ABSTRACT

Previous research has shown that there exist “harbinger customers,” who systematically purchase new products that fail and are discontinued by retailers. We extend this finding in two ways. First, we show that there are also some zip codes that are “harbinger zip codes”.  If households in these zip codes adopt a new product this is a signal that the new product will fail. Second, we show that households in these zip codes make choices that are systematically different from other households across a surprising array of decisions.
We identify harbinger zip codes using purchases of new products from one retailer.  We then analyze purchases from these zip codes at a different retailer. We show that the harbinger zip codes identified at the first retailer purchase products from the second retailer that other zip codes are less likely to purchase. They also purchase products that receive less favorable customer reviews and that other customers are more likely to return.
By studying households that change zip codes we are able to investigate whether harbinger zip codes result from households with harbinger preferences choosing to cluster together, or households learning these preferences by observing their neighbors.  The evidence strongly supports the first explanation. When households move from a harbinger zip code they tend to move to another harbinger zip code, while non harbingers do the reverse. However, we find no evidence that households’ preferences change after they move. It appears that harbinger preferences are a sticky trait, and the harbinger zip code effect is more due to where customers choose to live, rather than households influencing the preferences of their neighbors.
We then investigate whether harbinger zipcodes are predictive outside the retail domain. We use the harbinger zip codes identified at the first retailer to compare donations to congressional election candidates. Households in harbinger zip codes donate to different candidates than households in neighboring zip codes, and they donate to congressional election candidates that are less likely to win.  We also compare changes in house prices between 2002 and 2015. House prices in harbinger zip codes increased at slower rates than other zip codes. All of our results hold even when we compare five-digit zip codes with neighboring zip codes that share the same 3-digit zip code.

Duncan Simester, Catherine Tucker and Clair Yang