Studying Integration from Adolescence to Early Adulthood: Design, Content, and Research Potential of the CILS4EU-DE Data

Abstract

This data brief describes the German part and extension of the Children of Immigrants Longitudinal Survey in Four European Countries (CILS4EU-DE). The aim of CILS4EU-DE is to provide a data-infrastructure for theory-driven and evidence-based research on various aspects of integration among young people with and without an immigrant background in Germany. Key features of the survey are (i) its large and representative sample, with an oversampling of respondents with an immigrant background; (ii) the inclusion of a refreshment sample that maintains a convenient size, composition, and representativeness of the sample also in later waves; (iii) the focus on a young target population, allowing to investigate integration processes through a formative period of life; (iv) the longitudinal design with so far seven waves; (v) the focus on a wide variety of integration outcomes, accompanied by the use of innovative measures allowing to apply cutting-edge methods in integration research; and, not least (vi) the incorporation of CILS4EU-DE in the international CILS4EU-project, facilitating to study integration processes also in a comparative perspective in England, The Netherlands, and Sweden.

Ideas and Aims of CILS4EU-DE

The integration of immigrants and their children has become a key issue in almost all European societies, among them Germany; the topic will continue to challenge the social sciences for many years to come. In the light of the sometimes very emotionalized and politicized societal debates, the need for theory-driven and evidence-based research could hardly be more obvious. A basic requirement for respective contributions is strong, reliable, and representative data, which are notoriously hard to obtain. The German extension of the Children of Immigrants Longitudinal Survey in Four European Countries (CILS4EU-DE) aims to strengthen the data infrastructure for integration research, especially when it comes to the pursuit of questions that focus on younger people, refer to a wider range of integration-related aspects, and require longitudinal measures and analyses.

In recent years the number of elaborate quantitative empirical studies in the field has increased tremendously, and there has been great progress in understanding processes of integration. Empirical research has largely benefited from an increased availability of microdata from the official statistics and from a number of large-scale social science surveys, such as the German Socio-Economic Panel (GSOEP), which provide a sufficiently large sample size and oversampling strategies to study minorities. The major strength of general data sets like these lies in the opportunity to study an adult immigrant population, particularly those who experienced migration themselves. The data sets soon reach their limits, however, when researchers are interested in processes of integration among children and youth. Therefore, CILS4EU-DE focusses on immigrant offspring, which is a consistently growing group in almost all Western societies. In Germany, about one-third of all students have an immigrant background (Statistisches Bundesamt, 2017).

However, it is not only the increasing number that makes this group an important population to study. Adolescence and early adulthood are formative years, and hence, pathways of integration are especially telling and important within this period. It is the time when many crucial decisions relevant for the further life situation and life chances are made. For example, there are educational choices, transitions into the labour market, and first romantic relationships, to name only a few decisive events. Additionally, adolescence is a period in life when individuals experiment with a myriad of new social roles, question their identities, develop their religious beliefs, voice their own opinions about various topics, and so on.

The aim of CILS4EU-DE is to provide data on a broad range of these processes by surveying young people with and without a migration background during this important period of life over time. The first interviews were conducted when respondents were around the age of 14. Currently, they are about 22, and the plan is to follow them further until they will have reached age 26. Several thematic modules focus on aspects becoming particularly relevant at specific ages, like transitions in the education system, into the labour market, political and social participation, the formation of intimate relationships, marriages, and childbearing. While other large-scale educational studies like PISA (Programme for International Student Assessment), TIMSS (Trends in International Mathematics and Science Study), PIRLS (Progress in International Reading Literacy Study) (Klieme et al., 2010; Martin and Mullis, 2012), or the German National Educational Panel Study (NEPS; Blossfeld et al., 2011) already provide quite extended information for the dimension of structural integration, CILS4EU-DE has a very broad thematic framework and puts much emphasis also on social, cognitive–cultural, and emotional–cultural aspects of integration. A large set of key indicators on these aspects of integration is repeatedly measured over the years. All this makes it possible to contribute to answering open questions around causality in integration research, for example, on the interplay between the structural, social, and cultural dimension.

The data comprise over 5,000 respondents in the first wave of the panel. Moderate attrition and the inclusion of a refreshment sample in the sixth wave guarantee a sufficient number of cases also in later waves. By oversampling respondents with an immigrant background in Wave 1 and Wave 6, the survey ends up with about a half of the respondents possessing some form of migration background (up to the third generation). We are not concentrating on specific immigrant groups but rather aim at a representative sample of the specific age cohort. Both aspects, the oversampling of respondents with an immigrant background and the achievement of a representative sample, were the two guiding principles during the sample selection of the original and the refreshment samples.

The study contains a number of additional methodological features that are helpful to increase our knowledge about integration processes. Parental interviews are available and asking the same questions in the student and the parental survey allows to investigate true intergenerational change. The survey provides sociometric information on classroom networks in the first two waves. Furthermore, cognitive and language tests were administered in the first and the sixth wave. In the latter, also structural and partnership-related life history calendars were implemented, together with an accent measure. And, not least: CILS4EU-DE was incorporated into the international CILS4EU-project, with identical surveys in the first three waves in England, The Netherlands, and Sweden; so, for these waves, comparative analyses on the integration of immigrants are also possible (Kalter et al., 2018a).

