As we have alerted readers in the past, the definitions of race and ethnicity will be changing over the next couple of years. As with most major changes, this has both positive and negative effects on users.
On the plus side, the new definitions were outlined in the OMB ‘Statistical Policy Directive No. 15, published earlier this year. These call for a unified question on race and Hispanic origin, which should eliminate some confusion between what are two distinct tabulations at this point in time. As well, they will be adding a new “Middle Eastern or North African” race category, which has been long needed.
The manner in which the question is asked has a significant impact on the results, which initially misled many into believing that the 2020 census showed a massive decrease in the non-Hispanic white population. The reality was that the wording of the question itself was confusing to respondents as was clearly seen in the disconnect between the 2020 Census and the 2020 ACS numbers.
The new standards will be implemented in either the 2026 or 2027 ACS surveys, meaning that these will not be fully implemented until the 2026-2030 5-year estimates are released. The AGS plan will include a transitional period where we use the 1-year estimates using the new format in conjunction with the more spatially detailed 5-year estimates which will use the old format for several years to come. The new format will most certainly be used in the 2026 test census, and we would expect that the ACS will include this for 2026.
On a more troubling note, a recent Associate Press article (https://apnews.com/article/data-census-bureau-surveys-statistical-agencies-81d0581451b143fef10b87045848c095) entitled “’This is break glass in case of emergency stuff’: Analysts alarmed at threats to US data gathering” paints an ominous picture. Without getting into the politics of whether the census was intended to include illegal migrants, one of the key measures being discussed is that contact for survey purposes would be limited to two. The average ACS response requires at least three follow-ups to the original request, and it is believed that well over 10% of households in the 2020 census would have been missed entirely if this rule had been in place at the time.
Further, funding for the Census Bureau continues to be threatened and is certainly not keeping up with costs, so at some point, the data quality of surveys like the ACS, Current Population Survey, and even the Census are likely to suffer.
Given the disclosure avoidance debacle in the 2020 census, AGS is taking major steps towards minimizing our reliance on the decennial census, with a goal of being able to use the 2030 census as merely one of several comprehensive household counts. The ACS is likely to be spared for some time to come, as the bureau concluded that “the science does not yet exist to comprehensively implement a formally private solution for the ACS.” Translation? We would if we could, but we can’t so we won’t. This is probably the best news out of the bureau in some time, as we find the notion of implementing data muddying techniques on a sample survey to be nonsensical at best.
Speaking of muddied data, the DHC-A release last month includes up to 1500 ethnicity groups available at irregular geographies – in some cases state level only, in others down to a census tract level depending on the size of the group. The one area where this data will be very helpful is in the transition to the new race/ethnicity categories, since the Middle Eastern or North African tally should be available for most large counties.
Recent Comments