Chapter 2. Intergenerational mobility among young natives with immigrant parents

Family characteristics

• The literature shows somewhat inconclusive results on how the number of siblings can impact educational attainment, but mostly family size is not a particularly strong explanatory factor. Older siblings may also function as a resource to younger family members, yet little is known whether this improves educational outcomes.

• The amount of years parents have spent in the host country appears to positively affect the educational outcomes of their children, mostly due to the parents’ language skills improving over time. More generally, there is some evidence that good language skills of parents positively impact their children’s educational outcomes, particularly when they are young.

• Research shows that educational aspirations are generally high among migrant families. However, while educational aspirations may be a pre-requisite for educational upward mobility, by itself they are not sufficient, particularly when support structures and knowledge on how to attain these goals is lacking.

Growing up in disadvantaged neighbourhoods

• There is strong evidence from quasi-experimental research that growing up in poor neighbourhoods has a long-term impact on labour market outcomes in adulthood. Although these studies generally do not make a distinction according to migration background, these insights are particularly relevant for natives with migrant parents, as in many OECD countries, considerable shares grow up in disadvantaged neighbourhoods.

• There are often strong assumptions that the residential segregation of immigrant communities in and of itself presents an obstacle to mobility, yet the evidence on its impact on education and employment outcomes is not that clear-cut. Higher shares of co-ethnics may be advantageous when these contacts can provide job opportunities or knowledge about vacancies; however, when ethnic groups do not possess such collective resources, labour market outcomes may be poorer. Thus, the impact of residential segregation of immigrant communities strongly depends on group-specific social capital.

Determinants on a school level

• The often-observed negative relationship between educational outcomes and high shares of students with immigrant parents is largely driven by socio-economic disadvantage, often mirroring socio-economic disadvantage at the neighbourhood level.

• There is strong evidence that early childhood education – given that it is widely accessible and of good quality – can increase educational mobility.

• Although this depends on a number of contextual factors, the majority of research indicates that overall, school systems that stream students only at a later age, e.g. around the age of 15, reduce the importance of parental socio-economic background.

• There is evidence that parents’ familiarity with the education system is particularly important when parents can choose their children’s schools or have to make decisions regarding school streams early on. A lack of such strategic knowledge can thus become an obstacle to educational mobility, yet there is little evidence on how immigrant parents’ knowledge concretely influences decision making.

• A number of studies have sought to assess the extent to which teachers’ expectations and attitudes towards students with a migration background impact their educational trajectories. However, results are highly mixed and most studies cannot disentangle to what extent expectations are informed by attitudes towards ethnicity/migration status or (assumed) social class.

Pathways and obstacles for intergenerational mobility in the labour market

• Research has clearly demonstrated the importance of networks and personal contacts for finding employment. Fewer networks may be a factor that limits school-to-work transitions for natives with a migration background, particularly if their parents cannot provide them with useful contacts.

• Vocational education and training (VET) systems can, under certain circumstances, facilitate school-to-work transition and present a pathway for upward mobility. However, in countries with well-established VET systems, natives with a migration background tend to be under-represented.

• There is evidence, mostly from English-speaking countries, that some ethnic minorities are concentrated in low-paying occupations and also tend to receive lower wages than equally qualified white workers.

• Field experiments show that natives with a migration background and ethnic minorities experience discrimination in the hiring process due to their ethnicity, religion and/or gender. Quantitative evidence on discrimination during employment, e.g. with regard to wages, promotions and layoffs, is still sparse.

Number of siblings and birth order

Extensive literature has argued that parents may face a quantity-quality trade-off with regard to the investments they make in their children, as their resources, in terms of money and time, are limited. Thus, having multiple siblings could negatively impact a child’s educational outcomes (Becker and Tomes, 1976). Growing up in a large family would therefore decrease intergenerational mobility, and even more so for children from low-income families who have fewer resources to invest. Some studies have found such a negative correlation between number of siblings and educational attainment (Sieben, Huinink and de Graaf, 2001), even when taking into account that family size might capture the impact of unfavourable socio-economic characteristics of large families, such as limited financial resources (Meier Jæger, 2008). Overall, however, the evidence is mixed and strongly depends on the statistical model that is used (Angrist, Lavy and Schlosser, 2006). Furthermore, birth order may be an important factor. For Norway, the impact of family size becomes insignificant once birth order is taken into account (Black, Devereux and Salvanes, 2005) and similar results are found for the United States and the Netherlands (de Haan, 2005).

Literature on the intergenerational impact of family size and birth order that focuses on immigrant families is sparse, despite the fact that in most countries young people with non-EU parents have more siblings than those with native-born parents.3 In addition, young people in the EU and OECD areas with a migration background are more likely to grow up in poor households, meaning that their parents have more limited capacities to invest in their children. Studies focusing on the impact of siblings on children of immigrants present rather mixed evidence.

The number of siblings does not significantly affect secondary school outcomes in Germany (Kristen and Granato, 2007; Luthra, 2010). Similar results are found for Norway, where the number of siblings has only a very small effect on the educational attainment of natives with immigrant parents. Being the first-born child, however, increases educational attainment on average by about 0.4 years both for men and women (Hermansen, 2016). For France, compared to other family characteristics, relatively small negative effects are found of numbers of siblings on natives with a migration background (Domingues Dos Santos and Wolff, 2011).

In contrast, Bauer and Riphahn (2007) show that as the number of siblings increases, children born in Switzerland with at least one foreign-born parent are significantly less likely to be highly educated. Controlling for a number of family background characteristics, they find a negative impact of family size for young people with a migration background. Having less-educated parents and three or more siblings instead of none or one reduces the likelihood of being highly educated by 6 percentage points (from 21% to 15%). Similarly, for France and Germany there is evidence that the impact of family size depends on the number of siblings, with sibling size having a strong effect on educational outcomes only if students have three siblings or more (Meurs, Puhani and von Haaren, 2015). However, natives with migrant parents are less affected by family size (in Germany the impact is insignificant) than immigrant students and those with native-born parents.

However, family size could also impact educational attainment in the opposite way, as older siblings could have a positive impact on the educational outcomes of their younger siblings. Particularly in families with immigrant parents who have little knowledge of the schooling system or limited capacities to support their children, older siblings could partly take on this role by helping their siblings navigate the schooling system. Evidence that points to the importance of older siblings is still sparse and largely qualitative. Those studies, largely based on in-depth interviews, show that older siblings are often an important resource of help for younger children (see for instance Moguérou and Santelli, 2015).

Looking at native-born students with Turkish parents in Austria, France and Sweden, Schnell (2014) finds as the family size grows, the amount of school support provided by older siblings, such as helping with homework, increases. This might indicate that with a higher number of siblings, more responsibility is shifted from parents to older siblings. Furthermore, in Austria, support from older siblings decreases the likelihood of being an early school leaver and increases the likelihood of attaining post-secondary education. This correlation remains significant after controlling for parental education and involvement. In France and Sweden, however, the support of older siblings has no significant effect. The author argues that these findings are likely to reflect differences in the education system; whereas education is full time in France and Sweden, most schools in Austria operate on a half-day system, which renders the family a more important resource for school and homework support.

