MIFARE Survey. Migrants Welfare State Attitudes. Methodological Report. Hidde Bekhuis. Troels Fage Hedegaard. Verena Seibel.

MIFARE Survey Migrants Welfare State Attitudes Methodological Report Hidde Bekhuis Troels Fage Hedegaard Verena Seibel Daniel Degen

MIFARE Survey Migrants Welfare State Attitudes Methodological Report Hidde Bekhuis, Troels Fage Hedegaard, Verena Seibel & Daniel Degen In cooperation with Marcel Lubbers, Claudia Diehl, Christian Albrekt, Theresa Kuhn Lancee, & Jeanette Renema Nijmegen, January 2018 2

PROJECT PERSONAL Principal Investigator Prof. Dr. Marcel Lubbers Radboud University Nijmegen Netherlands Co-Applicants Denmark Prof. Dr. Christian Albrekt Aalborg University Germany Prof. Dr. Claudia Diehl University of Konstanz Team Members Dr. Troels Fage Hedegaard Aalborg University Daniel Degen, Msc. University of Konstanz Dr. Verena Siebel University of Konstanz Dr. Theresa Kuhn Lancee University Of Amsterdam Netherlands Dr. Hidde Bekhuis Radboud University Nijmegen Jeanette Renema, Msc. Radboud University Nijmegen 3

TABLE OF CONTENTS 0. BEFORE USING THE DATA 5 1. INTRODUCTION: PURPOSE OF THE MIFARE STUDY 6 2. RESEARCH DESIGN 7 2.1 Research design and selection of countries and groups 2.2 Sampling strategy, sampling method and sampling rates 2.2.1 The Netherlands 2.2.2 Denmark 2.2.3 Germany 2.3 Survey methods 3. SURVEY INSTRUMENTS 12 3.1 Topics of the questionnaires 3.2 Challenges of translation and national adaption 3.3 Questionnaire for natives 3.4 Pre-test 4. FIELDWORK 19 4.1 Time frame of the fieldwork 4.2 Coordination of the fieldwork 4.3 Strategies to increase response rate 4.3.1 The Netherlands 4.3.2 Denmark 4.3.3 Germany 5. RESPONSE RATES AND SELECTIVITY 22 5.1 Overview of response rates across all countries 5.2 The Netherlands 5.3 Denmark 5.4 Germany 6. DATA PROCESSING 37 6.1 Data cleaning 6.2 Variables in the data set 6.3 Constructed variables 6.4 Open-ended questions 7. QUESTIONNAIRE 43 8. REFERENCES 77 A1. APPENDIX A: PILOT QUESTIONNAIRE (separate file) A2. APPENDIX B: QUESTIONNAIRES (separate file) A3. APPENDIX C: CONSTRUCTION OF ISCED VARIABLES (separate file) 4

0. BEFORE USING THE DATA The data can be used for free by all those who are employed at a university, an academic research institute, or another non-profit organization. For others, permission must be obtained before using the data. To obtain permission, contact the research team at m.lubbers@maw.ru.nl. When using the data set, always include the following citation to the data set: Bekhuis, H., Fage Hedegaard, T., Seibel, V., Degen, D. & Renema, J. (2018). MIFARE Study Migrants Welfare State Attitudes. Dataset. DANS (Data Archiving and Network Services). KNAW. When using information from this report, cite the report as: Bekhuis, H., Fage Hedegaard, T., Seibel, V. & Degen, D. (2018). Design and content of the MIFARE Study. Methodological Research report. Radboud University Nijmegen, Netherlands. 5

1. INTRODUCTION: PURPOSE OF THE MIFARE SURVEY The MIFARE project ( Migrants Welfare State Attitudes ) is a comparative survey among immigrants in Europe which focuses on welfare state attitudes. The MIFARE project is funded by the NORFACE research programme Welfare State Futures. Coordinated by Marcel Lubbers (Radboud University Nijmegen), the MIFARE project has been conducted by researchers at Aalborg University, University of Amsterdam and University of Konstanz. The MIFARE study is the first cross-national survey that focuses on immigrants attitudes towards the welfare state. In Europe, the field of research on welfare state attitudes has paid little attention to the perspective of immigrants. Due to migrants socialization in different welfare regimes, and their often disadvantaged socio-economic positions, the immigrant perspective provides a unique opportunity to test the central theories in the field on the role of self-interest (Andreß & Heien, 2001; Gelissen, 2002; Van Oorschot, 2006; Jaeger, 2006b; Svallfors, 2012), group-loyalty (Esser, 2009; Maliepaard, Lubbers & Gijsberts, 2010) and of socialization in different welfare regimes (Esping-Andersen, 1990; Jaeger, 2006a; Larsen, 2008; Jaeger, 2009; Van der Waal et al., 2013). The MIFARE study aims to study immigrants welfare state attitudes, and to explain differences across migrant groups, as well as differences compared to the overall public opinion in the country of origin and the host country. In order to study migrants welfare state attitudes, and to explain differences across migrant groups new data are collected. The questions used in this new survey are partly based on the ISSP 2006 questionnaire Role of the government, the ESS 2008 questionnaire Welfare state attitudes and new questions which were piloted first. This document describes the data collection, the representativeness, the questions and response of the MIFARE survey. 6

2. RESEARCH DESIGN The MIFARE survey was designed to focus on immigrants who migrated to the receiving country at an age of 16 years or older, from different origin countries across three European countries. Efforts were taken to harmonize the data collection across the three countries under study. In this chapter the research design and the rationale for the selection of countries and groups included in the study is described (section 2.1). The sampling strategy and the method of drawing the sample are presented separately for each country (section 2.2). The survey mode is addressed in more detail in the last section of this chapter (section 2.3). 2.1 Research design and selection of countries and groups The MIFARE survey has been conducted in three countries: Denmark, Germany and the Netherlands. All three countries have the opportunity to sample from population registers, including immigrants. The opportunity to sample randomly from the registers enables us to test for representativeness of the survey, to approach migrant groups that are smaller in number, and guarantees comparable designs in the three countries. We proposed to sample immigrants from the age of 18 and older, and a native control group (to be able to compare between migrants and natives also for the questions specifically developed for the proposed survey). We chose 4 intra-eu origin countries and 6 extra-eu origin countries, including the most numerous migrant populations: (first generation) immigrants from Poland, Romania, Spain and the UK for the intra-eu origin countries. As for the extra-eu origin countries, we selected China (mainland only, excluding Hong Kong), Japan, Turkey, the Philippines (not in Germany due to sampling issues), Russia, and the US. China and Turkey are the only countries not included in either the ISSP or ESS when the welfare-state attitudes rounds were conducted. 7

