USA 2020 Census Results Thread (Release: Today, 26 April)

[jimrtex]

Columbus’ borders are worse than Madison’s! I didn’t think that was possible. Madison does, however, have border agreements in place with neighboring communities so that within the next 15-20 years or so they will follow roads and township borders to be clean. Is Columbus working on that too? Boy the city borders are nearly a crime against my eyes. Sheesh.

Ohio closed the barn door about 3 decades after the horse had left.

There is some sort of law that counties have to approve annexations - though it might not be discretionary if it meets certain requirements. Franklin County has approved about a dozen annexation this year, but most are small. The two largest were around 30 acres. Many were under an acre.

There were also three agreements to adjust boundaries between Columbus and other cities. I didn't see maps. These might be tweaks to get all of a road into one city or the other, which makes maintenance and policing simpler.

There are also some agreements where annexation to Columbus does not mean removal from a township. One problem with annexation is that it removes the territory from the township's tax base. If they have bond indebtedness it may make it difficult to service the debt.

Ohio redistricting law disfavors splitting of wards. It also disfavors splitting of cities and townships but that is not possible because of the size of Columbus. Columbus has lots of "wards", but doesn't use them for electing its city council. But they are drawn to help comply with the redistricting law. If you have a fragmented township forming islands, a Columbus ward will be drawn around the archipelago, so you can include all of the islands and that one ward in a district.

There is a similar system in Cleveland, where the wards are used for city council elections. They are drawn so you can connect to smaller cities on the edge of Cleveland.

[jimrtex]

Alright, here are some of the trends I noticed overall about this census:

2. Sun Belt ain't so hot:

This seems to be the story after every single census. For all the reports about huge growth in the sunbelt, it always seems to be overexaggerated. The estimates for areas such as San Antonio and Phoenix were noticeably bad, but in general most urban, suburban, and rural areas down south underperformed.

Bexar had been estimated to grow at 17.9% over the decade, and the Census only shows 17.1%.

That is not noticeably bad.

[jimrtex]

Anyone know how to convert a .pl file into a .XLSX or .CSV file?

The easiest way is to add an ".txt" to the end of the file name, and read into Excel.

Then say that the file is delimited with a pipe character/vertical bar ( | ).

What specifically are you trying to get?

You will likely be getting a lot more than you want.

If you just want total population for counties or cities in a state it is pretty easy.

[jimrtex]

Smaller Texas counties were overestimated. This appears to particularly pronounced in Oil producing areas such as the Permian Basin, Eagle Ford, and Austin Chalk.

It might be that the estimates were catching people living in temporary housing: Trailers, Hotels, and man camps. Towards the end of the decade this might have tailed off. The estimates of migration are largely based on income tax returns which lag population (e.g. 2020 tax returns are filed in 2021 and may not be processed until 2022).

An example is Zavalla which estimates showed increasing from 11.7K in 2010 to 12.3K in 2015 to 11.8K in 2020, but which the Census showed as 9.7K.

That is outside the range from 1930 to 2010. There might be significant migrant labor who were missed due to the delay in the Census.

Census Pop	Deviation
1M+	-0.38%
500K-1M	0.01%
200K-500K	-0.44%
100K-200K	0.05%
50K-100K	0.62%
20K-50K	-2.42%
10K-20K	-3.25%
5K-10K	-4.45%
2K-5K	-7.72%
1K-2K	-9.48%
0-1K	-7.16%

There may be systemic error, but it may be highly localized.  The largest counties are down, but only slightly. Some but not all ring suburban counties were underestimated.

