ACS Five-Year Income: The Last Year Is Not the Whole Period

A county income figure labeled “2020–2024 ACS 5-year” isn’t a measurement of 2024 alone. Before using it in a location comparison or business plan, keep the whole collection period attached to the number.

A dollar-year label doesn’t shorten the survey period

The Census Bureau’s explanation of ACS period estimates says five-year estimates pool and process records across the period. They aren’t the arithmetic average of five published annual estimates, nor a snapshot of the final day.

Income is inflation-adjusted to the final year’s dollar basis. “2024 dollars” identifies a purchasing-power basis; it doesn’t turn five years of collected responses into a 2024-only estimate.

Choose the product before copying the income

Census product guidance distinguishes standard one-year estimates, using 12 months of collected data for areas with populations of at least 65,000, from five-year estimates, using 60 months and covering all areas. One-year supplemental estimates are a separate product. These are collection windows, not a claim that every income response covers those exact calendar months.

The trade-off is currency versus precision: five-year estimates use a larger sample but describe a longer period. A missing standard one-year county row isn’t permission to substitute a five-year figure while keeping a one-year heading.

For a cross-county comparison, first find a common product and period available for every selected county. For an annual question, don’t silently replace missing annual evidence with a multiyear estimate.

Adjacent five-year releases aren’t adjacent annual observations

Census comparison guidance recommends nonoverlapping five-year datasets and says not to compare one-year estimates with five-year estimates. Check changes in geography and variables too.

The calendar below illustrates why the last year alone is an inadequate label. It doesn’t calculate income growth, statistical significance or a correlation between estimates.

Work through a three-window calendar

Original reading exercise: A planning folder contains three labeled windows. No actual county income values or current release results are supplied.

B and C share January 2020–December 2023: 4 × 12 = 48 months. That is 48 ÷ 60 = 80% of either calendar window. Twelve months leave and twelve enter. The 80% isn’t a measured share of respondents, a statistical weight or an independence calculation.

A and C share zero collection months. A and B still share 2019: 12 ÷ 60 = 20%. Changing just one endpoint in your notes can therefore hide an overlap.

For a five-year period comparison, A and C pass this calendar screen; B and C don’t. Passing doesn’t validate every other input. Match the household-income measure, population definition and geography, use a common dollar basis, and retain margins of error. A point-estimate difference alone isn’t a completed significance test.