When Workloads Are Emotional Labor: An Empirical Study of Livestreaming Productivity

Abstract

Livestreaming e-commerce has grown rapidly in recent years, yet little is known about how livestreamers' work patterns affect their performance and productivity. On the one hand, the prevailing view is that because livestreamers compete for limited customer attention, longer streams maximize their exposure, viewership, and sales-regardless of a livestreamer's productivity per hour. On the other hand, livestreamers engage in emotional labor, and their performances critically depend on factors like energy and affect, which may be highly sensitive to working extended hours. As a result, the relationship between workload, sales, and engagement remains unclear. This study leverages a natural experiment arising from localized policy changes in China that restricted excessive working hours on livestreaming platforms to examine the causal impact of workloads on performance. Using a difference-indifferences (DiD) approach, we empirically find that shortening streaming durations significantly increases both viewer engagement and sales, by respectively 24.6% and 42.1%, challenging the common assumption that "longer is better." To uncover the mechanisms underlying these results, we analyze live video feed data using artificial intelligence (AI) technologies to measure livestreamers' levels of emotional expressiveness and use of verbal tactics. We find that longer streams cause livestreamers to exhibit diminished expressiveness and use of effective tactics, negatively impacting viewer engagement and sales outcomes. These findings underscore how diminishing or negative returns-stemming from livestreamer fatigue-can outweigh the benefits of prolonging customer exposure. Our study highlights the importance of strategic bursts of content over continuous streaming and offers actionable insights for livestreamers, brands, and platform managers seeking to optimize livestreaming performance.

Keywords: livestreaming commerce, emotional worker, workload, natural experiment, causal inference Workloads in Livestreaming Commerce