Integration of Recently Arrived Underage Refugees: Research Potential of the Study ReGES—Refugees in the German Educational System

With the strong immigration of refugees at the end of the last decade, many minors came to Germany. Their integration into the German educational system will be one of the central tasks of German society and is important for the medium- and long-term integration of incoming young refugees in various areas of society. Data from the ReGES panel study enable analyses of the early integration process of refugees into the German educational system. The study focuses on two cohorts: 2,405 children of preschool age who are about to start school and 2,415 adolescents in lower secondary education who were followed during the transition into vocational training or upper secondary education. Each cohort was interviewed seven times over three school years from 2018 to 2020, with sampling procedures starting in 2017. This article provides an overview of the study design, contents, and research potential of ReGES. The large number of measures of individual characteristics, including objective competence test data (German language and basic cognitive skills) and information on the learning and living contexts of refugees provided by parents, teachers, and administrative staff, allow us to address many research questions.


Objectives of the Study
Over the second decade of the 21st century, the number of displaced people continued to rise. In 2019, a new high of 7.5 million forcibly displaced persons worldwide was reached (UNHCR, 2020). Although most refugees seek protection within their country of origin or in neighbouring countries, the number of those fleeing to areas farther away, especially to Europe, has increased sharply, especially in 2015 and 2016. Within the European Union, Germany hosts the most refugees in absolute terms. In 2014 to 2017 alone, more than 1.5 million asylum applications were submitted in Germany (Bundesamt fü r Migration und Flü chtlinge, 2020).
The sharp increase in refugee migration has not only presented politicians and local actors in Germany with severe challenges but also aroused the interest of the scientific community. Especially in the social sciences, new research projects have emerged relatively quickly (Kleist et al., 2019). Most of these projects focus on generating suitable databases for research on this group.
The study ReGES-Refugees in the German Educational System 1 -is a longitudinal panel study that focuses on children and adolescents. There are two reasons for this focus. First, there is a high proportion of minors among refugees. The proportion of first-time applicants under the age of 18 rose from just under a third in 2014 and 2015 to nearly half in 2017 (Bundesamt fü r Migration und Flü chtlinge, 2015Flü chtlinge, , 2016Flü chtlinge, , 2017Flü chtlinge, , 2018. Second, the acquisition of education is considered a particularly conducive condition for social and economic opportunities in later life for children and adolescents (OECD, 2020), especially for refugees (see, e.g., Nationaler Aktionsplan Integration, 2012). This relationship should be particularly pronounced in Germany since educational qualifications and certificates are essential for success in the German labour market (Bol and van de Werfhorst, 2011). The ReGES data allow us to describe the participation of refugee children and adolescents in the German educational system and to analyse conditions that foster or hamper (educational) integration. The study was designed following the German National Educational Panel Study (NEPS) (Blossfeld and Roßbach, 2019) to enable comparative analyses of children and adolescents of the majority population and other migrant groups. The study is interdisciplinary and aims to address questions of relevance for several academic disciplines working on issues of education.
In the following, first, the research design of the study is described (Design of the study section), and the sampling for the initial measurement and the panel study are discussed in detail (Sampling and response rates section). The response rates for the panel waves and instruments are also presented. Description of the sample and selectivity section describes the initial sample based on selected characteristics. Using representative data on refugees in Germany, the possible selectivity of the initial sample is examined, and attrition and selectivity within the course of the panel are discussed. In Content and special features of the study section, the central study contents of the ReGES study and the research potential of the data are described based on special features of the study. Finally, information on data access is presented (Data access section). Conclusion and outlook section provides concluding remarks and an outlook on the continuation of the study.

