Answers are grounded in this paper and cite their sources.
Empirical analysis reveals high statistical uncertainty in multiyear census estimates across geographic areas, highlighting severe limits on tracking community change over time.
Problem, research strategy, and findings The American Community Survey (ACS) is a fundamental data set widely employed in urban and regional planning, and its fine level of geographic detail makes it indispensable for many geographic analyses. With 20 years of data collected, it is now possible to compare several non-overlapping 5-year period estimates to study recent population change. Issues with statistical uncertainty in the estimates raise questions about the feasibility of making such comparisons. In this study I evaluated more than 300 variables from 25 data tables for two non-overlapping, consecutive time periods for more than 400,000 individual geographic areas. I calculated change and percentage change for each pair of estimates, which I evaluated for statistical significance and level of precision. The findings indicated that most estimates cannot be compared, and are of low reliability: Approximately 90% of the estimates from the two time periods were not statistically different, and for those that were, the estimate of change was of low reliability in about 90% of cases.Takeaway for practice The study illustrates the process planners and researchers should follow to evaluate estimates. Data users should be wary of interpreting change over time using the ACS, as the estimates are often not statistically significant and are of low reliability. It is important to calculate new margins of error, test for statistical significance, evaluate the reliability of results, and communicate the uncertainty of the estimates to stakeholders.
No takes yet. Share an insight, caveat, or question.
Frank Donnelly (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: