DeZIM.panel – Data for Germany’s Post-Migrant Society

1 Introduction

In recent years, most Western European societies have become increasingly diverse due to the ongoing immigration and naturalisation processes. Even though politics and society now mostly acknowledge the reality of post-migrant states in which immigration is an inevitable feature (Foroutan 2019), data collection infrastructures still have to catch up. Multi-wave access panels that allow researchers to analyse current trends in society, while still allowing for specific subgroup analyses, are still scarce.

The DeZIM.panel has been set up as such a data collection infrastructure that explicitly takes today’s post-migrant society into account. It started at the end of 2021, building on a large-scale offline-recruited first wave earlier in 2021 with more than 9000 participants, oversampling specific immigrant-origin groups: people from Turkey and from other majority Muslim countries, from states with guest worker agreements and re-settlers from the East.

The DeZIM.panel is a multi-thematic survey that includes topics relevant to sociology, political science, psychology, educational sciences, economics and other disciplines. It runs four regular waves per year, with additional flash surveys after current events such as the war in Ukraine (Mayer et al. 2022). The DeZIM.panel was initiated by the German Center for Integration and Migration Research (DeZIM), which is funded by the German Federal Ministry.

2 Aim of the DeZIM.panel

The general aim of the DeZIM.panel is to function as a multi-thematic, multi-wave access panel that allows researchers to track the impact of external events, as well as to track short- and long-term changes in public opinion of different societal groups with a specific regard to minorities, immigrants and their descendants. This allows us to capture both new developments and current trends. It also provides data of both the majority population and immigrant-origin groups, which is of specific interest to those studying minorities and immigrants. Such data include integration indicators, discrimination experiences, xenophobic and racist attitudes, and national and ethnic identities.

There are certainly panels that are already established in Germany that offer high-quality data based on random samples, which also often include at least subsamples with an oversample for certain groups such as refugees (e.g. IAB-BAMF-SOEP refugee sample, Brücker et al. 2016). However, as we intend to capture the impact of external events as well as short-term changes, we need a multi-wave panel that is not limited, as is the Socio-Economic Panel (SOEP), to running only once per year. As the DeZIM.panel should specifically add to the study of immigrant integration and societal conflicts, we also need indicators that capture concepts such as discrimination experiences or identities in a multi-faceted way that go beyond the scope of other multi-wave panels, such as the GESIS Panel (Bosnjak et al. 2018) or the German Internet Panel (GIP; Blom et al. 2015), neither of which focuses on these topics specifically.

Of course, commercial online access panels are available in Germany, which sometimes have at least 1000 persons with migration history in their databases and which may be used for specific subgroup analyses. However, these commercial panels do not rely on random samples and thus cannot be used for generalized statements. Although quotas might be applied for the general population to refer to the data as national samples, none of the larger online access panels offers this for the group of immigrants, as their numbers are too low to apply quotas.

Our chosen research design – the offline generation of an online panel – has been used before by the GESIS Panel (which provides ongoing possibilities for offline participation) as well as by the GIP (which also provides options for offline participants). So far, relying on offline recruitment for online panels offers the least amount of bias, as individuals’ addresses can be randomly drawn. However, the next step, going online, might still add bias because not every household might participate online, and the unit non-response is systematically distributed.^[1] To account for these biases, different approaches exist – such as providing an internet equipment or an alternative access form – with evidence that points in the direction that offering alternative access is the most fruitful (Cornesse and Schaurer 2021). In addition, we thus intend to survey by post those initial participants who never provided an email address. This will allow us to gauge the differences between online and offline participants regarding key indicators (e.g. socioeconomic status, migration background, age, education, gender).

3 Research Design and Oversampled Groups

The target population of the DeZIM.panel includes all persons with and without an immigrant background living in Germany within the age group 18–67. This age range was chosen because it matches the working-age population and reduces the need to account for many offline participants. Given the aim of the DeZIM.panel to oversample respondents from minority or immigrant background groups, we decided to employ a two-stage stratified sampling approach that uses name-based pre-classifications to oversample certain subgroups.

