Translate missing ACS 2005-2009 block groups into 2010 Census geography

[Lane]

When attempting to merge data into the crosswalk weights file in convert_00_geo.py, there are some data entries for block groups that are not present in the crosswalk with weights.

From Ashley:

I am attaching a csv file that can translate ACS 2005-2009 block groups into 2010 Census geography when there have been changes from the 2000 Census geography. There are 221 block groups in the file, 5 of which cannot be transformed. The variable "bkg09" is the 2005-2009 ACS block group and "bkg10" is the 2010 Census block group. It is easier to put the unmatched block groups in the ACS directly into the 2010 Census geography rather than translate them back to 2000 Census geography and then convert to 2010. This is because when the boundaries were re-drawn between the Censuses they often carried the new boundaries into the 2010 Census. The variable "nocompare" is an indicator for block groups that cannot be compared directly to either the 2000 or 2010 Census geographies. As mentioned above, there should only be 5 such block groups. These 221 block groups cannot be assigned to 2000 blocks. They can be transferred from the 2005-2009 ACS to the corresponding 2010 block group with a weight of 1.

changes_09acs_to_10cen.csv.txt

[Lane]

In addition to mapping some block groups from the ACS 2009 into 2010 GEOIDs, some of the block groups will also need to be mapped to 2000 GEOIDs.

They will be handled in this way:

ACS 2009 mappings to 2010 GEOIDs (changes_09acs_to_10cen.csv) Any block groups present in the 2009 -> 2010 GEOID mapping should be added to the weights dataframe in covert_00_geo.py with the 2009 GEOID in the GEOID00 column, 2010 GEOID in the GEOID10 column, and a count_weight and rate_weight of 1 because the data directly maps into 2010 geography.

ACS 2009 mappings to 2000 GEOIDs (changes_09acs_to_00cen.csv) The entries in this file should be updated immediately after fetching the data from the census in CensusDataStore. This ensures that the data will go through the weight calculations for 2000 -> 2010.