In the following sections of this data brief, we will provide information about the design of CILS4EU-DE and the development of the sample over time. We then give some basic descriptions of the respondents in the data. We continue by discussing thematic contents and some of the specific additional features, before elaborating on the opportunities for comparative research. Finally, we will provide a brief summary and an outlook on the future perspectives of CILS4EU-DE.

Design and Sample Development

In this section, we start by describing the selection of the initial sample in Wave 1, then report how the panel proceeded between Waves 2 and 5, and how the sample was refreshed in Wave 6 and continued afterwards. At the end we briefly summarize the basic overall development of the panel study.

The Initial Sample

The first three waves of CILS4EU-DE are the German part of the comparative CILS4EU study, which also included England, The Netherlands, and Sweden (Kalter et al., 2016a, 2018a, b). The first wave of data collection was conducted between Autumn 2010 and Spring 2011.

Sampling approach

In Germany, the target population for the first wave of CILS4EU was defined as all students being enrolled at that time in the 9th grade in secondary education, the usual age of the students being 14–15; in the other three countries corresponding grades focussing on the same age-groups were chosen.

The initial sample resulted from a school-based three-stage stratified approach. The first-stage sampling units were schools enrolling the eligible students, which in Germany means, schools with a 9th grade. The sampling frame was a comprehensive list of all schools enrolling this target population, allowing us to aim at a nationwide and representative sample. To include a considerable number of students with an immigrant background in the survey, schools with high minority shares were oversampled. More precisely, schools were sorted into four different explicit strata according to the expected proportion of children of immigrants (0–10, 10–30, 30–60, and 60–100 per cent). Besides the general idea to oversample schools in the higher strata, we followed considerations from the Neyman or optimal allocation approach to determine the exact distribution of schools over the different strata, to achieve the most effective sample (cf. Groves et al., 2004: p. 117; for more details see CILS4EU, 2016a).

In addition to the explicit stratification criterion ‘immigrant proportion’, implicit stratifiers—school type and federal state—were used to increase the face validity of the sample and enhance replacements of refusing schools by structural equivalents (Kish, 1965: p. 113 ff.; for more details, cf. CILS4EU, 2016a). After the distribution of schools over explicit and implicit strata, the targeted schools were selected with probabilities proportional to size, with the measure of size being defined as the (actual or estimated) number of students in the relevant grade level (Kish, 1965: p. 217 ff.). In the case of non-response on school level, refusing schools were replaced by schools, which are similar in immigrant proportions, school types, and federal states (CILS4EU, 2016a).

After the sample selection on school level, in the second stage of the sampling process two school classes in the respective grade level were randomly selected in schools with more than two school classes, while all classes were selected in schools with only one or two classes in the respective grade level. In the third and final stage, all students in the sampled classes were included in the gross sample, excluding those students who were mentally or physically not able to complete the survey or had major problems in the language of the survey country. However, such exclusion on student level occurred rather rarely (cf. CILS4EU, 2016a: p. 6, Table 1).

Given the disproportionate stratified sampling approach on the first level, the data set offers design weights allowing for representative estimates for the overall population of young people. Due to the three-level design of the original sample, these weights account for the (partly different) selection probabilities on the school, the class, and on the student level. Furthermore, the final weights also include adjustment weights, taking care of non-response on the different levels. The multiplicative combination of these weights results in a total student weight [totwgts], with the sum of weights over the cases in the data set approximates to the total size of the target population. The house weight [houwgt], which should in most cases be the weight of choice, is scaled in a way that the sum of weights over the different cases adds up to the sample size (for more details: Dollmann and Jacob, 2014; CILS4EU, 2016a).

Fieldwork and response

Following the sampling approach described above, 144 schools with 271 school classes agreed to take part in the first wave of CILS4EU-DE, with response rates on school level of 53 per cent before and 99 per cent after replacements. Once schools agreed to participate, the response rate on class level was almost 100 per cent (CILS4EU, 2016a).

The Data Processing Centre (DPC), which is part of the International Association for the Evaluation of Educational Achievement (IEA), conducted the fieldwork at schools. The survey was carried out during 2 school hours in the regular classroom context in the school year 2010/2011. All participating students in Germany received an incentive of 10 Euros. In total, 5,013 students agreed to take part, which corresponds to a response rate of 81 per cent (cf. Table 1 for the response rates of the first, but also of all other waves). This figure is rather high, especially when considering the fact that parents in Germany had to give their active consent to allow their child to take part in the survey. Out of the 5,013 students, 51 per cent have an immigrant background (defined as having at least one foreign-born parent or at least two foreign-born grandparents), which demonstrates the effectiveness of the oversampling strategy of schools with high immigrant proportions.

