The Official 2020 Census Thread

[jimrtex]

We can conclude that the 4th mailing to the self-response went out on time, or near to it.

No, we cannot.

I guess the NYT has insider information from the Census Bureau.

I guess that the April 8-16 mail questionnaires did NOT go out to all non-response households yet.

I guess that only some areas or states got them, but others not.

You only base your findings on MN, but not other states ...

I think there could be a major delay if they are understaffed right now.

Have you ever heard of the term "Afghanistanization" with respect to the NYT? This term goes way back, likely before the Russian takeover, where the NYT would send a reporter to a place in Central Asia, or Africa, or the US west of the Hudson, and explain it to their readers.

Would you prefer that I substitute, "Intelligent persons will conclude that the 4th mailing to the self-response went out on time, or near to it."

The Census Bureau contracts out their printing, and envelope stuffing is automated. The selection of addresses that have not responded is computerized, and most responses have been by the internet. The Census Bureau is not deploying the temporary field workers at this point. Large scale mailings don't use these workers. There is no reason that this would vary based on state.

Here are some results from other states, I chose the 6th, 16th, 26th, 36th, and 46th best performers (i.e. medians of the quintiles). The questionnaire at the 4th mailing was scheduled to be sent out from April 8 to April 16, so we are not seeing the full effect given the date of mailing, delivery time, response time, mailing-back time.

While internet responses dropped by half, other responses doubled this past week.

You can make you own conclusions.

Washington
Date	Total	Internet	Not Internet	Weekly Internet	Weekly Non Internet
19-Mar	13.6%	12.8%	0.8%	12.8%	0.8%
26-Mar	32.6%	31.1%	1.5%	18.3%	0.7%
2-Apr	46.7%	44.9%	1.8%	13.8%	0.3%
9-Apr	52.3%	50.1%	2.2%	5.2%	0.4%
16-Apr	55.4%	52.3%	3.1%	2.2%	0.9%
Oregon
Date	Total	Internet	Not Internet	Weekly Internet	Weekly Non Internet
19-Mar	14.3%	13.0%	1.3%	13.0%	1.3%
26-Mar	31.6%	29.2%	2.4%	16.2%	1.1%
2-Apr	44.5%	41.5%	3.0%	12.3%	0.6%
9-Apr	49.9%	46.1%	3.8%	4.6%	0.8%
16-Apr	52.8%	48.0%	4.8%	1.9%	1.0%
Nevada
Date	Total	Internet	Not Internet	Weekly Internet	Weekly Non Internet
19-Mar	15.5%	14.0%	1.5%	14.0%	1.5%
26-Mar	30.2%	27.5%	2.7%	13.5%	1.2%
2-Apr	41.9%	38.5%	3.4%	11.0%	0.7%
9-Apr	46.6%	42.6%	4.0%	4.1%	0.6%
16-Apr	49.9%	45.0%	4.9%	2.4%	0.9%
Georgia
Date	Total	Internet	Not Internet	Weekly Internet	Weekly Non Internet
19-Mar	11.4%	9.3%	2.1%	9.3%	2.1%
26-Mar	28.2%	23.8%	4.4%	14.5%	2.3%
2-Apr	40.5%	35.1%	5.4%	11.3%	1.0%
9-Apr	44.6%	38.8%	5.8%	3.7%	0.4%
16-Apr	47.2%	40.6%	6.6%	1.8%	0.8%
Wyoming
Date	Total	Internet	Not Internet	Weekly Internet	Weekly Non Internet
19-Mar	11.5%	10.1%	1.4%	10.1%	1.4%
26-Mar	23.2%	21.0%	2.2%	10.9%	0.8%
2-Apr	33.5%	30.8%	2.7%	9.8%	0.5%
9-Apr	37.9%	35.0%	2.9%	4.2%	0.2%
16-Apr	41.2%	37.1%	4.1%	2.1%	1.2%

[jimrtex]

50.5 (+0.3)

The final 2010 rate of 66.5% would be reached in ca. 50 days, if the average stays around 0.3 to 0.4 per day.

You may not have accounted for the correction in the denominator, nor the update/leave areas. We need additional information on the effect of the mailing of questionnaires.

[jimrtex]

Least Responsive Cities

In some cases, low rates may be due to second homes that are counted as possibly occupied on April 1. When several cities were tied, I chose the representative by lot.

1 to 9	Fontana Dam, NC	Graham	0.9%	The dam was built during WWII to supply electricity for Alcoa. It has now been converted to a resort, with between 106 and 177 cottages.
10 to 18	Pine Valley, NJ	Camden	0.0%	One of two towns with no response in this category. Often rated top golf course in country, with a very exclusive membership.
20 to 49	Rockham, SD	Faulk	0.0%	One of six towns with no response in this category. About 85 miles west of Watertown, with less than 1/10 of 1920 population
50 to 99	Wood Lake, NE	Cherry	0.0%	One of eight towns with no response in this range. Cherry is the big county in the Sandhills of northern Nebraska.
100 to 199	Batesland, SD	Oglala Lakota	0.0%	One of seven towns with no response for this size. On Pine Ridge reservation.
200 to 499	Barstow, TX	Ward	0.0%	One of three cities with no response in this category. Barstow is largest non-responsive town (2018 est. 380)
500 to 999	St. Francis, SD	Todd	0.5%	Tied with War, WV. On Rosebud reservation, with 1/200-odd responses.
1K to 2K	Spring Valley, UT	Sanpete	0.7%	Tied with Boley, OK. According to Forbes one of prettiest towns in America, with particular Danish Morman flavor.
2K to 5K	Quartzite, AZ	La Paz	1.7%	RV's congregate at Quartzite in winter. Low response rate may be due to confusion over which housing units are permanent.
5K to 10K	Big Bear Lake, CA	San Bernardino	3.5%	Lots of second homes, as population in area can swell to 100,000 on summer weekends.
10K to 20K	Jackson, WY	Teton	10.1%	Located in Jackson Hole, with lots of second homes for the ultra-rich.
20K to 50K	Paradise, CA	Butte	6.2%	May have lost 90% of its population in the 2018 Camp Fire, when 19,000 buildings were destroyed.
50K to 100K	Miami Beach, FL	Miami-Dade	29.6%	Lots of second homes for New Yorkers, Europeans, and Latin Americans.
100K to 200K	Hartford, CT	Hartford	33.5%	Cities are limited in Connecticut, making an easy commute from suburban towns. Hartford is 84% Hispanic or non-white.
200K to 500K	Laredo, TX	Webb	33.1%	Border city, 96% Hispanic.
500K to 1M	Detroit, MI	Wayne	40.7%	Has lost 64% of its population (over 1,000,000) since 1950. 92% Hispanic or non-white.
1M to 2M	Philadelphia, PA	Philadelphia	40.9%	Has lost 1/4 of its population since 1950. 65% non-white or Hispanic.
2M to 5M	Los Angeles, CA	Los Angeles	41.6%	Three cities in this range are tightly bunched.
5M to 10M	New York, NY	Kings, Queens, New York, Bronx, Richmond	39.5%	USA's largest city made up of Brooklyn, Queens, Manhattan, The Bronx, and Staten Island boroughs.

[jimrtex]

^^^

Seattle is 92% Hispanic or non-white?

Oops! I changed the description, but didn't change the name.

Seattle has the best performance in the 500K to 1M category at 61.0%. Detroit is the worst at 40.9%.

[jimrtex]

50.7% (+0.2)

If you break out the non-internet results, you will see that this reflects non mail-delivery on Sunday. 4/20 should be (+0.4) with at least half by mail.

