Research with Refugees in Fragile Political Contexts: How Ethical Reflections Impact Methodological Choices

Researchwithrefugeesposesparticularethicalchallenges,especiallyifdataiscollected inplaceswheremostrefugeestodaylive:namelycountriesneighbouringconﬂict,ones thataresometimesatwarwiththeircountryoforiginandwhererefugeesareexposedtodifferentdegreesoflegalvulnerability,posingsecurityriskstoparticipantsand researchersalike.Thesechallengesareexacerbatedwhendataiscollectedacrosscoun-triesandincludessurveyresearch.Thearticleaddstotheemergentliteratureonethics inforceddisplacementbyhighlightinghowsecurityprecautionsandethicalconsid-erationsinﬂuenceandshapemethodologicalchoices.Basedonrecentﬁeldworkwith SyrianrefugeesinLebanonandTurkeyin2018,thearticlediscussesamixed-methods approachcombiningin-depthinterviewswithanindividualsurveybasedonmulti-stageclustersampling,randomwalksandlimitedfocusedenumeration.Advocating forarefugee-centredapproach,itelaborateson:(i)howtonegotiate‘ethicsinpractice’; (ii)howrisksandviolenceinﬂuencethechoiceofﬁeldworksites;and(iii)howethical considerationsimpactinparticularquantitativeormixed-methodsstudies.It describestheadvantagesofincludingmembersofrefugeepopulationsinresearch teams,aswellasopenchallengeswithregardtorisks,informedconsent,conﬁdential-ity,sensitiveissues,positionality


Introduction
Research with refugees poses particular ethical challenges because of unequal power relations, legal precariousness, poverty, violence, politicized research contexts and the policy relevance of the research in question.These challenges are exacerbated if data is collected in places where most refugees flee to: namely neighbouring countries in the Global South, which are sometimes at war with and Turkey.Both countries were initially open to the influx of Syrian refugees and both have increasingly become relatedly restrictive.Lebanon officially hosts over 1 million registered Syrian refugees (UNHCR, 2017) and has one of the highest refugee ratios in the world: one-quarter of the country's population is such an individual.Turkey currently hosts the world's largest refugee population, with 3.5 million registered Syrians there in 2018 (UNHCR 2018).Both countries have been entangled in the conflict in Syria militarily and, in both, the 1951 Geneva Convention is not applied to Syrians either-leaving their legal status, to divergent degrees, fragile.
The first section of the article gives a brief overview of the state of the art and identifies gaps in the literature.The second part then discusses the legal and security contexts in Lebanon and Turkey, while the third highlights how these influenced the research design and locations.The fourth section elaborates on over-researched localities and suggests less-studied urban spaces engaged with through random walks as an alternative.The last part reflects on trust relations, sensitive questions, informed consent and how to include refugees in the analysis process, dissemination and advocacy.
The Missing 'Ethics in Practice' Literature in Fragile Political Contexts An emergent body of literature is now focusing on ethics in forced displacement (Bloch 1999;Jacobsen and Landau 2003;Guillemin and Gillam 2004;Rodgers 2004;Bakewell 2007;Bloch 2007;Dona´2007;Ellis et al. 2007;Mackenzie et al. 2007;Du¨vell et al. 2010;Hugman et al. 2011aHugman et al. , 2011b;;Block 2012;Sukarieh and Tannock 2012;Clark-Kazak 2017;Clark-Kazak 2019;Karooma 2019;Luetz 2019;Omata 2019).These works have mostly honed in on issues of voluntary, informed consent, confidentiality, the 'do-no-harm' principle, a favourable risk/ benefit ratio and respect for participants, as well as on data protection.In recent years, collaborative and participatory methods have been proposed as a way to address some of the ethical issues arising, that by involving marginalized and disenfranchised groups so as to ensure that refugees' voices are sufficiently heard (Ellis et al., 2007;Hugman et al. 2011).
There have also been efforts to further develop ethical guidelines for research on forced displacement.A recent example is the IASFM's aforementioned Code of Ethics.It is based on a broad definition of people in situations of forced displacement and adapts key ethical principles of voluntary, informed consent, confidentiality and privacy, and do no harm to the specific contexts of forced migration.The Code of Ethics upholds principles such as autonomy, equity, diversity, competence and partnership.It underlines in particular the importance of (i) actively including refugees as partners in research projects (ii) while recognizing the diversity of their experiences, (iii) mitigating the effects of intersecting, unequal power relations, (iv) using methodological approaches that are adapted to the contexts in which research is conducted and (v) contributing time and labour to projects and activities within the studied communities that are unrelated to the research itself (Clark-Kazak 2019: 14).
Yet, while ethical guidelines and complexities are well recognized, practical solutions to these challenges are less well described-especially with regard to: (i) how to negotiate ethics in practice; (ii) how risks and violence influence the choice of fieldwork sites; and (iii) how ethical considerations impact quantitative or mixed-methods studies.First, bodies charged with the ethical review of academic research, such as institutional ethics committees, are often ill-equipped to provide proper oversight of research involving refugees and internally displaced persons (IDPs) (Mackenzie et al. 2007: 300), especially if research is conducted far away.Second, ethical questions are not static and cannot all be solved before data collection begins.They need to be considered and reflected upon from the moment a research project is first conceived until its findings are eventually presented (Liempt and Bilger 2009: 13).Guillemin and Gillam (2004) suggest looking into ethics in practice, which involves responding to the everyday issues that can arise in the processes of research.Researchers are hereby faced with competing duties, obligations and conflicts of interest, with the need to make implicit or explicit choices between values and the diverging interests of different individuals and groups (Refugee Studies Centre 2007: 164).