Overall, evidence on the impact of family size for children of immigrants remains inconclusive. Although there seems to be some evidence that growing up in a particularly large family can be disadvantageous for educational attainment, in most studies family size is not a particularly strong explanatory factor. Therefore, it appears that it is not family size per se that has an impact, but rather other factors that are associated with growing up in a large family, such as limited economic resources. Therefore, institutional factors such as differences in education costs or school systems are likely to play into the impact of family size on intergenerational mobility. Moreover, little is known at this point about the extent to which older siblings can be resource for their younger siblings, and whether this translates into higher mobility rates for the younger siblings.

Parents’ length of stay in the host country

A wide sampling of the literature shows that both immigrants’ age of arrival and the years spent in the settlement country since migration strongly affect their own integration trajectories. Generally speaking, arriving at a young age and spending considerable time in the settlement country has a positive impact on indicators of integration, such as employment and language skills (Schaafsma and Sweetman, 2001; Böhlmark, 2008; OECD/European Union, 2015). Yet, very few studies assess the intergenerational effects of immigrant parents’ length of residence. This is somewhat surprising, considering that with more time spent in the country, parents may have better language skills, higher employment rates, more knowledge about the education system or more extensive networks than recently arrived immigrants, which in turn could render them better equipped to support their children.

In Canada, parental length of stay only modestly impacts their children’s vocabulary scores at the end of kindergarten, and has mostly insignificant effects on maths and reading scores at age 7 (Worswick, 2004). This may indicate that parental length of stay is mainly important for the transmission of language skills. The negative impact on vocabulary scores is slightly larger for parents who are neither English nor French native speakers and for children who scored among the bottom 10% in the vocabulary test.

Nielsen and Schindler Rangvid (2012) find for Denmark that parents’ years since migration have a positive impact on their children’s academic achievement. Mothers’ years since migration have a positive impact on exam scores in Danish language –and particularly so for their sons – whereas fathers’ years since migration do not affect Danish scores, but are found to positively impact math grades and decrease drop-outs.

A study of Sweden finds that parents’ length of residence in that country positively affects their children’s grades in Swedish and standardised language test scores, but has no significant impact on math scores. Assessing intergenerational effects according to the parents’ country of origin, the association appears to be somewhat stronger for parents whose origins are outside western countries (Smith, Helgertz and Scott, 2016).

Overall, it seems that the years parents spend in the country of settlement has a positive but mostly small impact on their native-born children. Furthermore, it appears that the advantage of longer parental residence mostly works through language skills transmission. Given that very few papers have assessed the intergenerational impact of length of residence and have only considered three countries (Canada, Denmark and Sweden), the evidence should be treated as tentative. Furthermore, there is currently no evidence whether parental naturalisation – which becomes more likely with more years spent in the country – may impact their children’s outcomes or how their initial motive for migration may affect their children’s mobility pattern.

Parental language skills

Although schooling and interaction with peers strongly impact language learning, there is evidence that children’s language proficiency remains associated with their parents’ language skills, as an important part of language learning takes place at home. Thus, parental language skills can be an important factor to explain why or why not natives with a migrant background do well in school and experience upward educational mobility.

Research has shown that language proficiency of one family member strongly correlates with the proficiency of other family members (Chiswick, Lee and Miller, 2005). However, language skills entail more than speaking a language correctly in day-to-day situations. Instead, students need to be able to read, understand and write texts to do well in school and the labour market later on. A number of studies have shown that children with less-educated parents – regardless whether foreign- or native-born – are more likely to have a smaller vocabulary and more difficulties in using academic language (Pan, Spier and Tamis-Lemonda, 2004; Becker, 2011), which can cause difficulties in reading comprehension, text production and ultimately, higher educational attainment. Therefore, studies that investigate how language skills are transmitted from immigrant parents to their children also need to take into consideration the impact of socio-economic background factors, so as to not confound immigrant-specific factors with lower socio-economic standing.

However, studies that examine how language skills are transmitted from immigrant parents to their children suffer from a number of limitations. The large majority of studies have to resort to rather imprecise measures of parental language skills, such as years spent in the country or self-reported fluency, which is found to be systematically biased for some origin groups (Edele et al., 2015 for Germany). Moreover, many studies look at the impact of language spoken at home and largely find that not speaking the country’s majority language at home negatively affects educational outcomes (Schnepf, 2007; Dustmann, Frattini and Lanzara, 2012; Sweetman and van Ours, 2015). Such findings do not reveal their children’s actual language skills or the language skills of the parents, and are therefore not a measure of language transmission across generations. Furthermore, language acquisition may be easier and faster for some immigrant parents, for instance when they are highly educated or speak a language that is linguistically close to the language of the host country (for an overview of the impact of linguistic distance on language learning, see Chiswick and Miller, 2005 and Isphording and Otten, 2013). Lastly, host-country language skills are not only transmitted from parents to children, but also vice versa. Children who improve their language skills at school may then in turn positively impact their parents’ language skills. At the same time, having children may also decrease parents’ language skills when it lowers their likelihood to work (Chiswick, Lee and Miller, 2005). Keeping these caveats in mind, a number of studies nevertheless indicate that limited language proficiency among parents negatively impacts the language acquisition and educational trajectories of their children.

In the United States, self-reported language proficiency of US-born children is positively impacted by parental language proficiency, yet this effect declines with the child’s age and reaches zero when children are in middle school (Bleakley and Chin, 2008).

Casey and Dustmann (2008) investigate how “language capital” is transmitted from immigrant parents to their children in Germany. Controlling for parental background characteristics such as education, income and years since migration, they find that parents’ language skills – here measured as self-reported fluency – remain associated with their children’s fluency. If parental language skills – coded from 0 (very bad) to 1 (very good) – increase by 0.1, their children’s language skills increase by about 2.5% when they are born in Germany and 3% when they are born abroad but arrived before the age of 10.

For France, Domingues Dos Santos and Wolff (2011) investigate whether immigrant parents’ self-reported language skills impact their ability to transmit their educational background to their native-born children. They find that parents’ proficiency in French has a strong, positive impact on the educational outcomes of their children. By introducing an interaction term between parental years of schooling and proficiency in French into the equation, they further show that returns to parental education on their children’s education are lower when parents report facing difficulties in speaking French.

Concluding, transmission of language skills is difficult to assess when there are only imprecise proxies for them. Moreover, language skills are not only transmitted from parents to children, but also from children to parents. Despite these caveats there is some evidence that parents’ good language skills positively impact their children’s language skills and educational attainment, and more so when children are still young.

Going to school with high shares of students with a migration background

If students go to a school in their neighbourhood, as is mostly the case, any concentration of disadvantage in the neighbourhood will also be present in schools. Immigrant students and natives with a migration background are often not evenly represented across schools; in the United States, United Kingdom and Canada, 60-65% of immigrant students would have to move to another school to achieve an even distribution across schools country-wide (Schnepf, 2004). This percentage is slightly lower for the Netherlands, New Zealand, Sweden, Germany, France and Australia (around 50%) and the lowest in Switzerland (40%). Moreover, schools are also split along socio-economic lines. Socio-economic segregation in secondary schools (according to parental background, but not to immigration status) is found to be particularly pronounced in Germany, Belgium and Hungary, somewhat less so in the United Kingdom and United States, and the least prevalent in Nordic countries (Jenkins, Micklewright and Schnepf, 2008).

