Collateral Underwriter- Too Much Too Soon
by David Braun, MAI, SRA

I would like to communicate some concerns and suggestions regarding the roll-out of the Collateral Underwriter (CU) scheduled forJanuary 26, 2015. This is not an attack on Fannie Mae or the CU. In fact, these suggestions support the potential of the CU in improving the loan process and the end result of having quality mortgages.

Let me introduce myself so you can understand my perspective. I have been a working appraiser for 39 years. I have been studying statistics for over eight years, specializing in regression analysis. I have no degrees or formal education in statistics. I call myself and people like me "valuation analysts," as I have one foot squarely placed in traditional valuation methods and the other in the statistical world. I have no axe to grind with Fannie Mae; I like many of the things it has done and dislike others. As an appraiser, I am highly attuned to its influence on how appraisers perform their work. I performed residential appraisals for the first 20 years of my career, and commercial appraisals the remainder of the time. In addition, I develop seminars and regression software for appraisers.

I have been using regression analysis in conjunction with appraisals and have run literally millions of scenarios of regression analysis in a laboratory setting to understand what it can and can't do for appraisers. When the 1004MC form was introduced, Fannie submitted an email to appraisers dictating how to count the listings for the month's housing supply (MHS) calculation. I published an article on the web indicating that the calculation method would often over-state the MHS by as much as 50 percent. I am not sure the folks at Fannie ever saw my article, but they did later retract the stipulated method, replacing it with a better one. The point being that as the old saying goes, two heads are better than one.

CU Upside
There are many potential upsides to the CU. Perhaps the most applauded by appraisers is that the CU is actively trying to identify the handful of unethical appraisers. Ethical appraisers would like to see the CU swat them like flies. I have personally seen appraisals where the comparable's data were fiction. Unfortunately, many of the appraisers who tend to inflate values are sometimes encouraged by a few underwriters. The CU is going to make this unethical group of underwriters uncomfortable, as there is now a check and balance system in place. You can see I support the use of the CU for a lot of reasons. Let's move on to some of the negative aspects, and what might be done to prevent them.

CU Downside
I have worked closely with underwriters over the years, and they do not pretend to be appraisers. They work with a simple checklist for review. I have personally been contacted from call centers in India who were the ones filling out the checklists. When the appraiser legitimately varies from secondary market guidelines there is a real problem explaining this to the holder of the checklist because they just don't understand the guidelines well enough. Now imagine these same underwriters and reviewers looking at the CU program on their monitor. I don't see how the CU is going to enable them to address the legitimate exceptions to the statistical analyses provided to them by the CU. The result will be an unacceptable addition of time and effort on the appraiser's part to explain the issues. Some appraisers may simply throw up their hands, and modify their appraisals to meet the suggestions of the CU. This path of least resistance is the last thing that should happen, not to mention it will tend to erode the appraiser's

independence.

There is a myth among appraisers and underwriters that when the CU uses holistic regression analysis methods, the outputs will be infallible. Fannie's training material clearly states that the CU analysis will produce some "false positives and false negatives." This means that the outputted indications of the CU will sometimes simply be wrong. Right or wrong, the appraiser can expect to get a call from the underwriter/reviewer asking them to explain why the CU is suggesting something different than he/she reported. In terms of line-item adjustments rates (coefficients), regression analysis is just a tool that provides evidence, not proof. This can be demonstrated by scrubbing the same data with a different method, omitting a couple of sales, or by omitting or adding variables. When these things are done the coefficient in question may change significantly. Also, it is not unusual for a 90 percent confidence interval for a coefficient for the gross living area to range from as much as $40 to $70. This wide range is further evidence that the coefficient is only statistical evidence, not statistical proof.

When there is a 90 percent chance that a coefficient will be within a specific range, this means that there is a 10 percent chance that the output is incorrect. From an appraiser's perspective receiving additional evidence for a line-item adjustment rate after the appraisal and report are completed is counterproductive. Consider a house painter who is told to paint a room a certain shade of brown. However, the mother-in-law would be visiting soon and if she didn't like the color he would be required to repaint the room a new color for free. The painter would say, if your mother-in-law wants to provide input about the color it must be done prior to my work, or I will have to charge for my additional work. If a client has any information for the appraiser, such as additional evidence of a line-item adjustment rate, they should provide it prior to the appraisal being performed, not afterward. If the appraiser performs adequate due diligence then he/she will have to charge extra to consider the additional information. I understand that Fannie Mae may have no intention for the underwriters/reviewers to require the appraisers to comment or provide additional support, however, if the CU provides it, the appraiser will almost certainly be asked to comment on it.

Jury of Your Peers
Many of the outputs of regression analysis are more reliable than the line-item adjustment rates. For example: identifying variables which have a relationship with value, identifying the random variance associated with a market, and final value estimates. These sorts of conclusions may be useful to underwriters.

The check that the CU will perform by comparing what the appraiser did to his peer group is problematic. This will penalize the very good appraisers who will be asked to explain why their conclusions are different from less experienced, inferiorly trained, or unethical appraisers. Considering a bell curve perspective, it is likely that you could divide appraisers into three groups, below average, average, and above average. This means that the majority of appraisers would be average or below. The implication is that the good appraisers should conform to the actions of the less credible appraisers. A reasonable answer from the good appraiser to the underwriter might be, "Yes, because I am a better appraiser than the overall pool of appraisers, I would expect my conclusions to often be inconsistent with the other appraisers." Again, I am not saying it is Fannie's intent that the underwriters will require an explanation from the appraiser, I am saying it will, none the less, be the reality appraisers are forced to deal with.