Regression Analysis

[hastalavista]

I think what is telling about your statement here, and many other statements in this thread, is that a reliable go-to method to extract and support dollar adjustments does not exist...

I believe I understand your position, and don't fully disagree with it. But I see a difference which you may or may not agree with.
No buyer I've ever met (except for an appraiser, perhaps) would ever try to make a purchase decision using the same methodologies we do. They may look at some metrics (usually, $/sf) but they do not break it down into the individual components that we do.
But, we are not analyzing the data for a particular buyer or the circumstances surrounding a specific buyer and seller. We are analyzing the market data as a whole, and trying to determine and identify those components between similar types of properties which can explain why they sold where they sold at, so we can form an opinion of the price-point that our subject should sell at (definition of market value, blah, blah, blah).
Certainly most (appraisers & non-appraisers) would agree that, in general, such things as...
1. Location
2. Size of the house
3. Configuration of the house
4. Condition of the house
5. Size of the lot

...are the major components of value for residential real estate. And, we have tools to analyze how these components (called "elements of comparison") explain why a home sold for such & such. Those tools include (but are not always limited to):
A. Paired sales
B. Statistical analysis
C. Market participant surveys
D. Cost

We are going to look at data that we believe represents the subject's market dynamic, and analyze those differences and determine how, based on the differences of the subject and comparables, what the market (not a buyer, but the market) reaction is. But even the best analysis leaves us with a result that we then need to apply our judgment to and conclude an adjustment (or no adjustment). We do this so we can explain how and why we came up with the value we did. I wouldn't call this an exercise in futility; I'd actually say that it is a very good process to explain why something should sell at $X rather than $Y.

The problem (IMO) we face is when stakeholders expect or demand a level of accurateness or mathematical precision to that process. My math is accurate and precise. The result, however, is based on the quality and/or the quantity of data I have available to complete the analysis. The lower the quality or the scarcer the data, the more my judgment needs to come into play.

I'm confident I can breakdown most markets and explain what components are valued in that particular market.
I'm confident I can analyze the data and provide evidence that my adjustments explain the differences between the subject and comparables.
I can use the above along with my judgment to conclude an opinion of value which is market-based, supported, and very credible.
What I cannot do is explain how any specific buyer or seller may react. What I can do is make a persuasive and compelling argument that what I opine is how the market reacts.

So, IMO, I don't need to devise a valuation or adjustment scheme which duplicates how most (if not nearly all) of the buyers and sellers actually make their individual purchase/sell decisions (I cannot; there is a lot of emotion and intangibles embedded in the individual decisions/transactions).
I can and do devise a valuation and adjustment scheme which can show how, in general and with a high degree of reliability, the market in reacts.

Significant elements of comparison do impact how a property is perceived and how it might sell within any given market.
Those elements can usually be analyzed and adjustments can be extracted or logically deduced from the market data.
Different types of tools will allow for different types of analyses. Our objective is to use the best tool for the specific problem/element. The more tools we have, the more options we will have to solve the problem.
It is rarely the case that there is only one tool available; sometimes it takes a couple of different tools to point us in the right direction; and then it is up to us, using our judgment, to reconcile those indications into a specific adjustment (or non-adjustment).
Sometimes, it is better to consider elements in the reconciliation (no specific adjustment; considering it in the placement within the adjusted range. That is the ultimate judgment call.).
Used together and correctly, all of the above results in a process that can explain why a house should sell for $X rather than $Y. What it actually sells for is outside our ability (and any other) to forecast.

 

[Deleted member 130081]

I believe I understand your position, and don't fully disagree with it. But I see a difference which you may or may not agree with.
No buyer I've ever met (except for an appraiser, perhaps) would ever try to make a purchase decision using the same methodologies we do. They may look at some metrics (usually, $/sf) but they do not break it down into the individual components that we do.

2 cents - I do not disagree, but to play Devils advocate, why then would we not mimic what buyers actually do? Just sayin there's a question mark on the relevance of a sales grid, especially in terms of being misleading. You said it yourself, buyers don't itemize like that when they decide on an offering price. So, why do we do it then? My answer to that (other than our masters told us to do it that way) is we conclude they are making these decisions subconsciously (a fairly large assumption if you ask me) and that further, we actually can dive into the market data and define the subconscious in terms of dollars.