We focused on four specific groups. First, respondents from Turkey because Turkish migrants make up the biggest group of immigrant-origin people in Germany. Second, we included people from other majority Muslim countries, enabling us to draw a broad picture of Muslim experiences in Germany. These countries were Afghanistan, Algeria, Bahrain, Egypt, Iran, Iraq, Jordan, Kosovo, Kuwait, Lebanon, Libya, Morocco, Oman, Pakistan, Qatar, Saudi Arabia, Syria, Tunisia, United Arab Emirates, and Yemen. Next, we oversampled people from states with former guest worker agreements (besides Turkey) because this group has a distinct migration history as working migrants and is a large part of the German community of people with migrant backgrounds. These states include Greece, Italy, Portugal, and Spain, as well as three of the successor countries of the former Yugoslavia: Bosnia and Herzegovina, Croatia, and Serbia. Last, we included another major group with a distinct history: re-settlers from Eastern Europe whose ancestors migrated in the 18th century to the then-Tsarist Russia and to other territories, and whose descendants mostly returned to Germany in the early 1990s. These respondents emigrated from the successor states of the former USSR as well as from Romania. All other people of immigrant-origin were not oversampled but were simply included as a fifth group of people with migrant backgrounds. For the setup of the panel and the pre-classification, we classified people from Poland as belonging to this fifth group even though re-settlers might have migrated back from Poland. Nowadays, the group of people of Polish origin in Germany consists mainly of non-ethnic Germans. It is important to keep in mind that these groupings are only preliminary, and we are of course able to group respondents later based on their actual answers to questions regarding their country of origin, ethnicity and religiosity.

4 Initial Recruitment Process

5 Panel Design and Survey Programme

In the future, we intend to field four panel waves per year with a median length of 25 minutes. In addition to these regular waves, we also have the survey format of the so-called short surveys (“OAP Kurzbefragungen”) with the median length of approximately 3–5 minutes and a field time of 7 days, with one reminder in between. This allows us to respond in a timely manner to important current societal and political issues.

The general idea of the DeZIM.panel was the offline recruitment of respondents with a switch to online-only for the subsequent waves, i.e. inviting respondents only via email and conducting the fieldwork via a web survey only. However, only about 52% of respondents with panel consent provided us with a valid email address at first. We, thus, contact the respondents for the regular waves via email or via post, with up to four online reminders and one postal reminder over the field time of the survey. We specifically ask respondents without an email address at the end of the survey to provide us their addresses. In the first wave, this procedure resulted in 1220 new email addresses after the survey.

Respondents who participate in one of the regular online access panel waves receive a €10 voucher for several online stores they can choose from.^[3] Furthermore, from wave 2 onwards, we will also offer the possibility of a bank transfer of the incentive. In order to increase the response rates at the end of the field time, we also implemented a lottery in the first panel wave in which, on each day over a 5-day period, all previous participants could win a €200 voucher. A similar lottery is organized for the short surveys: of all the participants, 20 could win a €100 voucher.

To date, two regular waves have already been conducted: the first panel wave from the end of November 2021 to the end of January 2022, and the second wave from the end of March to the beginning of May 2022. The next waves are planned for June to July 2022, for September to October 2022 and for December 2022 to January 2023.

The first short survey was launched on February 28, 2022, as a reaction to the escalation of the war in Ukraine that had started 5 days earlier. This survey included questions on affective reactions, preferred policies and engagement preferences, and attribution of responsibilities (Mayer et al., 2022). In total, 3682 respondents participated in wave 1 (completion rate = 55%)^[4] and 2663 respondents participated in the first short survey. In the short survey, respondents were invited to participate only by email (completion rate = 54%); the total number of participants are shown in Table 7.

In the future, we intend to field four panel waves per year. Each wave will include one of our four core modules. These remain largely the same every year: questions on current topics and scientific questions submitted by other researchers from the DeZIM Institute and Research Community. Prospectively, we also intend to open the submission process to other external researchers.

