But they kind of do as long as they can use an appraiser's analysis and opinions in a platform to punitively rate their work against the work of their peers. And if they don’t score well enough then they might get put on a watch list, or worse.
Welcome to Fannie Mae’s Collateral Underwriter. An automated warehouse of millions of appraisals, high level analytics, census track heat maps and various tools to improve appraisal quality and “… support proactive management of appraisal quality”.
That part about supporting proactive quality is an interesting use of terms since the only people that have access to this tool are lenders selling to Fannie Mae, and it’s in an ex-post-facto manner that the appraiser gets to know what’s wrong with her report after her report has already been run through the system and “graded” blindly.
Then she gets to fix it.
Now, not to be overly dramatic, there are some knowns in the process; 21 hard stops of key sections within the appraisal that if missed and/or not addressed will get flagged.
Other legacy guidance still holds true like the age old 15% net 25% gross adjustment rule. But what isn’t known are how the system determines that the comp selection is or isn’t appropriate, but even more importantly how adjustments are graded.
Are they measured against a model or are they measured against what other appraisers have used in the past within proximity to the subject? A combination of the two? Nobody knows, but we’re about to find out in January when Collateral Underwriter is officially rolled out.
There are all kinds of uses for data, and one of them is its use in real estate. Most of my conversations with people about real estate and analytics almost immediately reverts to Automated Valuation Models, those pesky little mathematical tools that were used ad nauseam at the height of the lending boom for everything from second mortgage decisioning to appraisal review and portfolio management.
The only problem with AVM’s is that they are programmed like a one way dead end street. They assumed that property values increase at a fixed rate.
That assumption worked out really well didn’t it?
There are over 40 different types of regression analysis and they all lend themselves to different predictive uses. Depending on the project, many types of regression will use multiple techniques to derive their solution but one in particular has a specific use. Hedonic Regression.
Hedonic regression breaks down a good or service into component values, which is precisely how an appraiser values real estate – individual component values are accounted for in a sales comparison approach to derive a final value estimate with a constant (the subject property ) and three or more variables (comps 1-3).
One can assume that if Fannie Mae developed a model to check comparable adjustments it would more than likely be a Hedonic model.
If that’s the case there are solutions for appraisers, and when properly applied one can also assume that the CU isn’t actually all that frightening.
In fact, if anyone should be worried, it may be the reviewer or underwriter being presented with an appraisal performed by a savvy appraiser using Hedonic Regression to substantiate their adjustments.
There is room for more discussion about what will happen in the industry once Collaterol Underwriter comes into play.
Whether or not it will help support or harm individual appraiser will soon be reveal on a case-by-case basis.
Which camp will you be in?