But, we are not analyzing the data for a particular buyer or the circumstances surrounding a specific buyer and seller. We are analyzing the market data as a whole, and trying to determine and identify those components between similar types of properties which can explain why they sold where they sold at, so we can form an opinion of the price-point that our subject should sell at (definition of market value, blah, blah, blah).
Certainly most (appraisers & non-appraisers) would agree that, in general, such things as...
1. Location
2. Size of the house
3. Configuration of the house
4. Condition of the house
5. Size of the lot

...are the major components of value for residential real estate. And, we have tools to analyze how these components (called "elements of comparison") explain why a home sold for such & such. Those tools include (but are not always limited to):
A. Paired sales
B. Statistical analysis
C. Market participant surveys
D. Cost

We are going to look at data that we believe represents the subject's market dynamic, and analyze those differences and determine how, based on the differences of the subject and comparables, what the market (not a buyer, but the market) reaction is. But even the best analysis leaves us with a result that we then need to apply our judgment to and conclude an adjustment (or no adjustment). We do this so we can explain how and why we came up with the value we did. I wouldn't call this an exercise in futility; I'd actually say that it is a very good process to explain why something should sell at $X rather than $Y.

2 cents - I agree 100% so far. We have made some big assumptions up to this point, but I think they are reasonable as well.

The problem (IMO) we face is when stakeholders expect or demand a level of accurateness or mathematical precision to that process. My math is accurate and precise. The result, however, is based on the quality and/or the quantity of data I have available to complete the analysis. The lower the quality or the scarcer the data, the more my judgment needs to come into play.

2 cents - You could not have said it any more true.

I'm confident I can breakdown most markets and explain what components are valued in that particular market.
I'm confident I can analyze the data and provide evidence that my adjustments explain the differences between the subject and comparables.
I can use the above along with my judgment to conclude an opinion of value which is market-based, supported, and very credible.
What I cannot do is explain how any specific buyer or seller may react. What I can do is make a persuasive and compelling argument that what I opine is how the market reacts.

2 cents - Yes to all. However please define support. That short, yet huge word is the monkey wrench. And when I ask to define it, I mean define it like our client defines it, cause after all, we do this for them, not us. The true issue is they do not have a clear definition of what they need, or want. It is up to us to provide them with what is possible, not provide them with an answer they want to hear, simply because it is the answer they want to hear, or confuse them into submission. You write very well, and if I reviewed your report, would conclude your are at least not an idiot, and therefore are probably a good appraiser too (an unsubstantiated conclusion, but a human one). However, good writing skills and tangible market evidence are not the same thing. Again, define support as our mortgage clients define it. I wonder how the tune of many appraisers would change if all appraisers were required to submit their source data and adjustment work sheets with the report, not unlike a 4th grade math teacher would require the student show their work - funny how no one ever asks for that, lip service and a signature is good enough (not saying I want that to happen for goodness sakes).

So, IMO, I don't need to devise a valuation or adjustment scheme which duplicates how most (if not nearly all) of the buyers and sellers actually make their individual purchase/sell decisions (I cannot; there is a lot of emotion and intangibles embedded in the individual decisions/transactions).
I can and do devise a valuation and adjustment scheme which can show how, in general and with a high degree of reliability, the market in reacts.

2 cents - Well, we might have to disagree on theory here a bit. Consider this, does the sales grid on a mortgage appraisal, in the eyes of the reader, reflect the reactions of buyers or not? You keep referencing a difference between a single buyer and the market as a whole, to which I understand and agree with. However, the market is comprised of buyers, not something else, so you can only differentiate that so far. At some point, you have to go back to the fact that the market is the buyer. USPAP says we must develop and communicate our conclusions to the client in a manner that is meaningful and not misleading. Do you think your thought process and subsequent actions pass that test? Do you think the reader understands the subtleties you employ? Just something to chew on.

