The Official 2020 Census Thread

[Tender Branson]

Is it theoretically possible to delay the finalization/presentation of the 2020 results to 2021, so that census takers can visit non-responding or rural households during the fall and the data then gets finalized during winter/spring next year ?

Usually, the results need to be made available by December 31 of the Census year ...

[Tender Branson]

Daily Response Rate Update:

33.1% (+1.5)

Top:

40.8% Wisconsin

Bottom:

19.7% Alaska

Among the big states, TX and NY are trailing the national rate by 5%. NC by 3.

All other big states are roughly in line with national response rates (or higher).

🇺🇸 USA 🇺🇸 wide, the response rate is ca. twice as high (!) right now, compared with the same time in 2010.

Also, half the 66.5% nationwide rate is now already in.

[Tender Branson]

If the trend keeps up, 2/3 of households in the US will have submitted their data by the end of April, passing the final 2010 rate.

But households still have until the end of May to respond.

Current trends indicate a 80-85% response among US households (up by 13 to 18 points compared with 2010) - which would be an all-time record.

[jimrtex]

WV is already pretty bad, but McDowell County has a 0.5% response rate so far.

Are there even any people left, or just abandoned housing units ?

I guess most of WV doesn’t have internet either to get counted.

Or they don't have home mail delivery. About 95.4% of housing units are being contacted by self-response, where an invitation is sent to a residence (not a resident). The resident can respond by internet, or request a phone call or paper form. In some cases, the initial invitation will include a paper form. This is used in areas with low internet access, or language or other issues (the paper form is bilingual English-Spanish, double sided with English on one side, and Spanish on the other.

In about 4.5% of the housing units (around 5.9 million) update-leave is used. In this method, primarily for housing units where mail delivery is not to the location of the housing unit. a census enumerator leaves a form as well as an invitation to use the internet.

But you will see that this type of enumeration has been delayed due to COVID-19.

2020 Census Operational Adjustments Due to COVID-19

Originally scheduled from March 15-April 17, it has been delayed from March 29-May 1.

Even under the best of circumstances, only about 13/34 of households would have been contacted. Because of the delay, none have. A person can respond online, by entering their address - but realistically how many people would know about that? TV or radio stations might not mention that "some" listeners won't be getting forms on schedule, particularly since they might not have gotten it by now anyhow. The Welch News is quite small, and the county government is only open for emergencies. Perhaps the library is closed.

This app shows the different Type of Enumeration Areas (TEAS). If you zoom in far enough, you can see the percentage of housing units in each census tract that utilize update-leave. McDowell County is the southernmost county in West Virginia.

2020 Census: Type of Enumeration Area (TEA) Viewer

I estimate only 0.4% of housing units got an invitation by mail. Based on households/housing units about 31% of housing units are vacant. Remembe McDowell has lost 80% of its post-WWII population. Even assuming that households are 1/2 the size of then (4.6 vs 2.3 persons/household), the number of households (and therefore need for housing has declined by 60%). You can take about 42% of housing units out of service, and still have 31% vacancy.

Thanks,

but why are these households not getting mail in the first place ?

Isn’t US mail supposed to deliver in all places - except in overly rural areas such as Alaska or Wyoming ?

McDowell County is very poor, but not exactly „very rural“, so I don’t understand why 99.5% of households there get no mail, not even once a week ...

I also read on Census Bureau’s „Quick Facts“ that 65% of county households indeed have Internet access, so the response rate should be higher.

But if they got no mail invitations in the 1st place ... not.

I read a lot about Update Leave (UL), Type of Enumeration Area (TEA), and Basic Collection Units (BCU) but don't really understand the criteria.

A BCU is an area that can be assigned to a field enumerator, whether for the initial contact or a non-response followup (NRFU). The concept is that enumerators can be assigned dynamically by supervisors, and that supervision will be remote, rather than face-to-face. I think enumerators can work somewhat like a gig employment, where an enumerator can indicate their availability, and that visits can be optimized geographically and temporarily. You don't want to make an in-person visit when no one is home, and you want to minimize travel time between visits (don't want an enumerator to be lost, or knocking on doors when the resident is at work), since the enumerator is paid by the hour.

BCU are supposed to have some homogeneity, so that various assumptions can be made about how they will respond. This is probably highly correlated with income, language, race, citizenship, age, etc. In 2010, an average enumerator did about 200 households, so I think BCU may be relatively small. Remember that in most areas it is hoped that only 20% (made up by me, I think this is typical) of housing units will need in-person followup. A 10-block area might only need 40 follow-up visits.

All of a BCU have a common TEA. It thus appears that there may be housing units that could receive their initial invitation by mail that get dragged into UL, and that this happened in places like McDowell.

