The Childbearing of Immigrants Who Arrived as Children: Understanding the Role of Age at Arrival for Women and Men

Abstract

It is well-known that childbearing is associated with age at migration, but most research has focused on foreign-born women who migrated as adults. Much less is known about male immigrants or immigrants who arrived as children, despite the importance of studying these groups to understand theories of adaptation and socialization. This study addresses these gaps with a case study of Sweden, using longitudinal whole-population data to analyze the role of age at arrival in determining childbearing. The results suggest that age at arrival affects fertility across the childbearing life course, although there is little evidence of critical ages at arrival. These results hold for women and men, particularly for immigrants from higher fertility origins, with more ambiguous results for immigrants from lower fertility origins. The main findings also persist after examining sources of selection and reverse causality using sex-specific family fixed-effects models and separate analyses for specific countries of birth. Therefore, the study provides evidence of an underlying process of childhood socialization, followed by adaptation, that is common for women and men who migrate. Theoretical implications are discussed, including the need for further work on the determinants and mechanisms of adaptation.

Introduction

A fundamental challenge for demographers is understanding the childbearing of immigrants and their descendants (Drouhot and Nee 2019; Dubuc 2012; Kulu et al. 2019; Parrado and Morgan 2008; Zhou and Gonzales 2019). The challenge is fundamental at the macro level because linkages between migration and fertility play a key role in determining population dynamics (Jonsson and Rendall 2004; Sobotka 2008). It is also fundamental at the micro and meso levels, especially for individuals and their families, because of the relationship between childbearing and other social processes (Balbo et al. 2013; Goldscheider et al. 2015).

Research examining the childbearing of immigrants and their descendants has a rich history (Adserà and Ferrer 2015; Andersson 2021; Kulu et al. 2019; Milewski and Mussino 2019; Myers and Macisco 1975; Parrado and Morgan 2008). A central concept in this literature is adaptation, which can be defined broadly as a process of convergence between immigrants (or their descendants) and the native-born population (or a subgroup of that population) (Andersson 2021; Goldstein and Goldstein 1983; Harbison and Weishaar 1981; Hervitz 1985; Kulu et al. 2019; Mussino et al. 2021; Parrado and Morgan 2008). Fertility adaptation can be conceptualized as three stages of convergence: an initial difference in fertility between the populations of interest, a narrowing of this difference, and the disappearance of this difference. Prior research has commonly claimed that fertility adaptation is occurring in general or for specific groups of immigrants (Andersson 2004; Dubuc 2012; Kulu and González-Ferrer 2014; Milewski 2007, 2010a). However, recent research has highlighted several uncertainties regarding our knowledge about fertility adaptation (Kulu et al. 2019; Milewski and Adserà 2023; Milewski and Mussino 2019; Mussino et al. 2021; Tønnessen and Wilson 2023).

A key source of uncertainty is that evidence of fertility adaptation is hard to establish, partly because fertility has two components: birth timing (tempo) and the number of children born (quantum). It is now understood that it is beneficial to examine both tempo and quantum, as well as their interrelationship, not least because the adaptation of fertility quantum does not necessarily imply the adaptation of fertility tempo (Mussino et al. 2021; Tønnessen and Wilson 2023; Wilson 2020). Even in studies examining both quantum and tempo, disentangling adaptation from other explanations has been extremely difficult, especially in studies of immigrants who migrate during adulthood (Adserà and Ferrer 2015). Fertility patterns that look like adaptation (e.g., lower fertility rates for longer durations of residence) might be the result of anticipatory analysis (Hoem 2013; Hoem and Nedoluzhko 2016) or processes other than adaptation, including selection or anticipation (Adserà and Ferrer 2015; Milewski 2010a; Toulemon 2004). As prior studies have argued, one way to avoid many of these issues is to study immigrants who migrated before they reached childbearing ages (Adserà and Ferrer 2014; Adserà et al. 2012; Mussino et al. 2021).

In this study, I aim to address these issues, building on and responding to prior research by carrying out a quantitative case study of Sweden using longitudinal data for the whole population. Much research has been conducted on the adaptation of immigrant childbearing. However, far fewer studies have examined childhood immigrants, and very few studies have included male immigrants (for examples of exceptions, see Kraus 2019; Mussino and Cantalini 2022), let alone examined the intersection between sex and age at arrival for childhood immigrants from specific countries of birth. No doubt, these gaps in research are due to a lack of sufficient data in most contexts. Nevertheless, they represent gaps in our knowledge about immigrant fertility, which are particularly pertinent for debates about adaptation and socialization, especially given the potential role of population heterogeneity (Milewski and Adserà 2023; Milewski and Mussino 2019).

To address these gaps, I aim to generate new insights about immigrant childbearing adaptation by studying the relationship between age at arrival and fertility for immigrants who arrived in Sweden as children. This study extends previous empirical research through the unique combination of (1) dynamic analysis of fertility at different stages of the childbearing life course; (2) tests of childhood socialization by examining the role of age at arrival for immigrants who arrived as children; (3) an analysis that avoids common sources of selection bias, specifically through the use of family fixed-effects models; (4) a focus on differences in fertility adaptation in specific origin groups, using data on the whole population; and (5) a comparison of women and men, which is rare in the literature on fertility adaptation or the literature on adaptation more generally.

