To Glance or to Peruse: Observational and Active Learning from Peer Consumers

This paper examines consumer social learning patterns in decision making. I propose a novel model that decomposes the learning process into two stages: observational learning, where a consumer quickly updates the belief about a product after observing its salient social-based characteristics (such as popularity), and time-consuming active learning through descriptive information content (such as consumer reviews). By demonstrating the interplay between the two stages, the model brings together previous literature that studies these separately. I characterize the optimal learning time, and provide comparative statics which show that an increase in the discount rate or in the product’s economic value drives consumers to rely more on observational learning. I test this model using unique transaction-level data for air purifiers sold on a Chinese online platform from January to March 2014. Exploiting an unexpected air pollution crisis in late February 2014, I find that past sales have greater weight as a reference for comparison among products during the pollution crisis than in regular times. I also document that, after the episode, consumers rely less on observational learning compared to periods before the crisis, which is consistent with the model’s predictions as sales made during the crisis convey less information.