The Official 2020 Census Thread

[cinyc]

2020 response rates by tract minus 2010 final mail response rate tract map gif for the past 3 days:

https://twitter.com/cinyc9/status/1245813750210408452

Red tracts are further behind their final 2010 mail response rates than blue or green ones. Yellow tracts will be ahead of their final 2010 mail response rates.

In NYC, the upper-income tracts near Central Park are way behind. Did they go to their 2nd homes and forget to fill out the Census?

[cinyc]

Hey cinyc, thx for the maps.

Could you do a state or county map too ?

BTW: what did you use as the „final“ 2010 rates ? The 66.5% recalculated rate for 2020, or the original 74% ?

I'd have to set up a new map to do county/state. Maybe in the next few days, if I have time.

I had to do some voodoo math to get 2010 response rates for the 2020 tracts. The 2010 tract map doesn't match 2020 in a lot of places. New tracts have been created since then. So I took the final mail total and the final total, multiplied by area within a 2020 tract for each 2010 geo, and divided. At least I THINK it was the original 74% - but I'll have to check my spreadsheets tomorrow. It's running ahead of the hardtocount maps in some tracts, so it's probably the 74%.

[cinyc]

Cinyc, on your Twitter Page you re-tweeted a misleading tweet of a NYT woman on response rates and daily comparisons between 2010 and 2020.

In her chart, you get the impression that 2020 rates so far are trailing behind the 2010 rates at comparable times, but that’s wrong.

2020 response is actually a bit better so far, because there were different times when invitation letters were sent out to households.

Besides, daily 2010 response was still based on the final 74% back then, not the re-calculated 66.5% rate that is used for the 2020 numbers as a comparison.

I can't really un-RT something I RTed. The damage is already done.

CUNY (i.e. the folks who do the hard-to-compare map) told me that they are putting out an article on how non-national this-time-in-2010 comparisons aren't accurate because of different methodology and other stuff. I'll RT that when they put it up.

The main focus with my maps are really to test which areas are lagging due to coronavirus or demographic change, not to show actual return rate. I'm particularly focusing on high-income areas in NYC and tracts adjacent to college campuses to see if there's any effect on return rate.

[cinyc]

It took a while, but I finally tracked down Census' official 2010 Final Self-Response rate data for 2020 tract, as opposed to my self-computed figure, which I think was the total mail-in rate after Census contact. Here's the revised tract gifs for the USA and NYC:

https://twitter.com/cinyc9/status/1246294190289543168

The Upper East Side rich tract problem still exists, albeit a little less stark.

Tender - I'll try to download the state/county data tomorrow to make additional maps.

[cinyc]

Here's the County gif requested by Tender Branson*:

https://twitter.com/cinyc9/status/1246666202627399685

*with a bonus Tweet of the Alaska tract closeup.

And the 4/4 data update (which I think is just technically through midnight 4/3):

https://twitter.com/cinyc9/status/1246518623302234117

[cinyc]

Daily update:

44.5% (+0.6)

Wow, yesterday’s returns were BAD.

If the daily responses are now deteriorating at such a speed, then the record-high response of 74%+ probably won’t be reached.

I've made my 2020 vs 2010 response rate differential maps interactive:

https://cinycmaps.com/index.php/2-uncategorized/52-2020-vs-2010-census-response-rate-mutlimap

Click on Menu to change the thing mapped. Default map = Today's Tract 2020 Response Rate vs 2010 Self-Response Rate differential map. The differential county maps are included in the multimap, too, as are maps of the daily and 2010 historic tract and county response rates.

[cinyc]

Daily update:

44.5% (+0.6)

Wow, yesterday’s returns were BAD.

If the daily responses are now deteriorating at such a speed, then the record-high response of 74%+ probably won’t be reached.

I've made my 2020 vs 2010 response rate differential maps interactive:

https://cinycmaps.com/index.php/2-uncategorized/52-2020-vs-2010-census-response-rate-mutlimap

Click on Menu to change the thing mapped. Default map = Today's Tract 2020 Response Rate vs 2010 Self-Response Rate differential map. The differential county maps are included in the multimap, too, as are maps of the daily and 2010 historic tract and county response rates.

Thanks 🙏

So you are using the re-calculated 66.5% national rate for 2010 ? And based on it all related state, county and tract data from this link ? https://2020census.gov/en/response-rates.html

If yes, then there are already some counties in SD and AL for example that have surpassed their 2010 rates.

Nationally, only 2/3 of the 2010 rates are in so far. 1/3 missing.

But the Census Bureau will mail out paper questionnaires this week to households that have not responded yet - so theoretically there should be another rate boost starting next week.