Significant elements of comparison do impact how a property is perceived and how it might sell within any given market.
Those elements can usually be analyzed and adjustments can be extracted or logically deduced from the market data.
Different types of tools will allow for different types of analyses. Our objective is to use the best tool for the specific problem/element. The more tools we have, the more options we will have to solve the problem.
It is rarely the case that there is only one tool available; sometimes it takes a couple of different tools to point us in the right direction; and then it is up to us, using our judgment, to reconcile those indications into a specific adjustment (or non-adjustment).
Sometimes, it is better to consider elements in the reconciliation (no specific adjustment; considering it in the placement within the adjusted range. That is the ultimate judgment call.).
Used together and correctly, all of the above results in a process that can explain why a house should sell for $X rather than $Y. What it actually sells for is outside our ability (and any other) to forecast.

2 cents - My favorite part of this last statement of yours is "logically deduced". Yup, that passes plenty of tests, but does it pass the test of current client expectations?

Thanks for taking the time to write this response - I always like reading your posts. I put my two cents in italics above. Its clear you take a second to think about what you are doing, and what you communicate to others. I wish I could write as well as you. Peace.

(Edit) PS - Crap, the entire reply is in italics!!! Sorry. Ill put 2 cents in front of my 2 cents.

 

[Michael P Jacobs MAI]

Excellent post - thank you. Funny you should mention rocket scientists, because I did an appraisal for a real rocket scientist last year ... he couldn't for the life of him figure out how we did it.

I think there's probably a little too much rocket scientist in appraisers who can't get reports out-the-door on time. You have to have a lot of tolerance for the randomness of real estate and buyer/seller variables to be a successful appraiser!

 

[timd354]

I think there's probably a little too much rocket scientist in appraisers who can't get reports out-the-door on time. You have to have a lot of tolerance for the randomness of real estate and buyer/seller variables to be a successful appraiser!

Some appraisers are just terrible at time management and also some appraisers cannot say "no" to new work even when they already are working at full capacity. My mentor, who is a very skilled appraiser, is one of those people....he was (probably still is) just awful at completing reports on time and there was really no reason for it other than he just did not manage his time very efficiently. He also has a habit of accepting too many orders.

 

[Michael P Jacobs MAI]

Some appraisers are just terrible at time management and also some appraisers cannot say "no" to new work even when they already are working at full capacity. My mentor, who is a very skilled appraiser, is one of those people....he was (probably still is) just awful at completing reports on time and there was really no reason for it other than he just did not manage his time very efficiently. He also has a habit of accepting too many orders.

I realize we're hijacking the thread here and apologize, but I completely agree. If he's commercial he should know how to raise fees during the good times and adjust for the slow. As I understand it, with residential appraisers are pretty much locked in. When I had a high-volume residential-only firm in the 1980s I was once told by a mortgage banker that they could fix stupid or wrong but they could never fix "late." When I began commercial a mentor of sorts told me that appraisal consumers pretty much just want two things from us - get it done on time and don't screw it up.

Today, our fees are too high for most banks but when they have a hard deadline they pay because they know we'll work around the clock if necessary to get their reports in on time. We can abbreviate write-ups keep it short, but we can't deliver yesterday today.

 

[leasedfee]

. . . what is a good resource for making graphs and doing statistical analysis on Excel?

MS Excel has a free add-on called "Data Analysis". It is already loaded onto your machine, but you have to turn-it-on, though the advanced Options menu. It may be called the Analysis ToolPak. When you have your columns of data, it will give you a pop-up wizard. It will do a very nice multiple regression analysis. It more than works for my needs, and is friendlier than semi-professional statistical software.

Make sure you have a large sample size.

Be very careful with this. One of the critical, major assumptions (overlooked by real estate peoples) of regression and general Anova is a homogeneous population. I.e., an apple is an apple. In appraisal, as we expand n, the sample size, we expand into less homogeneous properties. We start skipping into different neighborhoods, different classes, floor plans, sizes, different parcel sizes, property classes, ages, materials. You are increasing variables (i.e., elements of comparison and even highest and best uses) faster than you increase the sample size. The outcome is nonsensical statistics. Sadly, the academics of real estate have gotten this one completely wrong and have spewed out garbage articles for decades. Statistical analysis with a smaller sample size is far superior. If you wouldn't use it as a comp for the subject, it shouldn't be in your data set. My friend Steven Kane who wrote Practical Applications in Appraisal Valuation Modeling for AI touches on this and agrees with me on this.
------------
Match-pair analysis is regression analysis with a sample size of 2 and 1 degree of freedom, 100% correlation, and no statistical robustness -- but at least it maintains homogeneity. However, a dozen match-pairs have very strong statistical robustness, and would be an excellent model, known as bootstrapping, if one was so lucky to find such pairs. You could prove beyond a doubt, better than regression, that a pool adds $50,000 more or 15% more value.

