These data are part of a data portal that accompanies the special issue ‘Climate change adaptation needs a science of culture,’ published in Philosophical Transactions of the Royal Society B in 2023. To access the data portal, please visit 10.5061/dryad.bnzs7h4h4.

We worked with community leaders and local researchers in Andavadoaka, Madagascar, to develop the two focus group discussion guides and the survey that form this toolkit. Each is rooted in the central question, "What do you wish that conservationists and scientists asked you before suggesting changes for your community as you confront climate change and biodiversity loss?". This toolkit is designed to guide interviewers through assessing either 1) a community-produced project idea or 2) an external-generated prospective project idea or ongoing project (to assess if and how community and collaborator goals for this project could align). In our case, community leaders involved with the development of this toolkit used it to solicit problems and solutions from the community and explore a novel, community-selected project idea, rather than to collect feedback on an ongoing intervention. We then piloted the toolkit implementation process and present here the results alongside suggestions for the future use of these materials. The Agency Toolkit module is designed to be self-contained and assist the collaborator through each step of the process within a two-week period. We detail a suggested schedule in Supplementary Materials 5.

Before the project, authors met with community leaders to introduce the project and ask for their input in creating a tool to guide conversations on conservation and climate change adaptation initiatives. We explained both our desire for and ability to incorporate their feedback directly into the tool itself, as well as the desire to pilot the tool in the community. After gaining initial feedback and approval, we introduced these same study aims to all members of the focus group, who acted to disseminate this information to the broader community. Tools were translated into the local Vezo dialect of Malagasy, and back-translated to ensure correct meanings and cultural relevance of all questions.

For focus groups, G.M. and R.S. extended invitations to men and women in the community by visiting every household in Andavadoaka. Upon invitation, we informed participants that attendance was fully voluntary and that declining engagement in the study would not affect their relationship with the Morombe Archaeological Project or any of its affiliates. Every person who was willing to participate was selected for the focus groups. At the start of each focus group, we obtained informed oral consent and communicated to participants that they could withdraw at any point during the study, leave a discussion or interview early, or skip any questions they did not feel comfortable answering without affecting the compensation for their time or their relationship with the researchers. We received consent to audio-record focus groups. Members were provided with refreshments and compensation equivalent to USD 7 (based on a local average wage of 30,000 Ariary per day, yielding approximately $1 per hour of engagement with this project).

Similarly, we documented oral consent for all participants who agreed to individual interviews. Participants did not receive compensation for participating in individual interviews. To ensure the survey respondents’ anonymity, we removed names and identifying information from data made accessible for open-source access. The Columbia University Internal Review Board reviewed and deemed this research exempt from approval (Columbia IRB #AAAU3639; PSU IRB #00021067). We conducted this research with the requisite permits and approval from the Centre de Documentation et de Recherche sur l’Art et les Traditions Orales à Madagascar and l’Université de Toliara.

As part of the co-production of this tool, all survey questions were workshopped with local community members and subject matter experts. We recommend this workshop and adaptation process be a part of tool implementation to 1. ensure the instrument asks questions of interest to community members in ways that are 2. relevant to the local culture and topics of concern. The survey serves as a base framework that can be adapted with each use.

In this region, as in many regions where this toolkit may be implemented, an existing sampling frame (such as a census) was not available. We surveyed 149 individuals, 99 women (65.56%) and 50 men (33.11%); ~10% of the study population. This provided an 8.8% margin of error within a 95% confidence level. In order to effectively sample enough participants while also accounting for differing experiences across the population, we implemented a multistage stratified sample. We mapped our community by employing survey teams to document and number each household in the target area. Available demographic data from the recent regional census (2014–2015, available here) suggested the total population of Andavadoaka to be 1,355, approximately 51.7% female and 48.3% male (age subset information for Andavadoaka was not available). We then used a random number generator to select the numbers of households to be sampled. We marked these on each surveyor’s map and allotted a total quota of people per interviewer (one person per household). We subdivided this quota as half male and half female and across genders a roughly third of each gender in the following age brackets: 18–24 (young adult), 24–49 (adult), and 50+ (senior). If a household was not available, that interviewer marked the house for one follow-up but otherwise proceeded to the following house. At the individual level, each interviewer cycled through the substrata until each sub-quota was complete. This multistage stratified approach allowed us to cover large areas, while also sampling a diverse subset of the population.

Interviews took place inside or near each individual’s home, at the discretion of the participant, to protect the participant’s privacy. Researchers fluent in Malagasy obtained consent from participants and administered the survey in Enketo Webforms (KoboToolbox) on iPads. Survey questions are available in Supplementary Materials 1. Following the BCQ, barriers were measured on a Likert scale from "No problem – 1", "Small problem – 2", "Problem – 3," "Big Problem – 4." and "Very Big Problem – 5." 

The R code file (Barriers.R) processes and analyzes the data in Cleaned Data.csv. For a description of what variables and values mean, see the data dictionary.

Data and code are also available via GitHub (https://github.com/dcbuffa/agency-toolkit).