Use of Exact Matching to Examine Media’s Effect on Intended Behavior the Case of the Addition of the 2020 Census Citizenship Question

Abstract

Researchers have argued that events and surrounding media coverage shape attitudes and intended behaviors on topics related to the events (Langer et al. 1992; Hoekstra 2003). Such research relies on analyzing attitudinal trends using rolling cross-section designs or across data collections, but little published research measures whether an event covered by the media could shape a behavioral intention salient to that event within one data collection period. In 2018 the Census Bureau conducted the Census Barriers, Attitudes, and Motivators Study (CBAMS) survey. Five weeks into the survey’s field period, the Secretary of Commerce directed the Census Bureau to add a citizenship question to the 2020 Census, resulting in media coverage. The CBAMS survey asked likelihood to respond to the decennial census, which may have been affected by the external news environment. Treating this as a “natural experiment,” we match pre- and post-citizenship question-announcement respondents and report the results of a multivariate model predicting intent to respond to the census. We also examine differences between subgroups and their complements pre- and post-announcement. Although the citizenship question was not included on the 2020 Census, the odds that those responding after the citizenship question announcement were “extremely” or “very likely” to respond to the census were 20 percent lower than the odds of those responding before. Future research should examine the permanency of changes on intended behaviors, especially in cases where news coverage focused on an outcome—such as the addition of the citizenship question—that did not occur.