The goal of this project was to understand how migrants and other refugees make decisions regarding onward migration, stay, or return in Greece and Turkey. We used a mixed-methods approach of a migrant to migrant survey and qualitative interviews. This case study explains the steps taken in the research practicalities to implement the project including the selection and training of fieldworkers, questionnaire programming, and survey checking. As refugees and other migrants are a vulnerable group, the informed consent process and the use of reciprocity within this study are discussed. The research analysis explores how the mixed-methods approach brought forth different components of the results. In the conclusion, key results and lessons learned are explored.
By the end of this case, students should be able to
- Discuss the benefits of a mixed-methods approach
- Identify the steps required to create a fieldwork team
- Discuss the ethical issues of doing surveys with refugees and other migrants
- Identify elements of reciprocity in conducting fieldwork
Project Overview and Context
The “Understanding Irregular Migrants Decision Making Factors in Transit Project” was funded by the Australian Department of Immigration and Border Protection and the Australian National University’s Collaborative Research Programme on the International Movement of People. The goal of this project was to increase understandings of how refugees and other migrants make decisions within their migration journeys and in transit contexts. Most research on migrant decision-making focuses on the decision to leave the country of origin or the decision to return. Within the migration journey and transit countries, decision-making is more complex as an individual can choose to stay, migrate onward, or return. Furthermore, there are additional decisions on where to go, how to go, and when to go for those seeking to migrate onward. There is a recognition within migration research that we know too little about these details of migrants en route and what is happening between their origin and destination (Benezer & Zetter, 2015; Mainwaring & Brigden, 2016).
Increasing these understandings is important for the policy context. In 2015, over a million asylum seekers came to Europe, many of whom had been living in Turkey, Iran, or other countries for a significant period of time. This resulted in the European Union (EU)–Turkey deal being signed in 2016, which essentially created a one-for-one arrangement wherein all non-Syrians coming from Turkey to Greece would be returned, and for each Syrian who came and was returned, another Syrian would be resettled from Turkey to the EU. The effect was a significant decrease in the number of asylum seekers coming from Turkey to the EU in 2016 and 2017.
The data in this project were collected in Turkey and Greece from May to July 2015. The information gathered in this project reflects on the decision-making factors of refugees and other migrants in each country in deciding to stay, migrate onward, or return from each country.
The research design for this project included a mixed-methods approach. Most research on migration journeys and migrant decision-making is qualitative due to the complexity of decision-making and that it may change over time. In this project, however, we sought to use quantitative methods as we wanted a large enough sample to be able to conduct a regression analysis and understand correlations between different variables.
We decided to use a migrant-to-migrant approach, meaning that migrants would use computer-assisted personal interviewing (CAPI) methods to complete the questionnaire on tablets in face-to-face interviews. The benefit of this approach is that both the interviewer and respondent speak the same language, often have some similarities in experience in that they are both migrants, and migrant interviewers are able to gain access to locations that foreign researchers would find difficult (e.g., smugglers houses).
Recognizing, however, that closed-ended surveys exclude details on decision-making, we also incorporated qualitative interviews into the research design. These interviews were conducted with individuals that had already completed the closed-ended interview survey in an effort to gain further details on their experiences and decision-making.
There are several research practicalities that need to be navigated in implementing a migrant-to-migrant survey. The following steps were taken:
- Questionnaire development: The questionnaire used a life cycle approach, meaning that it began with the individual’s experiences before their migration, then continued to their decision to migrate, experiences during their migration, their arrival in the current country, current experiences, and then their decision-making if they wanted to stay, migrate onward, or return, and follow-up questions regarding this decision. In developing the questionnaire, we had an advisory team that reviewed the questionnaire and offered feedback. We went through several rounds of the questionnaire before finalizing it for the fieldwork.
- Informed consent: The questionnaire began with a detailed informed consent process. Considering that refugees and other migrants are a vulnerable population, it was important to be very clear about the purpose of the survey, how participation may or may not affect them, that it is confidential, and how the results will be shared and used. Within the preamble, fieldworkers explicitly stated “I also am not in a position to help you with the authorities or your situation here.” This was important as we wanted to ensure that individuals did not complete the survey with the expectation of assistance in return.
- Questionnaire pilot: The questionnaire was piloted in a planning trip to Turkey before the fieldwork. A key set of questions within this questionnaire was on the factors that influenced the individual’s migration decision. We had a list of 25 different variables in the onward migration sections. We tested two ways: asking the open-ended question and filling in the individual’s response and asking each of the 25 items for a yes or no answer. We found that when asking the open-ended question, we received only one or two items (even after probing for more), whereas when we asked the 25 items, we received a yes to roughly half of the variables. Considering that we wanted to understand the impact of these different variables, we decided to use the second approach and ask each variable.
