The fertility of male immigrants: a comparative study on six Western European countries

In the last decades, immigration has grown considerably in Europe, also in those countries, like Italy, Spain and Portugal, which in the past were considered ‘classic’ sending countries. This increase in immigration flows has profoundly affected the socio-demographic structure of European society (Sobotka 2008). Immigration, moreover, limited the downward trend of fertility observed across Europe in the last decades and contributed to a slight increase in period fertility rates since the mid-1990s (Héran and Pison 2007; Garssen and Nicolaas 2008).

An increasing number of empirical works started to study migrant families, but lack of quantitative data on migrant populations has limited the opportunities of analysing the fertility of immigrants in comparative perspective, focusing on its interactions with the macro-features of the host countries. As outlined by Kulu and Gonzàlez-Ferrer (2014: 428), ‘we lack truly comparative research on migrant and ethnic minority families in Europe, which would consider various institutional and policy contexts’. From this standpoint, this paper describes the fertility of male immigrants in six European countries: France, Germany, Italy, Spain, the Netherlands and United Kingdom. Moreover, empirical analyses incorporate the information on the time of residence in the host country, since various factors behind fertility behaviours may appear at different stages of the migration process (Singley and Landale 1998).

The paper also contributes to the literature with its empirical strategy, which analyses the fertility of migrants and natives by focusing on the male population. The literature on fertility has mainly studied parenthood from a female perspective, considering women as mostly responsible for childcare and the driving force behind fertility decisions. With few exceptions (e.g. Lundström and Andersson 2012; Wolf 2014), even empirical research on the reproductive behaviours of migrants focused on the female population (Roig Vila and Castro Martìn 2007; Milewski 2010), although foreign women often migrate for family reunion once the male partner has already faced and paid off the most difficult migration costs (Ballarino and Panichella 2018).

This paper considers primarily men, but it also includes in the model information about the partner and looks at the interaction between the employment conditions of both. On the one hand, this approach yields a different picture of fertility, highlighting the importance of the economic condition of men (Oppenheimer et al. 1997) as well as the reciprocal influence of both partners in the couple on reproductive choices (Winkler Dworak and Toulemon 2007). On the other, it allows more detailed study of the importance of disruption and, more specifically, investigation of how the fertility of migrants is affected by the fact that migration strategies often imply marital separation.

The paper is structured as follows. After this introduction, Section 2 discusses the mechanisms behind the fertility of immigrants, while Section 3 outlines the main research aims and expectations. Section 4 describes the data, variables, and methods used, as well as the analytical strategy. Section 5 sets out the empirical results, and Section 6 concludes.

According to the literature on immigrants’ fertility dynamics, there are four mechanisms/arguments linking migration and reproductive behaviours: socialization, adaptation, selection, and disruption (Goldstein and Goldstein 1981). The socialization mechanism stresses the crucial role of social norms and cultural values dominant in the childhood environment in driving fertility decisions once migrants have moved into the host society. Therefore, migrants are likely to maintain the reproductive preferences transmitted in the country of origin. On this view, immigration has contributed to the rise in the total fertility rate of many European countries because several groups of immigrants have been socialized in a context characterized by early marriages and high fertility (Rosero-Bixby et al. 2011), and their fertility remains higher also in the long-run (Andersson 2004; Coleman and Dubuc 2010; Milewski 2010).

The second argument is adaptation (or assimilation), which refers to the tendency of migrants to share the predominant reproductive behaviours of natives according to the duration of exposure to the host society (Kahn 1988). Indeed, their fertility levels may be primarily affected by the social and cultural norms of the society of destination (Caldwell 1982), as well as by its structural and macro context. This argument is related to the ‘assimilation theory’ developed by the Chicago School in the 1920s (Park and Burgess 1924), claiming that in the long run migrants gradually become similar to natives. If we focus on migrants from high-fertility regions, their adaptation to the local social context often involves a reduction of fertility (Schmid and Kohls 2009).

