ACS 2012 5% and PUMA codes

I’m using the old PUMA codes to look at the 2012 ACS 5% data but the PUMA codes that I’m using don’t align. e.g., in 2000 and 2010 the PUMA codes for New Haven County, CT (statefip=9), are 01400 thru 02000; not all of these codes are in the 2012 data.

I know that there has been discussion elsewhere about new PUMA codes for the 2012 data which will align with 2010 outcomes, but I can’t seem to find these listed or mapped out anywhere.

I am not sure I am completely understanding your issue, but I will try to answer your question. The PUMA, MIGPUMA1, and PWPUMA00 variables are the only representation of the new PUMA codes that were derived from the 2010 Census. It is not possible to apply previous PUMA codes to the 2012 ACS data since the geographic areas that the previous codes apply to may have changed. As of right now, the most comprehensive list of PUMA codes and the areas they represent can be found on the TIGERweb data files page. TIGERweb also has a web application that allows you to view the PUMA codes in an interactive map. Simply select PUMA from the Map Layers list and zoom into the area you are interested in. You can get the PUMA code for a specific area by using the Identify tool (the ‘i’ in a blue circle).

I hope this helps.

Thank you, this is very helpful. So to confirm, if I am using 2012 ACS 5% sample from IPUMS I should use the PUMAs located here (…).

Are these the same PUMAs for the 2000 5% sample? If not, where would I find those PUMAs? I have been using the files located here (…) to determine PUMAs. (The link says the following:The PUMA boundaries in the ACS/PRCS are identical to those in the 2000 5% census sample. This led me to think that the PUMAs would be the same but it’s clear that they’re not.

Thank you again.

You are correct. The 2012 ACS PUMAs can be identified using the resources on the TIGER/web site, and the 2000-2011 PUMAs are detailed in the Maps available on the IPUMS-USA website (using the link you provided). The PUMA boundaries and codes change every 10 or so years, corresponding to the most recent decennial census. So, data from the 2000 decennial census, and the ACS/PRCS samples from 2005-2011 corrispond to the 2000 Census, and now the 2012 ACS corresponds to the 2010 Census.