- Translation: Once the questionnaire was finalized, we had it translated into each of the required local languages for our target population (Afghans, Iraqis, Iranians, Pakistanis, and Syrians).
- Programming the questionnaire: We decided to use Kobo Toolbox as our survey methodology as it was a free software, could be used offline as well as online, and colleagues reported having a good experience with the software. The lead researcher learned the software, so that once in the field she could trouble shoot any problems that arose.
- Fieldworker recruitment: To recruit fieldworkers, a local project manager was hired in each location. This project manager went to non-governmental organizations (NGOs) working with migrants and introduced the project. They then asked for recommendations on individuals to hire for conducting the surveys and to also place an advertisement at the NGO. Local academics who had conducted research were also asked for recommendations regarding translators and migrants they had worked with. All applicants were then asked to come for an interview with the researcher and local project manager. They were interviewed on both their experience and their level of comfort in approaching people on the street to conduct interviews, time availability, and other practical issues required for conducting the research.
- Fieldworker training: Once the fieldworkers had been selected, a 5-day training took place. Mandatory attendance was required at all days to work as a fieldworker on the project (training was also paid). A training manual was developed that was given to each respondent and included information on
- Project background;
- The overall methodology;
- Overview of the questionnaire including guidelines on how to use the tablets, types of questionnaire questions, skip patterns, ranking questions, and scale questions;
- How to select and approach participants including informed consent;
- How to conduct an interview including building trust, privacy, objectivity/neutrality, interview behavior, and professionalism; managing sensitive issues; asking questions; and interpreting answers, how to handle long and elaborate answers, how to handle socially desirable answers;
- What to do after the interview including confidentiality;
- Potential problems and management;
- Field safety protocol;
- Daily reporting and expectations.
Four days of the fieldworker training were in a training center and included role-play, practicing the questionnaire, and testing of the fieldworkers.
- Pilot testing the survey: The final day of the fieldworker training was a pilot testing day for the fieldworkers to go out and complete interviews. On this day, all questionnaires were checked immediately at the end of the day and the fieldworkers reported back on any challenges they experienced or questions they had regarding the fieldwork. After this was completed, the official fieldwork began.
- Survey implementation: All surveys were implemented using CAPI methods. This allows for the surveys to be directly uploaded at the end of each fieldwork day and for the researcher to be able to check the data for consistency. All tablets were password protected, which also provides further data protection to the respondents than using paper surveys that could be potentially lost or stolen.
- Respondent recruitment: As there is no census available of the irregular migrant population in each country, it was not possible to use random sampling, and instead the research team used multiple points of entry to find respondents. Entry points included approaching migrants on the street (Greece = 226, Turkey = 115), referrals through a limited number of migrant organizations (Greece = 29, Turkey = 14), and the networks developed in the field (Greece = 25, Turkey = 187). From these entry points, snowball sampling was used to find further respondents (Greece = 249, Turkey = 213).
- Survey checking: After surveys were completed, survey checking was conducted on an ongoing basis. Approximately one in every three respondents interviewed was called to check the survey results. If there were any discrepancies between the information collected during the survey and the checking, this was recorded. An assessment was then made regarding each individual discrepancy, and in some cases, the original data were used; in others, revised data were used; and if the data were deemed unreliable, they were discarded (one case).
- Qualitative interviews: The final method used in this study was follow-up in-depth interviews with selected survey respondents. The main selection criteria were employment (we interviewed a mix of employed and unemployed respondents) and complexity of case (where the survey was not optimal for capturing the individual’s full story). We sought a balance between the different countries of origin and gender. All interviews were translated and transcribed into English and coded for analysis.
- Practicing reciprocity: Reciprocity is an important concept in refugee research. Mackenzie, McDowell, and Pittaway (2007) argue that although the principles incorporated under “do no harm” are critically important, they are not enough and “that researchers need to move beyond harm minimization as a standard for ethical research and recognize an obligation to design and conduct research projects that aim to bring about reciprocal benefits for refugee participants and/ or communities” (pp. 12–13). One way that we tried to do this in our study is to practice reciprocity by helping respondents whenever we could. It is important to contrast here, that we explicitly say in our recruitment we are unable to help people as we do not want them to participate in the survey with the expectation of something in return. However, all fieldworkers were given a list of NGOs working in the area and services that they offered to refugees and other migrants. When an individual was interviewed and if they were in need, fieldworkers would refer them to NGOs and provide them information on where they could go to access help. Finally, at the end of the project, we had small resources where we were able to bring some food and supplies to some of the most vulnerable individuals interviewed.