 

[Amy Perkins]

I found it useful for land sales when there were gaps in data, just used graph paper and drew my own trend line; excel also has this feature if you have lots of data. ACI analytics has regression analysis built into it. What is it for?

 

[Joe Flacco]

Be very careful with this. One of the critical, major assumptions (overlooked by real estate peoples) of regression and general Anova is a homogeneous population. I.e., an apple is an apple. In appraisal, as we expand n, the sample size, we expand into less homogeneous properties. We start skipping into different neighborhoods, different classes, floor plans, sizes, different parcel sizes, property classes, ages, materials. You are increasing variables (i.e., elements of comparison and even highest and best uses) faster than you increase the sample size. The outcome is nonsensical statistics. Sadly, the academics of real estate have gotten this one completely wrong and have spewed out garbage articles for decades. Statistical analysis with a smaller sample size is far superior. If you wouldn't use it as a comp for the subject, it shouldn't be in your data set. My friend Steven Kane who wrote Practical Applications in Appraisal Valuation Modeling for AI touches on this and agrees with me on this.
------------
Match-pair analysis is regression analysis with a sample size of 2 and 1 degree of freedom, 100% correlation, and no statistical robustness -- but at least it maintains homogeneity. However, a dozen match-pairs have very strong statistical robustness, and would be an excellent model, known as bootstrapping, if one was so lucky to find such pairs. You could prove beyond a doubt, better than regression, that a pool adds $50,000 more or 15% more value.

I have read your posts about regression in other threads and it is obvious that you know what you are talking about.

The thing is even a pool is not a pool. There are pools that cost $50k. There are pools that cost $250k. With matched pairs you are trying to find the value of a similar pool as the subject.

 

[Tom D]

Does anyone have any products that do a decent regression analysis directly from an MLS export? There are a ton available but all seem confusing to learn. Just want to make sure I choose a decent one if im going to put the work in learning it.

i live in an urban area where most of my comps are 1-4 blocks away, have 10-30 sales available so i am speaking from that standpoint. in the alamode store is a program call 'regression', which may have been mentioned here, that works with the sales i am looking at, exporting. i am not a regression believer, but for my situation it seem to work very, very, well in being close to my value. and if not, it's because of some out layer sales. it can be adjusted. i do a print screen of it, and put it into my addendum under regression analysis (looks quite appraisal important). for me it's one more aspect of supporting my value. the 'actual value' is on 1 page, but i now have 40 pages of info/support. for me recently, i love this thing. for suburban i don't know how well it works. for rural i don't understand an appraiser not shooting themselves. program can be tried for a week. don't like it, you get money back, no questions asked. worth trying. it gives you a number, not those stupid charts that no person can understand. i have attached a picture of what the program shows upon importing the data, and category comparison. nice neighborhood, very early in the morning.

 

[leasedfee]

The thing is even a pool is not a pool.

LOL, and you know far more about pools than I. You illustrate a weakness of statistical models: qualitative attributes. The likely datatable would simply dummy code No Pool =0 and Pool = 1. It'd average out the nice pools with the bad pools with the fabulous pools. Unless you sat down and tried to quantify the qualities: Big Pool vs. Small Pool (. . . and what is the functional size pool relative to the type of house?), or Simple Pool = 1, Good Pool =2, Great Pool = 3, Olympic = 4, and what to do with the hot tub and pool house. . . . As location improves, the homes can change. Homes with 2"x4"s and asphalt shingles in one neighborhood compare poorly to 2"x 6"s and slate roofs. Cross the street, and a new home in swanky neighborhood uses cheapo materials like EIFs and synthetic stone facade.