertility rate. We first estimate the change in aggregate fertility measures for a range of values of α when we artificially lower first-birth risks before age 30 (delay scenario). We then estimate how much first-birth risks after age 30 must increase (β) to compensate for the fertility loss due to a given increase in α for each of the aggregate outcomes (recovery scenario).

To disentangle the direct effect of varying first-birth risks on the aggregate indicators and its indirect effect linked to the change in exposure to the risk of conceiving a child, we extend the standardization and decomposition methods by Kitagawa (1955) and Das Gupta (1993) to these cohort parity progression indicators. In particular, we calculate the contributions of the rate component (direct effect) and the exposure component (indirect effect) to the difference between the proportion of women with at least one, two, three, or four children, PPR(0, k), where k = 1, . . . , 4, in the delayed fertility scenario and the original proportion in the baseline scenario. The parity progression indicator from parity 0 to k, PPR(0, k), which equals the proportion of women having at least k births or the number of births of order k, Bk, can be expressed as the weighted sum of kth-birth rates by age i, Mik:

PPR(0,k)=iBikN                 =iMik·Wi(k1)N,

where Wi(k1) denotes the number of person-years spent by women at age i and parity (k – 1), and N is the total number of women in the cohort. This formula allows us to decompose the parity progression indicator to parity k into a rate component at parity k, Mik, and an exposure component at parity (k – 1), Wi(k1)/N. The exposure at parity (k – 1) comprises all past birth experiences up to parity (k – 1), given that any change in births of lower birth orders will affect the age at which women reach parity (k – 1) and thus become exposed to the risk of having a birth of order k.

Using uppercase letters for the baseline cohort (base) and lowercase letters for the corresponding variables in the delaying cohort (delay), we decompose the difference in the parity progression ratios between two cohorts into a rate effect (R-effect) and an exposure effect (I-effect):

PPR(0,k)delayPPR(0,k)base=R-effect+I-effect        =[R(m¯k)R(M¯k)]+[I(a¯k)I(A¯k)],(1)

where

R(M¯k)=I-standardized indicator in baseline cohort           =iWi(k1)N+wi(k1)n2Mik,
I(A¯k)=R-standardized indicator in baseline cohort         =iMik+mik2Wi(k1)N,

and R(m¯k) and I(a¯k) accordingly for the delaying cohort. In Eq. (1), [R(mk)R(Mk)] measures the contribution of the birth rate differences to the difference in the parity progression indicator. In contrast, [I(a¯k)I(A¯k)] gives the contribution of the differences in the person-time at risk to a kth birth per woman to the differences in the indicator of progression to parity k (exposure effect). The exposure effect encapsulates distributional changes in exposure by the factor variables (i.e., change in exposure by age and partnership status in these analyses) and the overall change in exposure (i.e., change in average person-time at risk of having a birth at a given birth order). This conceptualization of the exposure effect contrasts with the composition effect in the decomposition of rates (Das Gupta 1993; Kitagawa 1955), which solely encompasses contributions due to distributional differences in factor variables.

Age and partnership status are highly correlated. In two-factor decomposition analyses, we disentangle rate and exposure effects by looking at differences in one factor while standardizing for the other factor. To further decompose the exposure effects into contributions of age and partnership status categories, we use a secondary decomposition, as proposed by Chevan and Sutherland (2009). The details of the two-factor standardization and decomposition that include both age and partnership are available in the online appendix.

Results

Scenarios of Delayed First Births

Table 1 presents the simulated parity progression indicators and completed cohort fertility rate according to diverse scenarios of delayed fertility (varying α). In these scenarios, first-birth risks are reduced by a coefficient of α at age 15 that linearly decreases to 0 at age 30. They remain unchanged for ages above 30 (see details in Figure A1, online appendix). As expected, the larger the α (i.e., the steeper the fertility decline at early ages), the greater the drop in aggregated fertility measures. Although we modified only first-birth risks, the progression ratios to second and (except for Italy) third births are more reduced than those to first births. Thus, the time to catch up during the reproductive life course seems mostly sufficient for delayed first births but insufficient for higher order births. The decline in aggregate fertility measures is twice as large in Great Britain as in the other countries (1.3% to 6.4% vs. 0.6% to 3.0% for the completed fertility rate), reflecting the particularly large contrast in Great Britain between the high first-birth risks at young ages and the low risks at older ages.

To deepen such findings, we explore the mechanisms underlying the fertility engine. We first inspect the simulated change in the parity progression indicators by age. Figure 1 shows the difference between the delay and the baseline scenario in the proportion of women with at least one, two, three, or four children by age, when α = .5.4 The lower first-birth risks at younger ages in the delayed fertility scenario initially decrease the proportion of women with at least one birth relative to the baseline scenario (Figure 1, blue line). Strikingly, the difference between the delay and baseline scenarios in the proportion of women with at least one child starts to narrow around age 26, even though first-birth risks remain lower until age 30. The narrowing of the difference arises because although first births are still delayed between ages 26 and 30 in our simulation, the “lost” births are outnumbered by the new births that occur because of increased exposure. The pattern is the same in all countries studied.