To allow for the analyses of intergenerational integration and transmission processes, not only information from adolescents but also from their parents was collected in the first wave. Parental questionnaires were handed out to the students during the classroom session with a request to return them via the school or directly per post. The questionnaires for the parents were available in the languages of the (numerically) most important immigrant groups in Germany, i.e. Turkish, Russian, Polish, Italian, Serbian, and Spanish, as well as in English. To raise the response rate, the German team followed parental non-respondents via a telephone survey, which resulted in an overall response rate in the parental survey in Germany of 78 per cent (n = 3,909; cf. CILS4EU, 2016a).

Furthermore, we also requested the form teacher to fill out a teacher questionnaire, in which we inquired about compositional characteristics of the school class, the school equipment, the individual background of the teacher, to name only a few items that were collected. As many as 92 per cent (248 out of 271 form teachers from participating school classes) of all targeted teachers participated in the survey (cf. CILS4EU, 2016a). However, it has to be noted that in 24 per cent of the classes, the form teacher was not available and another teacher filled in the questionnaire instead.

Panel Development from Wave 2–5

The fieldwork of the second wave started one year after the first wave (school year 2011/2012), and the aim was to administer the survey again in the school context, as most of the students were still enrolled in school. However, this default option could not be followed for students from 36 schools in Wave 1. Some schools refused to participate a second time (n = 10), while another 26 schools are lower secondary schools that do not enrol a 10th grade (CILS4EU, 2016b).¹ Students in non-participating schools were contacted using a mixed mode approach (postal, telephone, and Web survey), which was administered at the facilities of the University of Mannheim. Overall response rates of the inside and outside school survey were similar to those in the first wave (83 per cent; cf. Table 1).

As many of the respondents left school or changed schools between Waves 2 and 3, individuals were approached individually from Wave 3 onwards. The fieldwork in Waves 3, 4, and 5 was conducted in a mixed mode, using postal surveys, Web-based, and telephone interviews. The procedure slightly differed between the waves regarding the starting mode. All fieldwork was conducted by the German CILS4EU-team in Mannheim, again using the facilities of Mannheim University (telephone laboratory etc.). The response rate was lower in the third wave compared to the previous waves (67 per cent), which was mainly because this was the first survey for all students taking part outside the school context, and many students could not be contacted due to wrong addresses or telephone numbers. However, the sample achieved stability after Wave 4, with response rates of 79 and 86 percent in Wave 5 (cf. Table 1 and Figure 1; more details about the fieldwork can be found in the respective Technical Reports: Wave 3: CILS4EU, 2017; Wave 4: Olszenka et al., 2018; Wave 5: Sauter et al., 2018).

The Refreshment Sample (Wave 6)

Sampling approach

To guarantee a representative sample and a large number of respondents with and without an immigrant background in later waves, we drew a refreshment sample in the sixth wave in the year 2016. The target population were all persons in Germany of the birth cohorts 1994–1996, representing the modal birth years in the original Wave 1 sample. The aim of the refreshment was to bring the sample size back to its original size (as in Wave 1), with a similar distribution between adolescents with and without an immigrant background. As a consequence, it was again necessary to oversample adolescents with an immigrant background.

Given that respondents in this age group are usually no longer enrolled in schools, an identical sampling procedure as compared to the original sample was impossible. Hence it was decided to follow a sampling approach on municipality level. The first stage sampling units—municipalities—were drawn from a sampling frame including all German municipalities, with probabilities proportional to size, where the measure of size was the number of persons residing in the respective municipality who were born in the years 1994–1996. Prior to the sample selection, municipalities were sorted in the implicit strata defined by administrative district (‘Regierungsbezirk’) and administrative district size (‘Gemeindegrößeklasse’). The sampling process resulted in the selection of 62 municipalities, representing 63 sampling points (the city of Berlin represents one municipality, but due to its size, two sampling points were selected within the municipality; cf. Schiel et al., 2016 for more details).

To oversample adolescents with an immigrant background within the selected municipalities, information was needed to classify the units of the sampling frame obtained by the municipalities along this characteristic. Unfortunately, municipalities’ registration offices do not collect or share such information. Therefore, we decided to use a name-based approach to define the migration background of possible respondents a priori. The selected municipalities were asked to deliver all or a random selection of names and addresses of inhabitants of the relevant birth years. All names were then classified using an onomastic approach (Humpert and Schneiderheinze, 2000) into two categories: those cases with (most likely) having an immigrant background and those with (most likely) having none. The sampling frame with the information about adolescents’ names, addresses, and (estimated) immigrant background was then used as the basis for the selection of the gross sample within the different sampling points (for more details about the sampling process, see Schiel et al., 2016).

Like the original sample, the refreshment sample can be weighted to achieve estimates representative of the total target population. There is a weight accounting for the selection of municipalities as well as one for the selection of respondents. The latter considers the disproportionate design with higher inclusion probabilities of those persons in the gross sample being categorized as having most likely an immigrant background. The multiplicative combination of both weights resulted in the final design weight for the refreshment sample, which was further adjusted by accounting for non-response on individual level, using the information available for all cases of the gross sample (assumed immigrant background, federal state, and administrative district size; [refwgt]).