In 2010, the equivalent concept to UL was based on city-style street addresses. Rural roads may not have names, and there is really no need for street addresses, which were largely developed for postal delivery. If you go to a small town of 2000, there is no reason to ask for a street address, when you are looking for Mr. Jones. You ask someone along the street, especially if you are afoot, on horseback, or driving a wagon.

But during the last couple of decades there has been an increase in development of street-style addresses for use with 911 (emergency phone service for ambulance, fire, or police). In areas developed on a PLSS grid, the roads in a county might be numbered 1st, 2nd, ... for north-south roads, A, B, C for east-west roads (in some places these might be named with letters of the alphabet). Addresses might be based on distance along a road, with a house 0.23 miles east of 3rd Road as 323 B Road.

The USPS might or might not use the same system. On Rural Routes they would use Box Numbers, which would generally be in sequence, but might be assigned in an ad hoc basis.

But not everyone gets mail at their home. If you are in a rural area, you might not want anything of value placed in your mail box. It is easier to go into town and pick up your mail at the post office.

Anyhow, Update Leave designates the method of initial contact. An enumerator goes to each potential residence, and knocks on the door. If someone answers, they hand the person a packet, which includes a printed census form, but also includes an invitation to participate by mail. If possible, they also get a mailing address. They also verify locations - their device includes a GPS locator.

The objective of the census is to count everyone, count them only once, and count them in the correct location.

They can also delete units, and correct locations, add units.

Some of the quality control is based on enumerators deleting a large share of units, claiming that they were enumerating at large distances from where the house is supposed to be, doing several addresses while at the same location. High scores will trigger a review. Medium scores might trigger a sampled review. Low scores might trigger a sampled review at low frequency.

After the initial in-person visit, there will be additional mailings just like for the self-response areas where the there are five additional mailings. If the house is not visible from the street, or an unfriendly resident, they can be put on the follow-up program immediately.

It may be the large number of vacancies in McDowell County that pushing into Update Leave. If nobody responds because there is no one there, it is better to find that out initially than to send someone out later.

[jimrtex]

WV is already pretty bad, but McDowell County has a 0.5% response rate so far.

Are there even any people left, or just abandoned housing units ?

I guess most of WV doesn’t have internet either to get counted.

In about 4.5% of the housing units (around 5.9 million) update-leave is used. In this method, primarily for housing units where mail delivery is not to the location of the housing unit. a census enumerator leaves a form as well as an invitation to use the internet.

But you will see that this type of enumeration has been delayed due to COVID-19.

2020 Census Operational Adjustments Due to COVID-19

Originally scheduled from March 15-April 17, it has been delayed from March 29-May 1.

Even under the best of circumstances, only about 13/34 of households would have been contacted. Because of the delay, none have. A person can respond online, by entering their address - but realistically how many people would know about that? TV or radio stations might not mention that "some" listeners won't be getting forms on schedule, particularly since they might not have gotten it by now anyhow. The Welch News is quite small, and the county government is only open for emergencies. Perhaps the library is closed.

They should not have delayed that specific update-leave operation IMO:

If it’s just about leaving an invitation letter or paper questionnaires at a household, they can drive around from one house to another and drop off the materials there in the mailbox or on the porch - without having any contact with people, spreading the virus.

The same is true for later census taking: the 0.5 million census takers need to wear gloves and masks and keep 6 feet distance with the interviewed person and everyone’s OK.

Completely suspending the followup interviews would do incredible damage to the whole 2020 Census, because it would increase the polarization further: well-off households respond and get counted, urban/rural poor households, inner-city migrant areas & homeless are not.

The Census Bureau has delayed field operation another two weeks (until April 14).

Census Bureau Update on 2020 Census Field Operations

The original schedule for self-response was:

First letter: March 12-20
Second letter: March 16-24

If no response:

Post Card: March 26-April 3

Letter with paper questionnaire April 8-16

Final postcard April 20-27

Non-response follow-up was not expected until May 13.

The letters are spread out over nine days to lower workload for the post office, and traffic on the census bureau website, and for their telephone help lines.

I received my second letter several days after I had responded. I know someone who had received their post card on March 26.

Originally the Update Leave was scheduled from March 15-April 17. Perhaps, this would be tapering out, since you can't assume consistent rates of contact among all 50,000 enumerators. While you might be able to move someone to the next county, you can't move them to another state. If you assume that most will have been contacted by earlier in April, then you can work on the stragglers in mid-April.

IIUC, once the initial contact was made, UL households would be treated the same as the Self-Response areas, except there would be no need to send out a paper form on the fourth contact. This would have meant the final post card would be from April 23-May 26.

