Population Growth Patterns in Metro Areas, 2000-16

[jimrtex]

Now that I have the population estimate data for 2000-16 in a readable format, I can make gifs of the growth patterns in metro areas.  First, the greater NYC Metro Area:



Note how there was an exodus from the Inner suburbs until around 2008, followed by an exodus from the exurbs in the 2010s.

What metro should I make next?

The Nassau-Suffolk line is interesting. Also the Poconos. Is that people families moving out, or adult children moving away, as happened in Nassau in the 70s (there was the huge influx following WWII, and then 15-25 years later all the children were grown. The adults were still around 50-60, so not ready to vacate their homes, and there wasn't yet infill of apartments. The children who formed families had to move out to Suffolk. Those without, could move to NYC, or wherever they ended up for college, or North Carolina, Florida, etc. where they met a mate from somewhere else.

Is there something odd about the projection? It seems a bit stretched east-west.

[jimrtex]

Is there something odd about the projection? It seems a bit stretched east-west.

There might be.  I used the census default projection on QGIS without reprojecting it to Albers Conic or whatever.  It's NAD 83 (EPSG:4269).  But Long Island is really long, which might make it look like it's stretched (plus, my water file leads to more water than there really is being shown, which probably makes Long Island seem longer than it really is.  

What projection should I be using?  Albers makes things conic, so stuff isn't oriented the way you normally see it on a map.

NAD 83 is Latitude/Longitude so at NYC (roughly 40N), North/South is shortened by cos(40 degrees) or 0.76.

I notice it in the northern border of New Jersey and in diagonally oriented squarish towns, which become more diamond shaped.

There is something called False Mercator which QGIS switches to when display a Google satellite layer. I think this has something to do with the tiling of the Google images. For small areas you don't have to worry about area distortion. If you were interested in comparing the size of different metro areas (Miami v Boston) you might want to switch to more localized coordinate systems.

[jimrtex]

Great stuff. If you could do the DC-NoVA-Metro area next that would be awesome.

I will, after I fix the NYC Metro gif to a better projection (NAD 83 (HARN): Long Island looks better than what I have). 

Unfortunately, you're going to lose a lot of the granularity south of the Mason-Dixon Line and west of the Mississippi, though, since those states don't have MCDs that census provides data for, and there are no yearly estimates for fictional areas like CDPs, which replace MCDs there.

You could conceivably use ACS estimates for census tracts or block groups. You would have to use the five-year estimates, but could use 2011-2015 for 2013; 2010-2014 for 2012, etc.

[jimrtex]

Southern Wisconsin (Milwaukee & Madison) would be awesome!

Southeastern Wisconsin:

It appears that the Census Bureau is making estimates at the county level, and then distributing that among the towns and cities within the counties.

[jimrtex]

Boston:

Is (was) base housing for Hanscom AFB located in Lincoln.

[jimrtex]

Columbus, Ohio:



What's up with the wild population swings in Union Township, Madison County?

It has two prisons (London Correctional Institution and Madison Correctional Institution) which has a an address of London, but appears to be located in the township (Madison Township wraps around London, and the prisons are west and northwest of the city).

In 2010, 89% of the township population was male, and 78% was incarcerated.

[jimrtex]

Central Florida (Basically, the I-4 corridor from Tampa to Orlando to Daytona Beach, though part of Voluisa County got cut off in order to add the city of Sarasota):



The usually grey or blue cities near the middle of the map in Southern Orange County is Lake Buena Vista and Bay Lake, a.k.a. Disney World.  Very few people actually live in those cities, which is why the population gain usually doesn't reflect the rest of the map.

They do, but the Census Bureau discards forms from Micky Mouse and Donald Duck who claim that they live in the Magic Kingdom.

[jimrtex]

Denver:



Colorado Springs is on the map, too, since Park County is part of the Denver Metropolitan Area, and South Park County is parallel to it.  Boulder is its own metro, but is also on the map.  It's part of the Denver CSA, though, as is Weld County, which is partially on the map.

What is South Park County?

[jimrtex]

Denver:



Colorado Springs is on the map, too, since Park County is part of the Denver Metropolitan Area, and South Park County is parallel to it.  Boulder is its own metro, but is also on the map.  It's part of the Denver CSA, though, as is Weld County, which is partially on the map.

What is South Park County?

Sorry.  Southern Park County.  I shouldn't have capitalized the South.

South Park High School is in Fairplay, the larger dot in the west central part of the county. There are residential subdivisions just west of the Jefferson County line off of US-285. Commuting into Denver or just across the line to work in a grocery store pulls the county into the Denver MSA.

[jimrtex]

Providence:



The metro includes all of Rhode Island and Bristol County, Massachusetts.  It appears that Bristol County is growing faster than Rhode Island, thanks to growth in Boston's exurbs.

New Bedford and Fall River are drags on Bristol County, and aren't quite as close as Brockton or Lowell to have a post-deindustrialization recovery.

[jimrtex]

Is this plausible enough to be used for a reliable projection to 2020, for purposes of redistricting?



The holes on the map along the Schuylkill and around the Naval Shipyard are very small census tracts. The change ranges from 20% to -20%, from the 2010 Census to the 2015 5-year ACS (which would represent an averaged population from 2011 to 2015), if the rate of change were constant, this would be about July 2013.

[jimrtex]

Is this plausible enough to be used for a reliable projection to 2020, for purposes of redistricting?



The holes on the map along the Schuylkill and around the Naval Shipyard are very small census tracts. The change ranges from 20% to -20%, from the 2010 Census to the 2015 5-year ACS (which would represent an averaged population from 2011 to 2015), if the rate of change were constant, this would be about July 2013.

What would you do?  Triple the rate from the ACS?

I think there's an issue with post-recession rates of change being different than pre-recession rates.  For example, the Pittsburgh region seems to have actually grown a bit from 2008 to 2012 or so - and has declined since.

What I did for the county projections is determine the annual rate of increase from 2010-April to 2016-July, and project that forward to 2020. For county subdivision projections in Allegheny, Chester, and Montgomery, I did the same, but controlled the total population for a county to the 2020 county estimate.

The rate of increase in the estimates for Philadelphia have definitely slowed over the past 3 years, so that I am over projecting assuming that the rate for the remainder of the decade remains low(er).

What I have decided to do for Philadelphia is to aggregate the populations of the census tracts assigned to each district for the 2010 census and the 2011-2015 ACS, determine the percentage change, and then assuming that the 2011-2015 population is a July 2013 population, determine an annual rate of increase, project that to 2020, and then control to the 2020 county estimate.

So far, I've done the trans-Schuylkill. Initially the population was down based on the areas closer to the shipyard, but there is now an increase as I've got further west (Temple area?).

[jimrtex]

I'm working on a larger project with more historical maps.  For now, here's a gif of percentage population growth in the US from 1900-2016, using 2016 county lines throughout:

The 1920-1930 blue swath across Georgia was interesting. Reprise of Sherman's march to the sea?