Hi there, I am using ACS 5-year 2015 data, for basic demographic information (population, race, housing, education). I have about 350 tracts that are missing data, most seem to be in the 9900s (e.g. Census Tract 9900, Baldwin County, Alabama. Census Tract 9900, Contra Costa County, California, etc.). I’m wondering if this was somehow created for boundary changes? Or if there was some other explanation for why these exist at rows, but are empty outside GISJoin related information. Thanks!
There are two factors that might explain the missing data. First, the Census Bureau suppresses data for some records in ACS summary tables, as explained here. Second, NHGIS clips coastal water areas out of its shapefiles. This includes some tracts that are entirely offshore. Tracts that are comprised entirely of water area typically have 99## codes, and vice versa. There may be a few exceptions I’m not aware of, but I think that in all such cases, the population and housing counts are zeros.
Why do they have these coastal off-shore census tracts? What is their purpose?
As I understand, the purpose of off-shore tracts is mainly “filler” in order to simplify the partitioning of the U.S. into standard census areas.
The Census Bureau maintains a geographic database, the MAF/TIGER system, to define the boundaries of data collection and tabulation areas and to locate responses within those boundaries. They publish boundary data from this system in “TIGER/Line” files, and in turn, NHGIS shapefiles are based on the TIGER/Line files.
The spatial extent of the MAF/TIGER is set to a three-mile limit off the coasts, and it includes the full extent of U.S. territory within the Great Lakes (to the Canadian border). The MAF/TIGER subdivides all of this area among states (or state equivalents), counties (or county equivalents), and census tracts.
The Census Bureau’s 1994 Geographic Areas Reference Manual has a chapter on Census Tracts and Block Numbering Areas (BNAs). (BNAs are entities from 1990 and earlier censuses that are functionally equivalent to census tracts.) On page 15, it states:
One of the changes brought about by the TIGER System was the need to include all area (land and water) within a census tract/BNA. Rather than extending the census tract/BNA boundaries into the Great Lakes or out to the three-mile limit in coastal waters, the Census Bureau decided to close off the census tract/BNA boundaries along the shoreline or just offshore. The Census Bureau then assigned a default census tract/BNA number 0000 to the coastal and Great Lakes waters not assigned to any other census tract/BNA.
In later census years, it seems these “default” tracts typically have a 9900 number. I haven’t found any official documentation of that numbering being a requirement, nor have I found documentation explaining why the TIGER System entails a “need to include all area (land and water) within a census tract/BNA.” But it’s understandable given that this complete partition does simplify the model, ensuring that every area within the extent of the MAF/TIGER database is contained within a block within a tract within a county within a state.
Good to know, thanks!