Freddie Mac: Quality & Condition Ratings

[George Hatch]

So what's their data standard or confidence interval for "accurate"? Let's see it.

As it stands, if a reviewer thinks the value is $400k and the appraisal comes in at $405k the reviewer's choice for answering that question ("is the appraised value accurate?") either requires the reviewer to hold their nose and say "yes" or touch off a pointless catfight with the appraiser by saying "no". That catfight doing NOTHING for the legitimate interests of the user. Everyone loses in that dispute.

The reviewer's situation is only tenable if the question is framed in terms other than "accurate". Like ("is the appraisal credible within the context of the intended use?"). The answer to that one is that both the reviewer's opinion and the appraiser's opinion can fairly meet that criterion. Without lying, and without rolling around in the mud over opinions which can never be right or wrong to begin with.

I use the review example to illustrate the point but it applies equally to the appraiser. How confident can an appraiser be in their result if they think they have to meet exactly the number a reviewer will come up with? It's no wonder so many appraisers are paranoid, or think there's no such thing as a workproduct that can be defended against the critics. The unreasonable expectation of "accurate" for analyses, opinions, conclusions becomes an albatross.

 

[Terrel L. Shields]

That's about all any buyer can see without a detailed technical inspection.

That's, of course, right. The buyer sees no more than we do, and we are being very subjective - more so with quality than condition. We can catalog all the defects we see; we can look in the attic (and should); we can look in the crawl space.

When I worked for a civil engineering lab, I tested concrete both in cracking cylinders of samples and in situ testing. I couldn't tell what the strength of the concrete was without testing. The cylinders all look alike. 4000 psi or 2200 psi. Any of the home inspector videos will show you a lot of defects such as uneven insulation blown into an attic, trusses that were improperly joined, broken rafters, etc. It would help our reports if a home inspection was provided to us. But even so, lots of things are hidden or not obvious. When have you actually asked or known what brand of windows are installed? Actually measured the thickness of the walls or sheetrock (if you could even figure out by removing the cover of an electrical outlet. Doors - six panel solid, hollow core, whatever. My builder was used to Lowe's quality fiberglass tubs, but I had insisted an old classmate's plumbing company be used although he was more expensive. Sure enough, my friend's fiberglass tubs were quarter inch thicker material, and the carpenters had to plane off the extra from the framing so the tub would fit.

We cannot see the subflooring usually. We can see if the cabinets are expensive or not usually, we know the difference in maple floors and laminate products, etc. So, we make a very subjective assessment with what limited information we have but truly being able to quantify "quality" seems a fool's errand. We're just making an educated guess.

 

[George Hatch]

Well, we CAN see what we see. As for the rest, that's what the standard assumptions and limitations are for.

 

[Terrel L. Shields]

Some of the hardest for me are the older architecturally significant or period houses from 1920 and such, restored not just renovated - it is really hard to get a handle on that - as far as cost - though it can cost more to restore those kind of details then building new in some cases

Well, replacement cost does not require reproduction costs. But it is an issue. My old GF lived in a huge old 1900s house that had solid walnut pocket doors into several rooms. But would we build pocket doors today? Probably not. The house also had the original slate roof. 9' tall windows in 12' tall rooms. Single pane double hung. No one builds such windows today commonly. Almost all old houses have so much built in obsolescences it's tough to even catalog same. The house was later remodeled when her brother bought her out and the roof replaced, AC and other items replaced, and the home updated. It sold very well. And was on a premium lot overlooking the downtown. But they buyers will have a money pit in my opinion. It's nigh impossible to insulate walls cheaply in such old houses. The wiring is suspect. The plumbing and most everything else is not any standard that is current (like 2 wire electric, 2 wires on insulators in the attic)- ancient wiring, nor are any doors sized with stock sizes, etc.

 

[George Hatch]

So what's their data standard or confidence interval for "accurate"? Let's see it.

And while we're at it, if Fannie ever really thought there's only one true and correct way to pick comps and analyze for adjustment factors then why don't they spell it out? Show us the one and only true and correct way to quant so we can just copy/paste and hit the EASY button. Distribute the black box to the appraisalware vendors so everyone can use methods and techniques that neither they nor any of their non-GSE users understands.

You'll love this: I recently herd that some REAL professional quants at one of the big boxes figured out that if they pick a dataset that brackets the subject's dominant attributes and simply average it that result *usually* gets them within reason for a lending decision. Often, but not always. You might say that mode of analysis gets within a reasonable range for the decision at hand.

Isn't that weird?
--------------
One more rhetorical question for the gallery:

If a given mode of analysis is only usable some of the time or only under certain conditions then how can it be characterized or unilaterally designated as a minimum standard for all or almost all of the assignments? ​

 

[J Grant]

So what's their data standard or confidence interval for "accurate"? Let's see it.

