Protecting the Census
A DISCUSSION OFProtecting the Accuracy of the 2020 Census
Read Responses From
Census Day is right around the corner. The US Government Accountability Office agrees with the challenges facing the upcoming enumeration that Constance F. Citro identified in “Protecting the Accuracy of the 2020 Census” (Issues, Summer 2019). A complete count is inherently difficult given the size and diversity of the country, and the stakes couldn’t be higher: data from the decennial census, which is mandated by the Constitution, are used for such essential purposes as apportioning and redistricting the House of Representatives and allocating hundreds of billions of dollars each year in federal financial assistance.
As Citro points out, previous enumerations have had their risks and challenges, and 2020 is no exception. Over the past decade, our work has recommended steps the Census Bureau can take to ensure a more cost-effective and secure count of the nation’s population, and in February 2017 we added the 2020 Decennial Census to our list of high-risk government programs. This is because, among other things:
- The Bureau is using innovations that are not expected to be fully tested before being used in 2020 Census operations. These innovations, which include allowing the public to respond using the internet, show promise for controlling costs. But they also introduce new risks because, in part, they have not been used extensively, if at all, in earlier decennials.
- The Bureau faces challenges in implementing information technology (IT) systems. In July 2019, we reported that the Bureau was at risk of not meeting near-term IT system development and testing schedule milestones. These schedule-management challenges may compress the time available for remaining system development and testing, and increase the risk that systems will not function as intended.
- The Bureau faces significant cybersecurity risks to its systems and data. For example, as of the end of May 2019, the Bureau had over 330 corrective actions from its security assessments that needed to be addressed, including 217 that were considered “high risk” or “very high risk.”
- The Bureau is seeking to control the cost of the census, which has been escalating with each decade. According to the Bureau, the 2010 Census cost about $12.3 billion (in constant 2020 dollars), while the 2020 Census is estimated to cost approximately $15.6 billion.
Continued management attention and oversight will be vital for ensuring that risks are managed, preparations stay on track, and the Bureau is held accountable for implementing the enumeration as planned. As of July 2019, the Government Accountability Office has made 107 recommendations to help address these risks and other concerns, 32 of which have not been fully implemented. To ensure a high-quality, cost-effective, and secure count of the population, it is important that the Census Bureau continue to address these recommendations.
Gene L. Dodaro
Comptroller General of the United States
As Constance Citro ably catalogued, every decennial census has its challenges and controversies. The 2020 Census, I believe, faces a set of unprecedented challenges that collectively could create a perfect storm and threaten a successful census—that is, one that counts all communities equally well.
To be sure, there have been noteworthy advancements in census methods and operations, as well as measurable improvement in census accuracy, over time. Nevertheless, disproportionate undercounting of blacks, Latinos, American Indians living on reservations, renters, and children under age five persists, and recent censuses have overcounted non-Hispanic whites, as well as homeowners and older Americans in some race and gender cohorts. In short, the census is not yet an equal opportunity enumeration.
For the 2020 Census, the consequences of funding shortfalls and test cancellations throughout the planning cycle could fall hardest on activities specifically designed to reduce this disproportionate undercounting, such as language-appropriate promotion, community-based assistance centers, and a sufficiently large army of census takers to follow-up with reluctant and, yes, fearful households. Unfortunately, the current administration’s effort to add an untested citizenship question to the 2020 Census exacerbated concerns among immigrants that their census responses would be used to harm them or their families. Despite the US Supreme Court decision effectively quashing the citizenship question for 2020, President Trump’s July 11, 2019, executive order directing the Census Bureau to produce census data on citizenship and legal (immigration) status using administrative records continues to raise concerns about the administration’s motives. I am not surprised by the skepticism: the terms “census” and “immigration enforcement” should never appear in the same official document.The census is only as good as the public’s willingness to participate. If public confidence in the Bureau’s statistical mission and the motives for producing data falls, the agency’s ability to fulfill its mission could be in jeopardy.
