Is Autonomy Possible and Is It a Good Thing?

Abstract

Recently, Citro et al. published an article focusing on the autonomy, or lack of same, of the 13 principle statistical agencies of the United States. The authors are to be congratulated for raising an important topic—the concept of autonomy. Among their conclusions are: (a) existing autonomy protections are inadequate, (b) a lack of professional autonomy unduly exposes the principal federal statistical agencies to efforts to undermine the objectivity of their products and (c) agencies cannot completely rebuff these efforts. Their main recommendations are that the role of the Chief Statistician be strengthened and new statutory autonomy protections be legislated. Here, we consider the meaning of autonomy for a federal agency in general and for federal statistical agencies in particular. Additionally, we consider the benefits and limitations of autonomy for federal statistical agencies. We note that while additional legislation is useful to produce required autonomy, a powerful tool—and one which is possibly more readily available—is effective leadership. Finally, we suggest that the process used to select the leaders of the statistical system needs to be fundamentally changed.

KEYWORDS:

• Autonomy
• Federal statistical agency
• Government statistics
• Independence
• Leadership
• Official statistics
• Trust

This article refers to:

What Protects the Autonomy of the Federal Statistical Agencies? An Assessment of the Procedures in Place to Protect the Independence and Objectivity of Official U.S. Statistics

1 Types of Autonomy

This is not the place for a detailed description on the complex rules and operating procedures of the federal government. We note, however, that the agencies of the federal government do not, of course, operate as a network of independent entities. In general, funding, authority and mission are derived, inter alia, from many sources: the constitution, congressional actions, executive orders and judicial actions. Federal agencies—including federal statistical agencies—operate under prescriptions as to what they are allowed to do and how they are allowed to do it. (A broad but still useful distinction between the autonomy¹ of federal and private sector entities is that in the federal government agencies are permitted to do only what has been prescribed by higher authority and in the private sector an entity can do whatever it wishes provided those activities have not been proscribed by law.) The discussion on autonomy is best framed by categorizing the activities of an agency into two broad areas: the “what” and the “how.” The “what” can be thought of as the mandate of the agency. Details of the procedures used by the agencies in executing their mandate are often thought of as the “how” and usually, but not always, are left to the discretion of the agency. With respect to federal statistical agencies, the US Office of Management and Budget (2014) has laid out the responsibilities of these agencies.

This is the most critical authority that statistical agencies can use in setting professional autonomy boundaries across the government. There are four responsibilities cited:

• Responsibility 1: Produce and disseminate relevant and timely information.

• Responsibility 2: Conduct credible and accurate statistical activities.

• Responsibility 3: Conduct objective statistical activities

• Responsibility 4: Protect the trust of information providers by ensuring the confidentiality and exclusive statistical use of their responses.

We note that these responsibilities contain a mixture of the “what” (should an agency conduct surveys on income inequality?) and the “how” (the responsibility to conduct the survey). The OMB Directive clearly assigns to the statistics agency the responsibility to employ sound statistical practices. Moreover, the preamble to the Directive invokes the principle that statistics agencies should make their “how” decisions solely on professional considerations. It is a legitimate and a necessary function of the political process, for example, to determine the amount of resources devoted to the question of income inequality. It is the responsibility of the agency to make the statistical decisions to carry out that mandate.

It is useful in this context then, to consider two types of autonomy: administrative and professional. Administrative activities are those which are common across all bureaucracies and are not specific to the expertise and mission of the particular agency. These include, inter alia, areas such as personnel management and procurement and contracting services. Traditionally, information processing services would have been included in this area but these services have increasingly become inextricably bound to the specific and unique mission of the agency. In a federal statistics agency there are mathematical, scientific skills and knowledge which are developed through education and training. These are the basis for agency judgments on questions such as sample size, significance of results and applicability of results. This authority over the details is what is commonly referred to as professional independence or professional autonomy. There is a spectrum across the government which describes the extent of autonomy under which agencies have been permitted to operate. At one end of the spectrum are areas such as national intelligence and Federal Reserve activities. In addition to professional independence over methods of collecting and analyzing data, intelligence agencies usually possess a degree of autonomy—administrative autonomy—over areas such as human resource and contracting systems which most other federal agencies do not possess.

