Jump to ContentJump to Main Navigation
Growing up in Diverse SocietiesThe Integration of the Children of Immigrants in England, Germany, the Netherlands, and Sweden$

Frank Kalter, Jan O. Jonsson, Frank van Tubergen, and Anthony Heath

Print publication date: 2018

Print ISBN-13: 9780197266373

Published to University Press Scholarship Online: May 2019

DOI: 10.5871/bacad/9780197266373.001.0001

Show Summary Details
Page of
date: 30 May 2020

Dealing with Diverse Diversities: Defining and Comparing Minority Groups

Dealing with Diverse Diversities: Defining and Comparing Minority Groups

Chapter:
(p.62) 3 Dealing with Diverse Diversities: Defining and Comparing Minority Groups
Source:
Growing up in Diverse Societies
Author(s):

Frank Kalter

Anthony Heath

Publisher:
British Academy
DOI:10.5871/bacad/9780197266373.003.0003

Abstract and Keywords

In this chapter we use the CILS4EU data to investigate the precise generational status and the origin countries of the adolescent population in England, Germany, the Netherlands and Sweden. We describe the ethnic diversity in the study’s samples in more detail and show that it is very large in each of these countries. In addition, the composition of origin groups varies greatly across the four. This is a challenge for straight-forward comparisons between the countries, which is further complicated by the fact that generational status and origin countries are confounded. The chapter discusses the opportunities and limitations for the empirical analyses in the rest of the book. Basically, we argue for a strategy that regards country and group differences as phenomena of interest rather than of nuisance. They should be seen as descriptive facts and starting points of a search for explaining mechanisms.

Keywords:   comparison, diversity, generational status, origin groups, ethnic minorities

3.1 Key Challenges of Comparisons in Integration Research

A LEADING AIM OF EMPIRICAL integration research, and also of this book, is to detect the major mechanisms fostering or hindering the integration of ethnic minorities—relative to, and in interaction with, their majority peers. Within this overall endeavour the comparative perspective fulfils a specific and important strategic task: studying integration processes in different countries provides a much stronger empirical test for the generality of theoretical arguments and hypotheses. Comparisons might reveal common processes that generally operate in similar fashions despite the diversity of contexts, or they might show that processes operate somewhat differently in different contexts and that the validity of some mechanisms might depend on certain additional conditions. To know these conditions is particularly relevant also from a policy perspective as they point to institutional arrangements that could be more promising than others in leading to integration-related societal goals such as greater social cohesion or more equal life chances.

The well-known challenge to drawing the latter kind of conclusions, however, is that it is always hard to compare the minority populations between any two societies. Countries have distinct colonial and immigration histories and, though some are more similar than others, each country’s story turns out to be unique in the end, leading to a very specific minority-population composition (see Chapter 2). A first basic difficulty of comparison is the fact that some minorities have arrived in significant numbers only since 2000, while others have been widely present in some countries since the 1950s and 1960s. As a consequence, minority youth of the same age group might belong to different immigrant generations: while some were born and spent parts of their childhood abroad, others have parents or even grandparents who grew up in the country they live in; in other words, the strength of the ‘migration background’ can vary a lot. A second obvious fact and challenge for comparison is that the roots of minority members lie in very different countries of origin, thus providing rather diverse socioeconomic and sociocultural (p.63) starting points and conditions for the processes of integration. Both generation and country of origin are known to be associated with many integration outcomes, resulting in large variations between different groups even within a given destination country (Heath 2014). Thus, it is absolutely necessary to know how the two characteristics are distributed and to take them properly into account when trying to compare the fate of minority youth between destination countries.

In this chapter, we will use the Children of Immigrants Longitudinal Survey in Four European Countries (CILS4EU) data (Kalter et al. 2014) to explore how the population of adolescents is composed in terms of generational status and countries of origin in England, Germany, the Netherlands and Sweden. We can report a huge variance in these characteristics among the minority youth in each of the four countries, and we will assess how similar or dissimilar the countries are with respect to these mixtures. Given this basically descriptive picture, we will then explore the possibilities of and limits to achieving the mentioned general aims of a comparative approach and will discuss our strategies for pursuing our study of integration using a cross-country perspective.

3.2 The Generational and Ethnic Composition of Adolescents in the Four Countries

The first aim of this chapter is to give a comprehensive description of the ethnic diversity among the adolescents in the four countries studied in this book. We start by analysing the migrant experience of our respondents, or their families, usually captured by the concept of immigrant generation. Then we will look at the large varieties of origin countries in which the ancestral roots of those with a strong migration background can be found.

3.2.1 Generational status

The concept of generation is crucial for any study of immigrants’ integration. Its importance has been emphasised from the beginning of migration theory, most prominently in the writings of the Chicago School. The early relevance of the concept was based on the empirical observation that studies of the classical European immigrant groups revealed—in many aspects and indicators—a significant increase in assimilation to the American mainstream over generations, meaning from the immigrants themselves, to their children and their grandchildren. The notion of a ‘three-generation-assimilation-cycle’ (Price 1969: 204) is a telling expression of this idea. While much of the research activities since then have challenged the generality of this pattern—looking at specific aspects, other groups or other countries—the usefulness of the concept of generation as a descriptive category to study integration patterns has never been disputed. Even in countries like England, where, due to its colonial history, integration is traditionally studied more in terms of ethnic or racial minorities, relying on more (p.64) subjective measures of identity than on migration histories and countries of birth, the concept of generation has received more and more attention, not least for reasons of cross-country comparison (Waters 2014).

Basically, the concept of generation has been aptly described as capturing the ‘ancestral distance’ (Alba 1988: 213) from the point of immigration to the host society. The concept is especially clear-cut when looking at the intergenerational integration processes of groups that happened to immigrate within a particular narrow historical period, as has been the case, more or less, with the classical European immigrants to the USA. However, when migration flows extend over a longer period of time, and when there is a lot of replenishment, back-and-forth migration and intermarriage, the concept of immigrant generation can become fuzzier and more complex. Minority members around the same age might have very different ‘ancestral distances’ from their country of origin.

