The Realities of Regression
Current appraisal spotlight on regression
I recently attended the 2015 Valuation Expo which was held in Baltimore MD. As usual, it featured some interesting and useful CE sessions. Over the next few months, I want to share highlights and insights from some of the presentations that I think were particularly relevant to issues appraisers are dealing with right now. The first presentation I want to discuss was called “The Realities of Regression” by Ernest (Ernie) Durbin II, SRA, AI-RRS.
The use of Fannie Mae’s Collateral Underwriter (CU) tool by lenders has created some “hot topics” for appraisers. Durbin is encouraged by the use of CU and thinks it will be a good thing for the appraisal profession. One of these topics is the discussion around how to clearly show the methodology the appraiser used to arrive at the adjustment values. In this presentation Durbin proposes the use of “Regression Analysis” (RA) as a solution to this problem.
Durbin lamented there is a perception that appraisers make subjective, anecdotal, arbitrary, and unscientific adjustments. Now that CU will be looking at adjustments relative to other appraisals, appraisers are going to need to take a more transparent and scientific approach to arriving at adjustments. Currently appraisers have different reactions when confronted with the prospect of having to use regression analysis. Some appraisers ignore it, some hate it, some will play around with it, and some embrace it as a natural advancement of the profession.
Even though some appraisers may ignore or even hate thinking about using regression, Durbin points out the commentary underneath USPAP Standards Rule 1-1 (a), states, “each appraiser must continuously improve his or her skills to remain proficient in real property appraisal.” He says that there are really many good reasons, besides CU, that appraisers should be using regression. These include transparency for anyone reading the report, easily replicated results, the advantage of large data sets to support conclusions, the fact that you will learn something about your market, and that it will eventually save you time.
Even with all these reasons to be using regression analysis in your appraisals, Durbin did pause to explain that there are some drawbacks to using it. First, and this can be important, it doesn’t always work. This can happen for a number of reasons, but it’s usually due to data availability issues and/or unique properties. He also stated that regression is not always needed. The appraiser may have instances where the market indications are evident without it or where minimal or no adjustments are required. Finally, it may be misleading. You as an appraiser still need to apply your professional expertise and determine if the results make sense. Look to see if the adjustments are headed in the right direction and if the adjustment is reasonable in your professional judgement.
Durbin explained that there are actually many different tools available to help the appraiser with regression analysis. There are general “off-the-shelf” software packages such as SAVVI, Statwing, Redstone and Regression+. Academic software such as Excel, SPSS, Minitab and “R” are also readily accessible. Now, with the current appraisal spotlight on regression, many of your favorite appraisal software forms vendors have come out with new programs that will help you through the process. It is also important to note that not all your adjustments have to come from the regression. Durbin said that it is common to use the regression results for some adjustments and base others on different methodologies.
Now you may ask “how does all this fall in line with USPAP?” Well, Durbin basically looks at regression analysis as a tool for the appraiser so it can be thought of in the same way as an AVM. He discussed Advisory Opinion 18 (lines 97-103) and listed the five questions to which the appraiser should answer “yes” before deciding to use an AVM or RA in an appraisal or appraisal review assignment. These are:
- Does the appraiser have a basic understanding of how the AVM (RA) works?
- Can the appraiser use the AVM (RA) properly?
- Are the AVM (RA) and the data it uses appropriate given the intended use of assignment results?
- Is the AVM (RA) output credible?
- Is the AVM (RA) output sufficiently reliable for use in the assignment?
Just to be clear, Durbin pointed out that to answer “yes” to all these questions doesn’t mean that you have to know exactly how the tool you are using is built or the formulas it utilizes. What you need to know is how to use it, and be able to determine if the results it is producing are credible.
Durbin also provided some reconciliation commentary when you have used regression analysis in your report. Commentary will change with the results, but just to give you an idea here is one of the samples he provided: “In determining adjustments for differences between the subject and the comparables, regression analysis was utilized. This regression analysis was used as a starting point for adjustment factors and amounts. The values of the coefficients were then modified using sensitivity analysis. Below are the results of the regression analysis.”
If you need help getting started there are lots of sources for help these days. Durbin recommended taking more continuing education courses and/or online courses in statistics. He also recommended some Appraisal Institute texts:
- “A Guide to Appraisal Valuation Modeling”
- “Practical Applications in Appraisal Valuation Modeling – Statistical Methods for Real Estate Practitioners”
The bottom line is that it has always been good practice for appraisers to show how they calculated their adjustments. Now Fannie Mae and CU are shining the spotlight right on this issue. Durbin believes the profession needs to become more scientific and accountable. His hope is to get appraisers out of the “ignore it” and “hate it” columns and get them to at least “play with it” if not “embrace” the technology.
I will cover other interesting sessions from 2015 Valuation Expo in future issues.
By Steve Costello, a Relationship Manager at FNC working directly with the appraisal community since 1986. When not in the office, Steve can be found riding off-road motorcycles and doing home improvement projects. Steve can be reached on LinkedIn or Twitter. ~ Source AppraisalPort
The problem with RA is junk in and junk out. In my opinion until real estate agents are held to the same standards as appraisers and are required to enter everything accurately and similarly to each other across the grid, it will be mostly junk in. If Fannie Mae decided to hand the accurate date over to appraisers for our own use (since it is all appraiser verified) that they are using, it would be much easier and much more reliable. I am only talking about the data, not the review program. Ask appraisers across the nation what some of their biggest hurdles are? Typically, after finding comparables, it is verifying comparable information. This task is more difficult for some then others, but it is often like pulling teeth. Often the information provided in MLS is inaccurate or lacking.
I can almost guarantee you with virtual certainty that the foreclosure rate will decline by maybe a microscopic percentage if every appraiser did RA. They are trying to turn an art into a science, and they can’t.
Eric THANK YOU! Can you say that a little louder for the egg heads at the top! The day a computer tries to appraise anything is the day that’s worse than the 1929 crash.
i agree with everything that has been said above.
like KK said, if we are expected to provide accurate appraisals, we need accurate data. we dont see every comp we use, therefore we have to rely on the information we get from the realtors, AND, (taking this a step further), we also have to rely on the information we get from the assessors offices. until the realtors and the assessment offices are held to the same high standards we are, we will continue to get mostly incomplete and unreliable information, and every appraisal will be continue to be questionable. we again are in a catch 22, because we cant get the accurate information we need, but we open ourselves up to liability because we cant get the information we need. nice.
i have an idea? instead of wasting our time tweaking USCRAP to another insane level again, how doing something really productive like writing and enforcing rules and regulations that would hold realtors and assessors accountable for being lazy and not doing their jobs?
Like Eric and the article already says, RA doesnt work because rarely do i have a property in my area that isnt unique. appraising isnt like torquing a lug nut. appraising will never be an exact science, and we need to stop trying to turn it into one.
the bleeding continues . . . . .
Good luck getting this right, when residential real estate purchases are typically driven by emotion (My wife really likes this home and I want to make her happy) vs Commercial real estate which is strictly numbers driven with emotion taken out of the equation. Have fun trying to quantify the human emotional element in residential appraising!
the regression zombies are back again after destroying the “line of credit” market valuations. regression works very well if your subject is in the middle of the bell curve, and surrounded by identical homes in an identical market. however, regression will outlive us all, it’s where they want to go. sorta like the land of oz. does anyone want their children to be in this “profession”.