Property Condition: The Future Is Cost-to-Cure

Fannie Mae introduced its quality and condition ratings as part of the Uniform Appraisal Dataset (UAD) initiative just over a decade ago, an improvement over the idiosyncratic ratings that existed before it. The updates to the UAD have left this scale essentially unchanged. Previously appraisers described property condition and quality in whatever way they wanted, often using phrases whose meaning is not obvious: My favorite examples are “average-average” and “average good.” What is average good, exactly? Is it the average of the good properties, or is it between good and average? Average-average should, if my math serves me correctly, be average, but it is possible the appraiser was trying to convey something that eluded me.

However, this new system was far from perfect despite the best efforts of a number of talented people: The different stakeholders could never entirely agree on a guiding principle. Its designers chose to use an even number of categories to eliminate the ubiquitous use of “average.” This would force appraisers to decide how the house stood compared to others. Having only an even number of categories would cause appraisers to get off the fence and definitively rank houses as good or bad, at least in principle. The reality is that appraisers simply adopted a new ‘average,’ the ubiquitous C3-Q4. (Technically, it is a mode).

Multi-modal distribution and AVM’s

The intent was to produce a multi-modal distribution allowing the GSEs to improve their AVMs. If you have a variable that has minimal dispersion, a model cannot leverage it. If 80% of homes are rated average, then the condition variable can affect property values significantly only for the remaining 20%. It is unable to rank homes within that entire 80% block. Instead, in part due to policy decisions made by Freddie Mac and some unfortunate Selling Guide definitions, they faced the same problem they had before. The meaning of the ratings was now clearer, but appraisers had not stopped describing almost all properties in exactly the same way. 

The UAD accidentally presented appraisers with only three useable categories for most homes: C2, C3, C4. Freddie Mac decided not to accept appraisals with C5 ratings, making appraisers fearful of pushback if they ever used it. Both Fannie and Freddie refused to take C6 properties, and C1 properties are simply new—a rating that only applies to sales and is redundant with year-built: So, we effectively presented the appraiser with a new average, C3. A brief but interesting aside, REO appraisals do use the C5 rating regularly. However, when we looked at the subsequent purchase appraisal, they almost always used C4 even when we showed that Fannie spent no money rehabilitating the property. This demonstrates that Freddie’s policy decision did little more than encourage appraisers to lie. 

Read more: A Brief Introduction to Automated Valuation Models

One solution to this problem

One solution to this problem is to present appraisers with a broader variety of condition ratings, providing, let’s say, ten options instead of six. This could work if Fannie adopts a rule allowing a price adjustment only when the condition rating of the subject and comparable differ. However, it is likely such a rule would cause appraisers to simply contradict themselves regarding a comparable’s condition from one appraisal to the next, essentially moving them back into rating condition relative to the subject. One of the goals of the UAD was to move appraisers away from reporting condition in relative terms. 

The UAD’s definitions are problematic because they lump different issues into one heap called ‘differed maintenance.’ Let’s consider The Fannie Mae Selling Guide’s definition of C4: “The improvements feature some minor deferred maintenance and physical deterioration due to normal wear and tear.” Of course, the value impact of different kinds of deferred maintenance varies widely. The definition asks the appraiser to give us a rating whose economic meaning is hazy. 

C1-C6 Scale

The C1-C6 scale, while an improvement over the anarchy that came before it, has not delivered all that was hoped. A novel approach that has clear economic meaning is needed. Such a method exists, cost-to-cure, but appraisers have strenuously resisted it. The cost-to-cure method looks at the amount of money necessary to return the property to marketable condition. Appraisers have resisted this approach because they lack a contractor’s expert knowledge. However, the C1-C6 scale also bears no direct relationship to market value.

The cost-to-cure figure can be used to gut-check the reasonableness of condition adjustments and provide AVMs with information that is easier to leverage. Some might object that this creates a problem for C1 (new) and, more importantly, C2 (like new) homes. However, new and like-new homes are a small portion of the housing stock. This problem can be handled using C2 as the reference point for cost-to-cure rather than C3. These numbers would still be more interpretable and easier to verify than the existing condition ratings. 

Another objection is that appraisers do not have enough information about the comparables to estimate a cost to cure. However, appraisers have always had more trouble assessing the comparables’ condition than the subject’s. Furthermore, the GSEs could quickly solve this problem by making the UAD data available to licensed appraisers—at least the data associated with sales. While some would complain about privacy concerns, appraisers are already subject to strict licensure. No appraiser would risk his livelihood to tease a home buyer about having bought a fixer-upper. Any recently sold house that has already been subject to a degree of public scrutiny. While other assignment types might have privacy concerns, sales don’t, and all comparables must be sales or listings by definition.  

Read more: Unlocking the Power of Large Language Models in Real Estate

Conclusion

The most strenuous objection is that appraisers lack the detailed knowledge to use cost-to-cure. Of course, I do not believe appraisers can determine the market value of a property’s condition without some understanding of how much it would cost to fix that property’s issues. However, Kukun’s cost estimator makes expert contractor knowledge readily available to the appraiser. This was a far better objection back when cost estimates required using clucky reference manuals that soon became outdated. Modern web applications allow the appraiser to keep up with construction and repair costs. This is no longer the obstacle it once was. 

A cost-based approach towards condition ratings offers much that the UAD initiative hoped to achieve: An objective definition that is machine interpretable and granular enough to help differentiate property condition among a whole host of properties that are “pretty much the same.” It would also force appraisers to account for their adjustments for condition in a detailed manner that will help appraisal reviewers. It offers a clear improvement over UAD condition ratings.

Franklin Carroll

Franklin Carroll is Kukun's VP of Modelling and Analytics; After finishing his studies at the University of Chicago, he worked at Fannie Mae where he helped build Collateral Underwriter, roll out the Value Verify initiative, and specialized in collateral related and credit related policy.

Published August 16, 2023