*3.3. Data Analysis*

We began data analysis in summer 2020 with the intent of examining how narratives changed during the early months of the pandemic, specifically with documents from January through April 2020. We began by conducting a pilot analysis of approximately 10% of the data, taking notes on patterns that emerged from the text. This preliminary analysis served as the foundation of a codebook that the four-person research team developed across multiple meetings and rounds of testing. After reaching an intercoder reliability score

of 90%, we divided the remaining data between authors for systematic coding. We used NVivo, a qualitative software program that facilitates the coding process and provides a systematic way of analyzing data, for paragraph-level coding of documents that mentioned COVID-19 (or a variation of the term).

During this phase of coding, we continued to meet regularly to discuss concerns and periodically conduct further intercoder reliability checks. Through these conversations, we realized that, to fully address our research question on how groups have responded to the pandemic crisis, we needed to expand data analysis to include documents from the entirety of 2020. We also realized the importance of including documents that did not specifically address COVID-19 but were nonetheless published during this period, which allowed us to consider what these organizations were talking about when their conversations did not center on the pandemic. Thus, we expanded our initial data collection and analysis efforts to include all documents published in the year 2020. In addition, we supplemented our primary data with a sample of news media coverage<sup>6</sup> on both the pandemic and the immigration detention industry in order to contextualize the emergen<sup>t</sup> patterns.