Design of the Study
The ReGES study was a multi-informant, longitudinal panel study that focused on two cohorts that were facing key educational transitions.
Refugee Cohort 1 (RC1): The first cohort concentrated on the age group (4 years and older) that was not yet enrolled in primary school at the time of the first measurement. Participation in preschool plays an important role at this time point. Attending preschool facilitates the acquisition of host country language skills (see Lee et al., 2018;Becker, 2019) and offers the possibility of integrating the whole family (see Albers, 2011). Positive effects on parental participation in the host society are also discussed in the literature (Gambaro, Neidhö fer and Spieß, 2019;von Maurice et al., 2020).
Refugee Cohort 2 (RC2): This cohort consisted of adolescent refugees between the ages of 14 and 16 years who were attending lower secondary level (fifth to tenth grade) in Germany at the beginning of the study. The aim was to accompany this cohort during their transition to upper secondary level or vocational training in the course of the study. This transition is central to integration into the labour market in Germany as formal education largely structures later labour market opportunities (see Allmendinger, 1989;Breen, 2005).
A close-knit survey was employed in both cohorts because a high level of spatial mobility of the target persons was assumed. Moreover, frequent changes in educational contexts and episodes were conceivable, especially at the beginning of the integration into the educational system, if, for example, refugee students initially attended separate classes for newcomers or if the most suitable school form was not identified immediately.
At the time the study was designed, there were hardly any experiences with interviewing refugees in large panel designs. Therefore, different survey modes were used in ReGES to test which were particularly suitable for this target group. The core of the ReGES study was annual personal interviews (waves 1, 4, 7). These consisted of computer-assisted personal interviews (CAPI) combined with computer-assisted self-interviews (CASI). This combination was chosen because complex questions (e.g., educational biography, legal status) can be better processed in CAPI with the assistance of an interviewer who can ask for official documents (e.g., on legal status). By using an additional CASI, interviewer effects could be reduced, especially for sensitive aspects or those that are prone to social desirability. In addition to the annual personal interviews, one computerassisted telephone interview (CATI) (wave 3) and various computer-assisted web interviews (CAWI) (waves 2, 5, 6) were implemented. Overall, both cohorts were accompanied for approximately 2.5 years and interviewed in seven-panel waves (see Figure 1).
Several respondent groups were surveyed following a multi-informant perspective. Due to the age of the children, the main informants in RC1 were the parents. In RC2, the adolescents themselves were the main informants. Additionally, the adolescents' parents were asked about their social and educational background in the first wave. The interviews were supplemented at two measurement points (waves 1 and 7) by competence tests of the children and adolescents in the domains of German language and basic cognitive skills. Furthermore, in addition to the information on the family learning environment provided on the individual level, information on the institutional learning contexts (preschools and elementary schools in RC1; general education and vocational schools in RC2) and the regional contexts was provided by paper and pencil interviews (PAPI) of preschool/school heads and teachers and administrative staff in municipalities and collective accommodations, respectively. While the regional context persons were interviewed only once at the first measurement point, the institutional context persons were interviewed every school year. In the annual survey of the teachers, individual evaluations concerning refugee children and adolescents were also included. In the annual CAPI, the educational institutions currently attended and their contact details were recorded to enable the surveying of institutional context persons following the face-to-face fieldwork. Figure 1 presents an overview of the different surveys with the various respondents in the course of the ReGES study.

