The Census and ACS paint significantly different pictures. Somebody’s wrong.
For decades, the dominant trend in US demography has been rapid growth in major metro areas in the South and West, and stagnation in the Northeast and Midwest. For each of the past 10 years, annual estimates from the American Community Survey suggested that this trend was continuing at a significant rate, with northern states like New York, New Jersey and Pennsylvania seeing little-to-no change in their overall population, while southern states with large metros like Georgia, Arizona and Florida grew by a percentage point or more each year.
But the official 2020 Census state-level results, released Monday, suggest that the magnitude of this trend, particularly in the northeast, slowed in the past decade, and was greatly exaggerated by recent ACS estimates.
An unexpected bump for the Northeast
Almost every state in the northeast saw a significant boost from the 2019 ACS. In raw numbers, the largest bumps were New York, New Jersey and Pennsylvania, which posted a combined population 1.3 million higher than the 2019 ACS, and 1.5 million higher than the 2020 ACS, which showed a decrease from 2019 that pundits attributed to a Coronavirus exodus from dense cities.
All in all, the northeast region, defined here as New England plus every state that includes part of the NYC, Philly or DC metro area, saw a 2.6% increase from the 2019 ACS to the 2020 Census. That may not sound like much, but it’s remarkable given that the 2019 ACS showed only a 2.3% increase from the 2010 census.
Downwards revisions in the Sun Belt
On the flip side, booming Sun Belt states like North Carolina, Florida, Texas and Arizona mostly saw downwards revisions from their 2020 ACS figures, and some even saw downwards revisions from the 2019 ACS, while others saw only slight increases.
All in all, the “booming sun belt” states, a group I define as {AZ, FL, GA, NV, NC, SC, TX} saw a combined increase of just 0.14% from the 2019 ACS, after seeing a 13.3% increase from the 2010 census to the 2019 ACS.
The Great Lakes Region gets a Mini-Bump
The great lakes region is similar to the northeast in many ways, and it also fared better-than-expected in the 2020 census, though to a lesser degree.
Together, the Great Lakes states, defined as {OH, IN, IL, MI, WI, MN} saw a 1% boost from the 2019 ACS to the 2020 census. Again, doesn’t sound like much, until you consider that these same states saw only a 1.6% increase from the 2010 census to the 2019 ACS. And that 1% boost is more than 5 times greater than that of the booming Sun Belt states that have been draining population from the Midwest for decades.
A Hispanic undercount wouldn’t explain everything
Immediately following the census release, there was widespread speculation of a possible undercount of Hispanic populations, or immigrant populations writ-large. This is understandable given Trump’s attempts to add a citizenship question to the Census, and given that the three states that received 1 fewer Congressional District than expected were Florida, Texas and Arizona.
However, a systemic undercount of Hispanic or immigrant populations does not fit with all of the above trends. For starters, New York and New Jersey have significant Hispanic populations, and larger immigrant population shares than Florida, Texas or Arizona. Despite this, they significantly overperformed expectations.
In addition, southern states with lower Hispanic population shares, such as the Carolinas, also saw underwhelming census counts, more underwhelming than those of Florida and Texas, when compared to the 2019 ACS.
The most consistent pattern is not “lower than expected counts for states with large Hispanic populations”, it’s “lower than expected counts for booming sun belt states, higher than expected counts for northern states”.
Who’s wrong?
There’s virtually no plausible mechanism by which both the 2020 census and the past few ACS estimates could both be correct within a reasonable margin of error. So the basic question is, which one is more wrong than the other?
Possibility 1: The ACS overstated north-to-south migration
One possibility is that the ACS was consistently wrong over the past few years, by overestimating the degree to which people were relocating from the likes of New York, Philadelphia and Chicago to Sun Belt metros.
Possibility 2: The Census is Wrong
Another possibility is that the ACS was telling the correct story, and the Census is way off. The challenges the pandemic presented for the Census Bureau seemingly make this more plausible than it otherwise would be.
However, it’s difficult to explain how the pandemic would cause the above trends in the 2020 census counts, while still allowing for 2020 ACS estimates that looked even more grim for many northern states than the 2019 ones.
I don’t know much about census methodology, and don’t have much of an opinion on who’s wrong. I’d like to hear from others who do. And we’ll find out more as more granular data is released. But as a fan of the North, my hope is that the American Community Survey was a clown institution for the past several years.