Childbearing and Adaptation

A considerable body of research across a range of high-income destinations has shown that childbearing outcomes—the quantum and tempo of fertility—typically differ between foreign-born and native-born women (Abbasi-Shavazi and McDonald 2002; Adserà and Ferrer 2016; Andersson 2004; Dubuc 2012; Kulu and Hannemann 2016; Mussino and Strozza 2012; Parrado 2011; Tønnessen 2020; Toulemon 2004). Research has also observed differences between the children of foreign-born parents and those of native-born parents (Kulu and González-Ferrer 2014; Kulu et al. 2017; Kulu et al. 2019; Parrado and Morgan 2008). Certain measures of immigrant fertility exaggerate evidence of differential fertility owing to bias, notably when using total fertility rates that exclude years at risk of childbearing before arrival (Parrado 2011; Sobotka and Lutz 2011; Toulemon 2006). Nevertheless, childbearing differences are evident for some immigrant groups when unbiased measures are used, including completed fertility (Parrado and Morgan 2008; Wilson 2019). Researchers have proposed various explanations for why childbearing for some immigrants and their children differs from the destination norm (e.g., as measured using the native-born average), many of which center on the role of migration and its impacts on life after arrival in a new destination (Milewski 2010a).

The hypothesis of immigrant fertility adaptation typically asserts that immigrants’ fertility will initially (i.e., on arrival) differ from the destination norm and will gradually become more similar to this norm with increasing duration of residence (Harbison and Weishaar 1981; Kahn 1988; Milewski 2007; Schoorl 1990). Naturally, the extent of this difference will vary according to the combination of origin and destination. However, irrespective of the specific context, adaptation appears to make a similar prediction to that made by straight-line assimilation (Alba and Nee 2005; Kivisto 2017). The adaptation process might be driven by changes in attitudes, preferences, or norms, through a process often referred to as acculturation (Adserà and Ferrer 2016; Milewski 2010a), or attributed to changes in culture or cultural norms (Kahn 1988; Rumbaut and Weeks 1986). Fertility adaptation might also be driven by changes in socioeconomic behavior and decision-making (Hill and Johnson 2004; Lindstrom and Giorguli Saucedo 2002), sometimes called economic adaptation (Rumbaut and Weeks 1986).

Most empirical studies of immigrant childbearing over the last few decades have attempted to test a hypothesis of fertility adaptation directly or have at least referred to fertility adaptation when framing the analysis (e.g., Adserà and Ferrer 2016; Andersson 2004; Dubuc 2012; Kulu and González-Ferrer 2014; Kulu et al. 2017; Kulu et al. 2019; Milewski 2010a; Mussino et al. 2021; Parrado 2011). Nevertheless, efforts to derive an overarching conceptual framework for studying immigrant fertility adaptation have been limited. Exceptions include the frameworks created by Rumbaut and Weeks (1986) and Forste and Tienda (1996), which helped clarify the determinants of immigrant fertility adaptation. In another relevant framework, Genereux (2007) identified gender as the central means of connecting three spheres of influence: (1) origin (the sending country), (2) migration (the migration context), and (3) destination (the receiving country). Each of these three spheres contains factors that impact immigrant fertility specifically, and factors that influence fertility more generally (Balbo et al. 2013). Because adaptation occurs after migration, fertility adaptation can be conceptualized as part of the destination sphere, which also encompasses adaptation in other domains of life, as well as broader processes of acculturation (Berry 2005). Yet, adaptation might also be influenced by—or act alongside—the processes linked to the other spheres (origin and migration), as explained in the rest of this section.

Various studies have tried to summarize the theories and hypotheses typically used to study migrant fertility, beyond the standard literature review that precedes most empirical studies (Kulu 2005; Kulu and González-Ferrer 2014; Kulu et al. 2019; Milewski 2010a; Milewski and Mussino 2019; Zárate and de Zárate 1975). One of the most frequently used explanations for immigrant fertility that falls within the first sphere (origin) is selection into migration, which predicts that immigrants differ from individuals who do not migrate in ways that influence their fertility (Goldberg 1959; Goldstein and Goldstein 1981; Harbison and Weishaar 1981; Singley and Landale 1998). Related to this issue is reverse causality, which is essentially a specific type of selection based on previous childbearing—as, for example, when those who have children are less likely to migrate (Harbison and Weishaar 1981; Toulemon 2006). In essence, selection and reverse causality can be conceptualized as occurring before the migration event, even if they might impact fertility afterward.

These explanations can be contrasted with those more closely related to the second sphere (the migration event itself), particularly migration timing. Researchers have often proposed that migration timing will lead to childbearing disruption, perhaps owing to anticipation, often coupled with elevated fertility in the first few years after arrival (Milewski 2010a). Elevated fertility might be due to the interrelation of events, such as the link between migration timing and partnership, which could affect fertility (Milewski 2007). However, elevated fertility might also be due to other explanations, such as selection or adaptation, making it hard to isolate or identify the impact of migration (Wilson 2020), especially given the complexity of causal interrelationships between childbearing and migration (Hoem 2013).

