Is $25 per Square Foot Accurate?

What is the proper adjustment for differences in square footage on a 2,500 square foot home as compared to a 2,800 square foot abode? $25 per square foot? $30 a foot? $40? If you are the type of appraiser who makes an educated guess, throws it in the report, and then changes it if it does not work quite right, your grace period has ended.

AppraiserAdjustmentsBeginning in January 2015, Fannie Mae will be paying much closer attention to the UAD dataset and the adjustments that appraisers make in their reports. What is Miss Fannie looking for? Support for adjustments. What is the best (but not only) way of supporting your adjustments? Regression.  

I have written about the importance of regression in the past. The purpose of this post is not to tell you why regression analysis is essential or which tool I like best (currently, I am using my own Excel spreadsheet). Rather, my purpose is to tell you that this is coming. Fannie will be looking closer at your reports. Warning letters will be going out and blacklisting is a possibility. Is it fair? Again, a discussion for another day, but it is the new reality. If you use $25 a square foot, you had better have a reason for doing so.  

Happy Holidays!

Dustin Harris, Creating ‘Value’ for Real Estate Appraisers  

189 Comments on “Is $25 per Square Foot Accurate?”

  1. Pingback: Is $25 per Square Foot Accurate? | Appraisal Buzz

    1. Dustin: thanks do much for your very informative articles I always look forward to them. I was wondering if you could give me done info as to where to take regression analysis classes and forward me a regression analysis pages to me as well. Thank you

      1. I like everyone else would like one of the regression analysis pages; also which is the best class to take on this since it has been a long time.

      2. Hi Robin
        The Appraisal Institute have several classes on regression. To get started , I recommend going to YouTube and find Brandon Foltz. He has 4 videos on Linear Regression. Each one is about 30 minutes. This will help you understand the basic concepts.
        I have been using Linear Regression in my appraisals for about three years.

        Good luck, Roy

      3. I need your info as well!
        BTW, what is the MINIMUM sales population you would consider to be significant?

    2. Good Morning Dustin

      If you are sharing, your regression analysis page, would you kindly send me one. I would like to take a class but cannot find one in this area of Western MA.

    3. I also would like to have a sample of your excel spread sheet. I have use one not but am not real sure how accurate it is.

      Thanks

    4. Hi Dustin, I would really appreciate a copy of your regression analysis sheet. God love you thanks for your help.

    5. I, too, would appreciate a copy. I suppose with all these responses, you will be posting it to your blog. Sounds like there’s a lot of money to be made giving classes around the country!

    6. I too would like a copy of your regression analysis. There are so many variables in every house, exact “matched-pairs” is very difficult to find. Even new homes of the same floor plans will have different interior finishes and/or amenities.

      Merry Christmas and Happy New Year!

    7. Add me to the list of interested appraisers! Maybe you should host a paid webinar on regression and include your spreadsheet as an added plus! Appreciate you and all your input, as I have for several years now!

    8. What type of regression analysis are you promoting? Linear or multiple? If one, why not the other? While the point of the article is EXCELLENT, the practical application discussion is extremely weak. To the credit of your followers, the comments show a hunger from them for more information. Are you going to provide it?

    9. I would also like to request a copy or information regarding your regression analysis spreadsheet. I have not been able to find a suitable class about this topic. thank you

    10. Regression analysis is accurate if you have enough data to properly utilize that method. This adjustment has been discussed and cussed for years only to get short sheeted. What does the market/builders/developers say that it costs without land, appliances, fireplaces, built-ins, etc. Smaller markets must use all methods that provide consistency and clarity to the valuation.

    11. Hi Dustin,

      Appreciate all of your comments and information. Could you forward an example of your regression analysis pages to me? Also what classes do you recommend on this topic. Thank you, and have a safe and happy holiday season!

    12. Great points! And at the rate things are going, it will soon take 24 hours to do one appraisal and between what the AMC’s are paying and everybody looking over your shoulder to criticize every aspect of the report, it won’t be worth being in this business. At least, not for those who pass the bureaucrat’s insulting criminal background checks.

    13. Hey Dustin,
      This is what I use in my reports.

      “Living area adjustment valued at $X per sq. ft. after 100 sq. ft.(rounded) difference versus subject based upon 35% of the average price per sq. ft. of living area of the included comparables. This technique for determining GLA adjustments is described in the text Residential Sales Comparison Approach, by Mark Ratterman, MAI SRA, published by The Appraisal Institute, and suggests averaging the sales prices per square foot of GLA. Then multiply that average by 30%-40%(to account for features included in the sales price that are not included in GLA, such as garages, carports, basements, etc.) to determine an adjustment value.”

      Your thoughts?

      Also, may you and your family have a blessed Christmas.

      Brian- from the Salt Lake class last year.

      1. Brian, how long ago did he advocate this approach? My guess it is more than a decade and that he would not endorse his own position with the advances in computing power and software.

    14. “Happy Holidays” huh? Dustin…..thank you for the encouragement before Christmas. As if there isn’t enough bad news for appraisers floating around. I will continue to educate myself on these changes and choose not to listen to the Debbie Downers of the appraisal industry.

    15. If you are willing to share I would also like a copy. I have attempted this on my own without success.

    16. Dustin, please add me to the list of those seeking a copy of your excel regression template. Thanks and Merry Christmas, General Bob

    17. I do review work and it’s amazing how many reports use $10.00 per square foot for the GLA adjustment. $10.00 was used thirty years ago as an example in the education classes and that figure stuck with the young appraisers and is still used today. This won’t fly with FANNIE MAE

    18. I would like the regression analysis excel page; also which is the best current class to take on regression analysis? The true disappointment I have is the limited education and the discipline being promoted and patronized more than the education with respect what will meet and exceed FNMA requirements and expectations.

    19. I would also like a copy of your regression sheet. I should say, I would like to purchase a copy of your regression sheet, as I am sure you have put time into developing the strategies you utilized.

    20. I would appreciate a copy of your spreadsheet and an education on Regression. I am a member of The Appraisal Coach

      Thanks
      Terry

  2. Hi Dustin, I took one of your classes in LV. Could you e-mail me one of your regression analysis pages so I can see how to set this up?