Accordingly, the usage of generational categories in migration research is also a little fuzzy and the common practice differs slightly between countries. As a consequence, some terms might be associated with somewhat different meanings. While, for example, in most Dutch studies the category of ‘second generation’ also includes children of intermarriages between immigrants and native-born Dutch without any migration background, the same children would almost always be treated within a separate category of ‘mixed’ in studies for England or Sweden.

In this book, we distinguish the ancestral distances from the point of immigration, and thus from the country of origin, using strictly comparative and relatively differentiated categories to capture theoretically relevant differences. The scheme we use can be understood as a more fine-grained explication of the concept of generation. We have to emphasise again that—in the case of our CILS4EU study—we are always talking about adolescents of the same age who are descendants of a range of immigration cohorts. So if we find differences between generational categories any interpretation in terms of ‘change over generations’ must be tentative.1 To avoid such associations—and the confusion arising from such different praxis in different countries—we prefer using direct and unambiguous labels for specific generational categories wherever possible.

When we look at the more precise family migration histories of 14- to 15-year-olds in the four country samples of the CILS4EU study, we find that the past immigration trends as described in Chapter 2 have indeed led to a rich variety of patterns (see Table 3.1). Among all students in the sample, 2,092 have experienced migration themselves, that is, they were not born in the survey country. Taking the concept of generation literally, this means that these adolescents would usually be called ‘first-generation’ immigrants.2 When weighting the data so that (p.65)

Table 3.1. Generational status of students in the CILS4EU samples

ENG

GER

NETH

SWE

TOTAL

N(nw)

%(w)

N(nw)

%(w)

N(nw)

%(w)

N(nw)

%(w)

N(nw)

Child foreign born

  Arrived age 11+

199

3.9

103

0.8

36

0.4

222

2.8

560

  Arrived age 6–10

180

2.8

144

1.6

61

0.5

224

2.6

609

  Arrived age 0–5

183

2.6

266

3.6

170

2.3

179

2.4

798

  Age of arrival not known

47

0.7

22

0.1

28

0.3

28

0.4

125

Child born in survey country

  Both parents foreign born

543

5.3

1,232

12.8

671

5.1

1,017

10.4

3,463

  One parent foreign born

    Child of transnational marriage

298

3.8

227

2.5

98

1.5

170

2.5

793

    Child of intermarriage (mixed)

225

5.2

336

6.8

293

6.2

371

8.0

1,225

  Both parents born in survey country

    Some migration background (2–4 gps foreign born)

352

6.1

182

3.7

118

2.3

237

5.5

889

    No (sign.) migration background (0–1 gps foreign born)

2,169

66.8

2,421

67.5

2,868

81.1

2,516

64.4

9,974

Missing information

119

2.9

80

0.9

20

0.1

61

1.2

280

4,315

100

5,013

100

4,363

100

5,025

100

18,716

Note: (nw) = not weighted; (w) = design weighted; gps = grandparents. The categories above the dotted line constitute ‘children of immigrants’ or, synonymously, those with a ‘strong migration background’.

the stratified design of the study is accounted for, this corresponds to 9.9% of all students in England, 6.2% in Germany, 3.5% in the Netherlands and 8.2% in Sweden.

As age at immigration is a key factor for many integration-related processes, it has proved to be fruitful to further differentiate between foreign-born children in many empirical studies. In some chapters of this book we distinguish between those children who arrived at age 11 or older, those who arrived between age 6 and age 10 and those who were only 5 or younger when they arrived in the survey (p.66) country. Roughly, these categories correspond to arriving after, during or before primary education, although the age boundaries differ somewhat between the four countries.3 Table 3.1 shows that the four countries not only differ with respect to the relative share of the first generation, but also with respect to the composition of the three subcategories within the first generation. While the sample members in England and Sweden are relatively evenly distributed over the three subcategories, most foreign-born adolescents in Germany and especially in the Netherlands fall into the category ‘arrived at age 5 or younger’. So if we look at certain outcome variables in the remainder of this book, it is important that these differences in age at arrival are kept in mind.

Among those adolescents who have not experienced migration themselves a large share are children of two foreign-born parents. This means they are the second generation in the narrow sense of the term. To distinguish this from the wider sense usage of the term, we often refer to it as the ‘pure’ second generation.4 In our study this group comprises 3,463 of the unweighted cases. In terms of design-weighted proportions this means 5.3% of adolescents in England, 12.8% in Germany, 5.1% in the Netherlands and 10.4% in Sweden.

Next, we find a considerable share of young people who have one native-born and one foreign-born parent—a further 2,018 cases in the unweighted total sample. In many studies and publications they are also subsumed under the second generation in the wider sense of the word (e.g. Portes & Rumbaut 2001: 23). For some questions, however, it will be theoretically important to distinguish these respondents from those with two foreign-born parents. The group can further be differentiated according to the exact migration background of the native-born half of the parental couple: (a) this parent can, although born in the survey country, have a strong migration background her or himself, for example, belong to the second generation. In this case the respondent in our sample is the child of what is often called a ‘transnational’ marriage;5 or (b) if the native-born parent has no (strong) migration background we simply speak of the respondent as being the child of a ‘mixed’ couple or ‘intermarriage’. Again, for many outcome variables that are of interest in the remainder of the book these subcategories can make a huge theoretical difference, so the distinction as a rule might be necessary.

(p.67) In all cases discussed so far, the child, our respondent, or at least one of his or her parents, is foreign born. In these cases we are talking about ‘children of immigrants’ in the literal meaning of the word. We also describe these respondents as having a ‘strong’ migration background—a terminology we already implicitly used in the last paragraph. As a rule, the term ‘minority’ is also used as synonym for this group. This does not mean, however, that all remaining young people, the ‘majority’ members, do not have any migration background at all. We find that 889 of the remaining respondents have at least two foreign-born grandparents. Often, these people are classified under the ‘third generation’.6 We describe them as having some migration background, although not a strong one. This leaves 9,974 unweighted cases in our sample whom we describe as having no significant migration background, meaning that only one or none of the four grandparents were born outside the survey country.7

Table 3.1 illustrates that ‘having a migration background’ can by no means be understood as a dichotomous characteristic, as it quite often appears in public and also academic debates, but comprises categories reflecting very different familial migration histories. These tend to be empirically correlated with variables that refer to theoretical concepts that might be important to understanding some outcomes of interest. For example, the generational status is likely to be related to destination-country language proficiency in the family, which in turn is likely to affect the language and educational outcomes of the children in our study.8 The chapters in the book will study whether the specific generational status, indicating exposure to the host country, indeed makes a difference for specific outcomes (see Chapter 1). Table 3.1 also illustrates that the distribution of generational statuses varies across countries; this might be an important, though certainly not the only, variable in understanding country differences.

