Federal Statistical Agencies Struggle to Maintain Their Vital Role in the U.S. Data Infrastructure—Why and Whence?

Constance F. Citro, Ph.D., directed the Committee on National Statistics, National Academies of Sciences, Engineering, and Medicine, from 2004 to 2017. She is currently Senior Scholar with CNSTAT. Views expressed are her own.

In recent months, leaders from the federal statistical system have offered their thoughts on the state of the data infrastructure in the United States. Katherine Wallman, former Chief Statistician of the United States, aptly noted there is frequent discussion about the nation’s transportation infrastructure—roads, bridges, and rails—but less focus on the state of our nation’s statistical or data infrastructure. In this article, I offer my perspective about the struggle to maintain the vital role of federal statistical agencies in the United States data infrastructure.

Federal statistics are vital High-quality, objective statistics produced by federal statistical agencies are vital inputs for public and private decision making and informing the public (see National Research Council, 2017). The decennial census (now under threat from the novel coronavirus) not only:

• undergirds our democracy—most notably through its role in peaceful reallocation of power in the U.S. House of Representatives; and
• facilitates our federal system of government—through allocation of billions of federal funds to states and localities and other government uses; but also
• serves as the bedrock for many ongoing statistical series—providing sampling frames and statistical controls for surveys that produce the unemployment rate, the poverty rate, and other key information.

As noted by Steve Landefeld, economic statistics, including the gross domestic product (GDP), and other key national, industry, regional, and international statistics, are “hard-wired—by law and practice” into all manner of economic decision making. Statistics on educational attainment, crime victimization, agricultural output, the science and engineering work force, health conditions, and many other topics that inform policy and the public are the work of dedicated civil servants at the principal statistical agencies.

Agencies that produce federal statistics struggle Federal statistical agencies have a long track record of important innovations, including statistical sampling methods for population surveys, the National Income and Product Accounts, physical exams as part of health and nutrition surveys, and many others. Yet at least since the 1980s, they have been struggling to garner the recognition and tangible support—budget, staffing, and ability to coordinate and integrate their work to produce more useful, accurate, and timely statistics—they merit and the nation needs:

• budgets for many federal statistical agencies are flat or declining in real terms, and only a few agencies have budgets that are more than rounding error in their department;
• the chief statistician’s office in the U.S. Office of Management and Budget (OMB), which coordinates federal statistics, does a yeoman job with de minimis resources, but, inevitably, cannot play as robust a role as needed (the office has six staff members at full strength, plus short-term detailees from other agencies, and no budget to fund R&D that could benefit multiple agencies);
• staffing ceilings constrain agencies, leading in some instances to overreliance on outside contractors;
• laws, regulations, and differences in agency cultures make it difficult to innovate, particularly when innovation requires data sharing and other forms of interagency collaboration (e.g., the Bureau of Economic Analysis (BEA), Bureau of Labor Statistics (BLS), and Census Bureau may not share business information that comes from tax records);
• state laws and bureaucratic hurdles in underfunded state agencies make it difficult for federal agencies to acquire administrative data for statistical purposes even when the federal government pays program costs (e.g., SNAP benefits);
• federal statistical agencies are often “layered” in their departments (e.g., the National Center for Health Statistics in the Centers for Disease Control and Prevention, U.S. Department of Health and Human Services), making it difficult to raise their profile and gain needed resources;
• department-specific standards and “looks” constrain agencies in such matters as website development and format, making it more difficult for data users to find information;
• cabinet departments and White House staff sometimes fail to scrupulously respect the objectivity and independence that federal statistical agencies require for their work to be trusted by all parties
• growing distrust of government affects even the most nonpartisan, “good government” agencies;
• statistical agencies find it difficult to compete with the private sector for employees (e.g., those with qualifications in such fields as data science and artificial intelligence); and
• declining response rates threaten the quality of agencies’ survey-based statistical programs, while acquiring administrative and private sector information that might improve on survey responses in some areas (e.g., income statistics—see Citro, 2014) presents many challenges.

The agencies continue to produce their flagship and other statistical series and so may look as if they are in fine shape. In fact, they need substantial inputs: funding, staffing, elevation of the importance of their mission, and legislation to remedy long-standing problems. Two priorities for legislation in my view are:

1.    data sharing—why shouldn’t federal statistical agencies be able to share administrative records and other confidential data, with appropriate safeguards, as easily as in countries, which, by accidents of history, have a more centralized statistical system (e.g., Canada)?