With respect to the final sphere—destination—the most commonly used explanations are adaptation, as already defined, and childhood socialization, which predicts that an immigrant's fertility will depend on their exposure to childbearing norms and behaviors during childhood (Goldberg 1959, 1960; Hervitz 1985). For example, childhood socialization might explain differences between the childbearing of adult immigrants from different countries of birth because their childhoods are most likely to have been spent in different countries (Milewski 2010b; Mussino et al. 2021). Indeed, childhood socialization might explain fertility adaptation patterns (Adserà et al. 2012), at least for some origin groups (Mussino et al. 2021), which may explain why some authors see adaptation as a corollary of socialization (Stephen and Bean 1992). Regardless of whether it is the result of socialization during childhood, immigrant fertility adaptation might also be determined by processes during adulthood, not only acculturation (or the maintenance of cultural norms; Abbasi-Shavazi and McDonald 2002) but also adaptation in other life domains, including adaptation to the structures and institutions at destination (Andersson 2021). Thus, childhood socialization can be conceptualized as one of several potential moderators of fertility adaptation (Adserà et al. 2012), not least because the two are increasingly seen to be intertwined (Adserà and Ferrer 2014; Mussino et al. 2021; Stephen and Bean 1992; Wilson 2020).

Therefore, adaptation is one of the most central concepts in the extant literature on immigrant fertility. Yet, it has proved elusive to test for several reasons. As noted previously, patterns of immigrant fertility that look like adaptation might result from determinants before arrival in the destination, such as selection, rather than changes in the tempo or the quantum of fertility after arrival. Many authors have made this point, particularly when cautioning against the (over)interpretation of elevated fertility after arrival using measures that do not incorporate fertility before arrival (Toulemon 2006) and when comparing fertility before and after migration without a counterfactual (Hoem 2013).

One problem with studying the adaptation of immigrant fertility empirically is that it is extremely difficult to draw reliable conclusions about adaptation when analyzing immigrants who migrated as adults because (1) they have spent some of their childbearing career in a different country than the destination and (2) migration and fertility are endogenous (Mussino et al. 2021; Singley and Landale 1998; Toulemon 2006; Wilson 2020). Although adult immigrants might arrive with a different tempo or quantum of fertility from the native-born population, such that they do arrive with an initial difference that could narrow and disappear (in line with the definition of adaptation given previously), adaptation might not be possible because childbearing is a monotonic process. Perhaps the most obvious example of this issue is immigrants who arrive having already given birth to more children than the native-born average completed fertility. These immigrants could never adapt to this native-born (quantum) average because they could not have fewer children. Given this and the other aforementioned problems of studying fertility adaptation for immigrants who migrate as adults, the following section considers how researchers might move beyond this impasse.

Age at Arrival and Fertility

Most of what is known about immigrant fertility currently rests upon studies of foreign-born women who migrated as adults. Far fewer studies have addressed male immigrant fertility, and very few have focused on women or men who migrated as children (sometimes called childhood migrants, generation 1.5, or G1.5), despite scholars’ increasing recognition of the value of studying childhood migrants, particularly when trying to understand immigrant fertility (Adserà and Ferrer 2014; Adserà et al. 2012; Mussino et al. 2021).

Immigrants who arrive as children offer a unique opportunity to examine the role of childhood in the process of adaptation. Unlike immigrants who arrive as adults, childbearing and migration are not endogenous for those who arrive as children (Adserà et al. 2012). Unlike native-born descendants of immigrants, they arrive at different ages, which is a unique source of variation in childhood socialization (Bleakley and Chin 2010; Hermansen 2017; Mussino et al. 2021). As noted previously, childhood socialization might explain why immigrants from different origins exhibit different fertility patterns in the same destination, even if immigrants are selectively different from their origin populations (Hervitz 1985), especially if they conform to the norms of their origin country or the neighborhood in the destination where they spend their childhood (Hill and Johnson 2004; Wilson and Kuha 2018). At the same time, childhood socialization does not exclusively involve norms. It also impacts opportunities and constraints, including with respect to other life domains that might influence fertility (Mussino et al. 2021).

Childhood socialization might be determined by many different factors but is typically operationalized using exposure to origin or exposure to destination (Kulu et al. 2019; Mussino et al. 2021). Irrespective of the study population, the relevant exposure period for childhood socialization is during childhood and adolescence (Goldberg 1959, 1960; Hervitz 1985). Therefore, in studies of immigrants who migrated as children, the context of childhood socialization and its duration in each context will vary for those who migrated at different childhood ages. Put simply with reference to this case study, a lack of exposure to Swedish society might cause immigrants to follow family formation patterns that are associated with their origin country, rather than those of Sweden.

Children who arrive at later ages will spend less time in their new destination before reaching childbearing age, giving them less time to socialize in their new environment and adjust their family formation plans and behaviors (Adserà and Tienda 2012). A lack of exposure also implies fewer opportunities for adaptation in any life domain, including those that may impact childbearing preferences, opportunities, and constraints. For example, later arrival implies less time for language acquisition, the development of social networks, and interaction with institutional structures. Relatedly, the effect of age at arrival on socialization will be impacted by meso-level factors, particularly families and schools, which might facilitate or impede exposure to the destination (Rumbaut 1994). Age at arrival is likely to determine reception context in various ways, which is important because reception contexts are a well-known determinant of immigrant family formation (Milewski and Adserà 2023) and, more generally, immigrant incorporation (Portes and Zhou 1993; Zhou 1997). For immigrants who arrive as children, education is not only likely to be a key mechanism because the nature and duration of schooling after arrival will vary by age at arrival but also because age at arrival is linked to the nature and duration of schooling in origin countries (before arrival).

