The American Community Survey (ACS) Data Users Group has a new home—and a 12-year track record to build on. Before moving to the Federal Data Forum, the online community generated more than 5 million views across more than 5,300 discussion threads and replies, reflecting sustained demand for community-vetted guidance on one of the federal government's most widely used data products.
The chart below, drawn from 10 years of forum activity, shows what the community cared about most.
Housing, ZIP Codes, and Margins of Error: What ACS Users Asked About Most
Questions about specific subject matter—housing, income, poverty, and race/ethnicity—accounted for roughly one-third of all discussion threads. These questions grew steadily through 2022, when forum activity peaked at more than 800 threads and replies, before declining in 2023 and 2024.
Questions about geography were the second most common category and generated the most views per thread, averaging nearly 6,000 views each. ZIP Codes and ZIP Code Tabulation Areas (ZCTAs) were perennial topics, alongside posts about census tracts and block groups. ACS data users navigating differences among geographic concepts—or trying to align ACS estimates with other spatial datasets—continued to find these discussions useful long after they were posted.
Questions about accessing and working with ACS data products—including the Census API, data.census.gov, IPUMS, and the ACS Summary File—also grew steadily over time. By 2023, they accounted for 23% of all threads and replies. This trend likely reflects both the increasing complexity of the ACS data ecosystem and the growing number of researchers and analysts using these products.
Questions about statistical methods remained remarkably consistent, representing about roughly 10% of threads and replies each year. Margins of error, statistical significance, weighting, and Public Use Microdata Sample (PUMS) data were recurring topics that never went away, pointing to a persistent demand for technical information related to ACS data quality and measurement.
Finally, questions about the ACS program itself—including questionnaire changes, data releases, Summary File redesigns, and federal statistical policy—accounted for roughly 9% of all threads.
Differential privacy and response rates appeared too infrequently to qualify as perennial topics, despite being major policy issues—suggesting data users were more focused on practical data use than the underlying methodology debates and challenges.
Collegial and Solutions-Focused, With Frustration Peaking During the FactFinder Transition
Most discussions were practical and collegial. Practitioners asked specific questions, and other practitioners answered them. Frustration appeared from time to time, but it was the exception rather than the rule. When it did emerge, it tended to cluster around periods of significant change—most notably the 2019-2020 transition from American FactFinder to data.census.gov, when familiar workflows disappeared and replacement tools were still maturing.
The Future of Federal Data Expertise in an Era of AI
PRB recently launched a new ACS Data Users Group within the Federal Data Forum so that data users could continue the conversations that made the original community valuable. The way people access and work with federal data is changing rapidly, and the need for peer expertise has never been greater.
AI tools are increasingly part of how people approach data questions—for navigating complex datasets, drafting code, or getting a quick orientation to an unfamiliar table. These tools can be genuinely useful. They can also hallucinate, generating plausible but incorrect answers with complete confidence. An ongoing research project by the National Center for Science and Engineering Statistics (NCSES), in partnership with NORC at the University of Chicago and Georgetown University’s Massive Data Institute is currently examining how well large language models answer questions about federal statistical data and where they fall short.
Forum activity declined after 2022, for reasons that are likely varied. The growing use of AI tools may be one factor, alongside broader shifts in how practitioners seek information online. But the lesson of 12 years of ACS Data Users Group discussions is clear: getting federal data right requires human expertise. AI can surface an answer, but it takes a knowledgeable practitioner to evaluate whether that answer is correct.
That commitment to shared expertise is what this community exists to support. If you were a member of the original ACS Data Users Group, welcome back. If you are new, the discussion forum is a good place to start.