2.2 Sampling strategy and sampling method The sampling method chosen in the three countries depended on the national data sources available to identify immigrants. Although data from local or central registry offices could be used in all countries, there are differences in the possibility to select on migration age, which affects the sampling strategy. Since tailored sampling strategies were used in each country, the sampling procedure is described in separate sections for the Netherlands (section 2.2.1), Denmark (section 2.2.2) and Germany (section 2.2.3). 2.2.1 The Netherlands In the Netherlands, migrants who stay for longer than 4 months are required to register at the municipality. Statistics Netherlands (CBS) sampled the immigrants from these municipality based registrations. The aim was to have 300 questionnaires filled out per immigrant group and natives. Based on other surveys in the Netherlands, Americans, Brits, Dutch natives, Japanese and Spaniards were expected to have a response rate of 33.3%. While Chinese, Philippines, Polish, Romanian, Russian and Turks were expected to have a lower response of 27%. These expectations resulted in a sample of 900 immigrants from the first mentioned groups, and a sample of 1100 from the last country of origin groups listed. The sample for the survey is a stratified sample, one stratum for the native Dutch and one stratum for each of the ten migrant groups. From each stratum a simple random sample without replacement was drawn. Since people between 18 and 75 years were selected it was possible that immigrants from the Soviet Union were selected. Statistics Netherlands had also information on the area in which migrants from the former Soviet Union had lived. Only immigrants from the Soviet Union who lived in what is now Russia are selected. From the sample that was drawn, Statistics Netherlands successively requested the names and addresses from the National Identity Data (RvIG). When it turned out that a selected 8

person at an address had already been approached for a regular CBS survey in the last year, then this person was removed from the sample. If a person lived at an address of an institution then the person was also removed from the sample. The sample was drawn on October 23th 2015, 2 weeks before the first invitations were sent. Table 2.1 provides an overview of the group sizes in the Dutch population and in target population size in the MIFARE sample. Table 2.1: Natives and immigrants in the Dutch population and target population in the Dutch MIFARE sample 18-75 year olds in Dutch population on 1-1-2015 18-75 years old in MIFARE sample Native Dutch 9,616,462 900 China 42,891 1100 Japan 4,711 900 Philippines 11,026 1100 Poland 96,380 1100 Romania 15,046 1100 Russia 1 56,438 1100 Spain 19,780 900 Turkey 183,915 1100 UK 39,808 900 US 18,804 900 1 The sample of migrants from Russia is drawn from the population with a Russian origin; not all may have been born in Russia. Additional information about these migrants is used to select only those who were born in what is now called Russia. 2.2.2 Denmark All immigrants who have stayed in Denmark for more than three months, and have a permanent housing, can apply for the status of living in Denmark in the Civil Registration System (Det Centrale Personregister or CPR-register in Danish). This is unless they have a residence permit, which is necessary for migrants from some countries, in which case status of living can be applied for from the first day. The Danish respondents were sampled, using the Civil Registration System among Danes, and ten migrant groups. For the Danes the sample was drawn randomly from all who are born in Denmark, both parents are Danish citizens, at least 18 years old and are living in Denmark. For the migrant groups the sample was drawn randomly based on the following criteria: 9

The respondent must be born abroad, in one of the ten selected countries The parents of the respondent cannot be Danish citizens at the time they migrated to Denmark, that is, when the respondent obtained living status. The respondent must be at least 16 years old when they migrated to Denmark The respondent must have lived in Denmark (living status) for at least 12 months. The aim was to have 300 filled out questionnaires per migrant group and natives. Based on an expectation of higher non-response in some groups 900 Danes, Americans, Brits, Japanese and Spaniards where sampled, while 1000 Chinese, Philippines, Polish, Romanian, Russian and Turks where sampled (Font & Méndez, 2013). Only migrants from the Soviet Union who lived in what is now Russia are selected. Similar to the sample from the Netherlands this is also a stratified sample drawn separately among the natives, in this Danes, and the ten migrant groups. The sample was drawn in October of 2015 and therefore less than a month before fieldwork begun. Table 2 provides an overview of the group sizes in the Danish population and in the targeted MIFARE sample. For the Danish sample the name and address was provided. Table 2.2 shows that the sample sizes and the total population in some cases are not very different, e.g. among the Japanese immigrants. Furthermore, the total populations reported below are too large, in the sense that list drawn from Statistics Denmark online database cannot include all the selection criteria listed above. The result of this is that we sampled almost all Japanese who fit the criteria, and large parts of the Spanish and Russians living in Denmark. 10

Table 2.2: Natives and migrants in the Danish population and in the targeted Danish MIFARE sample Population Sample Native Danes 3,963,422 900 China 9,521 1,000 Japan 1,373 900 Philippines 9,690 1,000 Poland 31,561 1,000 Romania 17,532 1,000 Russia 5,047 1,000 Spain 4,783 1,000 Turkey 31,537 1,000 UK 12,543 900 US 7,243 900 Note: Based on FOLK2 from Statistics Denmark online database, for 1 of January 2015. Definition: Migrants are born abroad. No parents are Danish citizens born in Denmark. If there is no information on any of the parents, and the person is born abroad, the person is counted as a migrant. Note that is not fully identical with the sampling criteria described above. 2.2.3 Germany In Germany, migrants are required to register after 2 months of stay at their municipalities. Registration data are not available on national level but have to be acquired from each municipality separately. No information is provided for date of migration which led to oversampling in order to sample a sufficient number of migrants who migrated after the age of 18. To get a sample that is most likely to represent the whole population of the respective migrant (and native) group, we decided to cluster communities according to their size. Since we do not want respondents from bigger cities exclusively, we also added smaller communities. We decided to divide all German communities into four clusters (<50k; 50k-100k; 100k-500k; >500k). In a second step we used data from the Mikrozensus 2011, which was the newest dataset when it comes to individual characteristics for all German communities. 11

To make our sample representative, we decided to sample only in those communities of a certain size, where at least 10% of a migrant group is living (e.g. Chinese only from communities with at least 100000 inhabitants). Table 2.3: Distribution of migrant groups in Germany, by size of municipality Community Size Poland Romania Spain Turkey UK Russia USA Japan China Germany <50k 27.5% 39.1% 0.0% 26.4% 0.0% 42.5% 0.0% 0.0% 0.0% 48.9% 50k - 100k 16.4% 10.5% 0.0% 13.8% 0.0% 15.8% 0.0% 0.0% 0.0% 12.1% 100k - 500k 29.5% 32.7% 0.0% 28.8% 0.0% 22.9% 0.0% 0.0% 17.8% 20.4% >500k 26.7% 17.7% 100.0% 31.0% 100.0% 18.8% 100.0% 100.0% 82.2% 18.6% TOTAL 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% 100.0% Source: Microzensus 2015 Table 2.3 shows how many immigrants of each group should be sampled in the respective community size. We decided to sample those groups proportionally that the sample represents whether a group is rather living in big communities or not. Hence, around 27,5% of our Polish sample should be drawn from municipalities who count less than 50 000 inhabitants. For each group, we want to have 3 communities per group. We decided to choose those communities where most people of each group are living. In cases, where we were not able to obtain data we chose the community which was next on the list. After obtaining the data from the communities, we sampled the respondents proportionally to their group size in the respective city. Also, we limited the sample to communities who are in West Germany and excluded communities which are part of the former German Democratic Republic. Furthermore, we had to consider that respondents end up in the sample that have not migrated before the age of 18 or even second generation immigrants. Therefore, we oversampled for each community with the respective factor of having such respondents in the sample. In a final step to decide the group size, we oversampled certain groups due to previous research indicating a lower response rate such as migrants from Turkey or Poland. 12

Table 2.4: Natives and migrants in the German population and in the targeted German MIFARE sample 18-75 year olds in German population 2015 (in 1000) 18-75 years old in MIFARE sample Native German 81,700 900 China 87 1020 Japan 20 1602 Poland 800 1560 Romania 340 1133 Russia 1 514 1620 Spain 68 1279 Turkey 642 2051 UK 55 1114 US 55 1316 1 The sample of migrants from Russia is drawn from the population with a Russian origin; not all may have been born in Russia. Additional information about these migrants is used to select only those who were born in what is now called Russia. 2.3 Survey mode Postal surveys were conducted in all countries. Respondents were invited by mail to participate in the survey. The invitation letter was bilingual; in the language of the country of residence and in the language of the country of origin of the respondent. Respondents could participate by filling out a written questionnaire which was sent with the invitation. Or respondents could participate via Computer Assisted Web Interviews (CAWI ); the interviewee completed the questionnaire supplied to them via a website link with an unique login code. 13

3. SURVEY INSTRUMENTS In order to collect comparable data across the four countries the team developed a harmonized survey instrument. Apart from a few questions that were asked exclusively in selected countries or for specific ethnic groups, identical questions were asked in the Netherlands, Denmark and Germany. Many of them were adopted from established survey instruments, such as the International Social Survey Program, or the European Social Survey. This facilitates comparisons with other studies. Since it was anticipated that migrants were not all able to fill out a questionnaire in the receiving country language, the questionnaire was translated into immigrants native languages. It was assumed that this would reduce the non-response and would increase the accuracy of answers. This procedure included processes of translation and re-translation for testing the correct meanings of the questions in different languages. In the next sections, the topics in the questionnaire are discussed (section 3.1), the variation in the questionnaire between the Netherlands, Denmark and Germany (section 3.2). Section 3.3 describes how the questionnaire for native differs. Finally, section 3.4 outlined the pre-test which was done in Germany. 3.1 Topics of the questionnaire The questionnaire covered a broad range of topics: A. Demography and migration biography B. Government responsibilities C. Household and Health D. Language and Contacts E. Religion F. Media use and Political attitudes 14

G. Knowledge and opinions about welfare state use H. Experiences with residence country I. Education and employment J. Household assets In the first module, the respondents were asked about demographic characteristics and questions about their migration biography. The second module asked respondents opinion about the role of the government in the receiving country. The third module covered respondents household situation in the receiving country, respondents health, the health of their household members and relatives and the experience and satisfaction with health care services in the receiving country. The fourth module addressed language proficiency and social contacts with natives, other immigrants and people from the country of origin. In the fifth module, the respondents were asked about their religious affiliations, beliefs and practices. The sixth module covered respondents media usage, both from the country of origin and receiving country. As well as respondents political attitudes, including party preference, opinion about taxes, the right migrants should have to vote and the role the EU should have. In the seventh module respondents were asked about their actual knowledge about the welfare state rights migrants from their country of origin have, as well as the rights of EU and non-eu migrants have. In addition to the actual knowledge also the opinion about these rights were asked. In the eighth module topics about discrimination experiences and the role migrants have in the receiving country were covered. The ninth module looked into migrants, and their possible partner s, education and employment history. Finally, the tenth module examined migrants earnings, and the use of welfare state benefits in their household. On average, the fill out a survey lasted about 25 minutes. 15

3.2 Challenges of translation and national adaption The master questionnaire was constructed in English. Native-speakers, hired via the translation agency VVH translations translated the English version into the different immigrant languages. To control the quality of the translations, re-translations were performed by the translation agency. During the process of constructing, translating, and pretesting the questionnaire, further national and group specific adaptations were made. A major challenge was related to the different welfare state arrangement in Denmark, Germany and in the Netherlands. Some welfare state arrangements were present in one of the countries only, for example, only the Netherlands has the 30% tax rule for immigrants. While other arrangements have a, slightly, different meaning in the three countries. For example, what in the Netherlands is known as social assistance, is something different in Germany. The Dutch social assistance is called Hartz 4 in Germany. So, there are differences in the number of welfare state arrangements between the survey countries. And in some cases, the names of the arrangements look different but content-wise relate to the same form of arrangement. 3.3 Questionnaire for natives Besides respondents from ten different countries, also natives from Denmark, the Netherlands, and Germany were included in the survey. Questions regarding migration history, perceived discrimination and identification were straightforwardly not asked at natives. In general, the wording of the questions for natives was adjusted such that they could not derive from the questionnaire that it was a study on migrants. 16

