Does anyone have any suggestions on how we should test the accuracy of our HMDA-LAR? We have procedures in place that should be catching errors, but we know that some are slipping through. We will have about 3000 loans to report for the year... Thanks!

I believe the HMDA software that can be downloaded from the FFIEC website has a procedure that will not let you submit the LAR if there is a "validity error." If you are using this software, I would simply activate this feature.

Thank you for the suggestion. We are already doing that, but we're also looking for ways to make sure that what is found in the loan file is accurately reflected on the LAR.

You might want to do a random sample test and verify each piece of information on the LAR with the information in the file, for both originated and non-originated files. You want also want to double check the loans on the LAR with a loans closed report. Same thing with the non-originated loans. That way you'll know that you are capturing all of the loans that you should, and also none of the ones that you shouldn't. Go through each file and manually verify that the information matches -- file to LAR. If you have a high percentage of errors, you may want to do a total rescrub of all the files on the LAR. Better to do it now than January 20. I also found that it was easier to manage if it was done on a regular basis -- monthly or quarterly.

Thank you both!

Have you tried Excel pivot tables? I highly recommend using them because they allow you to isolate individual items like borrower race and compare it directly to borrower sex. So if you get a 5/5 (white/no co-applicant) you not only spot the error immediately, the pivot table will tell you which file it's on.

I know Andy Z wrote two Word Smarter, Not Harder articles regarding pivot tables in recent ABA magazines.

PM me if you'd like help and/or more info.

We perform a monthly audit of HMDA and CRA reported and potentially reportable loans using OCC guidelines for sample size, key fields and percentage of key field errors (which is currently 5%). This allows us to scrub any key field with an error ratio for that month in excess of 5%.

PM me if you would like more information.