According to the selection argument, the group of people who voluntarily emigrate is not a random sample drawn from the population; rather, it is selected according to observed characteristics, such as education, and unobserved factors, such as social mobility ambition or family orientation (Chiswick 1999). Hence, the fertility levels of migrants may not differ from those of natives because the former are a selected group of people with preferences towards family similar to those of people living in the society of destination (Chattopadhyay et al. 2006). This mechanism can be read in light of the ‘anticipatory socialization’ theory (Merton 1949), according to which the non-group members (immigrants) are likely to take on values and orientations of the group that they want to join (natives), easing their future integration into the group and the competent interaction.

Finally, the disruption argument focuses on the costs and the difficulties that migrants encounter after the geographical movement, such as the search for a house and a job, the entrapment in low-skilled and low-paying jobs, etc. The effects of these disruptive factors are particularly strong in the first years after migration (Jensen and Ahlburg 2004), while they tend to decrease over time, once migrants permanently settle in the new country and the disruptive costs are paid off (Stephen and Bean 1992), allowing them to resume their previous fertility levels (Sharma 1992).

One of the most important disruptive factor for migrants is the marital separation. In fact, families often reduce the risk of geographical mobility with a two-step migration: men migrate first to reduce the costs of migration, while women migrate later for family reunion (Ballarino and Panichella 2018).¹ Since, therefore, most migrations are gendered, staying temporarily alone in the host country prevents male migrants from having a child in the short run and forces them to postpone reproductive decisions until after family reunion. Moreover, several migrants come to work and live in the host country without their families and kids, who continue living in their country of origin. Although in this case there is not an actual reduction of the fertility of migrants, the disruptive effects of geographical mobility still limit their contribution to the population of the receiving country.

The effect of marital separation has been only indirectly confirmed by some works on women’s fertility, showing that female migrations driven by marriage or childbearing lead to high fertility immediately after the move (Andersson 2004; Kulu 2005; Mussino and Strozza 2012). However, since immigrant women often migrate for family reunion, when the costs of mobility are paid off, empirical analysis based only on women (and only on couples) tends to underestimate the effect of disruption.

Although many empirical works studied the effects of these four mechanisms, and how they evolve during the migration process (Singley and Landale 1998; Andersson 2004; Guetto and Panichella 2013), comparative research on this topic is still scarce. In fact, it is not clear yet if and how the consequences of socialization, adaptation, selection and disruption vary across countries with different institutional features, or if their role is ‘independent’ from macro structural factors (Kulu and Gonzàlez-Ferrer 2014). Focusing only on female migrants and second generations, the few comparative studies showed that migrants’ fertility behaviours tend to resemble cross-nationally those of the native population, despite some differences according to the country of origin (Milewski 2011; Kulu et al. 2015; Wilson 2018).

The next sections analyse this issue more in depth, describing the aims of this work and the cross-country comparison strategy.

The lack of longitudinal data containing detailed information on immigrant population – in both the ante and post migration periods – makes it hard to measure the causal effect of the mechanisms described above. The data analysed in this study are not an exception from this point of view, since they do not contain time-varying information on migrants covering the periods before and after the movement (see section 4 for details). Hence, this work does not aim to measure the causal effect of the abovementioned mechanisms, but rather to describe how the gap between migrants and natives evolves over time of residence in different host countries, and how such trend is consistent with the mechanisms linking migration and fertility.

The comparative design considers six Western European countries: France (FR), Germany (DE), Italy (IT), Spain (ES), The Netherlands (NL) and United Kingdom (UK).² These countries differ according to at least two dimensions. The first is the fertility level, which is important because it affects the magnitude of the difference between migrants and natives. France, the UK and the Netherlands are among the highest fertility countries in Europe, where the fertility levels range between 1.7 and 2.0 children per woman, while Italy, Spain and Germany are among the lowest fertility countries, where the TFR declined to 1.3-1.5 children per woman in recent decades. Hence, we can expect that the (possible) effect of socialization emerges more clearly in Italy, Spain and Germany, where the fertility rate among the native population is lower.

The second dimension regards the labour-market regulation and its linkages with the welfare regime. In UK the labour market is regulated very flexibly, hence migrants have easier access to it (Kogan 2007; Ballarino and Panichella 2015; Panichella 2018). The same reasoning may apply to The Netherlands, where flexibility has recently become a guiding principle of labour-market regulation, supported by a very generous welfare regime in which resident migrants have largely the same formal social rights as any native person (Esping-Andersen and Regini 2000). All these factors might reduce the disruptive effects of geographical mobility.

