Consistency and accuracy of USPS-provided undeliverable codes

Abstract

When survey mailings are returned as “undeliverable” in the US, we rely on information codes provided by the U.S. Postal Service (USPS) to determine the eligibility of a sampled household. Although little is published about the consistency and accuracy of USPS undeliverable codes, they are often accepted at face value and used to make operational decisions about further contact and the ultimate disposition of a sample case. For the Residential Energy Consumption Survey (RECS), we examine the quality of the USPS’s application of 8 different undeliverable codes. First, we review USPS codes from 3 mailings for the self-administered RECS National Pilot sent within a 12-day period to all 9,650 sampled households. We are particularly interested in codes that influence whether a case is disposed as an eligible occupied household or an ineligible vacant or seasonal household. Next, we compare USPS codes to field-verified status collected from the 2015 RECS main, field study. Then, we assess a model to predict the occupancy status of a case using auxiliary data that enhances known characteristics from the sampling frame and we assess the impact of that model on occupancy, coverage, and bias. Finally, we discuss the implications for operational tracking and decision-making on sample cases, the use of specific undeliverable codes in determining case eligibility, and the construction of sampling frames.