Incidence and Impact of Controlled Access Situations on Unit Nonresponse
Cunningham, D. B., Flicker, L., Murphy, J. J., Aldworth, J., Myers, S. K., & Kennet, J. (2005, May). Incidence and Impact of Controlled Access Situations on Unit Nonresponse. Presented at American Association for Public Opinion Research 60th Annual Conference, Miami Beach, FL.
The term “controlled access” applies to any situation where an obstacle keeps an interviewer from reaching the door of a potential respondent. Failure to reach controlled access dwelling units may introduce bias through systematic under-representation of certain sub-groups. For example, high-income and urban households are more commonly found in controlled access situations than other sub-groups in the United States population (Blakley and Snyder, 1999).
In recent years, the National Survey on Drug Use and Health (NSDUH), a federally sponsored annual survey that gathers data on substance use and abuse among the non-institutionalized household population of the United States, has seen an increase in the amount of nonresponse attributed to controlled access. Through this experience, NSDUH researchers have developed protocols for overcoming barriers to data collection. However, until 2004 the project had not collected the data necessary to analyze controlled access at the dwelling unit level nor analyzed the effect of controlled access on unit or item nonresponse. In order to develop these capabilities, the NSDUH now systematically captures housing characteristics, controlled access status, and the outcome of all controlled access situations for all 180,000 housing units in the annual NSDUH sample.
This presentation will summarize the incidence of controlled access by dwelling unit type and state, and introduce a model that predicts the effects of controlled access barriers on unit nonresponse. We predict that controlled access will be correlated with unit nonresponse. Finally, we will discuss ideas for investigating the role that controlled access barriers have on nonresponse error and data quality.