Given that first-birth risks are unchanged after age 30 in the delayed fertility scenario, the further narrowing of the difference in the proportion of women with at least one child is entirely due to the structural recovery of previously delayed first births. By the end of the simulated reproductive career, between 67% and 80% of the maximum difference in first births are eventually recovered, depending on the country.

Figure 1 also displays notable differences across countries. Italy shows the smallest simulated decrease in the proportion of women with at least one child relative to the baseline scenario. For Great Britain, which features particularly high fertility rates in the teenage years, the simulated proportion of mothers sharply drops in the late teens and early 20s in the delayed fertility scenario. Despite comparatively low first-birth rates in the late 20s and in the 30s, the “lost” first births at younger ages are overall recovered to the same extent as in Italy (67% of the maximum gap, compared with 70% for Italy and 80% for Norway and Sweden). In contrast to Great Britain, Norway and Sweden have first-birth rates that are low in the teenage years, steeply rise across ages, and eventually peak in the middle to late 20s. Hence, the simulated drop in the proportion of mothers in Norway and Sweden starts later, and delayed births are more extensively recovered in the late 20s and early 30s than in the other two countries.

We saw earlier that although only first-birth rates were changed in the simulation, births of all orders were affected (Table 1). Figure 1 also depicts the difference between the delay and the baseline scenario in the indicators of progression to higher parities. The pattern is similar for all parities: births initially decline, and the gap narrows at later ages, with the turning point (if any) occurring later with increasing parity. In fact, hardly any recovery is evident for third and fourth births. Finally, Figure 1 shows the difference for all parity levels combined, which approximates the difference in cohort fertility rate between the two scenarios across ages (thick gray line). It clearly illustrates the considerable country differences in the structural recovery of births: roughly half of the births missing in the late 20s are recovered by age 50 in Italy and the two Scandinavian countries, whereas approximately a third are recovered in Great Britain (scenario difference at age 50 of 0.11 children divided by the maximum difference at age 29 of 0.16 children). However, these figures provide only a lower estimate of the extent of structural recovery. As shown for first births, structural recovery begins when births are still delayed in the simulation.

The decomposition techniques presented in Eq. (1) allow us to quantify more precisely the extent of structural recovery. In particular, we decompose the difference in parity progression ratios between the delay and baseline scenarios into how much can be attributed to the reduction in first-birth rates at early ages (rate effect) and into how much was compensated by structural recovery due to increased exposure (exposure effect). Figure 2 shows the difference across ages between the proportions of women having at least one child under the delay scenario (only for α = .5) and under the baseline scenario (yellow line) and contrasts it with the same difference, but standardized for age-specific exposure (dashed gray line). Standardized here means that age-specific exposure is held constant at a standard level (which equals the average of the two scenarios’ age-specific exposure; cf. Eq. (1)) while rates vary according to the scenarios. The rate or R-effect in Eq. (1) corresponds to the dashed gray line at age 50. Standardizing for age-specific exposure, the cohort proportion of women with at least one child by age 50 falls by roughly 11 percentage points in Great Britain, Norway, and Sweden owing to the reduction in first-birth rates in the delay scenario (α = .5) relative to the baseline scenario. For Italy, the corresponding reduction is approximately 6 percentage points.

In contrast, the exposure or I-effect in Eq. (1) equals the difference between the nonstandardized and the standardized difference in the parity progression indicators (the difference between the gray and yellow curves at age 50 in Figure 2). It seems that the increase in exposure counteracts the reduction in the parity progression indicator at age 50 by roughly 4 percentage points in Italy, 8 percentage points in Great Britain, and 9 percentage points in Norway and Sweden, respectively. Thus, structural recovery has allowed a recovery of roughly 76% of all the first births lost to delayed fertility in Italy and Great Britain, and approximately 86% to 87% in Norway and Sweden. Accordingly, the larger decrease in the share of women with first births in Great Britain than in the Nordic countries (seen in Table 1) is due to a somewhat larger initial loss of first births and a smaller relative recovery. Italy experienced the smallest effect at both ends because of its generally lower first-birth risks.

The reductions in the parity progression indicators for parities 2 to 4 are due to the delay in previous births (age-specific exposure effect represented by the yellow line in Figures A3–A5 in the online appendix), with a small rate effect for parity 2. At all parities, Italy experienced the smallest effect of delaying first births, and Great Britain experienced the largest effect.

The decomposition also yields two small but important observations. The difference in standardized progression ratios to first birth narrows slightly after age 30, even though we assume that first-birth risks remain unchanged after age 30 in the delay scenario (Figure 2). Similarly, second births show a rate effect despite constant individual second-birth risks (Figure A3). This finding suggests that beyond reflecting the age-exposure effect, age-specific fertility rates at the macro level vary because of changes in the simulated population structure, which could be the effect of partnership status.

