The Official 2020 Census Thread

[jimrtex]

I updated my 2020 Census maps with data up through today. The maps are here.

It doesn't make a ton of sense to update every day, with things only going up .1% a day, if even. There's a lot more tracts and places ahead of their 2010 rate, though.

The color scale is confusing. I assumed that dark green would be the best response, but it isn't.

[jimrtex]

I updated my 2020 Census maps with data up through today. The maps are here.

It doesn't make a ton of sense to update every day, with things only going up .1% a day, if even. There's a lot more tracts and places ahead of their 2010 rate, though.

The color scale is confusing. I assumed that dark green would be the best response, but it isn't.

Gold is the best response vs 2010. It was added as an afterthought - like crap, I already used green for the best possible negative response rate. So I need something else to signify positive responses. Too late to change now.

It wasn't the Gold, but that you went Light Green to Dark Green to Light Purple to Dark Purple.

I could see something like Dark Green to Light Green to Light Purple to Dark Purple, with gold and red on the extremes. And then set the control point between the light green and light purple (with a default around the national deviation).

[jimrtex]

I updated my 2020 Census maps with data up through today. The maps are here.

It doesn't make a ton of sense to update every day, with things only going up .1% a day, if even. There's a lot more tracts and places ahead of their 2010 rate, though.

The color scale is confusing. I assumed that dark green would be the best response, but it isn't.

Gold is the best response vs 2010. It was added as an afterthought - like crap, I already used green for the best possible negative response rate. So I need something else to signify positive responses. Too late to change now.

It wasn't the Gold, but that you went Light Green to Dark Green to Light Purple to Dark Purple.

I could see something like Dark Green to Light Green to Light Purple to Dark Purple, with gold and red on the extremes. And then set the control point between the light green and light purple (with a default around the national deviation).

That's not how it goes, Jim. It goes from Purples to Reds to Blues to Greens on the negative side, Dark to light. Positive side is Gold, light to dark. Unless I screwed something up, which is always possible.

My thinking is that those areas close to reaching the 2010 response level should be dark green. I don't have a problem with those areas that have exceeded the 2010 rate being gold (or maybe it doesn't matter with this particular area).

Red doesn't really show up since that would be 40% below 2010.

[jimrtex]

Eleven-day unweighted averages:

Total 1.32%
Internet 0.69%
Paper&Phone 0.63%

Daily averages:

Total 0.12%
Internet 0.06%
Paper&Phone 0.06%

Week-over-week is 0.7%

Internet and Paper are almost balanced. It does appear that there is regional variation:

Paper:

Central: WI, MI, IL, OH, PA, KY, TN, MS, AL, (IN is even)
Carolina: NC, SC
Oddballs: HI, ID, RI, ME

Internet:

West: Everything west of Mississippi except HI, ID, and (SD, KS, MO which are even).
East: MA to FL, including WV, except NC, SC, RI, and (VA which is even)

VT and NH are also even.

[jimrtex]

Twelve-day unweighted averages:

Total 1.41%
Internet 0.73%
Paper&Phone 0.67%

Daily averages:

Total 0.12%
Internet 0.06%
Paper&Phone 0.06%

Week-over-week is 0.6%

Internet and Paper are almost balanced. It does appear that there is regional variation:

Paper:

Central: MI, IL, IN, OH, KY, TN, MS, AL, NC (MN, WI, PA, LA, SC, CT, MD)
New England: MA, RI, VT, ME
Oddball: DE

Internet:

West: Everything west of Mississippi except (HI, MT, MN, LA are even).
East: NH, NY, ND, DC, WV, VA, GA, AL (CT, MD, PA, SC are even)

There may be a processing delay in Phoenix. The west is Internet dominated, but also slightly lower overall (i.e. the Internet increases are similar, but the paper increases are slower).

[jimrtex]

Census Bureau to Resume Some 2020 Census Field Operations in Additional Locations

MAY 22, 2020 — The U.S. Census Bureau, in coordination with federal, state and local health officials, will begin a phased restart of some 2020 Census field operations in 10 additional states the week of May 25. With the addition of these area census offices resuming operations, 90% of the nation’s update leave workload will have resumed.

