by Tom Ruff of The Information Market
Ask three different agents what’s happening with home prices and you will get three different opinions. Oddly, they might all be correct. Prices have heavy upward pressure below $250,000, are moderate in the mid-ranges and are slow to negative in the luxury market. High demand and limited supply in the lower price ranges have pushed up the median sales price.
Turning to sales volume we have a bit of a conundrum. At the beginning of the year we changed our statistics methodology. We no longer include sales with zero cumulative days on market in our re-porting. Some Subscribers use the MLS to log their non-MLS sales by listing and immediately closing them. They report over as having a DOM of 0. Since the sales didn’t occur on the MLS, we feel removing these sales from our reports gives a more accurate representative of the MLS sales market. Note: reports in Flexmls remain the same, this change only affects the reports we publish on our statistics section of ARMLS.com.
The outcome of this change is seen when we compared sales numbers year-over-year in our charts. Around 200 faux MLS sales in the high season each month will not appear due to the new methodology, not because of the market. Compounding things is the fact that February 2016 also had an extra day this year.
Our headaches continued this month when we read about a study done by a Florida Gulf Coast University as reported by TheRealDaily. http://therealdaily.com/big-data/mls-errors-study/
The author addressed the question, how serious are the errors in the MLS sales price data? Quoting the article, “For years, appraiser, economists, and other experts have quietly questioned the discrepancies between multiple listing services (MLS) prices and the legal prices recorded on HUD-1 forms (and filed in local jurisdictions.)” The article is based on a study by Marcus T. Allen, PhD, Kenneth M. Lusht, PhD, MAI, SRA, and H. Shelton Weeks, PhD.
The study looked at 400 sales transactions over a period of four years in a single southeastern market. From this small sampling Lusht maintains that the problem of MLS errors is longstanding and certainly not limited to one or two markets.
“I would say that any appraiser who has a reasonable amount of experience already knew there were errors in the MLS. They knew that it is very possible some of the prices weren’t right, but no one has ever measured it before and found how frequent the errors were or how big they might be.” — Lusht
From their evidence, the authors suggested that Brokers may be gaming the system by intentionally inflating sold price information in the MLS, perhaps to make it appear that they negotiated a higher price for their clients.
“Whatever the cause of the errors”, the study concluded, “Regardless of the motivation or source of the error, the result is the same – a misstated price. Maybe someone else will get interested in it and look at different areas to see if what we found is typical or not. My guess is that what we found here, we will find in other markets, especially during down markets.”
We will take Dr. Lusht up on his challenge and evaluate MLS sale prices within Maricopa County. One of the obstacles mentioned in the article is the long delay for local government to post official sales price data. This is not the case in Maricopa County. Thanks to ARS § 11-1133, we have no such issues. When a home closes the county recorder will make the affidavit of value available for viewing within minutes. These closings are reported in the Monsoon tax system be-fore the start of the next business day. Quick links within Monsoon allow users to quickly view and verify the recording in less than 24 hours of the close of escrow.
For the purposes of our study we viewed the last 100,000 MLS closings in Maricopa County and compared the MLS sold price to the re-orded affidavit of value where the recording was not an A3 exemption or a Service Link transaction. Our sample size was exponentially larger than the 400 sales included in Lusht’s study. The reason A3 transactions were removed is because government transactions (HUD/VA) do not require an affidavit of value and Service Link trans-actions, where bank sales were handled by an out-of-state entity are obviously not familiar with ARS § 11-1133 (there were 233 A3 sales and 176 sales handled by Service Link in the past year). So what does our analysis tell us?:
94.567% of all sales as reported on the MLS precisely matched their affidavit of value. Of the 5,433 sales where the price varied, 1,934 had an MLS price that was lower and 3,499 had an MLS price that was higher.
Of the sales where the MLS price was higher: 1,204 sales were within 1% of the affidavit of value. 793 were between 1% and 2% 505 were between 2% and 3% 316 were between 3% and 4% 280 were between 4% and 5% 401 were greater than 5%
Of the sales where the MLS price was lower: 1853 sales were within 1% of the affidavit of value. 61 were between 1% and 2% 11 were between 2% and 3% 4 were between 3% and 4% 2 were between 4% and 5% 3 were greater than 5%
To quickly summarize, 94.567% of MLS prices were precise while 97.624% were within 1% and 98.478% were within 2%. Here are a few examples of the types of issues we saw when we viewed the small percentage of extreme differences:
MLS sold price was transposed, sale for $25,000 instead of $52,000 (These types of errors are caught by data integrity and corrected, appearing correctly in MLS today).
MLS sold price entered as $360,000 instead of $260,000.
Data processing errors matching the correct MLS sale to the correct affidavit of value where there were issues with improperly reported parcel numbers or where more than one sale took place on the same property and the MLS sale was matched to the incorrect transaction.
The original affidavit of value was incorrect and a subsequent correction was filed after the initial matching process.
Two affidavits were recorded on same transaction showing half of the sale price on each.
MLS price is correct, affidavit was filled out incorrectly.
New construction where parcels reported do not match the actual property sold, new construction accounts for 3.93 % of the sales and 8.37% of the differences between MLS price and the price given on the affidavit.
The study done at Florida Gulf Coast University drew some very broad conclusions based on a very small sample. I cannot argue with their math but I will question the apparent assumptions they made. Data is messy, read by machines and often input by humans. Humans make mistakes. When comparing two different data sources you have to be extremely cautious. I’m not comfortable with these assumptions that appear to have been made:
Anytime there was a difference between the HUD-1 data and the MLS they assumed the HUD-1 data was correct.
An assumption was made the correct HUD-1 was matched to the correct MLS sale.
An assumption was made that the assessor tax number and ad-dress were reported correctly in both sources. In their report they state the two data sources were matched by the assessor ID and
address. There is no mention of a third matching criteria. In our study we saw the difficulty in matching MLS data with the public record, particularly when a property is flipped on the same day or in a very short period of time and the second sale is at a much higher price than the first.
The one conclusion they made in the report on which I agree is the indisputable bias of more errors above the MLS sold price than below. We showed a difference of 3.499% where prices were over re-ported in the MLS while 1.934% where the prices were under. It’s very likely that the price pended in the MLS might have changed at the final negotiating table and the agent never went back and changed the data. It might also suggest that some gaming is taking place within the MLS but nowhere near the extent implied in the article. The issues we have in Maricopa County are a far cry from those stated in the study.
It is vital to understand the importance of reliable MLS and public records data, as well as their weaknesses. The depth and richness of MLS data makes it far more attractive to appraisers than public records data, nonetheless, the prices listed should be confirmed by viewing the affidavit of value. And finally, if you see errors in the public records data or the MLS, report them. They can be fixed making everyone’s job easier.
ARMLS Pending Price Index (PPI)
Our last Pending Price Index projected a February median price of $209,900 with the actual median coming in at $214,000, off by 1.92%. Sales volume in February as reported by ARMLS was 5,718 with 482 fewer sales than our projected volume of 6,200. Looking ahead to March, the ARMLS Pending Price Index projects a median sales price of $215,000. We begin February with 7,222 pending and 4,169 UCB listings giving us a total of 11,391 residential listings practically under contract. This compares to 10,502 of the same type of listings at this time last year. We expect sales volume in March to be very similar to the numbers last year with a slight increase in the median sales price. February was not a great month for projections as I still remain single and unengaged.
Posted on April 22, 2016 at 5:24 pm by Tamara Weber