[jimrtex]

Best Responding Cities

New leaders in 1K to 2K and 2K to 5K as Hudson, IL regains its lead in the 1K to 2K range, and Howard Grove, WI enters the ranks for the first time. Another household responds in Trout Valley, IL. The biggest increases (of 0.6%) were seen in Seattle and San Jose.

1 to 9	McMullen, AL	Pickens	60.9%	Small all-black town outside Aliceville, west of Tuscaloosa. Likely 14/23 response.
10 to 18	Bonanza, CO	Saguache	4.8%	Almost ghost town survived effort to have its government abandoned. At one time had 36 dance halls and 7 saloons. Likely 1/21 response.
20 to 49	Harbine, NE	Jefferson	73.1%	Southwest of Lincoln, a bit north of the Kansas/Nebraska line. Likely 19 of 26 response
50 to 99	North River, ND	Cass	95.3%	North of Fargo along Red River. Likely 22/23 response.
100 to 199	Oakland Acres, IA	Jasper	83.6%	Exurban Des Moines. Most of town is golf course, but not too exclusive ($20 greens fee on weekends). Likely 58/67 response.
200 to 499	Broecke Pointe, KY	Jefferson	82.5%	11 miles northeast of Louisville. Likely 80/97 response.
500 to 999	Trout Valley, IL	McHenry	86.2%	Southeast part of county near Cook panhandle. Was modular home showcase (169/196 response)
1K to 2K	Hudson, IL	McLean	80.4%	Just above Normal. It is probably apocryphal that it named for the burg in New York.
2K to 5K	Howards Grove, WI	Sheboygan	80.4%	Farming community in the northern part of the county.
5K to 10K	Huntington Woods, MI	Oakland	83.7%	Immediately northwest of Pleasant Ridge, includes part of the Detroit Zoo.
10K to 20K	Dardenne Prairie, MO	St. Charles	78.9%	Northwest St. Louis suburb, incorporated in 1983, has octupled in population since 1990 Census.
20K to 50K	Andover, MN	Anoka	77.4%	Just north of Coon Rapids, with half of the population added since 1990.
50K to 100K	Maple Grove, MN	Hennepin	73.6%	Northwest of Minneapolis, about 20 minutes in light traffic.
100K to 200K	Centennial, CO	Arapahoe	72.5%	Southeast of Denver, incorporated in 2001 as most populous incorporation at that time.
200K to 500K	Boise City, ID	Ada	64.5%	State capital, "city" is part of official name.
500K to 1M	Seattle, WA	King	61.6%	18th largest city, and fastest growing over 500,000 since 2010, slightly exceeding Austin.
1M to 2M	San Jose, CA	Santa Clara	59.7%	High-tech Silicon Valley giant, areas along southern edge are over 75% response rates.
2M to 5M	Chicago, IL	Cook, DuPage	44.4%	Three cities in this range are tightly bunched.
5M to 10M	New York, NY	Kings, Queens, New York, Bronx, Richmond	40.0%	USA's largest city made up of Brooklyn, Queens, Manhattan, The Bronx, and Staten Island boroughs.

[jimrtex]

Least Responsive Cities

In some cases, low rates may be due to second homes that are counted as possibly occupied on April 1. When several cities were tied I have chosen those shown by lot.

1 to 9	Fontana Dam, NC	Graham	0.9%	The dam was built during WWII to supply electricity for Alcoa. It has now been converted to a resort, with between 106 and 177 cottages.
10 to 18	Dering Harbor, NY	Suffolk	0.0%	Two towns in this category. Summer homes on Shelter Island off eastern end of Long Island
(do)	Pine Valley, NJ	Camden	0.0%	Often rated top golf course in country, with a very exclusive membership.
20 to 49	Rendville, OH	Perry	0.0%	Six towns in this category, including these two. Smallest village in Ohio, with 4% of 1890 population.
(do)	Rockham, SD	Faulk	0.0%	About 85 miles west of Watertown, with less than 1/10 of 1920 population
50 to 99	Alexander, KS	Rush	0.0%	Eight towns with no response in this range. Trading post on Santa Fe Trail. Namesake Alexander Harvey's daughter married William Bent.
(do)	Wood Lake, NE	Cherry	0.0%	Cherry is the big county in the Sandhills of northern Nebraska.
100 to 199	Anawalt, WV	McDowell	0.0%	One of seven towns with no response for this size. In 2010, 1/3 of housing units were vacant in coal mining town.
(do)	Batesland, SD	Oglala Lakota	0.0%	On Pine Ridge reservation, Lakota name of Íŋyaŋ Šála translates to "Red Stone".
200 to 499	Barstow, TX	Ward	0.0%	One of three cities with no response in this category. Barstow is largest non-responsive town (2018 est. 380)
(do)	Bryce Canyon City, UT	Garfield	0.0%	Incorporated in 2007 under quirky law that permitted incorporation by owners of 50% of land value, in this case resort owner Syrett family.
500 to 999	St. Francis, SD	Todd	0.5%	Tied with War, WV. On Rosebud reservation, with 1/200-odd responses.
(do)	War, WV	McDowell	0.5%	Mining town with 1/6 of 1950 population, with 2/400-odd responses.
1K to 2K	Spring Valley, UT	Sanpete	0.7%	Tied with Boley, OK. According to Forbes one of prettiest towns in America, with particular Danish Morman flavor.
(do)	Boley, OK	Okfuskee	0.7%	Founded by Creek Freedmen (former slaves) in 1905 when railroad was established.
2K to 5K	Quartzite, AZ	La Paz	1.7%	RV's congregate at Quartzite in winter. Low response rate may be due to confusion over which housing units are permanent.
5K to 10K	Big Bear Lake, CA	San Bernardino	3.6%	Lots of second homes, as population in area can swell to 100,000 on summer weekends.
10K to 20K	Jackson, WY	Teton	10.3%	Located in Jackson Hole, with lots of second homes for the ultra-rich.
20K to 50K	Paradise, CA	Butte	6.3%	May have lost 90% of its population in the 2018 Camp Fire, when 19,000 buildings were destroyed.
50K to 100K	Miami Beach, FL	Miami-Dade	29.8%	Lots of second homes for New Yorkers, Europeans, and Latin Americans.
100K to 200K	Hartford, CT	Hartford	33.7%	Cities are limited in Connecticut, making an easy commute from suburban towns. Hartford is 84% Hispanic or non-white.
200K to 500K	Laredo, TX	Webb	33.3%	Border city, 96% Hispanic.
500K to 1M	Detroit, MI	Wayne	40.8%	Has lost 64% of its population (over 1,000,000) since 1950. 92% Hispanic or non-white.
1M to 2M	Philadelphia, PA	Philadelphia	41.2%	Has lost 1/4 of its population since 1950. 65% non-white or Hispanic.
2M to 5M	Los Angeles, CA	Los Angeles	41.8%	Three cities in this range are tightly bunched.
5M to 10M	New York, NY	Kings, Queens, New York, Bronx, Richmond	40.0%	USA's largest city made up of Brooklyn, Queens, Manhattan, The Bronx, and Staten Island boroughs.

[jimrtex]

50.5 (+0.3)

The final 2010 rate of 66.5% would be reached in ca. 50 days, if the average stays around 0.3 to 0.4 per day.

You may not have accounted for the correction in the denominator, nor the update/leave areas. We need additional information on the effect of the mailing of questionnaires.

As of yesterday, 75 million households (or housing units ?) have submitted their data.

That's 50.7% of all households/units.

Which means 100% would be 147.93 million - which is different to the 150+ million during the 2019 address listing operation:

Another chart:

150.448 / 75 million = 49.85% response.

