ACS Data Users Group

Has anyone else been working with the race variables in the 2020 ACS 1-year PUMS with experimental data. The readme file says not to compare estimates using the experimental data with other PUMS. But race is one of the dimensions controlled for in the weighting, so it is natural to expect no surprise with the race variables. However, the RAC1P in the 2020 experimental data file shows 62.7% for "white alone". In comparison, in the 2019 1-year PUMS, "white alone" in RAC1P is 72.1%. At the same time, RAC1P in the 2020 experimental data has11.5% for "two or more races", compared to 3.5% in 2019 for the same category. These differences are huge considering the trends of very minor annual changes in previous years. Is there any social/demographic explanation for these huge differences between 2019 and 2020's experimental data or my calculation is wrong? Any help would be appreciated.

Hi, the big difference in the race estimates are likely due to the changes to the race and ethnicity questions in the 2020 ACS rather than the experimental weights. See https://www.census.gov/programs-surveys/acs/technical-documentation/user-notes/2021-03.html. These changes are similar to the changes made in the 2020 Census. One of the major changes is that the number of characters allowed in the write-in area has been increased from 30 to 200, allowing up to 6 distinct groups to be coded and tabulated from each line. Previously they assigned up to 2 codes/groups and prioritized coding race groups over Hispanic origin for the race write-in. This change could lead to the capturing of more multi-race and thus lower some of the alone categories.

Thanks Karen. That explains it. I checked the ACS document "AMERICAN COMMUNITY SURVEY 2020 ACS 1-YEAR PUMS FILES - Variable Changes and Explanations" (https://www2.census.gov/programs-surveys/acs/experimental/2020/documentation/pums/ACS2020_PUMS_Variable_Changes_and_Explanations.pdf) and it has no mention of the race variables being change in 2020. I wish they they had included the race variable changes in that document. A related question, is the change in the race question also implemented in the March Supplement of the Current Population Survey?

Hi Wei Yen, I'm not aware of changes in the race question for the CPS ASEC but please do share if you come across any notices or documentation of changes. I'm not seeing a big shift in the estimates of the White alone in CA using the 2019-2021 CPS ASEC. The data collection method for the CPS ASEC is different than for the ACS though. Unlike the ACS which is conducted primarily by mail using a self-response questionnaire with an internet response option, the CPS ASEC is conducted by interviewers via CATI or CAPI. I believe the interviewer reads the major race groups and asks the respondent to choose one or more from that list, and so you don't have the write-in and coding issues that are encountered with the ACS and 2020 Census questionnaire.

Thanks a lot!