Second, relatively little is known about what drives fieldwork choices from an ethical standpoint.Yet, our selection of fieldwork countries, regions and sites is in part determined by what is possible and safe-for those we study, but also for us who do research.The most repressive regimes within the authoritarian universe are less likely settings for field research because it would be too dangerous or simply because it is impossible: examples are Saudi Arabia, Turkmenistan, North Korea or lately Egypt (Glasius et al. 2018: 22-23).However, few articles reflect on how risks and violence influence the choice of fieldwork sites.The literature on qualitative research methods, for example, largely focuses on democratic regimes and not on the Middle East or post-communist countries (Rivera et al. 2002;Clark 2006).Clark's (2006) article is an exception.The survey, which she conducted with researchers working on the Middle East and North Africa, showed that only 16 per cent of respondents specified 'the political situation' and safety as contributing to their country choices.Other researchers reported having limited the research to certain areas considered safer or privileging certain methods over others.Vignal (2018), for example, interviewed Syrians in Lebanon about their lives and economic practices in wartime Syria as research inside their native country was not possible.Kastrinou (2018) chose to use mediated communication with friends and family inside Syria via telephone, Skype and Facebook.
Third, there is significantly less literature on ethics in forced displacement when it comes to quantitative research and, in particular, projects encompassing individual or household surveys.Many quantitative studies describe their sampling strategy without addressing ethics.Mandi c and Simpson's (2017: 76) study about anti-smuggling policies and migratory risks for Syrian refugees in five countries, for example, does not specify how individual respondents were chosen nor the potential risks involved.Similarly, Bohra-Mishra and Massey (2011) and Adhikari (2013) do not address how risks and ethics influenced their sample strategy when studying forced displacement in Nepal.An exception to the rule Research with Refugees in Fragile Political Contexts 2311 is Mulumba's (2007) article on Sudanese refugees in Uganda, in which she describes the challenges involved in getting clearance from the United Nations High Commissioner for Refugees (UNHCR) and district officials.This lack of discussion on ethics is particularly disturbing, as data collection via survey research is much more visible-also to authorities-than qualitative fieldwork, which focuses on one specific site and allows researchers to much more easily submerge in a research site and seek advice from local communities.
The literature is particularly thin when it comes to quantitative research in countries that witness regular violence and political tensions or in contexts where refugees live under acute legal vulnerability.Contrary to what one might otherwise think, survey research is indeed possible in politically fragile or authoritarian contexts-on some topics, in some countries.However, it raises additional challenges as, in survey research, one typically has no direct contact with one's respondents, no way of gauging their reactions and no chance to course-correct once data collection has actually started (Glasius et al. 2018: 45).Only a few articles address these issues.Doocy (2015), for example, conducted a needs assessment of IDPs inside Syria and stressed that reaching a representative sample was extremely difficult, as access to certain regions was complicated by security issues.The team excluded Governorates and communities if participation in the assessment would potentially present a security threat to respondents or interviewers.Clark (2006: 19) reported three main challenges among researchers who did survey research in Middle Eastern countries: (i) the reluctance of people to complete the surveys and/or the tendency to give potentially misleading answers (those in line with official rhetoric); (ii) their concern that participants might not be familiar with the meaning of intended concepts, given their relative newness in some of the countries of the region due to restrictions on freedom of speech; and (iii) tremendous variation across countries vis-a`-vis the ability to conduct survey work.

Working in Fragile Political Contexts with Different Degrees of Legal Vulnerability
First of all, it should go without saying that, in order to be able to take appropriate ethical decisions, researchers have to fully understand the political and legal contexts in which refugee communities live.Lebanon and Turkey are both major destination countries for Syrian refugees, are both militarily and politically involved in the civil war in Syria, and refugees from there are, as noted, to divergent degrees legally vulnerable in both countries as the 1951 Geneva Convention is not applied to them.Legally, both countries hence do not provide refugee status or subsidiary protection to Syrians.
Before the war, Syria had become an important business partner for Turkey: cross-border trade and investment flourished and a visa-free policy allowed citizens of both countries to move at will back and forth between the two.However, Turkey became increasingly critical of the Assad regime once the uprising started.In summer 2011, Ankara started to openly back both the Syrian National Council, the major civilian coalition in exile seeking the overthrow of Assad, and the Free Syrian Army (Balci 2012).In parallel with these developments, the Kurdish conflict grew into a core political issue-one that ended a period of relative rapprochement in 2015.Since then, the Turkish military has been directly involved in Syria, with three offensives in North Syria: a cross-border operation against forces of the Islamic State of Iraq and the Levant and the Kurdishdominated Syrian Democratic Forces in 2016-2017; Operation Olive Branch against Kurdish-controlled Afrin in spring 2018; and a military operation in autumn 2019 aimed at removing Kurdish-led forces from the border area and creating 'safe zones' for Syrian refugees.At the domestic level, since the failed coup d'e´tat in July 2016, Turkey's government has increasingly repressed its opponents, including critical researchers, civil society and media outlets-in parallel with a two-year-long state of emergency being declared.This had led to concerns that the Turkish government might move towards authoritarianism in certain parts of its territory, in the sense that it fails to conduct fully free and fair elections, curtails freedom of expression and of association, and that there is some arbitrariness to its governance too (Glasius et al. 2018: 6).
Legally, Syrians in Turkey have been granted temporary protection (TP) under the Law on Foreigners and International Protection enacted in 2014.The provisions of the TP regime include the right to legal stay, the principle of non-refoulement and a deportation ban (Article 6), the right to health services (Article 27), the right to education services (Article 28) as well as principles governing the operation of and support in the camps (Articles 36-41) (To gral Koca 2016).However, the regulations fall short of providing the explicit right to work.TP holders can apply for a work permit, but there is no guarantee that government authorities will grant it (Ineli-Ciger 2014: 32-34).Furthermore, due to a presidential decree, it is possible to circumvent the principle of non-refoulement under certain conditions: namely if there is a threat to public health, morality or order.This has been used, for example, to deport back to Syria-regime-critical Kurdish Syrians who support the People's Protection Units (YPG) as well as IS sympathizers (interview, human rights lawyer, Izmir, 24 November 2017).
Over the past few years, human rights organizations have also pointed out that Turkish border guards continue to carry out pushbacks and to injure and even kill Syrians as they attempt to cross the border (Human Rights Watch 2018).Furthermore, in early 2016, permits for travelling within Turkey were introduced.Finally, since 2017, several municipalities such as Istanbul and cities located along the Turkish-Syrian border have halted TP registration, leading to legal limbo for those who stay.Finally, the Turkish government does not allow returnees to seek refuge under TP again, as voluntary departure is considered grounds for the cessation thereof.The legal status of Syrians returned to Turkey under the EU-Turkey Deal remains unclear.Returned Syrians were detained for some weeks, officially for the purpose of identification and security checks (Tunaboylu and Alpes 2017).
Lebanon and Syria share a long-and difficult-history meanwhile.Lebanon has been a destination for labour and educational migration from Syria since the Research with Refugees in Fragile Political Contexts 2313 1950s.During the Lebanese civil war, the Syrian army had a long-standing presence in the country (including its security apparatus), which only ended in 2005.Some Lebanese refer to the presence of Syrian refugees in the country as a second Syrian occupation (France24 2017).Politically, Lebanon's political system continues to be fragile and dominated by confessionalism.Lebanon did not have a government between 2014 and 2016, partly explaining the closer involvement of international organizations in the management of refugee governance as compared to Turkey (Issa 2016).Lebanon remains divided between an anti-Syrian 14 March Alliance (uniting Sunni forces and a section of the Christian parties) and a pro-Syrian 8 March Alliance (Shia parties and the Maronite Free Patriotic Movement).The Shia party Hezbollah is an active part of the current Lebanese government, controls large areas of the country and continues to be substantially involved in the Syrian war on the side of the Assad government.On the other hand, the fear that Syrians-who are in a large majority Sunni-might change the confessional balance of the country has caused Christian leaders to become vociferous promoters of restrictive policies towards these refugees (The Daily Star 2018).
Lebanon itself has always refused to be an asylum or resettlement country, and regards itself as a transit one for Syrian refugees.Although the UNHCR can register refugees, the Lebanese government does not grant legal effect to the UNHCR's recognition of refugee status.Registered Syrian refugees are thus theoretically considered to be waiting for resettlement to another country.The Lebanese government also pursues a no-camp policy: according to Lebanese political discourse, Syrian refugees should not stay in the country as Palestinians once did-with the latter's camps having turned into permanent settlements (Dionigi 2016).Since October 2014, Lebanon has progressively implemented a restrictive legal framework.To enter the country, Syrians now need valid identification documents and one of the various temporary visas that exist.Other measures include an annual fee of USD 200 to apply for or renew a residence permit.Furthermore, Syrians are legally only allowed to work in the fields of construction, agriculture and environment (i.e.cleaning).Finally, UNHCR registrations have been suspended since 2015 and the right to employment barred for those already registered with the body.Equally, a Lebanese sponsor is required for those not registered with the UNHCR (Kahwagi and Younes 2016).This legal framework has resulted in a situation in which a vast number of Syrians lack valid residence permits and work in irregular conditions.Without a valid permit, refugees are considered to be in breach of the law-and thus may be detained by the security forces and forcibly returned to Syria.While Lebanon publically adheres to the principle of non-refoulement, it in fact engages in 'legal deportation'-consisting of the issuance of deportation orders informing refugees that they must leave the country.The chosen measure of refusing to grant residence permits seems to place Syrian refugees in such vulnerable positions that it might drive them to leave (Frangieh 2015;Janmyr 2016).
The remaining three sections elaborate on the ethical challenges that I faced when conducting fieldwork in such environments.During this time, I closely collaborated with a local team of young Syrians living in Lebanon and Turkey respectively.The project was based on qualitative interviews (trying to cover a wide array of different experiences) and an individual survey, which had the objective of ideally being representative of the fieldwork sites studied.

Building Partnerships and 'Doing No Harm' in Fragile Political Contexts
Given that I-a white and female researcher with a European passport conducting research out of a Dutch university-and Syrian research participants in Lebanon and Turkey might have at least partly disparate lifeworlds (Block et al. 2012: 71), I conceptualized the project from its very beginning as including in the team young Syrian refugees who live in the country where the research was to be conductedto discuss interview questions, data collection and data analysis, and to train them for potential future research projects.This was meant as a strategy to avoid the research potentially inflicting 'symbolic violence' through misunderstanding or misrepresenting research participants (Bourdieu 1996).A number of observers have criticized the fact that researchers often ignore the values, lifestyle and the cognitive and affective worlds of their subjects, instead imposing their own on them (Sieber, 1992: 129).While some scholars have pointed out that it is best to minimize the use of refugee interviewers as they might be perceived to be politically positioned in the conflict by participants (Jacobsen and Landau 2003), others have argued that including members of the same ethnic, cultural, linguistic and/or gender group might help to increase cultural sensitivity and facilitate relationships of familiarity and trust (Bloch 2007;Du¨vell et al. 2010).I considered that giving members of the researched community a voice in the project was ethically more important than the potential side effects of being perceived as politically positioned in the country and decided to seek out ways to mitigate these effects.
