Conflicts in the Use of the ACS by Federal Agencies Between Statutory Requirements and Survey Methodology

Abstract

The American Community Survey (ACS) has been fully implemented since 2005. The Census Bureau has released four 5-year datasets, the most geographically detailed dataset. Yet, government agencies still grapple with how to use the multiple datasets and estimates. The Census Bureau publishes guidelines for their use, emphasizing the need to balance timeliness and precision in choosing an estimate and encouraging the use of the margin of error. This study examines how three federal agencies use the ACS to implement programs to understand whether the published guidelines address the issues of importance to government agencies. These programs all use income data from the ACS, but for different ends: eligibility criteria, evaluation of federal guideline implementation, and allocation of funds among urban areas. After reviewing agency publications, studying changes in policy, and conducting personal interviews with officials, we find that the agencies consider not only precision and timeliness when choosing an estimate, but also statutory requirements, computational limitations, and geographic needs. Federal agencies are cognizant that these estimates are subject to error, but find that the margin of error adds complexity that does not necessarily result in better implementation of programs. The ACS offers users multiple dataset choices, with varying degrees of reliability, to estimate population characteristics. Current Census Bureau guidelines for the ACS do not meet the needs of many government agencies, as federal statutes are not designed with the current survey methodology in mind.