All time references are in CEST
Sampling migrants in general population surveys 1 |
|
Session Organisers |
Professor Francesco Molteni (University of Milan, Department of Social and Political Sciences) Professor Riccardo Ladini (University of Milan, Department of Social and Political Sciences) Professor Ferruccio Biolcati (University of Milan, Department of Social and Political Sciences) |
Time | Thursday 20 July, 09:00 - 10:30 |
Room | U6-06 |
Sampling migrant groups is one of the biggest challenges in survey research. Such a challenge has been tackled by relying on both probabilistic and non-probabilistic research designs focusing on specific sub-populations (often with a local focus). With very few and remarkable exceptions (i.e., the BHPS and the SOEP), what is lacking when studying migrant populations with survey methods is the comparison with the native population, which is a key issue when looking at the differences as well as the similarities between the groups. Because of this, one of the main approaches to perform such a comparison has been to refer to the subsamples of people of foreign origin from general population surveys samples. Although the strategy has various advantages, it presents several drawbacks mainly because the low numerosity of such populations, the potential selection bias leading to the under-representation of specific migrant groups, and the variety of sampling frames, often depending on the differentiated availability and composition of population lists across contexts. The latter becomes particularly relevant in comparative research. Moving from these considerations, the session welcomes papers focusing on strategies aimed at including subsamples of migrants in general population samples. Both contributions dealing with specific techniques (i.e., oversampling techniques, onomastic procedures, focused enumeration) or practical experiences will be appreciated. In addition, also contributions reflecting on the pros and cons of using immigrant sub-samples from already existent survey programs (i.e., the European Social Survey, the European Values Study or the European Union Labour Force Survey) are welcomed.
Keywords: survey, migrants
Ms Tanja Stojadinovic (Ipsos) - Presenting Author
Ms Sara Grant-Vest (Ipsos)
Ms Vida Beresneviciute (European Union Agency for Fundamental Rights)
Ms Rossalina Latcheva (European Union Agency for Fundamental Rights)
The EU Survey on Immigrants and Descendants of Immigrants and its predecessors EU-MIDIS I and II provide comprehensive (EU-wide comparable) data on experiences of discrimination and bias-motivated harassment and violence, used by policy-makers when addressing these issues and to assess progress over time as regards equality and non-discrimination in key areas of life. This round of the survey focused on the experiences of immigrants and descendants of immigrants from North Africa, Sub-Saharan Africa, Syria and Turkey.
In this presentation, we provide details on the random-probability sampling methods used that relied on the same sampling sources that are used for general public surveys. In France, Italy, Portugal and Spain a multi-stage clustered sample design, with screening for eligibility, was implemented using face-to-face data collection. The design oversampled areas of higher concentrations of the target populations, and in the latter three countries it used focused enumeration in areas of low concentrations. In Austria, Denmark, Finland, and Luxembourg population registers were used for direct sampling of eligible members of the target population for an online push-to-web survey. The push-to-web approach was also applied in Germany, in combination with an onomastic procedure that was used to identify the target group members, as the population register did not provide information on eligibility.
We discuss how the sampling sources containing the target population counts were accessed, approximations that were used to estimate concentration levels when available data did not align with the definition of the target groups, and how these compare to the eligibility rates recorded during data collection. We talk about the sample designs, the challenges of implementing them, and the final sample efficiency. We also present positive experiences of implementing online push-to-web designs with the target populations.
Mr Moritz Fahrenholz (infas Institute for Applied Social Sciences) - Presenting Author
Dr Merih Ates (German Center for Integration and Migration Research)
Dr Jörg Dollmann (German Center for Integration and Migration Research)
Mr Tae Jun Kim (German Center for Integration and Migration Research)
Dr Sebastian Link (infas Institute for Applied Social Sciences)
Ms Aneta Malina (infas Institute for Applied Social Sciences)
Mr Michael Ruland (infas Institute for Applied Social Sciences)
Sampling immigrants or persons with an immigration background is challenging, especially when it comes to small or very small populations.
In this paper, we describe and discuss the sampling approach of the NaDiRa.panel. The NaDiRa.panel is an online access panel with a probability-based offline recruitment. It comprises a sample of the whole German resident population (including immigrant groups) and four samples of small or very small groups of persons with an immigration background: people from Turkey, other majority Muslim countries, East- and Southeast Asia and African coun-tries.
The Sampling approach comprised three steps and a “double oversampling”: We made use of spatial context information to identify municipalities with a high share of inhabitants of the two smallest migrant groups. Based on this information we created two sampling frames of municipalities: (1) low to medium expected migrant share and (2) high expected migration share. We drew a sample of 250 primary sampling units from both frames. So half of the sampled population comes from municipalities with a high expected migration share. Since registration offices of the municipalities do not provide sufficient information on the immigration background, we applied name-based pre-classifications (onomastic approach) of about 1 million names in order to build and oversample the immigrant groups within the sampled municipalities.
In order to assess the success of the sampling approach in terms of size of the net sample and selectivity, we compare the share of our target groups in our oversampled gross sample, the share in our different primary sampling frames, the share in our net sample and official government data (German Microcensus). To assess the validity of our results we also compare the onomastic result and the migration background according to the survey data for our net sample.
Mr Carsten Broich (Sample Solutions BV) - Presenting Author
Mr Jacob Lagerstedt (Norstat)
Random Digit Dialing (RDD) sample consists of randomly generated phone numbers which can be sorted by different area, country or network provider. Thus, RDD sample is usually used for national representative studies, typically conducted by CATI (Computer-Assisted Telephone Interviewing). However, very often there is a need of targeting specific audiences during the survey sampling. In such cases, the pure RDD sample provides a low incidence rate and hence it is inappropriate due to high cost. Therefore, a need for other sampling methodologies or modification of the pure RDD methodology derives. Depending on the survey aim, the target audience could differ in terms of age, gender, ethnicity etc. Nowadays, living in a multicultural society, studies upon different ethnic groups are of immense importance. For the case of Sweden and Stockholm in particular, different migrant groups exist that came to Sweden for various reasons: economic reasons from outside the EU, economic reasons from within the EU but also conflicts make up the reason for a large number of migrants.
This paper outlines the sampling approaches of targeting different migrant populations. Furthermore, this paper outlines the advantages and disadvantages of each of the described sampling methodologies. Moreover, within this paper, the results of a pilot project for Stockholm is outlined. Consequently, the theoretical advantages and disadvantages are accompanied by those
Professor Ferruccio Biolcati (University of Milan) - Presenting Author
Dr Riccardo Ladini (University of Milan)
Dr Francesco Molteni (University of Milan)
The use of population lists is the golden standard for obtaining representative probabilistic samples of immigrant subpopulations. However, those lists are not always available to survey researchers. This is the case of Italy, where the use of population lists is not allowed for institutions (e.g. universities) and survey programs (e.g. EVS, ISSP) not adhering to the official statistical system. In these situations, the suboptimal strategy consists in the use of electoral lists, which do not include non-Italian citizens.
To overcome this issue and minimize the resulting bias, we propose a strategy for the inclusion of non-Italian citizens, first employed in the CoValues survey carried out in Lombardy in late 2022. Aimed at analysing the native-migrant divide in values, the survey employs the same probabilistic sampling design of the Italian edition of the European Values Study to collect interviews among the Italian citizens. To sample non-Italian citizens, it employed a focused enumeration strategy, namely, a spatial sampling design driven by respondents themselves. Each respondent was asked to name the three nearest addresses where he/she thoughts people with a foreign origin were living. This strategy allows a first screening of the immigrant population, by providing to interviewers addresses where to plausibly find non-Italian citizens. A rather rigid protocol was then employed for sampling those individuals. This contribution aims at analysing the pros and cons of this strategy, by evaluating its effectiveness and comparing the distribution of the immigrant sample with the distribution of the immigrant populations according to the more relevant characteristics. All in all, we will discuss to what extent this strategy could overcome the issue of sampling lists in general population surveys.
Ms Katrin Pfündel (Federal Office for Migration and Refugees Germany) - Presenting Author
Dr Anja Stichs (Federal Office for Migration and Refugees)
Surveys addressing immigrants usually lack an appropriate sampling frame, because public statistics are missing or unavailable. Sampling procedures for rare populations, such as the increasingly popular onomastic sampling technique can help. General population surveys can profit from the experiences within immigrant-specific surveys applying this technique such as the study Muslim Life in Germany, conducted by the Research Centre of the Federal Office for Migration and Refugees. It includes subsamples of persons with a migration background from Muslim-majority regions of origin. We use an innovative two-step procedure to establish a sampling frame for this population. First, we generate a representative national sample by selecting roughly 300 local resident registration offices. Second, we identify names, which match our specific study population using onomastics. Notice that our sample is not restricted to any place of birth or religious affiliation.
Even though several case studies exist that sample specific immigrant populations by the onomastic procedure, little is known about the reliability of this approach regarding the differentiation of immigrants with similar names. This paper contributes to the literature by evaluating the reliability of onomastics for similar populations, in other words, immigrants from five Muslim-majority regions (Turkey, Southeast Europe, Asia, Middle East, and Northern Africa). Subsequently, we describe the implementation of the method and evaluate the realised sample: we compare characteristics of target persons whose names have been correctly assigned to a specific region of origin group with characteristics of non-target persons who were falsely assigned to a group. In addition, we are able to examine the accuracy of the onomastic procedure for Muslim individuals from the aforementioned origin groups. For our analyses, we use different data sources: nationalities recorded by the resident registration offices, name-based classifications from onomastics, and various information from the survey and screening data set.