Sampling and Response Rates
In each of the two cohorts, the targeted sample size was 2,400 cases in the first wave to examine refugees in these age groups in detail and to have a sufficient number of cases to follow the refugees on their educational path over several years. This was particularly relevant in RC2 since after leaving the lower secondary school level, students can choose many different educational paths. However, the aims of the ReGES study were not only to analyse the relationships between individual characteristics such as social origin, residence status, or German skills and educational success but also to analyse the effects of different regional and institutional living and learning contexts on educational achievement. Therefore, it was necessary to provide sufficient case numbers in the individual regional, and institutional contexts. Thus, the ReGES study deliberately refrained from collecting data across Germany and instead concentrated on five federal states that were selected according to various criteria on the macro-level (for details, see Will et al., 2018b;Steinhauer, Zinn and Will, 2019): Bavaria, Hamburg, North Rhine-Westphalia, Rhineland-Palatinate, and Saxonia. This selection aimed to ensure that both rural and urban areas as well as municipalities with different economic conditions and regions with different levels of experience in the integration of migrants were included. Another important criterion was schooling regulations of (underage) refugees, which differ considerably between federal states in Germany. For example, while Rhineland-Palatinate provides inclusive schooling to new immigrants from the outset, other federal states initially separate immigrants who cannot speak enough German into special classes and only gradually integrate them into regular class settings (for a review of the regulations of the federal states examined in the ReGES study, see Will and Homuth, 2020; for an overview of all federal states, see Massumi et al., 2015).
A multistage process was used for the sampling. In Germany, the Auslä nderzentralregister (AZR; central register of foreigners) provides a register of all asylum applicants. However, the use of the AZR for sampling in scientific studies is only possible to a limited extent. However, the AZR can be used to identify rural and urban districts within the five selected federal states in which refugees from the main countries of origin (Afghanistan, Iraq, Syria) in the relevant age groups live (for details, see Steinhauer, Zinn and Will, 2019). Based on this information from the AZR, 20 rural districts and 40 urban districts were drawn proportionally to the number of refugees, with the city-state of Hamburg and all urban municipalities in Rhineland-Palatinate and Saxony set as sampling points. This procedure was necessary to ensure a sufficiently large number of cases in each federal state as well as the total number of cases of n ¼ 2,400 participants per cohort. In a second step, the four municipalities with the most inhabitants were selected from each of the 20 rural districts as it was assumed that these municipalities also have the most refugees. This resulted in a total sample of 120 cities and municipalities (40 cities and 80 communities within rural districts). In the last step, the registration offices of the selected cities and municipalities were asked to draw persons from the relevant age group (RC1: children born between 30 September 2011 and 31 October 2013; RC2: adolescents born between 31 October 2000 and 31 October 2003) who moved into the municipality after 1 January 2014, and who came from one of the following nations: Syria, Afghanistan, Iraq, Iran, Pakistan, Nigeria, Somalia, Gambia, Lebanon, Eritrea, unknown citizenship, or stateless. Based on these criteria, 18,735 addresses were drawn and randomly divided into parts (i.e., tranches), which ultimately were not all included in the sampling fieldwork. In total, 10,236 addresses (4,680 drawn via RC1-child; 5,556 drawn via RC2adolescent) were used for the initial measurement. To exclude other migrant groups or persons who had been living in Germany for several years, the interviewers checked whether the target children and adolescents drawn actually belonged to the target populations with a short CAPI screening.
To reach as many persons as possible in the first survey wave, the refugees were not only contacted personally by interviewers at home, but it was also ensured that the interviews were conducted only by interviewers who spoke the assumed mother tongue of the target persons. Instruments were available in the following eight languages: Arabic, English, Farsi, French, German, Kurmanji, Pashto, and Tigrinya (for further details on the selection of languages, the organization of fieldwork, and the translation, see Gentile, Heinritz and Will, 2019). In the second part of the interview, the CASI questionnaire could be answered in a completely audio-based format, and answers to open-ended questions were recorded so that illiterate persons could participate as well.
Response Rate in Wave 1 Figure 2 provides an overview of the addresses used, the identified target children and the interviews conducted in the first wave, differentiated according to all parts of the interview (interview, testing, consent that the institutional context persons may be interviewed) and willingness to participate in the panel. 2 In the first wave, the response rate (RR2) was 49.8 per cent. The most common reason for failure to participate was that the target persons no longer lived at the given address (23.9 per cent). Only 11.1 per cent of the cases refused the screening interview. Overall, the willingness of the refugee families to take part in the study was very high with a cooperation rate of 80.1 per cent European Sociological Review, 2021, Vol. 00, No. 0 (COOP4) 3 (for details, see Steinhauer, Zinn and Will, 2019).
Of the n ¼ 5,475 target children/adolescents identified, valid interviews were conducted for n ¼ 4,820 cases (n ¼ 2,405 RC1 children and n ¼ 2,415 RC2 adolescents). Participation in the CAPI and CASI survey for adolescents was a prerequisite for a valid case in RC2. For a valid RC1 case, parents had to participate in the CAPI and the CASI questionnaire. In the case of adolescents, the parent interview was optional.

Selection of the Panel Sample
After the initial interview in eight languages, further panel waves focused on the four so-called panel languages (Arabic, English, German, Kurmanji). Only participants who spoke one of these languages could participate in the competence tests in the first wave and could remain in the panel. A total of 95.9 per cent of RC1 parents and 94.1 per cent of RC2 adolescents were willing to participate in the panel; of these, 97.3 per cent of the parents and even 99.7 per cent of the adolescents spoke one of the panel languages. This resulted in the panel sample of n ¼ 2,251 RC1 children (in 1,935 families) and n ¼ 2,267 RC2 adolescents.

