Emotional interaction and conversion value in live streaming E-commerce

Multi-agent emotional interaction constitutes a core characteristic of live streaming e-commerce. Nevertheless, prior research has yet to build a unified analytical framework incorporating streamers, viewers and live streaming assistants, and rarely explored how emotional dynamics shape live streaming performance. Based on minute-level panel data of 4671 min retrieved from 40 Taobao live streaming sessions, this study adopts sentiment analysis to quantify the emotional expressions of streamers and viewers. Applying the Panel Vector Autoregression (PVAR) model and impulse response analysis, this study constructs a tripartite emotional interaction system among streamers, viewers and assistants, and investigates its dynamic linkage with core operational performance, including product sales and new follower growth. The empirical results are consistent with a unidirectional streamer-to-viewer emotional association, whereby streamer emotions are closely associated with viewer interaction and purchase decisions. By contrast, intra-viewer emotional association is not statistically supported, whereas viewer emotions are positively associated with new follower growth. Higher frequency of live streaming assistant intervention shows a dual association pattern: it is positively associated with viewer interaction frequency, but negatively associated with product sales. This study extends the application scope of emotional contagion theory in the live streaming context, enriches the role configuration of the live streaming e-commerce ecosystem, and offers practical and targeted operational strategies for live streaming management.