As other researchers working in the region also have (Clark 2006: 420), I consciously decided against outsourcing data collection to a research or survey company.Apart from financial restrictions and the lack of reliable companies working in Lebanon and Turkey, I had the feeling of not being in control of how data would be collected across two different countries-in regard to not only research ethics, but also data quality.I also hypothesized that collecting data with Syrian refugees living locally while being present and conducting interviews on a face-toface basis was needed to ensure the necessary trust was generated to gain access to people and to ensure data was interpreted correctly.As a side note, one of my research assistants mentioned to me one day in passing that he was surprised that I was part of the whole data-collection process.He then shared an anecdote about friends who had done research in Syria for international non-governmental organizations (NGOs) as enumerators and how easy it was to fake survey questionnaires.
Another important decision taken was seeking to mitigate the effects of intersecting, unequal power relations by conducting the entire data collection in Syrian Arabic-including the interviews conducted by myself.As preparation for the fieldwork, I thus strengthened my spoken skills in Syrian Arabic.Building on Research with Refugees in Fragile Political Contexts 2315 my previous studies of Arabic, I began to take online conversation lessons one year before the start of data collection with a Syrian Arabic teacher living in Europe.By chance, he had worked in the same institute in Damascus where I had taken my first language class in Syria in 2006.I discussed the project repeatedly with him, and he was later also responsible for the translation of the survey and interview questionnaire into Syrian Arabic.Leading up to data collection, I also volunteered 1 month respectively in each country with an NGO offering educational and recreational activities for Syrian children.I did not consider this to be part of the research, but a way to better understand the life realities of some Syrians in the two countries.It also brought me into closer contact with civil-society actors as well as Syrian families.
This, of course, did not solve all linguistic problems.I am not perfectly fluent in the Syrian dialect and strong dialectal differences exist between regions.However, I had the feeling that my attempts to speak Arabic with respondents created an atmosphere in which participants felt valued and respected culturally.Participants expressed sometimes surprise, sometimes gratitude that I had invested time in understanding their language and culture.Second, my spoken Arabic-sometimes dotted with mistakes and unusual formulations-also broke the hierarchy inherent to many interview situations.Working closely together with native speakers in such situations proved to be ideal, as I could check back with them if I did not grasp a particular expression or aspect of an interview.Many researchers are a long way from understanding their informants' innermost thoughts and feelings, even if they have language skills.However, they rarely address this problem openly.Borchgrevink (2003: 96) suggests as an alternative to acquire language skills while simultaneously working with interpreters, which not only helped him to yield a greater number of results, but also gave him qualitatively more salient data.
Working with assistants who had gone through the experience of displacement themselves also allowed me to discuss interview questions and answers in depth.On the one hand, it is of course crucial that concepts and questions are transferable across cultures (Bloch 1999).On the other, it was important to understand which questions were too sensitive-or, conversely, not-to ask in a survey questionnaire.Aspects that might be considered strictly confidential for a foreigner might be 'everybody knows about it, not a big deal' for a participant (Kabranian-Melkonian 2015: 718).This was, for example, the case with smuggling, which I considered initially too sensitive to address upfront.However, the practice formed such a typical part of refugees' life-with them often having paid smugglers to cross internal borders in Syria and/or to enter Lebanon or Turkey-that this was much less of an issue for them.However, ethnicity, religion and political affiliations were sometimes red lines and it was crucial to get assistants' insights into how such questions could and should be addressed.Assistants were, for example, given flexibility in terms of how to introduce the question on religious belonging.Some assistants explained to participants that it was an unusual question to ask a Syrian and emphasized that they had every right to refuse to answer.Some respondents expressed their disapproval of the question by declining to answer.
The ethical decision to actively include refugees as partners in the research project, as suggested by the Code of Ethics of the IASFM, meant that I had to make a series of difficult choices as a consequence: I had to ensure that not only would respondents not be harmed through their participation in the project, but nor would research assistants either.This included in particular considering risks, national regulations (residence, work and travel permits for assistants and research permits) and representativeness, and weighing them up against each other.My objective was to hire and collaborate with young Syrians who had experienced displacement themselves and who knew research sites well.Prior to the fieldwork, I started to contact local researchers in both countries and used my existing personal networks to have the job description shared among their contacts.I aimed at finding research team members who were fluent in Syrian Arabic and also spoke English, studied at university or worked with NGOs supporting refugees in the research locations.While the recruitment strategy for my prospective assistants was accepted by the Ethics Board Review of my university, it had left me unprepared for a series of issues.First, I did not address assistants' political opinions on the conflict when I interviewed them for the position.I considered it inappropriate and intrusive to do so, even if it would make data collection more complicated.During the fieldwork, I started to slowly grasp some of my assistants' opinionsbut we never engaged in a full-blown political debate about the conflict.Some assistants in Lebanon reported not feeling safe in certain areas, especially those controlled by Hezbollah, and so I decided to exclude those neighbourhoods.In a neighbourhood in Beirut, one assistant and I had the feeling that we were being followed by the local authorities as we walked through the streets.Consequently, we first sat down in a cafe´and then decided to stop collecting data for the day.I also avoided assistants collecting data in the neighbourhood in which they lived to ensure mutual confidentiality.
Second, requiring a valid work permit as a condition for employment would have excluded large numbers of Syrian refugees in both countries.My home university solely required a copy of an ID and a signed agreement from my assistants, specifying the assignment and the fee-leaving a rather large space for me to manoeuvre in.Legally speaking, I was operating in a grey area as I was not entirely sure whether assistants were officially allowed to work for a foreign employer in the country.While we all visibly carried name tags stating our names and that of my university throughout the data-collection period, we agreed to conceal the fact that assistants were paid in case of local authorities getting involved-reporting instead that they were doing an internship.Legal issues also had other unintended consequences.One of my assistants was trying to leave the country through the family-reunification scheme and I considered his appointments at the consulate and with local authorities more important than data collection.As a consequence, we tried to swap around shifts to accommodate his schedule.