Of course, exposure to the destination can be measured in different ways. However, as many scholars have argued, it is perhaps best summarized using age at migration, which determines the duration of exposure at a given age (Milewski 2010a). For childhood immigrants, age at migration can also be used to investigate the role of arrival at critical ages in childhood development (Adserà and Tienda 2012). For example, children who arrive after the onset of puberty might find it more difficult to socialize and adapt to life in their new destination, while arrival after becoming a teenager also lowers the likelihood of later partnership with someone who is native-born because of the difficulties of learning a new language (Bleakley and Chin 2010). Psychological research suggests that language acquisition becomes much more difficult after reaching a critical age (although there is some debate about this; see Birdsong 2006). Moreover, age at arrival might play a different role for women and men. Researchers do not appear to have compared the adaptation of immigrant fertility for women and men, but there are multiple reasons to expect sex differences. For example, boys who arrive as teenagers might find it more difficult to adapt to living in a new destination than girls arriving at the same age (Rumbaut and Portes 2001). Research suggests that immigrant women are more likely than native-born women to be in a “male breadwinning” and “female caregiving” relationship (Chuang and Tamis-LeMonda 2013; Pedraza 1991). However, differences between women and men will depend on the origin–destination combination, and early or high fertility might represent a deliberate and strategic choice for some groups of immigrant women (Hampshire et al. 2012). In any case, there are considerable grounds to propose differences between women and men, including differences in opportunities and constraints, and differences in the influence of parents, peers, and other social networks (Forste and Tienda 1996; Genereux 2007; Hampshire et al. 2012).

A Case Study of Sweden

Few studies have tested the influence of age at migration on the fertility of childhood migrants, and none appear to have tested whether the role of age at arrival differs for women and men. One Canadian study found that age at migration was associated with differences in birth rates between childhood migrants and the Canadian-born population (Adserà and Ferrer 2014). This result aligns with the general findings of a comparative study of France, Canada, and the United Kingdom (using the same data for Canada; Adserà et al. 2012). It also aligns somewhat with a more recent study of Sweden that focused on immigrants from origin countries with lower fertility and found considerable heterogeneity by origin (Mussino et al. 2021). However, none of these studies included male immigrants or attempted to model selection into migration, and the previous study of Sweden excluded immigrants from origin countries with higher fertility. Building on these and other prior studies, I carry out a quantitative case study of Sweden focused on four research questions:

1. Does age at arrival determine the childbearing of childhood immigrants?

2. Is this relationship the same for female and male childhood immigrants?

3. Is the relationship explained by common sources of selection?

4. How does the relationship vary by migration background—in particular, by refugee status on arrival, country of birth, or fertility level of the origin group?

I answer these questions with a comprehensive analysis of longitudinal register data for the entire Swedish population that links childhood conditions and migration background with information on childbearing from ages 15 to 45. The study takes advantage of Sweden's relatively large numbers of childhood immigrants, which provides enough statistical power to analyze individual ages at arrival, by sex, for different origin-country groups. The data also include parental identifiers that enable me to estimate family fixed-effects models. A similarly detailed analysis would not be possible in most other contexts, which means that Sweden is one of the few countries for which such a case study can be conducted.

Although this study focuses largely on the role of age at arrival for those who migrated as children (G1.5), it also compares G1.5 with immigrants who arrived in Sweden as adults (G1) and second-generation children of immigrants (G2). These comparisons focus on overlapping birth cohorts rather than subsequent cohorts, limiting the possibilities of drawing conclusions about intergenerational change. However, by comparing all three generations in addition to focusing on variation within G1.5, this study generates new knowledge about the possible trajectory of long-run adaptation beyond the first generation (Rumbaut and Portes 2001).

Data and Methods

Data are obtained via the collection of registers that are available for analysis by researchers at Stockholm University (under ethical approval from the Swedish authorities). These data cover the population of residents in Sweden from 1968 to 2017. Members of the population enter the register when they are born (if they are born in Sweden), when they receive a resident permit, or when they register their immigration. Registration is required to live in Sweden, and coverage of the population is close to 100% because it is very difficult to live in Sweden without registering.

Swedish population registers collect all demographic events and the dates of these events. Children can be linked to their parents using a register of personal identification numbers (as long as the parents live in Sweden or did so at some point in the past). I can therefore estimate the entire childbearing history of all women living in Sweden with a high degree of accuracy, including for immigrants who arrived in Sweden as children (who are highly unlikely to have had any children before arrival) and the second generation, defined here as those born in Sweden with two foreign-born parents. The data include all recorded immigrations, emigrations, and deaths, which enables me to calculate age at arrival for all immigrants and to exclude individuals who emigrated or died before I measure their fertility. Thus, my estimates exclude emigrants who returned, and I make no adjustment for over-coverage, although if it did have any impact then this issue would most likely impact the results for adult immigrants from Nordic origins (Mussino et al. 2024).

The data are longitudinal, allowing me to compare and contrast measures of fertility quantum—children ever born (CEB)—at any age. Quantum can be defined generally as the frequency of an event (e.g., number of births) and hence can be measured at any age (Ryder 1980). Initially (in Figures 1 and 2), the analysis focuses on completed fertility—quantum at the end of childbearing—measured here at age 45 for both women and men to facilitate comparisons. Relatively few births occur to women or men after age 45, albeit slightly more occur to men. For example, in the population analyzed in Figures 1 and 2, men's average CEB difference between ages 45 and 50 is 0.03, compared with a difference of <0.01 for women. As shown in the online appendix, this difference is almost the same for all the generational groups compared here, except for G1 (where the difference is 0.10 for men and <0.01 for women).