This distinction between professional and administrative autonomy is also referenced in the Citro et al. article. The rationale for professional autonomy is to ensure that, to the extent possible, partisan political concerns do not play a role in decisions about national intelligence data or central banking matters (e.g., interest rate fluctuations). The ceding of professional autonomy to these agencies is meant to give credibility that the information they provide and the decisions they reach are developed only through professional considerations and not due to partisan political allegiances. Statistical agencies also in many cases enjoy a degree of professional autonomy for similar reasons. As with the examples just cited, for statistical data to be believed and therefore used, stakeholders must trust that the data were produced solely according to professional standards and without partisan political influence.

The boundaries between professional and administrative autonomy are often vague and these areas often overlap. There is often disagreement, then, within the layers of the government with respect to the authorities each possesses. In fact it is precisely in this area—defining the amount of autonomy left to the agency in carrying out the “what” that is most problematic. It is the struggle that Habermann and Louis (2020) referred to:

“There is no easy solution to the struggle between statisticians asserting professional independence and politicians who see their authority as having primacy”

For example, the US Census Bureau is authorized and funding is provided by the Congress to collect, analyze and publish information on the extent of poverty in the United States. Decisions on whether even to carry out such a data collection program are not left to the Census Bureau. Professional decisions on survey design and execution are usually left to the Bureau. However, if the Census Bureau wanted to, for example, double the size of the sample it would need authority beyond its own. The situation is more complicated in other major statistical programs. The Congress has vested authority for the decennial census, not in the Director of the Census Bureau (a political appointee) but in his political superior, the Secretary of Commerce. There is no clear, unambiguous distinction, then, between, professional independence and the mandate of an agency.

The lack of an easy solution to the independence struggle referred to above is complicated by the decentralized nature of the statistical system. Citro et al. (2023) acknowledge that there is little evidence that this struggle over professional independence has resulted in systemic bias in the federal statistical system. We agree with them, however, regarding the critical importance of perception. There are two reasons for this. The first, as we have mentioned, is generating trust that the statistics produced by the federal agencies are reliable, accurate and unbiased. Without this trust the statistics will not be believed and consequently not used. To generate this trust users have to believe that decisions about the production of the required information are guided entirely by professional considerations and not by any partisan political considerations. The second reason involves trust considerations by those who supply information to the federal agencies. Information about individuals is the property of those individuals. It is not the property of the government and the government has to request individuals to provide the information. As a condition for individuals to provide the information, statistical agencies often pledge that the individual information will remain confidential. To a person who values the confidentiality of their data then, trust in the statistical agency is a condition for the provision of individual statistical data.² A recent controversy may help illustrate the differences between the how and the what and in what way these may impact the trust of the public.

2 2020 Decennial Census Citizenship Issue

After the Trump Administration proposed adding a question to the 2020 Decennial Census asking the citizenship of each respondent, several organizations sued the Administration in an effort to stop the plan.³ It is our view that the administration made a permissible attempt to answer the question of how many citizens were in the country. This is the “what” we discussed above. The Census Bureau had developed an approach, the “how,” which would have answered the question (in the opinion of the litigants) without damaging the accuracy and reputation of the decennial census. The Administration rejected the how proposed by the Census Bureau for, in the view of the plaintiffs, a more costly and inferior approach. The Secretary of Commerce, head of the parent agency of the Census Bureau, asserted that he had the authority and responsibility over the how and not the Census Bureau. It was the view of many of the plaintiffs that the approach proposed by the Secretary was developed to support a partisan political position and not, as required by the Fundamental Principles, developed solely according to professional statistical considerations. By not respecting the professional autonomy of the Census Bureau the Administration created a situation which damaged the credibility of the Bureau, and, if allowed to stand, would also have served to damage the trust in and accuracy of the decennial census.⁴

The Supreme Court found for the plaintiffs and did not allow the citizenship question to be included in the Decennial Census. However, the reasoning behind the Court’s decision did not include considerations of professional versus administrative autonomy that we have discussed here. Instead the Court ruled, in a 5 to 4 decision, that the Administration violated certain provisions of the Administrative Procedure Act. The question of the need, if any, for a measure of professional independence on the part of the federal statistics agencies was not considered by the Supreme Court.