In both countries, data collection necessitated transportation between the two cities from which we collected data: in Lebanon, this was a bus ride from Tripoli to Beirut or vice versa; in Turkey, as we collected data in Istanbul and Izmir, we needed to relocate once between the two cities.In Lebanon, assistants reported dreading the checkpoints dotted along the road; at the same time, this was described as being part and parcel of the daily difficulties that they face.In Turkey, some of the assistants had acquired citizenship-for them, travelling was not an issue.However, for one assistant, it turned out to be impossible to obtain a travel permit in time and we considered it too risky to travel without it.We hence had to find a replacement for him in a rather ad hoc way.Note: Red: advise against all travel; orange: advise against all but essential travel; yellow: pay attention to safety risks; green: no particular safety risks.
Research with Refugees in Fragile Political Contexts 2319 Third, there were significant differences between Lebanon and Turkey with regard to research permits to safeguard participants and researchers.In Lebanon, a permit was only obligatory to enter Palestinian camps, which we excluded as research sites.In Turkey, a research permit from the Ministry of the Interior was at least temporarily mandatory for conducting research on refugees.In the first phase of the project, I contacted local researchers who worked on refugees in Turkey and received mixed assessments: one group of colleagues thought that it was too risky to conduct fieldwork without a permit as a foreigner given the current political situation, especially when working with research assistants on larger data collection.One human rights lawyer told me that, in smaller cities, it might be risky for employees or members of refugee NGOs to be seen talking to a foreign researcher.Another human rights lawyer told me that the permit was officially not mandatory anymore, but that regional authorities still might arbitrarily ask for it.
The second conclusion was that my chances of getting a permit were quite low, but that they might be higher if I applied with a Turkish research partner.Another group thought that I should go ahead without a permit, operating instead in close cooperation with a local university or NGOs working in larger-and hence more anonymous-cities.However, they could not assess the danger for a foreign researcher exactly and doubted that it would be possible to remain invisible if I collected survey data.Moreover, I was later told by one assistant that several Syrians who had participated in an unauthorized research project had been threatened with deportation by the authorities.Before the start of data collection, my decision was hence to apply for a research permit in Turkey and change my fieldwork site from there to Jordan in case I was not awarded it.Following this decision, I applied for a research permit together with a Turkish scholar who had gone through this procedure before.This involved submitting my entire survey questionnaire for scrutiny.I also obtained affiliation to a well-known Turkish university beforehand.
The permit was, surprisingly, granted one month later.However, if it had not been-and there was a high probability that this could happen-it would have endangered the entire data-collection endeavour.This generated a lot of stress and unpredictability leading up to the fieldwork.Rather unexpectedly, the granted permit did not interfere with the content of the research except for forbidding one survey question about respondents' religious affiliation.While this was not particularly surprising given that, as in Syria, questions about ethnic and religious belonging have long been avoided-there has not been a counting of Syrians by religion since the 1960 census inside Syria (Van Dam 2017: 7-8)-I had expected interference with other questions inquiring about respondents' mother tongue (as a proxy for ethnic affiliation), freedom of speech, freedom of religion and Note: Red: advise against all travel; orange: advise against all but essential travel; yellow: pay attention to safety risks; green: no particular safety risks.
Research with Refugees in Fragile Political Contexts 2321 corruption in Turkey.These I had considered to be more sensitive issues.Other survey questions about living conditions were less sensitive, as they focused on access to health, school, work, perceived safety and gender equality.Political opinions were not addressed directly in the survey, but through two open questions (the reason for mobility aspirations and wishes for one's future) thereinand, in general, more closely in the qualitative interviews.I had initially chosen this strategy because I thought that it would increase participation, allowing respondents to address political issues in a sensitive manner.However, I am fully aware of the fact that, for other research topics, obtaining a research permit might not have been possible at all.

Studying Urban Spaces: Walking the City
Weighing up the do-no-harm principle and representativeness against each other was a crucial element when choosing research localities.Already, when I applied for the grant that funded the research, I had to explain how I was planning to safeguard my own security and that of those involved in the project given that both countries experience regular political tensions and violence-especially in border areas.Concerned that I would not be granted permission, I promised in my grant application to conduct research only in areas marked yellow in the foreigntravel advice (pay attention to security risks), orange if I had institutional support (advised against all but essential travel) and exclude areas marked red (advise against all travel).In Lebanon, this excluded border areas with Syria such as the Bekaa Valley, the southern suburbs of Beirut, the south of Lebanon and Akkar in north Lebanon.These are the locations at which Syrians have typically sought refuge in Lebanon-being mostly economically disadvantaged regions and ones that they share with the most vulnerable Lebanese (UNHCR 2018).In Turkey, the travel advice was less of an issue as only the immediate border regions along the Syrian-Turkish and the Iraqi-Turkish border were signalled as red, while some provinces in the south and southeast were marked as orange (Hatay, Tunceli, Bingo¨l, Mus, Bitlis, Batman, Siirt, Sirnak, Hakkari, Van).This would have technically allowed me to collect data in parts of some southern provinces with high numbers of Syrian refugees such as Gaziantep and Sanlıurfa (UNHCR 2017).However, as I was aiming for comparable data across the two countries, I excluded border areas in Turkey as well.In retrospect, I would be more hesitant to set such far-reaching exclusions in advance again, as the foreign-travel advice is not necessarily the best tool by which to judge which locations are safe to visit.Many local scholars still conduct research in these areas.Similarly, two of my assistants worked in the Bekaa region and regularly travelled there; I have myself visited there several times in recent years, without conducting research however.