Considering the US added more housing units since 2019, the response rate is closer to 49%.

US Quick facts

The 150 million housing units is an upper estimate for census planning purposes. Note that 138 million was the 2018 estimate. 2018 building permits were 1.3 million. Some would be replacement, others would cause housing units to no longer be viable housing stock.

150.448 x 50.7% is 76.277 million households have responded (there is no one to respond in vacant housing units). 76.277 / 119.773 households = 63.7% of households have responded. It will be a bit lower due to increase in households since 2014-2018 ACS. This is due to overall population growth and reduction in household size due to improved economy (if you have a job and are in your late 20's, you don't have to live with your parents, etc.) long-term declines in family sizes, and widowing of baby boomers.

You can represent a 1-person household. There is no such thing as a 0-person household.

119.7 mn x 2.63/household = 314.89 mn in households. Add the 8.0 mn in group quarters and you get close to the US population.

At some point, the Census Bureau will correct the number of housing units.

They still have to account for the housing that no longer exists (see response rate for Paradise, CA).

[jimrtex]

So how much longer until we have enough of the Census results in until we know how the House will look for the 2020's?

And to have complete-enough updated demographic data to make maps like this:
https://en.wikipedia.org/wiki/American_ancestry#/media/File:Census-2000-Data-Top-US-Ancestries-by-County.svg

The Census no longer has the long form which was sent to a large sample (about 1 in 6 households). The long form had detailed questions including ancestry. It was last used for the 2000 Census. The long form produced data that was statistically significant for block groups of around 1000 persons. But it would be out of date. Detailed data from the 2020 census won't be released until 2021 or 2022.

Instead it has been by the American Community Survey (ACS) in which the equivalent to the long form is sent monthly to a sample of households. Data from 12-months or 60-months is aggregated to produce statistically significant results.

Let's say that the long form was sent to 1/6 of 120 million households in the US, or 20 million households. Now instead, the ACS is sent out to 1/60 of those households, or 333,000 each month. Since it is being sent out on a regular basis, followup can be much quicker.

The Census Bureau is faced with tracking down maybe 40 million households who didn't respond to the 2020 Census in April, and doing in June, July, ...

They end up asking the neighbors. Do you remember the people who lived in that brick house across the street? "I just moved here in June" or "There were always people going and coming, I don't think that half of them actually lived there. They might have dealing."

You can take the data over a five-year period, say 2012-2016, and aggregate it, and use it much like the data from the long form. If the data is changing relatively slowly, you can treat the data as being representative for the midpoint of the period, in this case 2014.

The next year, you discard the 2012 data, and add in the 2017 data. So while the data might be a bit fuzzier because of the long period of collection (like a time-lapse photo rather than a snap shot) it may also shows trends better. The latest released ACS data is for 2018. cinyc has produced ancestry maps similar to the 2000 map from the ACS,
Census results for apportionment purposes are supposed to be released by December 31. That is by statute. The Census Bureau has asked for a change in statute to delay this until April 2021.

The Census is designed to collect data through July and then the second half of the year to finish up processing the data to make the apportionment. But because of COVID-19 they are unable to get out into the field, where they actually knock on doors, trying to track down people who have not responded. Those who have not responded are likely to move more, be minority, be younger, have language barriers, etc. Some may be anti-government or suspicious of people snooping around and asking personal questions.

Homeless people were supposed to be counted on April 1, by volunteers going out to soup kitchens, overnight shelters, under bridges and overpasses. So now it might happen on June 1, and they will have to try to figure out where they were "living" on April 1.

The Census Bureau does not release intermediate data, so its not like election night. They have made estimates up through July 2019, so it is possible to project the apportionment. But it will be pretty close for the final seats.

[jimrtex]

So how much longer until we have enough of the Census results in until we know how the House will look for the 2020's ?

Usually, the Census is to be completed by Dec. 31st (President gets the data).

But this year, the Census Bureau is seeking an extension because of the Virus (read last few pages).

Congress meets again in May or June and will likely approve the extension of 4 months - meaning that Trump will very likely never see the results because they are out on April 30 (assuming he loses).

It also allows for more time to self-respond (well into the fall), before census takers hit the streets.

Mail Response/Return Rates Assessment (for 2010 Census)(pdf)

There are tables at the end of this document that show the daily response rates. In 2010, the initial mailing included a questionnaire. While there was the possibility of Internet response, it was limited and somewhat experimental. Because forms had to be mailed back, there was a plateau each Sunday (the Census Bureau logged check-in dates that could lag actual date in the mailroom by up to 48 hours).

Because of that, the weekly dates shown were a Saturday in 2010, which happens to be a Friday in 2020. I made no attempt to reconcile the two censuses to account for operational differences.

Date	2020 Cumulative	2010 Cumulative	2020 Weekly%	2010 Weekly	Weekly Gap
20-Mar	16.7%	12.5%	16.7%	12.5%	-4.2%
27-Mar	31.6%	42.5%	14.9%	30.0%	15.1%
3-Apr	43.9%	54.0%	12.3%	11.5%	-0.8%
10-Apr	47.5%	59.5%	3.6%	5.5%	1.9%
17-Apr	50.2%	62.3%	2.7%	2.8%	0.1%
24-Apr		64.6%		2.3%
1-May		65.3%		0.7%
8-May		65.8%		0.5%
15-May		66.1%		0.3%
22-May		66.2%		0.1%
29-May		66.3%		0.1%
5-Jun		66.4%		0.1%

The 2010 Census pounded the 2020 Census for the March 21 to March 27 period by almost a 2:1 margin. This may have been due to the Census news being lost in the COVID-19 news. There was no Census Day surge in 2010. March 30-April 1, 2020 saw 6.6% response compared to 4.2% for the same three dates in 2010, which were Tuesday through Wednesday.

The official initial response period ended on April 19, 2010, with a response rate of 63.5%  Forms continued to trickle in, and eventually reached 66.5%, but by late May was down to 0.1% per week.

In 2010, a second questionnaire was sent to some areas, using one of three strategies: (1) blanket, send a second questionnaire to everybody; (2) targeted, send a second questionnaire to non-respondents; and (3) no second questionnaire. The second questionnaire boosted returns by a few percent. If households returned both forms, only the first was counted.

In 2020, the mail-out of the questionnaire should have concluded on April 16, so most of any effect would be seen by now. There has been a small increase in the share of responses by paper, but not enough to drive overall turnout. Last week through the 17th, the weekly response was 2.7%. In the first three days of this week 0.8% responded.

The 2020 Census appears to have a more spread-out design. In 2010, almost half of those who self-responded did so in the single week before Census Day (30.0% / 66.5% = 45.1%). In 2020, the 5th mailing of a postcard is just now going out (April 20-27). But it might not have high expectations. Imagine that you hoped for a 60% initial response by now, but found that a 5th mailing would finally get through to some people, let's say 2%.

But if it costs $1 to send out that postcard, and $50 to send someone out to knock on the door - remember that for some this might require multiple trips, and asking the neighbors, and travel time, then for every 100 houses, you would spend $40 for the post card mailing, but save $100 in enumerators.

The 2020 Census Has been hurt by the absences of update leave. If all responses have come from self-response TEA, then the rate for these areas is 51.0 / 95.4 = 53.5%. But even if the update/leave program had been conducted on schedule, it would not have had the same yield.

There are more vacant houses in update/leave areas. In 2010, the final response rate in mailout/mailback (MOMB) areas was 67.5%. In update/leave (UL) areas it was 56.%. In urban update/leave (UUL) areas it was 51.4%. UUL were not a subset UL. UL were generally rural areas without street addresses.  UUL included areas without individual mailboxes.