Figure 3 illustrates the decomposition of progression to parity 1 by age 50 for all countries into rate effects (gray bars) and exposure effects jointly by age (blue bars) and partnership status (green bars). The yellow bars on the left indicate the difference in the proportion of women with at least one child at age 50 between the delay scenario α = .5 and the baseline scenario for each country. The gray bars denote the same difference but standardized for the differences in the age- and partnership-specific exposure (R-effect in Eq. (2) in the online appendix). Note that in all countries, the rate effect decreases more when the proportions are standardized jointly for age- and partnership-specific exposure than for age alone (Figure 3 vs. Figure 2 at age 50; cf.Table A1 vs. Table A2, online appendix). It falls from −6 to −10 percentage points in Italy and from −11 to −14 percentage points in Great Britain, Norway, and Sweden. Accordingly, the exposure effect is 3–4 percentage points higher when we jointly standardize for age and partnership than when we standardize for age only.5 Hence, structural recovery depends on not only exposure by age but also the partnership status in which exposure increases.

To explore structural recovery further, we examine the contributions of age (blue bars) and partnership status (green bars) to the exposure effects (I- and J-effects), grouping the fine-graded age and rank-specific partnership categories into broader groups. Overall, these exposure effects have a positive impact on the proportion of women having at least one birth, largely counterbalancing the effect of the drop in the rate. More specifically, the partnership-specific exposure effect is mainly due to births in marriage in Italy but is outweighed by births in cohabitation in Norway and Sweden, reflecting the different partnership contexts of births across countries.

Figure 4 shows the decomposition of the proportion of women with at least two children by age 50 when we consider age and partnership status.6 When we jointly standardize for age- and partnership-specific differences in exposure, the small rate effect noted earlier and observed in Figure A3 disappears, as expected. Hence, given that we did not change individual second-birth risks, the change in the proportion of women with two children is entirely explained by changes in exposure (I- and J-effects) due to first-birth delays. Increased exposure from age 30 onward again increases the proportion of women with at least two children (dark blue bars), almost offsetting the drop before age 30 (light blue bars) in Great Britain and even surpassing it in Italy, Norway, and Sweden.

In addition, the cohort proportion progressing to a second child is negatively affected by exposure changes in all partnership states, where we differentiate by union type (cohabiting vs. married) and whether the first child is born within the current union or before (green shades in Figure 4). In Italy, the main contributor to the drop in second births is the change in exposure in marital first-birth unions. In contrast, cohabiting unions account for roughly the same share of exposure effects as married first-birth unions in Norway and Sweden. Furthermore, exposure changes in partnership states associated with union instability (not in a union; and married or cohabiting, with a first birth before the union) contribute markedly to the drop in the proportion of women having two children for the delay scenario in Sweden, Norway, and Great Britain. Likewise, changes in exposure of women with births in different union types markedly reduce the progression to third and fourth births, although such an effect is very limited for Italy (see Figure A6).

Scenarios of Delayed First Births and Fertility Recovery

We now assess the increase in first-birth baseline risks at later ages required to compensate for the earlier decrease. Figures 5 and 6 (and Figure A7 in the online appendix) show the share of women entering motherhood, the proportion of women with at least two children, and completed cohort fertility depending on the scenario of reduced first-birth risk before age 30 (α) and an increase after age 30 (β) imposed on the baseline risk of first birth. The green and purple shading of the surface indicates whether the aggregate fertility measure for the recovery scenario is, respectively, above or below the corresponding measure for the baseline scenario (indicated by a red dot on the vertical axis). The darker the shading, the greater the difference. In addition, the contour lines at the bottom of the graph represent α and β values with identical aggregate fertility measures. Again, when the first-birth risk falls at early ages (α > 0), the structural recovery at later ages (β = 0) is insufficient to maintain a constant proportion of women entering parenthood (measured along the axis labeled “Slope α” from left to right in Figure 5). To obtain the same fertility measures as in the baseline scenario, first-birth risks at later ages (β) would have to increase along the thick red contour line as α increases. Table 2 summarizes, for each country and fertility measure, the required level of β for the delayed fertility scenario specific to α = .5.

For instance, if we decrease risks of birth by α = .5 (i.e., by half at age 15 and gradually less and less until age 30), risks after age 30 would then need to be multiplied by 1.11 (1 + β) to recover all the missing first births in Italy (see Table 2 or the red contour line in Figure 5). Corresponding numbers for the other countries would be roughly 1.14 (Norway and Sweden) and 1.2 (Great Britain). Because the drop in the progression to first birth was largest in Great Britain and structural recovery in that country was not higher than in the other countries, the largest change in fertility behavior at later ages is required to compensate for the earlier decrease.

