TAF and USPAP - great analysis

[djd09]

thanks for demonstrating your inability to complete even the most basic tasks (cannot call them skills). You know bupkis.

Diversity, Equity and Inclusion​

The Appraisal Foundation, the foremost authority on the appraisal and valuation profession, believes that the appraisal industry should reflect the diversity of the nation. We are proud to be partnering with our sponsors and professional stakeholder organizations in working toward a more inclusive and diverse appraisal community.

are you proud...

 

[George Hatch]

Do you seriously believe using paired sales is going to cut it when you’re appraising multi million dollar properties? Maybe in some cookie cutter condo project.

Do you seriously think matched pairs is the only recognized method or technique in the GSE policies.

You know which method most appraisers believe is wholly unsuitable for use with such multi-million dollar properties? Think hard.

 

[RCA]

Jeff Bradfod, at today's Zoom AI Conference in Monterey, CA, said he is coming out with a new product called "Nighthawk" that will also sell regression models for neighborhoods. I recommend using R/earth (or MARS), - he isn't being that specific, although anything else would be a black box. He says he has Ph.D.s building the models, which is like him. I tried to sell him on integrating MARS into his software in 2007-8, but he passed. OF COURSE, getting it off the ground would have taken much effort and many years. He has been working on and off of AI photo analysis since about 2015.

Well, he has always had a not good reputation for carrying projects through to the end. Maybe now he will, as he has always had some cooperation from Fannie Mae. He won't answer any specific questions about what he is doing, but I guess he has money somewhere to hire a few programmers and statisticians. I think maybe I know who.

This is the scenario best I can make out:

1. NightHawk will generate reports in the new Fannie Mae format, expected to arrive sometime in 2026. Maybe later.

2. NightHawk will provide models by neighborhood. We don't know how often they will be updated or specifics. The models will be created by Ph.D. statisticians who are not cheap. Maybe he has the political weight to change the laws so statisticians can do appraisals without an appraisal license. -- Just one problem of many to be sure.

3. There is a liability problem. Other companies like Zillow and House Canary could also essentially sell such models -for thousands and thousands of appraisals across the US. They have thought about it, but have not carried forward for sundry reasons. Some such appraisal models would be messed up, absolutely for sure 100%. -- And then the question is, who will pay the price for the ****-ups. Jeff wants the appraiser to take full liability. He says he will have courses to teatch the apppraisers what they need to know. Good luck with that.

4. A good MARS regression model will predict all the measurable variables, such as GLA, room count, lot size, etc. In the SF Bay Area, that is 70-85% of the value -of the subject. The appraiser gets to determine the residual for the subject, which is added to the MARS estimate. So, if there is a significant problem with value, -> that points to Bradford Software, you can be sure they will be sued, if anyone. If only the appraiser gets sued, it won't take long for appraisers to become very shy of the software. In any case, there will be enough cases where the models will not be good enough for non-conforming homes- and then what to do? Jeff hasn't figured that out - at least, he didn't have an answer today. Of course, if appraisers sign an agreement to take full responsibility for the accuracy of the models, Jeff is off the hook. Oh yea. - Unless they start highering math and STEM grads to do the work, I think he will have problems.

Anyway, there isn't any value-added service appraisers can provide in this new system unless they are MARS/Appraisal savvy. Given that Stephen Milborrow has been working for 20+ years on R/earth, I don't think Jeff's programmers have much chance to create their own. Stephen Milborrows program is open source. If Jeff uses the code, he must provide it for free to the public. He could get a license from Minitab for their MARS - which doesn't work too well with their scripting languages, not as good as R in any case, but still good. Only one individual license is $16,000/year. I am unsure what small fortune he will pay yearly for several thousand appraisers.

It would be a good thing if Jeff could get his NightHawk to float. It will make headway for all who want to use advanced statistics. -- But we see what happened to Redstone and other projects he has started. The Glassdoor comments give some insight (they aren't from me, btw). - Jeff would instead go on a vacation to Russia or some such place rather than finish his startup projects if past patterns of behavior mean anything. But, on the other hand, maybe he felt the time wasn't ripe for change in the past, and now it is. Hmmmm.

