Social Contagion and Trial on the Internet: Evidence from Online Grocery Retailing

Abstract

For a traditional retailer, the size of the customer pool can evolve over time but is largely bounded in space. In contrast, an Internet retailer with the appropriate shipping infrastructure can draw customers from a wide-ranging geographical area (e.g., the entire United States). We examine the trial decision for customers shopping at an Internet grocery retailer. Drawing on literature in economics, marketing and sociology, we conjecture that the trial decision may be subject to social influence or contagion. That is, exposure to the actions of proximate others - either through direct social interaction or passive observation - influences the trial decision of individuals who have yet to experience the service.

This idea is tested in a discrete time hazard setting in which consumer trial decisions arise from utility-maximizing behavior. Moreover, our derivation allows use of region-level data to estimate a model consistent with individual utility maximization, even in the absence of individual level covariates. We find that the marginal impact of the so-called neighborhood effect is economically meaningful as it results in an approximately fifty percent increase in the baseline hazard of trial. This effect is robust to the inclusion of a broad set of covariates, region-level fixed effects, and time-dependent heterogeneity in the baseline trial rate. The model is calibrated on a unique dataset spanning: (1) 29,701 residential zip codes, (2) 156,049 customer transactions over forty-five months, and (3) zip code specific contiguity data from Geographic Information System (GIS) analysis. Substantive implications for customer base evolution and Internet retailing are discussed.

Keywords: Discrete Time Hazard, Neighborhood Effects, Retailing, Social Contagion