Household surveys typically begin with the creation of a household roster, which is a list of household members.  While creating an accurate list of household members is a straightforward task in many living situations, it can be difficult for some unique living arrangements, such as those with roommates, long-term guests, multiple homes, etc.  Inaccurately identifying household members can result in coverage error either by omitting applicable members (undercoverage) or by including inapplicable members (overcoverage). To minimize household coverage error, the American Community Survey uses four probing questions in the survey’s automated modes (Internet, Computer Administered Telephone Interview (CATI), and Computer Administered Personal Interview (CAPI)).  These probes are designed to reduce undercoverage by reminding respondents to include household members they may have initially forgotten, and reduce overcoverage by removing people who should not be counted as living or staying at the address according to the American Community Survey residence rules.  

The purpose of this study is to learn more about the impact of these coverage questions on the number of people in a household.  The research uses 2015 ACS data to explore the number and characteristics of households that added or deleted members as a result of the additional probing questions.  Furthermore, the research includes an analysis of survey paradata from the Internet mode to investigate how respondents interacted with these additional coverage questions, such as how long they took to complete them and whether they clicked on a help link.  Results from the research suggest that while the roster questions may help improve household coverage, especially for specific populations, they also increase respondent burden.  This research paper presents the results of this analysis.