3.4 Pre-test To test whether the questions were understood by a larger group, to check the mailing coordination from one country, the Netherlands, and to make a very limited prognosis about the influence of an unconditional incentive a pre-test was done in the German city Freiburg among native and American immigrants in August and September 2015. The aim of the national pre-test in Germany was not only to obtain information about potential pitfalls of the questionnaire, but also to get more information about how the sampling procedure works for Germany. Since in Germany, researchers had to contact every municipality we got useful information about the procedure of how to get in touch with the municipalities and how to contact them most efficiently. We decided to field two versions of the questionnaire, one for the native Germans and one for American immigrants. Furthermore, we obtained information about the different measures we wanted to apply to the full sample, like variance and item non-response. Additionally, information about the response rate, not deliverable mails and other logistic problems was obtained. In what follows, we discuss the sampling, fieldwork, and the response (section 3.4.1), as well as the actual questionnaire fielded (section 3.4.2). 3.4.1 Pre-test: Sampling, Fieldwork, and Responses We decided to field the questionnaire in a large city that would not be in our actual sample, since the immigrant group would be likely to receive the pre-test version as well as the actual version of the questionnaire. The city of Freiburg (around 225,000 inhabitants) was one city that matched the requirements. Additionally, the close distance to Konstanz would allow us to also travel to the city if something would go completely wrong, which was not necessary in the end. After the contact with the municipality we received a dataset containing the address data to launch the pretest. We obtained the random address data of 100 native Germans and 200 17

American immigrants. As in the main study, the questionnaires were printed by the Dutch company I&O Research. They were also responsible for the fielding by sending the mail to the potential respondents. The procedure was carried out according to Dillman (2000), by sending a postcard as a first reminder one week after the respondents had received the invitation. A second reminder containing the questionnaire and a letter was sent two weeks after the first reminder. Every participant was offered a 10 voucher of their choice (Amazon, Media Market, Müller). We started to send the invitation letter containing the questionnaire and the invitation code for the online version on August, 10 th 2015, followed by the first reminder on August 31 st 2015. The second reminder was then sent on September 14 th 2015. We were not able to track the time of return, since the post office in Konstanz just delivers the mail once a week. Also for those who participated using the online version, information about response date and response time was not obtained. 18

Table 3.1: Response rate pre-test Group Invited Letters not Respondents Naïve Response Response Rate delivered Rate Native German 100 2 (2%) 48 48% 49.0% American 200 85 (42.5%) 54 27% 47.8% Total 300 87 (29%) 102 34% 47.9% Looking at the returns and responses we must state that for the Americans 29% of the letters could not be delivered. For the Germans, we did not have those problems (2%). We assume that besides immigrants being a more mobile group (return migration) we additionally could have the problem that some of the potential respondents are exchange students, who have returned to America. This is likely since Freiburg is known as a university town. Hence, a large share of an uncommon immigrant group for Germany as the Americans might be consisting of students. Since the new semester starts in October it is likely that American students left the country in the time between we received the address data and the time the questionnaire was fielded. Nevertheless, the response rate indicated a high response rate. The actual response rate for native Germans and for Americans is close to 50%. The high response rate might be a result of the topic we are asking; it is higher than most other surveys conducted in Germany. 3.4.2 Questionnaire The aim of the pre-test was also to get more insights on problems that arose within the questionnaire. In conclusion, we adjusted the order, added, rephrased, and omitted questions according to the results. The questionnaire contained the following modules: A. Demography and migration biography 19

B. Government responsibilities C. Household and Health D. Language and Contacts E. Religion F. Media use and Political attitudes G. Knowledge and opinions about welfare state use H. Experiences with residence country I. Education and employment J. Household assets The questionnaire can be found in Appendix A. 20

4. FIELDWORK Despite frequent team meetings and central coordination, the data collection could not be done in exactly the same way in all countries. Partly due to different sampling strategies (see section 2.2), and due to different strategies to increase the response rate. In this chapter, the crosscountry time frame of the data collection is summarized (section 4.1) followed by details on the coordination of fieldwork (section 4.2). Country specific approaches to increase the response rate are discussed in section 4.3. The final section briefly outlined the strategies to check if the questionnaire was filled out by the invited respondent (section 4.4). 4.1 Time frame of the fieldwork Although the initial planning was to have the same fieldwork period in Denmark, Germany and the Netherlands, due to an extended sampling period in Germany, the German fieldwork was later than the Dutch and Danish period. Table 4.1 provides an overview of the fieldwork period in the Netherlands, Denmark, and Germany. Tabel 4.1: overview of fieldwork period Delivery invintation Delivery first reminder Delivery second reminder End data collection Netherlands December 1 2015 December 6 2015 December 19 2015 January 25 2016 Denmark December 1 2015 December 6 2015 December 19 2015 January 25 2016 Germany December 8 2015 December 13 2015 February 5 2016 April 2 2016 With the first mailing, all respondents received an invitation letter to participate in our research. This letter also included the website reference and unique login code to participate online. In addition to this login code, respondents received two questionnaires; one questionnaire in Danish, Dutch or German and one questionnaire in respondents mother tongue. The first reminder was delivered five days after the invitation. The first reminder was a (fancy) postcard that contains a reminder to participate, together with the website address and 21

login code to participate online (straightforwardly, natives only received the questionnaire in Danish, Dutch, or German). Since there was limited time between the invitation and the first reminder there was a likelihood of cross posting (respondents who already participate still received a reminder), which was noted in the text. The second reminder contained a reminder letter and again the two questionnaires. Since the likelihood that migrants visited their families in that period the decision was made to postpone the second reminder after the Christmas holidays. Due to some difficulties with the German mail the second reminder was delivered in Germany on February 5 th. In the Netherlands and Denmark, the data collection was closed four weeks after the second reminder. In Germany, this period was extended to seven weeks due to a lower response and the weekly, instead of daily in the Netherlands and Denmark, mail delivery at the university. 4.2 Coordination of the fieldwork Because of the budget, the data fieldwork was done as much as possible by the universities itself. For pragmatic reasons the coordination was done by the Radboud Univeristy. Although coordination with the fieldwork agency was done by the Radboud University, the Danish and German university provided the input to the fieldwork company directly about the respondents who needed a reminder and who not. The online questionnaire ran on a server at the Aalborg University. The three countries were each responsible for creating a national file of the online response. 22