The mail return rate was the percentage of forms returned after housing units that were vacant or deleted were removed from the denominator. That is, it is the percentage of households who responded voluntarily. There were a very tiny share of housing units where there was a response, but not a return (perhaps someone answered the mail, and responded nobody lives here). Perhaps someone in Florida responded that their usual residence was in New York.

The overall mail return rate was 79.2% (that is instead of 2/3 of households responding, almost 4/5 did. The Census Bureau still had to track down the 21 million housing units that were either vacant or did not exist, but that is presumably easier than tracking down people who failed to respond.

TEA	Mail Response	Mail Return	Occupancy
MOMB	67.5%	79.6%	84.8%
UL	56.2%	76.8%	73.2%
UUL	51.4%	69.5%	74.0%

The occupancy rate is the ratio of the response rate to the return rate. It is not a true occupancy rate, since the denominator for the response rate includes non-existent housing and housing not immediately suitable for use - if people moved back to a farming town which they won't, there is a house that has been kept closed up, but will need some work to make it habitable. The reason it is empty, is because there is better housing in the town. If there is little demand, you will choose the 1200 square-foot house built in 1950, over the 600 square-foot house built in 1910, which had running water and electricity added later.

People in the UL areas were almost as responsive as those in MOMB areas, but there were a lot more empty houses. So even if the response rate for self-response TEA is actually 53.5%, while that for UL TEA is 0.0%, that even under ideal circumstances the response rate for UL areas would be around 43.5%. Which would push the response rate back towards 53.0%.

It also appears that the Census Bureau is targeting more more "housing units". In 2010, the denominator for the response rate was 130 million. For 2020 it is 150 million. Population growth of 7% will require 7% more houses. A decline in household size would push this a bit higher. There should be an increase in the overall housing stock. If someone moves from Michigan to Texas, there will have to be new housing built, even though this does not increase the national population. It does not mean that the unnecessary house in Michigan is torn down (in some cases it will. If the mover is from Oakland County, someone from Detroit will move into that house, and eventually the old building in Detroit will become a vacant lot). But a 15.4% increase in housing to be filled by 7% more people seems incredible to me. The Census Bureau is counting iffier housing. If there are 8 million extra units, that only have 20% occupancy, that is still 1.6 million persons, and worth counting, but you are also driving the response rate down by a couple of percent.

So even under equivalent circumstances, getting to 64.5% would be reasonable performance, and if we had the update/leave areas were would be around 53%. But for 2020 that is a 11.5% gap to fill. In 2010 at this time, they were within 2.8% of the final mail response rate.

2020 could have a larger increase if there is a massive publicity effort, but I think it is going to end up short of 60%.

An interesting tidbit from 2020:

The return rate based on the the age of the householder:

18-24 55.3%
25-44 70.9%
45-64 83.1%
65+    89.9%

There is no comparable response rate calculation since vacant and non-existent houses don't have householders.

[jimrtex]

(Post)

Thanks for your long explanation, but I guess 2010 and 2020 are not really comparable.

Whereas 2010 was almost exclusively paper questionnaires and 2/3 completed by end-April, we now have seen mostly internet responses for 52%.

The paper questionnaires have only been sent out to non-responding households and A) not all might have gotten it by now (understaffing due to Coronavirus) and B) not all that were returned might have been processed as a result (also understaffing and security measures).

So, the responses will trickle in at a higher rate for the coming weeks, compared to 2010 when they also trickled in ... but at a virtually non-existing rate after April.

Besides, 2020 will have almost an additional 6 (!) months for households to respond, something that was not the case in 2010.

The rate will eventually go up to 70-80%.

Printing and distribution of paper questionnaires has been contracted out. It makes no sense to maintain high speed presses capable of personalizing mail and addressing (each mail-out has the unique 12-digit code assigned to the address along with the address) for mailing every 10 years.

Selection of addresses to be mailed to as non-responsive is automated. With most responses by Internet it was trivial to determine which address should receive the 4th mailing which included a questionnaire. You don't need field workers for that. The mail receipt handling might be contracted out. Only when you get to the actual OCR would the Census Bureau take over.

Incidentally, the bump in responses for 4/21 was due to mailed questionnaires with about 65% coming in that way - estimated based on unweighted average of state results, which had considerable rounding.

The questionnaire mailing was scheduled for April 8-16. Assuming 3 days mail out time to April 11-19; 3 days cogitation time, until April 14-22; and 3 days mail back time, until April 17-25; 1 day in mail bins and UV irradiating, April 18-26. And that pretty well matches the increase in mail returns on April 21.

The 5th notice postcard is just now beginning to be sent out.

The official response deadline in 2010 was April 19, 2010, at which time the response rate was 63.5%. The Census Bureau continued to record responses until September, when it reached 66.5%. 2.3% of the 3.0% extra came in the first week after the deadline.

Date	2010 Weekly
20-Mar	12.5%
27-Mar	30.0%
3-Apr	11.5%
10-Apr	5.5%
17-Apr	2.8%
24-Apr	2.3%
1-May	0.7%
8-May	0.5%
15-May	0.3%
22-May	0.1%
29-May	0.1%
5-Jun	0.1%
…
7-Sep	0.1%

In 2010, people could respond until September. They didn't. They pretty much had finished up by May.

We can expect a tiny bump once the update/leave happens. We know the response rate for self-response TEA is around 54.1% (51.6% / 95.4%) since it would require a special effort for someone in an update/leave TEA to to respond. But the update/leave areas will have a lower response rate because they have much more vacant or deleted housing. In 2010 the mail return rate for UL TEA was only 2.8% less than for the mail out/mail back (MOMB) TEAs. But the mail response rate was 11.3%.

That is, actual humans in UL areas were almost as likely to fill out a form, but almost no vacant or non-existent houses responded. There might be socio-economic reasons that the UL areas had a tad lower return rate. Someone living in a trailer on a dirt road might be less inclined to fill out government paper work.

So we are effectively at a 53.5% response rate once we get the UL areas. But the Census Bureau appears to have added in more addresses.

Remember the Census Bureau's goal is to count everyone, count them only once, and count them in the right place.

After the 2010 Census (and every census) they will try to figure out who and why they missed. One reason would be housing units that were missed. It is unlikely that the number of housing units has increased from 130 million to 150 million (15.4%) while the population has increased by about 7%. The number of housing units will likely increase at a faster rate due to decreasing household sizes - more widowed baby boomers, smaller family sizes, extended adolescence of millennials, and maldistribution of housing units: people moving to Arizona, Texas, and Florida need new housing, the housing they left in Michigan or New York or West Virginia is not necessarily eliminated. But I doubt these can account for the 8.4% difference.

Since the response rate includes vacant and non-existent housing in the denominator, this produces a structural bias against 2020. 2020 could have an identical rate of households (people responding), but a lower response rate because of more vacant or non-existent housing.

Factors favoring 2020:

(1) Delay of update/leave.
(2) Slightly extended formal schedule.
(3) COVID-19 fog pushing everything else out of the public consciousness eventually clears some.

Factors not favoring 2020:

(1) More vacant housing increases denominator for response rate..
(2) COVID-19 means that 2020 was wiped out for many people.
The bump due to questionnaires being returned is encouraging. April 21 was tied with April 6 as biggest increase.

My prediction is 60% to 62%.

[jimrtex]

Best Responding Cities

North River, ND becomes first city to reach 100% response, as all 23 households have responded.

