Selection of Sites: Where to Go and Who to Talk With
The selection of sites for research will vary with the context and purpose of the study. This section walks through a number of considerations in selecting a research site and shares how some teams have decided on where to go and when for analysis.
Considerations in Selecting Research Sites
Looking across CARE studies, most took place in areas where CARE worked, though in some cases, CARE researched villages where it did not work directly though has been active in the area (i.e. Uganda). For situational analysis studies, it is crucial that the study site is a place where CARE has a sustained commitment to working within the area in order to support social analysis and action, as well as ensure follow-up for the countless hours of time invested by the research team and community members.
To understand power dynamics and differences, a number of sites chose to research a more powerful and wealthy village and a poorer village to see how the situations of the two areas differed in terms of investment, voice and access to resources. This type of study also helped to highlight the inter-relatedness of people across villages.
Other considerations in research that teams consider include:
- Season: It is critical to consider accessibility of the study area, and risks of rain that may inhibit the ability of studies to take place. Further, for agricultural communities, teams must also consider whether community members have time to take part in exercises due to agricultural responsibilities.
- Security/Political Climate: The political climate is also one important consideration in scheduling research activities. Preceding elections, the political climate may be sensitive to questions concerning local government, local conflicts or other issues. Further, it is important to ensure that local leaders will support and make space for the study to take place. This may require the acquisition of a research permit in some contexts (Tanzania).
- Language and Culture: While this may not be a guiding factor in who to study, teams should account for how appropriate the research team members are for the study. This would include ensuring that the research team can communicate with community members in their local language. Further, teams may also consider how age, gender, ethnicity or religion may affect access to and dynamics in interacting with community members.
In general, teams found it easier to conduct research in areas where CARE was already familiar, and with whom CARE had developed a relationship of trust. The knowledge and understanding of staff in local communities allowed teams to make informed decisions on what types of questions would be sensitive, adapt exercises as appropriate, and be able to probe into more delicate and sensitive issues around cultural practices, violence and sexuality (Bangladesh). However, in some cases, teams felt it would be best for staff not to research in the same zone where they work (Uganda, Mali) as this can make it more difficult for them to act as researchers, given their familiarity and relationships within the community.
During these studies, teams found it critical to disaggregate communities across categories of: wealth, labor groups, kinship clusters, household type, ethnicities, castes, occupation, age or position within family, or other groupings the research highlights as relevant. Across categories, teams then probed into gender roles and identities, livelihoods, access to services and rights and other aspects of life differed or mirrored one another across groupings. Further, teams often probed into how different groups related with one another in order to identify opportunities for pro-poor alliances, as well as to identify key barriers and develop strategies with communities to overcome them.
Further, in researching the definition and indicators of empowerment, teams highlighted the importance of continuing to recognize the diversity across groups and across women in defining empowerment. Thus discussions on empowerment must take place across different categories of women, could involve various methods (i.e. in-depth interviews, field observation or body maps), and should be returned to over time as women’s aspirations for themselves and definitions of empowerment continue to change.
Comparing across Sites
Comparisons across study sites have provided useful perspectives for a number of types of studies.
As mentioned, situational analysis studies have explored richer and poorer communities to understand differences between them and their relationships to development opportunities (Bangladesh, Nepal, Tanzania, Uganda). This helped teams not only to see important implications for project targeting, but also to understand how interventions and development initiatives can be manipulated by local power holders to reinforce their own positions within a community. Further, this comparison helps identify key dimensions of exclusion and how they differ between better-off and poorer communities, and reveal opportunities for change.
Assessing the Effects of CARE's Work
In order to develop a sense of the effects CARE’s work may have, a number of sites compared changes in women’s empowerment:
- Across communities which CARE has engaged at varying levels: In India, the STEP project examined two sites, one where its project worked to mobilize groups, and another where a government initiative worked to mobilize women’s groups without CARE’s support. In Mali, the research team chose to study one village in each of three categories:
- A village where two VSLA projects both worked;
- A village where only one VSLA project (Project A) worked; and
- A village where only one VSLA project (Project B) worked.
- Across women and their households who have had varying levels of engagement with CARE projects: In Burundi’s Phase III SII study, the team conducted research in two project zone areas. Within these areas, solidarity groups (and their collines, or local administrative unit) were randomly selected to take part in the study. For the research, teams interviewed women and men who fell into the following categories:
- Group A: Participants who were highly involved in the project, such as social change agents, peer educators, or those who have participated for a long time;
- Group B: Participants who are involved to a lesser degree;
- Group C: women who have not been directly involved in the project; and
- Group D: project staff and other stakeholdersThe use of groups helped teams examine varying levels of changes in women’s empowerment and HIV protective practices across different levels of engagement in CARE programming. Based on analysis, CARE could then determine the association between its work and changes in women’s vulnerability to HIV and AIDS.