4.3 Strategies to increase response rate To increase the response rate both in the Netherlands and Denmark a conditional incentive was used. In Germany, an experiment with none, a conditional, an unconditional and the combination of conditional and unconditional incentives was implemented. 4.3.1 The Netherlands To increase the response rate in the Netherlands, a conditional incentive was used. Respondents who participated could choose to receive a 10-euro gift voucher from three (online) shops: Blokker, Bol.com, and Hema. 4.3.2 Denmark To increase the response rates in Denmark a conditional incentive of a movie ticket valued at 75 DKR (10 euro) was used. The incentive was sent out digitally after the collection of the survey was completed. 4.3.3 Germany For the incentive experiment, respondents were grouped into either receiving an unconditional incentive only, an unconditional and conditional, a conditional incentive only or no incentive at all. The unconditional incentive was a small handy-cleaner with the Konstanz University Logo imprinted. For the conditional incentive, respondents who participated could choose between a 10 voucher from Amazon, Media Markt, or Müller. 23

Table 4.2: Sample distribution by incentive group Origin country No incentives Conditional incentives only Unconditional incentives only Conditional and unconditional Total Native Germans 180 240 240 240 900 China 204 272 272 272 1020 Japan 321 427 427 427 1602 Poland 312 416 416 416 1560 Romania 227 302 302 302 1133 Russia 324 432 432 432 1620 Spain 256 341 341 341 1279 Turkey 410 547 547 547 2051 UK 223 297 297 297 1114 US 263 351 351 351 1316 Total 2720 3625 3625 3625 13595 After the first and second reminder, the response rate of people who received either none or only the conditional incentive was rather low (around 11.3%) compared to the groups which were promised a voucher (around 15%) which led to the decision to offer these groups a voucher as well in order to boost response rates after the second reminder. 24

5. RESPONSE RATES AND SELECTIVITY The response rate is defined as the proportion between the number of respondents who participated by filling out the questionnaire online or by paper and pencil in relation to the number of people approached (sample). In this section, a brief overview of the sample sizes and response rates across all countries is given (section 5.1). Based on this, response rates for each country are described in more detail (see section 5.2 for the Netherlands, section 5.3 for Denmark, and section 5.4 for Germany). 5.1 Overview of response rates across all countries Table 5.1 shows the overall response rate after cleaning for the Netherlands, Denmark, and Germany. For the Netherlands and Denmark, the response after cleaning is almost equal in both countries, where Denmark has a 1.65% higher response than the Netherlands. Germany has with a response of 18% the lowest response rate In section 5.2., 5.3, and 5.4 the response rates will be discussed in more detail. Table 5.1: Sample size, response and response rates Netherlands Denmark Germany Sample size 11100 10500 13561 Response 3672 3647 2234 Response rate 33,1% 34,7% 18,2% 5.2 The Netherlands The overall response rate of 33.1% in the Netherlands, which is relative high in comparison with other surveys among immigrants in the Netherlands (Andriessen & Kappelhof, 2016; Korte & Dagevos, 2011). Table 5.2 shows that there are large differences in response between migrant groups though. The response among native Dutch is the highest with 47.4%. The response from immigrants from Russia is with 43.18% almost as high as the response among 25

natives. The response from the Turks is the lowest, 19.3%, which resembles other surveys among Turkish immigrants in the Netherlands (Andriessen & Kappelhof, 2016; Korte & Dagevos, 2011). The expectation was that Americans, Brits, Japanese and Spaniards would have a response rate around 33.3%. While Chinese, Philippines, Polish, Romanian, Russian and Turks were expected to have a response around 27%. Americans (27.89%) and Brits (29.89%) have a somewhat lower response rate than expected. While the other groups, with exception of the Turks, have a (much) higher response rate. Table 5.2 also shows the difference between the response before cleaning and the response after cleaning. The differences are almost equal between all groups, and are caused because some people filled out the questionnaire twice (online and hard copy), or that a partner of the respondent also filled out the questionnaire. In section 6.1 the data cleaning will be discussed in more detail. Finally, table 5.2 shows that in general more than seventy percent of the questionnaires are filled out by paper and pencil and less than thirty percent online. Table 5,2: Dutch response rate Response Response rate % paper Invited Before After cleaning after cleaning pencil % online Native Dutch 900 442 427 47,44 69,09 30,91 China 1100 330 307 27,91 68,73 31,27 Japan 900 315 295 32,78 77,97 22,03 Philippines 1100 408 385 35,00 79,74 20,26 Poland 1100 377 353 32,09 79,04 20,96 Romania 1100 381 357 32,45 69,75 30,25 Russia 1100 518 475 43,18 74,32 25,68 Spain 900 367 341 37,89 63,64 36,36 Turkey 1100 237 212 19,27 80,19 19,81 UK 900 280 269 29,89 73,23 26,77 US 900 267 251 27,89 61,35 38,65 Total 11100 3922 3672 33,08 72,49 27,51 26