    1. Dustin, I too have taken a Regression Analysis course but have had difficulty in establishing an effective and credible data base which can be readily and easily called upon for use in a timely manor as needed for most appraisal assignments. Would you be so kind as to provide or direct me to the proper information sources to set up this system for making comp adjustments. Thank you.

  3. Hi Dustin. I also am hoping you can forward a sample of a regression analysis chart as it is apparent that our reports are going to get even tougher to defend. Thank you.

  4. I’d also like to receive your Excel regression analysis spreadsheet, if you’re sharing it.

    Thanks

  5. Good morning Dustin, I would also appreciate a copy or sample of your regression analysis, if possible.
    Thank you.

  6. May I have a sample of your Excel regression analysis spreadsheet also? Thanks and have a great Christmas and Happy New Year!

  7. If you could email me a sample of the regression analysis and/or a different example of determining square footage adjustments, I would greatly appreciate it. Thanks so much.

  8. Would you mind forwarding me a copy of the regression analysis chart. I would really appreciate it. Thanks.

  9. If you could just post a tutorial on your regression analysis tools, we’d all appreciate it…

  10. Dustin,

    I like all the others would like a copy of your excel worksheet or guidance on finding software. I have been trying to find regression software. I have spoken with several software companies but live in Missouri which is a non-disclosure state and our MLS doesn’t provide GLA data, etc for most of the counties I serve. I am at a loss. Thanks

  11. PLEASE ADD ME TO YOUR REQUEST FOR SPREADSHEET AND/OR GUIDELINES OF DOING MY OWN.THANKS FOR YOUR INFORMATIVE ARTICLES.

  12. I have been hearing this for months and sadly there is little training in the NE US area for regression analysis. I have looked for classes and taken one online webinar that was a recorded one and followed their method and as usual the data that they used comes out to a predetermined number that gave a 90-95% validity rate. When I used it in my area of Upstate NY the data comes out to 50% -75% at best. Perhaps more training to figure out were it is incorrect I do not know. What I do see is a lack of education on this for a lot of areas. I would like to use it and have used the Savvi regression analysis but that also is askew due to a $45,000 having 77% gross adjustments including a .5 car garage adjustment that was $17,000.

    I would like to see your spread sheet or if you offer some training on it Please let me know as I would sign up in a heart beat.

    Scott

  13. I’m jumping on the bandwagon! Would you mind forwarding me a copy of your regression analysis Excel chart. I would appreciate it. Thanks! And Happy Holidays!

  14. Like everyone else, I would also very much appreciate a copy of your regression spreadsheet. I have tried others without much success. Thanks for all the informative articles & Merry Christmas!

  15. I would like a copy of your regression analysis (excell spreadsheet) and any instructions you can give on how to set it up. Thank you and Merry Christmas!!

  16. I too would like a copy of your regression analysis spreadsheet. I have not found any classes in my area.

  17. Dustin,
    I too would like some information on the regression analysis. I purchased the Appraisal Tool and can not figure it out….I live in a very rural market and it is very difficult…Any help would be much appreciated….Thanks

  18. I personally do not think a regression analysis is necessary for residential properties. A matched pair analysis is the best way to determine adjustments. The adjustments used may change from one appraisal to the next. There are a lot of variables involved. I work out of Allegany County, Md where in Cumberland, a 2800 sf house may only be worth $60,000. We use $10/sf there. Other areas of this county, the value could be $400,000 which would mean a higher $/sf adjustment.

    1. There seems to be some confusion regarding accuracy vs. precision; $25 / sf is perfectly accurate with reasonable support (matched pairs are sufficient in nearly all cases, including most commercial reports), though it may not be as precise as you wish. On the other hand, single-variate regression analysis will give you an arbitrarily precise number which will often overstate the quality of the data and simply mislead the user of the report as to the precision of the estimate. It may be precise to say that the adjustment for the marginal value of additional living area is $23.67 per square foot, but if you have three data points with significant other differences (the usual – age, condition, quality, lot size), but it is unlikely to be accurate and, more importantly, is not a credible result. There is a reason that multi-variate regression analysis is used in cases of multiple significant variables where such precision is meaningful, but such advanced methods lie far beyond the scope of work for most appraisals and exceed the competence of typical appraisers, unless they have extensive training in multi-variate statistics.

      1. Thank you Mr Lutz. I had to scroll down numerous posts before a single person had the courage to state what is obvious to anyone who can actually analyze data and call it for what it is in terms of reliability. Bravo. I think it is unethical, per USPAP, for appraisers to jump on this wagon and simply comply to a method that is not credible as so many here appear to be doing. That said, once everyone does it, it will then be required by USPAP for all of us to do it, whether it is credible or not, as then it will be a recognized method. A horrible method of appraising homes (making and then placing weight on adjustments) is now about to get one step worse. While I recognize that a client has the power to add assignment conditions for their purposes, its way beyond time that someone in the industry take a stand and tell them what BS their conditions are in terms of adding credibility.

      2. Excellent point Mr. Geoffrey Lutz,
        Matched pairs and my all time favorite fail safe use cost approach local data cost sheets and of course Marshall and Swift. I believe two things. One, appraiser should be competent and accurate as much as technology and education allows and is provided. Two, this is another way to push appraiser out of the industry. FYI only the strong survive, aka “those that keep up.”

        “There is a reason that multi-variate regression analysis is used in cases of multiple significant variables where such precision is meaningful, but such advanced methods lie far beyond the scope of work for most appraisals and exceed the competence of typical appraisers, unless they have extensive training in multi-variate statistics.” Point taken and well said Mr. Lutz

        Personally I am taking a back seat working my other profession to see how this ever ongoing changing industry and new provision plays out. I will keep my eye on what appraisers are reporting their experiences aka (trails and tribulations).

    2. Seriously? Paired sales analysis belongs right next to the microfiche reader printer in one’s office. It is a dated virtually useless technique that was born of the days before computers and the ability to crunch serious numbers. Its results can easily be skewed, twisted and warped to fit any scenario. While that is still possible with regression analyses (figures lie, liars figure), for the honest professional regression analyses provides a far superior, more accurate and reliable tool for measuring values within a market.