15 of 23 housing units in McMullen, AL have responded, which is a bit odd given an estimated population of 9. The town supposedly declined from 66 in 2000 to 10 in 2010, and then was leveled by an EF2 tornado in February 2016. The current housing looks quite modest but could easily be new reconstruction. I'm kind of wondering why only 10 persons were reported in 2010.

In the 200-499 category, Broecke Pointe, KY is joined by Northbourne Estates, KY another Louisville suburb, and Frontier, ND another Fargo suburb.

St. Regis Park goes back on top in the 2K-5K range.

1 to 9	McMullen, AL	Pickens	65.2%	Small all-black town outside Aliceville, west of Tuscaloosa. Likely 16/23 response.
10 to 18	Bonanza, CO	Saguache	4.8%	Almost ghost town survived effort to have its government abandoned. At one time had 36 dance halls and 7 saloons. Likely 1/21 response.
20 to 49	Harbine, NE	Jefferson	73.1%	Southwest of Lincoln, a bit north of the Kansas/Nebraska line. Likely 19 of 26 response
50 to 99	North River, ND	Cass	100.0%	North of Fargo along Red River. All 23 households have responded.
100 to 199	Oakland Acres, IA	Jasper	86.6%	Exurban Des Moines. Most of town is golf course, but not too exclusive ($20 greens fee on weekends). Likely 58/67 response.
200 to 499	Broecke Pointe, KY	Jefferson	83.5%	3 cities are tied. 11 miles northeast of Louisville. Likely 81/97 response.
(do)	Frontier, ND	Cass	83.5%	Southern enclave of Fargo. Acre lots, big garages, windbreaks rather than fences. 66/79.
(do)	Norbourne Estates, KY	Jefferson	83.5%	6 miles east of Louisville. Likely 147/176 responses.
500 to 999	Trout Valley, IL	McHenry	86.7%	Southeast part of county near Cook panhandle. Was modular home showcase (170/196 response)
1K to 2K	St. Regis Park, KY	Jefferson	81.9%	11 miles east of Louisville.
2K to 5K	Howards Grove, WI	Sheboygan	81.9%	Farming community in the northern part of the county, seeing modest suburban growth.
5K to 10K	Huntington Woods, MI	Oakland	84.2%	Immediately northwest of Pleasant Ridge, includes part of the Detroit Zoo.
10K to 20K	Dardenne Prairie, MO	St. Charles	79.4%	Northwest St. Louis suburb, incorporated in 1983, has octupled in population since 1990 Census.
20K to 50K	Andover, MN	Anoka	77.7%	Just north of Coon Rapids, with half of the population added since 1990.
50K to 100K	Maple Grove, MN	Hennepin	74.1%	Northwest of Minneapolis, about 20 minutes in light traffic.
100K to 200K	Centennial, CO	Arapahoe	73.1%	Southeast of Denver, incorporated in 2001 as most populous incorporation at that time.
200K to 500K	Boise City, ID	Ada	65.4%	State capital, "city" is part of official name.
500K to 1M	Seattle, WA	King	62.3%	18th largest city, and fastest growing over 500,000 since 2010, slightly exceeding Austin.
1M to 2M	San Jose, CA	Santa Clara	60.4%	High-tech Silicon Valley giant, areas along southern edge are over 75% response rates.
2M to 5M	Chicago, IL	Cook, DuPage	45.2%	Three cities in this range are tightly bunched.
5M to 10M	New York, NY	Kings, Queens, New York, Bronx, Richmond	41.0%	USA's largest city made up of Brooklyn, Queens, Manhattan, The Bronx, and Staten Island boroughs.

Largest city to have exceeded certain thresholds.
 

100%=	North River, ND	55
90%+	North River, ND	55
80%+	Huntington Wood, MI	6312
70%+	Naperville, IL	148,304
60%+	San Jose, CA	1,030,119
50%+	Phoenix, AZ	1,660,272

[jimrtex]

51.8 (+0.2)

Back to crappy increase levels ...

It is interesting that they report this is a 0.3% increase. They apparently are independently rounding the total and daily change.

That is, it might have been 51.56% and now 51.84%, for a 0.28% increase.

If they were rounded independently it would 51.6%, 51.8%, and 0.3% as shown on their web site.

The unweighted state average for yesterday was 30.2%/51 = 0.59%, and for today 14.9%/51 = 0.29%.

Yesterday the unweighted state internet average change was 0.38%, and for today 0.16%. The non-internet change was 0.21% and 0.14%.

The paper share went from 35% to 46%. It might be a paper majority tomorrow.

The Census Bureau might have tried to push internet responses since they are cheaper and there can be some error checking when they are being filled out. Paper forms will have to use handwritten OCR and have mistakes such as missed answers.

If they had sent a questionnaire initially, many people would just gone ahead and filled out the form. You (or I) can imagine some conversations like this:

Daughter: What are you doing?
Mom: Just filling out the census form, do you want to do your page?
Daughter: So, like, isn't there an app?
Mom: I don't know - those are so complicated. I'd probably just take a picture of my thumb.
Daughter: (texting to a friend, "MIAMI", apparently code for My Mom Is An Idiot"
Mom: There done! Would you like some hot cocoa?
Daughter: What are you doing. Why are you licking that paper? Eeww! That is so gross!

But imagine if only the internet invitation had been sent.

Mom: (tossing the invitation aside): Sigh, I wish they would send the form to start with.
Daughter: (picking up invitation, enters code and begins filling out form): What does re-side mean? Is that like is our house brick or stucco?
Mom: Reside. How many people live here.
Daughter: Oh (enters "3" then enters "Snowflake", "Mom" and "Dad").
App: "When was Snowflake born?"
Snowflake: "September 3, 2005".
App: "Is Snowflake 14 years old?
Snowflake (frowns, enters): "18"
App: "Was Snowflake born on September 3, 2001?"
Snowflake: Yes.
...
App: "What relation is Mom to Snowflake?"
Snowflake (scanning list): "Non-relative"
...
App: "What race is Mom?"
Snowflake: "Other"
App: "Enter specific race of Mom"
Snowflake: "Martian" (then deleting that. with smirk) "Uranusian."
...
Daughter: (Clicking on submit): All done. (looking at messages): I'm going over to Emily's.
Mom: Are her parents home?
Snowflake: "rents?"
Emily "They're in quarantine, I've locked them in the basement. We'll have the whole rest of the house."
Snowflake: Yes. Don't worry I'll social distance from them.

There will some who will wait for the paper form in the fourth mailing, and when it finally arrives will fill it out and return it.

[jimrtex]

That’s funny, Jimbo.

Anyway, it’s update time:

52.4% (+0.6)

The unweighted average for the last 3 days:

Tuesday: Internet .384%, Other .208%
Wednesday: Internet .157%, Other .135%
Thursday: Internet .196, Other .420%

Share of responses by other than internet (you can also respond by phone)

Tuesday: 35%
Wednesday: 46%
Thursday: 68%

Only three states had an internet majority: AK, MA, NC
Four plus DC were tied: UT, PA, VA, WV, DC

[jimrtex]

Best Responding Cities

Argusville, ND takes the lead in the competitive 200 to 499 category. Motto of Fargo is apparently "City of Big Garages".

Lincoln, NE surpasses Boise, in 200K-500K range.

