The US Census Bureau has released undercount ond overcount rates for the 2020 census. Here is the press release: https://www.census.gov/newsroom/press-releases/2022/pes-2020-undercount-overcount-by-state.html?utm_campaign=20220519msc20s2ccnwsrs&utm_medium=email&utm_source=govdelivery
I want to applaud the Census Bureau for this study and getting it out to the public. Well done!!
Also, a question: The states with significant undercounts are Arkansas, Florida, Illinois, Mississippi, Tennessee, and Texas. There may be a geographic pattern: apart from Illinois, the remainder are all on the south. Including Illinois, here is the question: did the statitical challenges faced by southern states migrate north to with the population shift known as the Great Migration, which saw many from the South - mostly persons of color, who often face under-represedntation in official statistgics - migrate north, often to Illinois and Michigan? (NB: Michigan did not have a significant undercount but Detroit is reported to be challenging the results, claiming an 8% undercount). Are there demographic / sociological / cultural patterns in the undercount? What research has looked into this question?Thank you!!
Terry,Since the article is behind a paywall, I can't read much more than the headline and subheading:"Who Rigged the Census?Under-counts may have cost Florida and Texas another House seat."Again I can't read the full article, but the headline makes an accusation that the decennial census was rigged. That would require explaining why large numbers of Census Bureau staff---who (at least in my own experience) tend to be professional statisticians who deliberately cultivate a workplace culture of sincere impartiality and transparency---suddenly all went along with a partisan effort. Furthermore, it would also require an explanation for why someone would rig the census but not also rig the post-enumeration survey. Altogether this does not sound like a "simplest direct explanation" to me.A simpler explanation for the undercount in Republican-leaning states would be differential nonresponse, wouldn't it? The Republican president was quite open about trying to appeal to voters who distrust institutions. It follows that Republicans could have been more likely to distrust the census and thus less likely to reply.This also lines up with AAPOR's evaluation of why 2020 pre-election presidential polls underestimated the Republican vote share:"This hypothesis is not unreasonable, considering the decreasing trust in institutions and polls especially among Republicans (e.g., Cramer 2016). Trump provided explicit cues to his supporters that polls were "fake" and intended to suppress votes (e.g., Haberman 2020). These statements by Trump could have transformed survey participation into a political act whereby his strongest supporters chose not to respond to polls."https://www.aapor.org/Education-Resources/Reports/2020-Pre-Election-Polling-An-Evaluation-of-the-202.aspx
The simplest explanation seems to be the oldest one, namely that the census has always had a pattern of undercounting minorities and overcounting whites (and this time asians), and very much tied to having a stable address at which one could be reached. (So if those wealthy enough to have 2 houses & receive a census form for each, or college students studying away from home, are more likely to be double counted).
Normally a lot of effort goes into reaching out to people who did not respond to household surveys, or were not associated with a household, and many of us will recall how the Trump administration refused to extend the timelines for the Census to carry out that work of doing so (and made more difficult because of COVID). (And also recall that the administration fought to include a citizenship question on the census, which was expected to lower the response). At the time it was pointed out that Trump's decisions regarding the census would more likely hurt red states because of the demographics, and so it seems to have.