New approach for handling drop point addresses in mail/web surveys

Abstract

The purpose of this paper is to introduce the concept of drop unit substitution in
address-based samples for mail and web surveys. A drop point is a single US Postal Service (USPS) delivery point or receptacle that services multiple businesses, families, or households (USPS, 2017). Residential drop units are the individual housing units served by the drop point address. For the most part, address-based sampling frames list the number of units at a drop point address but will not contain information identifying specific units. Drop units comprise less than 2 percent of all residential addresses in the United States (McMichael, 2017), but they tend to be concentrated in certain large cities. In Queens, New York, for example, drop units constitute 27 percent of residential housing units. The problem with drop units for address-based surveys with mail contacts is that, without names or unit identifiers, there is no way to control which unit receives the various mailings. This limitation leads to distorted selection probabilities, renders the use of cash incentives by mail impractical, and precludes traditional methods for mail nonresponse follow-up, thus resulting in higher nonresponse. Alternatively, excluding drop units results in coverage error, which can be considerable for some subnational estimates. The authors propose a substitution approach when a drop unit is sampled—in other words, replacing the unit with a similar nearby unit in a non–drop point building.