As the original sample as well as the refreshment sample were selected using sampling frames comprising the whole population of our target persons in Germany (non-disjoint samples), a simple combination of both weights (e.g. by using both weights simultaneously) is not possible. Instead, the weights had to be combined considering the multiple inclusion probabilities in both samples (for more details, see Schiel et al., 2016: p. 51f). Prior to the combination, the weight from the original sample [houwgt] was adjusted to account for panel attrition between Wave 1 and Wave 6, using logistic regression to model the attrition process [panwgt]. After this adjustment, both weights—[panwgt] and [refwgt]—were combined through a convex combination of the two. The resulting weight was then calibrated using different characteristics form the German Microcensus 2013 (migration background, gender, highest educational degree, region, highest educational degree of parents, federal sate, and administrative district size), which led to the final weight [calwgt] (Schiel et al., 2016).

Fieldwork and response

The fieldwork in Wave 6 was carried out in a face-to-face mode by the infas Institute for Applied Social Sciences, a private and independent social research institute. The questionnaire was administered using Computer Assisted Personal Interviews (CAPI). Overall, the response rate in the face-to-face interview among the refreshment sample was about 23 per cent (Schiel et al., 2016). The response rate for the respondents in the existing panel was higher, but with 52 per cent considerably lower compared to the previous waves. Therefore, an additional, shorter survey for those persons who were part of the original sample but who did not participate in the CAPI-survey administered by infas, was administered at the facilities of the Mannheim University. Mixed mode (telephone, Web, and postal) of data collection was chosen and resulted in a participation rate of 65 per cent in this shorter survey. Overall, 5,820 participants were surveyed in the sixth wave. As many as 90 per cent of the respondents from the refreshment sample provided their consent to be contacted again in Wave 7 (Schiel et al., 2016: p. 38).

Current Stage and Summarizing Figures

The seventh wave of the survey was collected in 2018 (and is currently in its final phase, as of September 2018). In contrast to the previous waves, the survey started two years after the latest wave. The rationale behind increasing the intervals between the waves is to reduce the burden for respondents and to save survey costs. However, increasing the gap between the waves comes at price of lower reachability of respondents. As can be seen from Table 1, response rates have dropped compared to those from the panel sample before Wave 6 (68 per cent as of September 2018). This is not only due to the fact that we observe lower response rates among the refreshment sample, which has not yet achieved its stability. Rather, we also find slightly lower response rates among the original panel as compared to the response rates prior to Wave 6. The main reason for the nonresponse is that respondents could not be reached, as some have moved houses, while others changed their phone number and could not be contacted again.

Table 1 provides an overview over the number of cases, differentiated by the mode of the survey. The response rate in the last column presents an overall response rate for the fieldwork of the respective wave (for more information, see the Technical Reports of the different waves: CILS4EU, 2016a, b, 2017; Schiel et al., 2016; CILS4EU-DE, 2018; Olszenka et al., 2018; Sauter et al., 2018).

Figure 1 plots the development of the sample over seven waves for respondents with and without an immigrant background,² differentiating between the original sample and the refreshment sample. As can be seen, and as outlined earlier, the original sample achieved a considerable stability after the third wave (i.e. after the first time the survey was administered in the outside-school context). For Wave 6, we see a slight drop in the number of respondents with an immigrant background, reflecting difficulties faced by the survey company to get in contact with these respondents at home.

Composition and Quality of the Sample

In this section, we will provide some basic descriptions of the sample. We start with the composition in terms of migration background, which, of course, is among the most important information in the context of our study. In the second part, we will compare the CILS4EU-DE sample to data from official statistics, providing evidence on the quality and face validity of the sample.

Composition in Terms of Generational Status and Ethnic Background

There is no clear consensus about the definition of immigrant background in integration research. One of the characteristic features of our study is that researchers can themselves decide how to conceptualize migration background or ethnic minority status to suit their research question. The data contain information on subjective ethnic identity, language use within families, and on the country of birth of family members and hence provides various criteria to be used for a definition of migration background as used in the literature.

In the following we will rely on countries of birth within the family tree to give an impression of the sample composition in terms of migration background. Using the information for the student, the parents, and the grandparents, one can distinguish a rather fine-tuned classification of a child’s generational status. The data set provides generated variables [y1_generationG] and [y6_generationG], combining the information from the student and the parental survey about the country of birth of the respondents themselves, their parents, and grandparents (Dollmann et al., 2014). We thereby are able not only to classify respondents into the first, second, and third generation but further differentiate between the age of migration for those who migrated themselves. We can also distinguish respondents born to parents, of which one experienced migration (2.5 generation; for further constellations, see Table 2). Finally, we can also pay attention to the country of birth of each of the four grandparents, resulting not only in just a third but also in a 3.25th, 3.5th, and 3.75th generation, depending on whether four, three, two, or only one grandparent experienced migration. The generated variable also takes care of partial non-response on these different variables by imputing missing values based on some grounded considerations (for more details, see Dollmann et al., 2014).