My second consideration was that access to refugee camps might prove difficult in Turkey, where camps are placed under the supervision of the authorities.Concerned that we would not be able to collect comparable data in the two countries, I decided to exclude camps and informal settings.This decision was, however, not only driven by my fear of not gaining access.Focusing on urban refugees living in private accommodation in urban spaces also reflects the reality of life for the large majority of Syrians in both countries.In Lebanon, around 85 per cent of registered Syrian refugees are estimated to live in private accommodation (Regional Refugee & Resilience Plan 2016-2017 in Response to the Syria Crisis).In Turkey, it is estimated that only 7.1 per cent of the registered Syrian population live in camps (UNHCR 2018).Syrian displacement in both countries is largely an urban phenomenon: most Syrians in Lebanon and Turkey live in urban areas, specifically in rented apartments in residential buildings (Erdo gan 2017;Kabbanji and Kabbanji 2018).This is in contrast to the fact that there is generally a lack of research on urban self-settlement in refugees studies (Bakewell 2008).Moreover, my aim was also not to conduct research in over-studied sites (Clark 2008).Refugees are often both a hard-to-reach group in some instances (Faugier and Sargeant 1997;Jacobsen and Landau, 2003;Bloch 2007;Vigneswaran 2009) and over-researched in others (Sukarieh and Tannock 2012;Karooma 2019;Luetz 2019;Omata 2019).Overall, camps tend to be more researched than self-settlement (Jacobsen and Landau 2003;Bakewell 2008;Sanyal 2014); some refugee camps in Lebanon face the problem of overresearch (Sukarieh and Tannock 2012).A number of ethical issues emerge from working with over-studied groups: they are increasingly distrustful towards researchers and often feel extremely frustrated that research does not bring concrete betterment to their lives (Omata 2019).
As a consequence of financial constraints, I decided to choose two cities in each country: (1) the biggest urban metropolis hosting Syrians and (2) a second one considered to be a 'transit city' towards Europe.Following this logic, I had originally planned to conduct research in Beirut and Tripoli in Lebanon and Mersin and Istanbul in Turkey, respectively.However, after conversations with Turkish researchers, I decided to substitute Mersin with Izmir-the latter being bigger and more anonymous than the former, feeling hence safer as a research location.Beirut and Mount Lebanon have a population of almost 2 million, with 20787 registered Syrians in the former and 246356 in the latter (which covers some of Beirut's suburbs) (UNHCR, 2017).Tripoli is the second-largest city in the country, with a population of half a million and 148084 registered Syrian refugeesincluding in its five surrounding districts (UNHCR, 2017).Istanbul is the Turkish province with the largest number of refugees and hosted more than 401928 registered Syrian ones in 2016 (DGMM 2016 ).Izmir is Turkey's third-largest city, with a population of 2.5 million and home to 93324 Syrian refugees in 2016 (DGMM 2016).Feelings of safety were strongly localized.In Lebanon, Sunni interviewees reported feeling safer in regions overwhelmingly Sunni-in particular Tripoli, whose people have had long-standing political and social ties with central Syria-than in those where Christian or Shiite majorities prevail.In Turkey, some participants reported having moved away from the Turkish-Syrian border out of considerations of safety.
The next difficult decision that I faced was how to recruit participants, both from an ethical viewpoint and with representativeness in mind.Apart from the well-known problems involved with snowball sampling (Faugier and Sargeant 1997;Jacobsen and Landau 2003;Bloch 2007;Vigneswaran and Quirk 2012;Reichel and Morales 2017), having volunteered myself with local NGOs, I quickly understood that they are often over-researched localities, with their staff already being overburdened and working on limited resources.Colleagues within the NGO for which I volunteered also rejected the idea of using contacts made through volunteering for research purposes.Volunteering also made me realize that it was often economically more vulnerable refugees who sought help from NGOs.I hence decided not to go via NGOs, or use data provided by temporary accommodation centres-as other scholars have done (Bircan and Sunata 2015;Kohlenberger et al. 2017).
As a consequence, I contemplated alternative ways to recruit participants.Based on the existing literature (Jacobsen and Landau 2003;McKenzie and Mistiaen 2009;Reichel and Morales 2017), I chose to combine multistage cluster sampling, random walks and focused enumeration of the 'nearest-neighbour' technique.Focused enumeration is a variant of random household sampling, by which every household selected through conventional sampling methods is asked whether there are persons from the target population living in an adjacent one (cf.Mu¨ller-Funk et al., 2019 for details on the sampling strategy).This strategy meant walking the city as a team, getting to know neighbourhoods, locating clusters of Syrians within different ones, but also to knocking at people's doors, entering their houses and sitting down together.I chose this recruitment strategy after the training session with my assistants in Lebanon.Some of them had done outreach work with NGOs before and were convinced that it was possible to gain face-to-face access in such a way.In Turkey, the situation was different: none of the team had done outreach work before and very much doubted that Syrians would be trustful enough to open the door and choose to participate in the research.It was my previous positive experience in Lebanon that convinced them to try.However, it proved more difficult in Turkey to locate respondents, as houses were often locked and had intercoms, thereby not allowing initial direct face-to-face contact.