I then examine quantum at age 30 (in Figures 3 and 4), which indicates whether age at arrival plays a similar role midway (or approximately midway) through the childbearing life course (from ages 15 to 45) as it does toward the end (i.e., in the analysis in Figures 1 and 2). I then conduct a more detailed comparison of quantum and tempo using profiles of quantum differences from ages 15 to 40 (in Figures 5 and 6). These profiles compare the fertility of immigrants (by age) with that of Swedish-born women and men with two Swedish-born parents. Thus, birth timing (tempo) is not measured directly (e.g., using different birth parities). It can nevertheless be inferred by comparing different childbearing ages—in particular by examining the entire profile of fertility quantum (or quantum differences) from ages 15 to 45. To facilitate comparisons across fertility profiles, I calculate these quantum differences using the same study population at each age. Therefore, the population in each analysis includes only those individuals who reached the oldest age at which fertility is estimated (age 45 in Figures 1 and 2, age 30 in Figures 3 and 4, and age 40 in Figures 5 and 6) and who remained resident in Sweden until this age (excluding those who emigrated or died).

To the greatest extent possible, the case study uses data for the whole population of these cohorts. The study population in Figures 1 and 2 is composed of 3,498,649 individuals born during 1940–1971 who have been resident in Sweden from birth (for the Swedish-born) or first arrival (for the foreign-born) until age 45, which is the age at which CEB is measured. This study population excludes adopted children, Swedish-born children with one parent who is Swedish-born and one who is foreign-born, and those whose birth country is specified as “other.” It also excludes immigrants who arrived after age 35 to improve the estimate of childbearing for G1 in Figure 1, given that those arriving at older ages are more likely to have children who have never lived in Sweden. Some cases (5.7%) were dropped owing to missing data on key variables—notably, missing parental country of birth (3.9%), age at arrival (1.4%), and country of birth (0.4%). Further information about how the study population was obtained is provided in the online appendix. The composition of this population by sex, generational status, and country of birth is shown in Table 1. Because some countries are grouped in the data made available by Statistics Sweden, I use the most detailed classification available.¹ These origins reflect Sweden's migration history, including its receipt of large numbers of refugees since the 1970s (Statistics Sweden 2016). For some analyses, I use data on residence permits to analyze children of refugees separately from the children of other immigrants (although permit data are available only for later arrival years). For the second generation, ancestral origin is based on the mother's country of birth, although approximately 80% of G2 have parents from the same country of birth (or country group).

Overall, the final study population represents approximately 2.75 million Swedish-born individuals with two Swedish-born parents (1.41 million men and 1.35 million women), 311,000 G1 (153,000 men and 159,000 women), 99,000 G1.5 (48,000 men and 52,000 women), and 63,000 G2 (32,000 men and 31,000 women) for the cohorts born during 1940–1971 and analyzed up until age 45. For the analysis of these older cohorts up to age 30, the younger cohorts up to age 30, and those born during 1940–1976 who are analyzed at ages 15–40 (in Figures 5 and 6), the study populations of G1.5 and G2 are of a similar size, with some differences owing to censoring of those who died or emigrated.

All analyses are stratified by sex, and most analyses are based on the aggregation of individual-level data using arithmetic means. Otherwise, as in Figure 4, I model the number of CEB using generalized linear models (GLMs) with a Poisson link function. The models take one of two forms, which can be summarized as follows:

where $Yij$ denotes the number of CEB and $Xij$ denotes the explanatory variables for individual $j$ within family $i$⁠. On the left-hand side of the equations, $E(Yij|Xij)$ is the conditional mean of $Yij$ given covariates $Xij$⁠, and $E(Yij| familyi,Xij)$ is the conditional mean of $Yij$ given family fixed effect $i$ and covariates $Xij$⁠. By conditioning on the family, models based on Eq. (2) control for all confounders that are shared between siblings, such that the model estimates the within-family effect, which refers to a different population than the effect in Eq. (1).

Models without family fixed effects, based on Eq. (1), include covariates for birth cohort (in single years of age), mother's country of birth (grouped as shown in Table 1), and birth order (firstborn child, second-born child, or third-born or later child). Models based on Eq. (2) include the same variables except the mother's country of birth (which is the same for siblings). They also include family fixed effects (based on having the same mother) to control for many common sources of selection into migration: all factors shared between siblings (of the same sex, given that the models are stratified), including migration background (e.g., reasons for migration) and parental characteristics (e.g., parental education). Because all GLMs model fertility at a given age, the inclusion of birth cohort in the family fixed-effects models means they take account of differences between siblings in terms of period and cohort effects. None of the models include covariates measured after childhood, such as education or partnership, because they are endogenous with the fertility process (i.e., they might act as both causes and effects), which implies that their inclusion could bias the findings or at least make them hard to interpret. Family fixed-effects models include only those siblings who arrived in the same year (although <1.7% of siblings are dropped on this basis), who would likely have migrated for the same reason.

Childhood migrants are foreign-born individuals who arrived in Sweden (for the first time) aged 0–18. In the analysis by country, I classify them according to their own country of birth. The reference group in most analyses is those arriving at age 15, which is the age at which fertility researchers usually consider women and men to be at risk of childbearing. Nevertheless, most figures present the results for those arriving at older ages (16–18) to enable a fuller understanding of the role of age at arrival during childhood and adolescence (both of which can be defined up until age 18). I use G2 as a comparison group in the regression models. The inclusion of G2 in the family fixed-effects models is particularly important because it allows the identification of age at arrival and birth cohort, which would otherwise be colinear (for further explanation and a simulation study, see Wilson et al. 2021).