I got the 2010 data from Census' api. So, in theory, it's the same data as Census has on their website. They had to recalculate rates for tracts, so those might not add up to 66.5%. This is because 2020 tracts have different shapes from the 2010 tracts. I'm not sure if they recalculated the 2010 Final Self-Response Rate for counties - probably not as most counties haven't changed, but who knows?

Yes, if you look at the April 5 map, Bullock County, AL and Kinsgbury County, SD are ahead of their 2010 rates. That's why they're a shade of yellow on the interactive.

I just added CD maps to the interactives. Incorporated places and County Subs (in the states that have data) might be next, along with a simple margin map of the nationwide vs. the 2020 response rate on a particular date. Do you have the nationwide response rates for the past 6 days?

[cinyc]

Thanks. I think my dates are off by 1 compared to Census - my dates are the dates Census released the data, not the date of the data (there's a 15-hour lag or so).

I'll work on this tomorrow.

[cinyc]

Daily update:

45.1% (+0.6)

VA joins MN, WI and MI as blue states on the map, indicating 51%+ response.

Yeah - NOVA has really been filling out their forms quickly.

I've updated the maps with today's data and added the vs. Nationwide comparison maps.

Now, if I can only figure out which states have County Sub Data, I can add County Sub and Incorporated Place maps.

[cinyc]

I added the interactive map Tender was probably really looking for - 2020 Response rate minus 66.5%, the 2010 Final Nationwide Mail in Rate. (See, for example, 4/1-6-->Apr 6-->Tract vs 10 NW.

So now, not only can you compare how a tract/county/CD is doing vs its own 2010 Final Self-Response Rate, but 2010 overall, too.

The simple red/blue heatmap of the current day's 2020 reporting rate vs the nationwide 2020 reporting rate is under, for example, 4/1-6-->Apr 6-->Tract vs NW.

I had to move the probably less-viewed red-toned raw 2020/2010 reporting rate maps to their own tabs.

[cinyc]

I updated the maps with today's data and added incorporated place versions of the maps. The Place maps under the 3/31 A and 4/1-7 A tabs.

https://cinycmaps.com/index.php/population-change/tract-population-change/2-uncategorized/52-2020-vs-2010-census-response-rate-mutlimap

County subs/MCDs are next in the 20 or so states that Census provides data for them - probably tomorrow.

[cinyc]

I changed the Place maps to show the combined Counties, County Subs (MCDs) and Places that usually appear on my Incorporated Place Population Change maps. Unlike those maps, the County and County Sub (MCD) data includes the mapped Places, because unlike in the Incorporated Place Pop change data, there is no remainder-of-County/County Sub data available.

As with the incorporated place pop change maps, the data is more rich in the Northeast and Midwest than it is in the rest of the country.

The only thing I haven't mapped are tribal tracts, but getting the 2010 version of that data is nearly impossible. You have to search for it in the Census api by Reservation. Way too much work.

[cinyc]

45.7% today (+0.6)

Hopefully the paper questionnaires will lead to a boost next week ...

46.2% today (+0.5)

Getting slower and slower ...

Part of the problem is that the usual knock-and-drop operations in rural and other areas are currently on hold due to the coronavirus.

[cinyc]

48.1% (+0.2)

That’s a painfully slow daily increase.

Yesterday was Easter Sunday.

[cinyc]

https://twitter.com/hansilowang/status/1249790519796027392

[cinyc]

For what it's worth, I lived in university housing at the time of the 2010 Census and definitely received my own individual Census form that I mailed directly back to the Census bureau. (I was also accidentally double-counted in the 2010 Census as I learned later that my parents also listed me as still resident at their home when they completed the Census - oops.)

Group housing mechanics are probably going to be used more for group housing where the individuals occupying it can't reasonably fill out and mail their own forms (or complete a form online, this time around), such as prisons, asylums, nursing homes, etc., and less for places like military barracks or university housing.

On-campus college housing administrators have the option to fill out the forms themselves from their records or let students fill them out themselves.

[cinyc]

48.6% (+0.5)

Back to normal.

[cinyc]

https://twitter.com/hansilowang/status/1249790519796027392

Seems like the April and July 2021 dates would mess with states like Virginia and New Jersey that hold elections that year supposedly with the new maps.

Could they just reuse the old maps again or is that illegal?

They'd have no choice but to reuse the old maps unless they conducted their own state censuses, I think.

[cinyc]

jimrtex asked me to try to calculate an alternative 2010 final response rate metric. To do this, I had to estimate the percentage of HHs in each tract that are self-report tracts vs. Update/Leave or others. I then multiplied the 2010 Final rate in each tract by the 2020 Self Report rate. Here's the resultant map based on the most recent data released on 4/15, in gif form:

https://twitter.com/cinyc9/status/1250662104480264192

As you can see, many of the "problem" rural areas really aren't problems because Census effectively hasn't started there yet due to the virus.