Thus, the question is to what extent going to schools where the share of disadvantaged students is high impacts educational mobility for natives with a migrant background. The relationship between shares of disadvantaged students and educational outcomes is in fact unlikely to be a linear one. Instead, as a number of studies have argued, there may be a certain threshold or “tipping point” where a concentration of disadvantage becomes too high. Yet, it remains unclear where this threshold lies. In addition, this also depends on the capacity and preparedness of schools in responding to specific needs of students with low-income or immigrant parents (Szulkin and Jonsson, 2007; Andersen and Thomsen, 2011).

A number of studies on the impact of a high concentration of students with immigrant parents in schools demonstrate that it is not immigrant status per se, but rather a concentration of socio-economic disadvantage that has a negative effect on educational outcomes (Rumberger and Palardy, 2005; van der Slik, Driessen and de Bot, 2006; Lemaitre, 2012). The average socio-economic level of a school is therefore an important factor that mediates the intergenerational transmission of disadvantage, and as discussed below often impacts students with and without a migration background differently.

Varied literature seeks to determine the importance of peer effects in schools (for an overview, see Sacerdote, 2011). Yet, peer effects are often very difficult to measure because a large proportion of differences among student peers are an outcome of selection, such as ability grouping, parental choice or a school’s freedom to choose its students (Hoxby, 2000). Most studies examine how shares of immigrant students, natives with a migration background or ethnic minorities affect educational outcomes of natives without migration background (Brunello and Rocco, 2013 for a study of 19 OECD countries). Studies that consider both foreign-born students and natives with migrant parents are somewhat less frequent and often do not make a distinction between the two groups. Moreover, few studies assess the impact of diversity, i.e. the number and size of different ethnic groups in a school (Dronkers and van der Velden, 2013). However, ethnic diversity could affect educational outcomes differently than the overall shares of students with a migration background. Students attending highly diverse schools could, for instance, have more contacts with students from other language communities and therefore speak the language of instruction more often than in schools where one minority language group dominates. However, more ethnic diversity could also decrease schooling outcomes, for instance because teachers might be able to teach less effectively (Dronkers and van der Velden, 2013).

A meta-analysis – mostly of US studies – indicates that high shares of ethnic minority students have a larger effect on students from the same ethnicity than those of the majority group or another ethnic group (van Ewijk and Sleegers, 2010). However, they find only small effects, particularly when compared to the impact of parental socio-economic background. Effects on students with native-born parents appear to be close to zero. Nevertheless, it appears that the impact is not uniform for all minority groups. In the United States, shares of disadvantaged African-American students appear to have a stronger impact on educational outcomes than shares of immigrant students.

In the European context, the evidence of how the share of students with immigrant parents affects their peers with immigrant parents seems to indicate that the negative relationship is largely driven by socio-economic and school characteristics rather than immigrant status. Shares of students with immigrant parents are found to have no significant effect on the school performance of children of immigrants in the Netherlands (Veerman, van de Werfhorst and Dronkers, 2013), Spain (Cebolla-Boado and Garrido Medina, 2011) or Denmark (Jensen and Rasmussen, 2011), when taking into account socio-economic characteristics.6

Among native-born young people with Turkish and Moroccan parents in Belgium, Germany and Sweden, high shares of students with immigrant parents are found to have a small protective effect in Belgium and Sweden, increasing the likelihood of continuing with tertiary education as opposed to non-academic trajectories. In Germany, in contrast, native students with Turkish parents were less likely than their peers with native parents to attend university if shares of students with migrant parents were high (Baysu and de Valk, 2012). Fekjær and Birkelund (2007) look at schools’ ethnic composition in Oslo and its effect on grades. Controlling for the school’s overall socio-economic composition, they find a small positive impact on grades of students with and without a migration background. These results, however, are only applicable to upper secondary schools, and it is possible that students with a migration background in upper secondary school are a selected group that might have higher educational motivation or ability than students in other secondary streams.

Summing up, most studies find relatively minor or no effects of the share of students with immigrant parents on the educational attainment of other students with immigrant parents when controlling for socio-economic characteristics. Thus, the often-observed negative relationship between educational outcomes and high shares of students with immigrant parents is largely driven by selection of disadvantaged students who disproportionately happen to have immigrant parents.

Early childhood education

Across the OECD area, 69% of foreign-born and native-born children of immigrants who are between 3 and 6 years old were enrolled in preschool education in 2012, compared to 76% of their native peers. In most EU countries, however, differences are only marginal, particularly when services are free of charge. Notable exceptions are Italy, Norway and the Czech Republic, where participation rates differ by approximately 10% (OECD/European Union, 2015).

An extensive body of literature documents that participation in early childhood education can have significant positive effects on educational and labour market outcomes, especially for children from low-income and immigrant families (Heckman, 2011; Elango et al., 2015). For children with immigrant parents, preschool education has proved an important component in fostering language skills (Bleakley and Chin, 2008; Votruba-Drzal et al., 2015) and also positively impacts school performance later on (see for instance Spiess, Büchel and Wagner, 2003 for Germany; Magnuson, Lahaie and Waldfogel, 2006 for the United States; Drange and Telle, 2010 for Oslo; Schneeweis, 2011 for Austria).

Using results from the OECD’s Programme for International Student Assessment (PISA), which tests the mathematics, reading and science competencies of 15-year-olds, Figure 2.1 illustrates that attending pre-primary education has a strong impact on reading skills at age 15; a difference of 40 score-points is approximately equivalent to one year of schooling. In most countries, pre-primary education gives children of immigrants an educational advantage similar to one year of schooling when compared to their peers who also have foreign-born parents, but did not receive pre-school education. In the case of Italy and New Zealand, score differences even amount to the equivalent of more than two years of schooling.

Figure 2.1. PISA reading performance of students with immigrant parents, by attendance of pre-primary education, selected countries, 2012

Source: Adapted from OECD (2015), OECD Reviews of Migrant Education: Immigrant Students at School – Easing the journey towards Integration, https://doi.org/10.1787/9789264249509-en.

Intergenerational mobility can therefore be increased if early childhood education manages to equalise the playing field and ensure that children with immigrant parents are similarly prepared to enter elementary school. However, it appears that enrolment rates need to exceed 60% to have an equalising effect (Schütz, Ursprung and Wößmann, 2008). If participation rates are lower, it seems that it is mostly children from high- and middle-income families who attend preschool education, thereby increasing their educational advantage. Regarding elementary school education, a Swiss study assesses how intergenerational mobility is impacted by school starting age through regional differences in Swiss cantonal policies regarding age of entry into elementary school (Bauer and Riphahn, 2009). They find that earlier school starting age significantly impacts intergenerational mobility by decreasing the relative advantage of pupils with highly educated parents.