Results

Completed fertility (CEB at age 45) varies considerably among foreign-born women and men living in Sweden (Figure 1), with some origin country groups lower and some higher than the average for those who are Swedish-born with two Swedish-born parents. Among the population that I study, the mean completed fertility for those who are Swedish-born with two Swedish-born parents is 2.00 children for women and 1.80 for men (the dashed red reference lines in Figure 1). Figure 1 sorts every country of birth (or country of birth group) by G1 women's completed fertility. At the same time, it highlights (in darker tones) those countries with completed fertility more than 10% lower or higher than the average for Swedish-born women with two Swedish-born parents (see the online appendix for further details).² Countries that are 10% higher are considered to be higher fertility origins (in Figure 1 and in the analysis that follows), whereas those that are 10% lower are considered to be lower fertility origins. The advantage of this approach is that it classifies countries while taking account of selection among the G1. Even though most countries appear to be classified in line with the average fertility in their origin (e.g., when compared with Mussino et al. 2021), there are some exceptions. For example, the United States and Canada have similar levels of national female fertility as Sweden (Human Fertility Database 2021), but average completed fertility is clearly lower for female immigrants to Sweden from the United States and Canada. This difference might be due to differential fertility before or after migration, as well as the selective migration of certain types of women. Interestingly, however, a similar difference is not apparent for male immigrants from the United States and Canada.

Most origins exhibit similar patterns for women and men, but exceptions remain. Some origins, such as China, would not be classified as having higher or lower fertility if the classification were based on men. Similarly, some would be reclassified as having lower (Iran) or higher (India, Nepal, and Bhutan) fertility. The average completed fertility of immigrants who arrived as adults (G1) ranges from roughly 1.3 children for men who were born in Estonia to 4.6 children for women born in Somalia. For G1.5, the range is 1.1 for men from South Korea to 3.2 for women born in other Middle Eastern countries (not otherwise specified).

This article focuses on immigrant adaptation, but given the link between immigrant adaptation and intergenerational adaptation, it is interesting to note the differences between G1 and G1.5. For origins with higher and lower fertility—where G1 groups differ from the reference group of Swedish-born individuals with two Swedish-born parents—the completed fertility of both G1.5 women and men is more often closer to this reference group. The minority of exceptions to this general pattern include Pakistan and Bangladesh, North Africa (except Egypt), Iran, Bosnia-Herzegovina, South Korea (for men), and Iceland (for women).

Age at Arrival

Our first two research questions ask whether the fertility of childhood immigrants is determined by their age at arrival and whether any such relationship differs for women and men. Figure 2 provides evidence that age at arrival is associated with completed fertility to a similar extent for women and men. However, the association is much stronger for immigrants from higher fertility origins. Indeed, the results suggest no material association for immigrants from lower fertility origins.

Strikingly, the gradient in completed fertility differences is almost identical for women and men from higher fertility origins (see the lower panels of Figure 2). Compared with those who arrived as teenagers, immigrants who arrived at preschool ages have more than 0.5 fewer children (as do individuals of the second generation).

Quantum and Tempo

The patterns observed in Figure 2 for completed fertility might not be observed for other measures of childbearing, such as those that indicate variation in birth timing earlier in the life course. Moreover, the findings suggest that age at arrival is more important for higher fertility origins, perhaps partly because many immigrants from these origins are refugees. Therefore, Figure 3 examines whether the age-at-arrival gradient observed for CEB at age 45 (i.e., completed fertility) is also observed for CEB at age 30. It does so for individuals from higher and lower fertility origins (with both groups including children of refugees), as well as for the children of refugees at age 30. One advantage of analyzing CEB at age 30 is that it allows for examining younger cohorts (1972–1986) alongside the older cohorts (1940–1971) analyzed in Figures 1 and 2. For the children of refugees, I can analyze only the younger cohorts because data on residence permits are available only for those arriving in later years.

In general, the CEB patterns at age 30 are similar to CEB patterns at age 45 for women and men from the younger and older cohorts. Four additional findings are noteworthy. First, the results for children of refugees are very similar to those for higher fertility origins (comparing those born during 1972–1986). This finding suggests that refugee status might help explain the age-at-arrival gradient, which is also noteworthy because refugee origins are not exclusively those from higher fertility origins (e.g., Iran and Bosnia-Herzegovina). Second, birth cohorts differ in the age-at-arrival gradient. The gradient appears weaker for younger cohorts in higher fertility origins, but it is similarly modest (or almost nonexistent) in lower fertility origins regardless of the birth cohort. Third, the differences in the age-at-arrival gradient between women and men are not sizable. The gradient for men appears to be slightly shallower (implying a slightly weaker association) for higher fertility origins and children of refugees, whereas the opposite appears to be the case for lower fertility origins. Fourth, the differences in Figure 3 are considerable, even when compared with those observed for completed fertility (in Figure 2), implying that much of the impact of age at arrival on fertility occurs before age 30.

