FFIEC 2015 Census Data Released

The 2015 data has been released. It is available in the demographic state/county/MSA tables:

http://www.ffiec.gov/census/default.aspx

and in the downloadable searchable program and flat file (entire country in flat file)...these two are very large files.

http://www.ffiec.gov/censusapp.htm

There are several dozen tracts in the FFIEC census files that are incorrect. They are mistakes made originally by the Census Bureau and copied by the FFIEC. The Census Bureau has since corrected the tract identities but the FFIEC did not do so in 2014 after the mistakes were identified and corrected by the Census Bureau. This means that your software vendor may be using the correct tracts but they don't completely match the FFIEC file. The tracts are primarily concentrated in AZ and NY but there also are a few in some other states. So I suggest you contact your software vendor to find out if they are using the files that match the FFIEC, but not the Census Bureau. We had thought that the corrections would have been made by the FFIEC but in e-mails earlier this year (2015) I exchanged with them I have determined they did not make the corrections and I have been told they don't intend to correct them until 2017. The file specs and descriptions accompanying the new 2015 FFIEC Census files don't indicate they have made the corrections either.

Did you mean to post this in the thread about the geocoding error?

No KB - my point is that the FFIEC census tract file has errors in it with respect to census tract identities.

We discovered the errors when we were mapping an area for a GeoDataVision client. Typically we color code the tracts to represent the income class of each tract. But we noticed some tracts did not have any color. We wondered why and when we researched the issue we discovered that the mapping software had dozens of tracts not in the FFIEC software and vice versa.

Initially we thought the error was in the mapping software but after discussing the issue with the vendor and researching the matter we learned that the mapping vendor had updated their tract identities to reflect the corrections issued by the Census Bureau in 2013. We then contacted the FFIEC to point out the errors and they acknowledged the mistake but advised GeoDataVision they would not make corrections to the erroneous tract identities until the 5 year ACS changes which will be in 2017.

So anyone using mapping software and anyone using independent geo-coding may have discrepancies that ironically reflect errors in the FFIEC files, not in the vendor file. Nonetheless, because the FFIEC is not making the corrections it will be important to use the incorrect tract identities for CRA and HMDA purposes until the FFIEC adopts the correct tract identities.

The latest post in the CRA thread with respect to geo-coding has to do with the location of the property being on one side of the street or the other side when a tract boundary runs down the center of the street. That is a different type of error caused not by incorrect tract identities in the FFIEC file, but by a disagreement on where the property is located. I did not want to include this post in that thread because they are two totally different types of errors.

Thanks for clarifying. I only asked because you mentioned "your error" and coincidentally yesterday someone posted about a geocoding discrepancy with the FFIEC site, as you mention above. It seemed you were referencing someones specific situation vs this overall issue.

KP - I don't see where I used the phrase "your error". What I did say, "your vendor may be using the correct tracts but they don't completely match the FFIEC file" which points out a crazy situation in which vendors may have the correct census tract files but they won't agree with the official FFIEC files. So vendors will have to deliberately use the tract files with tract identity errors.

I guess that was it..your vendor...it just seemed you were speaking to a poster directly (other than me). Just how I read it.

The FFIEC is crazy not to make this change when it is so critical to analysis. Banks don't pay attention to tract income for fun.

So I urge all bankers reading this thread to contact their software vendors for CRA and HMDA and for mapping to discuss this issue and determine if they are using the "correct incorrect" (now there is an oxymoron) census tract file. It can happen only in the world of regulatory compliance!

The 2015 FFIEC estimated MSA/MD median family income figures are out: 2015 figures.

And those numbers are in the data linked at the top of this thread.