By contrast, disruptive factors should be stronger in the ‘coordinated market economies’ of Germany and France, where migrants encounter greater difficulties in finding a job, and social insurance is dependent on their labour-market participation. This pattern may be even more evident in Italy and Spain, where migrants are often included in the black labour market, which is informally regulated in a very flexible manner, leading to unskilled, poorly paid, and dangerous employment that does not give the right to access welfare services (Ballarino and Panichella 2015). Moreover, in the Mediterranean countries, immigration is a relatively recent phenomenon; hence, the disruption effect may have a stronger negative impact on fertility.

The empirical strategy consists in two steps. The first studies the average difference of the fertility levels between migrants and natives among the six countries considered, namely analysing the pooled sample. Then the analysis considers the cross-country variation, focusing on whether the mechanisms linking migration and fertility have different (or even opposite) effects in different host countries.

Moreover, the empirical analyses do not distinguish between children of migrants born in the host country and in the region of origin, rather we decided to describe the total contribution of migrant families to the European countries. There are two reasons behind this decision. First, we are interested in the contribution of migrant’s population to the European societies, which of course includes also those children not born in the host country (i.e. generation 1.5). Second, from a technical point of view, in absence of longitudinal data containing time-varying information on fertility and migration behaviours it is difficult to compare migrants and natives in a proper way. For instance, if migrants and natives are compared considering, among the former, only those children born in the host country, results systematically show a lower fertility level among migrants. Indeed, while for natives we would consider the whole life-course, for migrants we would consider only those children born in a specific stage of the life-course, namely the period after migration. One possible solution for avoiding such bias is to focus on a relatively young (and selected) cohort, for instance those aged 25-29, where the difference between the total number of children and the total number of children born in the host country is limited (mean: 0.12). Results of this analysis (available upon request) are consistent (but with higher uncertainty) with those shown in the next sections, which consider the total contribution of migrant population to the European countries.

As stressed above, the literature on fertility behaviours of migrants in comparative perspective is still in its first stages because of lack of data measuring both geographical origin and fertility in comparative terms. Actually, the European Social Survey (ESS) contains direct information on the number of children living in the household, but its limited sample size prevents robust estimation of the variation within countries. To cope with this limitation, data from the European Labour Force Survey (EU-LFS) are used (2005–2015). Although fertility is not the main aim of this survey, its large sample size makes it possible to study the fertility of immigrants in both the short and the long run and compare different European countries. Unfortunately, illegal immigrants are not covered by this survey, since they are not part of the resident population. This unavoidable bias may affect our estimation and it should be checked by proper sampling of illegal migrants. Another limitation of EU-LFS data is its cross-sectional structure, which does not allow to control the selectivity of migrants over time of residence in the host society.

The EU-LFS data, moreover, does not include individual-level information on fertility behaviours. Consequently, the dependent variables in our analysis were developed using the ‘own-child technique’ (Bordone et al. 2009). This procedure links children to their (supposed) mothers (or fathers) in the same household, assuming that minor children recorded in a household comprise all the children born, and still alive, to the parents in that household, even if the relationship is not directly specified. Lack of data obliges use of this technique, which of course enables only detection of those children still living, at the time of the interview, with at least one parent. We addressed this problem by including in our analysis only those individuals aged between 30 and 50 years old. By focusing on a relatively young sample of men, we could assume that there were no children living outside the household, and we were able to reconstruct the actual number of children indirectly.³ The own child technique, however, makes it difficult to disentangle the role of timing of parenthood, which is an important issue to study fertility behaviours (Cantalini 2017a), especially because it may differ according to migratory status and country of origin.

