New or Different Does Not Mean Bad

Whenever something new comes along there are a certain number of detractors.  It happened when the automobile was introduced, it happened when computers began to show up in homes around the world, and it happens whenever something different is introduced to the appraisal industry.

In the early 2000s, we were presented with Automated Valuation Modules (AVMs).  The cry of ‘foul’ went up collectively from the appraisal industry.  “Computers are replacing the appraiser!” they shouted.  Frankly, their complaints were justified in this case.  AVMs were premature and the technology was not sufficient to do the job they were tasked to do.  It was like taking a Model T to a monster truck rally.  They just were not built for the mission at hand.  Consequently, AVMs mostly died out… thankfully.

regressionappraisalRecently, we have been introduced to a slew of new appraisal tools to assist appraisers with valuations.  Though regression analysis has always been a part of appraising, there are new technologies which allow this type of research and examination to happen at a faster and more effective rate.  From stage left, enter the detractors.  I spend a fair amount of time on the internet forums dedicated to the appraisal industry and I am hearing a lot of screaming about these new products.  Once again, appraisers are feeling threatened and feel like their profession is being demised.  I believe they are using an old model on a new problem and their fears are unfounded.

Though regression tools (and the like) may remind us of the old AVMs, they are not the same animal and deserve different scrutiny.  Regression analysis is not a replacement for appraisals.  Rather, it is a tool to assist with better appraisals.  Where AMVs sought to replace the appraiser by shortcutting the human aspect of valuation, regression is designed to be used by human beings (appraisers) to develop better, more statistically sound appraisals.  Certainly, it depends on the market and the data available, but statistical analysis on a mass scale can shed a new light on the market and the specifics of your subject.

Lenders and users of appraisal services would be foolish to ever try to cut the appraiser out of the valuation process.  As we all know, the fourth approach to value (sitting on the curb and saying, “What the hell would I pay for this dump?”) is sometimes the best approach.  Computers have no instinct and sometimes instinct is needed to come to an accurate value of a property’s true worth.

On the other hand, it is just as foolish for appraisers to disregard these amazing tools to assist our valuation process.  Statistical regression can open windows of knowledge we could not have otherwise.  In a way, it is a fifth approach to value that can be priceless to the final determination of an accurate and well-supported appraisal.

Now, go create some value!

14 thoughts on “New or Different Does Not Mean Bad”

  1. I don’t disagree. In fact, since 2008 I’ve used various forms of regression analysis to formulate an opinion of market trends – after being exposed to this in training presented by David Braun and Patrick Egger. A graph demonstrating that is in every report I do. But I don’t use regression to determine individual adjustment amounts.

    The issue for most appraisers, IMHO, is regression ‘doesn’t work’ with the THREE sales on form page 1, and the TWO Pendings or Listings on the comps 4-6 page (if included). It also does not work for the few sales on the poorly designed and statistically incorrect MC Form. And they are exactly correct.

    For regression to work properly, two dozen or more data points are necessary, meaning that many comparable sold or listed properties must be found. In many places around the nation, those numbers are far less. My reports have scatter graph analysis of comparable property sales over a 5 year period in the selected neighborhood. With that I can see exactly what is happening for a specific type of property, for average price trends, and actual price trends tied to individual sales.

    Also, for regression to work properly, the correct data must be used to build the chart from which the regression is derived. My sense is that most appraisers are not well trained in statistical analysis, and thus just use ‘any sales’ in the particular area to determine trends, when in fact, the correct properties are those that are appropriately comparable to the appraised property.

    When I review residential reports, seldom are any charts or graphs included to give readers a “visualization” of market activity or other details. Why not? Well, because it takes more time to do those and appraisers are reluctant to spend the extra time to use a tool that makes their work product that much more credible and easier to understand. 100% text based reports are ‘good enough’ in their minds.

    Reports with graphs are “better” than average reports. Appraisers should learn to create and use them. It’s not difficult at all.