County	Population	Deviation
Coryell	83093	8.3%
Bastrop	97216	6.8%
Walker	76400	6.0%
Navarro	52624	4.1%
Van Zandt	59541	3.9%
Harrison	68839	3.6%
Taylor	143208	3.0%
El Paso	865657	2.9%
San Patricio	68755	2.7%
Kaufman	145310	2.6%
Starr	65920	2.5%
Jefferson	256526	2.4%
Orange	84808	2.2%
Potter	118525	1.9%
Galveston	350682	1.8%
Guadalupe	172706	1.7%
Ellis	192455	1.2%
Hays	241067	1.0%
Randall	140753	0.9%
Liberty	91628	0.9%
Parker	148222	0.8%
Brazos	233849	0.8%
Johnson	179927	0.6%
Bell	370647	0.6%
McLennan	260579	0.5%
Hunt	99956	0.4%
Lamar	50088	0.4%
Anderson	57922	0.1%
Collin	1064465	0.1%
Tom Green	120003	0.1%
Gregg	124239	0.0%
Harris	4731145	-0.1%
Nacogdoches	64653	-0.2%
Waller	56794	-0.2%
Montgomery	620443	-0.2%
Travis	1290188	-0.3%
Hidalgo	870781	-0.4%
Kerr	52598	-0.4%
Tarrant	2110640	-0.4%
Williamson	609017	-0.4%
Angelina	86395	-0.6%
Denton	906422	-0.6%
Bexar	2009324	-0.6%
Victoria	91319	-0.7%
Bowie	92893	-0.7%
Smith	233479	-0.7%
Cameron	421017	-0.7%
Rockwall	107819	-0.8%
Comal	161501	-0.8%
Maverick	57887	-0.9%
Dallas	2613539	-0.9%
Lubbock	310639	-1.1%
Fort Bend	822779	-1.3%
Ector	165171	-1.3%
Grayson	135543	-1.7%
Henderson	82150	-1.8%
Brazoria	372031	-1.9%
Hood	61598	-2.4%
Nueces	353178	-2.8%
Medina	50748	-2.8%
Wichita	129350	-2.8%
Wise	68632	-3.0%
Hardin	56231	-3.4%
Webb	267114	-3.8%
Rusk	52214	-4.0%
Midland	169983	-4.3%
Cherokee	50412	-4.6%
Polk	50123	-4.6%

[jimrtex]

Alright, here are some of the trends I noticed overall about this census:

2. Sun Belt ain't so hot:

This seems to be the story after every single census. For all the reports about huge growth in the sunbelt, it always seems to be overexaggerated. The estimates for areas such as San Antonio and Phoenix were noticeably bad, but in general most urban, suburban, and rural areas down south underperformed.

Bexar had been estimated to grow at 17.9% over the decade, and the Census only shows 17.1%.

That is not noticeably bad.

I thought it was pretty clear, since I used the city names of Phoenix and San Antonio instead of Maricopa and Bexar, that I was referring to the city of San Antonio, and not to Bexar county.

Here are the stats for anyone curious.

San Antonio projected growth: 16.6%
https://www.census.gov/programs-surveys/popest/data/tables.2019.html

San Antonio actual growth: 8.1%
https://www.kens5.com/article/news/local/census-breakdown-us-texas-satx-2020/273-ff46cb3a-3d7a-4430-8453-75235cc667d7

Personally, I'd say that seeing only half of the growth projected is noticeably bad, but to each their own.

San Antonio is dominant in Bexar County, with around 77% of the population, a share that according to estimates was maintained through the decade.

I had downloaded the Texas county results so used those.

If we believe the Census results then about 120,000 people disappeared from San Antonio, but 110,000 ended up in Bexar County outside the city.

There may be an error in how population estimates are distributed (city estimates are derived from county estimates adjusted for housing unit changes).

https://www2.census.gov/programs-surveys/popest/technical-documentation/methodology/2010-2020/2020-subco-meth_final.pdf

Or conceivably differences in how residence location are determined were a problem.

In any event I sent a note to the mayor of San Antonio suggesting looking deeper.

There is conceivably a similar situation in Phoenix. How close was the estimate for Maricopa? And what about the other cities in the county?

What evidence is there that urban, suburban, and rural areas underperformed enough to make any sort of generalization.

Of the largest counties in Texas > 50,000, 31 surpassed the estimate, 36 fell short and one (Gregg) was almost perfect.

[jimrtex]

Fun fact - Yakutat borough AK has the exact same population it did in 2010: 662.

10 years older?

[jimrtex]

Eagle Ford bust.

Relative to 2020 estimate:

Zavala -18.1%
McMullen -17.7%
Dimmitt -13.7%
Duval -11.4%
La Salle -11.3%
Frio -9.7%
Uvalde -8.2%
Jim Hogg -6.5%
Atascosa -5.2%
Jim Wells -4.0%
Webb -3.8%
Nueces -2.8%

The estimate methodology may overestimate "residence" or perhaps missed people moving out. Data used for migration includes IRS returns which lag.

There is not a lot of housing in the area, so people are commuting long distances or living in hotels or RVs or mobile homes or man camps. While multiple wells are drilled from a single pad there is still a lot of moving around to different job sites.

[jimrtex]

Interactives:

County:
https://cinycmaps.com/index.php/2-uncategorized/94-2020-census-county-population-change-map


County Sub (still somewhat experimental. There are a few issues with the data set):
https://cinycmaps.com/index.php/2-uncategorized/95-2020-census-county-sub-population-change-map

Have you done a comparison of 2020 estimate to 2020 Census?