3.2.2 Countries of origin

The students in our CILS4EU study differ not only with respect to their more fine-grained generational status but also in their geographical origins. In the following we will look at those 7,573 students who are literally ‘children of immigrants’, that (p.68) is, have a ‘strong migration background’ according to our terminology (Table 3.1). We discover that their ancestral roots can be found all over the world. If we identify the country of origin by looking at the country of birth of the grandparents,9 we end up with a total number of 172 different countries:10 121 different origin countries in England, 111 in Germany, 102 in the Netherlands and 118 in Sweden. Because of the distinctive migration flows to the different countries (see Chapter 2), the mix and the weights of origin countries differ greatly.

Table 3.2 shows the most frequent origin countries within all four destination countries as named by the respondents.11 The table lists all countries that either make up at least 1% of the children of immigrants over all four countries or at least 2% in one of the four countries (using unweighted frequencies). We sort these and all remaining countries into nine rough categories of world regions that largely follow geographical and regional principles, but occasionally combine these with some additional sociocultural aspects.12

The first larger category that we distinguish combines North, West and South European country origins, making up almost one thousand children of immigrants in our study. Table 3.2 shows that this category is an important one in each of the four destination countries, although the empirical relevance of specific origin countries differs between them. In each destination country there is one origin country in this category comprising over 2% of the children of immigrants. These are Ireland in the English sample, Germany in the Dutch sample and Finland in the Swedish sample—direct neighbours in each of the three cases. In the German sample the former guest-worker country Italy constitutes the largest group in this category.

The second larger category we distinguish are students of Eastern European origin. They constitute a large part of the minority population in German and Sweden. We find, however, that the mix within this category differs: while Russia and Poland are the most frequent origin countries in the German sample, Bosnia and Herzegovina as well as Kosovo and Albania prevail in the Swedish sample. Eastern European origin clearly plays a less important role in the other two survey countries. In England only students from Poland are a group of notable size in our data.

A similar logic of reading the table applies to the remaining categories and rows. The third larger category acknowledges the customary (and implicitly cultural) classification of North America, Australia and New Zealand into a (p.69)

Table 3.2. Origin countries of children of immigrants in the CILS4EU study

‘World regions’

EN

GE

NL

SW

Total

North, West or South Europe

182

284

150

345

961

  Finland

0

1

2

136

139

  Germany

30

48

41

119

  Ireland

55

1

1

1

58

  Italy

17

131

9

18

175

  Other

80

151

90

149

470

    (Austria 23, Belgium 28, Cyprus 14, Denmark 51, France 42, Gibraltar 1, Greece 66, Guernsey 1, Iceland 2, Luxembourg 3, Malta 3, Netherlands 23, Norway 35, Portugal 52, Spain 51, Sweden 3, Switzerland 15, UK 57)

Eastern Europe

95

666

53

506

1,320

  Bosnia & Herzegovina

0

36

5

127

168

  Kosovo/Albania

2

14

0

103

119

  Poland

42

166

11

67

286

  Russia

9

186

3

14

212

  Serbia

6

113

7

51

177

  Other

36

151

27

144

358

    (Belarus 3, Bulgaria 20, Croatia 48, Czech Republic 7, Czechoslovakia 3, Estonia 13, Former Yugoslavia 87, Hungary 15, Latvia 13, Lithuania 18, Moldova 3, Montenegro 16, Romania 57, Serbia and Montenegro 2 Slovakia 9, Slovenia 11, Ukraine 33)

Western, non-European

25

20

29

31

105

    (Australia 17, Bermuda 1, Canada 21, Israel 3, New Zealand 14, USA 49)

Caribbean

125

8

253

9

395

  Jamaica

90

0

0

1

91

  Suriname

0

0

165

0

165

  Dutch Antilles

0

0

36

0

36

  Other

35

8

52

8

103

    (Antigua and Barbuda 1, Aruba 26, Barbados 5, Bonaire, Sint Eustatius 1, Caribbean (unspecified) 2, Cuba 11, Curaçao 8, Dominica 1, Dominican Republic 15, Grenada 6, Guyana 8, Saint Martin 1, St Kitts and Nevis 4, St Vincent and the Grenadines 5, Trinidad and Tobago 8, Turks and Caicos Islands 1)

Latin America and Pacific

38

47

16

92

193

    (South America (unspecified) 2, Americas (unspecified) 26, Argentina 6, Bolivia 8, Brazil 30, Chile 44, Colombia 27, Ecuador 10, Fiji 1, Guatemala 1, Honduras 1, Mexico 7, Nicaragua 1, Panama 1, Paraguay 2, Peru 15, Tonga 1, Uruguay 6, Venezuela 4)

MENA+

382

1,150

649

791

2,972

  Afghanistan

16

28

30

31

105

  Iran

4

21

10

77

112

  Iraq

7

43

34

226

310

  Kazakhstan

0

64

1

0

65

  Lebanon

2

54

5

95

156

  Morocco

4

34

248

21

307

  Pakistan

304

14

9

5

332

   (p.70) Syria

2

13

5

88

108

  Turkey

11

826

269

133

1,239

  Other

32

53

38

115

238

    (Algeria 23, Arabian Country (unspecified) 1, Armenia 11, Azerbaijan 10, Bahrain 1, Egypt 29, Georgia 6, Jordan 5, Kurdistan 45, Kuwait 9, Kyrgyzstan 1, Libya 9, Palestine 41, Saudi Arabia 4, Tunisia 33, Turkmenistan 1, United Arab Emirates 3, Uzbekistan 4, Yemen 2)