    1. Hi Dave. Several thoughts occurred while reading your comments:
      1. A dozen data points? MY regression software suggests anything less than 200 samples (which may include FAR more ‘data points’) is going to produce questionable results.
      2. It (CANVAS) also suggests bracketing of comparables is essential in order to determine values of the various components. I don’t know. Im not an expert at regression analysis.
      3. An online source called “CourseRA” offers free educational offerings from places like Princeton and other major recognized universities, including statistical analysis.
      4. If you are NOT using regression to determine adjustment amounts, then what is the point? If it is only to develop a trend line for market increase or decrease, I can use DQNews results which include ALL transactions n the identified areas.
      5. I absolutely agree 1004MC is useless. Horribly designed by people that have no understanding of real estate markets.
      6. The ONLY reason for using RA in transactional SFR appraisals would be to support specific adjustments (IMHO) being made in the sales comparison. ANY other reason is pure puffery. Our state (CA) wants us to support individual adjustments and ‘suggests’ one way to do this is via RA. ANOTHER way is to ASK AGENTS and principles involved in specific transactions. So far no one has demonstrated that the former is more accurate than the latter in broad market wide application.

      I think RA holds lots of promise, but we need to understand and accept its limitations.

      I’d like to see RA software that is compatible with mls data for an entire city or county, and VERY user friendly. THEN I think we’d have something.

    1. Gary Bratton, ARA

      Appraisal Institute had a Real Estate Finance, Statistics and Valuation Modeling class that was good at introducing the concepts and some tools. It was available online through their website

    2. Dustin, I actually WAs responding to Dave’s post. Sorry for any confusion.

      I agree with your conclusions. Fourth approach IS as valid as any other; BUT within the limitations of design, we SHOULD consider other methods to augment our analyses.

      When I took my first Institute course back in ’86 (AIREA back then), the instructors made it a point to remind us all that we had one of the most efficient, information gathering and analyzing computes ever built right between our ears. It doesn’t always tell us the ‘how’ of its conclusions, but it does consider a lifetime of data input in arriving at those conclusions.

      I try to use ALL tools available to me, but I refuse to ignore the ‘best’ one.

  2. Yea…I aint with either of you on this one. You are exactly right, appraisers are reluctant to do more work. Are they just lazy? Don’t think so. In fact we are the opposite of lazy considering the amount of work we already do for the amount of pay we receive. When the computer and automobile came out, consumers were willing to pay a lot of money for the product, but they took it home after they paid their money. Can you imagine a loan consumer making the decision if they would like to “gamble” $1000 on an appraisal that may or may not work out for them? My guess would be $500 is about the high-end of the gamble threshold, less for many. The power of that number, the ultimate price a consumer is willing to pay for a mortgage appraisal, can not be underestimated. If you don’t understand what I just said you should keep reading and thinking about it until you do. The 1004 form is more exhaustive than it need be already and was designed with that consumer number in mind, from a few decades ago that is. Why screw around toying with adding more work for a “better” appraisal. Haven’t the underwriters added too much already? When have consumers been willing to pay more to get more? Why are appraisers willing to give the bar away for free? Give me a break. I can give you the most accurate appraisal any appraiser could give in a about 3 minutes of research on 99% of all property. It aint hard, so please stop acting like it is and that if we all just tried a bit harder things would be better for everyone. We get paid for our integrity and independence, not because valuing property is difficult or something our clients could not have figured out for themselves with just as much accuracy. Regression analysis….goodness…LOL….how about I do a regression on your face? I’m kidding of course, but I hope that punctuates the point.

    1. “M” I’m with you on a lot of what you say. What WE have to understand is that the people that buy and sell real estate loans do not know their ass from a hole in the wall when it comes to appraising. When bad decisions are made; or when end of year P&L reports are run, some genius sitting in a chair wondering how he or she can increase their annual bonus says “Obviously the problem is inadequate market trend data. If we only had that info, everything would be fine.”

