6

I am trying to gather the total population of Pittsburgh and overlay it onto a Google Map. I took the Census Tract .shp files from http://www.census.gov/geo/maps-data/data/gazetteer2010.html and put them in a fusion table and then ran the census API for total population

http://api.census.gov/data/2010/acs5?key=XXXX&get=B01003_001E&for=tract:*&in=state:42+county:003

Then I took that information and uploaded it into another google Fusion table and linked the tract information letting me pull in the population numbers and link it to the kml data.

I have my outcome hung here http://codepen.io/fpcond/pen/IvJDA?editors=101

My question is: How accurate is this? I ask because when I click on the polygons some of the population numbers seem really low for the size of the area.

Thanks

  • so is this accurate? – user3271518 Apr 30 '14 at 20:00
3

Note that your API request is for the population of all tracts in Allegheny County, not just in Pittsburgh. But I don't think the API supports tracts-in-place right now.

I don't know the region that well, but I don't see any problems with your method. Note that you're using 2010 ACS estimates, when you could be using 2012.

http://api.census.gov/data/2012/acs5?key=XXXX&get=B01003_001E&for=tract:*&in=state:42+county:003

Remember that the ACS data are estimates, so there is always a risk of sampling error. I don't think you can get the error as part of the Census API query, but you could get it from the American FactFinder, or from our project, CensusReporter: here is B01003 for all tracts in Allegheny county.

Our (Census Reporter's) advice to journalists is to use caution when the error is > 10% of the estimate. That is the case in 123 of the 402 tracts in Allegheny county. But also, the total of the tract-level population estimates from your API call is 1,223,066, which is quite close to the 2010 population from the Decennial census, listed as "estimates base" in the Allegheny County Quick Facts.

Do you have personal knowledge of the places with low population that makes you suspicious?

  • Well I am looking at tract 020100 which is downtown Pittsburgh and using my API it lists that there are only 5,325 people in that area. Also another tract that made me think my numbers were off was 422000 which is Fox Chapel a much larger area and yet the population is only 5,400. A side note is should i switch to block group for a more accurate picture? Taking a look at the CensusReporter my numbers seem to be correct. – user3271518 May 6 '14 at 19:00
  • I want to thank you for your answer BTW it has a lot of helpful resources. – user3271518 May 6 '14 at 19:01
  • 2
    smaller areas are more at risk of problems from sampling error, so you should only use block groups if you need to. The Census Bureau tries to optimize tracts at around 4000 people: before each decennial census they will split tracts that are much larger or merge those that are much smaller into adjacent tracts. / It looks like a lot of tract 422000 is park land, which would account for a smaller population than you expected. / Happy to help! – Joe Germuska May 6 '14 at 21:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.