Description:

People possess multiple identities, the relative salience of which responds dynamically to social context. Whether one thinks of oneself as a parent, an employee, a spouse, a particular gender, or an American varies from place to place and time to time. National identity in particular is a crucial master frame through which people understand their sense of collective belonging and their political choices. Yet, research on national identification has been limited to cross-sectional data, often gathered using single survey items, which fail to take into account the contextually of identity. With large-scale social media data a different approach is possible. Our study traces daily fluctuations in American identification using a random sample of U.S. Twitter/X users between 2012 and 2022. Because identifying tweets that engage with nationhood is akin to looking for a needle in a haystack, we employ few-shot classification using Llama 2, a large language model, combined with fine-tuning to further improve model performance. Our findings reveal that American identification rises and falls in a patterned manner, partly in response to nation-relevant events widely reported in the media. We show which events have the largest impact on nationalism time trends, estimate the collective half-life of identification events, and examine which events are politically galvanizing versus polarizing. In particular, we find that national identification gained chronic salience after 2016, when it first became a central point of contention in U.S. presidential elections.

Bio:

Bart Bonikowski is Associate Professor of Sociology and Politics at New York University. Using relational survey methods, computational text analysis, and experimental research, his work applies insights from cultural sociology to the study of politics in the United States and Europe, with a particular focus on nationalism, populism, and radical-right parties.