Funding shortfalls this decade also curtailed sufficient investment in early research and testing that could have fostered the “cumulative learning for evidence-based planning decisions” that Citro highlighted. She noted that there have been “sharp increases in costs per housing unit” over the past five censuses. As lawmakers work to spend federal tax dollars prudently, the Census Bureau certainly must do its part to keep costs in check, even when faced with a growing, diversifying population. Technology undoubtedly offers opportunities to improve cost-effectiveness and productivity; administrative and commercial data can streamline address verification and help fill in some missing data, although the jury is still out on whether administrative records can replace efforts to count entire households directly. Still, I do not think we can cut corners or settle for uneven results when the very strength of our democracy’s foundation is at stake. The digital divide could prevent a not-insignificant number of households from answering the census online. Administrative datasets often cover harder-to-enumerate population groups less well and do not include information consistent with Bureau rules for determining where people should be counted. Overall, personal outreach from community leaders and “trusted messengers” will still be necessary to improve participation in historically undercounted communities.
Are Americans ready for a markedly different census—one that does not rely significantly on personal outreach, persuasion, and response? I’m not convinced. The census is the nation’s largest, most inclusive civic engagement exercise. It gives most residents an opportunity to participate directly in an activity that empowers them and their communities. Everyone is counted—regardless of age, citizenship or legal status, or prior incarceration—which means everyone counts. And that is a powerful message for a nation whose cohesiveness and common understanding of our unique place in the modern world feels increasingly fragile and even fractured.
Terri Ann Lowenthal
Census consultant and former staff director, US House of Representatives census oversight subcommittee, 1987–1994
Constance Citro provides an excellent short history of United States censuses, along with detailed information on the successes and shortcomings of the five most recent ones. Her information is consistent with what I learned and observed during my tenure at the Census Bureau (1977–79, 1983–90, 1996–2004). In my last position there, I oversaw research, methodology, and quality for all Bureau programs. In particular, I oversaw an immense evaluation program for the 1998 dress rehearsal mandated by Congress and a later evaluation of the 2000 census, and was heavily involved in the discussions regarding coverage adjustment of the 2000 census.
The census was instituted to inform political processes—reapportionment and redistricting. At its heart, this constitutional rationale remains. Additionally, through legislation, census counts have come to be used to provide accurate estimates regarding funding needed to meet numerous societal needs. During the past five censuses, the political purpose of the census operation has occasioned conflict with efficient information-gathering.
I share many of Citro’s concerns. Conducting a mandatory census of the US population is an immense peacetime operation—with critical time constraints set in law. The relative infrequency of the decennial census leads to loss of institutional memory. In contrast, to meet mandated time deadlines, the Census Bureau has often been forced to return to past practices that were no longer technologically optimal.
Funding is always an issue, particularly as legislators never seem to understand that new procedures used in a large operation must be tested years before the operation. Familiar technologies have been adopted only tardily. In the 2010 census, even smartphones were used only hesitantly. To my knowledge, nobody has ever created a gold standard Master Address File. This is crucial, given that it is the dwelling that shelters the people being counted.
My greatest concern for the 2020 census is the potential for high levels of nonresponse and for coverage error. It is not clear that procedures are in place to deal with either under-or over-coverage. Both the dialogue concerning a citizenship question and the general fear of government have made it difficult to effect data collection and the outreach procedures needed to encourage response.
Internet collection will begin as a letter sent to each housing unit on the Master Address File that had been given an identification number (to access a census form). But individuals will be allowed to obtain and complete a census form without having been given an identification number. The Census Bureau will then match those forms to their housing units. I fear that this could lead to yet another type of over-coverage (in addition to what regularly happens with college students and children in joint custody families, among other cases)—and funding will not be available to correct this over-count.
Cynthia Z. F. Clark
Council of Professional Associations on Federal Statistics