The ACS 3-year Demographic Estimates Are History

by Ken Poole:

Census Director John Thompson and Acting Director for the 2020 Census Lisa Blumerman spoke with a group of Census stakeholders today about Census plans regarding the American Community Survey.   My take from the conversation: the ACS 3-year demographic estimates are history.  When Census first developed the ACS in the 1990s, the goal was to replace data from the Census long-form (available every 10 years) with 5-years of aggregated data combined to provide a “rolling average.”  Starting up a program with 5-year rolling averages took quite a long time to move from research and fieldwork (which all needed Congressional support) to useful data placed in the hands of practitioners.  Ultimately, we all knew that annual releases of 5-year estimates would make data available to communities of every size for each and every year (not every decade as before).

But, the fiscal headwinds at the time made patience a virtue that we ironically could ill afford.  The 3-year estimates were designed to get Census data out to many “medium-sized” communities (20,000 to 65,000) two years earlier than the first available 5-year estimates.  As a result, we created the three-part ACS with 1-year estimates for the largest communities, 3-year estimates for medium and large communities, and 5-year estimates for all communities.

Today, Census leaders described the hard decisions they have to make when allocating their limited resources.  Essentially, the 3-year estimates will not be published next fall for 2012-2014 due to the $15 million shortage resulting from the $124 million cut to the Census budget, all coming from the Periodic Programs and Censuses account.

The decision to eliminate the 3-year estimate was part of a series of moves including cuts to follow-up operations support, field representative refresher training, and ACS interviewer observations.   Furthermore, the President’s budget proposal released yesterday did not seek the estimated $2.4 million that would be required to re-instate the 3-year data product in the FY 2016 (so no 2013-2015 estimates would be made either).

Census plans to continue to release 1-year and 5-year estimates and to focus its resources on activities that ensure data quality.  During our discussion, the 5-year estimates were identified as the primary annual data element that ACS promises to provide to every community as part of its mandate.

It is unclear what advocacy on behalf of the 3-year estimates might beget from a Congress that continues to engage in debates about whether responses to the ACS should be voluntary or remain mandatory (or whether the ACS should exist at all).

Suffice it to say that, with no new 3-year averages being published, any users relying on those data will need to start making adjustments now.  The expectation is that once we adjust, the 3-year estimates will go the way of other data products remembered fondly.