Skip Header

Tucker McElroy, Natalya Titova, Chaitra Nagaraja
Component ID: #ti2025146740

The rolling sample methodology of the American Community Survey (ACS) introduces temporal distortions, resulting in Multiyear Estimates (MYEs) that measure aggregate activity over three or five years. They cannot be viewed as estimates of the final year, nor of the middle time point of the period; neither can they be viewed as simple averages of single year estimates belonging to the same time span. The U.S. Census Bureau has enunciated this principle forcefully, yet the question remains -€“ in light of the fact that users are likely to ignore official pronouncements -€“ how damaging are these unsanctioned viewpoints? In particular, can one quantify the impact of making such fallacious use of the MYEs? This paper answers these questions positively: yes, it can be quantified, and in general there is fairly serious degradation to the usability of MYEs when applying these faulty interpretations. We first offer a critique of the comparison of diverse MYEs via the published standard errors, and then we discuss a simple, general method based on relative percent discrepancies. This technique is illustrated on the test database of the ACS, from which we draw our general conclusions.

Back to Header