As expected, when first births are delayed (α > 0), fewer women transition to a second child (Figure 6). Further, as shown in Table 1, the reduction in second births might be even stronger than that for first births. Again, the largest reduction in the share of mothers having at least two children when first-birth risks are depressed at young ages is observable in Great Britain (see the darker purple shading in Figure 6). To fully recover second births as well, β values are required to move along the thick red contour line.

In the scenario α = .5, the increase in first-birth rates at later ages (i.e., 1 + β) shifts from 1.11 (necessary to recover first births) to 1.19 in the case of Italy (Table 2). A slightly higher incremental increase in the parameter β is required for Norway (from 1.14 to 1.27) and Sweden (from 1.14 to 1.25). In Great Britain, a much larger proportional increase in first-birth rates at later ages (from 1.2 to 1.46 for mothers with at least one child) is required to maintain the aggregate share of mothers with at least two children observed before the fertility decline.

Finally, a decrease in early first-birth risks more strongly affects completed fertility than the share of women entering motherhood because of the cumulative effect on the transition to first and subsequent births (Figure A7, online appendix). Accordingly, a larger increase in first-birth risks at age 30 and older is required to maintain the same average number of births per woman (thick red contour lines at the bottom of the figures). If α = .5 (cf.Table 2), the proportional factor 1 + β must be almost 1.2 to allow for complete recovery in Italy. The corresponding factors are 1.38 for the Scandinavian countries and roughly 1.54 for Great Britain, with the simulated drop in third and fourth births considerably larger in these countries (cf.Table 1). However, even if the aggregate cohort fertility level is maintained by a sufficient increase in first-birth rates at later ages, the resulting parity composition will be different. The required β values (noted in Table 2) lead to a simulated share of women with one or two children larger than in the baseline scenario (light green surface in Figures 5 and 6, respectively), but fewer women are then required to progress to third or higher births to maintain the cohort fertility level.

Discussion

We used a microsimulation approach to study scenarios of delayed first births and fertility recovery among the 1970–1979 birth cohorts in Italy, Great Britain, Sweden, and Norway. By simulating a decline in the risk of first births at earlier adult ages (before age 30) in our models, we investigated the mechanisms of the reproductive engine (Rodriguez et al. 1984). Specifically, we quantified structural recovery for completed cohort fertility, the share of women entering motherhood, and the share of women with at least two, three, or four children. To explore the mechanisms involved, we also assessed the impact of changes in age and partnership exposure on fertility indicators using decomposition techniques. Finally, we simulated scenarios of fertility recovery that increased first-birth risks at later ages in response to the decline at earlier ages and assessed the increase needed to keep macro indicators of fertility constant.

Our results offer two crucial pieces of information. First, we identified substantial structural recovery following a decline in first-birth risks at early adult ages, elicited by a rise in exposure linked to fewer births at earlier ages. The recovery capacity depended on a set of country specificities: the shape of the first-birth risk curve, family size (small vs. larger), and prevalence of births across partnership types. Second, full fertility recovery at later ages required increases in later (first) birth risks, which had to be more substantial in country settings not favoring structural recovery. For instance, in Great Britain, where baseline risks were relatively low at higher ages than in the other countries, achieving full fertility recovery would require substantial changes in childbearing and partnership formation behaviors that would increase the risks of late births. In contrast, in Italy, where economic uncertainty and labor market conditions have contributed to delayed entry into parenthood and very low fertility for a long time (Aassve et al. 2006), women are already accustomed to having fewer children, and later in life (Beaujouan 2020). Therefore, in the recent cohorts modeled, Italy loses fewer births to delayed fertility than the other countries.

When the risk of first birth is reduced before age 30, births show initial declines for all birth orders, but structural recovery occurs mainly for first and second births. The proportion of women having at least one birth is largely recovered, which becomes visible from around age 26 or 27. However, fewer higher birth orders are recovered, possibly because of the reduced time available during the reproductive period to recover them. Thus, even if only first-birth risks decrease, the consequences for completed fertility are magnified where families are initially larger.

Our approach allowed us to disentangle rate and exposure effects. Notably, in our scenario of a sharp first-birth decline at early ages (a linear decline from 50% at age 15 to 0% at age 30), the percentage of women with a first birth dropped overall by approximately 1.7% in Italy, Norway, and Sweden and by 3.3% in Great Britain. However, the decomposition analysis identified a rate effect that was about four times larger in Italy and Great Britain and seven times larger in Norway and Sweden. The difference between the rate effect and the decline in the simulated number of first births represents births that are inherently recovered because of the change in exposure. Thus, for first births, structural recovery plays a critical role. In a chain reaction, the change in exposure due to fewer first births decreases the proportion of women having two and then three and four children, although some recovery also occurs.

We also simulated that to maintain the same level of completed fertility rates as before the delay (again, with α = .5), first-birth risks after age 30 would have to increase by more than half in Great Britain, by around 40% in Norway and Sweden and by 20% in Italy. The increase is substantially higher than if we were willing to keep only first-birth levels stable.

