The 2018 Census Barriers, Attitudes and Motivators Study survey gave the U.S. Census Bureau a unique opportunity to explore whether early intention to respond to the 2020 Census aligned with actual response.

Before the 2020 Census, there were several research projects focused on public attitudes and perceptions about the census. One of the questions the Census Bureau asked in the 2018 CBAMS was “If the census were held today, how likely would you be to fill out the census?”

Most (67%) respondents said they were “extremely” or “very likely” to respond to the census. Previous analyses suggested that actual participation may be lower than that reported in opinion surveys. Once the 2020 Census results were available for analysis, we studied how intended response reported in the 2018 CBAMS survey compared to actual 2020 Census participation.

Households could respond to the 2020 Census online, by telephone or by mail. Throughout the country, areas were assigned one of two contact strategies: Internet First (invitation letter to respond online on the first contact) or Internet Choice (paper questionnaire and invitation to respond online sent on the first contact).

Households could also respond during the Nonresponse Followup (NRFU) operation or could have been enumerated using administrative records.

Of the 50,000 households in the CBAMS sample, a weighted 54.5% self-responded to the 2020 Census.

In this section, we focus on reported intent to participate in the census overall and how intent compared between Internet Choice (Figure 1) and Internet First (Figure 2) areas.

Table 1 shows the output for Internet Choice areas. There were 3,388 weighted 2018 CBAMS responses in the Internet Choice contact strategy (census responses). Households that had reported in 2018 they were “extremely likely” to respond to the 2020 Census self-responded at a higher rate (77.4%) than those who had stated they were “not at all likely” to respond (65.7%).Those who said they were “not too likely” to respond, still self-responded to the 2020 Census at a higher rate (76.0%) than the 2020 Census national self-response rate (66.3%).

Households that did not answer this question but responded to other questions in the CBAMS, either responded during the census self-response stage or were determined to be vacant.

Table 2 shows the outcomes for Internet First areas. There were 12,359 weighted households in Internet First areas. Households that had reported they were “extremely likely” to respond, self-responded to the 2020 Census at a higher rate (86.0%) than those who had stated that they were “not at all likely” to respond (74.2%).

Our research highlights the importance of public opinion surveys like CBAMS to study participation in the census. Our analysis is part of a report showing that respondents who said they were “extremely likely” and “very likely” to participate in the 2020 Census followed through and their household responded to the 2020 Census. Our CBAMS respondents voluntarily answered the survey, indicating respondents may be more willing to participate in other surveys. Interestingly, we also found the CBAMS survey can help identify households that are unlikely to participate in the census and that changed their behavior to actually participate in the census. Further research is necessary about self-response bias and factors explaining behavior changes for survey respondents who expressed they were less likely to participate in the census.