- Data storage: Ethics in data storage is also very important. Recalling that we collected phone numbers, these were all deleted at the end of the fieldwork and not stored in the data set. The data were kept under password protection and safely stored. It will be deleted after 7 years.
Method in Action
On the whole, the methodology worked well. The beginning of survey implementation is always difficult and challenging to find initial respondents. Fieldworkers experienced frustration and concern at the beginning regarding not being able to find respondents. This was managed by continued encouragement and support, and after the initial week, this improved and fieldworkers were able to find respondents. As many people were approached on the street, there were a high number of refusals to participate; however, this is to be expected and fieldworkers became used to this.
The CAPI surveys worked very well and few issues were encountered. In the initial few days, the fieldworkers reported that some response categories were not available in the questionnaire that respondents were stating. A few on the spot response categories were added to the questionnaire to avoid large numbers of “other” responses. This was possible as the lead researcher knew the software and could do this on the spot. Some days the fieldworkers were not able to connect to Wi-Fi right away and this led to a delay in the survey being uploaded.
At the end of the survey, respondents were asked for their phone numbers to follow-up and check survey responses. Some respondents did not have a mobile phone and were not able to give a number, and some did not want to share their phone number. This was anticipated from the beginning, but meant that these respondents could not be called for follow-up. This occurred, however, in only a small number of cases.
The analysis for this project had three components: (a) a descriptive overview of the survey results, (b) regression analysis, and (c) mixing of the qualitative data with the quantitative results to further examine certain topics. The first step was a cleaning of the data to check for incomplete or inconsistent respondents.
The descriptive results were informative in understanding the current conditions of refugees and other migrants in Turkey and Greece in 2015 and the extent to which they want to migrate onward or not. It was striking that more people were planning to migrate onward from Greece, an EU member state (74%), than Turkey (59%), a country usually described as middle income. Unsurprisingly, very few respondents wanted to return: only 2% from Greece and 7% from Turkey.
Returning to the issue in the initial questionnaire design of the 25 response categories, the results showed a range from 8% to 94% of respondents saying yes to some of the variables. Descriptively, this enabled a further understanding of what variables were and were not considered important to refugees and other migrants in making their onward migration decisions.
Due to the nature of the country contexts, a few of the respondent categories in the survey were not identical in Turkey and Greece. For example, in Turkey, respondents could apply to United Nations High Commissioner for Refugees (UNHCR) and request an interview for resettlement. The wait times for these interviews could be as long as 2 years. In Greece, this process was not in place. Therefore, in the Turkey context, it was relevant to have “resettlement wait times are too long” as a variable for explaining the decision to migrate onward. As the data sets were not exactly the same, we opted to not merge the data sets and conducted analysis on Greece and Turkey separately (Koser and Kuschminder, 2016). In a separate paper, only Afghan migrants were examined, and here all Afghans were merged as the analysis had a different form (Kuschminder, 2017).
We conceptualize migrant decision-making based on the question, “At this moment, do you want to (a) stay in Greece/Turkey, (b) migrate to another country, (c) return to your country of origin, (d) return to the country you were last living in?” For the Greece sample, a standard probit regression is used where the value 1 represents wanting to migrate onward and 0 represents wanting to stay in Greece. The model predicts the probability that an individual will migrate onward. The number of people selecting return in Greece was too small to include in the model. In Turkey, the return option, including both return to your country of origin or the country you were last living in, was large enough that we were able to use a multinomial probit regression to estimate the probability of wanting to stay in Turkey as compared with migrating onward and the probability of return as compared with migrating onward. A multinomial model is used as the dependent variable now has three options: migrate onward, stay, or return. This model results in two columns: first, stay versus migrate onward and second, return versus migrate onward. The models are estimated with robust standard errors and results are presented as average marginal effects.
Several independent variables were tested based on a literature review of factors influencing decision-making and the use of a model developed for testing migrants’ decision-making factors in transit. Conditions in the country of origin were assessed through the use of origin country and reason for the initial migration. Conditions in the country of transit were represented through current migration status, subjective living conditions, employed, experienced abuse, and speaking the local language. The destination country is represented through whether or not Turkey/Greece was the intended destination country in the initial migration.