Examining the mechanisms that link age at arrival and childbearing is beyond the scope of this analysis, but my findings offer some clues as to likely mechanisms. For example, the possibility that age at arrival has a greater impact before age 30 might be linked to factors such as partnership and education, which have been shown to impact the timing of childbearing by fertility research in general (Balbo et al. 2013). For example, perhaps younger age at arrival has a positive impact on enrollment in education (e.g., as found in Norway by Hermansen (2017)), which might lead to a postponement of childbearing (as it does for populations in general; e.g., see Ní Bhrolcháin and Beaujouan 2012). Given the findings, this explanation might be more pertinent to the groups that exhibit a stronger link between age at arrival and fertility, such as women and men from higher fertility origins, particularly those in the older cohorts. These older cohorts might exhibit a stronger relationship between age at arrival and other relevant mechanisms, such as partnership behavior or the other proximate determinants of fertility. Another possibility is that the partnership market differs (or operates differently) for older cohorts, including with respect to partner choice or a partner's characteristics, such as age or socioeconomic background.

Controlling for Selection

A potential challenge to the findings discussed thus far is whether they are explained (confounded) by common sources of selection (i.e., factors that jointly determine age at arrival and fertility) rather than age at arrival itself. One way to examine this possibility and answer our third research question is to estimate the two sets of models shown in Figure 4: (a) one with controls for mother's country of birth, birth cohort, and birth order; and (b) one with controls for birth cohort, birth order, and family fixed effects. The inclusion of family fixed effects is particularly useful because it controls for factors shared between siblings, including migration background (e.g., the reason for migration, year of arrival/migration cohort, and parental country of birth) and other parental characteristics (e.g., parental education and social class). It enables the analysis to include all origin groups—and thus to be more generalizable—while also accounting for many forms of selection before arrival. Given the magnitude of the findings for CEB at age 30, the models are estimated at this age (separately by sex and birth cohort) and are therefore broadly comparable with the analysis in Figure 3. Confidence intervals are shown in Figure 4 but they are based on data for almost the whole population so should not be considered as indicative of uncertainty with respect to population inference.

The results in Figure 4 show that at least some impact of age at arrival persists after controlling for different forms of selection. As in previous figures, the reference group is those arriving at age 15. Compared with these immigrants who arrived as teenagers, those arriving at earlier ages, especially those in the younger cohorts, have lower relative risks of CEB at age 30. As is expected, the age-at-arrival gradient attenuates somewhat after the inclusion of family fixed effects (which also changes the estimate itself), and the confidence intervals increase in size. However, the role of age at arrival in determining fertility remains clear for women and men, even if only clearly evident for the younger cohorts (the top panels of Figure 4). This can be contrasted with the results in Figure 3, which instead suggest that the gradient is weaker for younger cohorts, at least for those from higher fertility origins. Since the main difference between Figures 3 and 4 is family fixed effects (which control for immigrant selectivity), this suggests that selection might influence the effect of age at arrival differently depending on cohort and origin, even if the effect of age at arrival on fertility persists when differences between families are controlled.

Specific Countries of Birth

My final research question regards the extent to which the relationship between age at arrival and fertility is generalizable to different countries of birth. The averages in Figures 2–4 might well mask variation between origins (such as those shown in Figure 1). Moreover, such variation by origin has proven hard to examine in prior research (in the absence of highly detailed data). Figures 5 and 6 therefore use the Swedish data to illustrate the role of age at arrival for the eight largest individual birth countries with higher fertility and the eight with lower fertility (compared with the Swedish-born children of Swedish-born parents, as defined in Figures 1–3). The figures examine childbearing profiles across most of the childbearing life course and show the average difference in number of children born at a given age between the reference group (Swedish-born individuals with two Swedish-born parents) and members of the G1.5 who arrived at one of three age ranges: 0–6, 7–12, and 13–16.

For all eight birth countries with higher fertility (Figure 5), age at arrival and fertility are interlinked across the life course, with differences starting to manifest very early in childbearing, sometimes even before age 20 (e.g., for Turkey). However, one of the most obvious findings is that the role of age at arrival is not consistent across origins or fertility profiles. This finding is confirmed by analyzing countries with lower fertility (Figure 6), many of which show no discernible evidence of childhood socialization, possibly even evidence against it. Immigrants with younger arrival ages will experience greater exposure to Sweden during childhood and less exposure to their origin country (including compared with their siblings who arrive at older ages), so if they exhibit smaller differences compared with the destination norm then this can be taken as evidence of childhood socialization. Although this finding appears common for those from higher fertility countries, it is not evidently the case for those from lower fertility countries.

The lack of a specific pattern is also apparent from the interaction between origins and profiles. For example, by their mid-20s, women from the former Yugoslavia who arrived at older ages (13–18) had given birth to roughly 0.5 more children, on average, than the Swedish-born with two Swedish-born parents, whereas this difference is much smaller by age 40. This contrasts with the patterns for women from Lebanon, which show a gradually increasing difference with age. In both cases, age at arrival influences fertility, but the impacts on quantum and tempo are very different. The results for men are similar to those for women, and for some origins (e.g., Turkey), they are very similar. Perhaps the clearest contrast between women and men in general is that differences for men do not begin to manifest until slightly older ages (e.g., former Yugoslavia) and that differences are more often smaller in magnitude for men (e.g., Lebanon).

Discussion

This article carried out a case study of Sweden that extends previous research by combining dynamic analyses of adaptation over the childbearing life course with tests of childhood socialization. In examining the role of age at arrival for immigrants who arrived as children, the analysis avoided common sources of selection bias, analyzed patterns for specific origin groups, and provided detailed comparisons of women and men.