3 Benefits of the Political Process

As the previous example illustrates, partisan political decisions that replace decisions based solely on professional considerations can result in a loss of trust in the statistical system. Before leaving this area, however, we consider a major benefit of statistical agencies being a part of the normal political process. As we have indicated, trust in the unbiased character of the statistical system is critical. It is also vital that respondents and users have confidence that the statistical system is, in fact, focusing its efforts in the right places. There are many areas where the federal system could be devoting its efforts. These include economic, demographic, health and environment. Within each of these areas there are further considerations, for example:

• Should more effort be devoted understanding poverty or migration patterns?

• What is the relative priority of price statistics versus unemployment statistics?

The nation uses the political process to decide these questions, but if federal statistical agencies were, indeed, autonomous, these decisions would be made independently of elected officials. It is the political process which provides the authorization, justification and public support for statistical agencies to pursue their activities.

4 Fundamental Principles of Official Statistics

Previously, we noted that trust in official statistics rests on the practice and belief that production of appropriate statistical data rested on the foundation that these data would be produced based solely on professional and scientific standards. Internationally, the need for a set of guiding principles that could be used in preparing this foundation became apparent in the 1980s, when countries in Central Europe began to change form centrally planned to market-oriented economies. The Conference of European Statisticians adopted the Fundamental Principles of Official Statistics (FPOS) in 1991. It became clear that these principles had a much wider, global significance and that statistical systems throughout the world could benefit from their adoption. Finally, in recognition of their continued relevance and global nature, the United Nations General Assembly (2014) adopted the Fundamental Principles of Official Statistics on January 24, 2014. The words of the preamble in that adoption are germane to this question of autonomy and independence and we repeat them here:

The General Assembly,

Recalling recent resolutions1 of the General Assembly and the Economic and Social Council highlighting the fundamental importance of official statistics for the national and global development agenda,

Bearing in mind the critical role of high-quality official statistical information in analysis and informed policy decision-making in support of sustainable development, peace and security, as well as for mutual knowledge and trade among the States and peoples of an increasingly connected world, demanding openness and transparency,

Bearing in mind also that the essential trust of the public in the integrity of official statistical systems and confidence in statistics depend to a large extent on respect for the fundamental values and principles that are the basis of any society seeking to understand itself and respect the rights of its members, and in this context

that professional independence and accountability of statistical agencies are crucial,

Stressing that, in order to be effective, the fundamental values and principles that govern statistical work have to be guaranteed by legal and institutional frameworks and be respected at all political levels and by all stakeholders in national statistical systems,

Endorses the Fundamental Principles of Official Statistics set out below, as adopted by the Statistical Commission in 19942 and reaffirmed in 2013, and endorsed by the Economic and Social Council in its resolution 2013/21 of July 24, 2013:

It is significant that in its preamble the General Assembly of the United Nations affirmed the crucial importance of the professional independence and accountability of statistical agencies, and that the fundamental values and principles that govern statistical work have to be guaranteed by legal and institutional frameworks. Currently, guiding principles like the FPOS can be found throughout the world. Examples include the European Union Code of Statistical Practice published by Eurostat (2018), and in the United States the Office of Management and Budget Statistical Directives and the Fundamental Principles developed by the National Academy. With respect to the Fundamental Principles themselves, Principle 2 is most relevant to our concerns. According to Principle 2 it is quite clear that statistical agencies should make statistical decisions only according to strictly professional considerations. As we have discussed, however, the boundary between professional and nonprofessional statistical decisions is often murky and even with a strong legal foundation we can expect that there will be continuing disputes over the definition and enforcement of these boundaries. Moreover, we have seen that even strong legal frameworks can be ignored by partisan political leadership. Additional legislative attempts to define more precisely the boundaries between professional and administrative autonomy can and will be useful. However, history teaches that legislative changes—particularly in a decentralized system like the US—often entail a very lengthy process. While we support legislative efforts to more precisely define and protect professional independence, we note that there are some complimentary efforts which should be followed. In particular, we turn to an important subject which needs more consideration—the importance of leadership and the selection process to obtain those leaders.