Notable in your county map is how far the shadow of DFW, Houston, San Antonio, Austin extends outward. This is either exurbs or people retiring in the country.

[jimrtex]

Interactives:

County:
https://cinycmaps.com/index.php/2-uncategorized/94-2020-census-county-population-change-map


County Sub (still somewhat experimental. There are a few issues with the data set):
https://cinycmaps.com/index.php/2-uncategorized/95-2020-census-county-sub-population-change-map

Have you done a comparison of 2020 estimate to 2020 Census?

Yes. Click Menu-->vs Apr 20 Estimate (linear backdated July Estimate) or vs July 20 Estimate on the interactive. The County Sub version includes fewer geos because county subs outside the Northeast and Midwest are generally not included in the PEP estimates dataset.

The Census Estimates have both an April 1, 2020 and July 1, 2020 estimates (but maybe they simply interpolated and you knew that?).

Interesting they were way off for places like Cape Cod and the Islands. That suggests they were picking up people at their summer homes. The same thing is true in northern Wisconsin and the Adirondacks (Hamilton). Back in '020, these areas were notable for their low response rate which was based on responses per housing unit. If nobody was there in April, no one would respond. Someone from middle class Milwaukee might not be able to fully retire, but could run an ice cream stand up North, and stay through hunting season and the Holidays, and then head South for the winter.

The worst underperformers in Arizona were Pinal and La Paz, which likely have a larger share of winter visitors. There are actually people who live in Phoenix year round, as unlikely as that seems when it is 110F.

A couple of fun examples. Concho County, Texas.

2010	4101
2011	4121
2012	4082
2013	4118
2014	4080
2015	4065
2016	4134
2017	2701
2018	2679
2019	2749
2020	2827

There was (is) a private prison in Eden which had a federal contract that was not renewed at the end of the contract. The 2017 estimate reflects the closure. But the Census shows a population 3303.

Terrell County

2010	1007
2011	945
2012	917
2013	887
2014	906
2015	863
2016	815
2017	810
2018	792
2019	752
2020	702

It used to be the case that days of work for railwaymen was measured in miles rather than hours. Southern Pacific crews would work until Sanderson, when a new crew would get on and continue. The first crew would overnight in Sanderson and then work the return train the next day. As this practice has ended Sanderson has been dwindling away. In addition, transcontinental road traffic uses I-10, and few still use US 90.

The Census population was 760. It turns out the county had only lost 1/4 of its population, rather than 1/3.

[jimrtex]

Which Census estimates had April 1 estimates? I don't think it was in the PEP release from May -at least in May.

There was another release in July. I can only find them by clicking through the July 27 press release at:

https://www.census.gov/programs-surveys/popest.html

https://www.census.gov/newsroom/press-releases/2021/vintage-2020-populaton-evaluation-estimates.html

Then click on "evaluation estimates"

I just check the data set for Texas counties, and the April 1, 2020 estimate is not 3/4 of the way between July 1, 2019 and July 1, 2020.

I assume they don't consider the "evaluation estimates" the standard continuation of 2020 estimates, but a special version which they can compare with the Census.

BTW, I don't know if you remember back in '020 and Alabama had a relative high return for a southern state. It looks like a public effort to return forms may have saved the 7th congressional seat. Notice the Black Belt on the county map.

[jimrtex]

Interactive for Census Places, this time with a background map. Alaska is up top due to this.

https://cinycmaps.com/index.php/2-uncategorized/96-2020-census-place-population-change-2010-20

It is so mouse over Houston, TX. It is like when they use a fluorescent marker that binds to cancer cells.

I think there may be a systemic bias in how they distribute county estimates to cities within a county.

According to this (PDF) they create the county-level estimates using demographic methods, and then divvy it up among cities.

If you go to counties with lots of cities, you may find some that are overestimated, and others that are underestimated. The overestimated tend to be older cities that are landlocked and not many opportunities to add housing units.

This first showed up in Bexar County (which only has one city). Bexar County was overestimated by less than 1% bus San Antonio was overestimated by 8% and outside the city underestimated by 25% (about 3/4 of the county is in the city). So there allocation method was moving people into the city.

Maricopa: Phoenix and Scottsdale overestimated, while Buckeye underestimated.

Los Angeles: Santa Clarita, Lancaster, Palmdale underestimated.

Cities that have space for new housing units tended to be underestimated.

Cook (Illinois): Though Chicago was underestimated, all the cities outside the city were even worse.

