We are obviously not as careful on consumer loans as we are on HMDA. I have many PO Boxes, building numbers (on a military post), customers out of the country and out of state, etc.

Are there any guidelines on the examiner's review of your consumer loans for geocodes? Most of them in our trade area we have, out of areas have more errors and there are still the PO Boxes as this comes from the CIF address.
Yes, I am preparing for my first large bank exam.

Andy -
Since you do not have any response yet, I will attempt to answer your question hoping it's what you are looking for.
At our last CRA Exam (2001), we opted to have our consumer loans counted. We have a military base nearby and probably have had the same dilemma. We require all loans, regardless of type to have a the geocode information prior to boarding. The examiners looked in each file to verify how we obtained the census data. We do not accept PO's, always obtaining a physical address. All loans that originated used their current physical address - if they moved out of the country later it didn't affect our reporting.

The examiners spent one week on just verifying the data we submitted on the CRA LR. (Number, date, size, revenue category/annual income, census data) Fortunately we had no errors (what a relief!). They came back about three week's later to conduct the formal exam under CRA as well as Compliance.

Each month, all loans are reviewed (yes, every single loan). We're on ITI, so it makes the verification process somewhat easier when it comes to verifying collateral, type of loan etc. Fortunately we have CRA Wiz, and the data is downloaded which makes the process easier.

At this juncture I don't have many of those luxuries. They have increased the count of soldiers who are to deploy so we'll soon be losing 30,000+ of our customer base. While loans will decrease, I still have many, will continue to have many and couldn't possibly integrate that into my program. With this big a deployment, new staff isn't happening. We do several hunder new loans a month.

Of my 18,000 consumer loans I have about 2,000 that are not coded due to PO Boxes, having only a building number, etc. I hope there is some ratio that is acceptable, and that I can get into it.

Just a thought, but couldn't you send those records out to Rata or CRA Wiz or one of our member's who is in the business to geocode? You definately need a full time person assigned to CRA Data records.

Andy, we don't report consumer loans, but I keep hearing from many fronts that if you're providing the examiner with something they didn't require, they can't use it against you, only to help you. With 2,000 out of 18,000 (approximately 11%) error rate, I'm not sure they could claim that the data was meaningless due to the percentage geocoded.

We have a policy that no loan gets booked without a physical address - no exceptions, and we researched many P.O. Boxes before filing our LARs - it happens and you can't stop it!

I'd like to hear from someone who has used consumer loan info in an exam, and faced this issue - maybe Dawnie?

I too am waiting to hear opinions from Dawnie and Don.

The problem is, some customers do not have a physical address other than, for example, Bldg. 20067, Ft. Hood TX 76544. That is not geocodable, but it is where they live and where they get their mail. Building numbers don't seem to have a process so they can't easily be determined.

Also, with the state of readiness, asking someone on post to show me where these buildings are, how are they laid out, where is this particular unit, etc. may get me on some national security OFAC type of list.

I do not foresee the best method as being workable, that is, "Customer, here is a map, show me where you live on this".

This presents many opportunities to succeed. That is, challenges, that is problems.

If you absolutely cannot get a physical address, but know that your customer is somewhere in that zipcode, you can do a "zip centroid" geo-coding that is so-so on accuracy. The Zip Centroid is the Census Tract that is most central to that particular zipcode. It's better than nothing.

If that's the route you have to go, be sure you are ready to back-up your case on why this must be so - ex: Military base, military quarters not on any geo-code records, etc. If your examiner is at least partially reasonable, and he/she can see that you UNDERSTAND where you are lending and what you are doing (as opposed to those shops where the newly appointed Compliance/CRA "officer" freezes like a deer in the headlights before the glare of the examiner, then you SHOULD be okay.

Hint - make sure you guys have really good coffee during the exam!

The bank is responsible to determine an accurate geocode on all records it ""Reports"". In the case of no address, only building numbers on military bases that are large enough to have multiple census tracts, and in the case of Fort Hood also in two different counties this is not an easy task. It really comes down to an educated guess.

If an examiner can look on a map and determine the geocode by using cross streets or knowledge of the area, it means the bank can to, and its what's expected.

At Fort Hood It seems that over 98% of the barracks are located in one of the tracts. Using that tract exclusively will get you at least a 98% accuracy rate. In this case both tract 105.00 in Coryen Cnty and 232.00 in Bell Cnty which divide For Hood, are classified as Modetate income tract. So which ever one you pick will have the same analysis results.

By showing the logic as to why that tract is used you present a valid case, and as it is not an arbitrary pick examiners should accept it. You will be expected to prove that you have made a determined best effort to accurately geocode an address, and unless examiners can punch a lot of holes in your stated best effort, they will usually accept what you have. Remember, they do not have any different or better tools to determine a geocode than a bank does.