Sub-Saharan Africa

308

73

86

231

698

  Ghana

45

10

9

6

70

  Kenya

40

4

0

2

46

  Nigeria

63

5

4

5

77

  Somalia

38

4

12

112

166

  Other

122

50

61

106

339

    (Africa (unspecified) 27, Angola 18, Benin 2, Burundi 8, Cameroon 5, Cape Verde 20, Congo 29, Côte d’Ivoire 7, Djibouti 5, Eritrea 38, Ethiopia 31, Gambia 18, Guinea 2, Liberia 5, Malawi 3, Mali 1, Mozambique 1, Rwanda 1, Senegal 7, Seychelles 2, Sierra Leone 7, South Africa 33, Sudan 6, Tanzania 10, Togo 6, Uganda 15, Zambia 6, Zimbabwe 26)

South Asia

347

29

66

51

493

  Bangladesh

47

2

0

18

67

  India

220

8

13

18

259

  Indonesia

3

2

43

2

50

  Sri Lanka

46

16

6

8

76

  Other

31

1

4

5

41

    (Kashmir 2, Malaysia 13, Mauritius 15, Nepal 10, Papua New Guinea 1)

Total South-East and East Asia

93

51

55

148

347

  China + Hong Kong

46

11

25

24

106

  Other

47

40

30

124

241

    (Cambodia 3, Japan 10, Lao 1, Myanmar 5, Mongolia 2, Philippines 51, Rep Korea 26, Singapore 7, Vietnam 66, Thailand 70)

Not known

80

2

0

7

89

Total

1,675

2,330

1,357

2,211

7,573

Note: Country names as given by the respondents. Numbers that make up for at least 1% overall or 2% in one country are underlined. In brackets and in italic: list of countries that can be found in the residual category of ‘other’, number of cases in total (all four countries).

common ‘Western’ group (which all share predominant although not exclusive European heritages). Then we distinguish students within an overall category for Latin America; for the combination of North African, Middle Eastern and Central Asian countries (which we call MENA+); for sub-Saharan Africa; South Asia; South-East and East Asia; and finally some minority students who cannot be classified due to missing values. From these categories we can note the relatively (p.71) high density of immigrants with Surinamese, Turkish and Moroccan origins in the Netherlands; Pakistani and Indian in England; Turkish in Germany; and Iraqi and Somali in Sweden. Summarising all major groups, those with origins in MENA+ countries dominate our sample.

3.2.3 Comparing minority compositions among survey countries

Overall, Table 3.2 clearly demonstrates that in each of the four survey countries one finds a huge diversity in the group of young people with a migration background and that this composition largely differs among our survey countries. One way to express the differences between any two of these survey countries is to compute the index of dissimilarity.13 The index ranges from 0 to 1, reflecting perfect similarity and absolute dissimilarity, respectively. A straightforward interpretation of the index is that the value represents the percentage of observations in one of the countries which would have to be moved to another category in order to achieve perfect similarity to the marginal distribution in the second country.

Table 3.3 gives the dissimilarity between any two of the four survey countries expressed by this index. We find that the differences in the composition of the children of immigrants between the four survey countries are very large. Comparing England to Germany, for example, there is an extremely high dissimilarity value of 0.80. In general, England is the country that differs most from the other countries. Germany, on the other hand, is a little more similar both to the Netherlands and Sweden, but even here we find values of 0.58 and 0.61.

Note that the four survey countries not only differ with respect to the origins of their population with a strong migration background but also in the basic patterns of diversity. In Germany, at one extreme, the student sample with a migration background is clearly dominated by one group, those of Turkish origin, which makes up over a third of the children with a strong migration background in the sample. Adding the four next biggest groups (Russia, Poland, Italy, Serbia) already accounts for 61%. In the Netherlands three relatively large groups (Turkey, Morocco, Suriname) constitute almost half of the minority population, and adding the next two would cover nearly 57%. At the other extreme, in Sweden the five biggest groups (Iraq, Finland, Turkey, Bosnia and Herzegovina, Somalia) cover only one third of the whole minority sample, meaning that there is a more even distribution over more groups, that is, greater diversity. England stands somewhere in-between the Netherlands and Sweden, with two major (Pakistan, India), but no really dominating, groups. Here, the five largest groups cover roughly 48% of the minority sample. A more sophisticated way to express these patterns (p.72)

Table 3.3. Dissimilarity between the four survey countries with respect to origin-country composition among children of immigrants

England

Germany

Netherlands

Germany

0.80

Netherlands

0.78

0.58

Sweden

0.75

0.61

0.67

Note: Data not weighted.

would be to take the common diversity index based on the Herfindahl measure.14 According to this, the diversity at the origin-country level is 0.86 in Germany, 0.91 in the Netherlands, 0.93 in England and 0.96 in Sweden.15 A straightforward interpretation of the index value is that it represents the probability that two randomly chosen individuals, in our case two minority students, belong to different groups; so, for example, the likelihood that two minority students in Sweden have their roots in the same origin country is only 4%.

3.2.4 Combining origin countries into larger categories

An obvious technical problem for the empirical analysis of integration processes is that the great diversity of minority groups means that in available survey data sets many are represented only in small, some even in very small, numbers. In spite of the sample sizes and the oversampling strategy, the CILS4EU data is no exception (see Table 3.2). As a consequence, differences between fine-grained groups are often not so precisely estimated (i.e. regression coefficients in statistical models are surrounded by large standard errors). It is therefore necessary and common practice to combine individual origin countries into larger, more comprehensive groups. This is also what we will do in this book.

While it is sometimes adequate and most efficient to tailor the grouping to the specific analyses and their aims, the comparison of minority groups over different aspects of integration is one of the major aims of this book. We therefore develop a default scheme that serves, in a sense, as the common denominator for the analyses in the chapters of this book, although individual chapters may deviate slightly from it, either by distinguishing finer subcategories or by combining the default categories into larger ones, if the gain for the research question and analysis outweighs the loss of comparability to other chapters.

