Skip to main content
SearchLoginLogin or Signup

The U.S. Census as a Crucible for Trust in Government

Published onJan 31, 2020
The U.S. Census as a Crucible for Trust in Government
key-enterThis Pub is a Commentary on

1. The Integrity of Government Statistics as a Foundation for an Informed Society

Amid entrenched divisions and intense partisan rancor in the United States, Teresa Sullivan (2020) has provided a richly informative, thoughtfully nuanced, and much appreciated examination of the unifying role of the decennial census in American democracy. For me, her article elicited wide-ranging thoughts about the implications of the census for trust in government more broadly.

Since the early 1990s, my primary professional affiliations have been in university-based health science departments, but my long-standing interest in the role of the census at the interface of science and public policy has shaped my career in numerous ways. My PhD studies were intertwined with the census-undercount controversy, with summer jobs at the U.S. Census Bureau in the 1980s morphing into opportunities to work on the proposed adjustment of the 1990 census and to contribute to subsequent related discussions (e.g., Belin, 2000, 2001; Belin et al., 1993; Belin & Rolph, 1994). I also served from 2001 to 2006 as a member of the American Statistical Association (ASA) Census Advisory Committee, and my familiarity with the census fueled a collaboration with students to develop a one-number density-variation/compactness (DVC) score to evaluate redistricting plans based only on census data and geography in a way that could anticipate and potentially guard against political polarization (Belin, Fischer, & Zigler, 2011).

In addition to drawing on these experiences, I am also winding up two terms of service on the ASA Committee on Professional Ethics, and in that capacity I salute the connections made by Sullivan between the integrity of the census and the role of personal and professional integrity among statisticians. I believe the article’s reflections on the meaning of census-related constitutional provisions underscore that scientific accuracy has implications for equality and justice, that self-government is not a spectator sport, and that there is potential for the statistics profession to do better. Furthermore, in light of the reality of government statisticians in multiple countries having been the targets of outrageous attacks for refusing to bear false witness in their professional roles (see, e.g., Carriquiry, 2012; Georgiou, 2017; Langkjaer-Bain, 2017), there is a need for solidarity on matters of principle that transcend politics. Bridging these themes, I view the ASA’s Ethical Guidelines for Statistical Practice (, enunciating principles such as “The discipline of statistics links the capacity to observe with the ability to gather evidence and make decisions, providing a foundation for building a more informed society,” and “Above all, professionalism in statistical practice presumes the goal of advancing knowledge while avoiding harm,” as a reflection of this needed universal spirit that aligns with Sullivan’s perspective on the census. (In passing, I want to pay tribute to Howard Hogan, whose work on coverage evaluation at the Census Bureau, which included mentoring me when I worked there, foreshadowed his leadership of the ASA Committee on Professional Ethics in crafting the successful 2016 update of the previous 1999 version of the guidelines.)

In the balance of this discussion, I allude to other census-related experiences that I hope will complement the insights offered by Sullivan on the census as a crucible for trust in government.

 2. Census Enumeration in the Current Social Climate

An anecdote that a Census Bureau colleague shared with me in the summer of 2018 illustrates how policy conflicts can be channeled into the realm of personal lived experience through the census. The colleague described a recent report from a two-person address-listing team working in inner-city Baltimore as part of the Census Bureau’s “address canvassing” operation, a process summarized at as a way to refine the master address file that serves as the primary frame for the decennial census. Modern technology, including satellite imagery, allows much of the quality-control checking on address lists to be performed remotely in a central office, but the Census Bureau still hires tens of thousands of address listers across the country for field operations to pinpoint the elements of the nation’s housing stock. As reported to my colleague, in the wake of receiving no answer after knocking on a door, the address-listing team went down the street to collect information from another nearby housing unit, only to look back and see people dash out of the first housing unit carrying personal belongings, jump into a nearby parked car, and pull away. The clear implication was that government workers knocking on the door were regarded as a threat, and it is not hard to imagine that immigration-related law-enforcement fears fueled the dramatic reaction to Census Bureau address listers knocking at the door.

Regardless of one’s perspective on immigration policy, it is clear that many individuals will resist being counted due to a lack of trust in the government. Census Bureau assurances that the data will remain confidential for 72 years and will not be shared with law-enforcement agencies might help a little with enumerating hard-to-count populations, but there is a broader question of what is the Census Bureau’s responsibility with regard to noncitizen ‘persons’ who are living in the United States. It should be acknowledged that some observers are apt to ask, ‘Who cares if the census does not include some persons in the country illegally?’ It should also be acknowledged that the answer would include administrators of public schools and hospitals who are not legally allowed to refuse access to persons living in their communities, as well as state and local government officials whose budgets depend on their share of more than $600 billion per year in public funding tied to population counts. 