All of the area census offices in the following states will begin a phased restart of operations:

Connecticut
Massachusetts
Minnesota
South Dakota
Wisconsin

Some of the area census offices in the following states will begin a phased restart of operations:

California (13/30)
Maryland (1/4)
Michigan (2/5)
Texas (10/26)
Virginia (2/6)

In addition, the Census Bureau is expanding resumption of operations in New Mexico (Albuquerque) and New York (Albany, Buffalo and Syracuse).

Updates on the operations resuming by location are available at 2020census.gov. The Census Bureau will update this webpage weekly as 2020 Census operations resume across the United States.

https://www.census.gov/newsroom/press-releases/2020/2020-resume-additional-areas.html

New York is now (4/21) with Upstate (everything from Delaware, Greene, Columbia on northward reopening. 13 ACO are in NYC, 2 on Long Island, and 2 in the lower Hudson including Westchester. These will have relatively few UL TEA, or some second homes in the Catskills or Hamptons.

The 17 California ACO not resuming include San Francisco (1), Alameda(1), Contra Costa (1*), Sacramento(1), Los Angeles(7), Orange(2), Riverside(2), Fresno and Bakersfield. Only the latter have large swaths of territory in the San Joaquin Valley. Offices in Santa Clara and Ventura counties include the Central Coast, while those in Chico, Stokcton, and Santa Rosa have all of Northern California covered including Marin, but excluding Sacramento.

ACO in Santa Clara(2), San Mateo(1), San Diego(3), and San Bernardino(2) are re-opening.

The Concord ACO includes Solano and Napa where the 2700 UL TEA are likely concentrated.

The Maryland ACO that is reopening is in Hagerstown and covers the western part of the state beginning with Frederick, it also includes Montgomery. Note, it has no UL TEA. The Eastern Shore is covered out of the Towson ACO (just north of Baltimore), so consideration may be given to reopening the offices more than the UL operations.

Ironically, Census Bureau HQ is in Suitland, MD.

The two Michigan ACO that are reopening cover most of the area of the state, The three Detroit ACO that are closed do include Flint, Saginaw, Bay City, and the Thumb, but not Washtenaw (Ann Arbor) and Monroe.

The two Virginia ACO that are reopening cover the western part of the state. Two Nova offices cover Fairfax and Loudoun as well as Arlington and Alexandria. The RIchmond ACO includes some areas to the west as well as the Southside. Virginia Beach covers Hampton Roads.

The 16 Texas offices that are not reopening include Harris(6), Bexar(3), Dallas(3), Travis(1), and Tarrant(1) that are urban counties only. Two ACO with a larger footprint are El Paso which covers the Trans-Pecos and Midland-Odessa; and Laredo which covers a swath across to include Corpus.

With the latest openings 151 of 245 ACO will be reopened (about 62%). As the news release notes, this is 90% of the UL TEA which are concentrated in more rural areas.

ACO not yet reopening:

CA 17/30
NY 17/21
TX 16/26
IL 10
OH 8
NJ 8
VA 4/6
MD 3/1
MI 3/5
SC 3 (unknown why they are not opening)
DE 1
DC 1
HI 1
NH 1
NM-AZ(Navajo) 1

The ACO are physical offices with I think less than 10 people. Field Supervisors work from home as do the field workers who go out each day from their residence.

The last temporary worker count has about 25,000 that was about the same as it has been. It was through the 9th, which was just at the beginning of the UL reopening. We should see an increase in employment if the UL is really getting restarted.

There may be glitches in hiring. The UL was originally scheduled for March and April. Older workers might not be interested in working now. Workers require a background check.

[jimrtex]

Thirteen-day unweighted averages:

Total 1.48%
Internet 0.78%
Paper&Phone 0.70%

Daily averages:

Total 0.11%
Internet 0.06%
Paper&Phone 0.05%

Week-over-week is 0.6%

It appears that there is a slight preference for Internet responses (53%:47%), this has resulted in the paper-preferring states to become more scattered:

ME, VT, RI, PA, OH, KY, TN, MS, MT, HI, with previously preferring paper states switching to even. PA to MS are contiguous.

