Product Assortment and Competition: Evidence from Live-Streaming Platform Influencers

Abstract

This paper studies how an unexpected change in competition caused by the exit of a superstar affects the sales strategies of influencers on a live-streaming platform during live-commerce. We propose a novel multi-task graph convolutional network approach to measure competition intensity that can capture both direct and indirect interactions of market participants. We leverage a natural experiment in which the top influencer with the strongest market power quit the platform by accident. Our results indicate that, lower market competition leads to higher product variety, while retail prices are sticky in the short run. Heterogeneity analysis suggests that the effect is larger for less popular influencers and those experiencing greater competition relief. We also find evidence that influencers make their sales decisions mainly based on the overall market information rather than relative changes in sub-markets.

Keywords: Social e-commerce, Product variety, Platform economy, Deep learning

JEL Classification: D43, L81