Sampling Migrants and Other Mobile Groups as Hard-to-Reach Populations |
|
| Chair | Dr Joanna Napierala (Centre of Migration Research, University of Warsaw ) |
| Coordinator 1 | Dr Agata Górny (Centre of Migration Research, University of Warsaw) |
Respondent driven sampling (RDS) is a relatively new sampling method specifically proposed for sampling rare or hidden populations. RDS starts with the members of the target population and traces their social networks as well as the networks of those who are connected. Through chain referrals stimulated by incentivized recruitment coupons, RDS exploits the social networks for sampling purposes without screening.
With the pressing needs for studying rare/hidden populations despite the increasing expense and difficulty in sampling for rare or hidden populations, RDS has gained popularity rapidly. However, in contrast to the increasingly large volume of research using RDS data, its methodological assessments are very limited. Statistical inference for data collected under RDS designs are under-developed as their formal base sits on strong assumptions that can be easily violated in practice, such as respondents accurately reporting the number of rare/hidden population persons in their network and randomly choosing among them. Scarcity, if not near absence, of publicly available RDS data makes objective methodological assessments even more challenging, leaving the level of population representation through RDS under-scrutinized.
This study aims to provide the beginning of an already-overdue empirical investigation into the realities of RDS data collection through the total survey error (TSE) framework. Targeting foreign-born Koreans in the U.S. a rare group accounting for 0.3% of the general population, we implemented RDS in Los Angeles County and State of Michigan for a Web survey on general socio-demographic, social and health topics. Respondents were followed up after the main survey regarding their recruitment experiences. This study examines: 1) sampling productivity, 2) the nature of recruitment and recruitment chains, and 3) potential biases by comparing to external data using probability samples, such as the American Community Survey and the California Health Interview Survey.
The Respondent Driven Sampling (RDS) is a sampling method based on networks of respondents. In RDS we are following the respondents recruitment process which could be compared to a random walk through the network of migrants. As we collect the information on both the structure of the respondents’ networks and the characteristics of their members we are capable to research the migrant population profoundly.
The above advantage of the RDS method and the fact that it allows for sampling migrants ‘hidden’ to a researcher or ‘hard to reach’ due to high mobility, but accessible through network connections with other migrants, contribute to the growth of importance of RDS in migration studies. Its advantages in sampling ‘hidden’ and ‘hard to reach’ migrants are of particular value in ‘new immigration’ to Central and East European countries, like Poland, where volume of migrants is still small, temporary mobility prevails, significant groups of migrants have irregular status and consequently migration registers are of limited usage in sampling migrants.
The goal of our paper is to compare outcomes of four studies on Ukrainian migrants in Warsaw agglomeration carried out in 2010, 2012, 2015 and 2016 with the use of RDS. These studies, although largely similar, differed in selected methodological solutions such as type of incentives (cash or vouchers), validity period of a voucher and so on. Our aim is to identify how these differences in implementation of the RDS method impacted recruitment chains and final structure of the obtained sample. In particular, we hypothesise that implementation of a requirement of a delay (i.e. of two days) between an interview of a recruiter and an invited person allows for obtaining a more diversified sample. It is due to the fact that such a procedure slows down recruitment of those most quickly recruiting groups (i.e. students or unemployed) and, in this way, enables participation of other groups which are much slower in recruitment (i.e. workers) in the study.
The sampling of migrants in cross-national research is a challenging endeavor as sampling frames might not be available in all countries to the same extent and as the implementation of particular sampling strategies might be very costly. The use of name based (onomastic) telephone sampling is a relatively cost-efficient possibility, but it relies on a number of conditions (e.g., existence of telephone directories in the respective countries and inclusion of the target population). While it has become less and less common to register new numbers in the respective telephone books in most countries, the use of social networking sites (SNSs) has spread around the globe in the last 20 years. Research on transnationalism has shown that migrants make extensive use of such communication technologies as well. Therefore, we investigate whether a specific SNS, Facebook, can be used to sample migrants, with the ultimate goal to supplement other sampling approaches.
The presentation is based on data from a survey of Polish migrants in Austria, Ireland, Switzerland, and the United Kingdom. Respondents in all four countries were sampled with the help of Facebook advertisements which allow the targeting of specific populations within the network. The advertisements consisted of a standardized text and one of five different pictures. No incentives were offered to potential respondents. The study succeeded in sampling a total of 1,103 individuals who completed the relatively extensive questionnaire within a field period of eight weeks.
The approach can be described as highly accurate as 96% of the realized sample belongs to the target population. The results also suggest that the vast majority of the sample would not have been reached if onomastic sampling had been employed. With a total sampling budget of €500 the technique can be considered as highly cost efficient.