While there was no exact sampling frame available in both countries, I could use UNHCR data in Lebanon (cadastral level) and data (district-level) provided by the Directorate General for Migration Management (DGMM) in Turkey.Leading up to data collection, I also walked the city alone, trying to understand how and where Syrians lived in the four cities, based on available statistical data and informal conversations through volunteering and the people I met.As displacement has affected Syrians across all income and educational levels, I tried to reflect this by including districts/cadastres known to be hosting high numbers of lower-class Syrians as well as those home to middle-class Syrians for the multistage clustering.During these exploratory walks, I took photos, listened to the Syrian Arabic spoken in the streets and, in Turkey, was attentive to signs written in Arabic to locate Syrian shops and businesses-marking these locations on Google Maps as potential starting points for our random walks.Walking as an investigative method of social research 'encourages us to think with all our senses, to notice more, and to ask different questions of the world', as Bates and Rhys-Taylor (2016: 5) argue.Walking the city and entering participants' houses allowed me to gain a much more comprehensive and nuanced impression of respondents' life realities than survey data and in-depth interviews alone would have.
Within these areas, we started our random walks after we had located clusters through informal conversations.Assistants walked in pairs, but entered apartments alone-which presented a certain risk for assistants.One of my female assistants in Turkey, for example, told me later that she sometimes did not feel comfortable entering private houses by herself.However, I was constrained by financial considerations and could not let assistants interview in pairs, even if I knew that, in principle, something could potentially happen to one of my assistants.I could only accompany one assistant at a time.While this possibility likely exists in other-safer-contexts as well, host-community hostilities towards Syrians were on the rise when we conducted our fieldwork.I tried to counterbalance these risks by staying close by and having official documents ready in case.We were also in constant contact through a WhatsApp group through which we shared our locations and checked up on each other.
The assistants' knowledge of local customs proved crucial for locating clusters.They sometimes also knew someone in the district/cadastre who helped us to locate where Syrians were living.In Turkey, the process of finding clusters sometimes took a while-especially when districts were large.In Istanbul, for example, the proportion of Syrians in chosen districts was not higher than 5 per cent; in some cadastres in Beirut or Tripoli, contrariwise, almost half of the houses in a given neighbourhood were inhabited by Syrians.Ku¨cu¨kcȩkmece in Istanbul, for example, which hosts the highest number of Syrians (32011), had an overall population of 770393 in 2018 (DGMM 2016)-the proportion of Syrians there is thus a mere 4.1 per cent.We therefore had to resort to more focused enumeration, first asking locals in the area to locate Syrian shops and then asking shop owners to locate possible clusters.In Istanbul's large districts, we also occasionally resorted to approaching possible respondents in public places, such as parks.We hence adapted our approach in a flexible way to conditions on the ground.
While survey participants were solely recruited through random walks, my logic for recruiting those for the in-depth interviews was different.Interviewees were chosen with diversity in mind, specifically on the basis of different attitudes towards mobility, of gender, ethnic and religious affiliation, age and educational background.Some informants were recruited by asking survey participants whether they would agree to a follow-up in-depth interview should they want to share more than they could in the survey and correspond to the required profile.However, a large number of respondents were also selected by research assistants themselves through their own personal networks.This strategy gave assistants the possibility of including perspectives that they found important or to be missing from the sample.Assistants felt, for example, that, with the survey, we were not able to reach out to Syrians belonging to the upper class nor to those with high educational levels and tried to balance this by way of the qualitative in-depth interviews.Finally, I also sensed that this strategy increased the feeling of trust-as participants already knew at least one of us beforehand, which allowed us to talk about more sensitive political issues.

Research with Refugees in Fragile Political Contexts 2325
Open Challenges: Informed Consent, Sensitivities, Ethical Publications and Advocacy The last section of this article highlights open challenges with regard to informed consent, sensitive and potentially traumatizing questions, incentives offered and ethical publications.
Gaining trust and informed consent: trying not to harm while re-traumatizing refugees?All interviews were conducted on a face-to-face basis.I believe this helped immensely to increase participants' trust.My assistants also mentioned the positive effect of my presence-a foreign researcher from a rather neutral country (Austria) in the context of the Syrian conflict and one speaking the interviewees' native language-on gaining access.Survey interviews were conducted by assistants; nevertheless, I always accompanied one of them during data-collection days.In-depth interviews were conducted in a group of three: the interviewee, myself and one assistant, who helped in probing and asking questions, and with translation if necessary.The fact that in-depth interviews included three people created a more informal atmosphere than that experienced in a classic one-to-one interview situation.I had the impression that the at least partly shared destiny of respondents and assistants helped in creating a space in which the former felt safe to talk.The latter were sometimes perceived as university students by participants, with the project being seen as part of their education.Assistants acted as gatekeepers, providing legitimacy for the survey and for the independence of the research team.Using gender-matched interviewing for the survey was crucial to increasing the participation of women, who often expressed feeling more comfortable talking to a female interviewer.However, this felt closeness was also relative.Some of the assistants pointed out that the research allowed them to get to know the diversity of their own community, emphasizing important differences in social class, education and religious practice between interviewees and themselves.
I consciously decided to exclude questions about persecution and violence in Syria and regarding political opinions on the conflict from the survey.I considered it too sensitive and potentially re-traumatizing to address these topics in a context in which we were meeting participants for the very first time.In qualitative interviews, these questions came up if respondents chose to address them.Occasionally, assistants faced distrust from participants as they tried to understand whether the former were Lebanese or Turkish or where they stood politically in regard to the conflict-information often elicited through indirect questions about their original place of residence in Syria.One assistant also mentioned that he experienced it as challenging to interview someone with whom he did not agree or whose opinion he perceived as irrational, especially in regard to the conflict in Syria.As he did not have much information about the interviewee beforehand, it was difficult to prepare for such a situation spontaneously-as proregime remarks popped up only suddenly during the conversation.