Since the UL was worked on a geographical basis (by BCU) you can overlap the NRFU with the last of the UL initial contacts.

The Census Bureau is also hiring more temporary workers, some to replace workers who drop out (workers probably skew older - since those who are under 50 can find full-time employment in a 3% economy). There may also be some parents of school-age children who no longer are at school during the day. But they can also assign workers to reduce the extended completion.

Meanwhile, a worker at the Jeffersonville, Indiana (across the river from Louisville) has tested positive for COVID-19, and the facility is undergoing a deep clean. This has been the main form-processing center since at least 1960.

[jimrtex]

Is it theoretically possible to delay the finalization/presentation of the 2020 results to 2021, so that census takers can visit non-responding or rural households during the fall and the data then gets finalized during winter/spring next year ?

Usually, the results need to be made available by December 31 of the Census year ...

They try to hit March or early April for the PL 94-171 redistricting data. The 94 refers to the 94th Congress (1975-6), and not 1994. This is the law that requires the Census Bureau to coordinate with the States in producing data for redistricting. This has resulted in producing data at a block level to facilitate fine-tuned gerrymandering.

The December 31 is only state level used for apportioning representatives to Congress.

Many state legislatures only meet in the spring. A March 2021 release of data would permit redistricting to be completed in time for filing deadlines which are as early as December 2021.

There will likely be controversy over whether to impute data. 15% of Americans move in any given year. As you push later and later, fewer neighbors will know who lived next door, or be sure when they moved out.

3rd-party records may help. The IRS will know who filed in 2019 for example. But such tracking gets expensive.

Maybe as a result of the 2020 Census, a national registry and ID card will be implemented.

[jimrtex]

This shows the type of initial contact for self-response areas.

2020 Census: Mail Contact Strategies Viewer

Some areas received a paper form with the first mailing. These respondents could either respond by internet, mail-in paper form, or telephone.

In the internet first areas, they would have to request a paper form to be sent to them or wait until one arrives with the 4th communication (beginning on April 8).

I was looking at the county response rate for Missouri. These are becoming more interesting as counties begin to pass the 40% threshold. There are a couple of rural counties in NW and SW Missouri that have 40%+ response, with it almost all on paper.

[Tender Branson]

Daily Response Rate Update:

34.7% (+1.6)

Top:

42.7% Wisconsin

Bottom:

20.8% Alaska

[Tender Branson]

Daily Response Rate Update:

34.7% (+1.6)

Top:

42.7% Wisconsin

Bottom:

20.8% Alaska

4 WI counties are already at 51%+ response and are now shaded blue on the county map.

3 of them are in the Milwaukee suburbs.

Also, 1 in IA and 1 in NE are blue now - but surprisingly none yet in MN.

[Storr]

Most of WV is classified as an UPDATE/LEAVE area - which explains the low response in the Southern part.

It also explains the lagging rates in AK, MT, WY, NM and the Latino parts of TX.

It would be pretty bad if those regions were not visited by census takers because of the virus.

On the other hand, it is compensated by a higher response rate in other areas.

(Right-click for huge map)

McDowell County West Virginia, the county with the lowest life expectancy of all counties in the United States at 71.5 years for women and 63.5 years (!!!) for men, as well as the highest rate of drug related deaths of all counties in the United States at 141 per 100,000, currently has a total response rate of 0.6%! Though, I suppose this shouldn't be all that surprising at its total self response rate for the 2010 Census was only 38.2%.

[jimrtex]

Daily Response Rate Update:

34.7% (+1.6)

Top:

42.7% Wisconsin

Bottom:

20.8% Alaska

4 WI counties are already at 51%+ response and are now shaded blue on the county map.

3 of them are in the Milwaukee suburbs.

Also, 1 in IA and 1 in NE are blue now - but surprisingly none yet in MN.

Those in IA and NE are fairly rural, and a large response was by mail. In some areas, the initial invitation included a paper form. Respondents could still respond by Internet or phone. But it appears they were not uncomfortable doing the paper form, and it was not a hardship to mail it back - they are postage page.

A definition of Midwest might be areas with widespread high rates of census participation:

NE, MN, IA, WI, IL, IN, MI, OH.

ND, SD, KS are lower because large parts are Update Leave which has not began yet.

Congressional districts that are over 40% are 26D, 24D, and 1I (MI-3, Amash).

[Tender Branson]

Daily update:

36.2% (+1.5)

My guess for tomorrow is 38% for the data submitted until end-March (there could be a final rush before reference date April 1, printed on the letters).

That would put the rate on track for 70% by the end of April.

And 75%+ by end-May.

