Emotion Analysis in Narrative Economics

Abstract

The purpose of this quantitative cross-sectional design using a structural equation model was to determine if emotions mediate the viral spread of narratives with potentially far-reaching economic and political consequences. This dissertation addressed the problem of potential media manipulation through emotionally charged narratives that go viral and have a financial or political impact, such as meme stocks, cryptocurrency, or election interference. The objective of this dissertation was to see if emotions mediate the viral spread of narratives with potentially far-reaching economic and political consequences. A Big Data set from the COVID-19 pandemic years of 2020 until 2022 consisting of English news stories with emotional charge to go viral and trigger economic consequences was used for analysis. This data is from the publicly available GDELT database. The research methodology and design were a structural equation model (SEM) with a mix of latent exogen and observed endogen variables, specifically a partially latent structural regression. The results confirmed emotions as a significant (p < .01) mediator of the virality of English news stories within the selected timeframe for all three analyzed currencies (β = .23 USD, β = .06 EUR, β = -.07 GBP), as hypothesized by emotion dynamics theory and narrative economics. The economic consequences triggered by this emotion-mediated virality, although all significant (p < .01), were not in every case as predicted by prospect theory and narrative economics but only for the economic consequences with amounts in GBP (β = .02) and EUR (β = .06). For economic consequences with amounts in USD (β = -.01), the data did not correspond to the prospect theory rationale. EUR was the only currency for which all theoretical predictions of emotion dynamics theory, prospect theory, and narrative economics held true, while also being the only currency for which the dataset was normally distributed, and the ML estimator was sufficient to achieve good fit indexes.