The U.S. Census Bureau conducted the 2019 Census Test this summer to study the operational effects on self-response of including a citizenship question on the 2020 Census questionnaire.  The test was a nationally representative randomized field experiment designed to inform hiring levels for the Nonresponse Followup operation, as well as offer insight for the integrated partnership and communication campaign.

The major finding of the test was that there was no difference in self-response rates between forms with and forms without a citizenship question.[1]  The preliminary analysis suggests that in some areas and for some subgroups, there were lower self-response rates for the test form with the citizenship question than the test form without the citizenship question.  These differences were small.  Although the 2020 Census will not include a citizenship question, these test results indicate that the inclusion of a citizenship question would not have affected hiring for the Nonresponse Followup operation, which is designed to collect responses from households that do not self-respond.

The 2019 Census Test was sent to approximately 480,000 housing unit addresses across the country. These addresses were randomly assigned to one of two treatments and residents were asked to respond to the 2019 Census Test form. One test form included the question on citizenship; the other test form did not. Both test forms included additional questions that are planned for the 2020 Census: age, sex, Hispanic origin, race, relationship, and homeownership status.

Similar to the way the 2020 Census will be conducted, sampled addresses were sent up to five mailings, beginning June 13, requesting participation in the test by responding to the survey. Census Day for the test was July 1. Most sampled addresses were sent mailings that encouraged online response and received a paper test form in the fourth mailing, referred to as the Internet First contact strategy. The remainder of sampled addresses were sent a paper test form in the first and fourth mailings and could also respond online.[2] This is referred to as the Internet Choice contact strategy and is sent to areas with households that are historically less likely to self-respond online or have limited internet access. Households could respond in English or Spanish online, by mail, or by phone. Help was also available in English, Spanish, and nine other languages by calling the Census Bureau’s Telephone Questionnaire Assistance hotline. In the 2020 Census, online and phone response options will be available in 12 non-English languages. Since this was an operational test, there was no follow-up with households that did not self-respond by the end of the test on August 15.

As stated previously, the major finding of the test was that there was no difference in self-response rates between forms with a citizenship question and without a citizenship question. There were also no differences in self-response rates at specific points in time associated with the mailings. See Figure 1.

Self-response rates from this test were similar to other mid-decade tests, including the 2018 Census Test conducted in Providence County, RI, which had a 52.3 percent self-response rate. Note that the environment in which the 2019 Census Test was conducted will differ from the 2020 Census environment in terms of advertising and media attention. Thus, the 2020 Census self-response rates are expected to be higher than those associated with the 2019 Census Test.

Self-response rates were also examined by response mode: mail and internet. While overall self-response rate differences were not statistically significant, mail self-response rates were different between the treatments. There were no differences between forms with and without a citizenship question for internet self-response rates (which includes telephone responses). The mail self-response rate without the citizenship question was 0.3 percent higher than with the citizenship question (16.0 percent vs. 15.7 percent).

There were no differences in age or race of respondents between the treatments. There was a statistically significant difference in the proportion of respondents who identified as Hispanic between the forms (0.3 percent difference).

Although the 2020 Census will not include a citizenship question, results from this test may help inform operational decisions for future censuses and surveys. Had the citizenship question been included in the 2020 Census, it would not have affected staffing needs for the Nonresponse Followup operation, which is designed to collect responses from households that do not self-respond. Current plans for staffing for Nonresponse Followup would have sufficiently accounted for subgroup differences seen in this test. Note that this test did not include the Nonresponse Followup operation, so we are not able to measure the impact of a citizenship question for the completeness and accuracy of 2020 Census overall. Furthermore, the results of the 2019 Census Test will not trigger a major change in our communications campaign strategy, which was built on prior research that indicates that self-response differs across communities, and that some populations may be fearful about participating in the census.[3]

A full report of findings from the 2019 Census Test will be published when available.

[1] All of the comparisons in this blog have been tested for statistical significance at the 90 percent confidence level. For ease in presentation, “no difference” and “no statistically significant difference” are used interchangeably in this blog. In cases where the difference in estimates was statistically significant, the blog states that the estimates were different.

[2] Approximately 78 percent of the sampled addresses received the internet first contact strategy and 22 percent of sampled addresses received paper test forms in the first and fourth mailings.  These percentages are similar to the plans for the 2020 Census.

[3] The 2020 Census amount of funding allocated for the communication program is approximately $500 million.