The four core modules focus on political institutions, attitudes and behaviour in wave 1, on societal values and societal norms in wave 2, health and well-being in wave 3 and labour and education as well as discrimination experiences in wave 4. As we intend to have the same questions in the core modules, we are thus able to track trends over time. Questions about subjective well-being and the assessment of individual life situations are asked in each wave. Furthermore, we include scientific concepts submitted by other researchers such as affective polarization or racism. In addition, we ask questions about current topics, such as the Corona pandemic and the war in Ukraine. All questions should be taken either from other established surveys or should be pre-tested. They also need to be able to be given to the whole panel instead of only one or several subgroups, so that all answers are available for all respondents.

6 Composition and Data Quality of the DeZIM.panel

In order to assess possible biases during sample selection and fieldwork, we compare the distribution of our sample on specific central characteristics with the distribution of these characteristics in the German Microcensus, as displayed in Table 8.

The results show that from recruitment over panel consent to actual participation in wave 1, bias towards higher education and younger participants increases. With regard to gender, no conclusive patterns emerge. Conducting multivariate logistic regression on the likelihood of not participating in the panel or in wave 1 shows that besides age, education and gender, migrant status has a significant positive relationship. In the recruitment wave, the likelihood of not participating is 8% points significantly higher for immigrants than for natives (however, no substantial difference between natives and children of immigrants was found). In the first wave, it was about 14% points significantly higher for immigrants (compared with natives), and 5% points for children of immigrants (compared with natives, all p < 0.05). The results show the need to account for non-response characteristics with weights.

Data should be weighted for analyses, due to the differences in the selection probabilities and stratification of the groups. We, thus, offer a design weight that accounts for different selection probabilities [dweight], as well as a combined weight also accounting for unit-non-response in the recruitment sample (based on age groups, gender, federal state and size of municipalities [BIK]) [dweight_adj]. In future, we plan to provide specific panel weights to account for non-response between the waves.

7 Analytical Potential

Currently, the DeZIM.panel is the only access panel in Germany with an oversampling of immigrants and their descendants that offers a specific focus on topics relevant to integration research.

The DeZIM.panel allows the analysis of changes and trends over time by comparing different groups of respondents. Due the longitudinal survey design, the DeZIM.panel enables researchers to capture short- and long-term changes for individuals and between groups. Furthermore, the impact of external events with relevance for immigrant-origin groups – such as the war in Ukraine – can be studied with the help of flash surveys. Our long-term monitoring also makes it possible to detect whether – and how – policy changes, such as bans on head scarves or changes to immigration policy, affect the German population, and whether these effects are the same for all groups, while still taking individual factors such as socio-economic status into account.

For more in-depth analyses, contextual data on NUTS-2 or NUTS-3 (“Nomenclature des Unités territoriales statistiques”), such as employment rates or the proportion of refugees in the population, can be merged with the survey data based on official municipality keys (“Amtlicher Gemeindeschlüssel”), and the consequences of social contexts can be analyzed with multi-level analysis.

8 Data Security and Access

After a certain embargo time, the data from the DeZIM.panel will be made available to the scientific community within the DeZIM.fdz, the research data centre of the DeZIM institute (https://fdz.dezim-institut.de/). Because many questions touch sensitive topics such as health status, immigration history or political attitudes, the data need a high level of confidentiality and protection. In order to comply with strict national and international standards, a bundle of data security measures and a flexible data access plan have been established. We, thus, offer several different ways to access the data, with hierarchical levels of information sensitivity based on various levels of anonymization. Currently, Scientific Use Files (SUF) are available for secure download and secure on-site access. In future, we will also provide Campus Use Files with a subset of all cases and easier access specifically for academic teaching and for offering secure remote access. The SUF download version of the data has the highest level of anonymization, i.e. age groups and countries of origin are, for example, highly aggregated, and text answers are removed from the data set. The SUF are provided in Stata format with German and English labels and English documentation (questionnaire, codebook and method report). The SUF on-site version include more fine-grained details, and they also allow the addition of contextual data.

For any access to the DeZIM.panel data, the users have to first register for an account and then submit a signed user agreement (available in German and English). All users have to be affiliated with a scientific institution or a state organisation. Their applications need to include a brief description of the research project for which the data are intended to be used, the project time frame, and the other collaborators on the project. After the application has been approved, either the available SUF can be downloaded free of charge, or the data can be accessed on-site or from the secure platform.