I work as a statistical programmer for an economist at Princeton. We have been working with the ACS data using the PUMA variable starting with 2014 and working backwards in time. We got to the 2004 data, and realized that even though the PUMA data is there, it is not providing geographic information below the state level. We wondered if this information was available at all?
The ACS was not fully implemented until 2005, and during the demonstration period of 2000-2004 only a limited number of counties were sampled. Unfortunately, this means state is the lowest level of geography for that period of time. Please see Chapter 2 of Technical Paper 67 for more details.
I hope this helps.
Thank you so much for the quick response!