As previously stated, we focused only on men for two reasons: (a) migration is a gendered process, where women often migrate for family reunion (Ballarino and Panichella 2018); (b) male fertility is often neglected in the socio-demographic literature (Cantalini 2017b). This choice also had a limitation. By considering only men and their children within the household, we might underestimate the total number of children of both natives – given the higher likelihood of a child living with his/her mother in cases of separation or divorce – and (short-term) migrants – whose children may stay in the country of origin and join fathers later with family reunion. For this reason, the empirical analysis was replicated considering women (results available on request). After listwise deletion of missing values, the analytical sample included 2,007,547 cases.

Empirical analyses focused on two dependent variables. The first was a dummy measuring the presence of at least one child in the household, while the second was a continuous variable measuring the total number of children, born in the host country or elsewhere (see above), in the household. The independent variable was geographical origin, distinguishing migrants from native population according to the country of birth, except for Germany where we used nationality because descendants of German grandparents are automatically granted German nationality even when they are born abroad. Migrants were divided into nine categories: 1) Eastern Europe; 2) North Africa and near middle East; 3) South and Central Africa; 4) East Asia; 5) South and Eastern Asia; 6) Central America; 7) South and Latin America; 8) Western Europe; 9) North America and Australia. Empirical analyses excluded categories 8 and 9 because these migrants have peculiar features and are strongly selected in terms of education, skills, and so on (Ballarino and Panichella 2015) and their fertility patterns are similar to those in the countries of destination. These two immigrant groups were included in two ‘residual categories’, whose results are available on request.

We included a set of control variables in the analysis. The first was education, coded in three categories: a) lower-secondary or less (ISCED 0-2); b) upper-secondary or post-secondary non-tertiary (ISCED 3-4); tertiary (ISCED 5-6). The second was the employment condition, entered in dummy variables measuring the ISCO-08 code at 1 digit of the occupation. This variable also included two additional categories for the inactive and the unemployed. Models also controlled for year of the survey, age group (three 3-year dummies) and for dummies regarding the region of origin at NUTS2 level. When we studied the effect of the employment condition on fertility among couples (see below), we also controlled for the employment condition, the age and the country of origin of the partner. Both variables are defined as above.

We studied the first dependent variable, namely the presence of at least one child in the household, with a set of Linear Probability Models with robust standard errors (LPM), while OLS regressions were performed on the total number of children in the household. We divided the empirical analysis in two parts. The first focused on the pooled sample, while the second analysed the cross-country variations by estimating separate models for each country of destination, with the aim of considering the heterogeneity of fertility patterns among different origin groups (Wilson 2018).

Model 1 estimated the effect of migration status (MIGR) on the two outcomes considered, controlling for a vector of control variables concerning education and other socio-demographic characteristics (age, residence) (Z_i). This model also includes an interaction between survey year and region of residence (NUTS 2) to control the effect of several factors (economic crisis, employment opportunities and so on) related to every possible combination between calendar year and region of residence. In addition, the own occupational condition (OCC) is controlled for, to analyse whether the difference between migrants and natives is affected by their different inclusion in the labour market, as shown by previous studies (Ballarino and Panichella 2015; Panichella 2018).

Model 2 considers only those living in a couple at the moment of the interview, aiming at investigating the role of marital separation. This model includes the age, the country of origin and the occupational condition of the partner as control variables, as well as an interaction effect between the occupational conditions of both partners, which controls for the effects of the occupational homogamy of fertility behaviours. This specification allows studying not only the importance of the economic condition of the members of the couple, but also their reciprocal influence and interactions on reproductive choices.

In order to analyse the trends over time of migrants’ fertility, models 1 and 2 were estimated by splitting the immigrant groups according to the years of residence in the host country, considering six categories (1–2; 3–4; 5–6; 7–8; 9–10; more than 10).

We conducted a wide range of robustness checks with different models and specifications. They included: replication of the analysis considering women; replication of the analysis with a logit model and with an ordinal logistic model on an ordinal variable (no children, one child; two children; more than two children); estimation of the number of children with Poisson regression; estimation of the variation within countries by interacting geographical origin with country; estimation of a probit model with sample selection on the probability of having at least one child. Moreover, we also replicated models considering only ethnic endogamous couples.

