EXPRESS: Consumer Interactions and Peer Effects in Socially-Connected Digital Products
研究数字社交平台中用户流失的同伴效应，构建两阶段结构模型纠正内生性偏差，发现同伴效应显著放大干预效果（平均社会乘数2.48），且用户影响他人留存的能力差异远大于自身留存概率差异。
The paper explores peer effects in user churn within a digital socially-connected platform and investigates how firms can leverage peer effects and network structure for customer relationship management (CRM). The data come from a massively-multiplayer online game, where gamer social ties are endogenous. We build a two-stage structural model to capture network formation and peer effects in churn. Unobserved gamer heterogeneity inferred from network formation corrects for endogeneity bias in peer effects. The main findings are: (i) Peer effects substantially amplify intervention effectiveness, with an average social multiplier of 2.48. Critically, the full model with unobserved gamer heterogeneity reveals substantial cross-network variation in social multipliers that the restricted model largely masks. (ii) Users vary far more in how they influence peers' retention than in their own likelihood of retention. (iii) User's network value is driven primarily by network centrality measures, unlike user's direct value which correlates more with individual characteristics.