Has anyone encountered instances where the address entered on the FFIEC's site results in a census tract number, but when the map is pulled up the tract line runs along the street and the actual address belongs in an adjacent tract? If so, have the regulators said anything about such instances. We are contemplating pulling the map up each time just to make sure the property is not on the tract line.

Yes. This has resulted in some issues at one of our affiliates recently.

We use a third party sevice to do our geocoding. This service provider is one generally recognized by the regulators as a "trusted source" of geocoding data - their determinations are normally automatically accepted as accurate. An examiner with a little extra time "verified" some addresses against the FFIEC data base and came up with some errors. Some of the errors were bona fide, but others were wrong in the FFIEC system.

The reality is, no system is 100% foolproof. But, in my experience, if you can show an examiner that the FFIEC database is wrong, then they will accept it.

Be aware, compliance managers, some examiners are beginning some old assumptions about your data. Data integrity reviews and verification of the accuracy of your systems is a good idea.

Which third party do you use for the geocoding?

If you don't want to mention it here, could you please e-mail it to me? THANKS!
bdyess@golden1.com

I have spoken to folks at both the OCC and FDIC on this issue. They both assured me that as long as you can show them what you input and the geocode output, they will accept the geocodes you use. We have made it a part of procedure to print the geocode output from the website and place it in the file. That way if we have a geocode that doesn't match what they have, we can show what we input on the site and what geocode was given to us. I use the FFIEC website and the Eagle EZ Locate website.

I've had examiners test geocodes we got off the FFIEC website, and tell us they were wrong. We now use Centrax, and have never had one questioned.

Rick - one more post and you're a member of the 100 Club!

We found we had to be careful NOT to put the zip code in on the FFIEC site. In certain cases doing that sent us to the wrong MSA!

A part of me was hoping for that answer Rick. However, I just wonder how many differences the examiners would have to find before they tell us to go the extra step to look at the map. After all, if we knew it could happen and chose not to because of the extra few seconds it took - do we really have a valid excuse for reporting loans incorrectly? Regardless of the fact that we can show that the initial geoding results prove the data we reported. Besides, the correct tract may be a low or mod with the wrong one being mid or upper.

Check with Don Narup at CRA Info Services, he posts here and he should be able to answer your specific questions.

Census Tract boundaries run down the middle of a street. The tract number is determined by the odd and even number of the street address. Odd numbers on one side. even numbers on the other. So if you look at a map you have to know which side of the street has the odd or even number to determine the tract.

If a geocoder makes a match on the actual street address it has taken into account the odd or even nature of the address. It's accurate so go with it.

If the match is made using the zip +4 code then its a different ballgame, and something called a "Centroid" is used to determine the tract number. It would take several pages to explain all this, so just accept that a centroid is basically the "Center" of a defined area. This means that any address that is roughly within a 6-8 square block area (Zip+4 size) will be coded for this area, instead of the individual address.

Using a zip+4 geocode,if the address in within 6-8 blocks of a tract boundary, or a tract boundry runs through the area, you have a 50/50 chance of getting the tract the addresses is actually in. Thats because the geocoded tract number,is based on where the Centroid (center) is located within that area, and not on the actual address.

Geocoding down to just the 5 digit zip code can result in getting an incorrect county,FIPS code, as well as an incorrect MSA, as its usually a larger area which may cross county lines. It may also include both single or multiple MSA's or a non MSA area. Wherever the centroid is located is what determines geocoding information. I do not recommend using anything less than a zip+2 geocode. Unfortunately, the FFIEC on line geocoder does not tell you the level of the match its made.

Remember, the regulators have the same geocoding problems you do, and most examiners understand that differences will occur. If what you have reported is not way off and you can show a good best effort to obtain the information they will usually accept your code.

There are only a few companies that provide the databases used by geocoding search engines. While there may be differences in methodologies used to search for address matches by the various geocoding vendors and software, the actual databases used in the programs are probably the same.

Many differences in geocoding can be traced to the address format inputted for the search, and not to a problem in the search engine. Accurate geocoding is dependent upon an address being formatted to US Post Office Standards. We have geocoded thousands upon thousands of addresses, and find that about 12-15% of addresses in a banks database need to be corrected for spelling, prefix (N,S,E,W) and suffix (Blvd,St,Ln,Ave etc) in order to get good geocoding results. Plus incorrect zip+4 codes.Thats why we run an address check using the US Post Office database of all US Addresses, to correct spelling, and prefix/suffix data and obtain current zip+4 codes before geocoding. The USPO updates this file every 60 days.

I have made this offer before and if you have an address you just can't code, e-mail it and we will see if we can get it for you.

BOL was kind enough to post an article I wrote for a Federal Reserve compliance publication. The artice lists some solutions to geocoding problems. "Avoiding Common Pitfalls of Compliance Data Collection " It may help some of you.

www.bankersonline.com/compliance/datacollection.html

or

http://www.frbsf.org/publications/community/investments/cra02-3/data.pdf

I have checked the FFIEC information to many flood determinations and appraisals prepared by outside sources and have not found a problem with it being incorrect.

The three primary sources of data are GDT, eTak (aka Eagle aka Tele-Atlas), and Tiger.

FFIEC uses Tiger, known to have to most stale and inaccurate information of the three majors.

The FFIEC has contracted with GDT and also includes the Tiger database in its search engine.

If a geocoder cannot make a match on the street address it goes to the next level to find a match.

Here are the Match Levels in decending order of accuracy.
1. Street Exact Match
2. Street Near Match
3. ZipPlus 4 Centroid
4. Zip Plus 2 Centroid
5. 5 Digit Zip Centriod
6. 3 Digit Zip Centroid (In some areas this can be the geographical center of the whole State. Don't use this match for anything)

The zip plus 4 is an area of about 6-8 square blocks. A zip plus 2 is a larger area. the zip 5 could be several square miles depending on where its loctated.

If a Zip code match is returned, the geocode is based on the geographical center of that area. A 5 digit zip code could have several census tracts within its boundaries, and the tract reported is the geographical center of the zip code. That zip code could possible cross county and MSA boundries but only the tract in the geographical center is reported.

That's why the party received an inacurrate MSA.

Fortunately, if you are coding addresses in a metropolitian area, odds are you are going to get street level matches and no problems. If in a rural or a high growth area with a lot of new streets being built, odds are you are going to get a zip code level match on most geocoders, and a lot of no matches on the FFIEC geocoder.