Response Rates in the following Panel Waves
As mentioned in Design of the study section, various survey modes were used in the further course of the panel. These differed significantly with regard to response rates (see Table 1). The first CAWI (wave 2), announced only via a panel app, led to a very low level of participation (approximately 15 per cent), although willingness to install the necessary panel app was relatively high (see Figure 2). Therefore, the subsequent CAWIs were also announced by letter and e-mail, which increased the participation rate slightly to approximately 19 per cent (wave 5) or 24.4 per cent (RC1) and 29.6 per cent (RC2) in wave 6, respectively. As an important methodological result of the ReGES study, it can therefore be concluded that this mode is not very suitable for the refugee population. A telephone interview (wave 3) reached significantly more respondents (approximately 40 per cent) than the CAWIs, although the CATI received lower response rates than personal interviews. Overall, the response rates were highest in the face-toface fieldwork: a response rate of over 75 per cent was realized in the second face-to-face survey (wave 4) in both cohorts and approximately 58 per cent in the third face-to-face survey (wave 7). The proportion of participants reached in the last face-to-face interview was significantly lower than that in previous face-to-face waves as the fieldwork had to be suspended due to the COVID-19 pandemic. The survey mode was switched from CAPI to CATI to resume the survey, which resulted in lower response rates (for more details, see Will, Becker and Weigand, 2020).
Information on the regional context from the PAPI of the administrative staff of the municipalities was available for almost one-third of the children and adolescents. Regional context information on the level of collective accommodation was available for almost half of the RC1 children (46.1 per cent) and 37.7 per cent of the adolescents who lived there during the first wave. The participation of persons in the institutional context in the time before the COVID-19 pandemic was between 23.4 per cent and 30.1 per cent, which is comparable to the current response rates in large-scale studies (see, e.g., NEPS Newborn Cohort; doi: 10.5157/NEPS: SC1:7.0.0). However, since the identification of the attended educational institution proved to be very difficult in many cases as participants could not provide sufficient address information, it was not possible to invite the heads and teachers to participate in the survey in many cases. The proportion of children and adolescents for whom information on the institutional context was available is therefore significantly lower, at 15.5 per cent to 25.5 per cent (see Table 1). The last survey in the institutional context took place at the time of (pre)school closings due to the COVID-19 pandemic, and response rates were therefore slightly lower.

Description of the Sample and Selectivity
Description of the Sample In the following, both samples of the ReGES study are described on the basis of central characteristics (see Table 2 for an overview). Syrian asylum seekers represented the majority in both ReGES cohorts, followed by refugees from Iraq and Afghanistan. Most of the ReGES participants had a rather secure legal status and were granted refugee status at the time of the first interview. Approximately, 85.5 per cent of both ReGES samples were already living in private accommodations at that time.
Concerning educational background, the majority of the parents in ReGES had obtained no or only primary education in their countries of origin. However, approximately 20 per cent of both cohorts also had postsecondary or tertiary education. With regard to the selfassessment of German skills, we found great differences between the respondents. While the adolescents mostly rated their German skills as very good or rather good To assess the analytical potential with regard to the description of educational trajectories and the analysis of educational decisions, it is also relevant to consider the children and adolescents' educational stages in the course of the study (see Figure 3). By design, the RC1 children in the first wave were not yet enrolled in primary school. Nearly 80 per cent attended a preschool, and approximately 20 per cent were mainly cared for at home. One school year later, nearly half of the children were attending primary school, and in the third study year (school year 2019/2020), nearly 93 per cent had made the transition into primary school.
In RC2, by design, all adolescents were in lower secondary schooling (fifth to tenth grade) at the first measurement. The students of RC2 were distributed relatively evenly across all school types (see Table 2) but attended less-demanding types of schools and were in lower grades than their same-age peers in Germany (see Homuth, Will and von Maurice, 2020). Additionally, in the next school year (2018/2019), less than 20 per cent of the adolescents had left lower secondary schooling. At the time of the third face-to-face interview, 54 per cent of the adolescents were still attending lower secondary schooling, while 13.2 per cent had reached upper secondary schooling (eleventh grade or higher). Of those who had left the general education school system, the majority were in the transition system of the vocational educational system (total 21.3 per cent), and only 4.3 per cent had started vocational educational training (VET), while 6.6 per cent had left the educational system.

