City and Town estimates

[Nhoj]

I see that no one seems to have noticed The census bureau released new city and town estimates the other day.

http://www.census.gov/newsroom/press-releases/2016/cb16-81.html
http://www.census.gov/popest/data/cities/totals/2015/index.html

[Cubby]

Usually Tender Branson starts a thread on new census estimates. Maybe he's too busy voting for the new Green President of Austria today to do it. They used to release the city/town estimates in late June so this was unexpected.

They are full of bad news since 5 out of the top 11 fastest growing cities are in Tejas. It will take decades to turn that state atlas red so what is there to be happy about?

Georgetown, TX, a suburb of Austin, is the fastest growing in the whole country. I think Del Webb is going to or has started a massive mega master planned community there. Around here in CT we have SO many homes from the 1700's and 1800's so I think its ridiculous to just build thousands of houses and pray that people will buy them all. And Texas houses never have yards, all that available land and no lawns, strange.

If Austin reaches a million people by 2020 will the Texas legislature still gerrymander it into 5 districts stretching hundreds of miles like it is now?

The most shocking stat was that a San Fran/Oakland suburb made it onto the list of fastest growing cities. That region stopped booming in the 80's so I did not see that coming. Sure beats commuting to San Fran from Stockton though.

[Angrie]

If Austin reaches a million people by 2020 will the Texas legislature still gerrymander it into 5 districts stretching hundreds of miles like it is now?

Over the long run, so much of the growth being concentrated is a good thing for Democrats. It will help Texas turn blue a little bit faster, although it will still take another 20 years or so.

In the meantime, the GOP will probably divide Austin up among 6 or 7 districts.

[cinyc]

I've put together a maps of the percentage change from 2014-2015  and 2010-2015 in each of the 20 states that fully or partially uses County Subdivisions as their main muncipalities:
County Subdivision 2014-15 Change Map
County Subdivision 2010-15 Change Map

These states are exclusively in the Northeast and Midwest, running from Maine to North Dakota excluding Iowa - but coverage gets spotty west of the Mississippi for some reason.   I don't think it's a data integrity issue, but a lack of data for those County Subs issue (we have data for Places, as defined by the Census Bureau in some of those states, but not County Subs, and there's no data for some (but not all) UTs in North and South Dakota).

This map is preliminary.  I'm going to try to add percent change by Census Place in the other states, but a lot of the Census Places are CDPs that aren't incorporated, and the Census population estimate generally only included incorporated areas.  I'll update this post when I do.

[Angrie]

I've put together a maps of the percentage change from 2014-2015  and 2010-2015 in each of the 20 states that fully or partially uses County Subdivisions as their main muncipalities:
County Subdivision 2014-15 Change Map
County Subdivision 2010-15 Change Map

Very nicely done! It would be cool to see the high growth states in the south and southwest, even if it is not possible on the same level of detail.

The county lines really stand out in a lot of cases in a way which I presume is not realistic. There are a lot of sharp changes at county lines, whereas in reality one would expect more smooth variation. Did you do anything like use county level data if there was not municipal data, or is that a reflection of imprecision in the data?

[cinyc]

Very nicely done! It would be cool to see the high growth states in the south and southwest, even if it is not possible on the same level of detail.

The county lines really stand out in a lot of cases in a way which I presume is not realistic. There are a lot of sharp changes at county lines, whereas in reality one would expect more smooth variation. Did you do anything like use county level data if there was not municipal data, or is that a reflection of imprecision in the data?

I had to simplify the County Sub shapefiles because the original census shapefiles are over 100MB and I only have 250 MB in my whole CartoDB account, which has other maps.  That probably explains why some of the County Sub lines zig and zag more than you'd expect - or you might be looking at some County Subdivisions in states with bizarre annexation laws, which allows for cities with jagged lines.

I am working on using Census Places for the South and Southwest.  Since those areas don't cover most of the country, I'm going to try to use the U.S. County map as the base layer there, assuming I can overcome a technical hurdle.