On the basis of our results we conclude that the described method could be a valuable part of comprehensive sampling strategies in cross-national migration research projects.
This paper proposes a viable survey strategy for situations where a representative sample is warranted, yet a sufficiently complete sample frame is unavailable. Such situations are common when surveying immigrant populations, especially those comprising sizable shares of recent arrivals and/or people lacking residence permits. Typically, there is no reliable list of all members of such target populations from which to draw, with equal likelihood of selection, fully identified units – individuals or households – that could then be contacted for data collection. Therefore, generating a reasonably representative sample of such a population amounts to squaring a circle.
The solution proposed here to this conundrum consists of constituting a proxy universe, based on available information about some population characteristics, from which to draw a proxy sample of respondent search profiles. Short of full information on final units, researchers might still be able to obtain information on general parameters of the study universe, such as place of origin (specific countries or geopolitical groupings), gender distribution (within each geopolitical category) and prevailing socio-economic features of current location (municipality of residence), a proxy of employment prospects. Such information can be used for drawing a proxy sample – i.e., one that specifies each respondent’s characteristics, regarding each of those dimensions. No numerical match between proxy universe and target population is expected, nor is it warranted; the proxy sample only goes half the distance toward identifying any particular interviewee: his or her inclusion in the source file is not a prerequisite for selection as interviewee, but a questionnaire item instead (which a posteriori, will allow gauging the share of the target population included in that source). The only assumption made here is that the geographical distribution of those not included in the source file, in terms of socio-economic profile of their settlement areas, is similar to the residence patterns of those comprised in that source, which is used here for estimating the proportions of relevant population characteristics that define those interviewee search profiles.
For this approach to generate acceptable data quality, it is decisive how exactly interviewees are recruited by fieldwork staff; again, survey methodology’s standard requirement of equal selection likelihood for all individuals matching any given search profile is unrealistic. However, selection bias can be minimized by diversifying the contact points for interviewee recruitment, with strict quota to be met by fieldwork staff, and compliance enforced by various layers of quality control.
The procedures proposed here were first developed and employed by the Institute for Advanced Social Studies (IESA-CSIC) when surveying highly dynamic immigrant populations in Southern Spain (N=1,797). Thus, this study’s output is assessed, both by formal indicators and by comparison with data subsequently generated by third parties, with a view to gauging the representativeness obtained by the proposed procedures. The paper concludes by weighing the benefits of proxy sampling against its inconveniences, which include high costs and massive workload.
A core task of the European Union Agency for Fundamental Rights (FRA) is to collect and analyse objective, reliable and comparable data through scientific research and surveys to provide the relevant institutions, bodies, offices and agencies of the European Union and its Member States with assistance and expertise relating to fundamental rights.
Despite the continuing demand for data on immigrants and ethnic minorities and an increasing availability of socio-economic migration statistics, a considerable lack of data comparable across the EU on fundamental rights issues concerning immigrants and ethnic minorities persists. The reasons are manifold such as diverging definitions of the target groups (e.g. by ethnicity, country of birth and country of birth of parents, nationality and citizenship) and difficulties to properly cover the target population with traditional data collection methods.
One of the main challenges faced when surveying hard-to-reach groups is the lack of sampling frames or their incompleteness. A cross-country and/or cross-cultural survey design introduces additional complexity in surveying immigrants and ethnic minorities. The heterogeneity of applied methodologies (sampling, data collection modes, questionnaire design, translation and weighting), as well as regarding legal status, language proficiency and cultural norms of the target populations affects the results’ coherence across different immigrant groups and across countries. Moreover, standard questionnaire classifications, such as ISCED for educational attainment, cannot always be easily applied to immigrants and therefore call for the development and application of new concepts.
The paper discusses these challenges by outlining the approach of the second European Union Minorities and Discrimination Survey (EU-MIDIS II), which FRA conducted in 2015/2016 to assess progress over the past seven years since the first EU-MIDIS survey was carried out in 2008.
This second EU-MIDIS survey collected comparable data in all 28 EU Member States to assist EU institutions in developing evidence-based legal and policy responses to respect, protect and fulfil the rights of persons with immigrant or ethnic minority background, including Roma. The data will also serve to populate core indicators for measuring progress in the implementation of the EU framework for National Roma Integration Strategies. More generally, the survey will also provide data on selected indicators on immigrant integration. The survey covers topics such as experiences of discrimination in different areas of life (labour market, education, housing, health and other services), criminal victimisation (including hate crime), social inclusion and societal participation. The data are collected through a combination of different sampling methods designed to obtain representative samples of the target populations in the different EU Member States.