[Tender Branson]

What is the most likely scenario after May, when the online/mail/phone response dries up and around 75% of households have responded ?

Will the Census Bureau just wait until the Coronavirus is over in the US by September/October and only then send out their census takers for the non-response households ?

Will they do earlier in June/July, but with protection gear ?

Will they send additional mail reminders in June, July and August - that were previously not planned ?

Will they try calling non-responding households ? But where do they get the phone numbers from ?

More TV ads ? More digital ads ?

[Tender Branson]

Looking at the map above, I find it fairly interesting that both Iowa and Tennessee (large, rural states) are almost completely shaded blue - indicating almost 100% US-mail coverage and almost no yellow update/leave areas.

Even in Appalachian Tennessee virtually every household gets served by the US postal service.

[jimrtex]

Is it theoretically possible to delay the finalization/presentation of the 2020 results to 2021, so that census takers can visit non-responding or rural households during the fall and the data then gets finalized during winter/spring next year ?

Usually, the results need to be made available by December 31 of the Census year ...

It is a statutory requirement that apportionment be delivered within 9 months of Census Day (April 1 to December 31). Congress could conceivably change this.

Through 1910, Congress would receive the population data, and then determine the number of representatives and the apportionment method.

Following the 1920 Census, Congress could not decide on a number of representatives or an apportionment. Agricultural areas that had been filling up had stalled during the droughts of the 1910's. Industrialization had results in population increases in cities in the east. They couldn't prevent losses by expanding the House, so they did nothing. There were also concerns about dislocations due to the World War (I).

Around 1929, President Hoover pushed for determining a new method of apportionment in advance of the Census, so that it could be done automatically (and would be done at all). There were actually two methods permitted, but they produced the same result.

After the 1940 Census, there would have been a difference, so Congress chose the method that favored the Democrats (a seat in Alabama rather than Michigan).

Ever since it has been automatic, except when Alaska and Hawaii were added just prior to the 1960 Census. They were given a representative each resulting in a total of 437, increasing the size of the electoral college from 531 to 537 for the 1960 Election. But the House reverted to 435 after the 1960 Census, producing an Electoral College of 535. The 23rd Amendment was passed in 1961, so there never was an election with 535 electors,

[jimrtex]

Daily update:

36.2% (+1.5)

My guess for tomorrow is 38% for the data submitted until end-March (there could be a final rush before reference date April 1, printed on the letters).

That would put the rate on track for 70% by the end of April.

And 75%+ by end-May.

13 counties (or independent cities) are now over 50%.

6 in Wisconsin. Wow!
3 in Iowa
1 in Illinois
1 in Nebraska
1 in Virginia (Fairfax City)
1 in New Mexico (Los Alamos)

Best areas are in the Midwest, areas with a strong tech sector, and suburban counties.

[Tender Branson]

Daily update:

36.2% (+1.5)

My guess for tomorrow is 38% for the data submitted until end-March (there could be a final rush before reference date April 1, printed on the letters).

That would put the rate on track for 70% by the end of April.

And 75%+ by end-May.

13 counties (or independent cities) are now over 50%.

6 in Wisconsin. Wow!
3 in Iowa
1 in Illinois
1 in Nebraska
1 in Virginia (Fairfax City)
1 in New Mexico (Los Alamos)

Best areas are in the Midwest, areas with a strong tech sector, and suburban counties.

Also:

Idaho, which had a comparable return rate to the US as a whole in 2010, is now responding 4% above the US average.

I think that is significant (the same in Mississippi), because until recently ID was more known to be a government-hostile rightwing retreat.

[Tender Branson]

It's Census Day. Here's what you need to know about the 2020 count

(CNN) - Households across the United States have been receiving invitations to complete the 2020 census for weeks. But another milestone in the count is here.

April 1 is known as Census Day.

And around the country, local officials are encouraging residents to complete census questionnaires this week.

"Lots of people are home right now, as you should be. But the truth is, it's also a great time to fill out your census," Michigan Gov. Gretchen Whitmer said in a video message.

The results of the 2020 census will impact the lives of people around the country. And everyone living in the US plays a role in shaping them.

Here are some key things you need to know:

(...)

https://edition.cnn.com/2020/03/31/us/census-2020-things-to-know/index.html

[Tender Branson]

Daily response map update (data submitted until yesterday, end-March):

38.4% (+2.2)

TOP:

46.4% Minnesota & Wisconsin (for the first time, MN is ranked 1st)

BOTTOM:

23.6% Alaska

---

TX and NY are still lagging the nation by 5%, while other big states such as CA and FL are just a small amount behind the nation. IL, OH, PA, MI, VA are all ahead of the nation.

https://2020census.gov/en/response-rates.html

[Tender Branson]

Several large urban and suburban counties have already surpassed 50% response by now, such as Ada (Idaho), Macomb, Bay, Midland and Livingston (MI), virtually all of Minneapolis's suburbs (but not Hennepin and Ramsey).