The starting point of our scheme is the distinction between the nine rough ‘world regions’ in Table 3.2. As some of these nine categories still have rather small numbers, we merge those of ‘South Asian’ and ‘South-East and East Asian’ origin into one category for Asia, and we put those of ‘Western’ origin and those of ‘Latin (p.73) America and Pacific’ origin into a residual category of ‘others’, together with those whose origin is ‘not known’. As there are rather few ‘Caribbeans’ in Germany and Sweden, we also put them into the residual category in these two countries. We thus end up with a classification of the adolescents into seven broader categories altogether, including the large reference group with no strong migration background.

In the next step, we single out the most frequent individual origin countries in a given survey country. As a rule of thumb, we do this whenever the number of cases is larger than 100. This logic leads to the categories and frequencies outlined in Table 3.4. Note that the larger categories now refer to origin countries apart from the ones singled out.

We have to stress again that results based on this scheme have to be interpreted cautiously, especially when referring to the larger world regions and when comparing the four receiving survey countries. While, for example, Eastern Europe contains a considerable number of children of Polish origin in England, the Netherlands and Sweden, this does not hold for Germany as the group is singled out here. Moreover, as already shown in Table 3.2, the composition of specific origin countries within the broader categories can be very different between survey countries.

Table 3.4. Scheme of origin countries serving as the default for the multivariate analyses

Number of cases (not weighted)

ENG

GER

NETH

SWE

Total

Reference group (‘majority’)

2,640

2,683

3,006

2,814

11,143

North/West/South Europe

182

153

150

209

694

Finland

136

136

Italy

131

131

Eastern Europe

95

201

53

276

625

Bosnia & Herzegovina

127

127

Kosovo/Albania

103

103

Poland

166

166

Russia

186

186

Serbia

113

113

Caribbean

125

88

213

Suriname

165

165

MENA+

78

324

132

432

966

Iraq

226

226

Morocco

248

248

Pakistan

304

304

Turkey

826

269

133

1,228

Sub-Saharan Africa

308

73

86

118

585

Somalia

113

113

Asia

220

80

121

199

620

India

220

220

Other/not known

143

77

45

139

404

Total

4,315

5,013

4,363

5,025

18,716

(p.74) Following Table 3.4, there is only one individual origin country that we find in convenient numbers in more than two survey countries, namely Turkey in Germany, the Netherlands and Sweden. But even here the group can be composed very differently in terms of basic, theoretically relevant characteristics. If we look, for example, at religion, ethnic identity and language spoken at home we find some notable within-group heterogeneity and very different distributions for Turkish students in Sweden compared to Germany and the Netherlands (Table 3.5).

The upper part of Table 3.5 shows the religious within-group heterogeneity among youth of Turkish origin. Using the Herfindahl-based diversity index as a summary measure, it turns out that this is lowest in the Netherlands (0.08), and also

Table 3.5. Within-group diversity and between-survey-country dissimilarity for students of Turkish origin

Germany

Netherlands

Sweden

What is your religion? (%)

No religion

3.9

2.3

0.8

Christianity

3.6

1.5

50.4

Hinduism

0.1

0.4

Islam

90.2

95.8

47.9

Other religion

2.1

0.8

Diversity

0.18

0.08

0.52

Dissimilarity

GE–NL: 0.06;

NL–SW: 0.50;

SW–GE: 0.47

Which of the following groups do you feel you belong to? (%)

No group (other than survey country majority)

24.4

15.5

15.8

Other group (first answer)

    Turkey

66.7

78.8

30.1

    Kurdistan

3.1

2.6

10.5

    Assyria or Syria (or ‘Aramaic country’)

0.8

0.8

42.1

    Other (incl. no answer)

5.1

2.3

1.5

Diversity (if other group)

0.22

0.13

0.61

Dissimilarity (if other group)

GE–NL: 0.05;

NL–SW: 0.58;

SW–GE: 0.57;

Is there a language other than German/Dutch/Swedish spoken at your home? (%)

No

4.2

3.3

1.5

Yes—which language is this? (first answer)

    Turkish

84.8

89.6

42.5

    Kurdish

8.0

0.8

10.6

    Aramaic

1.1

0.8

34.1

    Other (incl. no answer)

1.9

5.6

11.3

Diversity (if other language)

0.24

0.14

0.67

Dissimilarity (if other language)

GE–NL: 0.07;

NL–SW: 0.50;

SW–GE: 0.43

Note: Percentages not weighted.

(p.75) relatively low in Germany (0.18), but rather high in Sweden (0.52). Remarkably, half of the youth of Turkish origin in Sweden are Christians while the vast majority of this group in Germany and the Netherlands are Muslim. Comparing the religious mix between countries we see that it is very similar between Germany and the Netherlands (dissimilarity index (DI) = 0.06), while it is very dissimilar between any of these two countries and Sweden (DI = 0.47, resp. 0.50).

Looking at the subjective ethnic identity (middle part of Table 3.5) and at the language spoken at home (lower part of Table 3.5) one finds that the exceptional role of Sweden is certainly related to the fact that half of the adolescents in the Swedish sample, who feel that they belong to a non-indigenous group, identify themselves as Assyrian or Syrian. Over a third of those who speak a language other than Swedish at home name ‘Aramaic’ as this language. So it is clear that when talking about minority youth with Turkish background in Sweden we are largely talking about a very specific sub-minority that is only of marginal importance in the other survey countries. And even if Germany and the Netherlands are relatively similar to each other with respect to the composition of ethnic identities and language spoken at home, there are differences in nuances, for example the subgroup of Kurds is found somewhat more frequently in Germany.

3.2.5 Confounding origin countries and generational status

In the course of this chapter we have argued that the differential composition of the children of immigrants with respect to their more precise generational status as well as to specific origin groups has to be properly accounted for in order to achieve the aims of comparative empirical analyses. A further challenge is that both of these aspects, origin and generational status, are often heavily confounded due to the unique migration history of each of our destination countries. This is also reflected in our data. Figure 3.1 shows, for example, the distribution of generational status for the three largest groups in each of the four survey countries.