As other scholars have also done (Du¨vell et al. 2010), we opted for oral consent-as obtaining written consent was impracticable.First, written consent with a personal signature and name would have raised suspicion among respondents if their names were truly to be kept anonymous.Second, we did not want to put participants in the potentially uncomfortable situation of admitting illiteracy.The fact that we met participants in person also allowed the team members to explain the aim of the project orally before asking for consent.The introductory part of the survey sometimes led to general explanations about what social science research is more globally.It was also crucial that we communicated to respondents that the research team was not part of an international organization, NGO or official national Lebanese/Turkish body, and that participation would not increase their chances of resettlement in Europe.
These combined measures might explain our high overall response rate for the survey (82.9 per cent in Turkey; 83.6 per cent in Lebanon) and the richness of our qualitative data.However, it cannot be ruled out that some participants still hoped that we might be able to provide assistance in the future.Some interviews also resulted in emotional distress for participants, especially when talking about experiences of torture, discrimination, humiliation, loss of family members and of social status in in-depth interviews.Nevertheless, some participants mentioned that having someone listening attentively to their stories was seen as a positiveand rare-experience, reflecting Powles's (2004: 18) observation that some experience the process of recording their personal narrative as empowering as they unburden themselves-sometimes for the first time-from very troubling experiences.
Incentives, ethical publications and advocacy.I have not found a solution to the potential risk of re-traumatizing participants through interviewing them, except for asking for professional psychological advice so as to learn how to best react in such situations and reflecting on how to support refugees in other ways.These ideas emerged 'in practice'-mostly when I felt helpless in the face of immense suffering.We had decided not to offer financial incentives to participants-some assistants considered it inappropriate, as doing so might comprise the freedom to participate in the study given that many respondents lived in difficult economic circumstances.Nevertheless, we tried to provide non-financial returns through access to information.We prepared a small leaflet with information about relevant NGOs that we handed to participants if they proved to be in need of support, especially in regard to medical services.Some of the research assistants also stayed in contact with survey respondents for these reasons after the fieldwork had come to an end.
Assistants also handed out my business cards to participants at the end of the interview, so that they could eventually access the results of the study.Ultimately, however, I am aware that this is a possibility for only a select few, as my academic profile is in English.A much more powerful strategy would be to translate key results into Arabic and Turkish, and publish in alternative outlets, such as via working articles-which are openly accessible and have different standards to academic journals.I also believe that, from an ethical standpoint, the academic community should think more seriously about strategies to include refugees in the Research with Refugees in Fragile Political Contexts 2327 writing process, to avoid misunderstanding or misrepresenting research participants and their lifeworlds.
As Sukarieh and Tannock (2019) note, subcontracted local research assistants often speak critically of their sense of alienation, exploitation and disillusionment regarding the projects that they work on.These could be overcome through actively incorporating interviewers in the writing process, through regular brainstorming or feedback loops or by developing policy recommendations together-strategies that I am currently in the process of trying out (cf. author et al., 2019).Such an approach could also help to avoid security breaches arising from unwanted confidentiality lapses.While I pay careful attention not to cite or narrate sections of interviews that could make participants identifiable, someone knowing the context from an insider perspective could provide a second opinion and counterbalance such a risk.It goes without saying that collaborative writing also has its limits-some of my assistants do not have a background in the social sciences, nor have they been trained in data analysis or drafting policy recommendations.As such, from an ethical standpoint, we should understand our roles as mutually-and constantly-learning from each other.

Conclusion
This article has aimed to add to the literature on 'ethics in practice' in forced displacement, with a focus on comparative cross-country mixed-methods designs in countries neighbouring a war.It has given an overview of the difficult choices that I made when following the IASFM's ethical guidelines as part of conducting a mixed-methods research project with Syrian refugees in Lebanon and Turkey in 2018.In doing so, the article has revealed the discrepancy between the ways in which qualitative and quantitative research approaches research ethics and representativeness.It has highlighted in particular how risks and violence influence the choice of fieldwork sites in politically fragile countries and in contexts in which refugees are, legally speaking, particularly vulnerable.
It is potentially easier to follow ethical guidelines when conducting small-N qualitative studies or long-term ethnographies, where overall representativeness is not a major objective of data collection.The article has carved out how ethically conceived mixed-methods projects can and indeed should weigh up ethical principles and representativeness against each other, such as by actively including refugees as partners in research projects, investing time in learning local languages and cultures while also considering security and national laws (employment, residence and travel regulations for refugees; research permits).It has advocated for a refugee-centred approach while avoiding over-researched localities, and highlighted the difficult and ad hoc choices that I faced so as to 'do no harm' to participants and team members.On the one hand, including refugees as research partners, spending long periods of time in data-collection locations and speaking the local dialect proved to be not only crucial for gaining trust among participants, but also helped immensely vis-a`-vis more precise analysis.On the other, the conscious choices taken also included the employment of refugees by bypassing national laws while also seeking official permission-done even despite the risk of jeopardizing the fieldwork and later the content of the research.Producing sound data that has been collected in an ethical way in fragile contexts takes time-a rare commodity indeed in academia today.This article is thus also a call for 'slow science' (Berg and Seeber 2016).
I do not hesitate at this point to note that the approach taken certainly does not provide a solution to all of the ethical challenges faced in the study of forced displacement.First, future projects could reflect on how professional counselling might benefit fieldwork on forced displacement-not only in regard to how research questions can and should be addressed, but also concerning how researchers can be best supported psychologically during data collection.One of my research assistants also suggested considering non-financial returns, specifically in the form of psychological support being provided to respondents.Second, more emphasis should be put on longitudinal research in future-and on the ethical and methodological challenges that this entails regarding data protection.Third, as part of efforts to strengthen collaborative research with local team members, a serious rethink about publication strategies is also required.The translation of key results into the languages of the communities that we study is crucial, as are alternative dissemination strategies too.However, such efforts often stand in contradiction to the types of publications that are valued by the academic community at present.