All states west of the Mississippi are Internet leaning, except LA, AR, KS, ID are even, and MT and HI are paper. The western Internet tendency now includes IL.

Remember that, in the USA, Monday is Memorial Day with no mail delivery. The holiday is also why the newest ACO to reopen for UL will do so 9 days after the previous group.

Note that overall, Internet has been used 80.7% of responses. It is only the recent responses that are balanced.

If the current 13-day rate is continued:

MI (15th), WA (21st), KY (22nd), and NV (24th) will reach their 2010 level in June. 23 states will reach in July, and 18 states will reach in August, 6 won't finish until after September 1.

If there is a 10% weekly decay.

Michigan will reach on June 20, three will reach in July, and five will reach in August.

I suspect that the final self response rate will be around 64%. It should be noted that planning for NRFU was based on a 60% self response rate. They may have low-balled the estimate so as to ensure they had sufficient capacity for NRFU.

[jimrtex]

Map of ACOs that are re-starting work, by week:



Too bad that South Texas is still not open.

It has a huge amount of Mexicans, who are lagging far behind with their responses ...

That map doesn't quite match
[https://public.tableau.com/profile/us.census.bureau#!/vizhome/StatusofCurrentOperations/ACOMap?publish=yes]Status of Current 2020 Census Operations[/url]

Where did you find it? I found earlier versions, but not a primary source.

In particular it shows that Rochester, MN; Gainesville, GA; and Fredericksburg, VA; and Allentown, PA as not reopening, yet.

Of interest is that the Window Rock, AZ ACO is specific to the Navajo reservation, excluding the Hopi reservation, but also extending into Utah. The North Dakota ACE (Bismarck) also includes all of the Standing Rock reservation in SD.

The ACO don't strictly follow county boundaries, at least in some locales. The Oakland, CA ACO includes western Costra Costa (e.g. Richmond), while the Concord, CA ACO includes the eastern part of that county, as well as including Solano and Napa. A Pleasanton, ACO covers southern and eastern parts of Alameda (e.g Fremont, Hayward, Livermore).

The Hidalgo-Cameron ACO is reopening. The Laredo ACO is not. Perhaps the Laredo city government is preventing the office re-opening. The Laredo police had a sting operation to catch someone doing hair-styling in their home. Note that Laredo and Corpus Christi, and San Antonio suburbs were mostly Self-Response. There are large areas of the Brush Country that are Update/Leave, but not heavily populated.

[jimrtex]

I updated my 2020 Census maps with data up through today. The maps are here.

It doesn't make a ton of sense to update every day, with things only going up .1% a day, if even. There's a lot more tracts and places ahead of their 2010 rate, though.

I came across an archive of daily response data:

https://www.gc.cuny.edu/Page-Elements/Academics-Research-Centers-Initiatives/Centers-and-Institutes/Center-for-Urban-Research/CUR-research-initiatives/Census-2020-Response-Rate-Analysis-Weeks-8-and-9

See link in "2020 Self-Response Data downloaded daily from the Census Bureau website and archived by the Center for Urban Research, CUNY Graduate Center."

It appears to be one large zip file of all the daily response rates.

[jimrtex]

Fourteen-day unweighted averages:

Total 1.52%
Internet 0.79%
Paper&Phone 0.73%

Daily averages:

Total 0.11%
Internet 0.06%
Paper&Phone 0.05%

Week-over-week is 0.5%


Seven-day unweighted averages (since 5/17/2010)

Total 0.54%
Internet 0.30%
Paper&Phone 0.25%

Daily average:

Total 0.08%
Internet 0.08%
Paper&Phone 0.04%.

I had miscalculated the projected date that a state might reach the 2010 self-response level, calculating the gap to be closed from 5/10/2020 rather than the current response rate (e.g. for Michigan I calculated a gap of 67.7 - 65.2 = 2.5; rather than 67.7 - 66.7 = 1.0).

But I have recalculated using the slower increase over the past 7 days.