How to reconcile clashing interests in a polarized political environment remains an abiding challenge. I agree with Sullivan (2020) that statisticians can make a difference by emphasizing that the public interest in an accurate census dovetails with the principles of statistical science. I also believe that public discourse would benefit from open recognition that policy depends not only on science but also on value judgments, as has been expressed in cross-national perspectives on the census-adjustment controversy in the United States by colleagues in the United Kingdom (Diamond & Skinner, 1994) and Australia (Steel, 1994). The American census itself serves as living proof that it is possible to balance competing interests across fractious divides, and with constitutional provisions empowering Congress to direct the census, the Constitution offers a framework within which clear thinking about core values can be translated into settled law. That, of course, is easier said than done, but as elaborated further below, it is difficult to envision closure on census controversies without some more specific guidance in the law that delegates census-taking to the Secretary of Commerce.

3. The Long Arm of Pretext in Debates About Census Accuracy

One of the most intriguing points made by Sullivan (2020) is the connection she drew between the recent Supreme Court decisions on gerrymandering in Rucho v. Common Cause and on census-taking in Department of Commerce v. New York. The interpretation of the Administrative Practices Act had been pivotal in court battles over the 1990 census, with a federal district court judge ruling that it was not “arbitrary and capricious” for the Secretary of Commerce not to adjust the 1990 census, a divided federal appeals court ruling that the validity of that decision should be evaluated according to the stricter standard of whether the constitutional rights of the plaintiffs suing for a statistical adjustment had been violated, and a unanimous Supreme Court reversing the appeals court and reinstating the more hands-off “arbitrary-and-capricious” criterion of the Administrative Practices Act as the applicable standard for reviewing the actions of the Secretary of Commerce.

In its recent Department of Commerce v. New York ruling, the bogus pretext that the Justice Department sought a citizenship question to enforce the Voting Rights Act proved to be too much for a narrow majority of the Supreme Court. But as Sullivan (2020) noted, future partisans might try to squeeze through a loophole lurking in the shadows of the Court’s decisions. I had not previously considered the scenario that courts might tolerate less obvious pretexts for operational decisions that adversely affect census accuracy, and I believe it is both realistic and alarming to anticipate that courts would accept contrived excuses or even openly partisan motivations for operational decisions about the census that have disparate impacts across different subpopulations in the same way that some governmental decisions that have the effect of purging qualified individuals from voter-registration lists have been accepted by courts. Sullivan deserves credit for anticipating such provocations, as there might be ways to guard against such possibilities and preserve the integrity of future censuses through forward-thinking planning.

Drawing on my own experience, I found myself reflecting on an exchange I had in the 1990s with Thomas Hofeller, who worked at the time as a staff member for a congressional committee and who had established himself as a critic of statistical adjustment of the census. Hofeller’s later work as a political operative was intertwined with both partisan redistricting activities and the effort to add a citizenship question to the decennial census. Back in the 1990s, as the discussant in a topic-contributed session at the Joint Statistical Meetings featuring talks critical of model-based adjustments for differential undercount, Hofeller described a vision of an “equal-opportunity” census as a framework for the government to fulfill its constitutional responsibility without entertaining statistical adjustments that he regarded as problematic. When the floor was open to questions, I asked him whether his vision of an equal-opportunity census extended to the idea that census-related advertising and coverage-improvement programs should be subject to equal allocation of resources across jurisdictions, to which he responded along the lines of, “I can see that my presentation provided grist for the mill for the other side.” His projection onto me immediately struck me as a window into a partisan worldview lurking beneath his nominally equity-anchored equal-opportunity census rhetoric.

After Hofeller’s death in 2018, it was revealed that Hofeller had written (in the context of political consulting work that generated more than $2 million in income from the Republican National Committee) that adding a citizenship question to the census “would be advantageous to Republicans and non-Hispanic whites.” It was additionally revealed that he had written the portion of the Department of Justice letter later recognized as a false pretext attempting to justify adding a citizenship question to the census. I find Hofeller’s behind-the-scene manipulations, trading on fears to discourage census participation by persons whose omission would advance a political agenda of his, to be similarly morally bankrupt and similarly corrupt as thinly veiled voter-suppression efforts. I also view the arc connecting his more recent machinations to earlier rhetoric about an equal-opportunity census as a not-so-woolly reminder that there is more than one way to dress up a wolf in sheep’s clothing. For the sake of not having the wool pulled over our eyes, I believe it would be advisable to build stronger standards into the laws governing census-taking.