Rural addresses are another story. Geocoding engines will not code every address and when the level of geocoding gets down to the 5 digit zip code, which it does for many rural areas in the country, the accuracy is not good.

The boundaries of a zip code can cover a large area. Within that large area, there may be multiple census tracts. A 5 digit zip code centroid is the center of the zip code and that center can be in any of the tracts inside the boundary of the zip code. The census tract geocoded has nothing to do with the actual address only where the center of the zip code is located.

Of course if the zip code has only one census tract in it its accurately geocoded.

The really difficult address require a map. A map the shows both census tract and zip code boundaries as well as streets. There is cracker jack inexpensive program on the market that we use exclusively to do this. I have sent maps to a number of BOL'ers using this program. If you want to check it out e-mail and I will send its name and web address.

It comes down to the bank doing what is necessary to obtain a geocode and to explain successfully any idiosyncrasies that cause records to not be geocoded, or why the bank has not geocoded them.

You cannot and should not rely on geocoding programs alone. It will require some time and indiviual efforts to obtain census tract on some addresses, and this is where the skill and knowledge of the proceses come into play.

I appreciate the feedback. It is all constructive.

I'm late in replying (class all day and smoozing all night).

I had suggested Andy contact the base for help (this is how I ended up geocoding my base accounts) but I was lucky enough to have a branch on base. My manager and the base commander worked out the geocode system for me. I did this prior to 9/11 though.

I don't like going with an approximate myself, but given the issue of security, if you've got no one you can get more info from Andy, Don's method (we'd discussed that before as well) is going to have to do. At least they are in the same income level so it won't appear you're swaying your numbers to one or the other.

I did not submit my consumer loans during my last exam. They didn't provide extra support to my cause, nor did they make up the bulk of my portfolio, so I felt prudent to remove them, in great part because I don't spend time scrubbing the portfolio. I know this is different for you. If you do have to include them, they would need to be supported with accurate geo-coding, (or at least a best effort as Don mentions) in order to give your banks efforts a fair examination.

I was just having a conversation today with an examiner about the issue of reliable data. He did mention that consumer loans were one of the worst groups to look at. I would imagine that your issue is one of the reasons. Even my anal retentive review process (ARRP) is limited to HMDA and CRA Small Business/Small Farm, and it leaves out a review of consumer data, and I'm famous for being picky as heck!.

While I haven't had my consumer deals reviewed, I have reviewed the examiners workbook and it states that 5% of a submittal (not 5% of any one category but an entire set of reported records) is considered to be "unreliable for an accurate review". 5% in any one catagory is reason to require resubmission. We don't require submission to the FFIEC of our consumer data, but if we were going to request consumer data to be in our CRA exam (or be required to do so) I would be concerned at 11%.

Hopefully you can reduce this ratio by dealing with the base as Don has suggested. I personally would get your examiner to buy off on the idea first if possible. I find if I work them slowly into an idea they're more accepting when it hits them in the face (view AnonRegulator's exchanges with me currently LOL..someday I'll win the challenge). Then I would suggest you do a very good scrub on the remaining deals. This shows your best efforts have been made.

I am having breakfast with an examiner tomorrow. I'll ask him his view of this issue for another take on it and share with you later in the day when I return to the office or home. As just a side note of interest, he is working with the OCC(perhaps more agencies as well...I'll know tomorrow) on ways to reduce the headache of exams. I was going to tell him I thought the process fine (shoot me, I really do ) but I'll be nice and say they're too hard on newly "big" banks!

Anyone have comments on consumer loans that are out of your area, as in out of state. I don't know how much time I should devote to cleaning these up if they get tossed out by virtue of the state code.

Andy
Not sure what you mean by tossed out. The number of loans outside your assessment area will obviously have an impact on the percentage of loans you have inside the asessment area.

Examiners generally like to see the number of loans originated inside you assessment area at around 70%. So, it is necessary to geocoding all you records to determine this percentage.

Were these loans actually made to out of state borrowers or originally to people in you AA who now live out of state.

Current mailing addresses may not be what you want to be using.

By tossed out I mean simply classified as out of the area and it doesn't matter that I geocode them.

They are existing customers, likely military and the loan may have originated with them in my market or not. The address I have is the current one on file. I am now refining the processes but started with our active portfolio.

I may end up getting a record of loans originated monthly or quarterly and appending them. That would give me the original address but requires more work.

Andy, it sure sounds like consumer loans are an important part of the bank's lending program, so you are on the horns of a real dilemma. At one point in my bank's history down here in College Station we had to rely on consumer loans to get enough credit under the lending test to get a satisfactory rating overall. This is no longer the case but we still collect the info on consumer loans. We pushed the cleaning up of the files out to the loan officers and processors with the support of upper management. I hope you are all set for the investment and service tests as well. A big bank CRA exam is a nightmare the first time around. The OCC should cut you some slack inasmuch as this is the first time.