Table 2 describes the variety of respondents’ generational statuses in our sample. It is evident that the weighted distributions differ only marginally between the initial sample and the refreshment sample. About 59 per cent of the respondents in the initial sample and 58 per cent of the respondents in the refreshment sample have no migration background (at least up to the grandparent generation). Among respondents with migration background, respondents who were themselves born in Germany, but with both of their parents being born abroad (second generation) comprise the largest group (13 and 14 per cent, respectively). About 6 per cent (10 per cent for the refreshment sample) of respondents have a migration experience of their own (first generation). This is also the category where the largest differences between both samples can be observed. Especially the recent arrivals (1.25th generation) are more prevalent in the refreshment sample, which may be a consequence of the recent influx of refugees to Germany. Respondents with parents being born in Germany but foreign-born grandparents (third+ generation) are rather rare. An exception in this respect are respondents having only one grandparent with a direct migration history (3.75th), a pattern that can be observed in the initial (9 per cent) as well as in the refreshment sample (6 per cent) and in the other three countries of CILS4EU (Dollmann et al., 2014).

The next obvious question is what specific ethnic groups of children of immigrants are found in the data. Again, there is no common way of defining group membership in integration research, and it can be operationalized in many different ways with the CILS4EU-DE data. In the following we identify the migration background of each child belonging, according to Table 2, at least to the 3.75th generation by the country of birth of the grandparents. If the grandparents were born in different countries (other than the survey country), we take the country with most nominations; in case of equal nominations we gave priority to the female line.³Table 3 shows the largest immigrant groups according to this definition in the first (initial sample) and the sixth (refreshment sample) wave (variables [y1_countorig_geG] and [y6_countorig_geG]).

We can identify over a hundred different immigrant groups, portraying the enormous ethnic diversity in contemporary Germany. Nevertheless, the immigrant youth population is still dominated by youths of the Turkish background. The remaining groups are either traditional labour migration countries of Southern Europe (Italy, Serbia, and Greece) or more recent immigrant groups from Eastern Europe or the Middle East. When comparing the weighted figures between the initial and the refreshment samples, it becomes evident that distributions are very similar.

Comparing the CILS4EU-DE Sample to Data from the Official Statistics

When applying the design-weights, the CILS4EU-DE sample can be compared to the data from the federal statistical office. In Table 4, we crosscheck the distribution of the immigrant and generational status in CILS4EU-DE to the distribution of the respective birth cohorts retrieved from the German Microcensus 2009. Once a coarser classification scheme (compared to the one applied in Table 2) is adopted, one-third of all respondents in our sample have an immigrant background. This share matches rather well with the information from official data. Among adolescents with an immigrant background, the vast majority are persons born in Germany to parents who both migrated (second generation). Furthermore, respondents with a migration background on their own (first generation) are a sizable group, with slightly higher numbers in the refreshment sample, which may—as already noted above—reflect the recent migration wave to Germany. Again, immigrants belonging to the third generation are a much smaller group, which is especially pronounced in our refreshment sample, but can also be observed in the Microcensus-data. CILS4EU-DE data slightly overrepresent respondents stemming from interethnic relationships.

In contrast to the figures pertaining to the generational status, it is more difficult to find relevant references to compare the distribution of minority groups by ethnic origin (as in Table 3) to official statistics. This is not least due to deviating or unknown definitions of an immigrant background in some of the official statistics. However, the largest immigrant group in Germany—individuals of Turkish heritage—comprises approximately 18 per cent of all persons in Germany with an immigrant background based on the German Microcensus 2010 (Woellert and Klingholz, 2014), which comes close to the weighted proportions in CILS4EU-DE (both samples: 19 per cent).

Next to oversampling students with an immigrant background, implicit stratifiers were used to increase the face validity of the sample. For both, the initial sample and the refreshment sample, geographical information was used to draw a sample proportionately across the German federal states. In the first wave, federal states were used as an implicit stratifier, while in the refreshment sample, information on a lower administrative level (‘Regierungsbezirk’) was used. In Table 5, we compare the weighted distribution of young people across the federal states to figures from the official statistics. The original sample was drawn without Bavaria due to the refusal at the federal state level, which is why we excluded this state from the calculation of the proportions from the figures from official statistics for the comparison with Wave 1 data.

Finally, school type was used as an implicit stratifier in the initial sample. As can be seen from Table 6, the weighted distribution of the students included in the initial sample over the different school types is rather similar to the figures available from the official statistics.⁴

Concepts and Special Features of the Survey

The basic idea behind the development of the instruments for CILS4EU-DE was to find a balance between the (i) inclusion of core modules in each wave or at least every other wave and (ii) the introduction of thematic modules. These thematic modules should focus on specific aspects of integration being relevant in the respective age group or using the opportunities provided by the survey mode to include specific modules. In the following, we will describe the content of the core and the thematic modules.

Core Modules

Regarding the core modules, we focus on four dimensions relevant for immigrant integration (cf. Berry, 1997; Esser, 2006; Kalter, 2008.) The cognitive–cultural dimension comprises language skills and other country-specific knowledge. Several questions asking about the language spoken at home and the subjectively evaluated proficiency in the language of the host as well as of the heritage country were included in the survey.

