TIGER2021 Census Tract Shapefiles vs. FFIEC.gov Census Flat File

Hi All,

I'm glad I found this forum, have a question for a current project that I'm working on.

A discrepancy in the total number of Census Tracts comparing:

This analysis only includes Census Tracts within the 50 states and DC.

For the 50 states + DC, there are:

  • 84414 Census Tracts in the TIGER2021 Shapefiles.
  • 74771 Census Tracts in the FFIEC.gov 2021 Census Flat File.
  • These two sources only have 61638 Census Tracts in common for the 50 states + DC.


  1. Retrieve the Census Flat File columns for all the unmatched Census Tracts in the TIGER2021 Shapefiles.
  2. Retrieve the Shapefiles for all the unmatched Census Tracts in the FFIEC.gov 2021 Census Flat File.



  • Hi CJ--

    The 74K numbers sounds a lot like the number of tracts up to 2020.  You should check and see if the supposed 2021 tracts actually match the 2019, which are the 2010 tracts with a few edits.


  • Hi Andrew,

    Thank you for your response.

    I need the Shapefiles to calculate the Census Tract centroids, and this is the only source for the Shapefiles that I can find:


    The TIGER2021 Shapefiles yield the 84414 Census Tracts mentioned above.

    I’m seeking a better source for the latest Census Tract Shapefiles, if the link above is the wrong one.

    Thank you,


  • That repository at www2.census.gov/.../ stores 2021 shapefiles. You can generally get to any other years' shapefiles just by changing the year in that URL. So to get 2019 shapefiles, use www2.census.gov/.../.

  • Also, you might like to try using IPUMS NHGIS, which provides census tract shapefiles based on TIGER/Line files, but in nationwide files (rather than split up by state), clipped at the coastlines and projected into an Albers equal area conic projection, suitable for nationwide mapping and analysis.

    Or for census tract population centers (which are more suitable for many types of spatial analysis than the shapefile centroids), see this Census web page.

  • Andy Beveridge <andy@socialexplorer.com>

    6:32 AM (39 minutes ago)
    to American

    Dear CJ--

    According to the FFIEC documentation, they are using the 2011 to 2015 ACS, probably actually updated.  Those tracts are the 2010 tracts which are being used for the ACS through 2019.  They are using the ACS for their info.  They could not switch to the 2020 tracts until the ACS 2016-2020 is released, which will probably be in early March.  That is the problem you have.  The Tract data from the Census Bureau for the 2010 blocks does have Centroid data for the tracts build in.  

    You should read the FFIEC documentation about what years of the ACS 5 year file they are actually using.  The actual file that is out is the 2015-2019,  I don't if this is a research project or what you are doing.  

    I would be happy to talk to you.  If you send me an email with your phone number.  


  • Hi Jonathan, thank you for this insight.