The second set of robustness checks focused on the heterogeneities among different immigrant groups. First, we estimated models by country of origin, showing that results are generally confirmed within each migrants group (Table A3, appendix). We also estimated further models aiming to control whether the heterogeneity related to different origins biases our results. Since it is not possible to include both natives and migrants in the same model and control for country of origin, we performed additional analysis studying the effect of time spent in the host country, considering only the migrant population in the model. In these models, we controlled for geographical area of origin with the highest level of detail available in the data, as well as for its interaction with the country of residence. Even this ‘internal validity’ confirmed our results, without finding any systematic composition effect related to the origin (Figure A1, appendix).

Regarding the probability of having at least one child, migrants from Northern Africa, Near and Middle East, South-East Asia and Southern America have higher probabilities of parenthood than natives. This positive difference remains stable when the conditional model is estimated, confirming that for these groups of immigrants the main disruptive factor is not the marital separation per se. Conversely, as shown by the comparison between unconditional models with and without occupation as a control, their disruption primarily regards their inclusion in the labour market in the host country, mainly occurring at the lowest strata of the occupational hierarchy (results available on request). Results for Eastern Asian immigrants are similar, although for this group the coefficient in the unconditional model is not statistically significant. However, also in this case the gap with respect to natives does not change in the conditional model, suggesting that the main disruptive factor is not related with the marital separation.

On the contrary, Central and Southern Africans are ‘penalised’ in the unconditional model, but this penalty totally disappears and turns to an advantage in the conditional model, confirming that the probability of parenthood of this immigrant group is primarily limited by marital separation. However, further analysis shows that also their poor occupational condition accounts for the disruptive costs of migration (results available on request).

Migrants from Eastern Europe do not substantially differ from natives, and Central American immigrants are slightly less likely to have at least one child, independently from the occupation. The similarity between these two groups of migrants confirms the importance of the socialization argument, since in both cases migrants come from countries with fertility levels similar to (or even lower than) those in European countries (e.g. Bulgaria and Romania in Eastern Europe, Cuba and Barbados in Central America). Another similarity between migrants from Central America and Eastern Europe regards the role of marital separation, which does not significantly reduce the fertility of male migrants, as shown by the comparison between unconditional and conditional models. This finding is the consequence of the family migratory strategies of these fluxes, which often consist in a family migration trajectory where women are first migrants and men move afterwards (Thadani and Todaro 1984; Morokvasic-Müller 2002). These results are further confirmed by models estimating the association between geographical origin and probability of living in a couple at the time of interview (see Table A3, appendix).

If we analyse the number of children in the household, coefficients are generally positive and statistically significant in the unconditional model, with the two exceptions of male migrants from Eastern Europe and Central America, who are ‘penalised’ in fertility with respect to natives. Again, their lower family size does not significantly change in the conditional models, confirming the peculiarities of these migration fluxes (see above). In both cases, moreover, the lower number of children of migrants is also related to their lower fertility level in the country of origin. Finally, marital separation and family reunion are again crucial especially among migrants from Central and Southern Africa (see also Table A3, appendix).

Trends are similar across geographical origins and show increasing probabilities of having at least one child with respect to natives over time of residence in the host country. For instance, Eastern Europeans have 6 percentage points less in the probability of parenthood than natives if they have been living in the country of destination for less than two years, whereas they turn to have 3 percentage points more in the probability of parenthood if they have settled in the new society for at least ten years. Those coming from Asia have already higher probabilities of having at least one child than natives immediately after migration, but these probabilities become much higher in the long-run (13 percentage points higher than natives).⁴ These findings confirm that migrants have to face disruptive costs in the first years after the geographical move – for a longer period among Africans and a shorter period among Eastern Europeans and Latin Americans – and they resume their previous parenthood propensities once they successfully settle in the new society.

The conditional model gives some insights on the role of marital separation as a disruptive factor for migrants in the short-run and family reunion as a driver of parenthood in the long-run. As shown in the previous section, these dynamics are crucial especially among Africans, since their ‘penalisation’ in the propensity of parenthood after migration disappears if the model is estimated considering only those who are living in couples. On the contrary, those disruption effects related to partnership status are less important for Latin and Southern Americans, since their curves are similar in the unconditional and conditional models.