The structural dimension focusses on the integration into important societal domains, like the education system or the labour market. In the beginning of the survey and due to the comparably young age of the respondents, the most important indicator in this respect was the placement into the different tracks of the German highly stratified education system. From that on, every wave contains an extended module on the current situation within the education system or on the labour market. Thus, it is possible, for example, to study whether respondents stay in general academic track or whether they follow more vocationally oriented tracks and enter the vocational training system. In the following waves, it should also be possible to investigate the access to the labour market and evaluate whether there are specific obstacles for respondents with an immigrant background.

The social aspect of integration can be operationalized by the extent to which children of immigrants get in contact with peers without an immigrant background (and vice versa). These contacts are manifold and comprise acquaintances, friendships, partnerships, and at older ages also marriages. In the CILS4EU-DE-survey, we ask extensively about the composition of weak and strong ties and about the ethnic and social background of potential partners.

Finally, information about the degree of ethnic or national identity, the religion and religious behaviour, and norms and values are subsumed under the emotional–cultural dimension. Whereas the aspect of religious, ethnic, and national identity gets particular emphasis in the survey, further questions about gender roles, the acceptance of different lifestyles and sexual behaviour, and opinions about whether immigrants should integrate into the host society are also captured by this module. Finally, and additionally to the items focusing on the integration of immigrants, we also include questions about respondents’ health status and their physical and mental well-being.

As outlined earlier, items of the core modules are repeated in an annual or biennial manner. This allows researchers to assess individual changes over time. To evaluate generational changes within these items, several items from the students’ core modules are included in the parental survey in an identical formulation, like self-evaluated skills in the language of the heritage and the receiving country, religious, ethnic and national identity, gender roles, to name only a few.

Thematic Modules

The core modules are complemented by thematic modules in the different waves, which are both displayed in Figure 2. For example, with the aim of capturing not only subjective evaluation of the skills in the language of the receiving country, an objective language test pertaining to the respondents’ lexicon was administered during the first wave of the survey (Heller and Perleth, 2000). Figure 3 presents some exemplary items from the lexicon test that was administered in the English context, which is highly comparable with the German version of the lexicon test.

The lexicon test in Wave 1 was accompanied by a language-free cognitive function test (cf. Figure 4), which is culturally fair and hence does not penalize participants with lower levels of German language skills (Weiß, 2006).

As the sixth wave of data collection was again organized in a face-to-face mode, it was possible to repeat both tests from the first wave, with the lexicon test being adjusted to a difficulty level suited for an older population.

The social integration was a core aspect in the thematic modules of the first three waves. Information on the ego-centric networks was collected in the first as well as in the third wave. We also took advantage of the survey being fielded in the school context during the first two waves and implemented a sociometric network measure. With this measure, it is possible to investigate different aspects of social relations within the classroom. Each student in the class received a number, and while answering survey questions, respondents had to relate to these numbers when describing their social ties. For example, respondents were asked about their best friend(s) in class or with which classmates they sometimes spend time outside of school, as well as who is sometimes mean to them or who they would not want to sit by. Figure 5 provides an overview over the patterns of social integration in two selected classrooms (taken from Dollmann and Jacob, 2014). The symbols represent respondents’ gender (circle: female, square: male) and their ethnic background (Black: no migration background; White: Turkish migrations background; Grey: other migration background).⁵ The connecting lines present friendship nominations and the arrows who nominated whom as a friend.

As can be seen, the classroom on the left-hand side is rather segregated, with contact mainly within the own ethnic background, accompanied by a segregation along the gender dimension. A similar pattern of gender-segregation can also be observed on the classroom on the right-hand side. However, here we find many more interethnic contacts.

In Wave 4, respondents were aged around 18, and partnerships should have become increasingly important for them. Therefore, we included a thematic module asking in more detail about the characteristics of partners, the role of parents and the family in the partner choice, and the importance of opportunity structures.

At the age of 18, German citizens are allowed to take part in federal elections, which is why we implemented a thematic module on social and political participation in Wave 5. Here, questions about formal as well as informal political participation, left-right orientation and political and institutional trust were complemented by questions asking about the degree of participation in social movements, etc. Furthermore, as this issue is also relevant in this age group, respondents without a German citizenship were asked about the possibility of applying for the German citizenship and the roles their families would play hereby.

In Wave 6, we implemented a life history calendar (LHC) collecting information on all relevant spells pertaining to the structural areas, like educational career and entry into the labour market, starting at the beginning of 2011 (which is slightly later than the start of the original panel-survey started). This makes it possible to analyse the initial as well as the refreshment sample in a similar way with respect to the structural integration processes of young people. Figure 6 provides an overview over the basic pathways between the ages 15 up to 20 of the students resulting from the data of the LHC, with more detailed information available within these basic spell types (not shown here). As can be seen, immigrants⁶ seem to stay longer in education, which could point to longer educational careers due to grade repetitions, but also to the existence of so-called positive choice effects, according to which immigrants are less likely to enter the vocational training system but rather aim for further schooling. This latter assumption is confirmed when focusing on the rates of entry into vocational education, which is consistently lower for respondents with migration background compared to the benchmark of those without an immigrant background. In contrast, the attendance rates in tertiary education are higher for native respondents, while the ‘something else’ category is more common among immigrant adolescents.