Assuming no decay,

MI June 7
WA June 14
KY June 15
NV June 18

12 states would reach in July
19 states would reach in August
16 states would reach after September 1

If a 10% weekly decay is used:

4 states would reach in June
0 states would reach in July
3 states would reach in August

[jimrtex]

Jim, I noticed something odd:

Between May 21 and May 22, WV had a 0.3% daily increase on the ranking page.

But it increased by only 0.1% on the response map.

KY had a 0.2% increase on the ranking PAGE, but the rate remained the same between the two days on the ranking MAP ...

That cannot be possible, even with rounding.

It's a mess.

On the ranking map, you can get the underlying data in a .csv format.

Warning: The csv file has the current date, while the map selects to the previous date.

That is, on 5/24 after the daily update has been issued (2 PM CDT, UTC-5) the map and ranking pages will show 5/23. But the .csv file will say 5/24.

I don't know how to get data for previous days, except that it can be displayed on the map, but not the rankings.

But the link I gave to cinyc has an archive of daily csv files that CUNY Center for Urban Research has downloaded.

The csv file that they have for 5/22 is missing columns. There is a lot of statistical information such as minimum daily increase, maximum daily increase, average daily increase, county ranges for a state, etc. This was not present for the 5/22 csv file. Either the Census Bureau posted an incomplete file, or the CUR downloaded only part of it.

But I just looked back through the files, and the daily response rate is inconsistent with the cumulative response rate:

	Cumulative	Reported	Calculated	Difference
12-May	62.5	0.2
13-May	62.7	0.3	0.2	0.1
14-May	62.9	0.3	0.2	0.1
15-May	63.0	0.2	0.1	0.1
16-May	63.2	0.2	0.2	0.0
17-May	63.2	0.1	0.0	0.1
18-May	63.3	0.1	0.1	0.0
19-May	63.4	0.1	0.1	0.0
20-May	63.5	0.1	0.1	0.0
21-May	63.6	0.1	0.1	0.0
22-May	63.7	0.2	0.1	0.1
23-May	63.7	0.2	0.0	0.2
24-May	63.8	0.1	0.1	0.0

The reported daily response rate is always the same as, or greater than the difference between cumulative response rates on consecutive days. While these could be different due to independent rounding, the difference should either be +0.1, 0.0, or -0.1, with balance between positive errors and negative errors.

There may have been an additional glitch on May 23 (csv date, May 22 display date)

This would suggest that the cumulative number of responses is not the total of daily responses. I don't know why.

Since I switched over to reporting multi-day changes, I have been using the cumulative response rate to calculate the change in response rate. For example, the cumulative response rates for Kentucky for 5/10, 5/17, and 5/24 are 62.2%, 63.2%, and 63.8%. I calculated the weekly response change as 1.0% and 0.6%.

Assuming that the cumulative response rates are accurate minus rounding errors, the rate of change should be accurate +/- 0.1%.

Prior to May 10, I was using the the daily response rates reported on the rankings page. Assuming that there was an error due to use of that number, it is likely to have been relative small.

[jimrtex]

Jim, I noticed something odd:

It's a mess.

I just checked the national rates.

From 5/12 to 5/24 the cumulative response rate increased from 58.7% to 60.0%, a change of 1.3%. The sum of daily response rates was 1.7%.

[jimrtex]

Eight-day unweighted averages (since 5/17/2010)

Total 0.57%
Internet 0.32%
Paper&Phone 0.26%

Daily average:

Total 0.07%
Internet 0.04%
Paper&Phone 0.03%.

There appears to be a slight shift towards Internet responses, perhaps due to the holiday weekend, perhaps due Update/Leave areas beginning to come in.

The late responses to the Self-Response TEA had more paper responses because many of those households did not receive a paper form until the 4th mailing in mid to late April. If they couldn't respond via Internet (or preferred not to, they could not respond at all).

While Update/Leave households received paper forms on first contact, they also received a Internet ID. While it may be perceived that the census worker walks up a dirt driveway past rusted out trucks and pieces of old stills with banjos playing in the background, this may not be wholly accurate.