4. How Can We Most Effectively Protect Against Census-Related Partisan Manipulations?

With clashes of interests and diverging political priorities being inevitable in a society as complex as ours, Sullivan (2020) does an exemplary job of anchoring the census to the legal imperatives of the Constitution and in particular to its provision that the census is to be conducted “in such Manner as they [Congress] shall by Law direct.” A reality of the existing legal framework surrounding the decennial census is that although the grant of authority is not unlimited, the broad discretion granted to the Secretary of Commerce in Section 141(a) of Title 13 of the U.S. Code includes language that “The Secretary shall, in the year 1980 and every 10 years thereafter, take a decennial census of population as of the first day of April of such year … in such form and content as he may determine, including the use of sampling procedures and special surveys.” One could envision Congress providing more direction by law than simply designating April 1 to be the decennial census date and leaving the details to the Secretary of Commerce. As an example, since the existing law was interpreted by the Supreme Court in Department of Commerce v. U.S. House of Representatives (1999) not to allow sampling for follow-up of initial non-response to the census, as was planned in the Clinton administration’s proposed “one-number census,” it would be possible for Congress to change the law to explicitly allow or even to require sampling for nonresponse follow-up. It might be hard to imagine such a change in the current political climate, but Congress has passed new laws in many other contexts when courts have not interpreted earlier versions of laws as Congress intended.

Sullivan highlights the negative impact of census error on trust in government, although it is hard to see bipartisan consensus emerging from that observation, as the politics of addressing the census not being equally accurate everywhere have been fraught. Another factor that has the potential to influence the politics of census-taking is cost containment. To that end, one could envision statisticians explaining that there are inherent tradeoffs between cost and precision in any measurement context, and that if lawmakers were to impose a strict budget constraint along with explicit permission to use sampling for nonresponse follow-up and/or approaches for combining data from multiple sources (hopefully with other built-in constraints to address information privacy concerns), one would anticipate changes to census-taking methodology to adapt to the new set of constraints. It is still difficult to envision bipartisan consensus emerging in the current political environment, but the long-run potential for interests in bias correction and cost containment to come into alignment should not be ignored.

While I agree with Sullivan that statisticians should always be ready to speak truth to power, I have qualms about her prescription being adequate to the kinds of maneuvers pulled by political operatives like Thomas Hofeller. In particular, I do not believe that the remedy for dysfunctional gerrymandering rests in census-taking methodology. But taking a broader view from a systems-engineering perspective, I do believe that statisticians have the ability to engineer consensus-building strategies for furthering constitutional democracy.

An illustration presented in Belin, Fischer, and Zigler (2011) illuminates the need for more structure to protect against partisan manipulation in redistricting than is provided by the one-person/one-vote constraints imposed by Supreme Court rulings cited by Sullivan. Considering a jurisdiction (e.g., a state) that is entitled to seven electoral districts and that has an even split between two parties (A and B) in voter registration, there are many scenarios for the district-level partisan divides in voter registration. For example, there could be three equal-sized districts that are 60–40 in favor of Party A, three that are 60–40 in favor of Party B, and one district that has a 50–50 split. Alternatively, the overall balance in voter registration could be manifested in four districts that are 60–40 for A, two districts that are 60–40 for B, and one district that is 70–30 for B. With a more aggressive strategy of what political scientists call ‘packing’ high concentrations of opposing-party voters into a small number of districts, it would be possible for Party A to position itself for an even greater partisan advantage in the legislature if it had the ability to form five districts that are 60-40 for A, one district that is 70–30 for B, and one district that is 80–20 for B.

We analogized such an aggressive packing strategy to a scenario where the National Basketball Association (NBA) championship, currently based on a best-of-seven series of 5-on-5 basketball games, would instead be structured to allow the team representing the conference that won the All-Star Game to choose the number of players per game subject to each game having 10 players but only requiring an average of five to a side across seven games. Accordingly, the team with numerical control could dictate that five games would be 6-on-4 in their favor, one game would be 3-on-7, and one game would be 2-on-8, inducing such lopsided advantages that the outcome of the NBA Finals series would effectively be determined by the outcome of the All-Star Game. If sports fans would not tolerate such a system for determining an NBA league champion, why should Americans tolerate an analogous system for determining public policy?