Furthermore, the LHC also contained information on all types of partnerships as of January 2011, together with characteristics of the partners, like his or her ethnic and social background.

In addition, we also measured accents in Wave 6. Respondents had to read a text aloud and afterwards engage in an oral conversation on the content of the questionnaire. Both audio-segments were recorded and subsequently evaluated by raters with training in linguistics. The resulting variables are measured on a nine-point Likert scale reflecting the strength of a person’s foreign and regional accent, respectively.

In Wave 7, a thematic module on interethnic relations and social distances was included. The module encompasses different measures, e.g. a repetition of the feeling thermometer already included in earlier waves, which asks how respondents feel about different ethnic groups (including the majority population). In addition to the opinions about immigrants from the ‘traditional’ sending countries, the item battery was extended to include some of the newly arriving immigrant groups. By including such measures, it will be possible to examine whether changes in the ethnic hierarchy ensue upon the arrival of new immigrant groups, with an ‘upgrade’ of established immigrant groups in contrast to groups entering Germany in recent times. Furthermore, we also repeated the ego-entered network measure from Waves 1 and 3 in the survey of the seventh wave.

Using CILS4EU-DE for Comparative Purposes

CILS4EU-DE is part of the international CILS4EU-project, which was funded in the framework of the NORFACE (New Opportunities for Research Funding Agency Cooperation in Europe) research programme on migration between 2009 and 2014. Other countries being involved in this project are England, The Netherlands, and Sweden. The aim of this project was to provide information about the integration of children of immigrants in a comparative perspective.

To ensure the comparability of the data, all countries followed the same sampling design. School based samples were implemented in all countries, following the same idea of oversampling schools with high immigrant proportions. Table 6 presents the number of cases achieved for Germany (cf. already Section 2), England, The Netherlands, and Sweden on school, class, and student level. As can be seen from the last row, the oversampling strategy worked in all countries, although to a lesser degree in The Netherlands. In total, CILS4EU provides information about 18,716 students, of which almost 46 per cent (n = 8,557) have an immigrant background (cf. Table 7; see also CILS4EU, 2016a).

However, not only the sampling process was strictly harmonised between the countries but also the fieldwork. In all countries, school-based surveys were a default in the first two waves, whereas in the third wave respondents were approached individually in mixed modes (telephone, postal, and Web). Furthermore, all country teams implemented parental and teacher questionnaires in the first wave as described for the German case in Section 2.

Regarding the survey content, the instruments were identical in the four countries over the first three waves, with some country-specific questions whenever this was necessary (i.e. questions regarding educational transitions that were possible in Germany already after Wave 2). To avoid item bias due to imprecise translations, we implemented the TRAPD-approach with its five consecutive and iterative steps: Translation, Review, Adjudication, Pretesting, and Documentation (Harkness, 2007). Furthermore, cognitive pretests were implemented in all countries prior to the survey to identify possible problems related to translations. Following the TRAPD-principles, not only the instruments of the student survey were translated into Dutch, German, and Swedish starting from the English master version but also the instruments for the parental surveys into the different minority languages (CILS4EU, 2016a).

Similar to the German case and its national CILS4EU-DE study, all country teams undertake efforts to implement national follow-up studies. To date, the Dutch team conducted four additional waves (CILSNL), while the Swedish team followed their respondents in one additional wave (with register data being available for a longer period of time). The English team is currently working on the implementation of additional waves, together with the selection of new cohorts of the same starting age as in the original sample. In contrast to the first three waves, the content of the survey is less harmonized due to specific research interests in the different country teams. However, there is still some consensus about core concepts that should be measured in each wave, allowing for comparative analyses also beyond the CILS4EU project.

Data Access

So far, six waves of CILS4EU-DE are available at the Data Archive for the Social Sciences (DAS) at GESIS (Gesellschaft Sozialwissenschaftlicher Infrastruktureinrichtungen) (Kalter et al., 2016a, b, 2017, 2018b), with the seventh wave planned to be published in Spring 2019. Furthermore, all three waves of the international CILS4EU-survey with the data from England, The Netherlands, and Sweden are also available. The actual and estimated data release dates are displayed in Figure 7, together with a roundup of the basic facts about number of cases and modes.

Data users need to apply for the data with a short description of their research projects.⁷ Once having access to the data, users are able to download different single data sets. The structure of these data sets follows the differentiation between the groups of respondents (students/adolescents, parents, and teachers) as well as the basic parts of the survey among students/adolescents (e.g. main part with the core modules, sociometric measures, achievement tests, and LHC). Furthermore, the data structure also differentiates between countries participating in the first three waves.

