How Many Have Figured Out

[RCA]

Bert, you have yet to post anything that shows your method of nonparametric regression. Why is that?

I've summarily described the method. A detailed description, I'm working on. But, I leave it as an exercise for others to figure out in the meantime. I keep posting that Powerpoint presentation I did at Salford Systems 2006 Data Mining Conference. That pretty much describes it --- up to a point. Yes, I have gone further than that. But you should look at that. http://docs.salford-systems.com/BertCraytor.pdf

 

[Rick Stillman]

Well humpf.

There's an hour of my life I'll never get back by reading this thread. But I'm glad I did, as I might encounter some appraiser that actually believes this stuff.

Here's how it works in the real world - yeah, the world we live in that is not predictable enough to be analyzed and charted on a graph.

First of all, who is going to be the one to "grade" these properties. That's defining the data that everything else will be based on. If we account for all the tangible features, and factor in time adjustments, then we are left with the intangibles. Those will have a precisely defined range, and we can presume the one of the very top was/is the nicest property in terms of overall market appeal, whereas the property on the bottom of the list must be the least appealing property. Well, that's fine. Or, not.

Because my market (Denver) is so unbelievably diverse, there is no rhyme or reason to the market if you look at it as a whole. I just appraised two new homes in the million dollar range that have about half the living area of homes within a half a mile. It's the "yuppie" factor. It is common to have Silicon Valley types move in and overpay for properties, just because they can - more or less. Yes, it's that simple. The one(s) at the bottom of the list were not unappealing. Surprise! Well, my data analysis just blew up in my face. They bottom sales were estate sales, for which I am going to presume the family discounted the properties for a quick sale. What's $5,000 among friends? So they each took a 5K hit, and low and behold, these homes sold for a good 20% under market. So I guess we throw that grading procedure in the junk pile. Now we're back to square one. Good old one on one analysis and grading. Damn - all that money on my TI 5750-XL Pro down the drain. That hand held miracle machine was supposed to do all my work for me ~

Secondly, it amazes me that anyone thinks for a moment that we can use a logical mathematical method to calculate the most illogical market in the world. I can't keep track of the times I've asked myself "why did they sell for that price" when I'm analyzing comps. There is no rhyme or reason many times. There is a VERY general trend, but prices can vary on a street by street distance - and often for no good reason. I've had my share of appraisals that came in significantly below the selling price. And I've admired the great buys buyers have gotten due to sellers being clueless about the market. That's not an anomaly - that's par for the market all the time.

It is important to realize, that there is no way a new appraiser could see the market data and filter it like I am able to. That is a skill that has taken many years to develop. And no algorithm is going to step in and do it, as we have seen in my prior examples. Some things just cannot be quantified by a formula. But let's put this to a controlled test to see how it performs under monitored conditions - just for giggles. What's the favorite color of most Asian women? Who knows - but if we could figure that out somehow by surveying every Asian woman on the planet, we might be able to analyze and chart that. Until the next fashion phase comes out and women decide they like orange. Who can possibly predict that? And that's so simple. Imagine the 1,001 factors involving real estate - many factors being completely hidden from analysis - such as being in close proximity to work, or being far from noise, or being close to a favorite shopping center, etc. to infinity. I've dealt with every one of those scenarios.

Well, that was interesting. Not really, but if we get tired of talking about the weather or politics, we have this on our list now.

Thanks for the lesson. I feel much more confident on this topic if and when I might encounter someone relying on it. Do you feel good enough to defend this on the stand Bert? I wouldn't have any fear of going up against this. How many holes does it take to sink a ship? I don't this could withstand too many of them.

 

[RCA]

Well humpf.

There's an hour of my life I'll never get back by reading this thread. But I'm glad I did, as I might encounter some appraiser that actually believes this stuff.

Here's how it works in the real world - yeah, the world we live in that is not predictable enough to be analyzed and charted on a graph.

First of all, who is going to be the one to "grade" these properties. That's defining the data that everything else will be based on. If we account for all the tangible features, and factor in time adjustments, then we are left with the intangibles. Those will have a precisely defined range, and we can presume the one of the very top was/is the nicest property in terms of overall market appeal, whereas the property on the bottom of the list must be the least appealing property. Well, that's fine. Or, not.