5 The Importance of Leadership

Peter Drucker is said to have made the distinction between management, which is doing things right, and leadership, which is doing the right thing. While leadership encompasses many responsibilities, one of the most important “right things” for a head of a statistical agency is that of maintaining professional independence. Of the several paths that a statistical agency head can take to achieve independence, the first is the willingness to be fired. It is critical that an agency head take every opportunity to speak out publicly on the importance of professional independence. One way to accomplish this is to cultivate relations with the media. A strong media relationship can help the agency establish its brand with the public and aid in providing its information in the most useful form. Establishing these lines of communication can, admittedly, be difficult when the parent agency wants to control communication, as is often the case. This is particularly difficult when the head of the agency is a career official. A political appointee, particularly one who must be Senate confirmed, can and should make clear the conditions under which he or she would accept the position. Moreover, he/she must be prepared, if the circumstances warrant—as in the citizenship example—to resign and explain the reason. Certainly this is not a cure-all and does not solve the issue of parent agencies usurping areas of appropriate professional independence. However, it will increase the likelihood that a parent agency will be reluctant to do so.

In the same vein, the position of Chief Statistician at the Office of Management and Budget should be changed from a career official to a term-limited, political appointee with Senate confirmation. Because of the decentralized nature of the US statistical system, the most important function of the Chief Statistician at Office of Management and Budget is to lead in setting a vision for the statistical system and for overseeing the federal statistical system.

The Chief Statistician should be able to set professional standards for the federal system and publicly communicate them. Under the present selection system this is best accomplished if the Chief Statistician is a political appointee with Senate confirmation. However, the primary allegiance of the Chief Statistician should be to the production of unbiased statistical information and not to the political party of which they may be a member. We recognize the difficult and delicate position into which this puts both the head of an agency and the Chief Statistician. Nevertheless, it is critical that every Administration have prominent public voices that can speak for the need for quality, unbiased statistical information. However, events such as those described above and more recently in the United Kingdom by the Washington Post (2023) demonstrate the inadequacies of a partisan political selection process for leaders of what are intended to be a-political agencies. If unbiased statistical information is the lifeblood of democracy then the leadership selection process becomes critically important.

We note that current agency heads are a mix of career and political appointee, some of whom have term appointments, and that problems with respecting professional autonomy can be found in all these environments. Leadership is often the distinction between an agency which preserves professional autonomy and one that does not. A key question, then, is how to select the best leaders. We suggest that the nomination and vetting of potential agency heads be accomplished through a public process. A respected organization such as the American Statistical Association, which represents statisticians across academia, industry and government, could be tasked with developing a set of recommendations for vacant agency head positions. This list would be public. While there is no guarantee would always result in the desired leadership it would, we believe, increase the likelihood of this happening.

6 Conclusion

As we noted, unbiased statistical information is a lifeblood of a democracy and so trust in the statistical system is essential. To generate trust, users must have confidence that societally desired statistics are produced solely based on professional statistical decisions. Inevitably, however, tensions will arise between elected political leaders and the statistical system over the boundaries between professional and administrative independence. While there is no simple and comprehensive solution to this problem, we submit that more procedural attention should be paid to the role of leadership of statistical agencies and of the Chief Statistician of the United States and their selection in preserving professional autonomy and thereby enhancing the impact of national statistics.

Disclosure Statement

The authors report there are no competing interests to declare.

Notes

1 The Oxford English Dictionary defines autonomy as

The possession or right of self-government

Freedom of Action: a structure that gives greater autonomy to employees

2 In some cases, such as the decennial census, the government has determined that the information is of sufficient critical importance that it deems response to a survey or census to be mandatory. Even in these cases, however, the respondent is promised that their responses will be confidential.

3 We note that one of the authors testified in the lawsuit.

4 In fact, some argued that even if the Administration’s action were reversed, irreparable damage had already been done to the credibility of the Census Bureau and the 2020 Decennial Census.