AVMs in Declining Markets

As house prices continue to stabilize in many of the nation’s housing markets, industry observers have expressed concerns about the predictive accuracy of automated valuation models (AVMs), particularly in areas where house prices are falling.

• Critics have expressed concerns that some AVMs may overvalue properties in cooling markets because of ‘data lag,’ or the interval between when a sales transaction occurs and the model receives this information in its backend database. According to these detractors, outdated sales prices will result in higher value estimates that do not reflect current market realities.
• Further complicating matters, there is a common misconception that AVMs struggle to ‘see over the hill’ and are restricted to making linear adjustments to update prices and counteract data lag.
• Finally, critics argue that sales concessions occur more frequently in declining markets, compromising sales data used by AVMs and increasing the probability of overvaluation. 

This article will examine the validity of these claims as well as whether traditional appraisals are necessarily a better option in areas afflicted by depreciation.

AVMs calculate a property’s market value based partially on market information contained in public records. Several larger counties deliver this data through online distribution channels as soon as the transaction is recorded. However, this timeliness comes with a price; online data is usually more expensive than alternative sources and delivery methods. In other areas, AVMs receive this information monthly on a CD Rom. In counties where sales transactions occur less frequently, an AVM developer might receive data on a quarterly basis.

In general, AVMs have online access to current market intelligence in high frequency sale transaction areas. AVM vendors receive thousands of valuation requests every month; hence the extra cost of online data per order is negligible. On the other hand, appraisers do not have the scale to invest relatively significant amounts of money to acquire the most current data. In counties where vendors receive sales information on a monthly basis, appraisers and AVMs are on par in terms of access to current data.

Appraisers do have an advantage in those counties where data is reported quarterly; however, in these areas there are not many sale transactions in the first place.

Click here to continue reading . . .

Overall, AVMs actually have a slight advantage over appraisers in terms of access to current data. The time lag in the underlying data collection process in large markets is not usually more than a month. Although the time lag in low volume markets can be as high as 3 months, AVM developers remedy this problem in several ways to ensure valuation quality.

Critics of AVMs frequently assume that these models can only compensate for data lag by making linear, time-based adjustments to bring older sales data up to date. In reality, the best AVMs also consider a number of macroeconomic variables when estimating a subject property’s market value. Even though real estate transacts in highly localized markets, regional and national factors still exert a meaningful influence on house prices.

For example, after real estate prices peaked in coastal CA, people who sold their properties moved inland; hence the prices in other parts of CA started to increase. After a while, this phenomenon spread to neighboring states like AZ and NV. What stopped the increase in real estate prices was the increase in interest rates, which affected homebuyers in every state.

AVM developers employ statisticians and econometricians who are experts in building sophisticated models to analyze the affects of local, regional and national factors on real estate prices. Few if any appraisers have the training and knowledge to identify these spatial autocorrelations in real estate prices; these advanced estimation methodologies are essential to accurately predicting the path of home prices.It is a misconception that AVMs cannot see over the hill and naively employ simple linear regression. On the contrary, AVMs use cutting edge analyses to see over the hill and beyond.

The last concern about the accuracy of AVMs is data related. In declining markets sales concessions are more common, potentially resulting in overvaluation. This is a serious issue not only for AVM developers but also for appraisers. However, vendors have access to vast quantities of data, sophisticated statistical software and enormous computing power. AVM developers can quickly and accurately analyze data and eliminate outliers as soon as a model receives an update.

Now that we have addressed several common misconceptions about AVM performance in declining markets, we will focus on how the housing downturn affects the accuracy of traditional appraisals.

Have you ever thought about why the number of transactions goes down and inventories go up in declining markets? There is a disconnect between market prices and actual listing prices. Homeowners and Realtors, the people who are supposed to have the ‘local knowledge,’ are slow to revise their asking prices downwards to reflect current market realities.

This has two consequences:

• first, the number of sales transactions declines significantly;
• second, some buyers (not many) will pay the inflated prices.

These buyers need mortgages to finance their purchases. This is where appraisers come into the picture. October Research surveyed 500 appraisers in 2003. Their survey indicated that 55 percent of respondents said they had experienced some form of inappropriate pressure by mortgage brokers to come up with ‘the right value’ to close the deal.

In 2007 they repeated the same survey with 1200 respondents. As expected, the picture got bleaker under current market conditions; 90 percent of the appraisers reported feeling pressured to restate, adjust or change values. Those appraisers who refused to give in to coercion also paid a steep price; 68 percent reported losing a client and 45 percent did not receive payment for their work.   

These are widespread, unethical practices instigated by unscrupulous loan officers and mortgage brokers.  In many quarters, automated valuation models are regarded as the most effective way for responsible lenders and investors to combat inflated property valuations. The inherent objectivity of AVM-based value estimates has prompted industry stakeholders to use these solutions extensively for quality control purposes.

Wholesale lenders frequently employ this technology to double-check the property values for loans coming through their pipelines, while investors rely on the batch-processing capabilities of AVMs to identify questionable appraisals. In situations where valuations are likely to be inflated, many lenders actually prefer an AVM’s valuations to traditional appraisals.

AVMs are econometric models that continue to evolve and become more accurate. They use innovative technology to overcome data problems and advanced statistics to estimate future path of home prices accurately. A careful analysis of AVM performance will reveal these tools’ ability to consistently follow market trends both in declining and increasing markets.

Still, although AVMs are a valid option in all market conditions, these models may not be suitable for all lending situations. Decision-makers should evaluate the different approaches to property valuation and identify which methods are appropriate given the risk in a particular transaction. Likewise, the accuracy of individual AVMs will vary by property type, price tier and geographic area; lenders must consider each model’s unique strengths and weaknesses to determine which AVMs match their business objectives and tolerance for risk.

AUTHOR: James A. Kirchmeyer, CMO - Real Info, A division of Zaio, Inc., 40 Gardenville Pkwy, Ste 100, Buffalo, NY 14224. Email: mailto:jakirch@real-info.com