We find notable differences in the generational composition of the three largest English groups. The Jamaicans, for example, consist of a relatively high share of children from mixed marriages. Indian youths are less likely to be foreign born than Jamaican or Pakistani. In Germany, having a Russian background relatively often means being foreign born, while having a Turkish background means belonging to the ‘pure’ second generation in over two thirds of the cases. Interestingly, in the Netherlands, Surinamese adolescents make up the highest share of foreign-born and mixed-marriage children. Sweden, finally, provides the most extreme compositional differences between its typical origin groups: while over 60% of all adolescents with a Finnish background are children of mixed marriages, more than 60% of those with Iraqi origins were foreign born. This reflects the fact that migration waves follow different historical events, such as the labour migration of Finns in the 1960s and 1970s, when the Swedish industry needed manpower, and the refugee waves from the Iraq War 2003–11 (see Chapter 2). (p.76)

Dealing with Diverse Diversities: Defining and Comparing Minority Groups

Figure 3.1 Generational status types among largest origin groups in each survey country.

Note: Design weighted.

This means that the more precise migration background is a natural candidate to explain, at least partly, the group differences in integration that we observe throughout this book. There could not be a more telling example to illustrate this point than the contrast between the Finnish and Iraqi groups in Sweden. If we were to find systematic differences in certain outcome variables, for example Swedish language proficiency, between Finnish and Iraqi youths one should not be surprised: while members of the first group most likely have one Swedish parent, members of the second group were born and spent parts of their childhood in a country that was culturally and linguistically very different.

Note that Figure 3.1 also allows us to compare the migration background composition of Turkish adolescents among Germany, the Netherlands and Sweden. Even though the compositions are more similar to each other than to those of other groups, they nevertheless include some minor nuances; young Turks in Germany, for instance, are more likely to have mixed parentage than those in Sweden. So, in addition to the findings in Table 3.5, this is a further hint and warning that talking about a nominally similar group in two different host countries can possibly mean talking about people with different characteristics. Accordingly, it is always likely that generational status differences account for differences in (nominally same) groups between countries.

While Figure 3.1 thus illustrates the need to account for generational status when group differences are the phenomenon of interest in research, a similar story holds the other way round. The impact of generational status has always (p.77) been among the key questions in integration research. Whether, for example, the second generation integrate to a greater extent than the first, whether children of transnational marriages differ from the ‘pure’ second generation or whether children of mixed marriages differ at all from the majority are crucial questions in many applications. With respect to these kinds of questions, one has to account for the fact that different generational status types comprise very different groups. This can be seen in Table 3.6, which—as the flipside of Figure 3.1—reports the top five origin countries separately for the four generation categories in each survey country.

In England we see that in three of the four categories a Pakistani background is the most frequent, followed by an Indian background. Among the children of mixed marriages, however, the Irish are the largest group and Pakistani or Indian backgrounds are less common. In Germany, Turkey dominates all of the subcategories, most often followed by Russia, then Poland. In the Netherlands the three largest groups overall, that is, with a Turkish, Moroccan or Surinamese background, have different weights in different subcategories, while other groups

Table 3.6. Top five countries of origin by type of generational status

England

%

Germany

%

Netherlands

%

Sweden

%

Foreign born

Pakistan

11

Turkey

19

Turkey

13

Iraq

22

India

7

Russian Fed.

18

Afghanistan

9

Somalia

6

Poland

6

Poland

8

Iraq

8

Poland

4

China

5

Kazakhstan

7

Suriname

8

Bosnia & Herz.

4

Sri Lanka

5

Iraq

7

Morocco

6

Kosovo-Albania

4

Native born, both parents foreign born

Pakistan

26

Turkey

47

Morocco

31

Turkey

11

India

19

Russian Fed.

6

Turkey

28

Bosnia & Herz.

9

Bangladesh

6

Poland

6

Suriname

15

Lebanon

8

Kenya

6

Serbia

6

China

3

Iraq

7

Nigeria

4

Italy

4

Cape Verde

2

Syria

7

Child of transnational marriage

Pakistan

32

Turkey

45

Turkey

35

Finland

21

India

26

Russian Fed.

12

Indonesia

13

Turkey

8

Jamaica

8

Poland

8

Suriname

12

Germany

8

Ireland

6

Greece

5

Morocco

8

Denmark

5

Nigeria

5

Serbia

4

Belgium

3

Bosnia & Herz.

4

Child of intermarriage

Ireland

13

Turkey

12

Germany

12

Finland

20

Jamaica

10

Italy

11

Suriname

10

Denmark

7

Germany

7

Poland

10

Indonesia

8

Poland

5

Pakistan

7

Austria

4

Morocco

4

Norway

4

India

5

Russian Fed.

3

Belgium

4

Thailand

4

Note: Percentages (not weighted) among non-missing origin countries.

(p.78) occasionally have a relatively high ranking in the table. Again, Sweden provides the most extreme view: while among the foreign born those from Iraq clearly dominate, Turkish origin is the most frequent background in the pure second generation and Finnish origin is by far the leading group among the children of transnational marriages and of intermarriages.

The strong confounding of origin countries and generational status among children of immigrants, as illustrated in Figure 3.1 and Table 3.6, has, therefore, to be taken into account when drawing conclusions about the impact of either of these characteristics. This means that when analysing origin group effects we should, next to other potential confounders, statistically control for generational-status effects and vice versa.

3.3 Dealing with Diversity in Comparative Analyses: Strategies and Limitations

Section 3.2 demonstrated the enormous diversity of minority adolescents in terms of countries of origin within each of the survey countries, as well as interesting differences between the four countries. These diversities are likely to go hand in hand with systematic differences in characteristics that are theoretically suspected or empirically known to affect the integration processes. Examples of such characteristics are socioeconomic background, language proficiency, general cultural distances and many more. Such systematic differences have consequences for the comparative analysis of integration processes.

In one important strand of comparative empirical research the heterogeneous compositions of the minority populations are basically seen and treated as a nuisance. Here, the main interest of the analyses is on the effect of destination country or its institutional arrangements (Z) on integration outcomes (Y) per se; for example, an important question in integration research is whether countries with early educational tracking, such as Germany, generate larger educational disadvantages for minority students.

