Results

Interviews were audio-recorded, fully transcribed and subjected to thematic analysis (Braun & Clarke, 2006; Joffe, 2011) using the software NVivo. This resulted in a coding frame comprising 20 top-level and 18 sub-codes, which were both theoretically driven and inductively grounded in the data. The coding frame was checked for inter-researcher reliability by independently coding two entire interviews by two researchers, which revealed high levels of agreement and clarified and refined code definitions. To facilitate comparisons across the case studies, the coded data were summarised using a number of matrices (based on the key themes emerging from the thematic analysis), containing all the coded data per theme ordered per case. This enabled iterative, interpretative analysis that investigated cross-case patterns and compared these internally and also with other themes. Furthermore, patterns and commonalities emerging from the themes were analysed in relation to various typologies of the nine projects, in particular, focusing on differences between the community, private, public and professional third sector.