A Computational Framework for Understanding Firm Communication During Disasters

Large firms are leaders in disaster response and communication. We study how firms communicate on social media during various disasters and the relationship between their communication and public engagement using a computationally intensive theory construction framework. The framework incorporates a novel natural language processing (NLP) approach, Semantic Projection with Active Retrieval (SPAR), as a key component of the method lexicon. Drawing on the two dimensions (internal versus external and stable versus flexible) of the Competing Values Framework (CVF) as our theoretical lexicon, we examine Facebook posts of Russell 3000 firms on multiple disasters between 2009 and 2022. We find that social media messages that are internal- and stable-oriented, or emphasize operational continuity, are more likely to elicit engagement from the public during biological disasters. By contrast, messages that are external- and flexible-oriented, or stress the innovations to adapt to the disaster, induce more engagement in weather-related disasters. The study offers theoretical implications and methodological support for the research and design of social media messages in disasters and other contexts.