Because we changed only first-birth risks in our recovery scenarios, maintaining the original completed cohort fertility level results in fewer childless women, more one- and two-child families in the simulated population, and thus a strong change in the parity distribution relative to the baseline scenario. In reality, however, a decline in childlessness is unlikely in these countries. Younger women in Italy are already more likely to be childless than women of their mothers’ generations (Fiori et al. 2017; Sobotka 2017a). And, if anything, delayed fertility is contributing to more childlessness in Great Britain (Berrington 2017). A recent decline in first-birth risks in Norway and Sweden raises questions about whether childlessness levels will remain low in those countries as well (Comolli et al. 2021; Hellstrand et al. 2020). Further recovery scenarios that postulate increases in second-birth risks at later ages could be explored to estimate how such changes would contribute to offsetting the decline in early first-birth rates.

Our findings highlight the close link between childbearing and partnership trajectories and their importance for fertility recovery scenarios. The partnership form most commonly associated with childbearing in each country is most closely linked to first-birth recovery: marriage in Italy and cohabitation or marriage in the other countries. Consistent with findings by Winkler-Dworak et al. (2017), we found that the highest contribution to fertility recovery in Norway and Sweden occurred in cohabitation, establishing parenthood in this partnership form as the norm. The limited flexibility of childbearing contexts in Italy could be seen as a limit to first-birth recovery, with any delay in marriage resulting in less time available to have a child. For higher order births, the exposure effect in unions was negative in all countries: delaying the first birth led to an overall decrease in the time at risk of having another child in all partnership states. Findings regarding the stability of different unions for fertility recovery are thus far inconclusive (Andersson et al. 2022; Fostik et al. 2023; Pelletier 2016; Thomson et al. 2019). In our study, we established that in Great Britain and the Nordic countries, separation largely contributed to the absence of recovery for higher order births. Thus, even in contexts where union dissolution and repartnering are common, their limiting effect on higher order births is pronounced in times of delayed fertility.

Our findings are contingent on the validity of the simulation model. Winkler-Dworak et al. (2021) thoroughly examined the model's replicative and predictive validity and concluded that the simulations replicate the family life courses in each cohort and country setting well. Further limitations might relate to the proportionality assumption postulated when modeling the scenarios: we modeled first-birth delay and recovery by adding splines multiplicatively to the age-specific baseline first-birth hazard rates, uniformly for all partnership states. However, a drop in early first-birth risks might be disproportionately lower in some partnership states than others. Similarly, other individual behaviors and characteristics (e.g., education, employment status, adherence to social norms) might disparately affect the occurrence of a fertility delay, and changes in exposure by these characteristics (along with age and partnership) might be relevant for fertility recovery. Another potential limitation is that we focused only on changes in first-birth risks when modeling fertility delay and recovery. However, it is not unlikely that individuals aspiring to have two or more children accelerate higher order births if the first birth is delayed. Such behavior, if adopted widely, would diminish the impact of delayed first births on the share of mothers with several children and on completed fertility.

Our analysis indicates that in contexts favorable to fertility recovery, structural recovery mechanisms are insufficient to recover the births lost to delayed fertility at younger ages. In addition to structural recovery, a large increase in the risk of having a first or second birth at a later age is generally required for the proportion of women entering motherhood and having at least two children to remain constant. Inflated birth risks have been observed between ages 30 and 35 in most postponing countries (Beaujouan 2023) and may be driven by an intensification of childbearing in the 30s or by changes in partnership behavior (e.g., increases in repartnering) at later ages. On the other hand, increases in childbearing might be moderated by limited or changing desires to have children (especially a second or third child) or by physiological difficulties in having a child at a later age (Beaujouan 2023). Overall, very few countries have experienced a full fertility recovery after a fertility delay, and our study provides new insights into why fertility recovery was limited.

Acknowledgments

The contributions of Eva Beaujouan and Maria Pohl were funded by the Austrian Science Fund (FWF), project Later Fertility in Europe (grant agreement P31171-G29).

Notes

1

Such cross-country differences are inherent in the position of each country at a different stage of the second demographic transition and the “postponement transition.” Thus, these countries differ vastly in their baseline risks of first-order and higher order births.

2

We estimated piecewise constant exponential models including age; birth cohort; the youngest child’s age; and detailed combinations of prior unions and births, differentiating births with previous partners from births with current partners. The hazard regression models also included several duration–cohort interactions with stepwise functions to represent linear splines. For full-model specifications and parameter estimates, see Winkler-Dworak et al. (2021). The hazard models do not incorporate variations in parental background, birthplace, education, or other experiences and characteristics that may influence life course choices, which may affect parameter estimation and henceforth simulation results. In Winkler-Dworak et al. (2021), however, we carefully validated the simulated family trajectories and found that they closely replicated the observed birth and union histories of the studied cohorts.