Because my market (Denver) is so unbelievably diverse, there is no rhyme or reason to the market if you look at it as a whole. I just appraised two new homes in the million dollar range that have about half the living area of homes within a half a mile. It's the "yuppie" factor. It is common to have Silicon Valley types move in and overpay for properties, just because they can - more or less. Yes, it's that simple. The one(s) at the bottom of the list were not unappealing. Surprise! Well, my data analysis just blew up in my face. They bottom sales were estate sales, for which I am going to presume the family discounted the properties for a quick sale. What's $5,000 among friends? So they each took a 5K hit, and low and behold, these homes sold for a good 20% under market. So I guess we throw that grading procedure in the junk pile. Now we're back to square one. Good old one on one analysis and grading. Damn - all that money on my TI 5750-XL Pro down the drain. That hand held miracle machine was supposed to do all my work for me ~

Secondly, it amazes me that anyone thinks for a moment that we can use a logical mathematical method to calculate the most illogical market in the world. I can't keep track of the times I've asked myself "why did they sell for that price" when I'm analyzing comps. There is no rhyme or reason many times. There is a VERY general trend, but prices can vary on a street by street distance - and often for no good reason. I've had my share of appraisals that came in significantly below the selling price. And I've admired the great buys buyers have gotten due to sellers being clueless about the market. That's not an anomaly - that's par for the market all the time.

It is important to realize, that there is no way a new appraiser could see the market data and filter it like I am able to. That is a skill that has taken many years to develop. And no algorithm is going to step in and do it, as we have seen in my prior examples. Some things just cannot be quantified by a formula. But let's put this to a controlled test to see how it performs under monitored conditions - just for giggles. What's the favorite color of most Asian women? Who knows - but if we could figure that out somehow by surveying every Asian woman on the planet, we might be able to analyze and chart that. Until the next fashion phase comes out and women decide they like orange. Who can possibly predict that? And that's so simple. Imagine the 1,001 factors involving real estate - many factors being completely hidden from analysis - such as being in close proximity to work, or being far from noise, or being close to a favorite shopping center, etc. to infinity. I've dealt with every one of those scenarios.

Well, that was interesting. Not really, but if we get tired of talking about the weather or politics, we have this on our list now.

Thanks for the lesson. I feel much more confident on this topic if and when I might encounter someone relying on it. Do you feel good enough to defend this on the stand Bert? I wouldn't have any fear of going up against this. How many holes does it take to sink a ship? I don't this could withstand too many of them.

Beautiful. I think you are the first one to grasp the essential idea. And it appears you are running into the same kind of messed up problems I have to deal with.

How tight are your comps? Guys over here brag when their comps are within 5% of each other (e.g. Bruce Hahn who uses R). I just finished a pretty difficult one with the comps within 5% of each other, R2 0.77. Everything objective except my subjective score of the subject property in terms of the features not covered by the regression of the objective features that came through as important contributors to value.

How much is your subjective judgment entering the picture? For all comps and the subject?

But it sounds like you are saying "Trust Me" and forget everything else. But the question remains, why should anyone trust your judgement when it apparently lacks any justification other than your say-so?

As far as Silicon Valley types coming over and paying higher prices. I don't see a problem with that. It pushes the market price upward and likely increases the variance in prices. You could add a new column on your Excel spreadsheet for "Idiot Silicon Buyer" Y/N. Run that through your regression and get an adjustment for Idiots. Well of course, you had better think of a fancy name for that feature in your report; it probably wouldn't go over too well with the buyer or any of the agents involved. "Buyer Anomaly"? "Market Degradation"? ... There must be something more polite.

On second thought I have to tell you something that I have seen. And I saw this in West University Place in Houston. This is an upscale area with lots of doctors, attorneys and wealthy people. Larger homes in the $2M, but some smaller older homes that at one time you could find for $400K or so. So I know someone who bought one of these homes and I actually advised against it and said they ought to think of a larger home in Sugarland or someplace. I figured the sort of people that would buy a home in that area would with money would only buy the property to tear down the old house and replace it with a new large one - and there is in fact a lot of that going on in West University. Now, the funny thing is after about 3 years they turned around and sold the property for almost $800K, after spending only $40K or so to fix it up. I was very puzzled by this, because the rest of the homes in the area only increased in price by 20% or so. Finally I figured it out. A "Silicon Valley" type with a Ph.D. bought the house from a retiring police officer. And by buying that home, he increased the prestige factor to make it worth more. It always was worth more. It was just that it was stuck in a kind of price inertia. He essentially rebooted the house price to where it really belonged. Now, I have seen this in other places.

So, what I would suggest, is that in your area, with Silicon Valley types coming in from California, the homes really are worth more; it is price inertia holding them back. The agents sense this and move on it, because the outside buyers are not nearly so influenced by local price inertia.

