Sullivan (2020) has done a great service to the statistical community with her tour-de-force review of the importance of the U.S. decennial census over its long and storied history from 1790 to 2020. She points out the threats to its integrity that have arisen in the past and that, arguably, are crescendoing today. She asserts that political checks and balances will not necessarily protect the census (or other federal statistical programs) and that (p. 1) “the professional integrity of statisticians is the best defense of the census.”
I broadly agree with her conclusions but am perhaps more pessimistic about the ability of federal statistics and statisticians to maintain independence from “political and other undue external influence” (National Research Council, 2017, p. 3). Can, in fact, statisticians defend federal statistics in general and the census in particular at this fraught time in our history, when the integrity of the federal government and the civil service is under attack and public trust in government is at a low ebb?
Over the past 50 years, substantial protections have been put in place for the independence and objectivity of federal statistical agencies and programs. These protections take the form of statistical policy directives issued by the U.S. Office of Management and Budget (OMB), legislation accompanied by OMB guidance for implementation, and statements of principles and practices for federal statistical agencies by the United Nations and the U.S. National Academies of Sciences, Engineering, and Medicine’s Committee on National Statistics (CNSTAT).
It is to the credit of the chief statistician’s office in OMB, working with other federal statisticians and with the support of statisticians outside the government, that these protections exist. Even more remarkable, the protections were achieved while the statistical system faced declining resources and staff (most egregiously, for the past several decades, the chief statistician’s office has had only half a dozen permanent staff).
Many of the specific protections were put in place in response to threats to federal statistics (locking the barn door after the horse had bolted in the hopes of preventing further escapes), while some were enacted proactively. Milestones in the construction of the current edifice for protecting statistical agencies and programs, such as the decennial census, are summarized in Box 1.
BOX 1 – Milestones in the Protection of U.S. Statistics
Statistical Policy Directive No. 3 This directive was first issued in the 1970s (and strengthened in 1985) in response to the efforts of the Nixon Administration to muzzle staff of the Bureau of Labor Statistics (BLS) because the president did not like their characterization of declines in unemployment as statistically insignificant. It identifies 38 “principal economic indicators” for which release schedules are set a year in advance and elaborate procedures are used to ensure that the indicators do not leak and are released in a policy-neutral way by the responsible statistical agency, separate from interpretation by policy officials. (President Trump has violated this directive twice in his administration to date—see, e.g., https://www.marketwatch.com/story/trumps-jobs-report-tweet-in-advance-of-release-appears-to-have-violated-federal-rules-2018-06-01).
Statistical Policy Directive No. 2 This directive provides standards and guidelines for statistical surveys, such as the need to analyze nonresponse bias and other elements of high-quality surveys covering design through data dissemination. The directive originated in the 1970s as two separate documents that were combined, updated, and reissued in 2006.
Paperwork Reduction Act Enacted originally in 1980 and reauthorized and expanded in 1986 and again in 1995, the act’s principal goal was to reduce the burden of filling out federal forms by businesses and individuals. It also codified important elements of the federal statistical system, including the chief statistician’s position and the Interagency Council on Statistical Policy (ICSP), chaired by the chief statistician and including heads of principal statistical agencies (e.g., BLS, Census Bureau). (The ICSP was recently expanded in response to the Foundations for Evidence-Based Policymaking Act of 2019—see below.)
Principles and Practices for a Federal Statistical Agency (P&P) In response to congressional questions about what constitutes a statistical agency, CNSTAT issued the first edition of its now well-known “purple book” in 1992. The first edition elaborated three bedrock principles and 11 practices. Beginning in 2001, CNSTAT has updated and reissued P&P every 4 years to inform political appointees to a new or second-term administration and to serve as a resource for statistical agencies and their staffs. The 6th edition, released in 2017, presents four principles: relevance to policy issues, credibility among data users, trust among data providers, and independence from political and other undue external influence. It also presents 13 practices, including professional advancement of staff and openness about sources and limitations of data. While having no official standing, P&P is widely recognized in Congress, statistical agencies, and the statistics profession as an important stake in the ground for the integrity of federal statistics. (See the endorsement by the Board of Directors of the American Statistical Association of the 6th edition; https://www.amstat.org/asa/files/pdfs/pressreleases/2017-Principles-Practices-Fed-Stat-Agency.pdf). OMB referenced P&P in the background information accompanying the issuance of Statistical Policy Directives Nos. 1 and 4—see below.)
