Estimating the impact of e-commerce on retail exit and entry using Google Trends

I address the degree to which variation in exposure to e-commerce is associated with establishment entry and exit in the retail industry at the county level. To measure exposure to e-commerce, I rely on within-state variation in relative search frequency for the phrase “amazon prime†as reported by Google Trends. To generate exogenous variation in this e-commerce exposure measure, I use within state variation in the relative search frequency for “porn†and “cat videos†. Fixed effects instrumental variable estimates suggest at least 10 of the 27 retail industry groups experience net exit with increasing e-commerce exposure, while at least 6 experience net entry. To address endogeneity concerns about my instruments, particularly that they are driven by a notion of “hipster-ness†, I conduct a robustness check to show that my results fail to replicate in consideration of a strategy to tease out this identification threat.