Abstract
Objective of Protocol: The primary objective of this protocol is to record the process of conceptualizing a semi-structured interview protocol, training enumerators on the protocol, collecting data, translating findings into English, and analyzing data in English and Bengali, in a study of family socialization among a stateless and conflict-affected population, Rohingya refugees, in Cox’s Bazar, Bangladesh. Research Questions of Study: (1) What are the socialization goals that mothers have for their children, and how do they perceive these goals in relation to their future or current roles as caregivers? (2) What are parents’ experiences with their children’s participation in Humanitarian Play Labs (education/child care provisions in the camps), and what are their perceptions of how their child’s participation in the programs has influenced their child/family? (3) What were the experiences of parents with their children as they migrated from Myanmar to Bangladesh? Design of Study: The design of the study involved a qualitative grounded theory approach based on an analysis of the participants’ responses to a semi-structured interview protocol. Study population: Participants included a purposive sample of 28 mother/father dyads in the Rohingya camps in Cox’s Bazar.
Keywords
Background
The Rohingya are one of the most persecuted minorities in the world (Kiragu et al., 2011). They have been escaping from Myanmar in response to persecution for multiple decades (Wali et al., 2018), with the largest influx entering Bangladesh in 2017 in response to a military crackdown in Myanmar (Messner et al., 2019; Ullah, 2011). Now, more than a million Rohingya people reside in camps in Cox’s Bazar in Bangladesh (Messner et al., 2019); this is one of the largest groups of stateless people in the world (Kiragu et al., 2011).
The Rohingya remain in a state of limbo, as the Bangladesh government prohibits social and economic integration of the Rohingya into Bangladeshi society. This prevents adults from securing legal employment and denies children access to schools in Bangladesh (Shohel, 2020). The Bangladeshi government provides the bulk of relief for the Rohingya, and a variety of NGOs and aid organizations are also working in these camps now, providing healthcare and educational services.
Global TIES for Children at New York University’s role in the research described in this protocol is as the research partner for a collaboration between LEGO Foundation, Sesame Workshop, and the NGO BRAC, to evaluate BRAC/Sesame Workshop implemented programming in the Rohingya camps (Mariam et al., 2021). Specifically, we are examining aspects of family and community life that can sustain remote, in-person, and hybrid early childhood services for children and families, focusing on the BRAC/Sesame Workshop program “Humanitarian Play Labs (HPLs)” (Mariam et al., 2021). The HPLs are designed to support playful learning for children aged 0–6. Humanitarian Play Labs for 2–6 year-olds are an education/child care provision, led by a “playleader,” (usually a local young woman) in which children play, sing, draw, and learn together (Foulds et al., 2021). These programs are conducted in Rohingya and incorporate traditional Rohingya games and rhymes to promote learning and build resilience (Foulds et al., 2021). The model aims to strengthen community engagement and involve community actors, in this case mostly women, and has the potential for replication in other under-resourced or humanitarian settings (Mariam et al., 2021). An evolution of this model is the introduction of “home-based” HPLs, which take place inside of people’s homes rather than in formally established centers. Global TIES for Children supports the development of tools for monitoring quality of this programming and is involved in evaluating various sub-models of these HPLs, though much of this work is temporarily halted/being redesigned given constraints of the COVID-19 pandemic.
Preparing measures for use in this context led us to realize there is a notable dearth in the literature in providing the kind of methodological details on what we, as researchers, required to collect data in the Rohingya camps. To our knowledge, in the 3 years since the major influx of Rohingya into Cox’s Bazar, only three qualitative studies involving this population have been published in the academic literature (Riley et al., 2017). This is perhaps unsurprising. Qualitative data is time-consuming and difficult to collect, analyze, and publish. The period immediately after the Rohingya were first displaced into Bangladesh was devoted to the provision of basic needs for survival and the setting up of housing, healthcare, and child protection; not to research (Tay et al., 2019). However, the situation has now stabilized and the crisis is recognized as protracted, leading to a confluence of organizations providing support and conducting research activity (Tay et al., 2019). As more organizations turn towards research with this population for academic and service-related purposes, learnings on the process of conducting rigorous qualitative research in this context will prove useful to disseminate.
