The Evolution of the AVM William E. King Veros Real Estate Solutions Director of Valuation Initiatives AVMs as we know them today were introduced in the 1990s, but the birth of the computer-generated valuation can be traced to the mid-1960s when computers were first used to assist in creating market value appraisals for tax assessment purposes in Michigan. By the early 1980s, computer-assisted mass appraisals (CAMA) systems had fully integrated the cost, income and sales comparison approaches to value and were ubiquitous in assessment offices across the U.S. These models focused on generating values for a universe of properties all at once, rather than attempting to precisely value a given property by itself. Property-specific, computer-generated valuations were introduced to the market in the early 1990s. These early models were primarily price-index models for looking at repeat sales to measure historic price change to calculate current value and predict future values. Although property specific, these AVM s were primarily used for analysis and prediction at the portfolio level rather than for individual property and/or mortgage decisions. By the mid-1990s, AVMs were being used at the individual property level by mortgage lenders. The AVM quickly became a useful tool for establishing a starting point for value at the point of loan application. Soon, better decisions could be made about which loan program might best suit a certain situation or whether mortgage insurance might be required. AVMs provided an alternative to waiting for appraisal results and quickly determined whether the originally contemplated loan was achievable by comparing against automated value expectations. By the late 1990 s, second generation AVM models were in their infancy. These second generation models included hedonic models that broke a property into its component characteristics and through regression analysis determined the contributory value of each part. The component values were then reassembled into a whole property value. Second generation AVMs often meld several methodologies using weighting and optimization techniques to produce more refined values. These models also incorporate the concept of a confidence score which rates the accuracy of the model s valuation. While the methodologies behind how each AVM provider generated their score varied, the inclusion of a rating system was an important step in the evolution of the tool and a key differentiator among first- and second-generation models. February 8, 2012 Page 1 of 5
Considerable benefits generated by second generation AVMs came in the form of objectivity and costefficiency, however, controversy arose around the issues of abandonment of traditional appraisals. Many appraisers feared that appraisals would be replaced because of the time and cost-efficiency of AVMs. As AVMs took hold throughout the late 1990s, the industry moved toward a silo approach by using specific types of valuations as an exclusive solution for specific needs. However, traditional appraisals remained the preferred valuation tool for most mortgage finance transactions. Market Turn Impact on AVMs The early years of the millennium were marked by an extended period of rising prices, falling interest rates and relaxed credit policies, all of which created an environment conducive to loan churning. Many home purchase decisions were speculative rather than shelter-based. Whether for purchase or refinance, most transactions required an appraisal. However, prices were rising and risk was perceived as low, consequently, little attention was paid to the quality of the appraisal. Following the market crash in 2008, the industry experienced a renewed focus on valuation quality, and most regulatory agencies issued new requirements and updated existing guidelines 1 emphasizing due diligence and accountability in valuation. In an effort to add greater transparency to mortgage transactions, the Federal Financial Institutions Examination Council (FFIEC) revised The Interagency Appraisal and Evaluation Guidelines (Interagency Guidelines) and gave more detailed information on how agencies expect lenders to deal with appraisals, broker price opinions (BPOs) and AVMs. For the AVM, the Interagency Guidelines provides necessary insight into model validations and guidance on when an AVM can be used as an evaluation tool. The Interagency Guidelines recognize AVMs as acceptable and useful tools to value collateral under circumstances where documenting current property condition is neither required nor practical. For example, the use of an AVM is an acceptable valuation technique for most loan modifications and for monitoring portfolios where there has been no deterioration in credit or change in market conditions. Additionally, AVMs may be used to monitor valuation trends on portfolio risk and as a tool for appraisal or evaluation reviews (which the Interagency Guidelines also establish as a standard of practice). The Dodd-Frank Wall Street Reform and Consumer Protection Act (Dodd-Frank) required AVMs to integrate quality control standards for a higher level of credible AVM estimates, guard against data manipulation and conflicting points of view, as well as require sample testing. Consequently, both the focus and use of AVMs continue to evolve in the ever-fluctuating housing market. In the post Dodd-Frank era, many lenders have elected to procure full appraisals in situations where they would previously have judged an AVM or drive-by type appraisal sufficient. While it is agreed in the realm 1 OCC Bulletin 2000-16, Supervisory Guidance on Model Risk Management was replaced with OCC Bulletin 2011-12 in April of 2011. February 8, 2012 Page 2 of 5