Thus there is strong evidence that early childhood education – given that it is widely accessible and of good quality – can increase intergenerational mobility, as it ‘intervenes’ in the education of children from disadvantaged backgrounds early on. Particularly for children of immigrants whose parents only have limited language skills, early childhood education is highly important for increasing their language proficiency and overall school readiness.

Early streaming in secondary school

For the most part, the literature indicates that early streaming, i.e. after elementary school, heightens the significance of family background (for a general review see Betts, 2011 and Burger, 2016). A number of studies have used quasi-experimental policy reforms to assess whether delaying early streaming has a causal impact on intergenerational mobility (Pekkarinen, Uusitalo and Kerr, 2009 for Finland; Meghir and Palme, 2005 and Holmlund, 2008 for Sweden). Although such policies were accompanied by other changes in the education system, such as longer compulsory schooling, it appears that later streaming contributed to higher intergenerational mobility independently of the other changes. However, there is considerably more controversy over whether comprehensive school systems decrease efficiency (for a discussion, see Pfeffer, 2015).

Moreover, the impact of early streaming depends on a variety of other factors, e.g. whether lower secondary education streams are (perceived as) a dead-end7 or to what extent different school streams are permeable and make it possible to switch between tracks easily. In addition, while comprehensive school systems are less selective, this may also reduce the signalling power of secondary school degrees to employers, i.e. the amount of information that degrees convey about the skills of a job applicant (Schröder, 2010). Furthermore, a cross-national study finds that while early streaming increases the importance of parental background, this impact is likely to be overstated when not including other selection mechanisms, such as school admission policies and peer effects (Raitano and Vona, 2016). The effect of streaming is largely diminished when controlling for school admission policies and the school’s social environment, yet higher socio-economic heterogeneity among students in schools is found to reduce the impact of parental background.

Relatively few studies have assessed how early tracking affects children of immigrants specifically. In Switzerland, educational mobility is found to be higher for native-born children of immigrants when streaming only occurs at a later stage in secondary school; yet the effect of early enrolment in kindergarten is still stronger than early streaming (Bauer and Riphahn, 2013).

A study covering 45 countries finds that early streaming in OECD and PISA-participating countries only negatively affects certain groups of students (Ruhose and Schwerdt, 2016). When comparing test scores in primary and secondary school, early streaming does not change test score differences over time between students with native and foreign-born parents. However, native-born students who do not speak the language of assessment at home, as well as foreign-born students who arrived only recently, are negatively affected by early streaming systems.

Another study based on PISA data, covering 11 countries,8 demonstrates that early streaming increases educational inequality between students with and without a migration background, partly because the impact of peers in school may be stronger in streamed school systems than in comprehensive schools (Entorf and Lauk, 2008).

Thus, while the impact of early streaming in secondary school remains a rather contentious topic, the majority of research indicates that overall, school systems that stream students only at a later age, e.g. around the age of 15, reduced the importance of parental socio-economic background on children’s schooling outcomes. However, there is some evidence that other factors, such as peer effects or enrolment in pre-school education, have a stronger impact on educational mobility.

Parents’ familiarity with the education system

Country-specific knowledge about the education system may have an important impact on children’s mobility, as little familiarity with these systems may render it more difficult for immigrant parents to support their children. For instance, parents from countries without well-established vocational education and training (VET) systems may not be aware of the potential benefits of this option for their children. In addition, in countries where parents can relatively freely choose their children’s schools or have to make choices regarding secondary education early on in their children’s education, children of immigrants may be at a disadvantage if their parents lack this type of strategic knowledge. Without distinguishing by immigration status, Pfeffer (2008) finds that educational mobility is lower in countries were secondary school streaming occurs early, and argues that this is the case partly because it requires parents to guide their children through these systems and make the right choices for them. Thus, limited knowledge may become a mechanism that strengthens the association between parents’ and children’s attainment.

In Germany for instance, Turkish parents are found to have less knowledge of the local primary school system than native-born parents and therefore pay more attention to a single school rather than other options, which is found to increase ethnic segregation in primary schools (Kristen, 2008). Moreover, parents’ familiarity with the education system is likely to matter beyond early education. Qualitative studies have demonstrated the often implicit need of “social know-how” to succeed in university, which is less available to students with parents who have not attended university themselves or obtained their university education in another country (Deil-Amen and Rosenbaum, 2003; Brooks, 2008).

Yet, how the knowledge of immigrant parents precisely impacts decision making and their children’s trajectories is more difficult to measure, and it is unclear at this point to what extent such knowledge increases over time. For instance, Dag Tjaden and Hunkler (2017) find that although natives with an immigrant background are more likely to decide against vocational education and training than comparable peers with native-born parents, this decision is better explained by high parental expectations, rather than a lack of information on the VET system (see also Section 3.4).

Programmes that foster parental involvement and give schools and teachers a proactive role in reaching out to immigrant parents are often proposed as a means to increase parents’ knowledge about the education system. However, policy evaluations are somewhat inconclusive, partly due to the large variety of programme content, which often also includes components to increase parenting or language skills, and because programmes often target low-income parents in general (for a review of US policy programmes targeting ethnic minority families, see Jeynes, 2003). Nevertheless, parental involvement overall appears to be positively associated with children’s educational outcomes regardless of socio-economic background or ethnicity (Schofield, 2006).

Teachers’ expectations

Teachers’ behaviour towards students with immigrant parents can be biased. Such behaviour can be comparatively visible, e.g. giving lower grades to students with immigrant parents compared to similarly achieving students with native parents. Yet, biased behaviour can also be more implicit, for instance having lower expectations, which in turn may discourage students and turn into a self-fulfilling prophecy of students performing worse (Boser, Wilhelm and Hanna, 2014). Furthermore, if students have less-educated parents, this may further decrease teachers’ expectations. In the United States for instance, a study finds that teachers treat minority students differently depending on whether their names are perceived to be typical “lower-class names”, adding to the evidence that ethnic and social class-based discrimination are intertwined (Figlio, 2005). Comparing siblings with mainstream and less common names also demonstrated that teachers had lower expectations of children with African-sounding names than Asian-sounding names. Other studies confirm that teachers’ expectations tend to vary for different immigrant/ethnic groups. Perceptions about groups that are seen as model minorities or problem groups appear to have an influence on why teachers tend to over- or under-assess their students’ academic potential (Burgess and Greaves, 2013).

However, ascertaining how to best quantify these often subtle forms of low expectations and prejudice remains a challenge. A number of studies assesses whether equally qualified students with foreign- and native-born parents have the same likelihood of receiving a recommendation for upper secondary school. In Germany for instance, a study finds that natives with a migration background are less likely to receive a recommendation for the highest secondary educational stream than peers with native parents (Lüdemann and Schwerdt, 2013). These differences persist when controlling for test scores in math and reading, yet become insignificant when taking into account socio-economic background, indicating the importance of social inequalities based on class. Similar results are found for the German-speaking part of Switzerland (Becker, Jäpel and Beck, 2013), whereas differences persist when controlling for individual grades and socio-economic background in Luxembourg (Klapproth, Glock and Martin, 2013).

