RCA
Elite Member
- Joined
- Jun 27, 2017
- Professional Status
- Certified General Appraiser
- State
- California
For those who are getting back into appraisal to earn some extra income in retirement (I am now 71), I just finished my first residential appraisal since when I quit back in 2008. It was a retrospective estate appraisal of an average home in El Granada, CA. I originally estimated two weeks to get it done (as I have other work as well and needed to do more work on setting up the infrastructure), but got it extended to three weeks. It was 70 pages, not counting all the attachments, and even then, I left out things I would have preferred to put in.
It turned out I was rusty on a few items. Of course, what you are rusty on, you’ve forgotten, so you don’t think you’re rusty. But it was relatively minor stuff.
Importantly, being able to go in and do exact (per ANSI) measurements of the exterior, as well as fairly exact and detailed measurements of the interior was still there. I now use Chief Architect’s Home Designer Pro, instead of the Premier package. But it's the same thing, - when it comes to the kinds of floor plans I do.
I still use MARS just in the way I presented it at the 2006 Salford Data Mining Conference (http://docs.salford-systems.com/BertCraytor.pdf). However, I use the latest SPM 8.2 version, which has some new features for automation and controlling the generation of the model. In particular, I like the ability to specify a variable as “non-transforming”. I did that with “age”, in order to make it a linear function, which people prefer; but also, there is a problem in El Granada, in that newer homes, less than 20 years old, have increasingly more upgrades and features. MARS, if not otherwise directed will break the age function at 20 years (inserting what is known as a “knot”) and throw in the contribution of all these upgrades to age. That messes up the 0.0 – 10.0 scoring of the subjective CQA (Condition-Quality-Appeal) variable. Most of that is kicked out of the “age” contribution by specifying it as non-transforming. The residual fairly accurately values your comps. They all have a sale price of course, and the difference between the sale price and the initial model predicted sale price, the residual, fairly accurately serves as a measure of the CQA for a comparable. The appraiser has to score the subject himself, percentage of homes in worse or better condition, quality and appeal. The human mind is very good at that sort of thing, we can look at a number of things like houses and pretty easily rank them from least to most desirable. I actually find this method is fairly accurate – especially if you can account for 80% of the value through objective quantifiable values like GLA, Lot Size, Age, Bedrooms, Bathrooms and so on. If you are off by 10% in estimating the remaining 20%, that is only a 2% error. You do need a good regression program to get that R2 of 0.80 or something close. But this is the future.
What I also noticed, is that valuing comparables in this area is becoming more complex, i.e. in non-homogenous areas like El Granada. It is absolute murder. Anyone who tells you different is talking BS. There are so many collinear effects, that you simple cannot compare homes feature by feature. San Mateo County is now the highest priced county in California. It’s recovery in home prices since 2007 is astounding. … Another story. Anyway, $700 is my minimum price for doing a residential appraisal in such an area. It should probably be $1000. We’ll see. And of course I don’t do Fannie Mae appraisals, as I can’t deal with the BS and constraints to valuation - or the crazy low pay.
It turned out I was rusty on a few items. Of course, what you are rusty on, you’ve forgotten, so you don’t think you’re rusty. But it was relatively minor stuff.
Importantly, being able to go in and do exact (per ANSI) measurements of the exterior, as well as fairly exact and detailed measurements of the interior was still there. I now use Chief Architect’s Home Designer Pro, instead of the Premier package. But it's the same thing, - when it comes to the kinds of floor plans I do.
I still use MARS just in the way I presented it at the 2006 Salford Data Mining Conference (http://docs.salford-systems.com/BertCraytor.pdf). However, I use the latest SPM 8.2 version, which has some new features for automation and controlling the generation of the model. In particular, I like the ability to specify a variable as “non-transforming”. I did that with “age”, in order to make it a linear function, which people prefer; but also, there is a problem in El Granada, in that newer homes, less than 20 years old, have increasingly more upgrades and features. MARS, if not otherwise directed will break the age function at 20 years (inserting what is known as a “knot”) and throw in the contribution of all these upgrades to age. That messes up the 0.0 – 10.0 scoring of the subjective CQA (Condition-Quality-Appeal) variable. Most of that is kicked out of the “age” contribution by specifying it as non-transforming. The residual fairly accurately values your comps. They all have a sale price of course, and the difference between the sale price and the initial model predicted sale price, the residual, fairly accurately serves as a measure of the CQA for a comparable. The appraiser has to score the subject himself, percentage of homes in worse or better condition, quality and appeal. The human mind is very good at that sort of thing, we can look at a number of things like houses and pretty easily rank them from least to most desirable. I actually find this method is fairly accurate – especially if you can account for 80% of the value through objective quantifiable values like GLA, Lot Size, Age, Bedrooms, Bathrooms and so on. If you are off by 10% in estimating the remaining 20%, that is only a 2% error. You do need a good regression program to get that R2 of 0.80 or something close. But this is the future.
What I also noticed, is that valuing comparables in this area is becoming more complex, i.e. in non-homogenous areas like El Granada. It is absolute murder. Anyone who tells you different is talking BS. There are so many collinear effects, that you simple cannot compare homes feature by feature. San Mateo County is now the highest priced county in California. It’s recovery in home prices since 2007 is astounding. … Another story. Anyway, $700 is my minimum price for doing a residential appraisal in such an area. It should probably be $1000. We’ll see. And of course I don’t do Fannie Mae appraisals, as I can’t deal with the BS and constraints to valuation - or the crazy low pay.
