The American Statistical Association’s Privacy and Confidentiality Committee, in collaboration with the Social Data Science Center at the University of Maryland, hosted a Privacy Day webinar on January 28 as part of Privacy Week.
Titled “Balancing Statistical and Non-Statistical Uses of Federal Data: Privacy, Governance, and Public Trust,” the event examined a longstanding principle in federal statistics—the distinction between statistical and nonstatistical uses of data. This distinction has evolved over more than a century of federal statistical practice and is now encoded in laws such as US Code Title 13 and the Confidential Information Protection and the Statistical Efficiency Act.
At the heart of the discussion was a deceptively simple question: What is meant by “statistical purpose”? Although the term appears frequently in legislation, policy, and public communication, it’s often left underspecified. As the speakers noted, statistics span a broad range of applications, making the meaning of “statistical purpose” far from self-evident. This webinar explored how this concept has evolved, how it is interpreted in practice, and why greater clarity is increasingly important in an evolving federal data landscape.
The first presentation was delivered by Sallie Ann Keller, chief scientist and associate director for research and methodology at the US Census Bureau, and Michael B. Hawes, senior statistician for scientific communication at the US Census Bureau. They examined the historical and legal foundations of “statistical purpose” within the federal statistical system. Tracing developments from early 20th-century legislation through the suspension of confidentiality protections during World War II and the establishment of modern frameworks like Title 13 and CIPSEA, her presentation explored how the notion of “statistical purpose” has evolved gradually and through contestation.
The concept of a functional separation was introduced in the 1970s and later codified into statute. As reflected in CIPSEA, “statistical” uses refer to uses of data aimed at producing aggregate information, without identifying the individuals or organizations that comprise such groups. “Nonstatistical” uses refer to using data to make decisions about identifiable individuals, such as in law enforcement. This functional separation has guided federal data practice and is intended to help protect confidentiality and prevent harm to individuals.
Building on this foundation, the speakers highlighted two key dimensions underlying “statistical purpose”: restrictions on access to identifiable information and limitations on how such information may be used. Together, these dimensions are intended to ensure data is protected from unauthorized disclosure and used only for appropriate, non-harmful purposes. They also noted that the statistical community has often been “nonchalant” about the term, despite its central role in maintaining public trust. In this context, the presentation referenced the 2025 executive order on “restoring gold standard science” and related guidance from the Office of Science and Technology Policy, which articulates nine tenets of scientific practice, such as transparency, reproducibility, and the communication of error and uncertainty. The presenters argued that these tenets strengthen the scientific and ethical foundations of “statistical purpose.” As new data sources, particularly administrative data, become more prevalent, the speakers emphasized the need to more carefully examine what the term entails and to improve how it’s communicated to the public.
The second presentation, delivered by Alexandra Wood, visiting assistant professor of artificial intelligence, policy, and society at Purdue University, framed “statistical purpose” as a special case of purpose limitation that defines the boundaries of permissible data use. Wood emphasized that despite its central role in privacy and data governance, the term lacks a consistent and operational definition across legal and regulatory contexts. The presentation also noted that, when adopted in consumer privacy frameworks like GDPR, “statistical purpose” is often used to balance data access with privacy protection. But this balance can become unclear when its original context is not preserved. As a result, the concept may be interpreted in ways that emphasize data access without corresponding safeguards, which contributes to ambiguity, inconsistent applications, and uncertainty on acceptable use.
To address this ambiguity, the presentation introduced a multi-dimensional framework that illustrates the discrepancies in how “statistical purpose” is constructed across regulatory definitions such as population-level analysis, aggregation, and exclusion of individual-level decision-making. The framework illustrated the risks of unclear or inconsistent interpretations through historical examples like the use of census data during the internment of Japanese Americans and post-9/11 tabulations. These cases highlight the need for clearer definitions and more consistent safeguards as data use expands in scope and complexity.
Taken together, the two presentations highlighted both the importance and the evolving challenges of “statistical purpose.” Keller and Hawes emphasized its role as a longstanding legal and institutional safeguard, while Wood focused on its conceptual ambiguity and the risks that arise when it’s not clearly defined. Both perspectives point to a common concern: Although the term is central to maintaining public trust, it is not consistently understood or applied. As data systems evolve and new uses emerge, these discussions highlighted the need for clearer definitions and more transparent communication about how data is used and protected.
Watch the webinar “Balancing Statistical and Non-Statistical Uses of Federal Data: Privacy, Governance, & Public Trust” on YouTube.
New Webinar Series: Statistical Confidentiality and Data Access
Interested in learning more about how to protect confidentiality when publishing statistics or expanding data access? The ASA Privacy and Confidentiality Committee—in partnership with the Federal Committee on Statistical Methodology, Confidentiality, and Data Access interest group—is sponsoring an upcoming seminar series on statistical confidentiality and data access.
This monthly series will explore topics such as assessing disclosure risk, commonly used disclosure avoidance techniques, governance of the disclosure review process, methods for generating and evaluating synthetic data, differential privacy, tiered data-access models, and more.
The series will kick off this summer with an introductory overview of disclosure avoidance across the federal statistical system. . To learn more, sign up for the FCSM Confidentiality and Data Access Interest Group’s listserv by sending an email to FCSM-CDAC-subscribe-request@listserv.gsa.gov.
Leave a Reply