3

For brevity, we refer to the risk of a conception leading to a first live birth as the risk of a first birth.

4

See Figure A2 (online appendix) for the difference in the proportion of women with at least one child for various levels of α relative to the baseline scenario.

5

The amount of the change in exposure effect is the simple opposite of the change in rate effect, because the same difference between the parity progression ratios of the two scenarios is decomposed, as seen in Tables A1 and A2 (online appendix).

6

See Figure A6 (online appendix) for the joint decomposition of the proportion of women having three and four births, respectively, by age and partnership status.

References

Aassve, A., Burgess, S., Propper, C., &Dickson, M. (
2006
).
Employment, family union and childbearing decisions in Great Britain
.
Journal of the Royal Statistical Society, Series A: Statistics in Society
,
169
,
781
804
.
Andersson, L., Jalovaara, M., Uggla, C., & Saarela, J. (
2022
).
Less is more? Repartnering and completed cohort fertility in Finland
.
Demography
,
59
,
2321
2339
. https://doi.org/10.1215/00703370-10351787
Balbo, N., Billari, F. C., & Mills, M. (
2013
).
La fécondité dans les sociétés avancées: Un examen des recherches [Fertility in advanced societies: A review of research]
.
European Journal of Population / Revue Européenne de Démographie
,
29
,
1
38
.
Beaujouan, E. (
2020
).
Latest-late fertility? Decline and resurgence of late parenthood across the low-fertility countries
.
Population and Development Review
,
46
,
219
247
.
Beaujouan, E. (
2023
).
Delayed fertility as a driver of fertility decline?
In Schoen, R. (Ed.),
The Springer series on demographic methods and population analysis: Vol. 56. The demography of transforming families
(pp.
41
63
).
Cham
:
Springer Nature Switzerland
. https://doi.org/10.1007/978-3-031-29666-6_4
Beaujouan, É., & Toulemon, L. (
2021
).
European countries with delayed childbearing are not those with lower fertility
.
Genus
,
77
,
2
. https://doi.org/10.1186/s41118-020-00108-0
Bennett, N. G. (
2017
).
A reflection on the changing dynamics of union formation and dissolution
.
Demographic Research
,
36
,
371
390
. https://doi.org/10.4054/DemRes.2017.36.12
Berrington, A. (
2017
).
Childlessness in the UK
. In Kreyenfeld, M. & Konietzka, D. (Eds.),
Childlessness in Europe: Contexts, causes, and consequences
(pp.
57
76
).
Cham, Switzerland
:
Springer Nature
. https://doi.org/10.1007/978-3-319-44667-7_3
Berrington, A., Stone, J., & Beaujouan, E. (
2015
).
Educational differences in timing and quantum of childbearing in Britain: A study of cohorts born 1940−1969
.
Demographic Research
,
33
,
733
764
. https://doi.org/10.4054/DemRes.2015.33.26
Billari, F. C. (
2006
).
Bridging the gap between micro-demography and macro-demography
. In Caselli, G., Vallin, J., & Wunsch, G. (Eds.),
Demography: Analysis and synthesis
(Vol.
4
, pp.
695
707
).
Burlington, MA
:
Academic Press
.
Billari, F. C., & Liefbroer, A. C. (
2010
).
Towards a new pattern of transition to adulthood?
Advances in Life Course Research
,
15
,
59
75
.
Buhr, P., & Huinink, J. (
2017
).
Why childless men and women give up on having children
.
European Journal of Population
,
33
,
585
606
.
Castiglioni, M., & Dalla Zuanna, G. (
2009
).
Mariage et reproduction en Italie après 1995: Convergence avec l'Europe de l'Ouest? [Marital and reproductive behavior in Italy after 1995: Bridging the gap with western Europe?]
.
European Journal of Population / Revue Européenne de Démographie
,
25
,
1
26
.
Chevan, A., & Sutherland, M. (
2009
).
Revisiting Das Gupta: Refinement and extension of standardization and decomposition
.
Demography
,
46
,
429
449
.
Ciganda, D., & Todd, N. (
2019
).
The limits to fertility recuperation
(MPIDR Working Paper, No. WP-2019-024).
Rostock, Germany
:
Max Planck Institute for Demographic Research
. https://doi.org/10.4054/MPIDR-WP-2019-024
Clark, W. (
2007
).
Delayed transitions of young adults
.
Canadian Social Trends
,
84
,
13
21
. Retrieved from https://www150.statcan.gc.ca/n1/pub/11-008-x/2007004/pdf/10311-eng.pdf
Comolli, C. L., Neyer, G., Andersson, G., Dommermuth, L., Fallesen, P., Jalovaara, M., . . . Lappegård, T. (
2021
).
Beyond the economic gaze: Childbearing during and after recessions in the Nordic countries
.
European Journal of Population
