Modeling cross-media agenda interactions in the international communication of Chinese culture using multi-source news time-series data

Abstract Modeling cross-media agenda evolution is challenging because media agendas involve both temporal dependence in issue salience and structural reconfiguration among issue associations. Most existing intermedia agenda studies rely on contemporaneous correlations or simple lag comparisons, which reveal synchronization but provide limited evidence of predictive precedence across heterogeneous media systems. To address this limitation, this study develops an integrated time-series framework combining transformer-based agenda extraction, network agenda representation, cross-lag correlation analysis, vector autoregressive modeling, Granger-type predictive-dependence testing, and transfer entropy estimation. Using multi-source news time-series data on the international communication of Chinese culture from January 2021 to December 2023, we construct outlet-level measures of issue salience and structural agenda density and examine temporal-dependence patterns among official international media, global mainstream media, and digital-native or transnational online media. The results reveal structured predictive-precedence gradients across institutional media categories and hub-mediated agenda connectivity centered on selected intermediary outlets. Models incorporating structural indicators achieve higher explanatory power and lower forecast error than salience-only models, indicating that network configuration provides incremental predictive information beyond issue prominence alone. Overall, the findings suggest that cross-media agenda evolution reflects structured temporal dependence and directed information association rather than coincidental co-movement or definitive causal influence.