There is a reason for everything, more often than not.

 

[Carnivore]

Bert I do understand what you are trying to accomplish here at the forum. I respect that and clearly you have a well above average grasp of the topic and possibly your an expert. Most of what you have said makes my head begin to spin, not because I am stupid or have a deficiency in intelligence, but mostly because I am just ignorant.

In the real world. metallurgy as an example of where this is very valuable; https://www.hutnickelisty.cz/en/application-of-specific-statistical-methods-in-metallurgy/

So i can see the value of regression analysis. OTOH what your opining is still subjective.

This is a pretty good article on different techniques and overall topic.

https://www.toptal.com/finance/real-estate/real-estate-valuation


Here is where I see a real problem is in this discussion. How does this improve my bottom line profit margin? I don't think it does for a variety of reasons.

1. The mortgage industry will not acknowledge or more likely chooses to ignore the investment we as appraisers have made in ourselves. We are not rewarded for professionalism. We are rewarded for Fast & Cheap!

2. If I educate myself to higher degrees of knowledge under the Fast & Cheap Model eventually any appraiser will leave the field and go to where it is better appreciated.

3. Lenders almost universally refuse to allow me the flexibility to have a trainee inspect. Even though FNMA says its OK.

4. Does it improve residential report credibility. I say yes, but only to sophisticated readers/users. Meaning that if my report is more credible by using regression then
that should translate into higher fee's(already mentioned) and more work. Carnivores reports are more credible than one of my friends at this forum hence Carnivore is always busy and his friends are less busy. That's only if I am priced the same as them. Fast n Cheap AGAIN!

6. The AQB doesn't wholly agree with you, because they accept even a BA in Puppetry as suitable for pre-educational requirements.

So I like the academic discussion you have presented. I hope i made some sense in what I said.

Question for you: Can Regression differentiate between tangible and intangible when it comes to Residential Appraising?

 

[DWiley]

Accuracy is broadly defined. What constitutes accuracy actually could take a book. And there are in fact books written on it. Sorry.

Yes, there are indeed books on accuracy. BTW, for rifle shooting, one of the best is Applied Ballistics for Long Range Shooting, by Litz I think you would like it, given your background.

As for appraisal - an appraisal is, by definition, an opinion. There are many definitions of "accuracy." Common ones include:

ac·CU·ra·cy
ˈakyərəsē/
noun

1. the quality or state of being correct or precise.
   "we have confidence in the accuracy of the statistics"
   synonyms: correctness, precision, preciseness, exactness, exactitude; More
   
   
   • TECHNICAL
     the degree to which the result of a measurement, calculation, or specification conforms to the correct value or a standard.
     plural noun: accuracies
     "the accuracy of radiocarbon dating"

In order to measure accuracy, one must have the "correct" answer/result as a benchmark for comparison. For example, one could measure the accuracy of your estimates regarding home ownership by appraisers by assembling the relevant home ownership data and comparing the results with your statements. Or, one could determine the accuracy of a rifle shot by measuring the distance between the impact point and the center of the target. What method would one apply to appraisal results?

In measuring the "accuracy" of an appraisal, what would one use as the "correct" answer to which the appraisal result would be compared? I have a valuer friend in the Netherlands who has devoted many years to an attempt to measure the accuracy of valuations by comparing them to sale prices. I am simply asking what measure you would use.

To me, measuring the accuracy of an opinion is oxymoronic; it is a task akin to trying to construct a round square. But, since you have specifically referred to appraisal "accuracy" numerous times, it seems you have a different view. I would just like an explanation of how you would measure accuracy for an appraisal. Given your many references to the phrase, you must have some view on the subject.

And, to be clear, I am not trying to debate or debunk any position that you have on the topic, I just want to understand your views. Given your background in both math and appraisal, I find your references to "accuracy" a little surprising. Thx

 

[J Grant]

Way too much attention is paid to prices in discussing "accuracy", and not enough attention to the value part. If we were providing price opinions, one could measure our accuracy with prices as the benchmark. But we are providing market value opinions. That is, the value comes first, expressed as a numerical dollar amount, not the other way around.

It's not price with value attached as a pesky tag along, it's value expressed as an equivalency of price.

An appraisal is not a formula that can be flipped,.It is not market value = price, thus price =market value. It is an opinion of market value at a most probable price equivalence point value ( per the MV definition used as defined )

Given a market value opinion can not be numerically "accurate", since no external benchmark exists as a measure, to represent it as such becomes misleading. The word accuracy also means truthful,/developed with care, attention to detail. So we may not be able to defend our value as "accurate" per a benchmark , hopefully we can say we developed it as accurately as possible- .