,
37
,
473
520
.
Das Gupta, P. (
1993
).
Standardization and decomposition of rates: A user's manual
(Current Population Reports, No. P23-186).
Washington, DC
:
U.S. Department of Commerce, Economics and Statistics Administration, Bureau of the Census
. Retrieved from https://www.census.gov/library/publications/1993/demo/p23-186.html
Esping-Andersen, G. (
1990
).
The three worlds of welfare capitalism
.
Princeton, NJ
:
Princeton University Press
.
Fiori, F., Rinesi, F., & Graham, E. (
2017
).
Choosing to remain childless? A comparative study of fertility intentions among women and men in Italy and Britain
.
European Journal of Population
,
33
,
319
350
.
Fostik, A., Fernández Soto, M., Ruiz-Vallejo, F., & Ciganda, D. (
2023
).
Union instability and fertility: An international perspective
.
European Journal of Population
,
39
,
25
. https://doi.org/10.1007/s10680-023-09668-1
Frejka, T. (
2008
).
Overview chapter 2: Parity distribution and completed family size in Europe: Incipient decline of the two-child family model
.
Demographic Research
,
19
,
47
72
. https://doi.org/10.4054/DemRes.2008.19.4
Gabrielli, G., & Vignoli, D. (
2013
).
The breaking-down of marriage in Italy: Trends and trendsetters
.
Population Review
,
52
(
1
). https://doi.org/10.1353/prv.2013.0005
Gaddy, H. G. (
2021
).
A decade of TFR declines suggests no relationship between development and sub-replacement fertility rebounds
.
Demographic Research
,
44
,
125
142
. https://doi.org/10.4054/DemRes.2021.44.5
García Pereiro, T., Pace, R., & Grazia Didonna, M. (
2014
).
Entering first union: The choice between cohabitation and marriage among women in Italy and Spain
.
Journal of Population Research
,
31
,
51
70
.
Heap, K. L., Berrington, A., & Ingham, R. (
2020
).
Understanding the decline in under-18 conception rates throughout England's local authorities between 1998 and 2017
.
Health & Place
,
66
,
102467
. https://doi.org/10.1016/j.healthplace.2020.102467
Hellstrand, J., Nisén, J., Miranda, V., Fallesen, P., Dommermuth, L., & Myrskylä, M. (
2020
).
Not just later, but fewer: Novel trends in cohort fertility in the Nordic countries
(MPIDR Working Paper, No. WP-2020-007).
Rostock, Germany
:
Max Planck Institute for Demographic Research
. https://doi.org/10.4054/MPIDR-WP-2020-007
Human Fertility Database
. (
2023
).
Rostock, Germany
:
Max Planck Institute for Demographic Research; Vienna, Austria: Vienna Institute of Demography
. Available from www.humanfertility.org
Kertzer, D. I., White, M. J., Bernardi, L., & Gabrielli, G. (
2009
).
Le cheminement de l'Italie vers les très basses fécondités: Adéquation des théories économique et de seconde transition démographique [Italy's path to very low fertility: The adequacy of economic and second demographic transition theories]
.
European Journal of Population / Revue Européenne de Démographie
,
25
,
89
115
.
Kitagawa, E. M. (
1955
).
Components of a difference between two rates
.
Journal of the American Statistical Association
,
50
,
1168
1194
.
Lappegård, T., Klüsener, S., & Vignoli, D. (
2018
).
Why are marriage and family formation increasingly disconnected across Europe? A multilevel perspective on existing theories
.
Population, Space and Place
,
24
,
e2088
. https://doi.org/10.1002/psp.2088
Leridon, H. (
2004
).
Can assisted reproduction technology compensate for the natural decline in fertility with age? A model assessment
.
Human Reproduction
,
19
,
1548
1553
.
Lesthaeghe, R. (
2020
).
The second demographic transition, 1986–2020: Sub-replacement fertility and rising cohabitation—A global update
.
Genus
,
76
,
10
. https://doi.org/10.1186/s41118-020-00077-4
Myrskylä, M., Kohler, H.-P., & Billari, F. C. (
2009
). Advances in development reverse fertility declines.
Nature
,
460
,
741
743
.
Myrskylä, M., Kohler, H.-P., & Billari, F. C. (
2011
).
High development and fertility: Fertility at older reproductive ages and gender equality explain the positive link
(MPIDR Working Paper, No. WP-2011-017).
Rostock, Germany
:
Max Planck Institute for Demographic Research
. https://doi.org/10.4054/MPIDR-WP-2011-017
Nicoletti, C., & Tanturri, M. L. (
2008
).
Différences entre pays Européens dans le retard à la maternité: Analyse des données de l'ECHP [Differences in delaying motherhood across European countries: Empirical evidence from the ECHP]
.