The primary objective of this protocol is to record the process of conceptualizing a semi-structured interview protocol, training enumerators on the protocol, collecting data, and translating findings into English for analysis in a study of family socialization among a stateless and conflict-affected population, Rohingya refugees, in Cox’s Bazar, Bangladesh. We hope that explicating the procedure, as well as challenges faced and lessons learned in the field, proves useful to other researchers working in the field of early childhood development in humanitarian contexts.
Objectives of the Study
The main objective of the study was to answer the following research questions: 1. What are the socialization goals that mothers have for their children? How do they perceive these goals in relation to their future or current roles as caregivers? 2. What are parents’ experiences with their children’s participation in the home-based HPL program, and what are their perceptions of how their child’s participation in the program has influenced their child/family? 3. What were the experiences of parents with their children as they migrated from Myanmar to Bangladesh?
Explanation and Justification of Method
A qualitative grounded theory approach (Glaser & Strauss, 2017) was used, based on an analysis of the participants’ responses to a semi-structured interview protocol. This was deemed an appropriate approach given the limited research base existing for this population, and we aimed to avoid preconceived ideas of parenting, experience of early childhood programs, or participants’ perceptions of their lives in the camps.
Study Period
Data collection for this study started with a brief (six participants) field test of the protocol in March 2020, after which full data collection was supposed to resume. This process, however, was halted due to the lockdown because of the COVID-19 pandemic. Data collection for the full sample ultimately resumed in November 2020 and concluded in February 2021.
Data Collection Partner
Global TIES for Children partnered with Innovation for Poverty Action (IPA) Bangladesh as the primary data collection partner for all LEGO-funded research in the Rohingya camps. Innovation for Poverty Action hires enumerators from the host community, that is, Cox’s Bazar, Bangladesh, most of whom have some level of university education and are fluent in both standard Bengali and Chittagonian Bengali (discussed further below).
The Semi-Structured Interview Questionnaire
The interview protocols were prepared in English and went through multiple rounds of review by a number of personnel, including researchers from NYU, BRAC, Sesame Workshop, and IPA. This English protocol was then translated by IPA translators into Standard Bengali. Bengali is a key intermediary step between the English used by international researchers and aid workers and the Chittagonian Bengali/Rohingya spoken by members of the host community/Rohingya themselves. Importantly, Chittagonian Bengali/Rohingya do not have a commonly used written script (Khan et al., 2019); therefore, surveys and interview protocols, for our studies as well as most research taking place in the Rohingya camps, are translated from English into Bengali. Given that the Bengali protocol was a reference document, since the protocol was not actually delivered in Standard Bengali, we chose not to back-translate from Bengali to English; however, the protocol was reviewed by a Bengali-speaking consultant.
The interview questionnaire consisted of several sections, covering parents’ perceptions of the characteristics of their children, features of parenting and play with child, activities in and impact of BRAC HPLs on children, the role of culture, religion, and norms, and their experience of forced migration from Myanmar to Bangladesh and living in the camps.
Accessing Rohingya Camps
There are a number of logistical difficulties involved in accessing the Rohingya camps for research purposes (Ainul et al., 2018; Chuah et al., 2018). The main city of Cox’s Bazar, where most researchers tend to stay, is a two-hour drive from the closest camps. In most cases, permissions from camp and governmental authorities are required to enter the camps; some of these required permissions change on short notice based on local/national guidelines and the day-to-day camp context. In our case, when we were interested in speaking to female caregivers, we required female enumerators, which involves added logistics of ensuring security. Outsiders are not allowed in the camps past 3:30 p.m. (Foulds et al., 2021) leading to a cap on the amount of work that can be done in any given day. The confluence of factors led primarily to the process of data collection taking often much longer and involving more logistical effort than was expected by the NYU team.
