*3.2. Methods*

Within the sampled villages, farming households were selected using stratified random sampling based on the gender and ethnicity of the household head, yielding a total sample size of 380 households1 (Kakum, *n* = 232; Ankasa, *n* = 148). The rationale for sampling the households along gender and ethnicity lines was to ensure that the study sufficiently captures important tenurial differences that exist within and across social groups in the study context [8,22,27,29]. In this regard, the majority of households surveyed were (320, or 84%) male-headed, reflecting the dominance of men in land-use decision-making, even in matrilineal contexts. To the extent that women dominate agriculture in Ghana, this suggests that the seemingly gender-disproportionate sample could be a limitation of this study [8,22]. Nevertheless, the careful constitution of the focus groups vis-à-vis the treatment of gender as an analytical category in this study afforded in-depth qualitative insights into the key differences between men and women regarding perceptions and experiences with local dispute resolution mechanisms.

The administered household surveys were supplemented with focus group interviews (in 8 communities), with farmers purposively selected for their in-depth knowledge of the communities, and key informant interviews with farmers, traditional authorities, farmer cooperative representatives and selected local- and national-level officials of key land sector agencies (including from the Assinman Customary Land Secretariat (CLS), Office

<sup>1</sup> In instances where some household members managed their landholdings independently, they were also interviewed to capture specific variations in tenurial dynamics. This approach afforded an opportunity to capture the views of such household members (particularly female plot managers within male-headed households) who otherwise would have been excluded.

of the Administrator of Stool Lands (OASL) and Ghana National Land Administration Project Secretariat), which afforded an opportunity to sufficiently explain and capture some local-level nuances on tenurial dynamics, which otherwise would have been difficult to capture in a wholly quantitative study. The semi-structured interviews with the key stakeholders in customary land management conducted in Twi2 focused on collating information on the prevailing tenurial situation in the studied communities, perceptions about land tenure security, land disputes and mechanisms for safeguarding land rights among others. The collated interview data from the semi-structured interviews and Focus Group Discussions (FGDs) were transcribed, followed by content and thematic analysis of the ensuing textual data guided by Miles and Huberman's [51] general strategy for qualitative analysis. The survey data were coded, entered, cross-checked for accuracy and analysed using SPSS (version 20). Cross-tabulation analysis and Pearson's chi-square tests were performed to detect significant differences (*p* < 0.05) between groups based on study area, gender and ethnicity. A dispute perception score was developed to ascertain the severity of land-related disputes in the studied communities. Using a five-point Likert scale, the opinions of respondents were measured by assigning a weight of 5 if 'they felt land disputes are a very serious problem', 4 if 'they felt land disputes are a serious problem,' 3 if 'they were uncertain,' 2 if 'they felt land disputes are somewhat not a problem' and 1 if 'they felt land disputes are not at all a problem'. Therefore, a high score above the median score meant a positive perception to the statement (i.e., severity of land disputes in a said community) and vice-versa. It was assumed that respondents who were uncertain as to whether land-related disputes are problematic in their communities would have a median score, which in this case is 3. The final (mean) score represents the cumulative perception of respondents on a particular statement of the severity of land-related disputes in the surveyed communities.