[cinyc]

jimrtex asked me to try to calculate an alternative 2010 final response rate metric. To do this, I had to estimate the percentage of HHs in each tract that are self-report tracts vs. Update/Leave or others. I then multiplied the 2010 Final rate in each tract by the 2020 Self Report rate. Here's the resultant map based on the most recent data released on 4/15, in gif form:

Does the data you have for response rates have population numbers?

If you click on the map on the Census Bureau response page, you end up on public.tableau.com, which is apparently a platform that permits creation of public data visualization.

The rankings page of the visualization allows filtering by population range.

It appears that the visualization believes that the population of The Highlands, KS (sic, it would alphabetized between Thames and Thomastown) and Highland, KS are both 1001. The Highlands is in Reno County and was newly incorporated in 2017. I can not find anything on the census pages, including the 2018 place estimates. Highland is in Doniphan County, and the estimate of 1001 appears to be accurate.

But the response rates for the two cities are different.

The response rate rankings would lead you to believe that The Highlands is the response leader in the 1000 and 1999 range, but it has nowhere the number of houses to support that. It is built around a golf course. Oddly The Highlands golf course is now Crazy Horse golf course. There is also a Crazy Horse in Salina, KS so I suspect the two are related.

Tableau looks interesting. I don't know how difficult the learning curve is.

I’ll have to check the data. I think population and household PERCENTS are in the tract relationship file, but not neccesarily raw counts. I’ve now calculated HH numbers, so that could be added for 2020 tracts.

I do have the population data from the 2014-18 ACS for places/MCDs, but I don’t think the data is in the shapefile I uploaded to Mapbox. I could use min show to put a floor in, but would have to program to put a ceiling in.

I’ve never gotten the hang of Tableau, but some find it easier to use. Mapbox requires cobbling together some Javascript to get the maps to work.

[cinyc]

jimrtex asked me to try to calculate an alternative 2010 final response rate metric. To do this, I had to estimate the percentage of HHs in each tract that are self-report tracts vs. Update/Leave or others. I then multiplied the 2010 Final rate in each tract by the 2020 Self Report rate. Here's the resultant map based on the most recent data released on 4/15, in gif form:

Does the data you have for response rates have population numbers?

If you click on the map on the Census Bureau response page, you end up on public.tableau.com, which is apparently a platform that permits creation of public data visualization.

The rankings page of the visualization allows filtering by population range.

It appears that the visualization believes that the population of The Highlands, KS (sic, it would alphabetized between Thames and Thomastown) and Highland, KS are both 1001. The Highlands is in Reno County and was newly incorporated in 2017. I can not find anything on the census pages, including the 2018 place estimates. Highland is in Doniphan County, and the estimate of 1001 appears to be accurate.

But the response rates for the two cities are different.

The response rate rankings would lead you to believe that The Highlands is the response leader in the 1000 and 1999 range, but it has nowhere the number of houses to support that. It is built around a golf course. Oddly The Highlands golf course is now Crazy Horse golf course. There is also a Crazy Horse in Salina, KS so I suspect the two are related.

Tableau looks interesting. I don't know how difficult the learning curve is.

I’ll have to check the data. I think population and household PERCENTS are in the tract relationship file, but not neccesarily raw counts. I’ve now calculated HH numbers, so that could be added for 2020 tracts.

I do have the population data from the 2014-18 ACS for places/MCDs, but I don’t think the data is in the shapefile I uploaded to Mapbox. I could use min show to put a floor in, but would have to program to put a ceiling in.

I’ve never gotten the hang of Tableau, but some find it easier to use. Mapbox requires cobbling together some Javascript to get the maps to work.

How do you get tract names? The csv file just has ID's.

1600000US2070300 Thayer, KS
1600000US2070357 The Highlands, KS???
1600000US2070625 Timken, KS

There's a Tract Crosswalk File here:

https://www2.census.gov/programs-surveys/decennial/2020/data/2020map/2020/

It's by county name, not by town name, I think.

The Place layer is in a separate shapefile that I used for my 2018 pop change maps. Those are in the names are in the 2018-sub-all csvs from the PEP.

I checked. I can easily get Population for Places/MCDs/Counties in the place layer. I've already done that, and will add it when I add the 4-16 data release to the maps tonight. I'll change the Min Show to be geo pop for now. Coding a max show will take some time, maybe over the weekend.

Unfortunately, it will take some time calcualting population for tracts. I'd have to get the data from the 2014-18 ACS, then sumif for the new tracts either by PCT Area or PCT Households (probably the latter). Unfortunately, the 2010-20 relationship file doesn't have raw population counts.