Selectivity of the Initial Measurement
To assess the quality of the data, an attempt should be made to provide information about the extent to which the initial measurement of the ReGES study is biased before attrition is discussed in the course of the panel.
Since there are only a few basic characteristics in official statistics on refugees in Germany (e.g., age, gender, country of origin), these data are only suitable to a limited extent for comparisons with the ReGES data. Data from the IAB-BAMF-SOEP survey of refugees (see Kü hne et al., 2019; hereinafter SOEP), which is representative of newly immigrated refugees in Germany from 2013 to 2016, are more suitable.
First, the majority of the refugees in Germany were single males (52 per cent), followed by families with children, who represented the second largest group within the newly arrived refugees in Germany (31 per cent couples with children, 6 per cent single parents; see Brü cker, Kosyakova and Vallizadeh, 2020). As the ReGES study focused only on families, some deviations were to be expected solely on the basis of this focus. However, the restriction to five federal states and the specific sampling could also have an impact. Syrian asylum seekers represented the majority in both ReGES samples, followed by refugees from Iraq and Afghanistan. These were also the three largest groups of origin in the SOEP and the main countries of origin of refugees in the years relevant for the sampling (Bundesamt fü r Migration und Flü chtlinge, 2020). However, while the Syrian group is overrepresented, other countries of origin are less represented in the  ReGES data. This composition of the countries of origin is due to the specific sampling of the ReGES study, which deliberately excluded countries from the West Balkan, for example (see Sampling and response rates section). Another peculiarity of the ReGES sample is that approximately 85 per cent of respondents were already living in private accommodations, which is significantly higher than that in the SOEP (see Brü cker, Kosyakova and Vallizadeh, 2020). Considering only refugee families in the SOEP with children of preschool age (3-6), the proportion of those who were already living in private accommodations also rises to over 70 per cent (see Homuth, Liebau and Will, 2021), but the proportion in collective accommodations is still almost twice as high as in the ReGES sample. This could indicate that in the ReGES study, access to collective accommodation did not work optimally. At least for Bavaria, there is evidence for this from the field (see Steinhauer, Zinn and Will, 2019). Compared to the SOEP sample (see Brü cker, Kosyakova and Vallizadeh, 2020), the parents in the ReGES sample achieved slightly better degrees. This is possible because the surveyed ReGES parents were, on average, somewhat older than the adult respondents in the SOEP. Additionally, the fact that ReGES respondents had already formed families may contribute to explaining these differences. Indeed, analyses of families with preschool children in the SOEP show that the educational distribution is similar to that of the ReGES samples (see Homuth, Liebau and Will, 2021).
Concerning educational participation, there are no significant differences between RC1 and refugee children of the same age in the SOEP (see Homuth, Liebau and Will, 2021). In RC2, due to the specific sampling in ReGES, direct comparisons of participation in education are not possible as adolescents who were not attending school were not part of the ReGES samples by design. However, analyses with the SOEP data also show that, at least after a certain period of residence, the majority of refugees under the age of 17 had entered the German educational system (see de Paiva Lareiro, 2019).

Panel Attrition
Selective bias can occur not only in the sampling and recruitment of study participants but also in the course of a panel study. The first step is the willingness to participate after the first wave. Analyses of the selectivity of panel participation (see Heinritz and Will, 2021) show a positive bias for higher-educated parents. This effect was much more pronounced in RC1 than in RC2. This is also expressed in the fact that in RC1, illiterate respondents were somewhat less willing or able to participate in the panel. While adolescents in RC2 were more likely to participate in the panel if they did not have a secure residence status, there was no such correlation in RC1. In both cohorts, Syrians were more likely to be included in the panel sample than refugees from Afghanistan, Iran, or other countries. This could be because Arabic and Kurmanji, two of the panel languages, are often represented in Syria (and in Iraq), whereas Pashto and Farsi, which are spoken in Afghanistan and Iran, could not be offered in the panel interviews. This explanation is also supported by the fact that this effect was less pronounced in RC2, in which the respondents participated in the panel in German far more often than in RC1. Once the respondents were part of the panel study, the selectivity was no longer as pronounced, although participants from more educated families had a slightly higher probability of continuing to participate in the course of the panel. Moreover, we observed strong mode effects. The samples of the online surveys were most selective since illiterate individuals and less educated families in particular had a lower probability of participating (see Heinritz and Will, 2021). 4 On the other hand, there is slight evidence that in RC1, people in collective accommodations were less likely to participate in CATI and CAPI. This mode-specific correlation was not found in RC2. Selectivity with regard to educational background was lowest in the face-to-face fieldwork (waves 4 and 7).

Content and Special Features of the Study
In the following, the contents of the ReGES study are presented. This section first describes how educational processes, educational outcomes, and returns to education were recorded before introducing the measurement of factors that can help to explain educational success. Finally, special features of the ReGES study are presented that enable additional in-depth analyses of the educational success of refugees. Over the course of the study, the contents were recorded identically between the two cohorts wherever possible, and age-or educational stage-specific aspects were added.