[cinyc]

I added incorporated Places in the South and Southwest to the existing maps.  You need to zoom in on them to really see.  They are mere islands as part of the larger state.

There may be a data integrity issue with some of these Places due to an Excel cut and paste error.  Let me know if something seems off.  Clicking on a geography will give you specific stats about it, assuming you're using a computer or tablet with Flash.

I'm working on the remainder of the counties layer now.  I'll probably have that up by tomorrow.

[cinyc]

I finished the county layer.  The data is mainly for the remainder of the county not in incorporated places, so you'll see unexpected declines in places like El Paso County, Texas, probably due to annexations by El Paso City.  The county layer can only be seen for the 30 non-County Sub states.

Let me know if you see any obvious data errors. 

[Angrie]

I finished the county layer.  The data is mainly for the remainder of the county not in incorporated places, so you'll see unexpected declines in places like El Paso County, Texas, probably due to annexations by El Paso City.  The county layer can only be seen for the 30 non-County Sub states.

Let me know if you see any obvious data errors. 

Aha.

There are a number of county remainders with unexpected declines like that. From the 2010-2015 map, some of the most striking ones are:

DeKalb County GA
Williamson County TX
Rockwall County TX
El Paso County TX
Utah County UT
Salt Lake County UT
Davis County UT
Riverside County CA
Santa Clara County CA
King County WA

These are pretty much all places where we should expect that there is high growth, but the county remainders have population declines. So I think your explanation regarding county annexations must be correct.

When you realize that a lot of the municipal "growth" is actually coming from annexations in certain municipalities, that makes the census press releases seem a little bit misleading. Some of the fastest growing cities on the census list may not actually be so fast growing in terms of additional people moving to the area. Instead, they are "fast growing" at least in part because they are just annexing already populated territory. That seems to be particularly the case in parts of suburban Texas and Utah, which dominate the list of "fastest growing" municipalities. That casts cases like Georgetown TX, which Cubby mentioned above, in a different light.

Anyway, the point is that for reapportionment purposes, maybe we should not expect such great shifts to those areas as we would have previously expected based on the census municipal growth press releases. To get a better idea of that, we should look more at the county level data.

[cinyc]

I'll take an in-depth look at one of those counties if I have time tonight to make sure I just didn't mess something up when doing the county conversions (some counties have no Places, and I had to combine those with those that do).  It would also be interesting to see whether this is due to annexation or something else, like the methodology Census uses to determine sub-county population growth.  I would expect they allocate to the Places first and the remainder of the county last, which could automatically lead to less population being allocated to the remainder than the incorporated Places.

[Cubby]

If Austin reaches a million people by 2020 will the Texas legislature still gerrymander it into 5 districts stretching hundreds of miles like it is now?

Over the long run, so much of the growth being concentrated is a good thing for Democrats. It will help Texas turn blue a little bit faster, although it will still take another 20 years or so.

In the meantime, the GOP will probably divide Austin up among 6 or 7 districts.

True, I've noticed that Collin & Denton Counties are not staying as strongly GOP as Montgomery County is. Williamson & Fort Bend also seem to be getting bluer, so there is hope at the Congressional level in the (somewhat) near future.

[cinyc]

There are a number of county remainders with unexpected declines like that. From the 2010-2015 map, some of the most striking ones are:

DeKalb County GA
Williamson County TX
Rockwall County TX
El Paso County TX
Utah County UT
Salt Lake County UT
Davis County UT
Riverside County CA
Santa Clara County CA
King County WA

Unincorporated Dekalb County, GA's "decline" is almost entirely due to the creation of a new city, Brookhaven, in 2012.  If I compared populations to the 2010 Census estimates (which Census revises backwards), DeKalb would gain population.

Unincorporated Williamson County, TX's decrease is likely due to a 15,000-resident annexation by Austin.  Unincorporated Rockwall County, TX only lost 440 residents from the 2010 Census, but since the population base is less than 10,000, that loss gets amplified. There probably were a few small annexations there. Unincorporated El Paso County, TX's population loss was due to the incorporation of San Elizario in 2013.