 

[RCA]

Bert I do understand what you are trying to accomplish here at the forum. I respect that and clearly you have a well above average grasp of the topic and possibly your an expert. Most of what you have said makes my head begin to spin, not because I am stupid or have a deficiency in intelligence, but mostly because I am just ignorant.

In the real world. metallurgy as an example of where this is very valuable; https://www.hutnickelisty.cz/en/application-of-specific-statistical-methods-in-metallurgy/

So i can see the value of regression analysis. OTOH what your opining is still subjective.

This is a pretty good article on different techniques and overall topic.

https://www.toptal.com/finance/real-estate/real-estate-valuation


Here is where I see a real problem is in this discussion. How does this improve my bottom line profit margin? I don't think it does for a variety of reasons.

1. The mortgage industry will not acknowledge or more likely chooses to ignore the investment we as appraisers have made in ourselves. We are not rewarded for professionalism. We are rewarded for Fast & Cheap!

2. If I educate myself to higher degrees of knowledge under the Fast & Cheap Model eventually any appraiser will leave the field and go to where it is better appreciated.

3. Lenders almost universally refuse to allow me the flexibility to have a trainee inspect. Even though FNMA says its OK.

4. Does it improve residential report credibility. I say yes, but only to sophisticated readers/users. Meaning that if my report is more credible by using regression then
that should translate into higher fee's(already mentioned) and more work. Carnivores reports are more credible than one of my friends at this forum hence Carnivore is always busy and his friends are less busy. That's only if I am priced the same as them. Fast n Cheap AGAIN!

6. The AQB doesn't wholly agree with you, because they accept even a BA in Puppetry as suitable for pre-educational requirements.

So I like the academic discussion you have presented. I hope i made some sense in what I said.

Question for you: Can Regression differentiate between tangible and intangible when it comes to Residential Appraising?

I think I agree with most of what you say. As to the last question, the quantities that you enter into the regression from the MLS database are typically tangible; such as GLA, bathroom count, lot size, garage cars, stories and so on. Sometimes you see something like "Quality of Construction" that comes from some rating the assessor did when the house was constructed - and you will find is largely useless. But mostly, what you are able to enter into the initial regression is what I would call tangible properties, because you can objectively specify those quantities without subjective judgement (other than reviewing to make sure that they are not erroneous or anomalies of some sort - like a probate sale). The regression creates a model on those known quantities. On the basis of that model you created you can input the comparable data and let it give you a "predicted" sale price. That sale price will invariably fall short or overshoot the actual sale price. The difference, under normal circumstances, is the value contribution of the features that you did not input, such as the Quality of Construction, Quality of the view, aesthetics, Condition, functional utility and so on. The actual Sale Price captures the value of the intangibles as the difference to the sale price predicted by the model. That is a measurement of the value contribution of the intangibles. It does include some noise, and it can include some tangible features that are missing in the input data. You need to account for that. But, mostly, in my area, it is going to be condition on the low end and quality on the high end. Note that on the high end, quality can relate to tangible things like the quality and configuration of appliances and countertops in a gourmet kitchen. Or showers, tiling, sinks, countertops, flooring in a master bathroom.

If you rank the sales by residuals, and look at the comps, you will invariably be able to confirm that the high positive residuals (where the model under-predicted the sales price) are the higher quality comps, and the high negative residuals on the other end are the fixers. If you see anomalies, then call the agent - and most likely you will find extenuating circumstances, like a probate sale by a sales agent who lives 50 miles away, and would rather keep the whole commission to herself rather than let a local agent sell the property. (Or possibly worse).

 

[RCA]

Yes, there are indeed books on accuracy. BTW, for rifle shooting, one of the best is Applied Ballistics for Long Range Shooting, by Litz I think you would like it, given your background.

As for appraisal - an appraisal is, by definition, an opinion. There are many definitions of "accuracy." Common ones include:

ac·CU·ra·cy
ˈakyərəsē/
noun

1. the quality or state of being correct or precise.
   "we have confidence in the accuracy of the statistics"
   synonyms: correctness, precision, preciseness, exactness, exactitude; More
   
   
   • TECHNICAL
     the degree to which the result of a measurement, calculation, or specification conforms to the correct value or a standard.
     plural noun: accuracies
     "the accuracy of radiocarbon dating"

In order to measure accuracy, one must have the "correct" answer/result as a benchmark for comparison. For example, one could measure the accuracy of your estimates regarding home ownership by appraisers by assembling the relevant home ownership data and comparing the results with your statements. Or, one could determine the accuracy of a rifle shot by measuring the distance between the impact point and the center of the target. What method would one apply to appraisal results?