Fundamental Principles of Official Statistics In response to the breakup of the Soviet Union and the desire on the part of statistical agencies in Western Europe to assist newly independent states in Eastern Europe to establish objective, credible statistical programs, the Conference of European Statisticians in 1991 drafted a resolution of fundamental principles. This resolution was adopted by the UN Statistical Commission in 1994 and by the UN General Assembly with a revised preamble in 2014. It articulates 10 principles, which support the value of official statistics and the need for their production and release to be objective and independent.
Confidential Information Protection and Statistical Efficiency Act (CIPSEA) Prior to CIPSEA’s passage in 2002, few statistical agencies had strong legal protections for preserving the confidentiality of responses to their surveys and censuses. This law represented the culmination of several decades of effort to pass confidentiality protection with teeth for the federal statistical system as a whole. (CIPSEA was strengthened by incorporation in the 2019 Foundations of Evidence-Based Policymaking Act—see below.)
Statistical Policy Directive No. 4 This directive was issued in 2008 in response to an incident in 2005 when a political appointee in the U.S. Department of Justice wanted the Bureau of Justice Statistics (BJS) to rewrite a press release accompanying the release of results from a regular survey of police traffic stops. The head of BJS refused to do this and was removed from his post. The directive requires statistical agencies to issue a schedule of release dates for their statistics and states that it is the sole province of the agency to draft press releases or other announcements of newly issued statistics.
Presidential Appointment Efficiency and Streamlining Act of 2011 This act provided that the director of the Census Bureau remain a presidential appointee with Senate confirmation but have a fixed 5-year term (with one renewal permitted) for terms beginning on January 1 of years ending in 2 and 7. Previously, the Census Bureau director served at the pleasure of the president. Other principal statistical agency heads lack fixed terms, with the exception of the commissioner of BLS.
Statistical Policy Directive No. 1 This directive was issued in 2014 to state the fundamental responsibilities of principal statistical agencies and recognized statistical units. It was labeled no. 1 because it was intended to provide a common foundation for core statistical agency functions. The directive lists four responsibilities of federal statistical agencies and charges departments to support the agencies in carrying them out: (1) produce and disseminate relevant and timely information; (2) conduct credible and accurate statistical activities; (3) conduct objective statistical activities; and (4) protect the trust of information providers by ensuring the confidentiality and exclusive statistical use of their responses.
Foundations of Evidence-Based Policymaking Act of 2019 This act incorporates Statistical Policy Directive No. 1 and CIPSEA and establishes heads of principal statistical agencies as the statistical officials for their department. Departments without a statistical agency are required to appoint statistical officials, who sit on the ICSP along with the principal statistical agency heads. Implementation of the act is in progress, but it promises to inculcate an appropriate culture for federal statistics throughout the entire executive branch.
Note. National Academies of Sciences, Engineering, and Medicine (2017, Appendix A), provides background on most of the protections referenced above.
All of these laws, directives, and statements of fundamental principles have helped federal statistical agencies build and protect high-quality staffs and a culture of objectivity, independence, and professionalism. Yet their effectiveness depends on everyone in the executive branch buying into this culture. Most often, departmental officials do protect their statistical agencies and programs, but not always. And it is difficult for a statistical agency to fight back effectively when a department is bound and determined to muzzle the agency or misrepresent the statistics.