      The year before, that same person said “Obviously, detrimental neighborhood conditions are the problem. If we only had THAT information, we’d make better loan decisions.”

      Then they said “If we only had TWO lines above the sales grid telling us the number of comparable sales and listings with their ranges, THAT will solve our problems.”

      Currently, we have a person or persons with their heads so far up where the sun never shown, that said “Obviously the rating system used by appraisers and understood by everyone except us is the problem. We need to create a uniform rating or measuring system that has its own two pages of explanations right inside the appraisal. THAT will fix all our problems” Especially if it eliminates the need for all common sense.

      Lastly we also have others with their hands aching to get back into the cash tills, that say: “Obviously the problem is those damned appraisals, with all that negative market trend data; negative neighborhood descriptions and easy to understand comparison grids. What we really need is a system that allows us to transmit ONLY THE PORTIONS OF THE APPRAISAL that our clients want their investors to see.” Thus was born XML appraisal reporting.

      PS to Coach That’s something “new” that I’m not fond of either.

  3. I find that all of these regression models use false assumptions up front to simply allocate a percentage of value to the different slots in the sales comparison grid. For instance, 5000 sq ft size difference in site does not always demand an adjustment so you have to “fool” the software into giving it a true market adjustment or no adjustment at all, by telling it the sites are of equal size. Also the percentage and adjustments for GLA are double or worse what is actually proven in the market. Then the remainder of any adjusted value differences between the comps is simply spread through the remaining fields and “normalized” to try to get the smallest deviation in the adjusted values. And they do all this with only the comps that you use in the grid. I’m sure this fools the underwriters and many appraisers into believing that these are true market adjustments, but to be truthful not every buyer is a robot with exactly the same tastes and financial strengths and not every Ferangi (Realtor Sales Agent) is as skilled at milking the uninformed buyers into paying far too much for any given property. If these regression tools could use a true market segment, instead of just 6 comps, and you could tell it what to allocate for any given area of comparison, then you may have a neat little tool to plot a chart of what you already knew. Dog and Pony Show coming to an appraisal near you! OOOH the glitter!

    1. Billyon – excellent post!

      CANVAS offers a software that seems to consider the broader market data. Not sure how reliable the results are. It says you need 200 samples. Most my mls will give me is 100. It ALSO says one MUST include dissimilar (better and inferior) properties or there is no way to isolate the value differences. Makes sense to me (Bracketing). I can use depreciated replacement costs to develop supported GLA adjustments that are FAR more accurate. I can use historic land to improvement ratios to develop overall prices per SF of site area, though that does not mean the market recognizes excess or inadequacy at the same rate.

      I can ASK Realtors and agents what THEIR experiences are with respect to two or three bedroom prices and either break that down to values per SF or rooms as the case may be. I am NOT about to give up tried and true methods of determining market adjustments just because some paper pusher or bean counter somewhere says “You must prove your adjustments using proven scientific methods that are duplicable”. THOESE people do not understand what an appraisal is and is not.

    2. LOL!!! I know what a Ferangi is and that might be the funniest comparison I ever heard. So what race are the appraisers? Certainly not the Captain Kirks of the world.

  4. Statistical analysis may have some role in adjustments, but knowledge of a particular neighborhood and market is usually a better indicator. Buyers don’t use regression analysis to decide how much they want to pay for a larger back yard.

Leave a Comment

Your email address will not be published. Required fields are marked *

Related Posts

Scroll to Top

Existing Members

If you have been a member prior to Jan. 1st 2024

Or, click on the right side to sign up as a new member (with a free month and added bonus material) and your existing membership will be automatically moved over and any extra payments credited. 

Or, click on the link below to sign up as a new member (with a free month and added bonus material) and your existing membership will be automatically moved over and any extra payments credited. 

New Members

If you became a member after Jan. 1st 2024 or are an existing member and want to move to our new system. 

Try the All-Star Team No-Risk for 30 Days Free!