As adumbrated, the basic problem for these kinds of analyses is that host-country differences might partly or even mainly result from different compositions of the immigrant population and thus from very different mixes of individuals’ characteristics. While some of these characteristics, for example, the educational and occupational attainment of the parents, can sometimes be directly measured and thus be controlled in the statistical models, others, such as general motivation and other personality traits, are usually unobserved in the available data sets. As a consequence, it is common practice to control for group membership in the hope that doing so will capture most of these important characteristics and lead to less biased estimates of the true destination-country effect.

Technically speaking, the group variable then serves as a proxy variable for the unobserved characteristics and the idea is basically equivalent to (p.79) a measurement-error model. The situation is likely to appear not only when researchers are interested in destination-country effects but in the effect of any factor Z on a given integration outcome Y. The general problem is that some other relevant factors accounting for different outcomes Y are unobserved (U) but have an impact on the factor(s) Z of interest. In the language of causal analysis one would speak of U being a ‘confounder’ of Z (Morgan & Winship 2007; Pearl 2012). Next to the direct impact of Z, there is then an additional so-called back-door path, so that the estimate of the effect of Z will be biased. This could only be avoided, that is, the path could only be ‘blocked’, if one was able to control for U. If this is not possible, researchers try to reduce the bias by controlling for G instead. The situation can be sketched like this:

Dealing with Diverse Diversities: Defining and Comparing Minority Groups

Indeed, the bias can be reduced if G is controlled for as a proxy. And the better the proxy, the more the bias is reduced (Pearl 2012). However, it is important to see that controlling for a proxy cannot really block the back-door path and thus really solve the problem completely: only a perfect correlation between G and U would lead to an unbiased estimate for the effect of Z. Simulating the situation, one can show that even very strong correlations between group membership and the unobserved variable(s) lead only to relatively modest reductions of the bias. However, the relation between theoretically relevant variables and ethnic origin group boundaries is usually fuzzy and fluent (Brubaker 2002; Wimmer 2009) and thus far from perfect. So, attempts to control for groups G in order to get the effects of other variables Z right, must be seen as a gradual improvement, but not as overcoming the basic problem.

There is, however, an alternative perspective and strategy of analysis: here, the heterogeneous compositions of the minority populations—and the destination-country differences between these compositions—are not treated as phenomena of nuisance but rather as phenomena of interest. These types of empirical studies start with the observation that there are differences in integration between ethnic groups (often including a reference majority population with no migration background), that is, that group membership (G) is associated with a certain outcome variable (Y).

While this is a descriptive fact, the aim is then usually to explain why these differences exist; or in other words, the group effects are treated as the explanan-dum in the research process. The task is to look for theoretical mechanisms that could possibly account for the relationship between group membership and the outcome of interest. These mechanisms point to theoretical concepts and variables (X) that might mediate or moderate the empirically observed group differences. Researchers introduce these variables (X) into the statistical models and check how far they account for the empirical relations.

(p.80) In the extreme case, the former effect of G vanishes upon taking X statistically into account, which would mean that the group differences are mediated by the variable(s) X. This would give support for the assumed mechanisms related to X. In the less extreme case the effect of G is reduced, but there is still a notable direct effect of G left. This means that X can contribute to explaining the statistical relation, and thus the theoretical mechanisms assumed are likely to be important for producing the group differences, but that there must be other unmeasured variable(s) U that affect the outcome Y, which are also statistically related to G, and thus additional theoretical mechanisms which are missing in the explanation of the group differences.

Dealing with Diverse Diversities: Defining and Comparing Minority Groups

An example for this strategy can be found in the analyses in Heath & Cheung (2007). The book looks at differences in labour market outcomes (Y) among common ethnic groups (G) in a number of countries, and studies how far these differences are due to differences in education (X). As a rule, education is responsible for a large element of group difference in labour market success, but there are still considerable direct effects of group membership left even when taking education into account. This leads to the conclusion that for a more exhaustive explanation of group differences additional mechanisms seem relevant too, for example, unmeasured skills, social capital, ethnic discrimination and so on.

Dealing with Diverse Diversities: Defining and Comparing Minority Groups

The example of the analyses in Heath & Cheung (2007) also shows how the comparative perspective can be integrated into this logic. Just as group differences, country (C) differences should be seen as descriptive facts and potentially interesting explananda. As long as the variables (X) related to available explanations—no matter whether they refer to the micro-level or to institutional characteristics—do not lead the differences between countries to disappear, another element (not yet fully understood) must be present in the conditions and the underlying processes. Within this logic of analysis, a specific interest often is whether the strength of certain mechanisms differs for different countries; statistically, this means it is important to look at interaction effects between countries and the effects of the mechanism-related variables X.

Treating group differences as phenomena of interest rather than of nuisance explicitly avoids any essentialising of ethnic groups, a not uncommon research practice that has rightly been criticised (Brubaker 2002; Wimmer 2009). It also eases concerns and discussions about ‘right’ or ‘wrong’ ways of defining ethnic (p.81) groups and combining single groups into larger categories, as it allows and pleads for a pragmatic approach. The research endeavour is about finding the underlying mechanisms: group effects—like country effects—serve as the motivation and as a litmus test for detecting such mechanisms. Starting with group and country differences is, in a sense, setting the bar to a certain level to see whether the theories are encompassing and general enough to jump over it. If not, there is further theoretical work to do.

A key contribution of the CILS4EU data to improving our understanding of integration processes among adolescents in the diverse European societies of today is that it contains many variables that try to measure key concepts related to assumed theoretical mechanisms. For many of them this has been undertaken, for the first time, in a strictly comparative way and in a more comprehensive way than usual, while some measures even pertain to concepts that are completely new to the integration research data infrastructure. Starting with group and country differences as phenomena of interest, a basic aim of the thematic chapters that follow in the rest of the book is thus to provide better tests of key explanatory mechanisms by introducing adequate measures (‘more X’) for concepts that often remained unmeasured (‘less U’) in previous comparative research.

References

Bibliography references:

Alba, R. (1988), ‘Cohorts and the Dynamics of Ethnic Change’, in M. W. Riley (ed.), Social Structures and Human Lives (Newbury Park CA, Sage Publications), 211–28.

