Typological Analysis of Pay-What-You-Want Donation Behavior in Virtual World

This study examines pay-what-you-want donation behavior among viewers of Twitch, the live-streaming platform, with the aim of explaining why monetization outcomes differ across streamer communities and what this implies for monetization in the metaverse era. Motivated by the growth of virtual-world support activities during the COVID-19 pandemic, the authors adopt service-dominant logic as their primary analytical framework, supplemented by uses and gratifications theory and social density theory. Each streamer-viewer community serves as the unit of analysis, with gifted subscriptions used as a proxy for donations, average concurrent viewers as a measure of viewer effort, and monthly hours streamed as a measure of streamer effort.

The dataset covers 100 consistently streaming Twitch partners tracked over 19 months, yielding 1,900 monthly observations. Descriptive statistics reveal very wide variation in donations, with monthly gifted subscriptions ranging from 37 to 33,869, which the authors take as evidence that communities differ in meaningful ways and warrant a typological approach. Applying k-means clustering, with the Bayesian Information Criterion indicating four groups as optimal, the authors identify four distinct community types: a large group of 63 communities characterized by balanced and stable mutual engagement; a group of 15 driven primarily by one-sided viewer effort and high concurrent viewership; a group of 15 driven by streamer effort and multifaceted content; and a small group of 7 that achieves above-average donations despite average viewership, attributed to distinctive streamer personalities and formats. The central conclusion is that examining streamer-side and viewer-side metrics together reveals community-level patterns that viewer-only analyses overlook, offering a basis for more targeted monetization strategies.