Data Collection Procedure
Interviews took place approximately simultaneously but in separate physical locations for the mother and father of the children. The mothers’ interviews always took place inside the home and were conducted by female enumerators, while the fathers’ interviews (with male enumerators) took place in the home if a separate room was available or in a nearby outdoor area. On average, the interviews lasted about an hour and a half. The interviews were conducted in Rohingya/Chittagonian Bengali (Rohingya and Chittagonian Bengali are considered mutually intelligible (Khan et al., 2019), with 70–90% overlap). Enumerators speak Chittagonian, while the Rohingya population speak Rohingya. All interviews were audio-recorded, then transcribed by the interviewers into written Standard Bengali. These Bengali interviews were later translated into English by IPA’s translators (see below for more details on this process). The NYU analytic team worked primarily with hand-written Bangla transcripts and typed English transcripts.
Directly after the interview, enumerators wrote hand-written field notes about their experiences of the interview. Images of the field notes were sent back to the NYU research team on each day of the data collection. This allowed the NYU research team to remain updated on the events of the field on a day-to-day basis. In some cases, events in the field necessitated direct discussion with the enumerators and their supervisors at IPA to respond with small changes in procedure or in the questionnaire itself.
Sampling/Recruitment
Participants
The participants were chosen using a purposive sampling technique. A total of 28 couples were originally targeted for the sample. These participants were chosen to fulfill several different criteria. Sixteen couples were chosen whose children were attending home-based HPLs, split across all the camps that the home-based HPLs were present in. Six couples were chosen from “target camps” (i.e., camps where BRAC was, at the time of designing the study, intending to scale up programming and where Global TIES would have conducted its evaluation; this is now on hold due to the pandemic). An additional six couples were chosen from camp locations without HPL programming/far from HPL programming.
Recruitment Strategy
For the caregivers whose children attended home-based HPLs, BRAC compiled a list of pre-screened households, who had children who attended HPLs and were willing and available to participate. From this list, IPA randomly selected households to participate while physically on site. BRAC playleaders or staff would often accompany IPA staff to introduce them to families but were not in charge of selecting or recruiting households.
For caregivers whose children did not attend HPLs, IPA attempted to recruit from a broad geographic radius as much as possible. In most cases, IPA started the search from a central location of the camp and spread out in multiple directions, selecting the first available/willing household.
It was often difficult finding households where both male and female caregivers of the HPL children were present
Data Handling/Analysis
Transcription and Translation
The team of enumerators who collected the data transcribed the audio-recorded interviews into written text. This transcription took place from the audio recordings of Chittagonian Bangla/Rohingya into written standard Bengali. The transcribers thus carried out an informal translation process during the transcription (note that this translation process is often overlooked in studies with the Rohingya, where it is often simply noted that “interviews were transcribed”). Accurate transcription is a central piece of qualitative work, but is often overlooked in the literature. Although formal rules and notation of transcription are well-established, they are usually for English (Weeks et al., 2007) and therefore not necessarily easily transferable or intuitive to apply to other languages (Beck et al., 2003; Wang et al., 2006). They also are not easily applicable in cases where translation occurs within the transcription process. The enumerators who completed transcriptions were unfamiliar with these rules and notations, which ultimately resulted in us revising each transcript in comparison to the audio for the first approximately 12 interviews. Over the course of this process, we developed a set of guidelines for transcription to assist enumerators with future transcriptions. Initially, the NYU research team worked with Bengali hand-written transcripts; over time this proved to be infeasible given that hand-written transcripts were not easily usable inside of any qualitative data analysis software. Hand-written transcripts were therefore then typed with enumerators either typing them by hand or using a phone-based voice-to-text application.
The Bengali transcripts were then translated into English transcripts by translators working for IPA. This process also faced challenges similar to transcription, because the translators, though fluent in both Bengali and English, were unfamiliar with the often-colloquial language and phrases used in the conversations these interviews represented. Given that translators play a central role in qualitative data analyses, (Van Nes et al., 2010), we worked to develop the capacity of these translators and engage them longer-term. This involved biweekly meetings and constant three-way communication between the NYU research team, the translators, and the enumerators who conducted the interviews to discuss particularities of the language used. The ongoing objective here was to create a researcher-translator partnership, and the process resulted in the creation of documents to guide the translation process (Iqbal & Sunny, 2020) as well as frameworks of common difficulties the translators faced.