      1. Michael Mathis,
        As you stated “Its results can easily be skewed, twisted and warped to fit any scenario. While that is still possible with regression analyses (figures lie, liars figure),” All numbers can be FUDGED. Math and numbers are not always fact. They are merely educated guesses that look pretty Superior to the average laymen. This is simply going to be about who and when they want to pick on someone for good reason or no good reason. PERIOD

        1. Wrong Anna. I would love to face any appraiser in court who chooses to use paired sales as his/her basis for an adjustment, when I have a 200-300 data point regression analysis as mine. In fact, I have already done so in depositions and guess what, the other side offered to settle knowing they were going to be decimated if they went to trial.

          1. Mike I would love to face you face to face in court with your computer print outs. I would break down your 200-300 data points and show the garbage included in this lump of data. Unless you are able to find 200-300 data points of comparable data within the same and similar neighborhoods and exclude neighborhoods and properties not similar to the subject along with eliminating non arms length transactions you would not have me worried. I would show the weakness in your assumptions. (Unless of course you had quality data and hand picked your 200-300 data points, which could be found in very few places in this country). The term garbage in garbage out, example Zillow. Regression Analysis has its applications but appraisals in areas with limited comparable data is not one of them. This is the lazy appraisers approach to valuation. This is similar to the real estate agents approach to valuation with the cost per square foot. Unlike many on this board asking for excel sheets of a process they know very little about I have knowledge in this area. Now based on your prior responses to others on this board I expect you to puff out your chest and let it fly.

  19. Hi Dustin, if you are sharing it, I would like a copy also. Thanks, I enjoy reading your articles, keep it up!

  20. Good stuff Dustin. If you decide to provide what these other folks are requesting then please add me to the list of requesters. (Regression spread sheet, instructions for use, etc…)
    Thank you

  21. Dustin,

    Like everyone else I would appreciate a copy of your regression analysis spread sheet.

    Thanks,

    Bill Ludenia, IFA
    Bill Ludenia Appraisals, Inc.

  22. I have to jump on the band wagon and ask if you would please share your regression analysis spreadsheet with me. Thank you so much.

  23. I think what gets missed in this discussion is that the marginal value of an additional square foot of gross living area is highly dependent on the elements of the particular comparable relative to the subject. For instance, the per unit marginal value of an additional 100sf of GLA between homes of the same room count should be relatively low. However, the same subject property when compared to a property that is 500sf larger with an additional living area will typically exhibit a much higher per unit marginal value for the additional space. It is the difference in the utility and market perception of the value of the additional space. An analysis of how appropriate the GLA consideration is for a particular subject property is highly dependent on the comparable selection and how the appraiser addresses differences in room count, design, and functional utility. While regression analysis does have some usefulness, it should only be considered a supplement to the actual paired sales analysis, which is what most accurately reflects the market reaction to the specific difference considered.

    1. Brent- good viewpoint and comments (imho). Personally I do not want to run EITHER paired sales or regression on every single SFR I do; and then have to document that for FNMA. It would be foolish of them to expect this also. If so, it will turn into the same auto populated software “analysis” that currently passes for a 104MC market analysis. Data will magically appear, but it wont actually be analyzed for accuracy or relevance by the appraisers.

      I’ll continue to run it for competitive market areas quarterly; and also continue to use alternate methods that more quickly approximate those same results.

  24. If you’re sharing, I’d sure like to see how you’ve set up your Excel template. If you could, please forward me a copy of your RA spreadsheet. Thank you!!

  25. I not only would like to view your spread sheet should you be sharing it, but would also like to take a class on regression analysis?????

  26. Please forward a copy of the regression analysis spreadsheet. Also interested in a class that would cover this. Thanks.

  27. I would like to view/copy of your spread sheet as well. If you have any suggestions as to what website to look on for additional information on regression analysis would be greatly appreciated. It appears we could use ALL the help we can get if we plan to stay in this business. Thanks!

  28. Dustin,

    I’m a new appraiser and have been in the field a year. I use regression at times but find that it is so unreliable when basing my adjustment solely on that. I maintain the analysis in my work file but there are so many other factors that play into deviations in a closing price / GLA regression chart. If there were no other factors, I would assume you are appraising in an area with track homes with similar condition, quality, site contribution, bed/bath count, etc, etc. If no other adjustments are necessary, I assume regression would be perfect. Regression has been a tool for me in supporting adjustments but has not be the only tool (in my limited experience). I am open to comments regarding this as there is very little help in this industry. Thank you for your blog!

    PS – I have not seen your spread sheet but am open to how you have it set up. Thanks!

    1. First, no measurement tool will yield perfect results every time. As appraisers we are measuring the actions of often irrational buyers/sellers and those same consumers’ unique and often bizarre decisions are going to be reflected in the data. The presence of irrationality is one of the chief reasons we should use regression analysis and not paired sales analysis. Paired sales has only two data points. What happens if one of those points was created by the actions of an irrational or uniquely motivated buyer or seller? You guessed it, with 50% of your data being bad your results are bound to be poor, but I digress. It is enough to know that all data has the potential to yield questionable results, which is the reason to scrutinize results every time. It is possible that the failure of the analysis is not controlling for significant differences. One can still use simple linear regression analysis to solve square footage adjustments, if an effort is made to filter out of the sources of confusion by removing properties to create a more homogeneous sample, e.g. create a sample only with or without pools , filter out all properties with extreme age or lot size differences. This is essentially your basis for stating you are more comfortable using tract data, which is by its nature more homogeneous. The fact is you can massage your data to create greater homogeneity. The more homogeneous the sample except for the item being measured, in this instance square footage value differences, the better chance that the results will be reliable. In short, the failure to get your results is not an excuse to abandon regression tools, but rather it is the reason to seek better methods for their application.

