Automated valuation models (AVMs), computer-based systems which provide instant valuations on residential properties, are increasingly being recognised as the default valuation method for mortgage lenders, replacing both the time-consuming and expensive drive-bys and physical valuations.
Prior to the recent government change of heart on the introduction of a mandatory Home Condition Report (HCR) within a Home Information Pack (HIP) in June 2007, it was widely believed the use of AVMs would increase further, as the HCR would have enabled lenders to use the information in it to compliment the AVM valuation. However, irrespective of the final decision on the introduction of HCRs, we firmly believe the use of AVMs will continue to rise, becoming a much more common part of the mortgage process and bringing with it savings in time, effort and cost.
ASUB
An AVM works by drawing on comparable data from a database that is created from multiple sources, such as Land Registry data and surveyor valuations. The AVM uses a computer algorithm, usually exploiting other metrics and datasets to calculate a market value. However, the performance of different models will vary significantly. For lenders to use an AVM as a truly effective risk management tool, it must provide maximum coverage and accuracy, ensure good predictability and guarantee full reliability during implementation.
Coverage versus accuracy
The coverage and accuracy of an AVM will ultimately determine the financial benefits that will stem from its implementation.
With some systems there is a trade-off between coverage and accuracy and a good system will provide the optimal, most flexible balance between the two. Therefore, to successfully compare the performance of different AVMs, a mortgage lender must not only look at the ability to value as wide a geographic area as possible, but also at the number of cases that achieve the desired accuracy, hence ultimately the lender needs to consider the proportion of high confidence results.
Naturally, there will always be a very small number of properties where the valuation cannot be derived due to the lack or extreme diversity of any comparable properties, but in a good system this will be minimal. Also, a good AVM is a living, evolving model, and the best providers will constantly be fine-tuning their systems, increasing coverage, adding data, developing the algorithms and ensuring the most accurate system is available to the customer at all times.
Predictability
The predictability of an AVM allows a mortgage lender to differentiate between the acceptable valuations and the least accurate valuations, possibly rejecting some of the latter in certain situations based on their specific risk appetite.
Good AVMs are calibrated so accuracy can be consistently predicted in terms of Forecast Standard Deviations (FSDs) – these are then translated into confidence level bandings, i.e. for Hometrack, 0-7; 0 being a low confidence valuation and seven a high one. This ultimately gives the user a statistical confidence as to the accuracy of a valuation and enables them to determine an appropriate level of risk against which particular business actions take place.
Reliability
For anyone using an AVM, reliability must be a key criterion too. The reliability of the AVM will, in the long run, affect the operational cost, and it is the speed and convenience of a system’s deployment and operation that assures the use is seamless with other lending processes at origination and in portfolio revaluation.
Sub-second response times, along with secure and highly flexible interfaces, can ensure the user minimises their technology installation and deployment costs. Whether it be regular portfolio revaluation, or point-of-sale mortgage provision, the form and efficiency of the service is crucial to its success.
In conclusion, each user of an AVM will have different requirements for a system, but there are certain key fundamentals that in any scenario are crucial. A good AVM provides accurate, immediate and objective property valuations for all properties in a given geographic area, with each valuation having a statistically predictable level of accuracy. It is down to the mortgage lender’s diligence in evaluating these criteria and assessing each provider’s commitment to professional competence, to determine which system should be implemented within their organisation.