An alternative approach to measuring expectations compares students’ grades with their performance in standardised tests, and assesses whether teachers predict their students’ performance in these tests differently based on migration/ethnic background. According to German data, teachers are found to generally overestimate their students’ performance, regardless of whether or not students have a migration background (Hachfeld et al., 2010). Studies that look at discrepancies between grades and standardised test scores yield somewhat inconclusive results. Teachers’ individual assessments of student performance tend to underestimate ethnic minority students in the United Kingdom (Burgess and Greaves, 2013), whereas in Sweden students with immigrant parents appear to be assessed more positively than their test scores would indicate (Lindahl, 2007).

A common problem with these studies, however, is the fact that it remains unclear to what extent teachers’ grading or assessments are based on personal, and possibly correct, knowledge of the student’s capabilities that are not captured in standardised testing. Therefore, a number of experimental studies seek to observe stereotype-driven behaviour of teachers towards fictitious students. But at the same time, social desirability is likely to be an issue in these studies: teachers may choose to respond in a way they perceive as more socially acceptable rather than stating their actual opinion.

Van Ewijk (2011), for instance, analysed how teachers’ grading of a written essay is impacted by the assumed ethnicity of the student. Randomly assigning Dutch, Turkish and Moroccan sounding names to these essays, which were then graded by about 100 elementary school teachers, did not reveal any direct grading bias. However, teachers are found to have lower expectations and negative attitudes towards students with immigrant parents. For instance, they were less likely to expect that students will continue with upper secondary education if the name on the essay is not Dutch-sounding. Thus in practice these attitudes could still negatively affect the educational trajectories of students with immigrant parents.

Using the same study design in Germany, essays of the same quality obtained significantly lower grades if they were assigned a Turkish-sounding name (Sprietsma, 2013). However, the effect is small and appears to be driven by a small number of teachers. For fictitious students with a Turkish name, teachers were also less likely to give a recommendation for upper secondary education.

Considering that all three approaches to testing discrimination and lower expectations have their drawbacks, it is difficult to draw conclusions about the extent to which ethnicity-based discrimination in schools is an issue. First, studies show that socio-economic status and migration background as reasons for differential treatment are closely intertwined, but difficult to disentangle with any precision. Second, teachers’ behaviours, such as lower expectations or differential treatment, can be unconscious and very subtle, thereby making their measurement rather complicated. Lastly, social desirability bias makes it difficult to estimate the “true” extent of discrimination in schools and their impact on educational outcomes of the children of immigrants. Therefore, overall the evidence of the impact of teachers’ expectations on educational mobility remains inconclusive.

School-to-work transition

The transition from school to employment has been highlighted in the literature as a critical point in young people’s lives, where spells of unemployment are found to have particularly negative long-term consequences on earnings, employability and career trajectories (Scarpetta, Sonnet and Manfredi, 2010). Moreover, there is some evidence that long-term consequences on earnings are less severe for unemployed young people with high-income parents than for young people from low-income families (Sirniö, Martikainen and Kauppinen, 2016), and that natives with a migration background are more strongly affected by high (youth) unemployment rates (Lutz, Brinbaum and Abdelhady, 2014).

In 2009, the average duration in the European Union between leaving school or university and finding employment is 10 to 13 months and comparable between natives and native-born young people with immigrant parents (OECD/European Union, 2015). By definition this only includes those who managed to find employment. Yet, in almost all OECD countries with available data, the shares of young people not in education, employment or training (NEET) are higher for natives with a migration background than for their peers without migration background (except for Australia, Canada and Israel). On average in the EU, around 20% of natives with a migration background fell into this category, compared to about 16% of those with native-born parents (OECD/European Union, 2015).

The relatively high shares of NEET natives with a migration background are partly due to their overrepresentation among the less-educated, who generally face more difficulties in finding employment. Although higher education helps to some extent in finding employment, a number of studies show that high educational outcomes do not necessarily translate into the respective jobs or higher earnings later on (Connor et al., 2004; Dustmann and Theodoropolous, 2010; Krause and Liebig, 2011). Other channels have been proposed that impact school-to-work transition, particularly for natives with migrant parents, including the importance of social networks as well as the role of vocational education and training (VET).

The effects of networks on entering the labour market

Although the importance of networks for finding employment is relatively well established, fewer studies have focused on their importance for young people’s first entry into the labour market. In addition, it should be kept in mind that the large majority of research on social networks cannot determine a causal effect. As Mouw (2006) argues, friends and acquaintances do not form networks randomly. Since people tend to know and befriend those who are already similar to them, effects attributed to networks may in fact reflect unobserved selection effects (see also Mouw, 2003).

Nevertheless, there is a relatively firm consensus that a broad social network helps young job seekers to obtain relevant information and better opportunities to apply to and get accepted for jobs, provided their social contacts are useful and can be called upon. This shows that it is crucial to consider a network’s composition and the resources accessible through these networks, rather than network size alone (Behtoui, 2015).

The definition of a social network is wide and includes essentially each social connection a person has – and that, in the context of job search, can help to find employment. As young people may not have built up their own professional network and therefore have to rely more strongly on their parents’ contacts, networks can be seen as a form of social capital that enhances status transmission across generations and put those with limited resources at a disadvantage. Young people with immigrant parents can thus be at a particular disadvantage if their parents either have a limited network and/or mostly contacts that cannot help with finding employment.

Putnam (2000) famously distinguished between bridging and bonding capital – links between different networks or socio-economic groups, and within a social group. Thus, young people with less-educated immigrant parents might be at risk of not having enough “bridging contacts” to connect them to opportunities outside their social circle. Burt (1992) also described this phenomenon as “structural holes” that mirror a person’s social position with potentially strong ties within their own network, but little access to networks outside their community. Studies have argued that such networks effects partly explain the maintenance of “ethnic minority businesses” across generations, or the strong prevalence of self-employment among immigrants and their descendants (Andersson and Hammarstedt, 2010).

Research has shown that ethnic minorities and natives with a migration background have fewer “bridging” contacts to people in higher social positions (Li, Savage and Warde, 2008 for the United Kingdom); get fewer job leads (McDonald, Lin and Ao, 2009 for the United States); and receive less help from their social network when applying for apprenticeships (Beicht and Granato, 2010 for Germany).

To date there is little empirical evidence on whether limited networks put natives with a migration background at a disadvantage when searching for a first job. In Belgium, social capital – here measured as respondents’ social connections to people with different occupations – is found to positively impact the likelihood to find a job after finishing vocational education and training (Verhaeghe, van der Bracht and van de Putte, 2015). Observed differences in social capital between those with Belgian- and foreign-born grandmothers from Morocco, Turkey or the Balkans are explained by socio-economic background.

Roth (2014) looks at the importance of social networks in finding an apprenticeship as part vocational training in Germany. Germany’s dual vocational training combines schooling in vocational schools with on-the-job training, and requires students to apply directly to companies for their apprenticeship; for a considerable proportion of young people, this represents the first transition into the labour market. Young people with Turkish parents are less likely to report that their network was able to help them during the apprenticeship search than those with native parents. After controlling for background factors and grades, they are also less likely to find an apprenticeship. Furthermore, only mothers’ networks – as opposed to young people’s own contacts – had a positive effect on finding apprenticeship, indicating the importance of parental networks for young people. Whereas the ethnic composition of mothers’ networks does not appear to have an impact, only those contacts working in low- to medium-skilled professions had a positive impact, which highlights that the usefulness of social capital is context-specific.