of valuation that more is almost always better than less in terms of detail, over-purchasing of valuation services as a singular solution to regulatory compliance may not be prudent. What Is The Best Reaction? Is an AVM the right tool for every job? No, it is not. It wasn t the right tool for every job pre-meltdown either. A valuation tool, AVM or otherwise, must provide an appropriate amount of detail to support the valuation decision. An AVM is often most appropriately used in the following circumstances: When the frequency and/or volume of the valuation(s) needed makes a full appraisal impractical. When valuation baseline is needed, recognizing that the decision-maker can escalate to higher levels of precision and investigation through supplemental valuation products or services. When maintenance and monitoring is necessary for an instant benchmark between two major decisions. For example, it can be used between the point of origination and the point of sale to an investor. The balancing act, when the decision is still unclear, lies within the degree to which one must rely on the valuation relative to other loan-level information such as credit. Although the terms accuracy and reliability have been used interchangeably by users of valuation services and valuation providers alike, they are distinct terms with unique and specific meanings. Accuracy implies the existence of a specific value for a given property on a given day that is right, rendering any other valuation wrong and denoting a degree of precision that simply is not achievable. Reliability infers the ability to be trusted and consistent quality in performance. The degree of valuation precision expected or required depends on many factors. For instance, why is the valuation needed? Who needs the valuation? What is the valuation being used for? How quickly is the answer needed? Where is the property? How is the value being developed? In short, how close is close enough to meet the needs of a given situation? As any seller who has had multiple offers for his or her property would agree, the odds of two buyers simultaneously assigning the exact same value to any property is remote. That said, the concept of market value by its very definition recognizes that there is a most probable price that a group member of active buyers would pay for a property on a given day. The question is, when do two or more different numbers become equivalent under the concept of most probable or how close is close enough? The answer: it depends. February 8, 2012 Page 3 of 5
Appraisals and BPOs often provide more detail than an AVM, but because people are involved, the possibility of bias is always present. AVMs are truly detached and independent. They don t know anyone or anything about the contemplated transaction. Most users of valuation services want not only independence and objectivity, but they also want to be able to trust that the value provided is appropriate for the decision at hand. A home equity loan on a free and clear property in an appreciating market presents remarkably different risk compared with a 97 percent LTV purchase decision in a depreciating market. It is critical in today s regulatory environment to clearly understand the level of valuation detail required for each transaction in order to demonstrate compliance and effectively mitigate valuation risk. The onus for the valuation product or method chosen ultimately rests with the lender, so it s imperative that reasonable and informed valuation decisions are being made and over-reactions on either side of the spectrum are avoided in order to prevent a repeat of past industry missteps. AVMs Today and in the Future Today we see AVM providers spanning the divide between a mature second-generation and the beginnings of the third-generation with improved data fueling its development. In addition to advancements in data (type and quality), AVM models have become more complex and have a greater capacity to incorporate trends relative to value estimates. By virtue of their instant availability, data-centricity and flexibility, AVMs can be efficiently paired with related analytics that provide another important element in the valuation equation and move lenders out of a silo-mindset. For example, significant benefits of AVM pairing can result when: An AVM is not independently sufficient to demonstrate current condition. It can be bundled with a condition inspection report before stepping up to a BPO and incurring greater costs. There is concern about default. An AVM can be bundled with a risk detection analytic that measures probability for foreclosures in a certain neighborhood, or with credit data to identify borrowers at risk of going into default. A decision must be made as to how long to hold onto a property or portfolio. An AVM can be paired with a market forecast and look at the market several months down the road. Rightfully, with improved data and more complex analytics, AVM users are coming to expect clearer and more reliable estimates of value than in years past, as well as the flexibility to pair with other data requirements and integrate into higher functioning valuation management systems. February 8, 2012 Page 4 of 5
AVM cascades provide yet more options and greater opportunity to customize for specific needs. It is also important to note the changes in perspective surrounding AVMs, namely their role relative to other valuation types and the regulation surrounding them, both of which are continuing to evolve and are producing a notable shift in the AVM space. Beyond this aspect of increased compatibility, the momentum created from the Interagency Guidelines and Dodd-Frank has added an emphasis on due diligence and AVM testing that is unlikely to be reversed. Much like the appraisal community experienced a renaissance of regulation which resulted in the formation of professional associations, codes of ethics, etc., AVM providers have all come to the table with new levels of due diligence testing and compliance in order to stay competitive in today s regulated environment. Cooperation and speed are essential to recovery in the real estate and finance markets. Never before has the industry seen so many typically distant entities working together for the common purpose of accurately measuring collateral risk. Appraisals, BPOs, and AVMs need to co-exist in order to support the varied needs of today s mortgage industry. Valuation products have become increasingly portable, standardized and data-centric. An AVM can provide valuation for a high volume of properties with an almost instant turn around, whereas an appraiser can take specific property characteristics and circumstances into account. No valuation method is perfect, but they all have application in various situations and each is credible when appropriately matched to a given set of needs. Technology leaders will keep expanding the bounds of what data and models are able to accomplish. Savvy lenders, servicers and investors will continue to be watchful of their service providers and their respective outputs. February 8, 2012 Page 5 of 5