... A problem with statistics a larger data sample is needed, thus results become based on a number of properties that are only marginally competitive to subject,.
It sounds like you are talking about parametric statistics, which is sensitive to the size of the data sample. In parametric statistics we are always concerned with how well the sample size REPRESENTS the underlying population, which is assumed to have the same usually normal distribution. Real estate data is lumpy, non-parametric. Use median, not average. We have different measures of spread. With non-parametric methods, the size of the sample really depends on the number of features you want to analyze. MARS typically requires 15 rows of data to get started; but you can duplicate some rows to get around that. I've built MARS models with 12 home sales. It creates an explanation for why there are differences in the sales, such as GLA or lot size, and provides estimates (adjustments) of their various contributions to value.
In addition, lots of raw data is "off" from MLS or public records. Is the appraiser going to verify and correct the data on 50 or 100 properties ? No.
Actually we do. You can put these and more into a database and write a script that goes through and makes changes to hundreds of rows. I go out and get GPS data for the addresses and merge that in. Also Census data. And create new columns off old ones that can't be analyzed directly. In fact, I have about 6 million sales in my database - and sometimes make changes to all of them at once.
Still, Statistics might shine for some adjustments such as market appreciation but be "tone deaf" for others.
Yes, that is a major point I make. But you have to remember, that on sales comps, the value impact of all features is "captured" by the sale price. That is your key to opening up the secrets to every sale. The Sale Price is absolutely critical.
An appraiser has a built up knowledge bank from being in the market every day and analyzing sales, over a period of years. So they never "just" analyze 3 or 4 sold comps in a vacuum. I might pull 20 sales to consider, narrow them down to 8 possibles, research them, then eliminate 4 and use 4 as my sold comps. So more sales were analyzed than just those 4 comps.
Well, it can be a lot worse than that.
Imo, the ideal adjustment is comprised from a number of sources: the sales comps themselves, right there on the grid, adjusted from paired sales and line item sensitivity to the subject.
Appraisers argue against paired sales. They are often hard to find, if they even exist. They mostly exist in text books and homogenous subdivisions. That is why the move to statistics.
Then one can compare an adjustment to cost approach/depreciated cost. One can get feedback from surveying RE agents and other parties. One can look at market activity of listings and pending sales and apply that perspective. One can use statistics or RA as well. In some assignments, certain methods are more reliable or credible or possible than others.
But watch out. It is easy to go off on a tangent on this stuff. Take your logic and apply it to some other sale to create a beautiful contradiction. The real problem is that calculating the depreciation for something in the subject is easy ... but exactly how do you do the same thing and be fair to the comps, for which you probably don't have detailed information. You can take depreciated cost and make an extraordinary assumption about market reaction, but that doesn't mean it is legit. Gas ranges cost from about $250 to over $25,000. For example, a high end 60" Thermador is $19K+. Some buyers don't like the maintenance costs associated with high end ranges. Many could care less. You in fact, can't just "assume" that market reaction reflects depreciated cost. (aka functional obsolescence) And so much of it depends on context. In a high end kitchen in a high end home, you just expect a high end range. But of course, some people are crazy enough to stick one of these in their run of the mill homes. And certainly expect an appraiser will add its depreciated value to the value of their home, without regard to "market reaction". And if you think you are going to find a $20K range in an average home in an average neighborhood, good luck on that. .... Of course this sounds like an impossible assignment. An experienced appraiser, would probably take 30% of the depreciated value and most appraisers would probably agree with that (I'm guessing). If the owner complains you have to tell him: "Well, when you move, take it with you. Or sell it." That's a clue. If someone buys the home, they could turn around and sell the damn thing. --- Yea but if they have to go to the trouble to sell it, that's a hassle, they don't have time to deal with .....
Perhaps in future appraisals can include an AVM or stats as part of the report. The biggest problem is getting clients to allow enough time/ pay to better develop /support adjustments .
Exactly WHY would they do that? AVMs have no intrinsic value, because you don't know how they came up with the vallue.
Selling your method by calling appraisers too stupid to understand it... when did that sales pitch ever work?
Unfortunately, a software company that wants to sell software to appraisers, may moe or less come to that conclusion in testing the options before massive investment. If "someone" asks them why they didn't take a more advanced approach that is EXACTLY the response you are likely to hear, whether justified or not.
The best inventions are simple to use.
God is or can be merciless. What is is what is. To some people something may seem ultra simple. To others, they don't get it no matter how much you explain. It's like Eratosthenes calculation of the circumference of the earth in 200BC. It was another 2000 years before most people understood it. And it is damn simple. The geometry you had in high school. Run a line through two parallel lines and the opposing angles are equal. Two parallel rays emanating from the sun. One hits Syene (Aswan), lighting the bottom of a deep well 100%, one day a year. At the same time the other hits any vertical object in Alexandria, 516 miles to the north and throws a shadow that is 1/50 the arc of a circle. Thus the arc from the center of the earth to Alexandria and Syene is 1/50 of a circle and the earths circumference is 50 times the distance from Syene to Alexandria.
To someone who can mentally create abstract representations of reality this is simple. But many people simply cannot create such abstract representations.
Electrical grids are complex but any idiot can turn on a light bulb / power up with a switch, which is why everybody is willing to pay for electricity.
But USPAP says you have to UNDERSTAND the statistical methods you use. If you want to use an AVM, you need to understand its internals - at least to some nebulous extent.
Do a few sample appraisals with your method perhaps and post results.
I can do that. But it is work, to do it good.
We are aware appraisals are not 100% "accurate" in the value estimate, that is an accepted premise of appraising due to the nature of properties and buyers/sellers. Appraising is about solving a marketability problem, not solving a math problem. Math can be used as part of the SOW but we do not set out to solve a math problem, that is why the human perspective in integrating the various parts of the SOW drives assignment results.
Of course. Although there are many reasons for valuation and measurement - not just marketing.