In this article, we study the effect of peer influence on products that exhibit positive network externalities to non-adopters (i.e., products that benefit adopters’ friends even if they do not adopt). Contrary to products that exhibit positive network externalities upon adoption, this structure of incentives likely results in negative peer influence: the more friends that adopted the product, the smaller the incentives to adopt. We measure this effect empirically by using observational data from a large mobile carrier serving 5.7 million users. We estimate the effect of peer influence across five different products of this type. A naive approach to do so results in a positive estimate for peer influence, due to unobserved homophily. We follow two approaches to address this issue. First, we suggest using the number of friends that end up adopting the product as a proxy for unobserved user fixed effects. Second, we control for homophily by applying a shuffle test (i.e., we compare the effect of peer influence from the original data with the effect obtained from comparable, randomly generated data without peer influence). We obtain negative estimates from both approaches, which provides robustness to our findings. Finally, we show that even for these products, the effect of peer influence associated with the first friends that adopt the product is positive, which arises because they still convey useful information about reducing uncertainty. The negative effect of peer influence arises only for the subsequent friends that adopt the product. These friends are unlikely to convey new information about the product, but each of them decreases the economic incentive to adopt, resulting in a negative aggregate effect of peer influence.

中文翻译：

---

在具有正网络外部性的产品中搭便车：来自大型移动网络的经验证据

在本文中，我们研究了同行影响对对非采用者表现出积极网络外部性的产品（即，即使采用者的朋友不采用也使他们受益的产品）的影响。与在采用时表现出正网络外部性的产品相反，这种激励结构可能会导致负面的同伴影响：采用该产品的朋友越多，采用该产品的激励就越小。我们通过使用来自为 570 万用户提供服务的大型移动运营商的观察数据来凭经验衡量这种影响。我们估计了同行影响对这种类型的五种不同产品的影响。由于未观察到的同质性，这样做的幼稚方法会导致对同伴影响的积极估计。我们采用两种方法来解决这个问题。第一的，我们建议使用最终采用该产品的朋友数量作为未观察到的用户固定效应的代理。其次，我们通过应用洗牌测试来控制同质性（即，我们将来自原始数据的同伴影响的影响与从没有同伴影响的可比较、随机生成的数据中获得的影响进行比较）。我们从这两种方法中获得了负面估计，这为我们的研究结果提供了稳健性。最后，我们表明，即使对于这些产品，与采用该产品的第一批朋友相关的同伴影响的影响也是积极的，这是因为它们仍然传达了有关减少不确定性的有用信息。同行影响的负面影响只对后续采用该产品的朋友产生。这些朋友不太可能传达有关产品的新信息，