1 to 9	McMullen, AL	Pickens	65.2%	Small all-black town outside Aliceville, west of Tuscaloosa. Likely 16/23 response.
10 to 18	Bonanza, CO	Saguache	4.8%	Almost ghost town survived effort to have its government abandoned. At one time had 36 dance halls and 7 saloons. Likely 1/21 response.
20 to 49	Harbine, NE	Jefferson	73.1%	Southwest of Lincoln, a bit north of the Kansas/Nebraska line. Likely 19 of 26 response
50 to 99	North River, ND	Cass	100.0%	North of Fargo along Red River. All 23 households have responded.
100 to 199	Oakland Acres, IA	Jasper	86.6%	Exurban Des Moines. Most of town is golf course, but not too exclusive ($20 greens fee on weekends). Likely 58/67 response.
200 to 499	Argusville, ND	Cass	87.5%	Another Fargo suburb on I-29 north of the city. Had a big boom between 2004 and 2008 when new subdivision was built (133/152).
500 to 999	Trout Valley, IL	McHenry	86.7%	Southeast part of county near Cook panhandle. Was modular home showcase (170/196 response).
1K to 2K	St. Regis Park, KY	Jefferson	83.7%	11 miles east of Louisville.
2K to 5K	Howards Grove, WI	Sheboygan	82.5%	Farming community in the northern part of the county, seeing modest suburban growth.
5K to 10K	Huntington Woods, MI	Oakland	84.6%	Immediately northwest of Pleasant Ridge, includes part of the Detroit Zoo.
10K to 20K	Dardenne Prairie, MO	St. Charles	80.2%	Northwest St. Louis suburb, incorporated in 1983, has octupled in population since 1990 Census.
20K to 50K	Andover, MN	Anoka	78.3%	Just north of Coon Rapids, with half of the population added since 1990.
50K to 100K	Maple Grove, MN	Hennepin	75.2%	Northwest of Minneapolis, about 20 minutes in light traffic.
100K to 200K	Centennial, CO	Arapahoe	75.1%	Southeast of Denver, incorporated in 2001 as most populous incorporation at that time.
200K to 500K	Lincoln, NE	Lancaster	66.3%	State capital and home of University of Nebraska.
500K to 1M	Seattle, WA	King	63.0%	18th largest city, and fastest growing over 500,000 since 2010, slightly exceeding Austin.
1M to 2M	San Jose, CA	Santa Clara	61.2%	High-tech Silicon Valley giant, areas along southern edge are over 75% response rates.
2M to 5M	Chicago, IL	Cook, DuPage	45.9%	Three cities in this range are tightly bunched.
5M to 10M	New York, NY	Kings, Queens, New York, Bronx, Richmond	41.9%	USA's largest city made up of Brooklyn, Queens, Manhattan, The Bronx, and Staten Island boroughs.

Largest city to have exceeded certain thresholds.

Dardenne Prairie, MO becomes largest city to surpass 80%.

100%=	North River, ND	55
90%+	North River, ND	55
80%+	Dardenne Prairie, MO	13,360
70%+	Naperville, IL	148,304
60%+	San Jose, CA	1,030,119
50%+	Phoenix, AZ	1,660,272

[jimrtex]

Least Responsive Cities

Miami Beach, FL sneaks by Deerfield Beach, FL in 50K to 100K category. Deerfield Beach is up the coast just south of the Palm Beach line, and has a large Brazilian and Haitian population.

Descriptions of non-responsive cities Putnam, OK; White Bird, ID; and Ratliff City, OK have been added.

1 to 9	Fontana Dam, NC	Graham	0.9%	The dam was built during WWII to supply electricity for Alcoa. It has now been converted to a resort, with between 106 and 177 cottages.
10 to 18	Dering Harbor, NY	Suffolk	0.0%	Two towns in this category. Summer homes on Shelter Island off eastern end of Long Island
(do)	Pine Valley, NJ	Camden	0.0%	Often rated top golf course in country, with a very exclusive membership.
20 to 49	Granger, MO	Scotland	0.0%	Six towns in this this category.  Declining farm town just south of Iowa.
(do)	Putnam, OK	Dewey	0.0%	West Oklahoma town, drying up for decades.
(do)	Rendville, OH	Perry	0.0%	Smallest village in Ohio, with 4% of 1890 population.
(do)	Rockham, SD	Faulk	0.0%	About 85 miles west of Watertown, with less than 1/10 of 1920 population
50 to 99	Alexander, KS	Rush	0.0%	Eight towns with no response in this range. Trading post on Santa Fe Trail. Namesake Alexander Harvey's daughter married William Bent.
(do)	Hitchita, OK	McIntosh	0.0%	Small town in eastern Oklahoma.
(do)	White Bird, ID	Idaho	0.0%	Panhandle town in decline after US 95 bypassed.
(do)	Wood Lake, NE	Cherry	0.0%	Cherry is the big county in the Sandhills of northern Nebraska.
100 to 199	Anawalt, WV	McDowell	0.0%	One of seven towns with no response for this size. In 2010, 1/3 of housing units were vacant in coal mining town.
(do)	Batesland, SD	Oglala Lakota	0.0%	On Pine Ridge reservation, Lakota name of Íŋyaŋ Šála translates to "Red Stone".
(do)	Deer Creek, OK	Grant	0.0%	Wheat farming town just south of Kansas.
(do)	Ratliff City, OK	Carter	0.0%	South Oklahoma, way off I-35.
200 to 499	Bryce Canyon City, UT	Garfield	0.0%	Two towns tied. Incorporated in 2007 under quirky law that permitted incorporation by owners of 50% of land value, in this case resort owner Syrett family.
(do)	Washingtonville, PA	Montour	0.0%	Largest town with no responses (274 est. 274). Borough is at largest population ever.
500 to 999	St. Francis, SD	Todd	0.5%	Tied with War, WV. On Rosebud reservation, with 1/200-odd responses.
(do)	War, WV	McDowell	0.5%	Mining town with 1/6 of 1950 population, with 2/400-odd responses.
1K to 2K	Spring City, UT	Sanpete	0.9%	According to Forbes one of prettiest towns in America, with particular Danish Morman flavor.
2K to 5K	Quartzsite, AZ	La Paz	1.8%	RV's congregate at Quartzite in winter. Low response rate may be due to confusion over which housing units are permanent.
5K to 10K	Big Bear Lake, CA	San Bernardino	3.7%	Lots of second homes, as population in area can swell to 100,000 on summer weekends.
10K to 20K	Jackson, WY	Teton	12.1%	Located in Jackson Hole, with lots of second homes for the ultra-rich.
20K to 50K	Paradise, CA	Butte	6.5%	May have lost 90% of its population in the 2018 Camp Fire, when 19,000 buildings were destroyed.
50K to 100K	Deerfield Beach	Broward	31.3%	Highest share of Brazilian (Americans) and Haitian (Americans) in the United States.
100K to 200K	Hartford, CT	Hartford	34.8%	Cities are limited in Connecticut, making an easy commute from suburban towns. Hartford is 84% Hispanic or non-white.
200K to 500K	Laredo, TX	Webb	34.8%	Border city, 96% Hispanic.
500K to 1M	Detroit, MI	Wayne	41.4%	Has lost 64% of its population (over 1,000,000) since 1950. 92% Hispanic or non-white.
1M to 2M	Philadelphia, PA	Philadelphia	42.5%	Has lost 1/4 of its population since 1950. 65% non-white or Hispanic.
2M to 5M	Los Angeles, CA	Los Angeles	43.5%	Three cities in this range are tightly bunched.
5M to 10M	New York, NY	Kings, Queens, New York, Bronx, Richmond	41.9%	USA's largest city made up of Brooklyn, Queens, Manhattan, The Bronx, and Staten Island boroughs.

Largest cities to not have reached certain thresholds.