A representative from Fannie Mae was present at the conference to discuss their great new report system. I let him know what I thought about their great progress in the past but that they still have their head stuck in the sand. -- No, I didn't say that, but kind of, the moderator of the Zoom conference was trying very desperately to cut me off. But I said enough. And somebody said: Well, he won't get any more Fannie Mae appraisals! Ho hum.

 

[Zoe]

I still say TAF should be voted in by all licensed appraisers. Licensed appraisers are east to contact with what you stand for and why they should vote for you. I don't know how many positions on TAF but give each licensed appraiser that many votes. 7 members....each appraiser gets 7 votes. I assume with the way BOT appoints TAF members, BOT should be voted in too by all licensed appraisers.

I think State appraisal boards should be the same way. Only appraisers in the State with capability to vote for people running for the position. See how hard it would be for an AMC to get elected to a State board.

It would shift power to all licensed appraisers vs many connected to certain biases. There could be short list of topics that candidates could email all licensed appraisers in order for the appraiser to decide on who they wanted to vote for.

 

[Francois K. Gregoire]

Diversity, Equity and Inclusion​

The Appraisal Foundation, the foremost authority on the appraisal and valuation profession, believes that the appraisal industry should reflect the diversity of the nation. We are proud to be partnering with our sponsors and professional stakeholder organizations in working toward a more inclusive and diverse appraisal community.

are you proud...

Understandable you may not like this. Serious question: Does this have a negative effect on you or your business? If so, how? Do you have a solution to propose? Who knows, there might be someone or some organization to lobby and get your solution implemented.

 

[Mejappz]

Jeff Bradfod, at today's Zoom AI Conference in Monterey, CA, said he is coming out with a new product called "Nighthawk" that will also sell regression models for neighborhoods. I recommend using R/earth (or MARS), - he isn't being that specific, although anything else would be a black box. He says he has Ph.D.s building the models, which is like him. I tried to sell him on integrating MARS into his software in 2007-8, but he passed. OF COURSE, getting it off the ground would have taken much effort and many years. He has been working on and off of AI photo analysis since about 2015.

Well, he has always had a not good reputation for carrying projects through to the end. Maybe now he will, as he has always had some cooperation from Fannie Mae. He won't answer any specific questions about what he is doing, but I guess he has money somewhere to hire a few programmers and statisticians. I think maybe I know who.

This is the scenario best I can make out:

1. NightHawk will generate reports in the new Fannie Mae format, expected to arrive sometime in 2026. Maybe later.

2. NightHawk will provide models by neighborhood. We don't know how often they will be updated or specifics. The models will be created by Ph.D. statisticians who are not cheap. Maybe he has the political weight to change the laws so statisticians can do appraisals without an appraisal license. -- Just one problem of many to be sure.

3. There is a liability problem. Other companies like Zillow and House Canary could also essentially sell such models -for thousands and thousands of appraisals across the US. They have thought about it, but have not carried forward for sundry reasons. Some such appraisal models would be messed up, absolutely for sure 100%. -- And then the question is, who will pay the price for the ****-ups. Jeff wants the appraiser to take full liability. He says he will have courses to teatch the apppraisers what they need to know. Good luck with that.

4. A good MARS regression model will predict all the measurable variables, such as GLA, room count, lot size, etc. In the SF Bay Area, that is 70-85% of the value -of the subject. The appraiser gets to determine the residual for the subject, which is added to the MARS estimate. So, if there is a significant problem with value, -> that points to Bradford Software, you can be sure they will be sued, if anyone. If only the appraiser gets sued, it won't take long for appraisers to become very shy of the software. In any case, there will be enough cases where the models will not be good enough for non-conforming homes- and then what to do? Jeff hasn't figured that out - at least, he didn't have an answer today. Of course, if appraisers sign an agreement to take full responsibility for the accuracy of the models, Jeff is off the hook. Oh yea. - Unless they start highering math and STEM grads to do the work, I think he will have problems.

