A Text-as-Data Approach to the Strategic Performance of Interstate Relations

• In Event: Mathematical Sociology Section Flash Talks
  In Refereed Roundtable (60 minute): Table 1. A Text-as-Data Approach to the Strategic Performance of Interstate Relations

Abstract

States routinely issue public National Security Strategy (NSS) documents outlining
their defense priorities. Are these merely bureaucratic exercises, or can they
provide meaningful insights into interstate relations? This paper argues that NSS
documents are more than cheap talk; they function as dynamic texts that both reflect
states’ perceived relational hierarchies and act as performative constructions
of state social signals. This paper introduces a novel, directed‐dyad index of affective
salience—combining prominence (salience) and evaluative stance (affinity)—
calculated using large language models, sentiment analysis, and stance detection.
Analysis of over 400 documents from 92 countries (1962-2024) reveals salience is
primarily driven by geographic proximity and material capabilities, while affinity
largely mirrors traditional indicators of alliance and rivalry. Affective salience
demonstrates significant predictive capacity for future international crises and interstate
diplomatic visits even when controlling for established factors, with performance
exceeding that of established measures like alliance portfolio similarity and
UN voting patterns. This research challenges the conventional dismissal of public
strategic documents as cheap talk, demonstrating that systematic computational
analysis offers robust insights into how states publicly articulate their relational
priorities.

Author

• Andres Gannon, Vanderbilt University