<50%	New York, NY	8,398,748
<40%	Miami, FL	470,914
<30%	San Luis, AZ	33,490
<20%	San Luis, AZ	33,490
<10%	San Luis, AZ	33,490
<5%	Big Bear Lake, CA	5,281
<2%	Quartzsite, AZ	3,766
<1%	Spring City, UT	1,067
<0.5%	Lyons, PA	477
0%=	Washingtonville, PA	274

[jimrtex]

The Census Bureau has issued further schedule adjustments.

Fact Sheet: 2020 Census Operational Adjustments Due to COVID-19 (Long Version) (PDF) As Of April 13, 2020

Most significant is delaying the apportionment counts from December 31, 2020 until April 30, 2021; and delivering of redistricting data PL94-171 from March 31, 2021 until July 31, 2021.

These changes require congressional approval and will have significant effects on redistricting. Many legislatures meet in the spring of odd-numbered years and draw new maps at that time. The Census Bureau attempts to push the data out to these early states as soon as possible in February of the '1' year. This will likely require some special sessions for legislatures.

An extreme case is Texas, where legislative redistricting must be completed during the regular session, which is limited to 140 calendar days, or the redistricting will be turned over to a board comprised of high state officers. Both of these provisions are in the Texas Constitution.

There may also be an impact on filing deadlines for the 2022 election, the earliest of which are in December 2021.

The Non-Response Follow-Up (NRFU) has shifted from May 13-July 31, to August 11-October 31. That is, it is not expected to be able to get field workers out into the field until four months after the Census.

The Self-Response "deadline" has been extended from July 31 to October 31. This does not mean that they expect self-response that late, but just that they will accept self-response that late. The original NRFU was May 13-July 31. a period of 80 days. But there has to be the possibility of multiple attempts at contact. So let's assume 50 days for the initial contact period, which for self-response areas would include discovery or vacancies and deletions. That is about 2% followups per day. If the Census Bureau gets to you first it is not considered a self-response (though if the enumerator leaves a note to call a phone number, the person might finally self-respond). By 50 days into the effort, there will have been no opportunity to actually self-respond.

In 2020, there will likely be a greater publicity campaign to get people to self-respond in May, June, and July. Any self-response is way better than a NRFU visit four months after the census date. Perhaps there will a sixth or seventh postcard, or maybe more community outreach (I think it would break confidentiality to reveal which addresses had not responded, so this will probably just be more publicity).

In 2010, Self-response had reached 64.6% by April 24, before the final self-response of 66.5% occurred before September 2010 when counting stopped. 2020 still has a 11%+ to make up.

Update-Leave will be delayed from March 15-April 17 until June 13-July 9. It appears that they have decided to leave off knocking on doors, and will just leave the questionnaire on the doorstep. This schedule may also result in multiple contact attempts before NRFU begins. This effort will improve the total self-response by 2 or 3% (5% of housing units are in update-leave, but more of them are vacant. The self-response rate will be under 50% i these areas).

Group quarters enumeration may be particularly difficult at places like college dormitories where the norm might have been for paper forms to be dropped off for individuals to fill out. Administrators may know little beyond the names of residents, some of whom were kicked out before April 1.

Enumeration of homeless people has been delayed until at least September. This was intended to occur on April 1, with volunteers going to soup kitchens and under overpasses and bridges to try to get a head count. At best this will produce a number that might be comparable. The same will be true for transitory locations such as hotels and RV parks.

I can still see a second Census happening in 2022.

[jimrtex]

Question:

Could they also adapt their outreach further, for example to send an additional wave of paper questionnaires in July or August to those households that have not responded by then ?

Or would this be too big of an adjustment ? I guess printing anoter wave of questionnaires would have to be contracted to the printing company months in advance ...

I think they printed enough questionnaires for the entire country, just in case nobody responded by internet, or the whole system collapsed like Obamacare enrollment did.

In the update/leave method, the enumerator is supposed to  attach a code to the questionnaire they leave at the house. The enumerator goes to the housing unit. With their laptop they capture the GPS location and the status, including non-existent  (deleted), not habitable (structure that looks possible turns out to be a barn), vacant, etc. They then attach a code to the questionnaire and leave it.

The updated schedule suggests that they won't knock on the door, and attempt to get an alternative mailing address. That might just be my reading, since under the original plan they would have left a questionnaire on the front door, regardless whether someone answered their knock. OTOH in rural areas, it might be considered unfriendly snooping around a house with a laptop, and not attempting to make contact. Perhaps if there is a mailbox, they can record that.

But if enumerators in the field can peel off a code and scan it so it is captured and associated with a location, then the same could be done back at the distribution facility (and automated).

My inclination might to send another post card and try to get people to respond by internet or phone. If they want a paper form it could be sent to them. They might also want to check whether the current household (or any members of it) was present on April 1 - which would be 2+ months before this hypothetical mailing went out.

I think they are doing the update/leave first is because it requires less person-to-person interaction, and also as a matter of equity.

[jimrtex]

52.8% (+0.4)

They show this as a 0.5% increase.

For the week ending Saturday, April 24, 2010: 2.3% cumulative 64.6%.

For the week ending Friday, April 24, 2010: 2.6% cumulative 52.8%.

If we start seeing weak responses on Sunday I will adjust the comparable week in 2020 to end on a Saturday.

2010 - 2020 gap : 11.8%.

In 2010, the next week dropped to 0.7%. It will be interesting to see what happens in 2020 (the 5th mailing - postcard is scheduled to be send on Monday April 27, 2020).

Unweighted average of states for Friday:

0.47% total
0.15% internet
0.33% other.

Other is now at 69% of daily responses.

[jimrtex]

If we start seeing weak responses on Sunday I will adjust the comparable week in 2020 to end on a Saturday.

Sunday, or weekend returns, are always much worse than weekday returns.

I'm still predicting a 4% increase for the 10-day period April 20-30 though, to 55%.

Then 3%, 2%, 2% on average until end-May. Which would be 62% by then (+/- 1%).

By the end of June, the final 2010 numbers should be reached and the trickle for the remaining 4 months should bump the rate to 70-75%.

For 2020

Tue, Wed, Mon, Thu, Sun, Fri, Sat

[jimrtex]

53.2 (+0.4)

Not bad for a Saturday ...

The USPS delivers postal mail on Saturday.

In 2010, the date used was the check-in date, the date the mailed-in questionnaire was actually logged.

Mail typically takes 1 to 3 days to be delivered. People tend to not mail on Sunday, knowing that the mail won't begin processing until Sunday (and there won't be home pickup on Sunday). People know when the last time mail is emptied out of collection mail boxes, since they know that mail will at least begin processing. Local post offices may hold local letters (or maybe not. Much sorting occurs at regional sorting centers. You can truck all the mail from post offices in an area to a sorting center near a major airport, where you have an overnight shift sorting mail to go on airplanes, or back to post offices in the region. When the planes arrive from other areas it is sorted in to post offices for delivery.

So a letter from City A will go to an airport facility and then back to City B a few miles away for possible delivery the next day.

People may choose not to mail on Friday or Saturday, knowing that there will be a lag day before delivery (businesses are exceptional since they will want mail to go out on the last day of the work week, rather than on the following Monday).

During the peak period of 2010, some mail was being processed at the Census Bureau incoming facility on Sunday:

This was the peak in week in 2010, with 6.1% response rate on Wednesday. Even Monday was on the downhill slope, but they did some processing of Saturday mail on Sunday.

Fri 6758K
Sat 4249K
Sun 449K
Mon 4719K

The following weekend, volume was down by 1/2 to 2/3, and they shut down on Sunday. The Monday work load was a bit higher than Friday or Saturday, due to back-logged mail from the USPS.

