We're having our first data integrity exam in a few weeks, and I'm trying to explain to management how the OCC calculates error ratios on the HMDA LAR based on key fields and line items. Before we submitted 2015 data, compliance identified two loans that were not on the LAR that should have been and two loans that were duplicated on the LAR. Those were corrected prior to submission. How do I properly calculate the error ratio in this instance?

On our last Data Integrity Exam we had a loan that was secured by 20 properties that was reported 20 times and was not split up properly, therefore we had 20 incorrect entries to the LAR, which was obviously over the error threshold. I was just looking at our error ratio report from our examiner and she didn't even include the 20 errors. It could be the fact that there were so many other errors that she focused on the real errors as opposed to just this one loan. (Trust me, this Data Integrity Exam was a nightmare - I'd just started working with the bank and went into the exam just a few weeks later, and found out all of this fabulous info. They made us re-submit 2012, 2013 & correct all of 2014)

However - I would count each field on your missed loans or duplicated loans as an error, my rationalization for that is the fact that you're only going to get tested on 60 loans no matter how many loans you have submitted on your LAR. You can only have a 5% error ratio in "non-critical fields" (Every field except for application date, action date & income) or you cannot go above a 10% error ratio overall. So you can make only 3 errors in any one field or 6 errors overall. If you've missed 2 loans off your LAR, and duplicated 2 more, well you've got 4 errors in one field and that's over your threshold!

I may not be right but I'll be darned if my bank messes up another data integrity exam! Trust me, I'd rather be overly cautious on this especially with HMDA data leading into Fair Lending, and the reliance of this data for CRA.

Best of luck to you in your exam.

First off, if you corrected all four of the loans prior to submission, you have ZERO errors on your submission.

Error ratios are not only based not only on the size of your LAR but on the number of loans that have a certain answer in any given field.

Read this for more on the topic.

Note: Randy sums it up nicely in the last post in that thread.

I know the errors have been corrected; I corrected them. As I mentioned above, I'm trying to explain to management how the OCC calculates error ratios had those been identified in the data integrity exam. The bank is a new HMDA reporter; we've submitted two years of data. Management does not yet understand the importance of HMDA accuracy. The scrub was terrible. We had errors in 21 of 180 lines of data--not including the ineligible loans or the loans that should have been included but were not. I know how to calculate single field errors based on key fields and line items. What I do not know is how to treat those that were added and omitted. My materials tell me there are 24 critical fields, so when doing the calculation, do I count 24 errors for each of the loans?