European Journal of Population / Revue Européenne de Démographie
,
24
,
157
183
.
Pelletier, D. (
2016
).
The diffusion of cohabitation and children's risks of family dissolution in Canada
.
Demographic Research
,
35
,
1317
1342
. https://doi.org/10.4054/DemRes.2016.35.45
Perelli-Harris, B., & Sanchez Gassen, N. (
2012
).
How similar are cohabitation and marriage? Legal approaches to cohabitation across western Europe
.
Population and Development Review
,
38
,
435
467
.
Rendall, M., Couet, C., Lappegard, T., Robert-Bobée, I., Rønsen, M., & Smallwood, S. (
2005
).
First births by age and education in Britain, France and Norway
.
Population Trends
,
121
,
27
34
.
Rodriguez, G., Hobcraft, J., McDonald, J., Menken, J., & Trussell, J. (
1984
).
A comparative analysis of determinants of birth intervals
(WFS Comparative Studies, No. 30).
Voorburg
,
the Netherlands: International Statistical Institute
.
Sassler, S., & Lichter, D. T. (
2020
).
Cohabitation and marriage: Complexity and diversity in union-formation patterns
.
Journal of Marriage and Family
,
82
,
35
61
.
Sobotka, T. (
2017a
).
Childlessness in Europe: Reconstructing long-term trends among women born in 1900–1972
. In Kreyenfeld, M. & Konietzka, D. (Eds.),
Childlessness in Europe: Contexts, causes, and consequences
(pp. 17–53).
Cham, Switzerland
:
Springer Nature
. https://doi.org/10.1007/978-3-319-44667-7_2
Sobotka, T. (
2017b
).
Post-transitional fertility: The role of childbearing postponement in fueling the shift to low and unstable fertility levels
.
Journal of Biosocial Science
,
49
(
Suppl. 1
),
S20
S45
.
Steele, F., Kallis, C., Goldstein, H., & Joshi, H. (
2005
).
The relationship between childbearing and transitions from marriage and cohabitation in Britain
.
Demography
,
42
,
647
673
.
Thomson, E., Winkler-Dworak, M., & Beaujouan, E. (
2018
).
Cohabitation and parental separation: Cohort change in Italy, Great Britain, and Scandinavia
(Stockholm Research Reports in Demography, No. 2018:23).
Stockholm, Sweden
:
Stockholm University, Demography Unit
.
Thomson, E., Winkler-Dworak, M., & Beaujouan, É. (
2019
).
Contribution of the rise in cohabiting parenthood to family instability: Cohort change in Italy, Great Britain, and Scandinavia
.
Demography
,
56
,
2063
2082
.
Thomson, E., Winkler-Dworak, M., Spielauer, M., & Prskawetz, A. (
2012
).
Union instability as an engine of fertility? A microsimulation model for France
.
Demography
,
49
,
175
195
.
Tomkinson, J. (
2019
).
Age at first birth and subsequent fertility: The case of adolescent mothers in France and England and Wales
.
Demographic Research
,
40
,
761
798
. https://doi.org/10.4054/DemRes.2019.40.27
Van Bavel, J., Jansen, M., & Wijckmans, B. (
2012
).
Has divorce become a pro-natal force in Europe at the turn of the 21st century?
Population Research and Policy Review
,
31
,
751
775
.
Vignoli, D., Bazzani, G., Guetto, R., Minello, A., & Pirani, E. (
2020
).
Uncertainty and narratives of the future: A theoretical framework for contemporary fertility
. In Schoen, R. (Ed.),
The Springer series on demographic methods and population analysis: Vol. 51. Analyzing contemporary fertility
(pp.
25
47
).
Cham
:
Springer Nature Switzerland
. https://doi.org/10.1007/978-3-030-48519-1_3
Wagner, M., Huinink, J., & Liefbroer, A. C. (
2019
).
Running out of time? Understanding the consequences of the biological clock for the dynamics of fertility intentions and union formation
.
Demographic Research
,
40
,
1
26
. https://doi.org/10.4054/DemRes.2019.40.1
Winkler-Dworak, M., Beaujouan, E., Di Giulio, P., & Spielauer, M. (
2017
).
Union instability and fertility: A microsimulation model for Italy and Great Britain
(Vienna Institute of Demography Working Papers, No. 08/2017).
Vienna
:
Austrian Academy of Sciences, Vienna Institute of Demography
. https://doi.org/10.1553/0x003ccffe
Winkler-Dworak, M., Beaujouan, E., Di Giulio, P., & Spielauer, M. (
2021
).
Simulating family life courses: An application for Italy, Great Britain, Norway, and Sweden
.
Demographic Research
,
44
,
1
48
. https://doi.org/10.4054/DemRes.2021.44.1
Zeman, K., Beaujouan, É., Brzozowska, Z., & Sobotka, T. (
2018
).
Cohort fertility decline in low fertility countries: Decomposition using parity progression ratios
.
Demographic Research
,
38
,
651
690
. https://doi.org/10.4054/DemRes.2018.38.25

Supplementary data