Overall, Internet responses have been 80% of all responses. It it is only recently been closer to 50:50. If Update/Leave areas are responding 60:40 Internet, it could still have an effect of tilting the recent response rate higher.

[jimrtex]

Starting next Monday, the Census Bureau will only update their response map 5x per week instead of daily.

They don't say which days, but I guess MO-FR.

Not sure if they keep updating their ranking page daily or also just 5 days ...

That is their original announced policy in their FAQ. I don't know whether they will adjust it because NRFU won't begin until July.

You didn't tell me the source of your ACO map.

[jimrtex]

Starting next Monday, the Census Bureau will only update their response map 5x per week instead of daily.

They don't say which days, but I guess MO-FR.

Not sure if they keep updating their ranking page daily or also just 5 days ...

That is their original announced policy in their FAQ. I don't know whether they will adjust it because NRFU won't begin until July.

You didn't tell me the source of your ACO map.

The source is on the map, below Alaska.

CUNY:

https://twitter.com/Census2020Map/status/1265313047318536200

They have updated their map today, so it looks pretty similar to the CB map.

I've sent CUNY an e-mail to correct their map.

[jimrtex]

Starting next Monday, the Census Bureau will only update their response map 5x per week instead of daily.

They don't say which days, but I guess MO-FR.

Not sure if they keep updating their ranking page daily or also just 5 days ...

That is their original announced policy in their FAQ. I don't know whether they will adjust it because NRFU won't begin until July.

You didn't tell me the source of your ACO map.

The source is on the map, below Alaska.

CUNY:

https://twitter.com/Census2020Map/status/1265313047318536200

They have updated their map today, so it looks pretty similar to the CB map.

I've sent CUNY an e-mail to correct their map.

Here is an update to the map credited to an "astute reader".



Incidentally, there are other cases where an ACO extends across a state border so as to encompass all of a reservation.

Window Rock, AZ ACO covers the Navajo Reservation in AZ, NM, and UT, but excludes the Hopi Reservation enclaves which are under the Flagstaff, AZ ACO.

The Bismarck, ND ACO includes all of the Standing Rock Reservation by extending into SD to do so.

The Overland Park, KS ACO takes in a piece of Richardson County in extreme southeastern NE to include all of the Sac and Fox Nation and Iowa(KS-NE) reservations. Otherwise, all of NE is Lincoln, NE ACO.

The North Las Vegas, NV ACO extends into Owyhee County in far southwestern ID to include all of the Duck Valley Reservation.

The Bakersfield, CA ACO takes two small areas in Nye and Esmeralda counties across the state line in NV which are part of the Timbi-Sha Shoshone Reservation.

The Tucson, AZ ACO extends across the Colorado River into San Bernardino and Riverside counties in CA and Clark County in NV to include all of the Colorado River Indian Reservation and the Mojave Reservation.

[jimrtex]

Jim, I noticed something odd:

It's a mess.

I just checked the national rates.

From 5/12 to 5/24 the cumulative response rate increased from 58.7% to 60.0%, a change of 1.3%. The sum of daily response rates was 1.7%.

There may be a systematic bias in the rounding of the daily response rate.

Let's assume that the cumulative rate is displayed as 60.1%. It is reasonable to assume that the actual rate is uniformly distributed between (60.05% and 60.15%) and the rounding error will range from (+0.05% to -0.05%).

But the same is not true for the daily rate, particularly as it approaches and falls below 0.1%.

It might be much more likely to be between 0.05% and 0.10%, than it is to be between 0.10% and 0.15%, and thus more likely to be rounded up repeatedly over several days.

If the rounding error for the cumulative rate and the daily rate were both uniformly distributed from (+0.05% to -0.05%) then we would expect that 1/8 of the time the daily rate would round to a value 0.1% greater than the difference between the cumulative rate for two consecutive days; 1/8 of the time the daily rate would round to a value 0.1% less than the difference between the cumulative rate; and 3/4 of the time the daily rate would  be consistent with the change in the cumulative rate.