      1. I want to say I agree with what you said here 100%. Though I would bring up a couple of points. First, while you are 100% correct that “massaging” data in order to isolate features will result in a “better” result, the question for me is what sample size is left over to analyze? Very few properties are exactly similar. In fact most, if looking at the 26 different “adjustable” features on the 1004 form, are very different. By the time you widdle down the data to find properties exactly similar, except for the item you wish to extract an adjustment for, there are not enough properties left over to analyze, unless you go out of neighborhood. However once you do that, you break the golden rule of real estate, location, location, location. Only in large tracts of very homogenous homes can this type of analysis have meaning. Case in point is the 1004MC form. Different analysis, same result. Does it work sometimes? Yes. Does it work most of the time? Not really. When you analyze the market in terms of a subject-specific sub-market, and lets not lose track of the fact that we are hired to analyze just that in an appraisal, the end result is too often inconclusive or meaningless do to a lack of available data. Keep in mind that the 1004MC tracks only a few points of data, not the 26 that are tracked in the sales grid. In fact, the whole purpose of the MC is to support the Neighborhood Section which in turn supports the time adjustment made on the grid – we are talking about 1 adjustment out of 26. Please explain to me how the mathematical equation is going to meaningful for 26 adjustments, many of which are interconnected to each other, when the equation becomes that complex? Which brings me to point #2 and just plain common sense, buyers don’t purchase property like that. Isn’t the Sales Grid supposed to be a reflection of buyer reaction? In my opinion the problem is that lenders and subsequently appraisers believe that because the FNMA forms have spaces for potential adjustments on the Sales Grid, that adjustments are therefore appropriate.

      2. Michael,
        Thanks for your response and I agree with it. Neither paired sales nor regression should not be used blindly. I agree that regression should not be abandoned but used as an arrow in your quiver of arrows to arrive at the target. When I pull the data for regression, I use the same data pulled for the 1004mc. (Is this what you do or is it narrowed further?) The mc data is already narrowed and filtered by the characteristics of the subject; but not filtered so far down as to have only a few points of data – then the mc may be unreliable (that might be another topic). But I also don’t want to have only a few points of data for regression analysis, then your example of irrational buyers/sellers comes into play as well. When regression is used in conjunction with paired data analysis, it begins to tell the story. Ultimately, I have found that the more information to back up an adjustment the better. However, it is still our value opinion…right? We use our experience backed up by data to come to a conclusion. And anyone could argue against it with their experience and other data. 🙂 Again, I enjoyed your responses and am open to more.

    2. Dustin, I can understand your reluctance to share your spreadsheet, if for any reason, potential lawsuit exposure. But you can at least clarify what type of “regression” analysis you are advocating. Are you using a linear regression analysis? If so how, if at all, are you controlling for significant variations within your sample (pools, over sized lots, wide age variations)? Or are you using some form of multiple regression analysis? Do you cull that sample data in anyway prior to running your analysis? You have raised an important issue, your followers to their credit are interested, please step up and provide more coaching.

  29. Would appreciate an opportunity to see a copy of your excel spreadsheet and any instructions for proper set up and use. Thank you for the heads up.

  30. Please forward me your spread sheet also. I appreciate the articles and help. Have you looked into Compcruncher or Redstone regressive analysis systems?

  31. Further discussion on regression would be appreciated. The spread sheet would help. Please forward. Thank You, Tim Black

  32. I to would like a copy. I deal mostly with rural areas and the information is very limited.

  33. It appears that everyone and their grandma would like a copy of my Excel regression sheet. I am flattered, but will not be sharing for several reasons. First (and most important), it is a bastardized tweaking of another, copywrited spreadsheet and I do not have permission to share. Second, it only works with my MLS system/area. Third, there are MUCH better worksheets and programs out there for regression. Finally, I would rather give you the names of others that I know such as:
    Compcruncher
    Regression Tool (on a la mode’s store website)
    SAVVI
    Redstone
    http://www.avttools.com/
    http://www.compflo.com/
    Appraisal Tool

    If others would like to self-promote those you have to offer or give us some feedback on those you have used, I would love to crowdsource here and come up with what everyone feels is the best tool available right now. Please feel free to share.

  34. Good stuff Dustin. If you decide to provide what these other folks are requesting then please add me to the list of requesters. (Regression spread sheet, instructions for use, etc…)

  35. I am going to say what I have already said on previous threads and have touched a bit on here. Dollar adjustments to comparable sales have always been and will continue to be unreliable. A ramped up method of deriving them from the market data, such as a regression analysis, will/does not make them more reliable. Why? Because the method of deriving adjustments is flawed from the get-go; people simply don’t buy property like that. For instance, a buyer would rarely say, “that house is great, but I am going to pay $25 a square foot less for it than the property next door that is larger”. LOL. How about we take a survey of realtors and ask them the last time they had a buyer who behaved like that. The 1004 form was designed as if buyers typically pay more or less for homes based on a host of individual features and appraisers have simply followed suit with the form like blind sheep. This started as an assignment condition, then morphed into a recognized method after the majority of appraisers followed suit. The regression analysis will be no different. The best thing that could happen is that appraisers will stand up and admit they have been adjusting comparables on the sales grid simply because the form asks for it, the “support” we have for these adjustments is marginally reliable at best and that we don’t really weight the final value on adjustments. Will that happen, I doubt it.

    1. By your reasoning Candlestick analysis, often used in stock market predictions, cannot work and is useless, as it does not take into account earnings, market conditions, industry news, etc., since this method only measures the REACTIONS of buyers/sellers observed in stock movements to the information they have available. The adjustments appraisers make are no different. We measure REACTIONS. The more closely we hone our tools to measure the valuation differences created by an item or amenity, the more closely we measure the REACTION of buyers/sellers to these units of value. We do not, nor should we, attempt to duplicate their thought processes, rational or irrational as it may be, rather we should seek to extract the numerical breadcrumbs behind them. Regression analyses does this better than most tools and should be used, but as with all other tools with intelligent application.

      1. LOL. I think the intelligent application is to throw the method in the garbage, along with all quantitative adjustments in appraisals that can not be directly supported (which means almost everything except reported concessions). I am well aware that my argument will fall on mostly deaf ears. I am well aware that intelligent people will continue to think this course makes sense and follow it through to the end. The mob rules and always has and always will, don’t mean they’re rational, but they will rule. I think its very hard for people to admit they are and have been wrong. I just think in a business where we get paid for our honesty, more of us should be honest about what a crock quantitative adjustments are. Ill end with this. Not only is there not enough available recent data to make a regression analysis meaningful, there are too many variables within the equation. I think the Powerball has less variables. Appraisers should be rallying to eliminate adjustments in appraisals, not searching for a magic mathematical equation to justify them. How about we have some rocket scientists play appraiser for a day and see what they think. Id bet my livelihood their conclusion would be that the results are inconclusive, or at least unreliable, due to insufficient, conflicting and irrational data.