A recent very telling example was the success of the U.S. Department of Agriculture (USDA) in diminishing one of USDA’s two statistical agencies—the Economic Research Service (ERS). Not liking some of the work that ERS has traditionally done, such as estimating the likely effects of trade policies and, more recently, of climate change on agricultural production and income, the department first assigned the administrator to another USDA agency. It then announced and has carried through its intention to move ERS (and another USDA research agency) to Kansas City, ostensibly to be closer to the farm community, when ERS’s policy clientele is mostly Washington, DC–based. The move resulted in many staff members leaving, also clearly something desired by the department to minimize ERS’s ability to perform important analyses. Supporters of ERS enlisted Congress to fight this move, but to no avail.1 Indeed, the decentralization of U.S. statistical programs among over a dozen principal statistical agencies (many research, policy analysis, and operational agencies also have sizeable statistical programs) means that the removal of an agency head can lead to little or no media attention or public outcry.
The success of the 2020 census, as measured by a low net undercount and, more importantly, a low differential net undercount between Whites and minorities, hangs in the balance given heightened public distrust, particularly among the immigrant community. The controversy over whether a question on citizenship would be added to this census only exacerbated distrust, which may not diminish sufficiently for a high-quality census. While the Census Bureau never fails to follow up on addresses where no one in the household self-responds (by Internet, telephone, or mail), research conducted by Census Bureau staff indicates that the quality of the response deteriorates the more effort the Census Bureau needs to expend in follow-up visits by enumerators. Self-responses have very few duplicates or omissions; in contrast, responses obtained by enumerators, particularly if multiple visits are required and if the enumerator must turn to a proxy (e.g., a neighbor or landlord), are much less accurate in both respects.
Should there be indications of a large differential net undercount, there will undoubtedly be lawsuits by states and cities demanding statistical adjustment of the data to use in redistricting congressional seats, no matter how long it takes to prepare reliable coverage estimates and adjusted census figures. While such suits are playing out, there is the risk that political appointees may try to shut down the work of Census Bureau analysts prematurely or in other ways interfere with the process. The statistical community will need to be vigilant and prepared to vigorously combat such efforts. And Census Bureau executives will need to be prepared to resign if political interference is blatant and irremediable.
In the worst case, support by professional statisticians inside and outside the government may not be enough to protect the census, the Census Bureau, and the federal statistical system as a whole, just as it was not enough to protect the ERS. To maintain and build on past gains in fostering independence and objectivity of federal statistics, statisticians, like every U.S. citizen, will need to become educated politically with respect to how the U.S. federal system of government works (and doesn’t work) and support needed reforms, such as nonpartisan redistricting, clear and uniform voting rights and procedures, and other actions that can help everyone in our system adhere to the rules of fair play and equal justice. Statisticians can help by analyzing the properties of various proposed reforms (e.g., multiple-choice voting schemes), by commenting loudly and often on the need for the federal government to provide such public goods as weather data, scientific research, and federal statistics in an as objective and professional a manner as possible, and by making their voices heard through representations to Congress and, in their individual capacity, in the voting booth.
Principles and practices, whether codified in law, directives, or other official ways, are only as strong as the men and women who are in charge of upholding them. We are fortunate in the United States to have dedicated professional civil servants in charge of the census and other federal statistical programs. We must give them our wholehearted support, not only by speaking out on their behalf, but also by working in other ways to make our political system worthy of public trust.
National Academies of Sciences, Engineering, and Medicine. (2017). Principles and Practices for a Federal Statistical Agency. Sixth Edition. Committee on National Statistics, C.F. Citro, ed. Washington, DC: National Academies Press. https://doi.org/10.17226/24810
Sullivan, T. A. (2020). Coming to our census: How social statistics underpin our democracy (and republic). Harvard Data Science Review, 2(1).
This article is © 2020 by C.F. Citro. The article is licensed under a Creative Commons Attribution (CC BY 4.0) International license (https://creativecommons.org/licenses/by/4.0/legalcode), except where otherwise indicated with respect to particular material included in the article. The article should be attributed to the authors identified above.