[jimrtex]

Interactive for Census Places, this time with a background map. Alaska is up top due to this.

https://cinycmaps.com/index.php/2-uncategorized/96-2020-census-place-population-change-2010-20

With all the data you've collected thus far, is there anyway to "unify" a national map (i.e. include unincorporated areas' population changes along with the above map)?

That would involve math to figure out the "balance of" each geo. I'm not sure whether that's even possible. Census didn't provide that data in the release. I'd have to calculate it on my own.

To be clear, I was referring to the unincorporated remainder at the county/parish/borough level (and not each tract, block, etc). Maybe this is what you inferred and maybe it would still be difficult, but I just wanted to clarify. Minus any cities that may cross county borders, I would think this would be relatively simple to calculate.

There are some incorporated places that cross county lines. I'd have to think about how to calculate that. That should be in the dataset, but I'd have to figure out the summary level.

Doing what I'd like to do - within County Sub - seems to not be available, but I'll have to double check.

155 State-Place-County

It is kind of inverted, but you want intersections of Place and County anyway.

How close are the CDP's between the the two Censuses?

[jimrtex]

Interactive for Census Places, this time with a background map. Alaska is up top due to this.

https://cinycmaps.com/index.php/2-uncategorized/96-2020-census-place-population-change-2010-20

With all the data you've collected thus far, is there anyway to "unify" a national map (i.e. include unincorporated areas' population changes along with the above map)?

That would involve math to figure out the "balance of" each geo. I'm not sure whether that's even possible. Census didn't provide that data in the release. I'd have to calculate it on my own.

To be clear, I was referring to the unincorporated remainder at the county/parish/borough level (and not each tract, block, etc). Maybe this is what you inferred and maybe it would still be difficult, but I just wanted to clarify. Minus any cities that may cross county borders, I would think this would be relatively simple to calculate.

There are some incorporated places that cross county lines. I'd have to think about how to calculate that. That should be in the dataset, but I'd have to figure out the summary level.

Doing what I'd like to do - within County Sub - seems to not be available, but I'll have to double check.

155 State-Place-County

It is kind of inverted, but you want intersections of Place and County anyway.

How close are the CDP's between the the two Censuses?

Yup. I found and calculated the data. I'm getting ready to update the map.

How close the CDPs are between the 2 Censuses I don't know. There may be boundary changes, and the raw 2010 Census data won't pick that up. I'd have to use the April 2010 estimates base, which I've been loathe to do, except to fill in blanks due to no 2010 pop. And the 2010 estimates base doesn't have CDPs at all, unless you know of a source other than the PEP.

When I re-make the county sub map with the "official" April 1, 2020 estimates, I'm going to have to think about all of this. But I will update the incorporated place map with county remainders soon.

I don't think the Census Bureau makes estimates for CDP's.

I think for comparing Census 2020 with Estimates 2020 you will have to filter incorporated cities on the PLACECC field. All "Cn" are incorporated places. All "Un" and "Mn" are CDP's.

I'd then include all the unincorporated areas in balance of county.

You can compare CDP between 2010 and 2020, though I think there may be a major mismatch between the two. CDP's may become incorporated or decimated by annexation.

I think that the Census Bureau has grown indifferent to CDP's over time. They started in the 1950s when they started defining urban areas, which was when growth began outside cities. Previously, cities tended to expand to include all developed territory. You might have the central cities and a few suburban cities, and then some areas that looked like cities as far as land use.

But now that they have census tracts and census blocks and urban areas being defined automatically, the Census Bureau could care less about defining statistical cities. If the States want to do so they are free to do so. Now a county might want CDP's for all the colonias or settlement areas. It is hard enough to keep up with official annexations. It is hard to define boundaries of CDP's from the Census Bureau perspective. It requires too much context.

So I was thinking it might be fairly hard to match 2010 CDP's to 2020 CDP's..

I would be kind of curious to compare the 2020 Estimates to 2020 Census for counties where cities are substantial (say 20% to 80%) and the counties had substantial growth 10%+. I think the way they estimate city populations is wrong.

[jimrtex]

At Jim's suggestion, I updated the Place with Remainder map to show Counties minus incorporated places as the balance of counties for the estimates comparison map (now using the official April estimates, I think). I haven't fixed the shapefile to take out the CDPs yet. Pretend they don't exist for now. Maybe later.

There is a consistent divergent distribution in Texas, where the cities were overestimated, while the area outside the cities were underestimated.

I think there is a systemic error in Census Bureau methodology.

Do you mind if I pass this along to the PEP people?

