County level estimates

I may not have understood your original question. Are you asking how to create county-level estimates from the PUMS data (which does not contain count identifiers)?

Exactly.

OK. This is a bit involved, and there are several approaches to getting estimates for a political geographic area from the PUMS data. For starters, I would suggest going to the Missouri Census Data Center. You can use an application on that site to create a file that gives you the proportion of population in any county that is entirely or partly contained in a given PUMA. Here's the url: mcdc.missouri.edu/.../geocorr14.html. I have run this for counties in all states and have pasted in the first few records in the .csv output file below (including the column header records). You will see that PUMA 00100 in Alabama contains 4 records for the proportion of each county covered by that PUMA (based on recent pop estimates). By sorting the file on the "county" field and "puma12" within county, you can use the "alloc factor" field to estimate the records from each contributing PUMA to allocate to the county. However, that assumes a uniform geographic distribution of characteristic (= measures) across the contributing PUMAs.

There are several steps that could be taken to refine this approach using known estimates at the county level from the aggregated data published from the ACS (I assume you want this for all counties which means you would have to use the aggregated data from a 5-year period for the ACS). You can read about one approach to this done by the Federal Reserve Bank of Philadelphia at this site: www.philadelphiafed.org/.../methodology.pdf. Also, if you can get access to the book Millenials in America by Rovert Scardamalia (Bernan Press, 2015), the preface of that book describes a method for allocating PUMS data to other geographic areas.

Hope this is useful.

Doug


"state","puma12","county","stab","cntyname","PUMAname","pop14","afact"
"FIPS state","puma12","county","State Postal Code","cntyname","PUMA12 Name","Pop 2014 estimate fr county level ests","puma12 to county alloc factor"
"01","00100","01033","AL","Colbert AL","Lauderdale, Colbert, Franklin & Marion (Northeast) Counties",54543,0.292
"01","00100","01059","AL","Franklin AL","Lauderdale, Colbert, Franklin & Marion (Northeast) Counties",31601,0.169
"01","00100","01077","AL","Lauderdale AL","Lauderdale, Colbert, Franklin & Marion (Northeast) Counties",93096,0.498
"01","00100","01093","AL","Marion AL","Lauderdale, Colbert, Franklin & Marion (Northeast) Counties",7725.125,0.041
"01","00200","01083","AL","Limestone AL","Limestone & Madison (Outer) Counties--Huntsville City (Far West & Southwest)",90787,0.494
"01","00200","01089","AL","Madison AL","Limestone & Madison (Outer) Counties--Huntsville City (Far West & Southwest)",93157.849,0.506
"01","00301","01089","AL","Madison AL","Huntsville (North) & Madison (East) Cities",124425.297,1

OK. This is a bit involved, and there are several approaches to getting estimates for a political geographic area from the PUMS data. For starters, I would suggest going to the Missouri Census Data Center. You can use an application on that site to create a file that gives you the proportion of population in any county that is entirely or partly contained in a given PUMA. Here's the url: mcdc.missouri.edu/.../geocorr14.html. I have run this for counties in all states and have pasted in the first few records in the .csv output file below (including the column header records). You will see that PUMA 00100 in Alabama contains 4 records for the proportion of each county covered by that PUMA (based on recent pop estimates). By sorting the file on the "county" field and "puma12" within county, you can use the "alloc factor" field to estimate the records from each contributing PUMA to allocate to the county. However, that assumes a uniform geographic distribution of characteristic (= measures) across the contributing PUMAs.

There are several steps that could be taken to refine this approach using known estimates at the county level from the aggregated data published from the ACS (I assume you want this for all counties which means you would have to use the aggregated data from a 5-year period for the ACS). You can read about one approach to this done by the Federal Reserve Bank of Philadelphia at this site: www.philadelphiafed.org/.../methodology.pdf. Also, if you can get access to the book Millenials in America by Rovert Scardamalia (Bernan Press, 2015), the preface of that book describes a method for allocating PUMS data to other geographic areas.

Hope this is useful.

Doug


"state","puma12","county","stab","cntyname","PUMAname","pop14","afact"
"FIPS state","puma12","county","State Postal Code","cntyname","PUMA12 Name","Pop 2014 estimate fr county level ests","puma12 to county alloc factor"
"01","00100","01033","AL","Colbert AL","Lauderdale, Colbert, Franklin & Marion (Northeast) Counties",54543,0.292
"01","00100","01059","AL","Franklin AL","Lauderdale, Colbert, Franklin & Marion (Northeast) Counties",31601,0.169
"01","00100","01077","AL","Lauderdale AL","Lauderdale, Colbert, Franklin & Marion (Northeast) Counties",93096,0.498
"01","00100","01093","AL","Marion AL","Lauderdale, Colbert, Franklin & Marion (Northeast) Counties",7725.125,0.041
"01","00200","01083","AL","Limestone AL","Limestone & Madison (Outer) Counties--Huntsville City (Far West & Southwest)",90787,0.494
"01","00200","01089","AL","Madison AL","Limestone & Madison (Outer) Counties--Huntsville City (Far West & Southwest)",93157.849,0.506
"01","00301","01089","AL","Madison AL","Huntsville (North) & Madison (East) Cities",124425.297,1

Thank you.