Vocational education as a means of facilitating school-to-work transition

In some OECD countries – notably the Netherlands, Germany, Switzerland and Austria – apprenticeships have been found to facilitate school-to-work transition and even more so for children of immigrants (OECD, 2012). However, in a number of countries, such as Switzerland, Austria and Denmark, children of immigrants are also at a higher risk of dropping out of apprenticeships compared to young people with native-born parents (OECD, 2012; Schindler Rangvid, 2012). There is also evidence that qualified young people with a migration background encounter difficulties in finding apprenticeship places, often reflecting a complex interaction between limited social networks, discrimination in the hiring process, and competition in the sector to which they applied (Helland and Støren, 2006 for Norway; Schneider, Yemane and Weinmann, 2014 for Germany).

In countries where VET programmes are generally less appreciated by employers and often perceived as a “dead end”, children of immigrants tend to be overrepresented in vocational streams. In the French community of Belgium, for instance, students with foreign-born parents are overrepresented in the vocational track – in 2004/05, more than 30% of students enrolled in the vocational track had a foreign nationality (OECD, 2008).

Furthermore, Brekke (2007) finds for Norway that natives with non-Western migrant parents are slightly less likely to secure employment after graduation than those with native parents and comparable socio-economic background (64% vs. 68% respectively), and that the likelihood for foreign-born graduates is considerably lower (57%). In Denmark, a study shows that natives with a migration background have fewer job offers after finishing vocational education and higher layoff rates than comparable peers with native-born parents (Datta Gupta and Kromann, 2014).

In addition, in some countries children of immigrants tend to be overrepresented in VET pathways that lead to comparatively low-skilled and low-paid jobs. In Canada, Crocker et al. (2010) find that immigrant and minority women – while being underrepresented in VET in general – largely take up apprenticeships in the areas of hairstyling and food production. A survey on students with foreign-born parents in Germany also found that they are strongly underrepresented in those job areas that have the highest satisfaction rate among respondents (Haggenmiller, 2015).

Hence, whether VET promotes school-to-work transition among natives with a migration background appears to vary across countries. The literature has sought to explain why in some countries VET systems are more successful than elsewhere in facilitating entry into the labour market for natives with immigrant parents. Generally, schemes that combine part-time schooling with firm-based training are generally found to be particularly successful in facilitating school-to-work transition compared to VET systems that are based on full-time schooling (Wolter and Ryan, 2011). A strong work-based component may help to signal applicants’ practical skills to future employers and provide first work experience. This could particularly help applicants with immigrant parents if they do not have large networks or face the risk of discrimination (Schröder, 2010).

Sorting into occupations and wage gaps within occupations

There is strong evidence that natives with migrant parents and ethnic minorities are disadvantaged in the labour market, and that these differences cannot be explained by age or educational attainment. Furthermore, in most European OECD countries with available data, natives with immigrant parents are slightly more likely to be overqualified for the type work they are doing than those with native-born parents (OECD/European Union, 2015). However, these “ethnic penalties” (Heath and Cheung, 2006) differ across ethnic groups as well as countries.

Why this is the case is more difficult to determine. Whereas immigrants are likely to have foreign qualifications that employers may value less, natives with a migration background went through the same schooling systems as their peers without a migration background, and should therefore be equally able to benefit from their education as their peers with native-born parents. There are a number of explanations why earnings tend to be lower for natives with a migration background compared to similarly educated workers with native-born parents: natives with a migration background may work in lower-paying occupations or sectors, but they may also be paid less within a given occupation than workers with native-born parents (Altonji and Blank, 1999). If these wage gaps persist when controlling not only for education but also for factors such as work experience, age and location, discrimination may be a factor. Moreover, discrimination in the hiring process can also decrease representation in certain job fields, making “sorting into occupations” not an active choice, but an outcome of discriminatory practices.

Detailed analysis of how children of immigrants are distributed across occupations and to what extent pay is different within a given occupation is still limited. There has however been considerable research into what explains wage gaps between similarly qualified men and women in the same occupation. There is evidence for the United States, for instance, that within-occupation inequality for women is more pronounced in some jobs than in others and that for occupations where wage gaps are the smallest, median wages tend to be lower (Baxter, 2015). Furthermore, it has been argued that wage gaps partially reflect the fact that women are more likely to have jobs with flexible work hours, particularly when they have children, and that more flexible hours come with high wage penalties (Goldin, 2014). However, there is also evidence that once women become more strongly represented in a given occupation, wages drop (Levanon, England and Allison, 2009).

Although it is unlikely that such mechanisms are the same for natives with a migration background, it is still plausible that some of the findings on gender wage gaps could also apply to natives with a migration background – for instance, that wages might be lower in occupations with high shares of workers with a migration background.

A study in the United Kingdom investigates why ethnic minorities (self-identified without taking into consideration country of birth) are more likely to be paid below the living wage and how this is related to occupational choices (Brynin and Longhi, 2015).9 They find that wage gaps compared to white British workers are limited within a given occupation – except for Pakistani and Bangladeshi minorities whose wages differ from white British workers within a specific occupation – but that ethnic minorities are concentrated in low-paying occupations. This may indicate that selection of or access to occupations is an important factor in explaining wage differences. Moreover, having a university degree appears to benefit ethnic minority workers and white British workers to a similar extent, here measured in hourly wages compared to those without university education. White British workers do maintain, however, a small advantage (a 52% higher wage than those without university education compared to around 48% for other ethnic groups). Studies in the United States (Grodsky and Pager, 2001) and Canada (Hou and Colombe, 2010) also demonstrate that wage gaps between ethnic minority and white workers persist within occupations, pointing to the issue of discrimination.

Discrimination during the hiring phase and employment

While discrimination in the labour market is difficult to quantify precisely, a considerable number of studies indicate that natives with a migration background are disadvantaged in the hiring process due to their ethnicity, religion and/or gender (Heath, Liebig and Simon, 2013; Valfort, 2015; Arai, Bursell and Nekby, 2016). Different field experiments have been developed to assess the extent to which discriminatory hiring practices bar minority job applicants from accessing the labour market. These include audit studies where actors play similarly qualified minority and majority applicants and are sent to job interviews, as well as correspondence studies that measure call-back rates for fictitious CVs that are sent to employers. The latter method is generally considered to yield more reliable outcomes, as it removes personal interaction between applicant and interviewer that may impact recruitment decisions beyond observable characteristics. At the same time, correspondence studies by definition are limited to jobs that are filled through a formal, written application process. In other sectors where applications are more often made in person, it remains unclear to what extent discrimination is an issue. Moreover, they cannot yield evidence whether discrimination occurs after the interview stage, e.g. with regard to wages, promotions or layoffs.