[jimrtex]

Nine-day unweighted averages (since 5/17/2010)

Total 0.61%
Internet 0.35%
Paper&Phone 0.25%

Daily average:

Total 0.07%
Internet 0.04%
Paper&Phone 0.03%.

The Paper response has been essentially dead for the last two days due to the holiday weekend. Tomorrow should see an uptick as they process a two (or three) day backlog.

[jimrtex]

Estimated effect of Update/Leave. CA, MI, NY, and TX have a substantial majority of UL TEA restarting. MD and VA have some UL TEA restarting. DE, DC, HI, IL, NH, NJ, OH, and SC have not had any UL restarts. AZ and NM exclude effects of Window Rock, AZ not being opened.

State	Responses	Self-Response	UL HU	SR HU	UL HU Share	SR TEA Response	SR Response Rate	Estimated UL Responses	Estimate Response Rate	Change	Restart
Alabama	1400	58.2%	82	2323	3.4%	1396	60.1%	42	59.8%	1.6%	4-May
Alaska	130	40.8%	110	209	34.5%	125	59.7%	56	56.6%	15.8%	4-May
Arizona	1800	56.5%	270	2916	8.5%	1787	61.3%	141	60.5%	4.0%	18-May
Arkansas	840	54.9%	77	1453	5.1%	836	57.6%	38	57.1%	2.2%	4-May
California	9300	61.3%	273	14898	1.8%	9286	62.3%	145	62.2%	0.9%	27-May
Colorado	1600	62.8%	196	2352	7.7%	1590	67.6%	113	66.8%	4.0%	18-May
Connecticut	1000	63.8%	7	1561	0.4%	1000	64.1%	4	64.0%	0.2%	27-May
Florida	5900	57.6%	172	10071	1.7%	5891	58.5%	85	58.4%	0.8%	11-May
Georgia	2700	56.2%	112	4692	2.3%	2694	57.4%	55	57.2%	1.0%	11-May
Idaho	490	62.5%	71	713	9.0%	486	68.2%	41	67.3%	4.8%	4-May
Indiana	2000	64.9%	39	3043	1.3%	1998	65.7%	22	65.5%	0.6%	18-May
Iowa	990	66.9%	14	1466	0.9%	989	67.5%	8	67.4%	0.5%	18-May
Kansas	860	64.2%	33	1307	2.4%	858	65.7%	18	65.4%	1.2%	11-May
Kentucky	1300	63.8%	79	1959	3.9%	1296	66.2%	44	65.8%	2.0%	11-May
Louisiana	1200	54.4%	92	2114	4.1%	1195	56.5%	44	56.2%	1.8%	18-May
Maine	390	50.3%	112	664	14.4%	384	57.9%	55	56.7%	6.4%	4-May
Maryland	1700	64.2%	0	2648	0.0%	1700	64.2%	0	64.2%	0.0%	27-May
Massachusetts	2000	62.2%	69	3146	2.2%	1997	63.5%	37	63.3%	1.1%	27-May
Michigan	3200	66.7%	100	4697	2.1%	3195	68.0%	58	67.8%	1.1%	27-May
Minnesota	1800	70.0%	87	2485	3.4%	1796	72.3%	53	71.9%	1.9%	27-May
Mississippi	800	55.5%	55	1387	3.8%	797	57.5%	27	57.2%	1.7%	4-May
Missouri	1800	60.5%	131	2844	4.4%	1793	63.1%	70	62.6%	2.1%	11-May
Montana	300	51.2%	113	473	19.3%	294	62.2%	60	60.4%	9.2%	4-May
Nebraska	590	66.5%	30	857	3.4%	588	68.7%	18	68.3%	1.8%	18-May
Nevada	780	59.3%	48	1268	3.6%	778	61.3%	25	61.0%	1.7%	18-May
New Mexico	500	47.6%	168	883	15.9%	492	55.7%	79	54.3%	6.7%	27-May
New York	5000	55.0%	200	8891	2.2%	4990	56.1%	96	55.9%	0.9%	27-May
North Carolina	2800	56.4%	165	4800	3.3%	2792	58.2%	81	57.9%	1.5%	11-May
North Dakota	240	59.2%	42	364	10.3%	238	65.4%	23	64.4%	5.2%	4-May
Oklahoma	1000	53.7%	188	1675	10.1%	991	59.2%	94	58.3%	4.6%	4-May
Oregon	1200	63.0%	27	1878	1.4%	1199	63.8%	15	63.7%	0.7%	11-May
Pennsylvania	3800	63.1%	197	5825	3.3%	3790	65.1%	109	64.7%	1.6%	11-May
Rhode Island	290	58.2%	6	493	1.1%	290	58.8%	3	58.7%	0.5%	18-May
South Dakota	250	59.8%	50	368	12.0%	247	67.3%	29	66.1%	6.3%	27-May
Tennessee	1900	60.2%	9	3147	0.3%	1900	60.4%	5	60.3%	0.1%	4-May
Texas	6600	54.8%	334	11710	2.8%	6583	56.2%	159	56.0%	1.2%	27-May
Utah	760	63.7%	89	1104	7.4%	756	68.4%	52	67.7%	4.0%	4-May
Vermont	180	50.9%	61	293	17.2%	177	60.5%	31	58.9%	8.0%	4-May
Virginia	2400	65.2%	35	3646	0.9%	2398	65.8%	19	65.7%	0.5%	27-May
Washington	2200	65.4%	92	3272	2.7%	2195	67.1%	53	66.8%	1.4%	11-May
West Virginia	470	47.7%	289	696	29.4%	456	65.4%	161	62.6%	14.9%	4-May
Wisconsin	1900	67.4%	93	2726	3.3%	1895	69.5%	55	69.2%	1.8%	27-May
Wyoming	150	50.2%	67	232	22.5%	147	63.3%	36	61.2%	11.0%	18-May