Recording of Educational Progress and Educational Outcomes
The core content in ReGES is the documentation of the educational biographies of RC1 children and RC2 adolescents. Thus, a detailed recording of the educational situation (including after-school childcare and language support) took place in the annual face-to-face surveys (see Figure 4, waves 1, 4, and 7). Therefore, detailed information on the situation of the ReGES children and adolescents in the German educational system is available at least once per school year. In addition, all educational episodes were recorded retrospectively beginning with their arrival in Germany (in the first interview) or with the last face-to-face survey (in the panel waves). To capture the complete educational situation, refugeespecific aspects were also taken into account, including whether students were attending newcomer classes and the availability of specific preparatory offers for refugees in vocational training. In the waves between the face-toface waves (with the exception of wave 2), brief status updates on the educational situation were also recorded.
In addition to the detailed documentation of the educational biographies, which make it possible to investigate refugees' educational participation and decisions, information on educational achievement was assessed (grades, educational qualifications, competencies). Moreover, returns to education were considered. Classic returns to education, such as the characteristics of the professional position, play a subordinate role for the children and adolescents in the educational stages of the ReGES study and only gained importance for the adolescent cohort in the course of the study. However, in situations in which the adolescents took up employment, various job characteristics were queried in ReGES. For most participants, however, the focus was on nonmonetary returns, such as health and satisfaction in various areas of life. In addition, indicators were measured that provide information on the cultural, social, and emotional dimensions of social integration (see Esser, 2000; for details on the indicators, see Will et al., 2018a).

Recording of Central Factors for Explaining Educational Success
Another central aim of the ReGES study was to provide data on factors that facilitate or hinder refugees' educational success. The starting point for the ReGES study was the notion that no new theories were needed to explain the integration processes of refugees and that developed theories in the context of integration research could also be applied to the integration of refugees (see FitzGerald and Arar, 2018;Kogan and Kalter, 2020). Transferred to educational research, this means that established mechanisms of social and ethnic educational inequality could also help to explain the educational success of refugees (see Edele et al., 2021). For this purpose, ReGES relied on the conception of the NEPS, a representative German sample (including migrants) that covers the entire life span and thus the educational stages that are relevant in ReGES.

Social Background
To capture the social background of the family, the parents' educational level and their socioeconomic status in the country of origin were recorded in detail. The recording also allowed a calculation of the selectivity of the migrants compared to the society of origin. Furthermore, the educational experiences of children and adolescents in the country of origin were recorded. To obtain information about the resources of parents in Germany, the parents' current employment and participation in educational offers were covered in RC1. To take a closer look at the effects of social background on education, additional constructs included variables ranging from educational aspirations and expectations to questions that captured knowledge about the German educational system and specific questions about the perceived costs, returns, and expected probabilities of success of various educational alternatives. In addition, the social capital of the newly immigrated refugees in Germany was recorded.

Migrant-Specific Mechanisms
In addition to known mechanisms of social inequality, it should be possible to test mechanisms of ethnic educational inequality using ReGES data. A particular focus was on the cultural dimension and the detailed recording of German language acquisition. Parents were asked about both their own and their children's German skills. The adolescents were asked to rate their own language skills. Participation in language training, the media used for language acquisition, and language usage in various areas were also queried. The area of social capital was supplemented by aspects of ethnicity so that the ethnic composition of the network could be taken into account. Different orientations, such as return orientation, identity, and religiosity, were also measured.

Personality and Motivation
Motivational aspects (intrinsic vs. extrinsic motivation, self-concept, vocational interests, meaning of work) as well as various aspects of personality that are relevant for educational processes were measured. These included, for example, prosocial behaviour, personality traits (Big Five inventory), willingness to take risks, patience, and tolerance of uncertainty.

Learning Environments
The focus of the interviews with parents and adolescents was on the detailed measurement of the family context (e.g., family climate) and the familial learning environment as well as selected aspects of the institutional learning context. More detailed information on contexts was collected through the questioning of contextual persons outside the family. Heads of educational institutions and teachers gave information on structural characteristics of the institutional learning environment, such as the social and ethnic composition at the institutional and group/class level as well as selected structural quality aspects of the institution. In addition, features of the regional context (e.g., educational offers in the municipalities) and collective accommodations were recorded by surveying the staff in the municipalities and collective accommodations.