The populations of the unincorporated portions of two of the three Utah counties are fairly small.  Unincorporated Davis County, Utah has a very small population - around 3,600 now, down from around 3,800 in 2010.  Unincorporated Utah County had about 10,000 people as of 2010, and 8750 today.  My guess is some annexation occurred. Very few people live outside of those Utah counties' cities.  Unincorporated Salt Lake County has many more residents, about 145,000, only down slightly from about 146,000 in the 2010 Census.  Minor annexations probably explain this minor decrease.

Unincorporated Riverside County, CA's decrease was due to the incorporation of two cities, Jurupa Valley and Eastvale.  Unincorporated Santa Clara, CA's relatively small decrease was likely due to an annexation by San Jose.

I can't pinpoint the reason for the significant Unincorporated King County, WA drop.  Three cities appear to have annexed some territory, but the population difference might not be enough to explain the whole change.

I could always base the percentage change for the county remainders on the 2010 population estimates, if you think that would make more sense.  I'd use the estimate numbers for the Incorporated Places, too, to be consistent.

[cinyc]

I updated the maps to fix data integrity issues.  Of County Subs, only data for one town was wrong.  Data for a number of Places in the alphabetically first of the 30 other states was off, though.  It's all been fixed.

While making the update, I also calculated percentage change from the 2010 Population Estimate for jurisdictions that didn't have 2010 Census data, instead of leaving them as N/A.  These are most likely places incorporated after 2010.  I also put the data for all of Honolulu County instead of just the remainder, since the map doesn't include the Urban Honolulu CDP (even though Census included estimates for that geography).

[jimrtex]

I updated the maps to fix data integrity issues.  Of County Subs, only data for one town was wrong.  Data for a number of Places in the alphabetically first of the 30 other states was off, though.  It's all been fixed.

While making the update, I also calculated percentage change from the 2010 Population Estimate for jurisdictions that didn't have 2010 Census data, instead of leaving them as N/A.  These are most likely places incorporated after 2010.  I also put the data for all of Honolulu County instead of just the remainder, since the map doesn't include the Urban Honolulu CDP (even though Census included estimates for that geography).

I decided to answer your question about QGIS here, even though you had solved it.

I'm assuming you are using base shapefile(s) of census places. I can only find state-based files, so if I were creating a national map, I'd have to merge the layers.

Anyhow one of the fields in the shapefile is "CLASSFP"   Incorporated cities are coded with "Cn", while CDPs are coded as "Un" or "Mn" (CDP's associated with military bases).

QGIS permits selection  by expression, so you could use something like

left("CLASSFP",1)<>'C'

select all the non-cities, and then when in edit mode press the Delete key.

Another way to do it would be to start with your estimate data, and join to the places shapefile, using the ID for the place as the join field. You can then isolate the places without estimate data (which is really what you are wanting to get rid of) and delete them.

If you use American Factfinder there is a Product Type called Geographic Header, which includes things like CLASSFP for places.


I think that the Census Bureau is a bit ambiguous in its use of MCD. Lots of states have MCDs (legally recognizes entities at a subcounty level, but don't use them. North Carolina for example has townships created by the carpetbaggers. After the restoration of local control, they stripped them of any functional use, but they still exist, and the census bureau still produces data for them. States can go back and forth between the two. Tennessee switched to CCD (Census County Division) and then back to MCD.

The eight other states with MCD are: AR, LA, MS, TN, NC, VA, WV, and MD.

The 21 states in the northeast might be better characterized as "area that census bureau produces subcounty estimates".

For South Dakota, I matched the MCDs with estimates, to the 2010 MCDs. A common pattern, though not universal was that an entire county with unorganized territories would be skipped in the estimate. MCDs for parts of counties were incorporated cities.

In states where incorporated cities are independent of townships, the census bureau is sort of forced to treat them as MCDs, even if they aren't formally. But in some counties in South Dakota it appears that the cities are just floating in the counties. Conceivably you could treat these areas the same as in the south and west.