Table 5.3 shows the selectivity of the response in relation to sex and age. It appears that significant more Chinese, Japanese and Polish women filled out the questionnaire than expected on the male female ratio in the sample. Regarding age, native Dutch, Chinese, Japanese, Brits and Americans who participate are significant older than native Dutch, Chinese, Japanese, Brits and Americans in the sample. Table 5.3: Dutch response rate, split by sex and age Sample Response Male:female Mean Male:female Mean ratio age Male Female ratio 1 age 1 Native Dutch 0,95 46,89 203 224 0,91 48,88 China 0,77 34,10 97 210 0,46 33,25 Japan 0,73 38,97 85 210 0,40 39,93 Philippines 0,18 38,69 47 338 0,14 39,09 Poland 0,84 35,12 128 225 0,57 35,82 Romania 0,65 33,99 115 242 0,47 34,73 Russia 0,34 38,89 104 371 0,28 38,79 Spain 0,79 33,38 142 199 0,71 33,51 Turkey 0,88 38,88 92 120 0,77 39,51 UK 1,62 40,77 147 122 1,20 43,47 US 0,83 37,37 110 141 0,78 39,93 1 figure in bold = significant difference between sample and data p <.05 Table 5.4 combines above descriptive results by showing the results from logistic regression analyses predicting various reasons for participating. The results show to what extent people with certain characteristics in the sample are more or less likely to participate than others. Women are more likely to participate than men. The same applies for older people. Native Dutch are more likely to participate than other ethnicities. 27

Table 5.4: Odds ratios from logistic regression predicting reasons of participation (reference category is particiapte) Overall Native Dutch China Japan Philippines Poland Romania Russia Spain Turkey UK US Sex (male = ref.) 1.468 *** 1.069 2.021 *** 2.405 *** 1.465 * 1.763 *** 1.563 ** 1.377 ~ 1.171 1.186 1.635 *** 1.172 Age 1.011 *** 1.015 *** 0.987 * 1.020 ** 1.007 1.013 1.010 0.997 1.003 1.009 1.026 *** 1.023 *** Ethnicity (Dutch = ref.) China 0.477 *** Japan 0.570 *** Philippines 0.570 *** 28 Poland 0.582 *** Romania 0.585 *** Russia 0.835 * Spain 0.762 ** Turkey 0.282 *** UK 0.526 *** US 0.463 *** Constant 0.452 *** 0.428 *** 0.395 *** 0.131 *** 0.300 *** 0.220 *** 0.258 *** 0.668 0.508 *** 0.150 *** 0.118 *** 0.150 *** ~ p <,05; * p <,01; ** p <,005; p <,001

5.3 Denmark As shown in Table 5.1 the overall response for Denmark is 34,7 per cent, which is very similar to the results from the Netherlands, and must considered good considering that migrants surveys often have lower response rates than surveys of the general population (Font & Méndez, 2013). Table 5.5, below, shows the differences in response rate between the Danes and migrant groups. Table 5.5: Response rates by country of origin and method of answering Response before cleaning Response rate after cleaning % paper Invited After cleaning pencil % online Denmark 900 416 397 44,1 68,3 31,7 Poland 1000 312 293 29,3 74,0 26,0 Russia 1000 432 408 40,8 70,6 29,4 Romania 1000 284 277 27,7 59,9 40,1 Turkey 1000 228 216 21,6 81,1 18,9 Great Brittan 900 421 402 44,7 70,5 29,5 Philippines 900 296 280 31,1 72,6 27,4 Japan 900 393 379 42,1 72,0 28,0 China 1000 367 346 34,6 69,5 30,5 Spain 900 352 339 37,7 55,4 44,6 USA 900 319 310 34,4 60,2 39,8 Total/average 10.400 3820 3647 34,7 68,6 31,4 The goal was to get at least 300 respondents in each group, which we have not quite succeeded in, since there are fewer from Turkey, Poland, Romania and the Philippines. These are all groups with oversampling, but in retrospect this should have been higher in these groups. For Russians and Chinese the oversampling was, however, not needed, as there were over the 300 with good response rate. Generally, looking at response rates they were relatively high, with over 40 per cent, for Russians, Danes, Brits, and Japanese. On the other hand, they are relatively low for Turks and Romanians, which was as expected. 29

Table 5.5 also shows the difference between the response before cleaning and the response after cleaning. The differences are almost equal between all groups (between 9 and 24), and are caused because some people filled out the questionnaire twice (online and hard copy). When duals existed they were compared, and if both where filled out, paper where preferred over online. Finally, the table shows that paper questionnaires was generally preferred by about two third of respondents, while about one third used the online option. Though there is a little variation between the group this pattern is quite persistent 5.3.1 In depth non-response analyses Denmark Using register data, in Denmark it was possible to compare respondents to the full population regarding to gender, age, household type, citizenship, labour force position, social economic position, income, received benefits and migration history. The differences between respondents and non-respondents are remarkably small. There are some differences between respondents and non-respondents on labour market position and other socioeconomic factors none of them are remarkably different, and often they vary between group, e.g. unemployment leads to higher response rates for some groups and lower for other. Tables 5.6 to 5.18 compare the respondents with the non-respondents in Denmark. 30

Table 5.6: Gender. Presented as row percentages and total number of respondents. Man Woman N Man Woman N Respondents 43,2 56,8 3577 Non-respondents 42,7 57,3 7179 Denmark 48,8 51,2 396 Denmark 49,1 50,9 602 Poland 46,5 53,5 288 Poland 49,7 50,3 696 Romania 55,8 44,2 271 Romania 53,1 46,9 700 Spain 53,7 46,3 328 Spain 53,8 46,2 637 GB 73,5 26,5 397 GB 70,0 30,0 589 Turkey 45,3 54,7 216 Turkey 46,5 53,5 780 USA 57,2 42,8 297 USA 54,2 45,8 669 Japan 26,3 73,7 373 Japan 23,7 76,3 574 China 39,0 61,0 336 China 37,8 62,2 644 Philippines 8,1 91,9 271 Philippines 10,3 89,7 700 Russia 20,3 79,7 404 Russia 19,9 80,1 588 Note: Based on KOEN Table 5.7: Age. Presented as mean scores by group. Respondents 1972,0 Non-respondents 1972,3 Denmark 1966,2 Denmark 1963,3 Poland 1972,6 Poland 1974,3 Romania 1981,8 Romania 1981,8 Spain 1977,9 Spain 1977,8 GB 1965,0 GB 1965,5 Turkey 1966,3 Turkey 1967,8 USA 1972,2 USA 1970,3 Japan 1969,0 Japan 1969,8 China 1975,8 China 1976,3 Philippines 1975,3 Philippines 1974,9 Russia 1973,0 Russia 1971,8 Note: Based on aldernov. Age on the last work day of November. 31

