Online social and information networks play an important role in risk perception and economic decision-making. However, analyzing the cognitive and behavioral effects of online social learning is challenging because of the absence of scalable network metrics to capture the topic-specific information flows. We propose a data-driven approach to model the social media information connectivity across space, leveraging high granularity Twitter data and Elastic-Net augmented variance decomposition. Our method produces the directionality and intensity of information spillovers between each location pair (state/MSA) during the two-month period surrounding Hurricane Ida landfall. We find that the structure of the hurricane information network is influenced by both universal factors (such as geographical distance, friend linkages, and socio-demographics) and hurricane-specific factors (such as hurricane severity and climate risk). Our information network will be useful for understanding the drivers and barriers of climate information transmission and for modeling the effects of online social learning on offline climate adaptation behaviors.

Speaker: Yichun Fan