Policy incentives and disincentives are represented by having previously attempted to migrate onward and having used a smuggler in the initial migration. In essence, if someone tried to migrate onward and was prevented from doing so because of a disincentive (border controls or police restrictions to movement, etc.). These restraints act as a proxy for a policy disincentive. Smugglers are used to bypass border controls, which are a manifestation of a policy disincentive for migration. Finally, individual factors are represented through age, gender, marital status, educational attainment, and if the migrant is from a rural or urban background.
The qualitative results were important in this study in providing further context and understanding of the individual’s experiences and decision-making. First, the qualitative interviews revealed the challenges of the migration journey to Turkey for both Afghan and Pakistani migrants. These interviews showed that both Afghans and Pakistanis can be subject to kidnapping and torture for ransom extortion in Iran. For example, one participant stated,
They were torturing me and one other guy very badly … By putting red hot rods on our legs and I can show you the marks of those red rods on my legs as well. Anyways, they forced us to call to Pakistan to our parents; they were torturing us so much and beat us really hard so that our voices might travel so that they might arrange money to send them. They were demanding around 1000 dollars minimums at that time from our families. (Pakistani male, Turkey)
Second, through the qualitative results it was found that some respondents with refugee status in Greece planned to stay in Greece only long enough to receive a refugee passport, on which they could legally migrate to Germany. Once in Germany they planned to re-apply for refugee status, with the goal of staying in Germany for the long term.
Third, the interviews revealed frustrations with the system and the current process and how policy revisions could change decision-making. For example, one respondent stated,
I find it strange that they make us come, we who want to travel to other countries, they leave us in a country in the middle, where we lose all our money and then we go empty-handed to those countries. Why can’t we be directly accepted in one of those countries, while we still have money and still in control? Why? And even those who go illegally, they welcome them, and that’s already someone who’s breaking the law, and they accept them. Why don’t they accept those who obey the law? (Iraqi Male, Turkey)
This quote shows that this respondent did not want to migrate irregularly and wanted to wait for a regular opportunity such as resettlement. Therefore, if more opportunities were available for resettlement, individuals would feel less of a need to migrate onward irregularly.
The addition of the qualitative research brought forth issues that would not have been identified in the quantitative component and provided a strong complementary approach.
Practical Lessons Learned
This study highlights the multiple steps that need to be taken in implementing a migrant-to-migrant survey and the benefits of using a mixed-methods approach. Implementing a survey takes a lot of preparation and planning. Pre-trips were held to each field-site before arriving for the actual fieldwork. Based on previous survey implementation, several lessons were incorporated into this study such as having a mandatory paid training, establishing high standards for fieldworkers to meet at the training, and not continuing with fieldworkers that did not meet this standard for the actual implementation. It is essential to ensure a strong team of fieldworkers to have a good implementation.
Furthermore, it is important to make sure your fieldworkers are treated well during the implementation. A key practical lesson in this research was that these were very difficult interviews to conduct for the fieldworkers. This is because often respondents shared beyond the survey questions discussing issues that they experienced of rape, torture, abuse, family separation, and other topics. In the training, we discussed that interviews may be difficult and to discuss experiences within the field team for de-briefing and to take time off between interviews that were challenging.
Finally, recognizing that refugees and other migrants are a vulnerable population, practicing reciprocity and encouraging the fieldworkers to assist people where possible are very important. In several instances, fieldworkers walked respondents to an NGO and introduced them so that they knew the services available. In one case, a fieldworker took someone to the hospital who needed medical assistance and stayed with them to translate so that they got the care they needed and were calm because they knew what was going on. Although clearly as researchers we cannot intervene in respondents’ situations, when individuals needed help and the team could assist in these small but meaningful ways, it was always encouraged and supported.
This project resulted in the collection of 1,056 surveys and 60 qualitative interviews collected with refugees and other migrants from May to July 2015 in Turkey and Greece. The timing was important, as it unexpectedly coincided with the beginning of the so-called “European migration crisis.” Our findings helped to elicit understandings of why refugees and other migrants wanted to move onward from Turkey and Greece at this time.
The methodology we used was a mixed methodology of migrant-to-migrant questionnaire interviews and follow-up qualitative interviews. The approaches combined provided different insights into refugees and other migrants’ decision-making for onward migration, stay, or return in Greece and Turkey. In working with refugees and other migrants, several steps were taken to ensure protection for the respondents and the fieldworkers. We used reciprocity whenever possible to assist respondents. The methodology worked well and the resulting data set is still being used for further analysis.
Exercises and Discussion Questions
- How did the mixed-methods approach contribute to the findings?
- What are the steps to consider in completing a migrant to migrant survey?
- What are the benefits of CAPI?
- How can questionnaire question design affect your findings?
- How can you practice reciprocity in your research?