Due to data protection regulations, not all information is available in the download version available at GESIS. For example, countries of origin that are rare in the data are aggregated to higher-level regional information (‘Algeria’ is aggregated to ‘Northern Africa’); only a two-digit ISCO-code is provided,⁸ or extreme values on scales like body height or weight are aggregated. Regardless of these changes, and due to our own experiences, almost all analyses are possible with the off-site version of the data. For very specific questions, where the complete, non-aggregated information is necessary, it is still possible to make use of the on-site version of the data, which is available in the Secure Data Centre at GESIS.

The data are accompanied by comprehensive documentation material, including technical reports (CILS4EU, 2016a, b, 2017; Schiel et al., 2016; CILS4EU-DE, 2018; Olszenka et al., 2018; Sauter et al., 2018), codebooks (CILS4EU, 2016c), and the questionnaires used in the different waves. Furthermore, we also provide some more specific documents, like a documentation on how to best use the sociometric data (Kruse and Jacob, 2016) or information on how the evaluations for the accent measurement were conducted (Weißmann and Dollmann, forthcoming).

Concluding Remarks

The aim of this data brief was to outline the strengths of the CILS4EU-DE survey for the research on immigrants’ integration into their host societies. With the variety of thematic issues, the data offer manifold research options, including also aspects that have long been neglected in large-scale quantitative research, like the measurement of foreign and regional accents. It is possible to assess cognitive–cultural, structural, social, and emotional–cultural processes of integration, or the interplay between different dimensions. Moreover, it is not only possible to focus on the correlation between different dimensions of integration, but due to the longitudinal design it is also possible to disentangle the causal interplay between different dimensions, i.e. whether social integration is a precondition or a consequence of emotional-cultural integration. This longitudinal perspective hopefully does not end with the seventh wave, which is the latest wave described in this data brief. Within the overall period of the German follow-up study being funded as a long-term project by the German Research Foundation (DFG), two additional waves are being planned, which will take place in the next funding period, two and four years after the seventh wave.

Furthermore, the data allow to study integration processes in a comparative way, juxtaposing adolescents with an immigrant background in different European countries, representing also different systems in various respects. For example, regarding structural integration, we can examine academic achievements of students with and without an immigrant background in choice-driven education systems (like in Sweden or England) versus selective systems (like in Germany or The Netherlands). This internationally comparative approach is thereby not only restricted to the case of the four countries, but, given that Belgium and Norway fielded similar surveys (Phalet et al., 2015), can be extended to include other countries, which implemented or might decide to implement similar studies as well.

The CILS4EU-DE data as well as the data from the larger CILS4EU-project are extensively used and so far, about 60 articles were published based on the data, some of them in highly ranked journals. These studies comprise very different kinds of substantive research questions, like the emergence of positive educational choice effects among immigrants (Dollmann, 2017), the mental health advantage among immigrant children (Mood et al., 2017), neighbourhood effects on ethnic friendship formation (Kruse, 2017), segregation in friendship and online networks (Smith et al., 2016; Hofstra et al., 2017), and gender roles and their influence on achievement (Salikutluk and Heyne, 2017). Methodological issues can also be studied with the data (Engzell and Jonsson, 2015). Nevertheless, and regardless of the amount of articles published so far, there is still an enormous potential not yet being utilized in the data, making it an interesting source for contemporary integration research.

Frank Kalter is a Professor of Sociology at the University of Mannheim and Director of the German Centre for Integration and Migration Research (DeZIM) in Berlin. He is a Fellow of European Academy of Sociology (EAS) and served as its president from 2011 to 2015. He is the Principal Investigator of the CILS4EU study. His major research interests are in migration, the integration of ethnic minorities, social networks, and formal modelling.

Irena Kogan is a Professor of Comparative Sociology at the University of Mannheim. She is a Fellow of EAS. Her research interests are in the areas of social inequality, migration research, and labour market sociology. In particular, her research focuses on structural and cultural aspects of immigrant integration, transition from school to work, as well as the role of human capital, social, and cultural resources in these processes. Interest in the role of countries' institutional characteristics in explaining societal processes drives her research, which is largely internationally comparative. In her studies, she relies on modern methods of quantitative research and survey data.

Jörg Dollmann is a Postdoctoral Researcher at the Mannheim Centre for European Social Research (MZES). He was the former international project coordinator of Children of Immigrants Longitudinal Survey in Four European Countries (CILS4EU) and currently organizes the German part of the project (CILS4EU-DE), which became a long-term project of the German Research Foundation (DFG). His research interests are social and ethnic inequalities, institutional settings and educational success, and the integration of immigrants and ethnic minorities in a comparative perspective.

Footnotes

Acknowledgements

The authors would like to thank Markus Weißmann for his assistance in preparation of the tables.

Funding

CILS4EU-DE is funded as a long-term project by the German Research Foundation (DFG; KA 1602/8-1/2; KO 3601/8-1/2), starting with Wave 4 (as of 2014). The first three waves, which were also conducted in England, The Netherlands, and Sweden, were funded within the NORFACE ERA NET Plus Migration in Europe-programme (2009–2013).

References