Calculation: Self-Response rate and Number of Responses is provided by the Census Bureau. From this the total number of housing units can be calculated:

     Response_Rate = Responses / Total_HU

     Total_HU = Responses / Response_Rate

Census Bureau is providing UL HU in reopened(-ing) ACO. If a Housing unit is not an UL_HU it is a SR_HU.

     SR_HU = Total_HU - UL_HU

UL_HU share is the percentage of Total_HU that are Update/Leave.

     UL_HU share = UL_HU / Total_HU

SR responses are estimated based on an assumption that 5% of UL HU have responded.

    SR_Responses = Total_Responses - UL_HU * 5%

This percentage may be low. Persons in UL TEA may respond by the Internet, but they were not individually solicited. They might have read something on the Internet, or in local media. Now that UL has reopened (2+ weeks in earliest re-opened areas) some persons have been contacted and forms and Internet invitations left at residences. Once initial contact is made, reminders are sent just like they were in the Self-Response areas. If the 5% estimate is low, then the effect of Update/Leave will be less.

Given as estimate of Self responses in SR TEA, and the number of SR HU we can calculate the response rate in SR TEA.

  SR_Response_Rate = SR_Responses / SR_HU

We can then estimate what the response rate will ultimately be in UL TEA, based on an assumption that it will be 85% of that for SR_TEA. This estimate is based on the 2010 Census. UL TEA areas have more seasonal and vacation HU. They tend to also have more vacant housing. Part of the purpose of "Update" is to determine whether a Housing Unit even exists or is habitable. In rural areas with declining population, there may be housing that is habitable, but with no demand for it. It may take decades before it it is abandoned, vandalized, and bulldozed. In 2010, after vacant and deleted housing was removed from the denominator the Return Rate (the percentage of households that have responded) as opposed to the Response Rate (the percentage of housing units for which returns were made) were almost equal.

  UL_Response_Rate = SR_Response_Rate * 85%
  UL_Responses = UL_HU * UL_Response_Rate

From our estimates of SR_Responses and UL_Responses we can estimate what the total response rate will be at the end of June or so, when UL TEA areas have been contacted.

  Total_Response_Rate = (SR_Responses + UL_Responses) / Total_HU

If responses trickling in now are from SR TEA, this rate will also increase, but the change due to UL will not marginally change. If responses trickling in now are from UL TEA, then the ultimate rate will drop slightly, as will the change.