Refugee-Specific Mechanisms
As most refugees have considerably different preconditions and live under exceptional conditions, questions on refugee-specific aspects were added. Key points of the escape history were elicited through questions that focused primarily on educational aspects such as duration, longer stays at transit locations and the educational opportunities there. An update of residence status was regularly queried, and information was obtained on current accommodations (private vs. collective accommodations). These objective characteristics were supplemented by subjective aspects, such as the perceived security of one's residence status, perceived refugee-specific support and perceived discrimination. The recording of the general state of health was supplemented by further aspects. These included, for example, the measurement of stress, symptoms of post-traumatic stress disorder (PTSD) and a PTSD diagnosis. As the importance of refugees' resilience has been repeatedly emphasized in the literature on refugee migration (see, e.g., Fazel et al., 2012), a measure of resilience was included. Figure 4 presents an overview of the recording of the individual constructs in the course of the panel as well as the rough allocation of the constructs to the various modes used.

Special Features of the ReGES-Data
Due to the main aims of the ReGES study, the data have several special features that allow for in-depth analyses of several mechanisms of integration into the German educational system.

Comparison with Other Data Sets
The central concepts of ReGES were selected with possible comparisons in mind, namely, two very important studies in Germany: the NEPS and the SOEP. ReGES is based on the conceptual design of the NEPS, and core constructs of the NEPS can be found in ReGES. Depending on the specific research question, RC1 could be compared to the NEPS Starting Cohort 1 (Newborns) or Starting Cohort 2 (Kindergarten), and RC2 could be compared to the NEPS Starting Cohort 3 (Grade 5) or Starting Cohort 4 (Grade 9). In addition, some refugeespecific aspects were surveyed in ReGES that were not included in the NEPS. Whenever possible, an attempt was made to keep these items identical or at least comparable with the SOEP so that comparative data analyses with the two datasets were possible. This involved, for example, the recording of educational qualifications in the country of origin, regions of origin, and educational aspects such as reasons for not attending preschool. This procedure has the advantage that refugees in ReGES can be compared with newly arrived refugees in the SOEP dataset and with migrants who have been living in Germany for some time (including refugees) and the majority population (see, e.g., Homuth, Liebau and Will, 2021).

Recording of Regional and Institutional Contexts
The inclusion of different contexts, in which refugees are living and learning in, is central to understand refugees' educational trajectories. The additional information provided by persons in regional and institutional contexts allowed us to analyse the impacts and interdependencies of macro-, meso-, and microlevel factors on the educational trajectories of refugees in different age groups. In the first wave, the survey of administrative staff in refugees' residence districts and the survey of group accommodation staff provided information on the broader living conditions of refugees with information on the refugee population in the district, group accommodations and relevant infrastructure. Other important aspects were the yearly surveys of the heads of educational institutions and teachers of the educational organizations (preschools; primary, secondary, and vocational schools) that the refugee children and adolescents attended. These surveys provided information on the structural aspects of the organization and classes (e.g., the share of refugees and other compositional aspects), resources and programmes to support immigrants in general and refugees in particular (especially language training). Furthermore, an individual performance evaluation was conducted for every ReGES student by the class teacher (with a focus on socioemotional aspects, school achievement, and language competencies).

Assessment of Competence Data
Another special feature of ReGES is the assessment of objective test data from the children and adolescents at two measurement points: a baseline measurement in wave 1 and an outcome measurement two years later in wave 7. The data on language competencies comprise German receptive language competencies, including both vocabulary and grammar. In addition to language, domain-general cognitive abilities were assessed as control variables, including perceptual speed (NEPS-BZT) and figural reasoning (NEPS-MAT) (Lang et al., 2014). All tests that were used had previously been established in the NEPS. However, age groups and test modes are not always identical, so the use of the NEPS data as a benchmark for the ReGES test results must be weighed for the specific research question.
The objective language test data were supplemented by an external assessment of the refugees' language skills (via parents in RC1 and teachers for both cohorts) and a fine-grained rating of subjective language competence in RC2. For this assessment, adolescents were asked to rate not only their competencies in understanding, speaking, reading, and writing in German but also their competencies in a list of given tasks (e.g., introducing themselves, understanding a simple newspaper article, or reading literature). The availability of objective test data as well as different external and subjective ratings opens up vast research potential, for example, on deviations between the different measurements. Moreover, the effects of competencies on educational decisions and integration can be analysed in great detail. In addition, language skills are an important indicator of the cultural dimension of social integration and of educational success. Information on language skills was supplemented by a detailed recording of language training within and outside of educational institutions, which allowed for analyses of the effectiveness of the provision of different services.