Taken together, these findings are consistent with the disruption and socialization arguments. Indeed, among all migrants the probabilities of parenthood and the number of children increase over time spent in the host country. This scenario primarily applies to migrants from Africa and Middle East and from Asia, although the factors behind are different. Among the former, marital separation and family reunion account for much of the ‘disruption’ in the short-run and in the ‘socialization’ in the long-run, respectively. Among the latter, disruption is higher if only cohabiting partners are considered, confirming that living in a couple does not favour their fertility immediately after migration. This scenario is also suitable to interpret the trend of the Eastern European migrants, although their number of children does not become higher than that of natives in the long-run. However, following the socialization argument, we should not expect more fertility from these migrants, coming from countries characterized by low fertility levels.

Eastern Europeans do not differ from natives with the only exception of Italy, where they have lower probabilities of parenthood. Finally, Southern and Latin Americans show more mixed results, since they are more likely to have one child than natives in France and Spain, less likely in Germany and similar to natives in The Netherlands. Italy is again an exception, since these migrants turn to be more likely to have a child in the conditional model, confirming that processes linked to family migration are crucial in this country, even for a group of migrants that usually follows migration strategies where women move before men.

Results show that migrants are ‘penalised’ with respect to – or do not statistically differ from – natives in fertility immediately after the geographical move, but they significantly offset the disadvantage over time in all the receiving countries, except The Netherlands. Therefore, we can confirm both the socialization and disruption arguments, with some differences across countries. This pattern is especially true in Italy and Spain: migrants living in these two countries for less than two years have much lower probabilities of parenthood – and number of children, in Italy – than natives, but those living there for several years totally reduced these penalties. According to our expectations, disruptive costs are hence higher in Southern European countries, since migrants are often included in the black labour market, which is informally regulated and leads to unskilled and poorly paid employment that does not give the right to access welfare services. Moreover, among the disruptive factors marital separation is more important in Italy than in Spain (and all other countries), as shown above and supported by the comparison between unconditional and conditional models.

Findings from The Netherlands show relatively stable differences between migrants from Africa, Asia and Latin America and natives over time, pointing to a support of the socialization argument without disruption. This is consistent with the structural characteristics of this country, where the labour market is regulated in a flexible way and is supported by a generous welfare regime giving migrants the same social rights as natives. However, the pattern of migrants’ fertility resembles that of other countries if we consider only those who have a partner at the time of interview, confirming a (small) disruption immediately after the move and a socialization – linked to family reunion – in the long-run.

European countries have witnessed increasing immigration streams and ethnic heterogeneity of their populations (Castles and Miller 2003). Improvement of social cohesion and the effect of immigration on social, cultural and demographic trends have become major issues in Europe and significant topics of research among social scientists. Nevertheless, few empirical studies investigated these processes from a comparative perspective (Milewski 2011; Kulu and Gonzàlez-Ferrer 2014; Kulu et al. 2015). This work has contributed to this discussion by analysing the fertility of male migrants in six European countries. The first important result obtained in the first part of the analysis concerned the relevance of disruption, which prevents male migrants from ‘transferring’ their fertility potential in the short run. Our analysis showed that disruptive factors change according to the country of origin. For instance, among immigrants from Northern Africa, Near and Middle East, Asia and Southern America the main disruptive factor is their entrapment in low-skilled and low-paying jobs once they enter in the labour market of the host country. This disruptive cost is crucial among male migrants from Central and Southern Africa as well, but they face an additional disruptive factor, namely marital separation. In this case, family migratory strategies where men migrate first leaving the partner in the country of origin limit migrants’ fertility in two ways. First, they tend to postpone reproductive behaviours to a period when most migration costs are paid off and the partners can join them. Second, women can join their husband in the country of destination together with their children left in the country of origin at the time of the geographical move, contributing to increase the population of the host society through the process of family reunion.

These findings also suggest that the effect of disruption is not constant over time; rather, its strength depends closely on time since immigration. Migrants are ‘penalized’ immediately after migration, but their probabilities of having at least one child and the number of children in the household increase once they have permanently settled in the new country. This pattern supports both the socialization and disruption arguments and applies to male migrants from any country of origin, especially to those coming from Africa, Middle East and Asia. It even applies to Eastern European migrants, who increase their fertility over time since migration, but they do not overcome that of natives, consistently with the very low fertility levels in their countries of origin.