Noting that DVC scores distinguish plans with greater variation in population density across districts (consistent with safely ‘blue’ and safely ‘red’ districts that could be anticipated to produce polarized government) from plans with less variation in population density across districts (giving rise to more competitive districts across a range of realistic scenarios for shifts in overall partisan balance), Belin, Fischer & Zigler (2011) further argue that if information on DVC scores were routinely made available to the public as redistricting plans were being considered, it would be possible for consensus to build around plans that could reasonably be expected to yield a broader “center” and consequently less polarized government. In my view, statisticians should not be shy about helping to engineer systems that respect established democratic norms, that avoid partisan bias, and that are responsive to shifts in public opinion, all of which are worthy goals in a constitutional democracy. We should anticipate that there will always be proverbial wolves in sheep’s clothing seeking to gain partisan advantage, and we should be ready with freedom-of-information strategies such as publishing DVC scores as a way of honoring the public interest.

5. Concluding Remarks

In my work over the years with career professionals at the Census Bureau, I have been impressed with their commitment to what might be called the “Horton Hears a Who” principle, namely that “A person’s a person, no matter how small” (Seuss, 1954). Sullivan’s (2020) Harvard Data Science Review article is a worthy addition to that tradition, highlighting challenges we face in preserving the integrity of the decennial census in the United States and in preserving the integrity of our societal statistical infrastructure more broadly.

I commend Sullivan’s focus on personal integrity and ethics, and I share her concerns about unwarranted smears on the credibility of government personnel and disinformation in the political sphere more broadly. I also applaud the renewed attention that Sullivan has given to historical violations of confidentiality in the United States, as recent events in other countries (Argentina, Greece) underscore how damaging it is to trust in government when the integrity of official statistics is in doubt. To me, the public interest perspective advocated by Sullivan aligns with the universal principle embodied in the commandment “Thou shalt not bear false witness.” In highlighting parallel efforts within the statistics profession to cultivate identity development around ethics, I hope to reinforce the message that statisticians should continue to emphasize the importance of bearing witness in our contributions to public policy debates.

It is not possible to legislate one’s way out of all problems, as self-government necessarily relies on individuals being willing to act in good faith to further the public interest. But I also believe that gaps and deficiencies in existing laws can be and should be remedied, and that statisticians have an important role to play in contributing to those discussions. More broadly, I agree with Sullivan that we have more work to do on the decennial census in our collective efforts to form a more perfect union, establish justice, insure domestic tranquility, provide for the common defense, promote the general welfare, and secure the blessings of liberty to ourselves and our posterity, and I salute her contribution to the Harvard Data Science Review for moving us forward toward those shared goals.

Disclosure Statement

Thomas R. Belin has no financial or non-financial disclosures to share for this article.


Belin, T. R. (2000). Do we have to count one by one? [Book review]. Science, 287(5451), 239–240.

Belin, T. R. (2001). Can we improve upon an attempted headcount? Society, 39(1), 34–41. 

Belin, T. R., Diffendal, G. J., Mack, S., Rubin, D. B., Schafer, J. L., & Zaslavsky, A. M. (1993). Hierarchical logistic-regression models for imputation of unresolved enumeration status in undercount estimation. Journal of the American Statistical Association, 88(423), 1149–1166.

Belin, T. R., Fischer, H. J., & Zigler, C. M. (2011). Using a density-variation/compactness measure to evaluate redistricting plans for partisan bias and electoral responsiveness. Statistics, Politics, and Policy, 2(1).

Belin, T. R., & Rolph, J. E. (1994). Can we reach consensus on census adjustment? Statistical Science, 9(4), 486–537.

Carriquiry, A. (2012). Graciela Bevacqua. Significance, 9(6), 34–36.

Diamond, I., & Skinner, C. (1994). Comment. Statistical Science, 9(4), 508–510.

Georgiou, A. (2017). Towards a global system of monitoring the implementation of UN fundamental principles in national official statistics. Statistical Journal of the IAOS, 33(2), 387–397.

Langkjaer-Bain, R. (2017). Trials of a statistician. Significance, 14(4), 14–19. 

Steel, D. (1994). Comment. Statistical Science, 9(4), 517–519.

Seuss, D. (1954). Horton Hears a Who! New York: Random House.

Sullivan, T. A. (2020). Coming to our census: How social statistics underpin our democracy (and republic). Harvard Data Science Review, 2(1).

©2020 Thomas R. Belin. This article is licensed under a Creative Commons Attribution (CC BY 4.0) International license, except where otherwise indicated with respect to particular material included in the article.

1 of 9
No comments here
Why not start the discussion?