Brubaker, R. (2002), ‘Ethnicity without Groups’, European Journal of Sociology, 43(2): 163–89.

Dollmann, J., Jacob, K. & Kalter, F. (2014), ‘Examining the Diversity of Youth in Europe: A Classification of Generations and Ethnic Origins using CILS4EU Data (Technical Report)’, Mannheimer Zentrum für Europäische Sozialforschung: Arbeitspapiere Nr. 156, Mannheim, Universität Mannheim.

Heath, A. (2014), ‘Introduction: Patterns of Generational Change: Convergent, Reactive or Emergent?’, Ethnic and Racial Studies, 37: 1–9.

Heath, A. & Cheung, S.Y. (eds) (2007), Unequal Chances: Ethnic Minorities in Western Labour Markets (Oxford, Oxford University Press for the British Academy).

Heath, A., Schneider, S. & Butt, S. (2016), ‘Developing a Measure of Socio-Cultural Origins for the European Social Survey’, GESIS Papers 2016/16, accessed 6 February 2018, http://nbn-resolving.de/urn:nbn:de:0168-ssoar-49503-6.

Kalter, F., Heath, A., Hewstone, M., Jonsson, J.O., Kalmijn, M., Kogan, I. & van Tubergen, F. (2014), ‘Children of Immigrants Longitudinal Survey in Four European Countries (CILS4EU)’. ZA5353 Data File Version 1.1.0, Cologne, GESIS Data Archive. DOI: 10.4232/cils4eu.5353.1.1.0.

Morgan, S.L. & Winship, C. (2007), Counterfactuals and Causal Inference: Methods and Principles for Social Research (London, Cambridge University Press).

Pearl, J. (2012), ‘On Measurement Bias in Causal Inference’, in P. Grundwald & P. Spirtes (eds), Proceedings of the Twenty-Sixth Conference on Uncertainty in Artificial Intelligence (Corvallis OR, AUAI), 425–32.

(p.82) Portes, A. & Rumbaut, R. G. (2001), Legacies: The Story of the Immigrant Second Generation (Oakland CA, University of California Press).

Price, C. (1969), ‘The Study of Assimilation’, in J.A. Jackson (ed.), Migration (London, Cambridge University Press), 181–237.

Rumbaut, R. (2004), ‘Ages, Life Stages, and Generational Cohorts: Decomposing the Immigrant First and Second Generations in the United States’, International Migration Review, 38(3): 1160–205.

Waters, M.C. (2014), ‘Defining Difference: The Role of Immigrant Generation and Race in American and British Immigration Studies’, Ethnic and Racial Studies, 37(1): 10–26.

Wimmer, A. (2009), ‘Herder’s Heritage and the Boundary-Making Approach: Studying Ethnicity in Immigrant Societies’, Sociological Theory, 27(3): 244–70.

Notes:

(1) True change over generations can be studied with the CILS4EU data by comparing the outcome variables of the adolescents with those of their parents; see Chapters 10 and 12 for examples.

(2) Note that speaking of a first generation in this sense might also include a few children who might have been born abroad only ‘by accident’ during a shorter stay of their mother abroad; in the total sample we find n = 68 children who were born abroad, but nevertheless have two parents who are native born. This could include, for example, the children of parents in the British forces, stationed in Germany after World War II.

(3) These categories are sometimes referred to as the ‘1.25th generation’, the ‘ 1.5th generation’ and the ‘1.75th generation’ (Rumbaut 2004: 1167).

(4) Relating to the numerical approach mentioned in footnote 3, one could also speak of the ‘2.0th generation’.

(5) This configuration even breaks down into further subconfigurations, which are—as a rule—not further distinguished in the remainder of the book. If the native-born parent is 2.0th generation, as defined in footnote 4, the child, i.e. our respondent, could be categorised, as is sometimes done, as belonging to the ‘2.5th generation’. The case might be more complicated if the parent her or himself belongs to the thus defined 2.5th generation, in which case it would make sense to subsume the respondent under something like a ‘2.75th generation’. In both subcases the native-born parent has at least one foreign-born parent her or himself.

(6) According to whether four, three or two grandparents are foreign born, one could further differentiate between a 3.0th, 3.25th or 3.5th generation.

(7) To be sure, any classification like this suffers from potential missing information on the country of birth of relevant ancestors. While we made some grounded imputations for certain patterns of missing values (Dollmann et al. 2014), there are some cases left (n = 280) that cannot be plausibly subsumed under any specific category. Note, however, that it might be possible to classify some of these cases when using broader categories. For example, those children (n = 23) who state that their parents are foreign born, but neglected to give the information on their own country of birth belong either to the first or the second generation. Likewise, children with native-born parents and missing information on the grandparents definitely do not belong to the first or the second generation.

(8) This example shows that further to the additional differentiations discussed in the text, one could think of capturing the migration background in even more detail, e.g. by distinguishing according to the parents’ age at migration.

(9) If the grandparents were born in different countries we take the most frequent one among the four. In case of parity we give priority to the maternal line. If values for grandparents are missing we take the countries of birth of the parents—with similar rules—as a proxy, or—as the last means—the country of birth of the child (see Dollmann et al. 2014 for the exact algorithm).

(10) Note that some respondents give ‘countries’ that do not exist in the legal sense of the word, like ‘Kurdistan’ or that no longer exist, like ‘Former Yugoslavia’. As we cannot unambiguously identify this with an existing country we treat them as different categories.

(11) See the remarks in footnote 10.

(12) This classification broadly follows that proposed for classifying cultural and ethnic groups in the European Social Survey (Heath et al. 2016).

(13) The index of dissimilarity is defined as Dealing with Diverse Diversities: Defining and Comparing Minority Groupswith k being the number of distinguished categories (here: origin countries), Ai and Bi being the number of individuals from sets (here: destination countries) A and B falling into category i, and A and B being the overall number of individuals in set A and set B.

(14) The Herfindahl-based diversity index is computed as Dealing with Diverse Diversities: Defining and Comparing Minority Groupswith q being the number of groups and pi being the proportion of individuals belonging to group i.

(15) Note that this is the diversity within the (unweighted) population of children of immigrants and not in the total sample of adolescents.