Nevertheless, correspondence studies show that due to discrimination at the hiring stage, ethnic minorities and children of immigrants face additional hurdles in entering the labour market. A meta-analysis of 22 studies in 16 OECD countries shows that job applicants with a minority background have to send out more – in many cases, twice as many – applications before they receive a positive reply compared to equally qualified, white candidates (Heath, Liebig and Simon, 2013). Similar evidence is available for applications to apprenticeships (e.g. for Germany, Schneider, Yemane and Weinmann, 2014). A number of studies also find considerable variation in call-back rates between ethnic groups (e.g. Booth, Leigh and Varganova, 2012 for Australia; McGinnity and Lunn, 2011 for Ireland; Wood et al., 2009 for the United Kingdom). However, an international comparison or “ranking” based on these studies is not feasible, as they apply different methodologies and consider different occupational sectors, jobs and ethnic groups.

Based on register data on name changes in Sweden, Arai and Skogman Thoursie (2009) show that annual earnings increase after individuals change their surnames from African, Asian or Slavic names to Swedish-sounding names. However, this is not the case for persons changing Finnish-sounding to Swedish sounding names, or those who changed from one non-European name to another non-European name. The authors find that a large part of the effect is due to higher chances of being employed, and conclude that discrimination in the hiring stage is an important explanatory factor for wage differences.

Furthermore, ethnic minority status often intersects with other grounds for discrimination, such as religious belief and gender, potentially leading to particularly unfavourable situations for certain subgroups. Evidence from France demonstrates that religious minorities who signal in their applications that they are practicing their religion experience considerable disadvantages in the labour market (Valfort, 2015). Practicing Catholics are 30% more likely than practicing Jews and twice as likely as practicing Muslims to receive a positive response. Furthermore, practicing Muslim men experience particularly high rates of discrimination. Whereas call-back rates among practicing Catholic women are 40% higher than among practicing Muslim women, for male practicing Catholics the call-back rate is almost four times higher than for practicing Muslim men.

It remains somewhat unclear whether ethnic minority men or women experience more discrimination and to what extent this is due to gender, ethnicity or (assumed) religiosity. In the United Kingdom, for instance, 13-16% of ethnic minority women stated they were questioned about their plans for marriage and children in job interviews compared to 6% of British white women (Equal Employment Commission, 2006).10 In Sweden, however, it appears that women with Arabic-sounding names can “compensate” initial hiring discrimination (measured in call-back rates) when they have more work experience, which is not the case for male applicants (Arai, Bursell and Nekby, 2016). At the same time, women may also be more strongly affected by discrimination due to religious dress. In Belgium, for instance, around 44% of employers indicated that an applicant wearing a headscarf would impact their hiring decision (Lamberts and Eeman, 2011). In Germany, women with a Turkish-sounding name who wear a headscarf are found to be strongly disadvantaged in the hiring process (Weichselbaumer, 2016).11 They received positive feedback only from 4% of the employers that were contacted in a correspondence study, compared to 14% for women with a Turkish-sounding name and 19% for women with a German-sounding name. For management roles as compared to secretary roles, differences were even more pronounced. This indicates that the prevalence of discrimination can also depend on the level of seniority of a given position. Overall, however, it is unclear whether discrimination during the hiring phase is more widespread for low- or high-skilled jobs, or how this differs across sectors and countries.

Compared to discriminatory hiring practices, there is less and mostly indirect evidence on how discrimination impacts later career trajectories, such as wages, promotions and layoffs, also because the impact of discrimination is more difficult to quantify. However, a considerable body of literature has developed that seeks to identify the magnitude of “ethnic penalties”, i.e. differences in labour market outcomes that remain after controlling for relevant background factors, such as education, age, sector or work experience (Heath and Cheung, 2006). Although these penalties cannot be linked directly to it, discrimination is still likely to be an important factor considering evidence from correspondence studies.

Regarding wages, Hou and Colombe (2010) find that Canadian-born visible minorities are overall paid less for similar jobs in the private sector than white Canadian workers with similar education levels and work-related characteristics, such as work experience. Whereas for Chinese and South Asian minorities this gap is between 3% and 6%, it was found to be considerably larger for black minorities (11% and 16% for women and men, respectively). For public sector jobs, no significant differences are found. A study in the United States also shows that for black minority men working in the private sector – without distinguishing by country of birth – wage differences become larger compared to white workers when occupations are more highly paid (Grodsky and Pager, 2001). This association remains when controlling for a number of human capital occupational characteristics. Thus, wages differences appear to also depend on average occupational earnings. This relationship is not found in the public sector.

Furthermore, there is some evidence for the United States indicating that even when taking workers’ performance reviews and ratings into account, salary growth, promotions and layoffs continue to differ by gender and ethnicity (Castilla, 2012). However, these findings are based on the employment history of around 6 000 employees working in the same company and therefore cannot be generalised. In addition, firm downsizing appears to affect ethnic minorities more strongly than other workers (Couch and Fairlie, 2010 for the United States), and also when they are in managerial positions (Kalev, 2014).

Family characteristics

Evidence on the impact of the number of siblings is inconsistent, ranging from a small or no negative impact to considerable negative effects for larger number of siblings (Luthra, 2010; Hermansen, 2016; Riphahn and Bauer, 2007). Mostly, however, family size is not a particularly strong explanatory factor, when other characteristics such as income are accounted for. Moreover, very little is known at this point about the extent to which older siblings can be a resource for their younger siblings, and whether this can translate into higher mobility rates for younger siblings (Schnell, 2014).

Moreover, the amount of years the parents have spent in the host country appears to positively affect educational outcomes of their children, mostly due to the parents’ better language skills. However, the impact is small and evidence is only available for a few countries (Worswick, 2004; Nielsen and Schindler Rangvid, 2012; Smith, Helgertz and Scott, 2016). While the parents’ reason for migration may also impact intergenerational mobility, there is currently no research that disentangles how different reasons for migrating or legal status may impact intergenerational mobility.

Intergenerational transmission of language skills is difficult to assess when there are only imprecise proxies for language skills, such as years spent in the country or self-assessed language skills. Moreover, language skills are not only transmitted from parents to children, but also vice versa, making it difficult to rule out reverse causality. Despite these caveats there is some evidence that parents’ good language skills positively impact their children’s educational attainment, and more so when the children are still young (Bleakley and Chin, 2008; Casey and Dustmann, 2008).

Educational aspirations among immigrant parents and their children are generally found to be high (OECD, 2015). Whereas high educational aspirations may be a prerequisite for overcoming initial disadvantage, they do not appear to be sufficient when concrete knowledge on how to attain these goals is lacking (Gresch et al., 2012; Cummings et al., 2012).

Growing up in disadvantaged neighbourhoods

While there is evidence that growing up in a poor neighbourhood – not accounting for migration background – generally has long-term, negative effects on labour market outcomes (Chetty et al., 2016; Rothwell and Massey, 2015), it is less clear how a high concentration of immigrants at a neighbourhood level impacts mobility of natives with a migrant background. Literature that has aimed to capture immigration-specific factors of residential segregation shows that its impact strongly depends on the (often group-specific) economic and social resources of migrant communities (Zuccotti and Platt, 2016; Grönqvist, 2006).