        1. “M” has finally expressed the real problem we appraiser’s face every time we attempt to “fill in the blanks” on the 1004 form; the decision to purchase a HOME, by the typical buyer, is not based upon the square foot costs, or the value of individual features. The purchase of one’s HOME is an emotional experience based upon OBJECTIVE features including, location, bedroom count, bathroom count, GLA and amenities AND, SUBJECTIVE features such as, style, paint color, room layout, furniture placement (why do you think “staged” homes sell quicker than vacant homes?), “feel” (try to quantify how a house feels to one buyer vs. another), etc. As a former real estate broker, I learned very early in my short career to not, not show a colonial to a buyer desiring a cape because if the colonial had the right “feel”, that buyer might just want it. No regression analysis or matched pair analysis can account for or quantify the subjective features present in one home vs. another. Yet we are required to estimate an opinion of market value based solely upon OBJECTIVE data only. How many of you have purchased a house? In the process of comparing two or more homes for potential purchase, did you adjust the asking prices accordingly for the differences in GLA, bathrooms, garage bays, fireplaces, porches, decks, patios, etc., and then make an offer based upon that resulting adjustment? Probably not!

          1. The sharp appraiser won’t compare colonial comps to cape comps. Again a quantitative solution is present, just takes a little thinking. Many of the adjustment issues tossed out by M can be addressed by regression, though, with a diminishing return as the items become less quantifiable and more qualitative. The inability to perfect quantification does not excuse ignoring its uses, anymore than the inability to get back to my college shape and weight justifies me to ignore the caloric count and subsequent impact on my waist of my daily food. You apply the math as often as it is reasonable and possible. The fact is that it may be applied far more often than most appraisers currently recognize. Next point: The only person to state that the appraisals should be made “solely upon OBJECTIVE data” is M. That position wasn’t espoused by Dustin or any other contributor today. There is ALWAYS going to be an element of subjectivity in appraising — how do you adjust for brand new perfectly applied black and purple interior paint throughout a house? We all know it will pose a barrier to mainstream marketing, but saying so is an experiential call, a subjective call and one that cannot be quantified as data sources do not allow for it. But just because objective data won’t solve that issue, it nonetheless, is still very useful in narrowing down the ranges of many adjustments. The reality is that the valuations made without an attempt to quantify adjustments probably will be more inaccurate than those that work with the data using all mathematical tools available. What is the correct adjustment for a pool in a given market? Is it $5K or $10K or? Mathematical analysis will not give you that answer either, but it will allow one to find the mean value for a pool in a given area. Then the subjective takes over, but again balanced by the objective. For this example, the subjective side notes that the subject pool is over the top in quality and features and certainly will garner more from the market than the mean. How much more, well actual cost might bear some indication of that difference, but only an indication. There still will be a subjective call to make, but that call is guaranteed to be more accurate than the one made without such analysis. Bottom line — there simply is NO BASIS for rejecting using modern statistical tools to enhance valuation accuracy. NONE! To attempt to argue otherwise is willful ignorance of advancements in our field and will, as Dustin correctly pointed out, and as it should be, lead to problems for appraisers who attempt to do so. The final support for this position was inadvertently given by M who stated with certainty that staged homes sell faster than vacant homes. It simply is not true. As one who personally staged over 50 homes down to the flowers, paintings and knick-knacks for investors in the California market, and who had direct access to one investor who flipped over 1,000 houses, I know that the data did not support this common misconception. Another proof that one who just KNOWS based on subjective beliefs will never be able to present answers that match the reliability of one who gathers and analyzes the actual objective data.

  36. Hi,
    Bobby Crissp of Texas as a program called “Appraiser Companion”. It does provide tools for much of what you are seeking. Also, Bradford Technologies has a similar program that is a little more complex and is available. Google and Check them Out.

    1. Now there you go. Regression analysis, the future or appraisal credibility, brought to you by alamode!!!

    1. Has everyone forgotten “People can sale their house for what they want. Buyers chose to pay that price or the educated buyer will investigate what “Fair Market Value” is for that neighborhood, or get played right into a bad investment. Apples for apples, oranges for oranges is the golden rule. Some of you appear to be losing you wits about common sense and common appraisal standards, practices and how to measure worth/value.

      1. Ms Jones I totally get your common sense advocacy, but your credibility would be greatly enhanced if you write “people can SELL their houses” rather than “people can SALE their house”…… “Sale” is a noun, not a verb.

    2. Nice work Gary Kristensen,

      Surely that chart is for one neighborhood, and or, target area. Do you have similar charts for all the neighborhoods you cover??? If so what if Fannie comes back with a different chart in those areas that to their point prove/show/report totally different numbers. We all pick and chose our numbers chose wisely.

  37. I have all kinds of reasons why sq ft is what I state it is. For all neighborhoods. Simply put buyers and sellers determine the market not 10 years of regression charts. My lenders want 12 months, or and 6 months or less of reported data. Sometimes 90days or less. That’s all I have is what is reported in the MLS and other closed sale reporting agencies. I print and supply all my research data.

    I had one subject that I was asked to appraise in a twelve month period. The first lender I appraised the property of $130K 8 months later a different lender request an appraisal on the same subject. I report the standard “yes I have appraised this property before, blah, blah, blah.” As well as BEHOLD, RESPECT and REPORT what buyers and sellers are dictating. I appraised the same home a second time for $160K. Yes, a neighborhood that is in demand, it seems everyone wants to live is this GO TO in the NOW community. The demand is raising value faster than I can finish my morning cup of coffee. I state that in my report. DEMAND,DEMAND, LOCATION, LOCATION,DEMAND,DEMAND. Demand from and by Buyers and Sellers fuels price/value and per sq ft.

  38. We all need a second profession, hope you all have one in your back pocket. Why??? Apparently, and I do believe, DRONES will replace us EVENTUALLY. That’s for sure!!!