Linkage of Regional and Register Data
Two explicit linkage possibilities of the ReGES data are planned. The first extension is the possibility of linking regional data, such as the proportion of migrants or social assistance recipients in the residential area. In this way, it is possible to analyse how factors at the macro level influence refugees' educational integration. 5 The second extension is the possibility of linking the ReGES survey data to the participants' social security data in the form of the Integrated Employment Biographies (IEB) of the Institute for Employment Research (IAB) of the German Federal Employment Agency. In RC1, this applies to the parents, and in RC2, it applies to parents and the adolescents themselves as soon as they turn 18. Approximately, 85 per cent of the parents (RC1 and RC2) and adolescents (RC2) consented to the data linkage.

Data Access
The ReGES data will be available to the scientific community free of charge starting in June 2021 with the release of the first two waves under doi: 10.5157/ReGES: RC1:SUF:1.0.0 and doi: 10.5157/ReGES:RC2:SUF:1.0.0. Further waves will be added within the next 2 years at regular intervals. Access to the data in the form of scientific use files is provided in three ways based on the sensitivity of the data: (i) secure download via the website of the LIfBi Research Data Center, (ii) in a remote data access environment, and (iii) onsite in Bamberg. A valid contract between the data user and LIfBi is mandatory to access and use the data. Due to the high sensitivity of the IEB data, it is only possible to access this linked data product onsite at the IAB.

Conclusion and Outlook
Both ReGES datasets provide the potential to analyse processes that refugees experience in the German educational system and refugees' integration into German society. To obtain deeper insight into the underlying mechanisms of the educational processes of refugees, aspects of social educational inequality are supplemented with aspects of ethnic educational inequality with a focus on the role of the German language as well as motivation and personality aspects. Furthermore, information concerning educational programmes is included, such as programmes for refugee integration as well as more general language training offers within educational institutions, ranging from preschool to general schools and vocational training. With a multiinformant design, the data provide options to analyse the complex interrelation between refugee children and adolescents, their family environment, educational institutions and further informal supportive structures within communities being built for refugee support.
ReGES has proven to be a rather successful endeavour. Of the children and adolescents in the panel, n ¼ 2,111 (RC1) and n ¼ 2,059 (RC2) participants remained in the sample after the last measurement in 2020. As ReGES focused on the initial phase and medium-term integration processes of refugees in the German educational system, information on long-term educational trajectories is still missing. Therefore, a new BMBF-funded project, Bildungswege von geflü chteten Kindern und Jugendlichen ('Educational trajectories of refugee children and adolescents'), will accompany the ReGES participants of RC1 during their next steps in primary and early secondary school and the participants of RC2 on their way through upper secondary school, vocational education, or tertiary training. New survey and assessment waves are scheduled for 2022 and 2024 with intensive measures for panel care in the intervening time.

Funding
The project is funded by the German Federal Ministry of Education and Research (BMBF) under grant number FLUCHT03. The content of this publication is solely the responsibility of the authors.

Notes
1 The ReGES project is based at the Leibniz Institute for Educational Trajectories (LIfBi), and the surveys are conducted by the infas Institute for Applied Social Science. 2 Although all analyses in this article are based on final data, minor changes may still occur until the SUF is published due to data cleaning and participants' requests for data deletion. This particularly applies to the later waves of the ReGES study. 3 The response rate (RR2 according to AAPOR standards) relates to all cases that belong to the population, while the cooperation rate (COOP4 according to AAPOR standards) indicates the proportion of interviews conducted given that the family belongs to the population and could be contacted personally (see American Association for Public Opinion Research, 2016). 4 The first short CAWI in RC1 is an exception. Here, illiterate people have an increased probability of participating. This may be because the interviewers were asked to install the panel app together with the respondents at the request of the respondents. Illiterate people may have made use of this more often, and it cannot be ruled out that the short CAWI was also conducted jointly with the interviewer directly after the installation. 5 For an example, see Homuth, Liebau and Will (2021), who show that daycare and preschool attendance rates of refugee children positively correlate to regional daycare and preschool supply.