Avm's, Are They Reliable?

[Mike Simpson]

I've had clients jump ship & go to Countrywide. They get there and give me a call; "Mike would you be willing to do appraisals for us?" "Nope," I tell them, "I'm as busy as I could be with full paying clients."

I once attended a class "The Appraiser as an Expert Witness," in which the instructor told everyone he'd been to a 'Countrywide seminar.' According to the instuctor; some big Muckity-Muck (his name escapes me) from Countrywide began the seminar by pointing to everyone in the crowd (predominately appraisers), and said, "It's my goal to put you all out of business." I quit doing appraisals for them & their AMC shortly thereafter!

Now, the question is--if you put all the appraisers out of business (if that's their goal) then where is your data going to come from? Who's going to analyze the whatever data there is, wherever it comes from? Who's going to reconcile the data? Are they going to force tax assessor's Countrywide (pun intended) to toe the line so their data will be accurate?

Anybody wanna put an immediate stop to this? STOP DOING WORK FOR COMPANIES WHO'D PUT YOU OUT OF WORK!!! It's not that hard to find good clients!

-Mike

 

[Restrain]

Every time I've gotten the raw data from an AVM (of course it was higher), the sales data was corrupt. The AVM is only as good as the data and the regression analysis built into the computer. However, 4 blocks either way can totally change the market and AVMs cannot handle such issues.

That being said, if you're doing a 70% LTV where the homeowner has been in the home 10 years, and an AVM gives an acceptable value, you're doing a risk analysis that is probably not a bad risk. However, high LTVs are going to be incorrect and very high risk with an AVM.

I was involved in testing one of the first AVM systems in Texas and they never could get the data to an acceptable level of confidence. In areas like California, it could work due to the availability of data, but even then required appraiser tweaking.

Roger

 

[Carnivore]

I posted this in an earlier thread but thought it appropriate to post again. Comments!

Found this and thought it might promote some lively conversation.

http://realdata.com/learn_frame_regression.htm

Heres is the best paragraph in my opinion and drives right to my contention of where Banks and other big lenders have been Hoodwinked by Slick quick talking Salesman peddling there AVM's.

Using Linear Regression in Real Estate Market Analysis
and Projection
By Frank Gallinelli


"We need to keep in mind that, as with most projections, the quality of our output is entirely dependent on the quality of our input. We certainly have to make appropriate choices for our comparables. Otherwise we can't reasonably expect to achieve meaningful results. In addition, the kind of index we select must relate consistently to value. If we find tax assessments to be unreliable, we may want to try gross living area or experiment with a scoring system (X points for each bedroom, Y points for each bath, etc.). "

 

[Pamela Crowley (Florida)]

:rofl:

G I G O

:rofl:

 

[Austin]

Least sum of the squares multiple regression analysis is a mathematical procedure. If you have a perfect data set, the regression will give a perfect results, point being there is no flaw in the regression process. I read this article and I have read many others by experts in regression. Here is where the problem lies that most people can’t seem to understand: Its all in the algorithm or the order and independent variables used in the analysis. It is like adding, if you put the wrong numbers in the adding machine you will get the total but the wrong total. The problem presented in that article was totally out of touch with the underlying theory because the variables analyzed were the wrong ones. You can be the greatest mathematician in the world but if you don’t know the correct underlying appraisal theory then you can’t solve the problem.
Some examples: I have done thousands of regression analyses. Any time you see a regression with independent variables like room count, view, style, or similar non quantitative independent variables, I know the regression will not work for two reasons: 1. The sequence of adjustments is wrong. 2. These variables are covariant meaning they are all correlated to each other. This is called multicollinearity and when it exists in the variables the math will not work and if it is bad enough the regression will not even give an answer. The present accepted method of using a marketing grid completely ignores MC and just papers over it not to mention that the sequence of adjustments is not even addressed.
It takes two things to make regression work: 1. Knowledge of regression, which is not complicated. The computer does all of the calculation. 2. Understand the correct underlying theory so you can use the correct sequence of adjustments and thus analyze the correct value influencing independent variables.

 

[Sean Wagle]

Programmers have been trying to predict the stock market on computers for decades. Don'cha think there's enough incentive to try it?? But a human money manger (or two, or several) still need to look at the output, and make adjustments.

My uncle is a weather observer and could say a thing or two about how accurate the automated, computerized readings are, which the National Weather Service uses. Yeah, they're OK for temperature and wind speed, but they just don't "see" clouds and fog like people do. Major airports still need humans to take measurements. B)

Title companies have been trying to get rid of humans for a while now in title research... and they might succeed, someday, when the majority of courthouses are using a unified system. But it won't happen anytime soon with the tight state and county budgets we're facing.

I say, if computers can "solve" appraising, it will be after these things are working. Nobody reading these boards today will be "replaced".

 

[Fred]

Posted by
"If you have a perfect data set, the regression will give a perfect results"

If you have a perfect data set, ANY correctly employed "method" of analysis will give perfect results. If you have a perfect data set, any incorrectly employed method (and this includes regression) will not produce perfect results - except by accident.

 

[Austin]

Steven:
To prove to myself the power of regression analysis I once devised a perfect data set consisting of about 50 sales. Highly complex problem with about five independent variables with 10 location influences and quality of construction variance built in corrlating to GLA but perfectly mathematically balanced with no variance. The regression solved the problem in less than one second and the graph of the results was a picture of perfection. Amazing! If you and 1,000 other appraisers spent then next ten years using conventional methods I will bet you any amount of money you could not solve the problem. I have toyed with the idea of sending this problem into an appraisal magazine and let them publish it and give three months to see who could solve the puzzle. No one would come close in 1,000 years-except by accident. Oh you would get an answer but it wouldn’t be the correct answer but you would never know how correct or incorrect your answer really was.
You can dig a 300-foot well with a shovel and get a perfecly good well, but why would you want to do that when a machine can do in half a day?