By answering four research questions, this study generated the following main findings. First, in general, age at arrival determines the childbearing of many childhood immigrants, at least in the Swedish context. Second, the relationship between age at arrival and childbearing is similar for women and men, despite some differences, such as those found for younger cohorts from lower fertility origins. Third, this relationship is not entirely explained by selection, at least not for factors shared between families who migrated at the same time. Fourth, the link between age at arrival and fertility is neither ubiquitous nor homogeneous. For example, age at arrival is more positively associated with completed fertility for childhood immigrants from high-fertility origins. Yet, even for this group, the precise nature of the relationship is very heterogeneous, particularly when considering the interaction between origin, sex, cohort, and age.

In summary, the results of this study suggest that age at arrival influences childbearing for both women and men, particularly for immigrants from higher fertility origins and for the children of refugees. With respect to theoretical implications, these findings support an underlying process of childhood socialization, followed by adaptation, that is common for many women and men who migrate. Indeed, the impact of childhood socialization might have been underestimated in this study, at least when sex-specific family fixed-effects models were used, given that these models control for aspects of childhood socialization that are shared between siblings (of the same sex).

In general, a stronger impact of age at arrival tends to be observed at earlier childbearing ages, which highlights the importance of taking a life course perspective when analyzing adaptation. Indeed, this study highlights the importance of time—and different measures of time—in determining adaptation. Time is an integral component of the stages of adaptation and is a potential obstacle to the study of adaptation with respect to the timing of events, such as whether migration occurs before or after the end of childhood. Another important aspect of time, as evidenced by the results, is cohort. The role of age at arrival in determining fertility is more clearly evident for younger cohorts in family fixed-effects models (see Figure 4) but not in the earlier analysis, suggesting that the role of selection in determining the effect of age at arrival might vary by cohort. Finally, with respect to time, the study revealed little evidence in support of critical ages at arrival. Nevertheless, it suggests that future research might benefit from a more rigorous consideration of the role of time and dynamic processes in determining adaptation. Future research on fertility adaptation might benefit from a more thorough conceptualization of the role of temporal processes, not only for theory but also for empirical research design.

This case study has several limitations, including the fact that fertility is measured on the basis of children resident in Sweden. This is unlikely to impact the results for G1.5 and G2, but the estimates of fertility for G1 may be biased slightly downward, even though I excluded those G1 immigrants who arrived after age 35 to minimize this issue. Given that it is likely to be negligible, this bias is unlikely to impact the classification of origins greatly, although it may have some impact on the comparisons between G1 and G1.5 in Figure 1. Further, the study has a nontrivial amount of missing data, which could have some impact on the estimates. Yet, the level of missingness is very similar to that found in other register-based studies of immigrants and their children, and missingness is most material for parental country of birth, a type of missingness known to be far less common for immigrants arriving as children than for those arriving as adults (Wilson 2024). Thus, missingness is most likely to impact the estimates in Figure 1, the only figure that includes the G1.

Although the Swedish data are highly detailed, this study was limited in its ability to examine factors that measure selectivity and characteristics before migration. The family fixed effects control for many of these, but this modeling strategy does not enable such factors to be examined. In addition, the study did not explore the role of various factors measured at arrival, or after arrival, that might determine childbearing. These factors include those related to reception context, which I examined only indirectly via the comparison of birth cohorts and could be a promising avenue for future research. A further limitation is that the findings relate only to Sweden, which some might view as atypical or at least different from other contexts. Sweden is often perceived as providing more comprehensive support for newly arrived immigrants than most other countries (OECD 2016a, 2016b), which might counteract the impacts of age at arrival on adaptation. Thus, the impact might be different in other contexts, and in any case it could be beneficial for research to try to replicate this study in other contexts.

To conclude, this study found evidence that age at arrival can determine the childbearing of immigrants who arrive as children, just as it appears to determine many of their other outcomes, such as education, employment, occupation, language acquisition, partnership, mortality, and segregation (Åslund et al. 2015; Bleakley and Chin 2010; Böhlmark 2008; Hermansen 2017; Kilpi-Jakonen and Heath 2012; Mehta et al. 2019; van den Berg et al. 2014). Indeed, outcomes such as partnership or education might be the mechanisms that explain how age at arrival impacts childbearing—its tempo, its quantum, or both. However, more research is needed to clarify the interlinkages between different domains of adaptation. Such studies might generate deeper insights about the mechanisms of adaptation by, for example, examining the factors in different domains that mediate the effect of age at arrival on fertility. In addition, studies of adaptation might usefully examine multiple outcomes in a multiprocess framework. Indeed, these approaches are not mutually exclusive, but it is only via such developments that the roles of age at arrival and childbearing in determining the lives of immigrants and their descendants can be more fully understood.

Acknowledgments

The author would like to thank all those who helped him to develop this research, in particular Wendy Sigle, Alícia Adserà, Ana Ferrer, Jouni Kuha, Gunnar Andersson, Eleonora Mussino, Marianne Tønnessen, and Jan Saarela, as well as everyone else whom he has discussed this topic with over the last 15 years.

This research was supported by funding provided by the European Research Council Starting Grant 948727 (REFU-GEN), the Swedish Research Council (Vetenskapsrådet, via projects 2017-01021 and 2024-01481), and the Swedish Initiative for Research on Microdata in the Social and Medical Sciences (grant 340-2013-5164), as well as financial support received from the Swedish Research Council for Health, Working Life and Welfare (grants 2016-07115, 2016-07105, and 2018-00310). The author also acknowledges financial support from the Swedish Foundation for Humanities and Social Sciences (Riksbankens Jubileumsfond, RJ; grant registration M18-0214:1).

Notes

References