  39. This is ridiculous. Getting your comparables to adjust into a certain range requires “playing” with the adjustments to obtain a credible result. We do not live in Levittetown where all houses are similar and you can actually adjust based on what one house does versus one that doesn’t and come up with that value for that feature. That never happens in the real world. When is someone who actually appraises in the field going go have some input on changes? And just because the house is 2,800 Sf, the number could be $25 and in a much nicer area be $65. Whil i see what point you are trying to make on your article please refrain from scaring anyone who may be a novice in the industry, which sadly there are more than I care to count who continuously accept low fee work which is supposed to be “reasonable and customary”.. Maybe a uniform report, with uniform guidelines should be the first thing, along with increased fees for the increased work we have to do would be a start. I’m tired of having 6 different guideline on a report for 6 different lenders.

    1. So true and how to deal with the line, net or gross adjustments that are out of whack because a regression analysis model is telling you to adjust for $123.22/sf in a $100,000 neighborhood…..how are we NOT creating a misleading report?

  40. Why does all this sq ft business matters??? IDK, but I can tell you this those who are supporting and supply technology bamboozleness are only adding fuel for the DRONES to take over. ALL the banks need is to save money. Cut us/appraisers out the picture and use computer generated information, You regression analysis “wise guys” should think twice about what you are saying and stating. You are putting yourself right out of a job for a computer and DRONE to do.

  41. This is a lot of good comments and concerns. Would like to see your excel setup. I currently use macros within an excel form, but always willing to learn. Reference: Who moved my Cheese??

  42. Just say no, stop work for all Fannie related work. The first 1 week shutdown starts 1/26, if they continue on the crazy regression idea and refuse to share the data, threaten another week (lets set a date in advance), other, other, etc. What will / can they do. They cant fire us! Striking is the only way to combat bad labor practices. Wake up, we are under 100K Nationwide! Remember, you could be worse off, I appraise in Cuomo territory! SHUT IT DOWN LADIES & GENTS!

  43. John Steinbeck .. I couldn’t agree anymore… I’ve been appraising for 22 years and have never completed a regression analysis for one report. Never had a lender ask for it.. Never completed a RELO where a RA would be required… Now we have rumors going around the Fannie is requiring every adjustment in a report to have a RA done.. Can you imagine the extra time for all that and who’s going to pay for it…

  44. Just read above all the people that are asking for the spreadsheet info. Just tells you how many are actually doing it.

  45. It appears that everyone and their grandma would like a copy of my Excel regression sheet. I am flattered, but will not be sharing for several reasons. First (and most important), it is a bastardized tweaking of another, copywrited spreadsheet and I do not have permission to share. Second, it only works with my MLS system/area. Third, there are MUCH better worksheets and programs out there for regression. Finally, I would rather give you the names of others that I know such as:

    Compcruncher
    Regression Tool (on a la mode’s store website)
    SAVVI
    Redstone
    http://www.avttools.com/
    http://www.compflo.com/
    http://abc-mgt.com/
    Appraisal Tool

    If others would like to self-promote those you have to offer or give us some feedback on those you have used, I would love to crowdsource here and come up with what everyone feels is the best tool available right now. Please feel free to share.

    1. and most importantly the data used in your SS is from your market area and most likely is of no relevance to the markets of others…

  46. I think the message of this article is more geared towards “be prepared to defend your position on how you obtain the numbers for your grid adjustments”. I would much rather say I used droneian regression and paired sales analysis than say I simply closed my eyes and let the force be with me. LOL In all seriousness though I do agree that regression is one of the best methods, if not the best method, in obtaining a square footage adjustment, however, I disagree that you can use such a large range as has been shown here via link to get a reasonable result. That regression analysis is simply cost per square foot and has nothing to do with how much more a knowledgeable person is willing to pay for a specific property based on a small difference in square footage count versus a comparable listing or sale. Granted, the larger the number of homes used to obtain the statistical data in a regression analysis will give you a more credible accurate result, but only if they are fairly similar and comparable to each other, in not only location, but size, features and cost. I don’t think you should use every sale on the MLS in the past year, in your market area, and derive a singular figure to adjust for when making square footage adjustments. I also find it best to throw out the largest and lowest homes in the range due to the fact that they generally reflect non-typical transactions. I find paired sales analysis better in determining such things as time adjustments, special feature adjustments, and room count adjustments etc. than determining a square footage adjustment. But everyone has their own methods but be prepared to defend them. I can also surmise that if Fannie Mae is going to be asking these questions then every AMC review “drone” will also be asking these same questions. What this article didn’t touch on that it should is that Fannie Mae is also going to be looking at your UAD ratings for each home and comparable on each report and if you use that same comparable or the subject of your appraisal report in another appraisal it should reflect the exact same UAD ratings! If not Fannie will want to know why.

  47. I would like to review a copy of your spread sheet as well. If you have any suggestions as to what websites to look on for additional information on regression analysis would be appreciated. It’s always good to know several ways to obtain useful information.

  48. Proving adjustments – So we’re going from an “ART” to a “Science”. We’re going from a “Printed Form” to a “Mini-Narrative”. And, now if we screw up we will be Blacklisted – then our careers are over, just like that. All this for a $400 appraisal and some do them for much less. The risks of being in this business will skyrocket. The expenses are going up every year – fees, licensing, education, insurance and on and on. We need to raise our fees accordingly to cover all this risk. If they want an appraisal that is dead-on target as far as value is concerned then they will need to pay up. But I know that discussion has had little consideration. BTW all this for a home purchase that is mostly made on emotion. Do you have a spreadsheet formula for that?

    1. We need a formula that’s for sure! Maybe cost of fuel/MPG of vehicle * distance travelled + fixed costs + time to completion + risk + time spent answering AMC questions + all other associated costs = about $1,500 per appraisal if we billed it as others in business do. Problem is AMC’s use the least expensive appraisers to maximize profits, Lenders want poor quality work so down the road if they loose money on the foreclosure of the property they can successfully sue the appraiser, because their poor quality work which will make it easier to do so. Basically an insurance policy for the lender.

      1. Joel, you are 100% right! Frankly I think the minimum fee for a conforming, non complex FNMA transaction should already be $650.000 (to the appraiser, plus whatever the AMC needs on top). If I need to run (and monitor / adjust / weight) regressions software data to assure it retains validity; AND respond to CU generated inquiries I think $1,500 sounds just about right.

        Of course if it is COMPLEX, then its a whole new ballgame!

