Analysis
We assigned different roles to researchers to increase our ability to analyze case evidence in multiple iterations from divergent viewpoints (Eisenhardt, 1989). To record our observations and enfold them with extant theories, we made extensive use of memos (Lempert, 2007). First, we wrote 17 single-spaced pages of field notes beginning in March 2020 (Van Maanen, 2011). Second, we wrote 31 single-spaced pages of memos while analyzing secondary documents for each case. Third, we wrote 77 single-spaced pages of additional memos on topics such as the sampling process, interview insights, potential themes and variables emerging from the data, data collection and analysis procedures, and relevant literature.
Structured analysis of our interview data proceeded in two ways. First, we analyzed the quantitative and textual data provided through Qualtrics. We generated a database of responses encompassing Likert scales (e.g., How well did the types of PPE produce match what recipients needed? Where 1=very poor match, and 5=very good match), dates (e.g., first involvement in PPE production), raw numbers (e.g., amount of PPE produced), percentages (e.g., percentage of PPE provided to different types of recipients) and lists of organizations and names people interacted with. While statistical analysis of this data was not appropriate due to the sample size, this approach allowed us to capture comparable data in clear categories while also applying techniques from network studies, such as a name generator (Burt, 2011). This approach enabled us to analyze network ties and interactions as well as to create case history timelines in combination with dates found in secondary documents.
Second, we imported and categorized interview transcripts in NVivo for coding and constant comparative analysis (Eisenhardt, 1989). We engaged in open coding (Strauss & Corbin, 1998), reading the data line by line to generate 147 codes. Interim findings were discussed within the author team, helping us identify 59 additional codes related to two main themes that emerged: resourcefulness and legitimacy. Codes were then categorized into first order concepts and second order theoretical themes. From this process, we derived aggregate theoretical dimensions that include resource constraints, legitimacy deficits, network emergence and evolution, learned resourcefulness, and learned legitimation. As the final step in our analysis, we developed a theoretical model of these dimensions and their relationships. Throughout these steps, we continually revisited the literature to identify prior constructs and explanations of relationships related to our emergent findings. See Appendix C for additional analysis details.