We have been reviewing the new Census 2010 tracts and have noticed there are dozens of tracts that are underwater tracts even though the FFIEC has stated they don't include 209 underwater tracts in their Census database.

The inclusion of underwater tracts has confused a number of banks who are clients of GeoDataVision and who have gone to the FFIEC website and seen those tracts included in the FFIEC census files. These banks believe they should include the underwater tracts that have not been dropped by the FFIEC in their Assessment Area delineations. Apparently not all underwater tracts were deleted by the FFIEC.

We have communicated this to the FFIEC and received a response that they will continue to include certain underwater tracts within their census files. All these tracts that we have discovered are numbered 9900100. So if you see tracts inside your AA numbered as per above they probably should be omitted. If you have any doubt we suggest you go to the Census website and look up the tracts in the official Census maps to see if the tracts are underwater.

What do you mean by "underwater" tracts? I'm not familiar with that term in this context.

Tracts that there is no possible way, short of a boat, to make a loan in!

Examples include bays, large lakes, etc.

They are usually numbered as tract 9900.00. I think Len put in a 1 instead of a decimal.

Tract 9900.00 is for bodies of water. They used to be all 0s and are usually removed by the FFIEC; banks don't usually come across them because for our purpses they are meaningless.

Different numbers mean different things:

Some ranges of census tract numbers in the 2010 Census are used to identify distinctive types of census tracts.  The code range in the 9400s is used for those census tracts with a majority of population, housing, or land area associated with an American Indian area and matches the numbering used in Census 2000.  The code range in the 9800s is new for 2010 and is used to specifically identify special land-use census tracts; that is, census tracts defined to encompass a large area with little or no residential population with special characteristics, such as large parks or employment areas.  The range of census tracts in the 9900s represents census tracts delineated specifically to cover large bodies of water.  This is different from Census 2000 when water-only census tracts were assigned codes of all zeroes (000000); 000000 is no longer used as a census tract code for the 2010 Census.

The additional underwater tracts all appear to be numbered 9901.00 or 990100 depending on your naming convention. The problem is that banks are including these tracts within their Assessment Areas because they are included on the FFIEC site. Unless you are financing a shrimp boat with a permanent mooring in the middle of a bay or body of water you will never have loans in those tracts. Consequently they will appear as lending gaps when you do lending gap analysis. This is downright silly of course.

To make matters worse, the FFIEC just e-mailed me and told me they will add back all the underwater tracts that were omitted in last year's file. This will exacerbate the problem of course. I cannot explain why the Agencies are adopting this approach. I believe there was a misunderstanding between the Census Bureau and the FFIEC regarding the identities of underwater tracts and this was the genesis of the problem. But now with the original tracts that were omitted in the 2012 files being added back in the 2013 files there is sure to be more confusion.

Thanks for the explanations, everyone. I was going to make a snarky remark in my original post about the tracts being lakes or streams. Little did I know that was the actual case!

Only the federal government could list an area without any census population as a census tract.

Honestly, that makes sense to do. The problem comes when they then expect you to LEND in those areas!

By adding them back in, banks will have to fully understand what those tracts represent.

As will examiners.... wonder who is most likely to mess it up?