[cinyc]

I updated the maps to fix data integrity issues.  Of County Subs, only data for one town was wrong.  Data for a number of Places in the alphabetically first of the 30 other states was off, though.  It's all been fixed.

While making the update, I also calculated percentage change from the 2010 Population Estimate for jurisdictions that didn't have 2010 Census data, instead of leaving them as N/A.  These are most likely places incorporated after 2010.  I also put the data for all of Honolulu County instead of just the remainder, since the map doesn't include the Urban Honolulu CDP (even though Census included estimates for that geography).

I decided to answer your question about QGIS here, even though you had solved it.

I'm assuming you are using base shapefile(s) of census places. I can only find state-based files, so if I were creating a national map, I'd have to merge the layers.

Anyhow one of the fields in the shapefile is "CLASSFP"   Incorporated cities are coded with "Cn", while CDPs are coded as "Un" or "Mn" (CDP's associated with military bases).

QGIS permits selection  by expression, so you could use something like

left("CLASSFP",1)<>'C'

select all the non-cities, and then when in edit mode press the Delete key.

Another way to do it would be to start with your estimate data, and join to the places shapefile, using the ID for the place as the join field. You can then isolate the places without estimate data (which is really what you are wanting to get rid of) and delete them.

If you use American Factfinder there is a Product Type called Geographic Header, which includes things like CLASSFP for places.

Yeah, I had to merge the state shapefiles to create a 30-state Place shapefile.  I then deleted the CDPs by doing a query for all places except CDPs (LSAD <> "57" [57 is the LSAD code for CDP]), and saved the shapes selected, all using MapWindow GIS.  I'm now redoing my work to create a 36-state Place shapefile (adding the 6 states West of the Mississippi using MCDs/County Subdivisions to the Places shapefile, since many of those have blank counties), and eventually a national Place shapefile.  This time, I'll keep the Urban Honolulu CDP in, since there is data for it.

I then simplify the shapefiles using mapshaper.org to significantly shrink their size.  Otherwise, I'd have no room left on my CartoDB account.  For the current map, I used 3 layers - one for Counties and County Remainders (the standard Census U.S. County Shapefile), one for Places without CDPs (merged from 30 state Places shapefiles with CDPs deleted) and one for MCDs/County Subs (merged from 20 state County Sub shapefiles).

[cinyc]

I revised the maps to use Places/County Remainders instead of County Subs in the 6 states West of the Mississippi with limited County Sub data.  I also added Places to the remaining 14 County Sub states, and changed the population in the non-incorporated MCDs there to reflect the balance of the population in that jurisidiction.  This effectively adds villages to places like Long Island, New York, adding another level of granularity.

Finally, I created a third map, 2010-2015 versus the 2010 Estimates base, which generally smooths out annexations by comparing the populations of current areas to what they were in the 2010 Census regardless of what Place, MCD or County Remainder they were a part of back in 2010.

Here are the maps:
2010-2015 Percentage Change in Population, Using 2010 Census Data
2010-2015 Percentage Change in Population, Using Estimates Base
2014-2015 Percentage Change in Population

Note that I added county lines in black, County Sub lines in white, and Place lines in grey.  I'm not sure whether they are more distracting than they're worth.  Please let me know what you think.

I can make other maps on request - population change by number of new residents, instead of percentage, total 2015 population, Places and County Remainders only, County Sub only for the 14 states in the Northeast, etc.  Let me know if you want any.

[cinyc]

I made three more maps, this time showing the numerical change in Cities, Towns and County Remainders:
Numerical Change 2010-2015 versus 2010 Census
Numerical Change 2010-2015 versus 2010 Estimates Base
Numerical Change 2014-2015

Again, the main difference between the 2010 census and estimates base seems to be that the estimates base is backward-looking toward annexations, incorporations and the like.

[Kevinstat]

From an e-mail I sent to a friend who shares my interest in redistricting (although his interest is perhaps not quite as strong as mine):

You must be logged in to read this quote.

I later clarified my comments on Augusta with the following e-mail:

You must be logged in to read this quote.