Table 5.8: Household type. Presented as row percentages and total number of respondents. Single man Single woman Married couple Other couple Other households Respondents 6,7 11,0 40,9 11,5 29,9 3577 Denmark 11,4 17,7 44,7 13,4 12,9 396 Poland 4,9 11,8 35,1 12,2 36,1 288 Romania 4,4 7,0 25,1 14,4 49,1 271 Spain 7,0 6,4 24,1 18,3 44,2 328 GB 11,1 5,8 42,6 20,7 19,9 397 Turkey 6,0 13,0 50,9 3,2 26,9 216 USA 10,1 7,4 40,4 12,5 29,6 297 Japan 6,2 13,1 47,2 5,1 28,4 373 China 3,3 7,1 40,2 6,9 42,6 336 Philippines 1,5 11,1 44,3 5,9 37,3 271 Russia 5,0 17,8 51,5 10,2 15,6 404 Single man Single woman Married couple Other couple Other households Non-respondents 6,5 9,7 40,5 10,2 33,3 7179 Denmark 12,0 17,9 42,2 15,5 12,5 602 Poland 7,0 9,5 29,7 14,8 38,9 696 Romania 4,4 4,9 25,3 12,7 52,7 700 Spain 7,9 4,7 24,3 16,8 46,3 637 GB 14,3 6,6 40,6 17,0 21,6 589 Turkey 5,0 10,6 50,5 3,9 30,0 780 USA 9,1 6,9 47,1 8,1 28,9 669 Japan 4,0 9,9 49,0 8,9 28,2 574 China 5,0 9,3 37,9 5,6 42,2 644 Philippines 0,6 8,9 46,1 3,9 40,6 700 Russia 3,2 18,4 53,7 6,6 18,0 588 Note: Based on hustype. A household is defined by the address. A household cover all persons in the CPR/citizen register. Other households cover any household consisting of more than one family. 32

Table 5.9: Danish citizens, among migrant groups. Presented as row percentages and total number of respondents. No Yes N No Yes N Respondents 88,4 11,6 3246 Non-respondents 88,7 11,3 6709 Denmark NA NA NA Denmark NA NA NA Poland 88,1 12,0 293 Poland 91,1 8,9 706 Romania 92,8 7,2 277 Romania 94,0 6,0 721 Spain 95,0 5,0 339 Spain 96,2 3,8 661 GB 92,5 7,5 402 GB 94,0 6,0 598 Turkey 76,9 23,2 216 Turkey 79,3 20,7 783 USA 93,2 6,8 310 USA 94,9 5,1 689 Japan 90,0 10,0 379 Japan 92,2 7,9 586 China 88,7 11,3 345 China 85,5 14,5 655 Philippines 80,0 20,0 280 Philippines 79,9 20,1 720 Russia 83,0 17,0 405 Russia 82,0 18,0 590 Note: Recoded from STATSB. Citizenship for Danes is not included. 33

Table 5.10: Position in the labour force. Presented as row percentages and total number of respondents. Not registered While collar Skilled worker Unskilled worker Other Worker N Respondents 59,4 14,2 11,0 7,6 7,7 364 2 Denmark 52,3 20,2 15,4 3,8 8,3 396 Poland 43,0 7,5 22,5 18,1 8,9 293 Romania 57,4 5,8 7,2 15,2 14,4 277 Spain 65,5 19,8 6,5 3,5 4,7 339 GB 52,7 19,7 14,9 3,5 9,2 402 Turkey 65,7 6,9 8,3 13,0 6,0 216 USA 62,3 18,7 9,0 4,5 5,5 310 Japan 74,7 11,4 5,5 2,9 5,5 379 China 60,3 17,4 5,2 6,7 10,4 345 Philippines 60,7 1,8 14,6 17,1 5,7 280 Russia 59,8 18,0 11,6 4,2 6,4 405 Not registered While collar Skilled worker Unskilled worker Other Worker N Non- Respondents 60,6 13,7 10,8 7,1 7,9 731 3 Denmark 56,1 21,0 14,2 3,8 4,8 604 Poland 51,8 8,1 16,9 12,5 10,8 706 Romania 54,9 4,3 9,3 13,7 17,8 721 Spain 64,6 17,6 8,9 3,5 5,5 661 GB 51,8 23,9 12,5 4,2 7,5 598 Turkey 65,0 7,2 8,7 12,0 7,2 783 USA 62,7 23,2 7,1 1,2 5,8 689 Japan 75,9 11,8 7,7 1,0 3,6 586 China 57,7 16,0 7,9 5,5 12,8 655 Philippines 66,5 2,5 14,2 12,5 4,3 720 Russia 58,8 20,0 11,4 4,4 5,4 590 Note: Recoded from STILL 34

Table 5.11: Register based unemployment in percentage of time over the last year. Respondents 3,02 Non-respondents 3,19 Denmark 2,12 Denmark 1,75 Poland 5,00 Poland 4,82 Romania 1,97 Romania 2,68 Spain 2,32 Spain 2,45 GB 2,02 GB 2,81 Turkey 5,74 Turkey 6,64 USA 2,89 USA 1,49 Japan 1,85 Japan 2,47 China 3,72 China 2,49 Philippines 2,05 Philippines 2,14 Russia 4,26 Russia 4,15 Note: Calculated from on ARLEDGR Table 5.12: Total DKR received in unemployment and social assistance in 2015 Respondents 19240 Non-respondents 18300 Denmark 15146 Denmark 9534 Poland 24440 Poland 20823 Romania 26328 Romania 17667 Spain 13114 Spain 10954 GB 13021 GB 16609 Turkey 50563 Turkey 39852 USA 8160 USA 9586 Japan 6005 Japan 4551 China 15644 China 17066 Philippines 14477 Philippines 11223 Russia 36431 Russia 40156 Note: Based on DAGPENGE_KONTANT_13. 1 Euro 7.45 DKR. This includes social assistance, unemployment benefits, integration benefit (integrationsydelse), parental leave and sick leave. 35