- Across time (if information from baseline studies could be mined to identify change in women’s empowerment): In Niger, the Strategic Impact Inquiry studied changes across the stages and evolution of its VSLA intervention. For this approach, teams reviewed the assessment and evaluation reports across various phases of the intervention, as well as conducted focus group discussions and field visits to understand changes over time.
- In comparison with ‘desk controls’ or evidence from similar zones where CARE does not work: Through reports, interviews with informants from other organizations, etc., ‘desk controls’ may also be used as a point of comparison. In this process, teams study secondary reports for evidence of similar outcomes in zones where CARE does not work, or interviewed informants from organizations working elsewhere to see if they are seeing similar results in their zone of operations.
For studies, it is critical to consider sampling carefully and take serious consideration of research ethics in addition to research design needs.
For large-group participatory exercises, such as well-being analysis or village-level mapping exercises, it is important to ensure that each geographic cluster of the community being mapped is represented within the exercise.
For focus group discussions and smaller exercises, teams select participants based on the key group on whose lives they would like to explore (i.e. poor or very poor women or men). Within this group, the research team may also ensure that at least 2 of each 'type' of household is represented (i.e. poor female headed households, poor women in polygamous households, etc).
Teams should also be careful not to return to the same households for multiple exercises, and respect the time and voluntary nature of involved in people's participation. Further, exercises should be structured in a way that provides a safe space that enables researchers and participants alike to reflect and learn by mutually exploring gender and power dynamics.
For quantitative methods such as questionnaires or semi-structured interviews, many teams have used sample size calculators (a number are available online) to determine the number of people to survey in a study area. If the study aims to select a representative sample from a large population, it may be helpful to use the “30-Cluster” Sampling Methodology developed by WHO.
Random and Stratified Random Sampling
To select respondents, teams have used a random approach to select households – this may be based on a list of households within a village or area.
Many of the approaches described for gender analysis first stratify households based on the well-being analysis to identify poor, middle and wealthier households within the study area and then specifically select respondents according to their socio-economic status or other key characteristics (i.e. female-headed household, ethnicity, religion, age, etc.) within a community. For group discussions and interviews, teams often also make sure to select separate exercises with women’s groups, men’s groups, youth groups, etc.
Once stratified, teams often use random methods to select respondents, which may include:
- Drawing from a hat: with separate cards for each household or respondent within a given category (i.e. rich, middle, poor), teams randomly select respondents.
- Sequential sampling: teams list all households by number and then divide the total number of available households by the desired sample size. Then picking households by the quotient intervals (i.e. if there are 100 households and a desired sample size of 25, the team would select every 4th household).
- Using a random number table: Looking at the last digits of a random number table, in order, to select household numbers to sample.
Sometimes, the people teams seek to work with or study cannot be found through this method because they are very small in number, or difficult to identify. As a result, some studies have purposefully selected to ensure that female household heads or other key groups of interest are included in the exercises.
In the SII studies with sex workers, the underground nature of the work also made it very difficult to find respondents. Thus, teams worked through sex worker organizations to identify respondents, and also used respondent-driven sampling, where sex worker respondents were asked to inform three others about the study and how to participate in the research.
- B Bode (2009). CARE Nepal: Underlying Causes of Poverty Analysis and Contributions towards a Program Approach. CARE Nepal.
- B Bode (2009). The Causes and Conditions of Poverty in Acholiland, Northern Uganda. CARE Uganda.
- J Iredale and D Ntacobakinvuna (2009). Gender, Sex and Power: The Implications of Empowering Women at Risk of HIV and AIDS. CARE Burundi.
- T Klouda (2008). Links between ‘Empowerment’ and HIV Risk. CARE USA.
- B Bode (2007). Power Analysis in the Context of Rights-Based Programming. CARE Bangladesh.
- B Bode (2007) Disaggregating Community through Social Analysis. CARE Bangladesh.
- F Maiga, AK Coulibaly, RF Ngampana, M Fofana, O Keita, B Diallo and A Koné (2006) Facilitation Guide for In-depth Interviews and Focus Group Discussions with Women. CARE Mali.
- CARE (2007). Gender, Sex and the Power to Survive: the impact and implications of empowering women at risk of HIV and AIDS, Global Research Design Framework.
- CARE (2006). Global Research Framework for CARE’s Strategic Impact Inquiry on Women’s Empowerment.
- CARE India (2006). Strategic Impact Inquiry: Sustainable Tribal Empowerment Project.
- T Barton (1998). Program Impact Evaluation Process: M&E Tool Box. CARE International – Uganda.