[cinyc]

The census maps have been updated with today's data:

https://www.cinycmaps.com/index.php/2020-census

If you open the Options menu on the top right, Min Show will do the following now:

For tract maps: putting a decimal between 0.1 and 1 will white out precincts that have less than that 2020 response rate.
For place/MCD maps: putting an integer > 0 will white out areas with a 2018 population less than what's inputted.

Hit Options again to close the options menu.

I'll have to think about how to add a Max Show if I have time over the weekend.

[cinyc]

Well, the map is of incorporated places, and The Highlands were a CDP in 2018. As you know, Census doesn't provide population estimates for CDPs as part of their Population Estimates Program, just incorporated places. I'd have to use the ACS data for that, probably 5-year in a lot of cases. That's stale.

Is Census providing PCT return data for CDPs or just incorporated places? I haven't checked. If they are, I might be able to add a new map this weekend.

[cinyc]

Well, the map is of incorporated places, and The Highlands were a CDP in 2018. As you know, Census doesn't provide population estimates for CDPs as part of their Population Estimates Program, just incorporated places. I'd have to use the ACS data for that, probably 5-year in a lot of cases. That's stale.

Is Census providing PCT return data for CDPs or just incorporated places? I haven't checked. If they are, I might be able to add a new map this weekend.

The Highlands was incorporated in 2017, but the Census Bureau may not have had that information in time for 2018 estimates. The Census Bureau response rate map is actually a Tableau window embedded in the 2020census.gov web page.

Once you click on Rankings, you are on an actual Tableau page.

I think that they are just using percentage values (rounded to 0.1%) because there was a ranking tie between Waukesha, WI and Carver, MN which is quite unlikely without rounding.

But there has to be population data, because the filters would not work without it.

I first set a filter of 1000 to 1999, and The Highlands shows up as the best responding city. So I did a little research and came across articles about its incorporation. But their is no Wikipedia page or much other information. I added a link on the Reno County Wikipedia page, and was going to produce at least a minimal page for the city. The city does have a Facebook page, so I might have asked them to fill in some more information. If you do Google search, you are just as likely to come across articles about Highland, which is an older town on a square grid in Doniphan County. The Highlands appears to just be the houses around the golf course on about a section of land.

If you start trimming the filter down, you eventually get to 1001 to 1001 where the only cities in Kansas are The Highlands with a response rate of 77.9% and Highland with a response rate of 34.9%. 1001 is the 2018 estimate for Highland.

I think the relationship between the Census Bureau and Tableau is likely to be somewhat fuzzy. The Census Bureau realize they aren't nimble enough to do good presentation. They concentrate on data collection. If they contracted out the response rate app, there would have been cost overruns and it wouldn't be live on time. The Census Bureau hosts hackathons to produce apps accessing their data.

It is good advertising for Tableau to be hosting a Census Bureau visualization. They would probably be willing to do some hand-holding to help the Census Bureau get an visualization running. Maybe it was an intern.

Maybe if there was no population for The Highlands (N/A not 0), the visualization would crash and someone tweaked the data.

I don't know. I have contacted Tableau and they said it is the Census Bureau's visualization, which is quite reasonable. I emailed the Census Bureau email address for the response rate page, but just today and haven't got a response yet.

The data set (.csv file) has a Geographic ID that would fit between Thayer and Timken, KS, so I think that is the ID for The Highlands, but there is no population data on that site, or any other Census Bureau web site that I can locate.

I just found it on the 2019 shapefiles for places, just over a square mile at 109.04W 46.02N. No population numbers.

The response rates don't appear to include CDP's. It calls them "cities", which is colloquial and not something the Census Bureau would use since the meaning is very state-specific.

p.s. I can't seem to get the crosswalk file to load. Excel says it is trying to repair it, but can not.

I thought you were talking about Highlands, TX, which is a CDP. Neverhtheless, The Highlands, NV? are in my data, but not in the 2018 generalized Place shp. I'd need to download the 2019 Place shape, but unless Census recently added it, they haven't released a generalized 2019 file yet.

Here's a chart of all Places (including CDPs, I think) with a response rate over 77%:



Key:
DecFSRR2010 = 2010 Final Response Rate
A16 = 2020 Response Rate from 4/16 Data
DiffA16 = 2020 - 2010
NWA16 = 2020 - 0.494, the national response rate on 4/16/20
10NWA16 = 2020 - 0.665, the national 2010 Self-Response Rate


I'll see if I can find the Crosswalk file on my computer and e-mail it to you.

[cinyc]

Census extension could help get better count of diverse communities

Roll Call

BTW: When is Congress supposed to pass the extension ?

When the House comes back into session, which likely isn't until May.