[jimrtex]

Is there a list of population of the outgoing congressional districts?

The PL 94-171 data is tabulated for congressional districts (and state legislative districts)

The Top 10. After redistricting, DE-AL will be tops followed by WV-1 and WV-2.

MT-AL	1,084,225
DE-AL	989,948
TX-22	972,309
FL-9	955,602
ID-1	954,891
TX-26	943,106
TX-10	937,982
TX-31	933,772
TX-3	933,008
TX-8	916,386

The Bottom 10. After redistricting the two MT districts will be bottom, below the two RI districts, and the the three Nebraska districts.

ME-2	658,148
SC-6	646,463
VT-AL	643,077
WV-2	621,081
WV-1	602,044
NE-3	600,683
WY-AL	576,851
WV-3	570,591
RI-1	554,307
RI-2	543,072

The closest statewide range are:

HI: 4391
RI 11K
NH 17K
NM 19K (3)
OR 40K (5)
ME 46K
CT 47K (5)

The largest statewide ranges are:

TX: 264K
FL: 228K
VA: 188K
GA: 186K
SC: 172K
NC: 171K
NV: 152K
AR: 152K (it was pretty surprising to find AR on the list)

[jimrtex]

More detailed info on the next Census products that will be released:

Demographics and Housing Characteristics File (DHC)

Release Date May 2023

The DHC will include many of the demographic and housing tables previously included in the 2010 Census Summary File 1 (2010 SF1). Some tables are repeated by race and ethnicity.

Subjects: Age, sex, race, Hispanic or Latino origin, household type, family type, relationship to householder, group quarters population, housing occupancy, and housing tenure.

Access: data.census.gov.

Lowest level of geography: Varies, with many tables proposed at the census block level.

Detailed Demographic and Housing Characteristics File A (Detailed DHC-A)

Release Date August 2023

Subjects: Population counts and sex by age statistics for approximately 370 detailed racial and ethnic groups, such as German, Lebanese, Jamaican, Chinese, Native Hawaiian, and Mexican, as well as about 1,200 detailed American Indian and Alaska Native tribal and village population groups, such as Navajo Nation.

Access: data.census.gov.

Proposed 2020 geographies: Nation, state, county, and American Indian/Alaska Native/Native Hawaiian (AIANNH) areas. We’re evaluating the feasibility of adding places (cities and towns) and census tracts, per data user feedback.

This was from an April 2022 pre-release. The August 2023 was the forecast then. With more detailed information there is a greater risk of revealing individual responses. Supposedly a prototype based on 2010 data was to be release in January 2023. I could not find any such prototype (until now January 31).

This will be the first census data that permits a breakdown based on racial origin. I suspect the Census Bureau was trying to have an informal MENA category by suggesting "Lebanese" as subtype of White. Among the write-in choices were German, Irish, English, Italian, Lebanese, and Egyptian. This could eventually supplant the ancestry question.

Product Will Provide Detailed Racial and Ethnic Groups and American Indian and Alaska Native Tribes and Villages

Since they were using 2010 data, they were limited to Asian, NHPI, and AIAN groups where the Census Bureau had detailed subgroups. Much of the report are about statistical methods - in essence they want to know if this data is good enough for you to use.

This xlsx shows the type of detail they may have.

Proof_of_Concept_Examples (xlsx)

For the Navajo Nation (286,531 persons) they provided 5-year subgroups for the USA and Arizona. For Texas, it was 10-year subgroups. For Idaho it was broad groups (Under 18, 18-44, 45-64, 65 and up). For Alaska it was simple total count.

For Singaporean (5,347 persons) they provided 5-year subgroups for the USA and California. Kosraen (906 persons had broad categories), Tokelauan (107 persons had only head counts).

[jimrtex]

Image Link

-Mormons
-Amish
-Native Americans
-A smattering of plains counties with tiny populations?

Only certain types of Native Americans (the Navajo Nation for example does not stand out at all - even though most of the population in those counties that is neither Navajo nor Hopi is Mormon).

As others pointed out, some Hispanic populations (particularly farmworker areas in the Central Valley and the Tri-Cities and meatpacking workers in SW Kansas - maybe also that spot on the Iowa/Missouri border?) stand out as well.

Some odd patches in the South too that don't seem immediately explicable. And Ocean County, New Jersey stands out likely because of Hasidic Jews.

Patterns in the South look a bit suggestive of a correlation with the military.

Forts Benning, Campbell, Polk, and Hood stand out. It may also be true for Fort Drum and Riley though not in the South.

Williston Basin and Permian Basin.