  49. Put me on the list for learning regression analysis. I would like the template too. Thank you!

  50. Hi Dustin, Regression Analysis is merely the current ‘soup du jour’ for appraisal support. As for being the ‘best’ method to support an adjustment, I have to give you an argument. The ‘best’ method (imho) is still matched pairs from our subject’s submarket or immediate competitive area. Admittedly these are scarce; and in the past were subject to so much liberal interpretation of what was a “match”, as to have become almost meaningless.

    Regression is a phenomenal opportunity for software and educational course providers to dip into the appraisers pockets even further. Yes, we should all know how to use it, but I don’t want to see it hyped up to the point where it is presumed to be the cure all for all facets of appraisal. Heck, if it is so reliable, then why bother with the appraiser at all? Just run regional regression analyses on all aspects and characteristics of property value and then have automated software create a value indication…you know, like ZILLO or Trulia; or other similar paragons of accurate property value estimates.

    IF regression were the sole answer to the question of how much to adjust, then wouldn’t follow that using FNMA’s even larger national database, that national or regional prices per SF could be determined and published as being acceptable? Obviously that wouldn’t relate to LOCAL markets though, would it? It also would not adequately measure things like high cost custom development areas (Hollywood Hills, Canyons, bluffs etc.) where the intangibles cannot be uniformly quantified by software. Views REQUIRE informed subjectivity. So do usable site areas; and condition differences (despite UAD belief that all property will neatly fit its predetermined definitions).

    Regression is merely another tool. BEFORE I commit to paying money for Redstone or some other magical pixie dust software, I want uniformity in the marketplace as to just how many ‘competitive’ properties are required for a valid regression analysis; and Id want clearly understood (widely agreed to professionally) concurrence as to what minimum levels of property diversity are going to be used.

    In Southern California; dating back to approximately 1970; Land to Improvement ratios approximated 75% land: 25% improvement. In the crazy days of 2005-2008 we saw LTIR as high as 90%, but the NORM was traditionally in a range of from 70% to 80%. During the collapse these were skewed the other way, with ratios as low as 30% or 40% much like many other parts of the nation.

    Now, in recovered areas, the old traditional ratios are becoming evident again. Using your 2,500 sf example and a reasonable MV (L.A. County, semi upscale neighborhood) of $750,000 a 30% IV would give you a demonstrated market indication of $90.00 SF. It would go up or down depending of prevailing ratios. Your 2,800 sf comp would have sold for approximately $27,000 more-all other factors being equal. IF it were rounded in the marketplace it may have gone for $780,000 so your sf adjustment would have actually been $83.57 (reflecting economy of scale) if measured from the larger comp. If measured from the small comp it would equate to $100 sf [ $30,000 / 300 = $100]. ANY of these adjustments could be supported just as well as a number derived by regression. So too, could use of numbers derived by depreciated replacement costs. Regression pretends to provide scientific certainty in an art that is also subjective; in a market that is imperfect.

    A free regression product that already works with our mls data is available through CANVAS. They caution however that you really DO need 200 transactions MINIMUM for reliability. Now ask your readers how many of their competitive markets have 200 CURRENT transactions? We ARE concerned that the data be current aren’t we?

  51. Michael-
    I respectfully disagree that regression is merely the current ‘soup du jour’ to quantify adjustments. The fact of the matter is, its been discussed within the appraisal literature for over 30 years, however, interestingly, while it has been referenced within the Appraisal of Real Estate Text, the 14th Ed. incorporates numerous examples of using graphic analysis, using basic regression, to quantify adjustments. And includes an entire “chapter” discussion within the addenda. The issue is not the validity of the concept, its more about the users. More than half the appraisers in the industry have over 20 years experience, which means they learned how to appraise before products like Excel were readily available. The industry has been slow to adapt to it, but with its recent inclusion in the 14th Ed., I suspect we will be seeing MUCH more of this technique. Thank you, Justin, for mentioning this important subject. …and based on the volume and tone of these responses, there are many people very interested in learning more! Best.

    1. I question the validity of using a simple regression analysis based on a whole market in determining a square footage adjustment. It is ludicrous to think that a regression analysis on a large market with a huge range in variables will give you a reliable figure to adjust for in all of your appraisals. There are simply too many other things to consider. An example: Lets say we do a regression analysis on a particular market that gives us a nice round figure of $100 per square foot. This is hypothetically based on homes in the market from $50,000 to $800,000 and square footages ranging from 800 sf to 7000 sf. So you are trying to convince to me that we throw demographics and common sense aside and decide to use this number on all appraisals? You think someone that is buying a 800 sf home for $50,000 can afford and would be willing to pay $10,000 more for a home with 900 square foot versus the 800 sf home they are also considering? Of course not and on the other hand do you think someone buying a home for $800,000 would expect to pay a mere 100 per square foot for a home that is 500 sf larger than the 6500 sf home they are looking at? I am sure they would jump at that chance. That being said if you were to find a median sales price and median sf for homes used in the statistics to create the regression analysis then you would probably be close to being correct with that 100 per sf figure somewhere in the middle. You also must consider the law of diminishing returns. As the square footage differences increase so does the price one will be willing to pay for each additional square foot. It is NOT a straight line graph. This is why you should only use similar comparable homes when creating a regression analysis and you must also consider that the fewer homes you have to input into the analysis the lower the reliability of the data you will receive from it. So there is no one size fits all number for any large market and no best one way to determine gla adjustments.

      1. “As the square footage differences increase so does the price one will be willing to pay for each additional square foot.”
        oops that should read As the square footage differences increase the price one will be willing to pay for each additional square foot will decrease.

  52. Dustin, always love reading you articles/input. Please add me to the list for receiving a sample of your excel spreadsheet – regression analysis…always looking for new ideas help in this crazy world of appraising.