Anyway, there isn't any value-added service appraisers can provide in this new system unless they are MARS/Appraisal savvy. Given that Stephen Milborrow has been working for 20+ years on R/earth, I don't think Jeff's programmers have much chance to create their own. Stephen Milborrows program is open source. If Jeff uses the code, he must provide it for free to the public. He could get a license from Minitab for their MARS - which doesn't work too well with their scripting languages, not as good as R in any case, but still good. Only one individual license is $16,000/year. I am unsure what small fortune he will pay yearly for several thousand appraisers.

It would be a good thing if Jeff could get his NightHawk to float. It will make headway for all who want to use advanced statistics. -- But we see what happened to Redstone and other projects he has started. The Glassdoor comments give some insight (they aren't from me, btw). - Jeff would instead go on a vacation to Russia or some such place rather than finish his startup projects if past patterns of behavior mean anything. But, on the other hand, maybe he felt the time wasn't ripe for change in the past, and now it is. Hmmmm.

A representative from Fannie Mae was present at the conference to discuss their great new report system. I let him know what I thought about their great progress in the past but that they still have their head stuck in the sand. -- No, I didn't say that, but kind of, the moderator of the Zoom conference was trying very desperately to cut me off. But I said enough. And somebody said: Well, he won't get any more Fannie Mae appraisals! Ho hum.

Bert can you get good results in your market using paired sales, sensitivity analyses, grouped Analysis or cost data?

 

[djd09]

Understandable you may not like this. Serious question: Does this have a negative effect on you or your business? If so, how? Do you have a solution to propose? Who knows, there might be someone or some organization to lobby and get your solution implemented.

asking me...call relman or one of your paying sponsor like the one that wrote ao-16 and they can do all the lobbying they want...but i know you are a bible thumping trump supporter

 

[Francois K. Gregoire]

The law firm cannot explain if this TAF initiative has a negative effect on your business. FTR, AO-16 is no longer part of the ASB Guidance.

You still do not know me. Does using pejoratives reinforce your identity, sense of self, or does it give you a feeling of power?

 

[RCA]

Bert can you get good results in your market using paired sales, sensitivity analyses, grouped Analysis or cost data?

We are talking about SCA, so throw out the cost data except possibly for new construction. I was just in an AI conference yesterday that had a couple of high-volume real estate agents discussing real estate along the California Central Coast. In Carmel, new construction SFR costs $4,500/sf for many homes. The agent said he recently sold a 2,500 sf home for $10.5M in Carmel. Santa Cruz - Monterey/Carmel real estate has been booming since the pandemic. Many wealthy people from Silicon Valley started moving in that direction. He said people see a house, fall in love with it, and are often willing to pay whatever it takes to get it (more or less).

The set of good sales comparables is usually very limited, so you have to deal with large differences and a number of differences that all interrelate. It is a complex dynamic, and it is very, very difficult to deal with features in isolation. GLA interacts with room counts with lot size with location and so on. If you adjust one up or down - you need to adjust all the others. It is a situation where all the impact of all features on sale price have to be handled as a whole. Paired sales treats these features in isolation, under the assumption they do not interact with other features. That is a MAJOR failure of paired sales. Linear Regression assumes that a given features impact on price is constant throughout its range of value. That is absolutely not the case, especially if we have to expand outward from the subject beyond its original subdivision (if it were even part of some subdivision originally). Of course in older areas with many updates, you can have major differences from one house to the next neighboring house.

The only thing that works consistently and is NOT a black-box approach (such as Random Forest, Neural Networks, ...) is MARS (aka "earth"). Earth can be used to increase your understanding of a neighborhood or market area. It will tell you that, for example, The value of a bathroom, if a house has more than 3 bathrooms, is impacted by whether or not the house has an ADU. Yes, it can tell you that and give you an approximate value for the combination of ADU and bathroom count as an adjustment to the value contribution on top of the contribution from the total bathroom count. It is that precise. I can say, I can go into a completely new market area, run Earth and walk away with a better knowledge of market value in that area, than an appraiser who has been working there for 20 years.