  Change = (estimated) Total_Response_Rate - Current_Response_Rate

In the included states, the current cumulative rate is 59.8%. The estimated effect of UL completion is 1.7%.

[jimrtex]

Why the f**k does the ranking page have KY at +0.6%, but the map only +0.1% ?

They are reporting data found HERE

It is not the display that has an error, but the data input into the display.

Now this is where it gets weird.

Earlier I had downloaded the .csv file from the above and was trying to figure out an explanation (the states with the high daily response rate KY, TN, IN, WV, VA have counties with very high daily response rates), but those are probably messed up as well.

The USA county with the reported highest daily increase (Crittenden, KY) has barely moved over the past several days. It was at 61.9% on 5/22 and is only up to 62.0% five days later.

Whatever is happening is specific to certain states, and the daily reported response is wildly inconsistent with the cumulative change.

When I started writing this I was checking the link above, to make sure it was the correct one. But now the .csv file was 3/24/2020. It was like someone had reported a bug and they attempted to fix it by uploading an old file.

But now it is back to the 05/27/2020 file.

Maybe there are hackers from some country like Bulgaria or Austria.

[jimrtex]

Ten-day unweighted averages (since 5/17/2010)

Total 0.67%
Internet 0.39%
Paper&Phone 0.28%

Daily average:

Total 0.07%
Internet 0.04%
Paper&Phone 0.03%.

I think we are starting to get some update/leave responses, since Internet is beginning to become more dominant. Alternatively, some sleepyheads are responding to Internet ads.

[jimrtex]

Eleven-day unweighted averages (since 5/17/2010)

Total 0.75%
Internet 0.43%
Paper&Phone 0.33%

Daily average:

Total 0.07%
Internet 0.04%
Paper&Phone 0.03%.

A slight uptick in paper forms now that we are past the weekend.

At recent rate, Michigan will reach 2010 response rate on June 14, 16 days from now.
Washington on June 19
Nevada on June 24
Kentucky on June 25

6 states in July
20 states in August
21 states later.

[jimrtex]

KY once again increased by 0.2% on the ranking page, but by 0.0% on the map.

There were 327 counties (in 32 states) with a discrepancy of 0.2% or more between the daily response rate, and the change in the cumulative response rate between 5/27/2020 and 5/28/2020.

The average (per county) discrepancy was greater than 0.10% in 6 states:

IN 0.36%
TN 0.18%
MI 0.17%
KY 0.15%
LA 0.15%
IL 0.11%

The error is showing up at the county level and is widespread, but is concentrated in a few states which causes the error to also show up at the state level.

There were 13 counties that had cumulative increases of 0.3% or greater. These may indicate the effect of Update/Leave.

Aleutians East, AK
Hoonah-Angoon, AK
Gilchrist, FL
Grant, NE
Mineral, NV
Harding, NM
Tyrrell, NC
Adams, ND
Jones, SD (0.5%, all others are 0.3%)
Schleicher, TX
Throckmorton, TX
Val Verde, TX
Cumberland, VA

[jimrtex]

The Census Bureau has confirmed they will send additional reminder letters/postcards in the next months before the census takers start door-knocking on August 11.

Maybe this will increase the declining responses again.

Source?

When NRFU does start, if contact is made people are given a chance to self-respond. If they are not home, a notice is left on the door, and a reminder letter is sent a week later.

Up to 6 attempts may be made during NRFU. After the 3rd attempt, the Census Bureau can use proxies such as neighbors.

[jimrtex]

The Census Bureau has confirmed they will send additional reminder letters/postcards in the next months before the census takers start door-knocking on August 11.

Maybe this will increase the declining responses again.

Source?

When NRFU does start, if contact is made people are given a chance to self-respond. If they are not home, a notice is left on the door, and a reminder letter is sent a week later.

Up to 6 attempts may be made during NRFU. After the 3rd attempt, the Census Bureau can use proxies such as neighbors.

https://www.calendarwiz.com/calendars/popup.php?op=view&id=144104849&crd=cens1sample

I don't understand that. Are they saying that there will be a news release on June 1?