Data Analysis
We were initially interested in exploring a process whereby the transcripts would be analyzed in both existing languages (English and Bangla) and findings compared, acknowledging that this poses methodological challenges given that the researchers’ positionality and orientations differ in addition to the language they speak. Between the field test (March 2020) and the resumption of in-person data collection (November 2020), the NYU analytic team met every one or two weeks over the course of several months and carried out the following activities: a mapping analysis to pinpoint problematic questions (detailed in the rigor section below), a preliminary content analysis (in two languages) to extract major findings, and returning to enumerators to further understand their experience in order to inform the revision of the interview protocol for subsequent pilots.
The team approached the transcripts using a combination of grounded theory (Charmaz, 2006; Glaser & Strauss, 2017) and thematic analysis (Braun & Clarke, 2013). We first did multiple close readings of the text, in both English and Bengali (in the case of the Bengali comparing the text to the audio files of the interview to ensure comprehension). We then each developed an analytical matrix; specifically, spreadsheets containing each interview’s major sections and notes on emerging themes. We then met as a full group to discuss these themes and shared findings back and forth until consensus was reached.
When analyzing data from the field test, analysts focused primarily on Bengali transcripts, with small additional sub-groups of analysts examining English transcripts. After the full set of data had been collected, analyses shifted primarily to Bengali transcripts, with English analysts reading occasional English transcripts to provide feedback in ongoing group discussions. These group discussions were used to arrive at consensus for major themes the data should be coded for across a number of projects. The Bengali transcripts were entered into the qualitative data analysis software Dedoose, and two Bengali analysts (the first and second authors) coded the data across these projects. These analysts met with enumerators, translators, and members of the NYU analytic team on a regular basis to discuss emerging findings and discuss whether these findings accurately represented the conversations taking place in the interviews.
Ethics
IRB
All protocols used for this study were approved by the Institutional Review Board (IRB) of Innovation of Poverty Action, and further reviewed by New York University’s IRB (which then provided a reliance agreement for use of IPA’s IRB).
Referrals
The interview protocol covered questions on maternal physical and mental health, parenting stress, and other topics that could potentially impact the respondents’ state of wellbeing or raise concerns that the respondent may require consultation from service providers offering psychosocial counseling, therapy, or shelter care. For this, we deferred to IPA’s referral policy, which entails providing referrals to respondents who show cause for concern. These referrals contain contact information and addresses for local psychosocial counseling centers and shelters for women, children, and other vulnerable populations. The enumerators give out this information on a piece of paper and verbally, directing them to appropriate nearby services and encouraging them to contact for help.
The process for referrals was decided in consultations with a number of organizations providing relevant services in the refugee camps since the influx in 2017. International organizations such as IOM, BRAC, Oxfam, MSF, and local NGOs such as Mukti and Pulse Bangladesh have been leading service providers, and their contact information was provided to the respondents who seem at risk. Innovation for Poverty Action Bangladesh itself does not have the jurisdiction to provide direct health intervention by admitting respondents to counseling centers, nor can respondents’ confidentiality be breached by informing health authorities of the respondents’ state of health. Innovation for Poverty Action consulted with BRAC, Oxfam, and Pulse Bangladesh while defining their protocol for referrals and these organizations specified that their referral protocols similarly state that respondents need to self-admit to obtain required health services. Innovation for Poverty Action was advised to provide the address and contact information of the local health/counselling provider, namely the organizations mentioned earlier, in print, but more importantly verbally, as many of the refugees are unable to read. Innovation for Poverty Action was also advised against moving a respondent from their household, accompanying them to health service centers, or pointing health providers to specific households, as any of these actions may incite risk for staff and participants and be damaging for the future programs of the project and the organization.
COVID-19 Safety Procedures
Once data collection for this protocol commenced in November 2020
Cultural/Linguistic Considerations
We faced a number of logistical and cultural challenges in the process of designing this protocol, accessing the camps, and collecting interview data. Two of the most prominent of these challenges, and, where relevant, how we attempted to mitigate them, are listed here.