In the SF Bay Area, the value contribution of GLA changes as you go up in value - and usually at specific points or "knots." So you may have $700/sf up to 1200sf, then $750/sf for each additional square foot up to 2500/sf, then something like $500/sf for each additional square foot over 2500 sf and up to 4000 square foot and then have it level off beyond 4000 sf or even drop off. You may get a nice R2 of 0.80 or higher from such a model. On the other hand, a linear regression model may do no better than 0.40 or 40%.

You see, in real estate, at least in a large mixed Metro area like the SF Bay Area, there are many different neighborhoods, which are either mostly custom-built or originate from older subdivisions with houses that have been updated over many years. There is certainly no single underlying parametric distribution like the normal distribution. No - the variances in property values change over all feature ranges. The most you can do is just fit a model to the whole mess - a model that is good at replicating previous sale prices and does good on tests for robustness on the value of unknown sales ( We do this 100 times: Randomly divide our 200 sales into 10 partitions of 20 sales, then randomly choose 9 out of the ten partitions to build a new Earth (aka MARS) model to test on the remaining 10th partition and see how close it comes to estimating the sale prices in that 10th partition. Then, choose the best model - or average the best models into one. )

 

[Mejappz]

We are talking about SCA, so throw out the cost data except possibly for new construction. I was just in an AI conference yesterday that had a couple of high-volume real estate agents discussing real estate along the California Central Coast. In Carmel, new construction SFR costs $4,500/sf for many homes. The agent said he recently sold a 2,500 sf home for $10.5M in Carmel. Santa Cruz - Monterey/Carmel real estate has been booming since the pandemic. Many wealthy people from Silicon Valley started moving in that direction. He said people see a house, fall in love with it, and are often willing to pay whatever it takes to get it (more or less).

The set of good sales comparables is usually very limited, so you have to deal with large differences and a number of differences that all interrelate. It is a complex dynamic, and it is very, very difficult to deal with features in isolation. GLA interacts with room counts with lot size with location and so on. If you adjust one up or down - you need to adjust all the others. It is a situation where all the impact of all features on sale price have to be handled as a whole. Paired sales treats these features in isolation, under the assumption they do not interact with other features. That is a MAJOR failure of paired sales. Linear Regression assumes that a given features impact on price is constant throughout its range of value. That is absolutely not the case, especially if we have to expand outward from the subject beyond its original subdivision (if it were even part of some subdivision originally). Of course in older areas with many updates, you can have major differences from one house to the next neighboring house.

The only thing that works consistently and is NOT a black-box approach (such as Random Forest, Neural Networks, ...) is MARS (aka "earth"). Earth can be used to increase your understanding of a neighborhood or market area. It will tell you that, for example, The value of a bathroom, if a house has more than 3 bathrooms, is impacted by whether or not the house has an ADU. Yes, it can tell you that and give you an approximate value for the combination of ADU and bathroom count as an adjustment to the value contribution on top of the contribution from the total bathroom count. It is that precise. I can say, I can go into a completely new market area, run Earth and walk away with a better knowledge of market value in that area, than an appraiser who has been working there for 20 years.

In the SF Bay Area, the value contribution of GLA changes as you go up in value - and usually at specific points or "knots." So you may have $700/sf up to 1200sf, then $750/sf for each additional square foot up to 2500/sf, then something like $500/sf for each additional square foot over 2500 sf and up to 4000 square foot and then have it level off beyond 4000 sf or even drop off. You may get a nice R2 of 0.80 or higher from such a model. On the other hand, a linear regression model may do no better than 0.40 or 40%.

It's a shame that some voices on the forum still hold strong beliefs in the effectiveness of paired sales despite their evident challenges in accurately reflecting the complexities of real estate markets. Your insights from the AI conference, especially regarding the booming real estate in Carmel and Santa Cruz due to Silicon Valley migration, shed light on how unique each transaction can be.

It's clear that traditional valuation methods struggle to capture all the nuances involved, such as the interaction between GLA, room counts, lot size, and location, which all influence property values in intricate ways.

I just completed an analysis where the bedrooms and living area showed a strong correlation of 0.74, both were on a curve. The bedroom adjustment was higher than expected, but everything resolved smoothly when I restricted the selection to sales with exactly two bedrooms, matching the subject property. The subject property was in the $1,200,000 range, and the model's error was only 1% off the actual price.