The last part of the analysis studied whether the fertility gap between migrants and natives as well as the pattern over time since migration (disruption in the short-run and socialization in the long-run) change across countries with different institutional features, fertility levels, and immigration histories. Results showed that, on average, male Africans and Asians have higher fertility than natives in all countries of destination. However, even this group of migrants turns to have significantly higher propensities of parenthood and number of children than natives only some years after the geographical move, confirming the socialization and disruption arguments. Although this pattern is clear in all European countries considered, it is especially evident in Southern Europe, especially in Italy. In these countries, migrants from Africa and Asia have to face three sources of disruption, related to their specific institutional characteristics and to migrants’ family migratory strategies. The first refers to the inclusion of male migrants in the black labour market and in the poorest and unskilled occupations that does not give the right to access welfare services (Ballarino and Panichella 2015). The second regards their short immigration history, making these countries less capable to favour migrants’ integration through specific policies and dense ethnic networks. The third concerns marital separation, since most male migrants moving to Spain and especially Italy leave their family in the country of origin, postponing the reunion once the costs of migration are paid off. Southern European countries provide a ‘disadvantageous’ context – at least in the short-run – even for migrants from Latin America and Eastern Europe. Moreover, our results showed that processes linked to family migration – marital separation in the short-run and family reunion in the long-run – are crucial there, primarily in Italy, although these two groups of male migrants usually adopt peculiar family migratory strategies and move after the wife.

There are still some points requiring further investigation. First, other studies (and other datasets) are needed to overcome the main limitations of the EU-LFS data. Using cross-sectional data as we did in our analysis is a first step towards understanding this complex relationship, but it has some shortcomings. First, we could not observe the life-course of migrants after the move and look at the real pattern of disruption, selection, socialization and adaptation over time; rather, we simply compared different migrants who had been in the host countries for different numbers of years. This means that the changes in migrants’ fertility might depend on their selectivity over time of residence in the host country, as in the case of short-term migrants who can decide not to have children immediately after migration – or leave them in the country of origin – because they expect to return to the country of origin within few years. Second, we did not measure the exact fertility of migrants in the countries of destination, since we did not distinguish their children according to their place of birth. Nevertheless, considering all children should not be considered a limitation, since migrants contribute to the total fertility of the host country not only with children born after the move but also with those born in the country of origin. Given the limitations of cross-sectional data, longitudinal data with complete migration and birth histories are better-equipped and hence strongly suggested to study the relationship between migration and fertility further.

Finally, theoretical considerations often depend on the time-span considered in the analysis. This study has shown that once migrants settle in the new country, they tend to reproduce the fertility dominant in their country of origin (high fertility for those coming from Africa, Middle East and South-East Asia, and low fertility for Eastern Europeans and Central Americans). This is true for the first generation of immigrants, whereas it is likely that the mainstream society affects primarily the second generations, who are socialized into or adapts to the values, norms and behaviours of the native population, as shown by other empirical studies at a comparative level (e.g. Kulu et al. 2015). However, the second generation of migrants cannot be detected in the EU-LFS data. Therefore, the results of this study need to be replicated not only in other host countries but also with longer times after migration. Even this point can benefit from the use of longitudinal data, which would enable researchers to study several transitions simultaneously and to gain a ‘holistic’ picture of the family lives of migrants (De Valk and Milewski 2011).

No potential conflict of interest was reported by the authors.

Stefano Cantalini is Post-doc Researcher at the Department of Social and Political Sciences at the University of Milan. His current research interests include: fertility and family formation, social stratification and mobility, educational returns, gender inequality, sociology of the family.

Nazareno Panichella is Assistant Professor of Economic Sociology at the Department of Social and Political Sciences at the University of Milan. His current research interests include: social stratification and mobility, inequality of educational opportunities, migration, fertility, labour market dynamics and quantitative methods for social sciences.

Stefano Cantalinihttp://orcid.org/0000-0003-2493-9461

Nazareno Panichellahttps://orcid.org/0000-0002-7326-6817