Remove Negative Remarks About the Client

[Bama Bayou]

What do you mean by 50% off?

You might consider writing that the "incorrect" values had something to do with bad data. The valuation model might work just fine.

Steven, here it was I did. I printed out a listing of every closed SFR MLS sale in my county for the preceeding thirty days. There were about 150 sales. Then I compared the sales price with the Zillow values. Zillow produced a value for about 70% of them. Some they could not geocode, and some they stated that property recently sold and it would not produce a value. Zillow claimed the data for my county had been updated about two weeks before I did the comparison so I think it was a fair comparison. I did this little study out of curiosity not to "dis" them.

I then simply calculated the $ and % difference between the "Zestimates" and the actual closed sales price.

I then compluted the average and median % difference (there was not much difference between median and average). The Zestimates were, on average, off by about 50%. It was not like they were typically off by 50%. The Zestimates were were all over the place but consistently high. Of 150 sales two of the Zestimates were below actual sales price and the rest were above it.

These are not good results by any standard.

As far as the model being good but the "incorrect" Zestimates being the result of bad data, well, tough. The Zestimates for this county are completely unreliable.

I have no idea whether it is because of bad data or their model because, like every other AVM I have seen, it is a black box AVM. There is no way to know the strengths and weaknesses of it unless they publish their methodology.

The accuracy of an AVM is not difficult to determine.

 

[JSmith43]

Bama, does the state department of revenue publish sales ratio studies? Of course, there is the time lag problem, but I bet the assessors are doing a better job.

 

[Bama Bayou]

Bama, does the state department of revenue publish sales ratio studies? Of course, there is the time lag problem, but I bet the assessors are doing a better job.

No, they don't. We have elected assessors in AL that are supposed to conform to the state dept of revenue guidelines.

In an unprecedented move, the state took over our county assessor office last fall because the values were all over the place. The tax assessor decided not to run for reelection after being in office for 16 years. He would have lost.

We are a non-disclosure state but you can back into the value by using the deed tax and mortgage amount if there is a mortgage. The deed tax is $1 per every thousand dollars of downpayment.

If the deed tax is $10 that means they put 10K down. Just add the mortgage amout to the 10K and you have your sales price. Pay 100K cash and your deed tax is $100. We have a data service that covers our county and produces the "sales price" this way. They are very good IMO and cover a good bit of FL also. Their data is updated over the web monthly and it appears that is the data Zillow is using for my county.

Zillow had recent sales, including several of the ones that I was trying to test. That was not the problem. I don't know what the problem was. I just know the results were not good at all.

 

[Kali the Boston Terrier]

The easiest test on how accurate an AVM is, every time a property goes into pending status on the MLS...run an AVM on it. When the property finally sells, record the price adjacent...do this about 100 times over. Theoretically the AVM should hit the number dead on...reality is...well you know.

 

[Greenback]

I would have said the AVM was "misleading".

I was thinking the same thing. But, sometimes, misleading can lead to fraud. Or, maybe pressuring an Appraiser through misleading AVM reports can lead to fraud.

Or, misleading may equal an attempt to fraud?

I'm not a lawyer, but it sure can mess up a market...

 

[Caterina Platt]

An AVM itself is not fraudulent. It may be flawed to the core, but fraud involves intent to deceive and that's not accomplished by a machine.

Now, a developer of an AVM who knowingly provides inaccurate information can commit fraud, as can a person who would manipulate the input for skewed results. Either way, you have to have pretty good proof before you use the 'f' word. Better off to stick with words like inaccurate or unreliable.

 

[Bama Bayou]

An AVM itself is not fraudulent. It may be flawed to the core, but fraud involves intent to deceive and that's not accomplished by a machine.

Now, a developer of an AVM who knowingly provides inaccurate information can commit fraud, as can a person who would manipulate the input for skewed results. Either way, you have to have pretty good proof before you use the 'f' word. Better off to stick with words like inaccurate or unreliable.

My bold. The machine can't produce the fraud but the person behind it sure can. The Wizard of Oz would be a good comparison.

It's not likely manipulating the input that would be causing skewed results. It most likely would be manipulating the output. One line of code in the "model" could increase (or decrease) the values by any percentage you want.

 

[timd354]

I stated that information provided by the AVM was outright fradulent. Now the AMC wants me to remove the negative statement about the lender. Is this a negative jab at the lender or an unfortunate truth backed up by concrete evidence? Should I hold my ground or remove the statement?

Before you go kicking around accusing clients of fraud, maybe you ought to have actual evidence of fraud. The only thing the "evidence' you have cited is that you were provided with an AVM that used data that you discovered to be incorrect. You have no evidence as to why the data is incorrect and you have no evidence whatsoever that anyone with the lender or entity that issued the AVM is even aware of the incorrect data. In other words, you have absolutely no evidence whatsoever that a fraud has been perpetrated. You should remove your accusations of fruad immediately and just make a statement sticking to the facts that you know. Making unproven accusations of criminal conduct is unprofessional at best, and, like, is defamatory.

You should also hope that this incident does not result in you getting blacklisted by that client or that AMC. I know if I was a client or ran an AMC and you made an allegation that the client committed fraud with absolutely no evidence, I would absolutely never allow you to perform any work for my company again.

 

[Caterina Platt]

Well, in the case of Zillow, something sure is happening to the input.

Actually, what I'm meaning to state in the 'input' is probably more manipulating your search criteria as a user. Granted, I've not used an AVM so I can't tell you exactly how one might accomplish that. I do know that if I were a less than ethical appraiser, I could certainly manipulate my MLS search criteria to spit out some whopper one liner CMA's. I envision an AVM to be similar. Either you want the truth, or you want to play with your criteria to obscure the truth.

 

[Bama Bayou]

Caterina,

From a practical standpoint it is much easier to manipulate the output on an AVM than the input. For a national AVM it would take much work to manipulate the input. Doing that to the output would take about 15 seconds.