  53. WOW a lot of opinions. In the spirit of true disclosure I developed and sell a regression program. I sold my first unit about 5 years ago. I am a bit of a numbers guy and pretty good at Excel. I managed a 20+ appraisal office for 30 years and sponsored at least 20 appraisers. I have worked long hours for meager pay for years- so I am one of you.
    On regression: Don’t panic buy any regression program because of the impending CU (which I am very suspicious of). Stats is really just analyzing more than a hand full of data at a time. Most of you are doing that now. Some are trending things in Excel, like SP to Date of sale, if you are then you are currently performing simple regression as Excel uses the sum of the least squares to position the line. So, little by little you are performing statistical analysis. Heck finding an average SP/SF from 20 sales is descriptive analysis. So here we are. I suggest everyone set an 18 month plan to become better trained in stats (including regression). I developed a 7 hr online class “Fundamental concepts of analysis” that is available at McKissock school- its a good start. It introduces statistical concepts and proves why things like paired sales really don’t often work. As appraisers you all know more about stats than you think right now. But as analysts you do need to know a lot about probability, collinearity, law of big numbers, how to work a spread sheet- and yes understand how regression works. But each of us will have to decide how far to take it. I am sure I have gone overboard.
    I am working on “an Appraisers guide to analysis” real examples not theory. I will try to put on two free I hour webinars the first of January to introduce regression analysis- they will be advertised. I did two webinars 3 years ago and they were well attended.
    A percentage of appraisers should learn a lot about stats (they need my program), but not every appraiser will want to attack regression- but even they will have to learn a little more about statistics.
    I honestly can’t say most of you should buy my R+ program. It is a one time purchase and comes with 100 pages of training material + 18 videos. But, all of you should have my program called USL Documenter which documents any appraisal report for USPAP.
    There is a lot of info at my site that will help no matter who’s program you use.
    Thanks

  54. Sure are a lot of appraisers out there who sound like they lack the competence to make a simple Excel spreadsheet that calculates simple regression analyses to support their adjustments. Hell you don’t even need a spreadsheet – you can do it on graph paper if you want to ! But the apathy bothers me. However, it is encouraging that there are a few commenters on here that do seem to understand the statistical tools available to them. But the number of less sophisticated practitioners is disturbing.

    If appraisers can not put in a little time and effort to learn exactly what it is they are doing, and are content using someone else’s program, then all they have is a black box that spits out answers. How will they ever know what the black box is doing if they don’t study up on it ? They won’t. Oh but hey, they will have someone else’s spreadsheet ! Hopefully it is idiot proof enough to CYA.

    Another bother is that this is yet another move by lenders toward removing any and all subjectivity from the value opinions Appraisers were once paid for. Our industry should push back. If not, by the time the regulators are done, a robot will be doing all the appraisals and many of you will be out of a job. The punch line is that the valuations will not necessarily be significantly better – and may actually be worse than they were before, in many cases. Ha !

    Sorry if my tone offends anyone, but having worked in mass appraisal for about 20 years, it is very frustrating to see so many “professionals” so eager to copy someone else’s work instead of doing it themselves. If you do it yourself, you will learn something !

  55. Wow! It is amazing how divided we all are on this issue. It is disappointing that this topic of regression is all being introduced on the off year of continuing ed. It seems as if no matter what side you choose on the issue, everyone could use some instructionals on how regression works, myself included. And now we will have to suffer through an entire year of flagged appraisals and rebuttals since it sounds like most of us have no idea how to use regression correctly. I have trouble grasping that they would introduce this now, with no instructionals on the issue. It has been 35 years since I have taken a statistics course in college. I continue to stay in the appraisal business, not because I love the industry, but because it is a source of income and a way to pay bills. Everytime something is introduced, it costs me more in BOTH time and actual dollars. When the MC 1004 was introduced, people said it would only take about 15 minutes to fill out. Well, that is an additional 15 minutes. When electronic delivery was introduced through Mercury or whatever software you use, they said it would only be an additional $10-15. Well, that’s another $10-15 PER APPRAISAL. When background checks were introduced, they said it would be about $50-75. Well, that is for one client. I wasn’t told that different clients only accept background checks from specific companies, thereby requiring more than ONE background check. And how are we supposed to quote on an appraisal job before we walk into the home, even after doing on-line research. Can you tell a plumber to quote you on a job before he comes to see it? Can you imagine what a plumber would charge in the amount of time it takes us to complete an appraisal? Now we are expected to purchase regression tools that further cut into the minimal appraisal amount that I am getting. It is impossible to pass on to the management company as they will only pay $300-350 for an appraisal or move on to the next appraiser. I am making less than I did 30 years ago on an appraisal and doing more work and taking on more liability. If the appraisal foundation would read some of the responses to the blogs as to how lost everyone is, it tells something about the industry. They want to make the appraisal industry an exact science and it just never can be.

  56. GLA adjustments should be market based and abstracted directly from the data in the subject’s market. There is no formula or regression technique that will provide the accuracy or support for adjustments better than graphed sales analysis of properties within the subject’s market. Once the graph line is created for that market showing a decrease in cost per square foot as size increases (all things being equal) then it becomes simple to compare any two points on the line to find the GLA adjustment. Adjustments should generally show a lower cost/SF amount when adjusting from larger houses to a smaller subject; and a higher cost/SF amount when adjusting from smaller houses to a larger subject. The use of a single number ($25 or some other “fixed” number) will not reflect the market reactions to changes in GLA.

  57. I would appreciate a copy as well. We might as well do this in concert with each other. Thank you.

  58. Dustin,

    I would like to see a copy of the spreadsheet if you are sharing. I’m attempting to make my own but would like to compare/use as a guide if possible
    Thanks
    Jon N

  59. This begging for an excel spreadsheet is embarrassing to our profession, thank you Dustin for not sharing it. Please before you all line up and follow the bright star in the sky educate your self on regression analysis. If you do believe this is the answer why not just use Zillow which is based on regression analysis, I wonder how many appraisers verify their values using Zillow? In most if not all areas of this country regression analysis will fail. Regression analysis is based on the law of big numbers which means the accuracy is based on large amounts of reliable and related data. Paired sales when available within the market area or a similar and competing area are the best and most defendable when data is limited, as it is in most areas of the country if not all. Dustin and I may have different views on this subject but I give credit to him for educating himself and most of all not sharing his spreadsheet. I assume everyone on this board holds yourselves out as real estate appraisal professionals, it will be a lot easier to create your own spreadsheet once you understand what regression analysis is before asking for a spreadsheet you think will do the work for you.

  60. I would also like to have copies of the regression sheets, or any other tools to use for regression. I have done the CVR’s but not sure if they are really accurate.

Leave a Reply

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

This site uses Akismet to reduce spam. Learn how your comment data is processed.