Wednesday, January 14, 2015

Big Data and the Collateral Underwriter

I would like to express my concern about the implementation of the CU in January.  Unlike major changes in the recent past this one is being sprung on appraisers in just a few short months, and few are aware of its impact.  Fannie Mae is serious about this, and the Appraisal Institute's recent article revealing that Fannie is doing away with the 10, 15, and 25% rules confirms this.  

If someone is not familiar with multi-variable regression they may not see the correlation between the two, so let me explain.  Regression is backed by complex mathematical formulas that extract components of value by comparing a large number of properties from multiple sub-markets.  On the surface this sounds fine, but I use regression on a regular basis, and even with directing the process, the regression will occasionally output values that are not typical in our markets.  For example, I performed an analysis on a home located on Audubon Pkwy in Louisville, and one of the regressions output was an adjustment for fireplaces to the tune of $14,000 which was not used in the report.  Adjustments like this rapidly increase line, net, and gross adjustments which is why Fannie has rapidly done away with these rules.  Gross adjustments are still a good indicator of overall similarity.  Regression is rapidly being considered the 4th approach to value.  The problem with this is we don't know if the data is accurate, and Fannie has no intention of allowing us access to it.

According to Fannie Mae, "CU leverages an extensive database of property records, market data, and proprietary analytical models to analyze key components of the appraisal including data integrity, comparable selection, adjustments, and reconciliation."  What this means is that reports ran through the CU (or VA's version, FHA will not be far behind) will be scrutinized based off these models.  If you are not using your "$14,000" fireplace adjustment in your report, then it will be flagged.  I've also heard that the CU will be comparing our data and adjustments to our fellow appraisers.  On the surface that sounds fine (which is at the heart of being peer reviewed), but Fannie's statement doesn't mention any of this.  It's extensive database it the only thing that was contributed to by our fellow appraisers.  

If Fannie is comparing all adjustments to the "analytical models" then we have a fundamental problem.  In their recent update to the selling guidelines concerning our analysis of adjustments, " The appraiser’s adjustments must reflect the market’s reaction (that is, market based adjustments) to the difference in the properties."  I don't see where we are to base our adjustments on statistics or analytical models.  In my practice, I use statistics and analytics to support my market data, not the other way around.  

Although the use of the CU is not mandatory, I'm sure lenders will jump on board.  What does all this mean?   1. If an appraiser is not using regression as their primary approach to value, then their reports could be riddled with flags.  Bradford Software conducted an experiment after they got access to the CU.  They submitted a non-regression derived report from a reputable appraiser, and the CU assigned it a risk level 5.  The same report was then submitted after regression was introduced and the risk level returned at 2.  Bradford did not say if there was a significant difference in value or if there was a change in comp selection.  2. We will now be reviewed and scrutinized not our peers, but by a mathematical algorithm.  I'm sure lenders will copy and paste these flags as revisions and send them back to the appraiser for explanation.  Even if the appraiser took due diligence in deriving their adjustments from the market, they will be inundated by a tidal wave of revision requests.  Lastly, I see appraising as part art part science.  What will the outcome be when the CU tries to analyze an appraisers "gut instinct"?  

I have been using a regression tool for several years, but in light of the upcoming changes I’ve been looking into other companies that provide this service.  If you have a favorite regression product, feel free to share and why you are pleased with it.

In conclusion, Fannie Mae, VA, and FHFA are all gearing up for some form of analytical review.  We will not have access to this data, and our reports will now be compared to the CUs database.  Our adjustments will be judged against statistics and analytics instead of the market's reaction.  Lastly, appraisers who do not use regression as the primary approach to value could find themselves swimming in a pool of revision requests.  Thank you for taking your time to read this.  Stay tuned for a series of articles on what it going on in the Hardin County area in regards to the changes happening on Ft. Knox.


1. http://cloudtimes.org/

No comments:

Post a Comment