2.    confidentiality protection—in the internet age, it is not possible to guard against any and all reidentification of individual respondents without greatly curtailing the availability of data for public use and thereby undercutting the fundamental purpose for a federal statistical agency; legislation should place the onus for protecting confidentiality on the user in addition to the agency.

Finally, agencies need to find ways to build a virtual integrated federal statistical system that works more effectively toward providing high quality, timely, and relevant statistics to the nation. They already collaborate in many ways, yet there are important areas where more effective collaboration is essential: 

• expanding the range of definitions and concepts for which there are common standards;
• dovetailing one agency’s programs with other agencies’ related programs so they are not telling different stories (except where needed—e.g., the household and establishment surveys of unemployment and payroll provide complementary information on the labor market); and
• heightening the agencies’ collective visibility so that the public and policymakers recognize them collectively as the source of “gold standard” statistics and do not assume the private sector can do it all .

My point about the need for an integrated system hints at a prime reason why federal statistical agencies are so challenged, so layered, and so often taken for granted—namely, that as a nation we recognized the value of federal statistics but developed them in a manner that led to many separate agencies and not to a Statistics USA. It is unlikely, for many reasons, that consolidation of agencies could or should occur in the United States (although the idea of integrating Census, BEA, and BLS, which the late BLS Commissioner Janet Norwood supported, is currently on the table). But new legal developments may make it possible to move toward virtual integration in many important areas of federal statistics.

But first, why and how did we get to such a decentralized federal statistical system? The collection of statistics by the federal government is part of our founding: Article 1, Section 2, of the U.S. Constitution mandates a decennial census. From that beginning and as the federal government increasingly assumed responsibilities heretofore left exclusively to states and localities, it chose the path of setting up a new cabinet department or agency, which typically included an office to provide statistics to inform policymakers and the public (see Citro, 2016). Some milestones in the development of federal statistics include:

• pre-Civil War—State and Treasury Department offices set up to collect statistics on foreign and domestic trade;
• 1863—Division of Statistics created in the new U.S. Department of Agriculture;
• 1867—Office of Education Statistics established in a new Department of Education (the office soldiered on even though the department was disbanded in 1869, not to be reestablished until 1980);
• 1884—Bureau of Labor created to collect statistics on the growing industrial labor force;
• 1902—permanent Census Office established (previously, an office was set up for each census and then disbanded);
• 1930s—position of chief statistician of the United States established at the Bureau of the Budget (now OMB) and codified in the Paperwork Reduction Act of 1980;
• 1977—Energy Information Administration set up in the new Department of Energy;
• 1991—Bureau of Transportation Statistics established in the Department of Transportation to focus on transportation flows across modes (air, highway, railroad, water);
• 2002—Confidential Information Protection and Statistical Efficiency Act of 2002 (CIPSEA) enacted, which strengthened confidentiality protection for statistical programs across the federal government; and
• 2018—Foundations for Evidence-Based Policymaking Act enacted, which provides for a chief data officer and statistical official in every cabinet department and independent agency.

Looking to the Future At present there are 13 principal federal statistical agencies, which have statistical activities as their principal mission, 100 or more statistical programs throughout the government, and, now, a statistical official in every cabinet department (the head of a principal statistical agency serves this function in departments with one or more statistical agencies). Statistical agency heads and officials sit on an expanded Interagency Council on Statistical Policy (ICSP), chaired by the chief statistician in OMB. Enhanced interagency communication and leadership through the expanded ICSP, together with other provisions of the 2018 legislation, offers hope in at least three respects:

• elevating the mission of producing high quality, relevant, and timely federal statistics as a critically important public good;
• creating new arenas and advocates for federal statistical programs; and
• building on current interagency collaborations to achieve a more unified federal statistical system in fact if not law.

Reactions? Comments? I have made a number of assertions and suggestions about the history of our federal statistical agencies, the challenges they face today, and some of the steps that I believe would help strengthen their ability to respond with the statistics we need for the future. The ability of the agencies to be innovative, maintain the confidence of policy makers and the public, and produce “gold standard” statistics on a timely basis is even more vital in the face of the societal shocks from the novel coronavirus and likewise the dislocations of people from natural disasters exacerbated by climate change. We—collectively and individually—need to give the agencies all the help we can.

I ask readers to state their agreement/disagreement with one or more of my points and the basis for same—and to ask me to further justify my take where it is not clear. I particularly hope that readers will offer ways in which we can create a virtually integrated federal statistical system from our dispersed statistical agencies, whose staffs work exceptionally hard and devotedly to produce the bedrock information the nation requires.

I hope you will comment below and continue the conversation in the CountOnStats LinkedIn group!