Fri 2270K
Sat 2018K
Sun 0
Mon 2700K

By the following week, they had shutdown a shift on Saturday, and made it up on Monday.

Fri 889K
Sat 306K
Sun 0
Mon 1201K

By the next week, Saturday is almost closed, and Monday is almost twice as busy as Friday.

Fri 738K:
Sat 15K
Sun 1K
Mon 1524K

Overall volume continues to decline, and there are three-day shutdowns over Memorial Day, 4th of July, and Labor Day weekends.
Unweighted averages for 2020.

Total: 0.42%
Internet: 0.12%
Other: 0.31%.

The other was only down from 0.33% on Friday.

Total mail response for 2020 is about 1/8 of what it was in 2010, when almost all response was by mail. Though they might be able to shutdown on Saturdays, they may have scaled down their mail handling facilities. They may have changed their operations to maintain social distancing, such as spreading the work out over more shifts, doing more thorough disinfection of equipment, monitoring employee temperatures, etc.
Through the 4th Saturday in April, 2020 is 11.4% behind 2010.

Through April 24, 2010 64.6%
Through April 25, 2020 53.2%

My prediction for the upcoming week is 2.0% to 2.5%.

[jimrtex]

53.4 (+0.2)

1.8% to 2.3% for remaining 6 days of week to hit my prediction.

Unweighted state average.

Overall: 0.22%
Internet: 0.15%
Other: 0.07%

Internet was 98% of previous 3 days (Thu, Fri, Sat)
Other was 21% of previous 3 days.

This is generally consistent with USPS not delivering on Sunday. It is not quite as dead as 2010, when Sunday was literally under 100, while other days were 100s of thousands.

For 2020, this is around 100,000 as compared to about 500,000 each of the previous 3 days. There may be process changes due to social distancing in mail processing centers. There are two mail centers (one in Jeffersonville, IN which is opposite Louisville, KY and the other in Phoenix, AZ). The questionnaire has a return envelope with the preferred address.

It is possible that they are also getting phone responses. I assume this is counted as a self-response even though you are talking to an enumerator on the other end of the line.

The NYT may have been less muddled about the Census Bureau being muddled.

Don’t Forget to Respond: 2020 Census Reminder Postcards Arriving

The original schedule was:

(1) March 12-20. (1a) Internet invitation letter or (1b) Internet invitation letter and questionnaire. (1b) Was sent to areas less likely to respond by internet - older population, low past response, language barrier, less high speed internet access, etc.

(2) March 16-24. Reminder letter.

To non-respondents:

(3) March 26-April 3. Reminder postcard.

I know this went out on schedule based on someone receiving one, and their knowing that a questionnaire would soon be arriving.

(4) April 8-16. Questionnaire. I don't know whether this was sent to (1b) non-respondents who received the initial questionnaire. In 2010, it was found that it may have been more productive to simply send two questionnaires, and ignore a second response from some.

The press release says that non-respondents will "receive" a questionnaire by April 30. The Census Bureau can't assure that will happen unless the last questionnaires were sent by April 27. However, the author of the press release might not make that distinction. They may be simply treating sent ("the check is in the mail" and receipt as the same).

When I tried to determine whether there was increase in paper responses, I could if I made some very generous assumptions about mailing, questionnaire sitting in the mail pile, etc. It may have been that it was delayed. Perhaps there was a split between the (1a) and (1b) groups, with second questionnaires being delayed.

(5) April 20-27. 5th Postcard reminder.

The news release says this has been delayed until April 27-May 9 (8 days originally to 13 days now).

[jimrtex]

Descibe what these mean to you (don't simply translate):

Volljährige Person

Wohnung

Baby

Gemeinschaftsorganisator

Notfallwiederherstellungsprogramm

Wohnung in zwei Stockwerken

Cousin/Cousine

Vorname

schwer zu zählende Teile der Bevölkerung

Hotel

Wohneinheit

Säugling

Nachname

Wohnwagen

Motel

Neugeborenes

im Eigentum stehend und frei von jeglichen Lasten oder Rechten Dritter

Radio

Zimmer- oder Hausgenosse

Straßenname

Fernsehen

Wohnanhänger

[jimrtex]

It’s a few terms from the German-version of the US census questionnaire.

Actually, from the German-language glossary. There is no German-language questionnaire. As part of the language question on the ACS, they ask about how well the person can speak English. The response on that question determines whether the language must be included in ballots.

14a. Does this person speak a language other than English at home?

If NO skip to question 15a

14 b. What is this language? __________

For example: Korean, Italian, Spanish, Vietnamese

14 c. How well does this person speak English?

[ ] Very well
[ ] Well
[ ] Not well
[ ] Not at all

A response of "well" is treated for this purpose as not being able to comprehend a ballot. Since the language test is based on whether the person understands a ballot, this is used to inflate the numbers.

In any case, European languages (other than Spanish) do not qualify because they are not discriminated against by the Aryan majority. New York has its own law that requires use of Russian, but this is not a federal law.

I assume that the glossary would be used by a census language-assistance operator to make sure that they used Census Bureau-approved terms. In some cases the glossary added a parenthetical phrase in English, which suggests that the German term has a different connotation.

Volljährige Person

Why is "person" added. If these types of persons are recognized as a unique class in Germany wouldn't there be a word for them?

Wohnung

I was interested in the connotation, what you think of.

Baby

Is this a German word?

Gemeinschaftsorganisator

I'm not sure what this type of person would actually do. I'd think he was some sort of lay-about. Is this used in Austria?

Notfallwiederherstellungsprogramm

These may have a different name in Austria.

Wohnung in zwei Stockwerken

This indicates that there is a different understanding of the word "wohngung" than its English equivalent.

Cousin/Cousine

What relationships does this include?

Vorname

Why was this clarified as meaning first name?

schwer zu zählende Teile der Bevölkerung

These must not exist in Germany or Austria, since there is not a term, but this has to be explained.

Hotel

Is this a German word?

Wohneinheit

How many of these are there in your town?

Säugling

It is not clear that Americans recognize the meaning of this word.

Nachname

Why was this clarified as last name?

Wohnwagen

How big?

Motel

Is this a German word?

Neugeborenes

Is their clear distinct from a Baby or a Säugling?

im Eigentum stehend und frei von jeglichen Lasten oder Rechten Dritter

There must no be an idiom for this?

Radio

Both the medium and the device?

Zimmer- oder Hausgenosse

This included an English clarification. Perhaps these are not the same types of persons.

Straßenname

I thought that the use of "ß" had been deprecated?

Fernsehen

Both the medium and the device? Why don't Germans have TV's while they have radios?

Wohnanhänger

What is the difference between this and a Wohnwagen?

[jimrtex]

* paper questionnaires are still being mailed out:

On or between April 8-30 - Extended from April 8-16 to allow more time for addressing the mail packages in the midst of social distancing.

* reminder postcards are out later too:

On or between April 27-May 9 - Pushed back from April 20-27 to allow the paper questionnaire package to arrive several days ahead of the final postcard.

https://www.census.gov/newsroom/press-kits/2020/2020-census-mailings.html

Thank you.

I notice that households who originally received a paper questionnaire were sent a second questionnaire. I wonder if that was a consideration about when the 4th mailing went out. Households that had been promised a paper questionnaire if they didn't respond to the first three mailings may have been expecting the questionnaire which was originally scheduled to go out two weeks after the 3rd (postcard) mailing.

The 5th mailing (postcard) which is just now being sent out is more imperative, warning respondents that the only way to prevent an interviewer from being sent is to respond.