As it stands, if a reviewer thinks the value is $400k and the appraisal comes in at $405k the reviewer's choice for answering that question ("is the appraised value accurate?") either requires the reviewer to hold their nose and say "yes" or touch off a pointless catfight with the appraiser by saying "no". That catfight doing NOTHING for the legitimate interests of the user. Everyone loses in that dispute.

The reviewer's situation is only tenable if the question is framed in terms other than "accurate". Like ("is the appraisal credible within the context of the intended use?"). The answer to that one is that both the reviewer's opinion and the appraiser's opinion can fairly meet that criterion. Without lying, and without rolling around in the mud over opinions which can never be right or wrong to begin with.

I use the review example to illustrate the point but it applies equally to the appraiser. How confident can an appraiser be in their result if they think they have to meet exactly the number a reviewer will come up with? It's no wonder so many appraisers are paranoid, or think there's no such thing as a workproduct that can be defended against the critics. The unreasonable expectation of "accurate" for analyses, opinions, conclusions becomes an albatross.

A reviewer is supposed to be competent, and part of competence is understanding that the URAR standard of Accuracy for an OMV is at odds with USPAP and the limitations of what an appraisal is.

If a OMV in an appraisal of 400k is well supported, it would be inane for a reviewer to disagree because they think it should be 405k.

However, if the OMV of 400k was poorly supported and the review concludes that 375k is the well supported value, then a competent reviewer would disagree with the 400k .

 

[George Hatch]

A reviewer is supposed to be competent, and part of competence is understanding that the URAR standard of Accuracy for an OMV is at odds with USPAP and the limitations of what an appraisal is.

If a OMV in an appraisal of 400k is well supported, it would be inane for a reviewer to disagree because they think it should be 405k.

However, if the OMV of 400k was poorly supported and the review concludes that 375k is the well supported value, then a competent reviewer would disagree with the 400k .

The question is about the application of accurate, and more directly, how the GSEs define that. (obviously the term is also not defined in USPAP)

You know that user-driven requirements are add-ons and cannot replace or undermine our minimums, and in the example I am using it is the reviewer's conduct WRT accurate that we're talking about.

My read of the situation is that WRT the analyses, opinions and conclusions the GSE don't actually expect a literal application of accurate. Binary, go/no-go, red light/green light. Moreover, they know they can't get "accurate" most of the time. AFAICT they actually expect reasonable and sufficient to purpose, the same as any other user that is thinking about it.

All the noise they're making now about appraisers showing their work basically comes down to a simple
"Do SOMETHING, stop doing nothing and saying otherwise." They're still not telling you which methods to use. I disagree with their use of the one term mostly because I don't think they mean it exactly like that; not that they're acting in bad faith. Whatever else they're asking for seems within reason for those types of assignments.

 

[jay trotta]

Republicans have sought to privatize Fannie and Freddie for years, noting that the federal conservatorship was never intended to be permanent. They have also blamed these institutions, and their
mandate of making home loans more accessible to lower-income people, for causing the 2008 financial crisis in the first place. This is not true, as Fannie and Freddie's share of the highest-risk category of mortgages actually decreased during the housing bubble.

Who is in the "highest Risk" category, by their definition, is the question

 

[Mejappz]

Accuracy can indeed be measured. Here is an example of 57 sales I analyzed for an assignment last week. And an explanation. Where do you think the subject is going to land?

1. Both the median and average ratios are 1.00, meaning the appraised values and sale prices are pretty much identical when you look at the overall group.​

2. The COD (2.02) and COV (2.43%) are super low, which means most appraisals are very close to the sale prices. In real estate, anything under 10 is really good—this is much better than that.​

3. Almost all (98.25%) of the appraisals are within 5% of the sale prices. This shows the appraisals are super accurate for nearly all the sales.​

4. Whether the house is expensive or cheap, the appraisals are fair. There’s no favoritism toward one end of the price range.​

5. On average, the difference between the appraisal and the sale price is less than 2%. That’s incredibly close.​

6. The average sale price and appraised value are the same ($405,284), so the numbers line up perfectly when you look at the big picture.​

 

[George Hatch]

Can you do this for all of your assignments? Can local appraisers do this for all SFR assignments across the nation? Can you do this for non-SFRs? Do you have any clients who are capable of reviewing the analysis and develop their own opinions of the credibility of that analysis? Are any of them willing to pay you extra for doing this work?

IRL, nobody is stopping you from making the effort. If you can sell the extra charts and graphs for an additional $5 per assignment then I think that's terrific. Good for you. Live long and prosper.

What you seem to be annoyed about is that the IRL minimums are lower and that other appraisers are getting paid the same as you even though they've been getting away with doing less. And therein lies the rub - the expectations of the users are established by the users and demonstrated by what they're willing to buy. Not by the appraisers based upon what they want to sell. The lenders ask and the appraisers respond; not the lender do what the appraisers tell them to do.