Table 5.13: Income in DKR in 2015 Respondents 203017 Non-respondents 192337 Denmark 212680 Denmark 206181 Poland 199960 Poland 196238 Romania 169058 Romania 177056 Spain 238653 Spain 197812 GB 303783 GB 323121 Turkey 100333 Turkey 110451 USA 269296 USA 291010 Japan 170255 Japan 156551 China 176637 China 177709 Philippines 119500 Philippines 120490 Russia 204142 Russia 186317 Note: based on LOENMV_13. 1 Euro 7.45 DKR Table 5.14: Total received DKR from any type of public pension in 2015 Respondents 12221 Non-respondents 12980 Denmark 35250 Denmark 45980 Poland 8070 Poland 9124 Romania 1713 Romania 2864 Spain 4539 Spain 4966 GB 13739 GB 13257 Turkey 32131 Turkey 29940 USA 8632 USA 6801 Japan 18786 Japan 15342 China 1982 China 3454 Philippines 6394 Philippines 7377 Russia 3536 Russia 4600 Note: Based on OFFPENS_EFTERLON_13. 1 Euro 7.45 DKR. Total public pension s covers the public pension, early retirement, and any supplements to these. 36

Table 5.15: Total DKR received in social benefits from in 2015 Respondents 42180 Non-respondents 41921 Denmark 62342 Denmark 67107 Poland 43967 Poland 41179 Romania 36853 Romania 30646 Spain 25816 Spain 24147 GB 32384 GB 35292 Turkey 98406 Turkey 84022 USA 23962 USA 23118 Japan 33775 Japan 29808 China 25400 China 29036 Philippines 31346 Philippines 30501 Russia 61787 Russia 63690 Note: Based on OFF_OVERFORSEL_13. 1 Euro 7.45 DKR. Table 5.16: Year immigrated to Denmark Respondents 2003,0 Non-respondents 2003,0 Poland 2004,2 Poland 2005,9 Romania 2009,4 Romania 2009,5 Spain 2006,6 Spain 2006,7 GB 1997,2 GB 1997,8 Turkey 1992,3 Turkey 1993,9 USA 2002,8 USA 2001,7 Japan 2000,9 Japan 2001,2 China 2005,5 China 2005,5 Philippines 2004,3 Philippines 2004,1 Russia 2004,9 Russia 2004,1 37

Table 5.17: Socioeconomic status in 2015. Presented as row percentages and total number of respondents. Self-employed employed Unemployed Student Pensioner Other N Respondents 3,71 50,11 9,8 6,4 10,52 19,47 3642 Denmark 3,84 56,16 8,22 28,77 3,01 365 Poland 1,37 68,73 10,65 6,87 12,37 100 Romania 3,2 70 11,6 1,6 13,6 250 Spain 3,48 55,75 8,71 4,53 27,53 287 GB 4,86 59,85 6,14 16,37 12,79 391 Turkey 7,44 31,16 29,3 24,19 7,91 215 USA 5,5 51,55 3,78 8,59 30,58 291 Japan 4,12 36,54 3,3 17,03 39,01 364 China 5,5 57,61 9,71 1,94 25,24 309 Philippines 2,92 43,8 6,93 5,84 40,51 274 Russia 2,15 54,57 22,31 4,3 16,67 372 Self-employed employed Unemployed Student Pensioner Other N Non-respondents 3,8 50,09 9,34 6,33 11,25 19,19 7313 Denmark 3,44 52,8 4,52 38,16 1,08 553 Poland 3,4 69,28 9,6 7,83 9,9 677 Romania 0,62 74,07 7,92 2,8 14,6 644 Spain 2,26 52,96 5,75 5,23 33,8 574 GB 7,61 57,96 8,65 14,01 11,76 578 Turkey 6,85 38,5 23,13 23,39 8,14 774 USA 6,13 53,29 5,21 9,19 26,19 653 Japan 5,45 36,73 2 15,82 40 550 China 4,93 55,92 10,2 4,28 24,67 608 Philippines 1,16 44,27 5,81 6,97 41,8 689 Russia 2,55 53,27 24,18 5,09 14,91 550 Note: Recoded from PRE_SOCIO. Students excluded from the by country tabulation to comply with rules on privacy. The pensioner s category covers public pensions, early retirement and disability pension. 38

Table 5.18: Percentage who is married. Presented as row percentages and total number of respondents. Respondents 47,6 3642 Non-respondents 45,9 7313 Denmark 54,0 396 Denmark 54,5 604 Poland 44,7 293 Poland 52,0 706 Romania 56,7 277 Romania 56,9 721 Spain 69,0 339 Spain 68,8 661 GB 46,3 402 GB 48,7 598 Turkey 27,8 216 Turkey 24,8 783 USA 51,3 310 USA 42,8 689 Japan 43,5 379 Japan 40,6 586 China 41,5 345 China 37,9 655 Philippines 48,2 280 Philippines 44,9 720 Russia 36,8 405 Russia 34,4 590 Note: Based on recoding civst 39

5.4 Germany The overall response rate in Germany was 18.15% which was lower than expected. The response rate was highest among native Germans, migrants from China, Spain and the UK with overall 28%. Response rates were particularly low for migrants from Poland (16.25%) and Turkey (10.85%). Overall, respondents preferred pen and pencil over the online tool. Table 5.19: German response rate Response Before Response After Valid Observations Response Rate Pencile and Paper Group Invited Online Native Germans 900 250 233 866 26.91% 80.69% 19.31% China 1,020 263 244 889 27.45% 70.49% 29.51% Japan 1,602 364 316 1,465 21.57% 78.16% 21.84% Poland 1,560 253 190 1,455 13.06% 87.89% 12.11% Romania 1,133 200 150 1,021 14.69% 78.00% 22.00% Russia 1,620 311 217 1,493 14.53% 77.88% 22.12% Spain 1,279 328 298 1,117 26.68% 66.44% 33.56% Turkey 2,051 224 138 1,907 7.24% 86.23% 13.77% UK 1,114 297 237 963 24.61% 65.82% 34.18% US 1,316 294 211 1,131 18.66% 58.77% 41.23% Total 13,595 2,784 2,234 12,307 18.15% 74.17% 25.83% Looking at the response rate by incentive group we see that we achieved the highest response rate when people were promised a voucher in case they would participate in the study. The unconditional incentive, however, had almost no effect on the response rate. 40