Also St. Charles (MO, St. Louis suburbs) and Holmes (OH).

Morgan and Davis in UT.

And in WI a total of 14 counties already, among them Dane and the Milwaukee suburbs.

[Cashew]

What metric do they use to calculate what percentage of households are resounding?

[Tender Branson]

What metric do they use to calculate what percentage of households are resounding?

If you look at the map I posted on top of the page, you see the classifications.

There are ca. 135 million households in the US right now, of which 95.5% are designated self-response households.

That means that 95.5% of all households in the US were sent invitation letters to respond first by internet (but they also get the chance to respond by phone. Or after repeat letters to respond online, they get paper forms mailed. Or you can call the hotline and request a paper form).

In 4.5% of US households, the update/leave is used - mostly rural areas not served by US Postal service. A census employee was supposed to show up at those households and leave a notice to respond or a paper questionnaire. This is not the case right now because of the virus, but they could be visited and counted in the fall.

All other rural areas only account for 0.1% of the US population (Alaska and Maine).

[Tender Branson]

What metric do they use to calculate what percentage of households are resounding?

Technical documentation here and here (FAQs).

[jimrtex]

Looking at the map above, I find it fairly interesting that both Iowa and Tennessee (large, rural states) are almost completely shaded blue - indicating almost 100% US-mail coverage and almost no yellow update/leave areas.

Even in Appalachian Tennessee virtually every household gets served by the US postal service.

These are characteristics of Update Leave areas.

UL can occur in geographic areas that:
- Do not have city-style addresses.
- Do not receive mail through city-style addresses.
- Receive mail at post office boxes or at drop points.
- Have been affected by major or natural disasters.
- Have high concentrations of seasonally vacant housing

On Page 15 of this document is the comparable map for 2010.

2020 Census Detailed Operational Plan for: 35. Update Leave Operation (UL) (pdf)

Audubon County, Iowa has named streets (100th to 350th from North to South, with an increment of 10 per section). The streets from west to east are named for birds (Avocet to Swift). If on the section line it is an Avenue, on a half-section line Place. The towns: Audubon, Exira, etc. have their own historic streets. If you go to Google Earth, the road grid is obvious from a long way out, with farm locations obvious perhaps five acres set aside for a house, barns, and other outbuildings, with the remainder of the section in crops (corn or soybeans).

There are two tiny areas in Audubon County that are Update Leave. The one in Kimballton appears to be a small apartment complex with perhaps a dozen or so apartments. Maybe they don't have individual mail delivery. There is another area in the city of Audubon that appears to have been a trailer park. If you look at historic pictures using Google Earth there were 20 or so single-wides under trees in low resolution USDA images. As you go forward in time, there are fewer and fewer. One winter picture shows just trees. And now it appears that it a new industrial building.

You may be focusing on the "Leave" and not on the "Update". Ideally, the Census will survey everywhere that a person might live. But they have to have an accurate inventory of all housing, and it really can't be done in a haphazard fashion. Before the 2010 Census, census workers walked the entire country matching addresses with GPS locations. Sometime addresses get mixed up or are irregular. Before the 2020 Census, they compared satellite images and determined only about 30% of the country needed to be updated due to changes.

McDowell County, West Virginia does not have a street grid, with roads running along the creeks. They might get mail on a mail route where the address is a box number, or perhaps they have a box in a post office. It is possible that the large number of vacancies is triggering update/leave. It is unclear how you measure "seasonal vacancies"? Perhaps it is any month with over X% of housing units vacant. In McDowell this is January-December.

[Tender Branson]

Daily response map update (data submitted until yesterday, end-March):

38.4% (+2.2)

TOP:

46.4% Minnesota & Wisconsin (for the first time, MN is ranked 1st)

BOTTOM:

23.6% Alaska

---

TX and NY are still lagging the nation by 5%, while other big states such as CA and FL are just a small amount behind the nation. IL, OH, PA, MI, VA are all ahead of the nation.

https://2020census.gov/en/response-rates.html

27 states are now responding above the national average, 23 below (+DC, +PR).

4 are less than 1% below the national average, might catch up quickly: CA, FL, NV, DE.

MI has already reached 66% of its 2010 response rate, with 2 months to go.

Another 17 states have reached more than 60% of their 2010 rates already.

6 states have not even reached 50% of their 2010 rate yet.