After the once-a-decade census is complete the Census Bureau conducts a follow-up to gauge the accuracy of the count. The follow-up of the 2020 census was delayed a bit because of the pandemic but the results are now out and they show that in 36 states the tally was pretty accurate. However in 14 other states the numbers were off. Six states had an undercount and eight states had an overcount.

That there were a roughly equal number of states with overcounts as undercounts seems reasonable but the geographical and political breakdown of where those happened is interesting:

The results showed that six states — Arkansas, Tennessee, Mississippi, Florida, Illinois and Texas — likely have a larger population than was officially counted.

Eight states likely have fewer residents than were recorded, the survey found: Hawaii, Delaware, Rhode Island, Minnesota, New York, Utah, Massachusetts and Ohio…

The post-mortem will not change the official state-by-state results of the census, which said that 331,459,281 people were living in the United States in 2020. Nor will it alter the allotment of seats in the House of Representatives or in state and local political districts, even though its findings arguably could have affected those decisions.

So, the bottom line is that it’s too late now for any of this to matter. Still, it’s hard not to notice the pattern. Arkansas, Tennessee, Mississippi, Florida and Texas are all red states. Here’s the breakdown of who these states send to the House of Representatives:

• Arkansas – 4 Republicans
• Florida – 16 Republicans, 11 Democrats
• Mississippi – 3 Republicans, 1 Democrat
• Tennessee – 7 Republicans, 2 Democrats
• Texas – 23 Republicans, 13 Democrats (though one, Filemon Vela, resigned in March)

That leaves one outlier in the undercount category, Illinois, which sent 5 Republicans and 13 Democrats to the House. Still, it’s basically red states, including two of the largest states in country, where the undercounts happened. The size of the undercounts was pretty significant is some cases.

Those error rates are percentages. So a 3.5% error rate in Florida (population 21,070,000) works out to nearly 3/4 of a million people. As you can see above, Florida has had 27 House seats but under reapportionment based on the 2020 Census it gained a seat and will now have 28. That means the number of people per seat is about 3/4 of a million people.

Texas already did gain two seats in reapportionment. It had a smaller estimated undercount (as a percentage) but it has more people so in that case we’re talking about well over half a million people. And of course there’s a margin of error on these estimates. If the undercount was in the upper range then you could be talking about another House seat worth of people.

On the other side of the equation you have the opposite story with Hawaii, Delaware, Rhode Island, Massachusetts and New York all blue states. Minnesota leans blue and Ohio and Utah are red states. But again, it seems noteworthy that one of the largest blue states in the country had an overcount. Here’s the estimates of how big the overcount was in each state.

New York’s overcount was about the same size (as a percentage) as Florida’s undercount. New York did wind up losing a House seat but it was very close.

New York’s congressional delegation will shrink by one seat after the 2022 election, the Census Bureau announced on Monday, but the state came excruciatingly close to snapping an eight-decade streak of declining representation in Washington: It was 89 residents short, to be precise.

The figure, revealed during the Census Bureau’s announcement of which states would gain and lose seats in the House of Representatives after the decennial population survey, meant that if New York had counted 89 more people last year than the 20,215,751 who were tallied, it would have held on to the House seat. Instead, it went to Minnesota, which came close to losing a seat.

In fact, New York’s actual population probably missed the cut off by around 600,000 people. Again, none of this will matter.

During a press briefing on Wednesday, bureau officials emphasized the follow-up survey’s estimates, which were delayed largely because of the pandemic, will not change each state’s share of representation in the House or the Electoral College.

The Supreme Court ruled in 1999 that statistical sampling, which would be needed to factor in the states’ over- and undercount rates, cannot be used to produce the census data for reapportioning Congress.

As for the obvious question, why this happened the way it did, the Census Bureau doesn’t seem to know. And while apportionment won’t change, they are considering using the error estimates to guide future funding.

UPDATE: Census Bureau has "pulled together a team within the population division" dedicated to figuring out if/how 2020 census over/undercounting rates can be used to improve population estimates that guide funding, official Karen Battle told Census Scientific Advisory Committee https://t.co/HabGBtiOr9

— Hansi Lo Wang (he/him) (@hansilowang) March 17, 2022