Language barriers and communication difficulties
The enumerators, local to the Cox’s Bazar region of Bangladesh, speak Chittagonian Bengali, which is thought to have 70–90% overlap with Rohingya (Khan et al., 2019). Hiring Rohingya themselves would have been an excellent way to conduct this research—they would have familiarity with the language and culture of the participants, and have the ability to be involved in development of the protocol and research topics. This, however, comes with its own set of challenges: Rohingya are not officially permitted to be employed by the Bangladeshi government, and most of them, in particular the women, do not have the necessary education/exposure to be able to communicate directly with foreign research teams. In addition, the Rohingya are a culturally conservative people, with a high prevalence of women ascribing to the concept of “purdah”: not interacting with people outside of their immediate family.
Because of the imperfect nature of conducting research with enumerators external to the community, the NYU research team is exploring methods of participatory research and collaborations with local organizations who work directly with volunteers from the Rohingya community. However, these collaborations do not map onto the tight timelines existing in evaluation research.
Communication barriers were not only about language. Our protocol included questions about early childhood and parenting through conceptualization of these ideas that come primarily from a research literature stemming from WEIRD contexts (Henrich et al., 2010), one that may not be relevant for these parents. In many instances, the caregiver was reluctant to answer the question or did not want to explicate beyond a one- or two-word answer. Culturally, Rohingya women do not typically speak to outsiders and are not meant to leave the home (Islam et al., 2021; Rahman, 2021); they are unaccustomed to speaking to outsiders and answering questions (Rahman, 2021). Part of the process of beginning the protocol with a field test was to begin to work through these communication barriers, so a certain amount of difficulty communicating was expected. However, despite repeated urging from the enumerator to speak up if they did not understand questions, participants were often reluctant to communicate with enumerators, instead choosing to remain silent. These instances potentially point at the power dynamics between researcher and participant, and raise questions around the extent to which we may have been placing the respondents in uncomfortable situations by asking them to respond to these questions. Beyond issues of language barriers, participants have good reason to be unwilling to share full information with individuals from NGOs, researchers, aid workers, or members of the government of the host community (see Carlson et al., 2018; Khalaf-Elledge, 2020), in the interest of their own safety and privacy. In addition, although we did not observe anything concerning in our enumerators, almost all enumerators come from the host community in Bangladesh, which is the exact population that was most adversely affected by the arrival of the Rohingya in Cox’s Bazar (Uddin, 2020). There may therefore exist potential intergroup prejudices between these enumerators and the Rohingya participants which could impact interactions between enumerators and participants.
Spending time developing trust and rapport was crucial to conducting successful interviews. One enumerator reported that he conducted the entire interview (with some difficulty), then at the end, when chatting with the family, realized that he knew some of the family’s relatives. When he expressed this, they immediately became warmer and more open, urging him to stay for lunch. Of course, in most cases enumerators will not have familiarity with respondents’ families and will not always be able to use these connections to build trust. However, the point remains that this trust greatly eased communication once it had been established, indicating the need for spending time on building rapport.
Privacy during interviews
Field notes from the interviews indicate a steady stream of involvement from family members, children, and community members curious about the interviews. Especially in the case of women, community members (such as the playleader, or the respondent’s mother, or mother-in-law) often provided input to the respondent on what an appropriate answer might be. This was especially the case given that for the female caregivers, interviews always took place inside the home, and the homes were often small and had many other members of the household present. We worked with enumerators to find culturally appropriate and polite ways to request household members to refrain from answering questions for the participant, though in general we had to accept that our participants would be afforded little privacy during their interviews.
Rigor
Given the complexity of conducting research in this context, we took a number of steps to attempt to increase the rigor of our research. The steps that proved most important were field-testing the protocol and extended training of enumerators, explicated below.
Field-Testing
The protocol first underwent a period of piloting or “field-testing” (March 2020) in which the enumerators, after undergoing an initial training, collected one respondent’s worth of data each. Directly after this field test took place, in-person data collection halted because of the COVID-19 pandemic and did not resume again for 7 months. This period of time allowed us to closely examine the field test data and conduct in-depth focus groups with the enumerators regarding their experiences in the field. Based on this, we undertook an extensive and thorough process of redesigning the protocol to improve the flow of questions and specify the language where there was difficulty in asking (on the enumerators’ part) or understanding (on the participants’ part) the questions.
