2020 ACS PUMS with Experimental Weights - CTPP

28 Jan 2022

Questions:

1) Will the Census Bureau be developing and releasing a 2019 (yes, 2019) Public Use
Microdata Sample (PUMS) using Experimental Weights?

I’m not sure this is a simple-to-answer question, but it bears asking.

The Census Bureau released (November 30, 2021) the Year 2020 ACS PUMS using Experimental
Weights. And the iPUMS site just this week (1/25/22) launched the 2020 ACS PUMS-X on their
wonderful/amazing website!

Also on 11/30/21, the Bureau released the 54 tables that will be the sum total of regular
tables on the 2020 ACS. There will be NO 2020 single-year estimates available either on
data.census.gov <http://data.census.gov/> or through the API (application
programming interface, like the R-package tidycensus.) These 54 tables are only available
at the national and state level. Region, county and large place level data will
(apparently) not be released for these 54 tables. 

I also read the working paper by Census Bureau staff: "Addressing Nonresponse Bias in
the American Community Survey During the Pandemic Using Administrative Data” 

https://www.census.gov/library/working-papers/2021/acs/2021_Rothbaum_01.html
<https://www.census.gov/library/working-papers/2021/acs/2021_Rothbaum_01.html>

The Rothbaum working paper goes into incredible detail on the need to replace the
traditional weights used in the ACS with experimental weights / new weights / “entropy
balance weights” / “experimental entropy balance weights” / ACS experimental weights.  The
paper shows a lot of comparisons between older ACS data, typically 2009 to 2019, and ACS-X
data with the experimental weights, for 2019 and 2020. Yes, the Census Bureau has
implemented the experimental weighting procedure for 2019.

The Bureau is pretty clear about data users NOT comparing the ACS 2005-2019 using the
“standard weights” with the ACS 2020 using the experimental weights. Well, at least that’s
their advice/admonition/plea.

Year 2020 data is incomparable. It’s like 61* (Maris) versus 60 (Ruth). 

I think we need ACS-X data (ACS with Experimental Weights) for multiple years in order to
make some sense out of this bedlam: 2019, 2020 and 2021. There’s obviously a major time
and cost with creating these data, but it will be worth the investment.

A followup question is:

2) Will the five-year 2016-2020 data, tentatively scheduled for release March 20, 2022
(plus or minus weeks?) be weighted using ACS standard weights? ACS experiment weights? A
hybrid?

I scoured the Census Bureaus’ website, read the transcript from the 11/30/21 webinar,
scoured the state data center and ACS data community websites, and couldn’t find answers.

Help?

Chuck Purvis