Evaluating three ways to handle drop points in address-based sampling frame surveys

Abstract

Address-based sampling frames are commonly created from lists derived from the United States Postal Service’s Computerized Delivery Sequence file (CDS), a comprehensive list of addresses in the United States. While most addresses in the CDS represent only one household, some addresses—known as drop points (DPs)—are delivery points for two or more households. The drop point units (DPUs) therein do not have secondary unit designators specified in the CDS, which is a challenge for self-administered surveys because they cannot be contacted specifically. An earlier paper by the authors examined differences between responses of DPUs and nearby non-DPU substitutes in the 2021 Healthy Chicago Survey of sociodemographic characteristics and key health outcomes. This paper is a follow-up analysis comparing the effects of including, excluding, or substituting DPUs on the entire sampling frame, and thus entire study area, to quantify the magnitude of point estimate differences between those methods. No statistically significant differences were found in the distributions of key health outcomes for the three methods. This is true even when focusing on the 11 community areas where more than 30 percent of the homes were DPUs.