This process of redesigning consisted of designing a spreadsheet to assess the types of difficulty that we saw coming up for various interview questions and consisted of us mapping the following common problems we observed in the protocol: • Rephrasing: the caregiver does not understand the question and either expresses that they did not understand or remains silent, and is therefore asked the question again in different language. • Misunderstanding: The caregiver gives an answer that is clearly unrelated to the question. • Induced response: The interviewer answers the question themselves to try to elicit responses from the caregiver.
Two analysts coded the six interview transcripts from the field test for the above errors, and we revisited the protocol section by section, in some cases revising the questions themselves and in other cases having conversations with the enumerators about their experiences attempting to ask these questions and working together with them to come up with better ways to ask them.
After this, the final questionnaire was again reviewed by a Rohingya-speaking sociolinguist for ease of comprehension; on his recommendation we further added several alternative Rohingya words/phases for many of the terms used in the written Bengali protocol.
Training of Local Enumerators
Two members of the NYU research team (both Bangladeshi; one Research Scientist and one external research consultant, the first and second authors of this protocol), in conjunction with IPA staff, trained six IPA enumerators (in-person) on use of the protocol for the initial field test, then subsequently re-trained them (using a hybrid of in-person and online modalities) prior to data collection.
The training was a bi-directional capacity building exercise, because although the NYU staff brought knowledge on how to do qualitative interviews, the enumerators had extensive experience working in the camps, and were able to provide crucial contextual knowledge. The training consisted of the following elements:
Refining language
The most important part of the training was the process of clarifying the language to be used in delivering the interview questions. As mentioned, the written form of the protocol was in Bengali, not in Chittagonian/Rohingya that they were meant to be delivered. The enumerator team therefore went through each question during the training phase, speaking it out loud and clarifying language to be used. We wrote down key phrases and terminology to be used, but did not translate the entire document. Partners at IPA report that protocols written in Bengali script with Rohingya phonetics have not proven helpful, with Bengali-speaking enumerators finding these written documents difficult and non-intuitive to use.
Key features of conducting qualitative interviews
This included themes such as building rapport, empathetic listening, giving the respondent time to answer, not asking double-barreled or leading questions, and how to probe.
Brainstorming potential issues and how to solve them
We discussed potential logical issues that could come up during data collection. Due to crowded conditions and the social atmosphere of the camps, we anticipated lack of privacy during the interviews, and discussed potential strategies of dealing with the presence of the caregiver’s family members, such as her husband, mother-in-law, or children, during the interview.
Role playing for practice
Once we had developed a protocol that we were comfortable with, the enumerators role played it with each other for practice. This was an important step as it helped the enumerators become comfortable with delivering the questions while further clarifying aspects of the language that had to be refined.
Conclusion
The primary objective of this protocol is to record the process of conceptualizing an interview protocol, collecting data, and translating findings into English for analysis, and analyzing data in both Bengali and English, while attempting to ensure rigor in each step. Our primary recommendations to other researchers working in this context include dedicated attention, beginning in the design phase, to issues of language, access, and cultural specificities of the camps. We anticipate that the details provided here might be helpful for other researchers in understanding what to expect and challenges are faced while conducting qualitative research with this population. These on-the-ground experiences are crucial to research design. The study itself is one of the first in-depth explorations of the experiences of male and female caregivers and their perceptions of early childhood programming in the Rohingya camps. We hope that some of our process and recommendations can be helpful for others.
Footnotes
Acknowledgments
This work was produced using resources and staff expertise from NYU Global TIES for Children in collaboration with Sesame Workshop, BRAC, and Innovations for Poverty Action, Bangladesh, as part of the Ahlan Simsim and Play to Learn programs generously funded by the John D. and Catherine T. MacArthur Foundation and the LEGO Foundation. We thank NYU Abu Dhabi Research Institute for core support to the Global TIES for Children Center, which provides intramural support for research activities that directly contributed to this paper. We would also like to thank independent consultant Andrés Motta for his contributions to the methodology in this paper.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by LEGO Foundation [grant number A19-0170].