Determinants of intergenerational mobility on a school level

In the majority of OECD countries, natives with migrant parents are overrepresented in schools with high shares of pupils with a migration background. However, the literature finds a relatively minor or no effect of the share of students with immigrant parents on educational attainment when controlling for socio-economic characteristics (Lemaitre, 2012; Veerman, van de Werfhorst and Dronkers, 2013). Thus, the often-observed negative relationship between educational outcomes and high shares of students with immigrant parents is largely driven by socio-economic disadvantage.

There is strong evidence that early childhood education – given that it is widely accessible and of good quality – can increase intergenerational mobility. Especially for children of immigrants with limited language proficiency, early childhood education is highly important to increase their language skills and overall school readiness (OECD, 2015).

Regarding early streaming, i.e. sorting students into different educational tracks according to their academic ability, the evidence is somewhat less clear-cut. Yet the majority of research indicates that overall, school systems that stream students only at a later age, e.g. around the age of 15, reduces the importance of parental socio-economic background, including for children of immigrants (Meghir and Palme, 2005; Pekkarinen, Uusitalo and Kerr, 2009; Ruhose and Schwerdt, 2016).

Parents’ familiarity with the education system is likely to have an impact on how well they can support and guide their children through their educational career, particularly when parents can choose their children’s schools or have to make decisions regarding school streams early on (Pfeffer, 2008). The lack of such strategic knowledge can thus become a mechanism that limits the educational mobility of their children, but little evidence is available on how limited familiarity concretely impacts parental decision making.

Lastly, obtaining an accurate picture of teachers’ expectations and potentially discriminating attitudes towards students with a migration background is difficult, not only because such behaviour can be very subtle and difficult to quantify, but also because students’ social class plays into teachers’ expectations (Figlio, 2005; Lüdemann and Schwerdt, 2013). As a consequence, the impact of teachers’ – potentially biased – attitudes towards children of immigrants is highly mixed (Burgess and Greaves, 2013; Lindahl, 2007).

Pathways and obstacles for intergenerational mobility in the labour market

The transition from school to work has been highlighted in the literature as a critical point for natives with a migration background, who are often less successful in finding employment. In most countries these differences are not explained by differences in educational attainment. Fewer networks may be a factor that limits school to-work-transitions for natives with a migration background, particularly if their parents cannot provide them with useful contacts (Li, Savage and Warde, 2008; Beicht and Granato, 2010; Roth, 2014). In some countries, vocational education and training systems can facilitate school-to-work transition for natives with a migration background under certain circumstances, and can present a pathway for upward mobility (OECD, 2012).

Sorting into low-paid occupations or receiving lower wages for a given job than their colleagues may also hinder mobility in the labour market. Detailed analysis of how children of immigrants are distributed across occupations and to what extent pay is different within a given occupation is still limited. There is, however, some evidence, mostly from English-speaking countries, that certain ethnic minorities are concentrated in low-paying occupations and also tend to receive lower wages than equally qualified white workers (Hou and Colombe, 2010; Brynin and Longhi, 2015), pointing to the issue of discrimination in the labour market. Yet, it should be highlighted that in these studies no distinction is made whether persons are foreign- or native born.

Field experiments show that natives with a migration background and ethnic minorities can experience discrimination in the hiring process due to their ethnicity, religion and/or gender, and often have to send out considerably more applications before being invited to an interview (Arai and Skogman Thoursie, 2009; Heath, Liebig and Simon, 2013; Weichselbaumer, 2016). Studies on discrimination during employment, e.g. regarding wage differences, promotions and layoffs, are less frequent, yet there is evidence for some countries that ethnic minorities receive less pay than similarly qualified peers (Hou and Colombe, 2010; Grodsky and Pager, 2001).

Directions for future research

Reviewing the literature reveals that there are still considerable research gaps in a number of areas. Whereas these gaps are often due to data limitations – either because data are not collected or because the population of natives with a migration background is too small – the following highlights research areas to be pursued further within the intergenerational mobility literature. Some of them are already well developed, but have not yet focused on immigrant families.

First, little is known about how intragenerational mobility, i.e. how the social mobility of immigrant parents in the host country and intergenerational mobility are connected. It seems plausible that they can be interrelated; intergenerational mobility in immigrant families might be different if parents are themselves upwardly or downwardly mobile compared to parents who do not experience mobility during their lifetime. Thus, taking into account that family circumstances are often not static and connecting these two aspects of mobility would require tracking immigrant parents’ outcomes over time and assessing how this might impact their children’s trajectories. Relatively few studies have assessed the relationship between intra- and intergenerational mobility (see, however, Plewis and Bartley, 2014), and so far none has considered this relationship for immigrant families.

Second, sibling studies can be an interesting avenue, as shared genes and a similar social environment while growing up makes them more similar than other members of society. Sibling correlations are therefore arguably a broader measure of the impact of family and neighbourhood than intergenerational estimates based on parental income or education, and have become an increasingly frequent means of assessing drivers of intergenerational mobility (Black and Devereux, 2011). Other studies have compared adopted children and twins in an attempt to determine causal effects (see Holmlund, Lindahl and Plug, 2011 for an overview; Black et al., 2015), yet this approach may not be feasible when focusing on immigrant families only. Nevertheless, sibling correlations could prove to be a promising avenue for future research on natives with a migration background.12

Third, there is still inconclusive evidence regarding the relative impact of mothers’ and fathers’ socio-economic background on their children’s mobility, and whether this also depends on the gender of the child. Moreover, relatively little is known about women’s intergenerational earnings mobility, as the majority of studies observe father-son pairs. While limited data may still present a considerable challenge, much could be gained from having a better understanding of which characteristics of mothers and fathers have a stronger effect on their children, and whether this differs between immigrant and native families.

Fourth, a number of studies show that intergenerational mobility tends to be overestimated when only taking into account two generations (Pfeffer, 2014). With regard to the children of immigrants, this is a particularly important issue. Whereas children of immigrants growing up in disadvantaged families are largely found to do better than their parents’ generation, it is often (implicitly) assumed that this trend will continue in the grandchildren’s generation. However, the extent to which this is the case remains largely unclear in European countries. Although administrative data in Scandinavian countries as well as a number of recent surveys have been used to identify respondents with foreign-born grandparents (Andersson and Hammarstedt, 2010; Fick et al., 2014), multigenerational mobility of immigrant families remains under-researched.13 More studies with a multigenerational approach exist in the United States, Canada and the United Kingdom, but these largely rely on ethnic self-identification rather than the grandparents’ country of birth, which renders mobility analysis difficult (Duncan and Trejo, 2016). Focusing on the grandchildren of immigrants also raises questions of how to define and identify them in the data. Indeed, it remains debated whether for a “third generation immigrant” migration status or ethnic minority status is the more insightful characteristic to explain social mobility. Moreover, across three generations, within-group differences are likely to increase as well – e.g. due to intermarriages or internal mobility – and would require additional research attention paid to the internal heterogeneity of ethnic minority or immigrant-origin groups (Alba, Jiménez and Marrow, 2014). Although the numbers of grandchildren of immigrants are still rather small in many European countries, multigenerational approaches may become a relevant research topic in the future and provide a more long-term perspective on the intergenerational mobility of immigrant families.