Hey Andy, I just looked your bank up in the Texas Red Book. Our banks' profiles are very similar, a mixture of rural and secondary major markets. We are actually direct competitors in a few markets (Bryan, Dallas, Denton, Fort Worth, Houston, Irving, Midland, Odessa, Plano, San Angelo, Stephenville, and Wichita Falls). How do you accomplish CRA with all those communities ?

I'll tell you after the exam!

See if you can create some new fields on the mainframe system for CRA address. When the loan is boarded enter the address to be used for CRA purposes, even if its the same as the mailing address.

This way you will always have the address to use for CRA purposes no matter how mwny times the customer changes mailing addresses. Same applies to those Small Business loans where the proceeds were used at an address diffetent than the mailing address.

The IT folks may say its not possible as all the available fields are being used. If that occurres, it time to check to see if all the fields are really necessary or just something someone set up years ago that has outlived it usfulness.

I talked with Craig today about the percentages and he told me this. If he is examining a bank and finds that one in 10 records is flawed, it's very hard to give a decent rating or even produce a good exam. He endorsed the suggestion on the base loans, using the geocode where 98% of the housing is and being sure to mention your reasoning to the examiners to avoid any appearance of swaying the data towards a low or mod track. Also, since you won't have any loans in the other track, the base needs to be covered in your gap analysis, clearly defining the base tracts. No one expects you to be rocking the world of lending on a military base, but you have to tell them it is there

If that leaves you with a ratio of errors (or non-geocoded deals) above 5% it is very likely (if the data is checked for integrity) that your exam will have a note in it mentioning the weakness of the data and the examiners lack of ability to produce an accurate assessment of the bank's performance due to this weakness. It won't kill you, but it's not something you want to haunt you in the future if you can avoid it. When the next exam rolls around, the examiner in charge of that exam will read this while doing his homework and seeds of doubt could be placed in his mind.

Craig also mentioned that it is difficult to give a bank an outstanding rating when the examiner is uncomfortable with data integrity. The above note in your exam would likely ensure that you would not receive an outstanding, even if the numbers leaned towards one, because the examiner would be testing you on unreliable data. So, he said you should think about what kind of result you hope to achieve in this exam and take action from there. If you're boarder line satisfactory, the risk could be to end up with a needs to improve and a note on the lack of data integrity. If you're feeling pretty close to satisfactory and that is enough for you this time around, clean up the military loans and go forward from there if the resulting deals are less than or close to 5% in error.

If you want outstanding, you need to do an agressive scrubbing of the period being examined quickly.

Whatever you do, if you're going to leave the records as is, it's important to document your future plans to eliminate these weaknesses. This will show the examiner in charge that you are taking the issue to heart and while it might not get you outstanding this time, it could keep the examiner from mentioning the data integrity in the exam.

We had one other bank there at the meeting and she was once told to scrub her data on consumer lending due to geo-coding and lack of income documentation. The compounding issues might have been the reason but it was not clearly communicated as to if it was due to one category of reporting or the two combined.

Any suggestions on how to geocode when the physical address provided is a ship (e.g. USS Anchorage)?

That is what is known as a floating census tract, mobile.

If the boat is docked relatively permanently in a harbor you could always use the geo-code for that. (It is a code with a .99 on the backend of it). I believe (Don Narup can chime in on this because it's only a guess from finding a few of these) that the assignment of a geo-code to a harbor stems from having living quarters (i.e. houseboats or more permanent morage options) in the harbor. I notice that some of ours have them (the largest harbors) but others do not. If you know the geo-code of the land adjacent to the harbor the .99 extention might do it for you. I'd check with your geo-code provider to make sure that harbor is really a designated area.

In the 1990 census any tract the ended with a .99 extension was identified as a civilian or military ships.

In the 2000 census tracts with .98 and .99 extensions appear to have been eliminated, and and ship docks and piers are incorporated into regular census tracts of the area.

Use the census tract of the location the ship was docked at the time of origination

My soldiers have a residency where their unit is stationed, or at least that is what I have used. In the case of a Navy ship, it is a real possibility that the person lives ON the ship, that is where they get their mail and that ship may be at sea when the loan is made.

If these are not done by a bank that uses "military" as its designation (all military are "in", non-military are "out") then I see a conundrum unless you can base it on the home port.

The Bureau alleges that they counted the population on military bases.

This is necessary because the information is used to re-apportion Congressional districts.

If the base is so large that it spans multiple census tracts, would you expect that they have about the same
income levels?