In measuring the "accuracy" of an appraisal, what would one use as the "correct" answer to which the appraisal result would be compared? I have a valuer friend in the Netherlands who has devoted many years to an attempt to measure the accuracy of valuations by comparing them to sale prices. I am simply asking what measure you would use.

To me, measuring the accuracy of an opinion is oxymoronic; it is a task akin to trying to construct a round square. But, since you have specifically referred to appraisal "accuracy" numerous times, it seems you have a different view. I would just like an explanation of how you would measure accuracy for an appraisal. Given your many references to the phrase, you must have some view on the subject.

And, to be clear, I am not trying to debate or debunk any position that you have on the topic, I just want to understand your views. Given your background in both math and appraisal, I find your references to "accuracy" a little surprising. Thx

OK, what I typically do is look at a past appraisal I did, that sold at a later date, adjust for change in the average market price and see how close I was. It is not always that accurate. But by doing this, I have convinced myself in some cases that I should have made adjustments based purely on my subjective opinion, that I didn't because of lack of evidence at the time. But, clearly there is a risk in doing that.

One can also argue that a value conclusion is likely to be more accurate if you have adjusted comps that fit in a tight range, if your adjustments are well supported and objective. But that doesn't really measure accuracy directly. But seems to be an improvement that reduces the probability of error in estimating value. We could split hairs, but I don't have time.

 

[Carnivore]

Bert I appreciate and understand your reply to my post above. You said nothing that surprised me or that I don't already recognize as a factor. Intangible benefits to real property can be minimized to some degree by an individual appraiser. None the less intangible benefits may not be easily recognized and segregated due to additional noise(more intangible) and then either be discarded or accounted for in the analysis.

I chuckled a little bit when Dwily and Jgrant jumped on your use of the term accurate. What was funny about this to me was your analogy was incorrect.

Lets change this up a bit. We know what a precision shooter does before they shoot at a target. Lets go though all the same motions, except we don't have a target yet!

Input all the data, then point down range, fire, and observe where the round impacted. Walk down range with a target(subject) and place it where the data input caused the round to impact.

Lets do this in another way. Same scenario. We take a rifle to the range. A Large Piece of Blank Paper(That is our neighborhood) is set up. You fire a round with unknown weight(subject). We then fire three different rounds. One round is 150 grain. 2nd round is 165 grain and the third is a 175 Grain(comps). Maybe we could estimate the weight of the subject round by drawing some intersecting lines.

Added Comment: I change this after J grant responded below. I think I have made my point clearer. Maybe not! I can further complicate the analogy. The three rounds of known quantity(weight) were fired on three different days(months even), and different atmospheric conditions.

 

[J Grant]

Bert I appreciate and understand your reply to my post above. You said nothing that surprised me or that I don't already recognize as a factor. Intangible benefits to real property can be minimized to some degree by an individual appraiser. None the less intangible benefits may not be easily recognized and segregated due to additional noise(more intangible) and then either be discarded or accounted for in the analysis.

I chuckled a little bit when Dwily and Jgrant jumped on your use of the term accurate. What was funny about this to me was your analogy was incorrect.

Lets change this up a bit. We know what a precision shooter does before they shoot at a target. Lets go though all the same motions, except we don't have a target yet!

Input all the data, then point down range, fire, and observe where the round impacted. Walk down range with a target(subject) and place it where the data input caused the round to impact.

Your analogy is correct, we are not "shooting" at a target that thus can measure accuracy by how far "off we were, rather we are developing the target ( developing an opinion of market value from the SOW, it's MV unknown) . Prices form the parameters we work with. however they are not the sole benchmark of MV because we are not providing price opinions . We are providing value opinions with a price equivalency ( most probable price equivalency to the MV definition used in appraisal )

There's a saying, "price is what you pay, value is what you get" that sums it up well.

Thus, value can never be "accurate"...however it can be measured 1) against an internal factor- how credibly and well supported it was developed in the appraisal. and 2) by external factors - the attributes or defects of a property relative to $ paid, comparison to prices paid for similar properties, influence of market conditions, alignment with historical and pending trends, cost to build/replicate new, income or cash flow (or loss) generated.