Property Valuation: Drivers of change
The financial crises and the challenging economic climate that followed heralded a period of regulatory change designed to restore financial stability and rebuild consumer confidence in the financial services industry. These changes have also had significant implications for the mortgage lending business, a bedrock of financial services in most developed and developing economies.
Mortgage lending depends on an accurate, transparent, efficient method of property valuation to provide a low-risk, low cost financial service to its consumers be they house buyers or property portfolio investors. The inherent risks associated with mortgage lending can only be managed by ensuring the underlying asset – property – is accurately valued.
The European Mortgage Federation’s (EMF) 2017 review recommends a broad set of risk related criteria underpin a robust property valuation including market, construction, location, fiscal and legal risk as well as an assessment of tenants. This can be a time-consuming, expensive and of course subjective process given valuations have historically been carried out by surveyors who combine a first-hand, physical assessment of the property with secondary information such as census data with socio-economic indicators (crime, health, education, transport).
Enter: Automated Valuation Models (AVM)
Introduced in the US in the 1980s and the UK in the 1990s, AVMs are machine learning platforms that use powerful algorithms to analyse large amounts of comparable property data –including historical property prices, socio-economic statistics and other commercial indicators that can affect property valuations – to value a certain property at a certain point in time. The underlying data are continually updated to ensure valuations remain current and accurate within a range usually provided by the vendor.
AVMs were developed as an alternative to the traditional labour-intensive, full valuation process which is expensive and time-consuming. AVMs are cost-effective and can significantly speed up the valuation process. For instance, completing a valuation at point of sale would effectively telescope the application to completion process to a fraction of the time it currently takes.
AVMs: Trends in the UK market
A recent paper by the European Mortgage Federation (EMF) indicates a consistent upward trend in AVM usage in the UK since 2008. Used primarily for mortgage origination for remortgages or when providing additional lending, they create a virtuous cycle for this segment of the market by expanding the data set of comparable valuations and therefore providing greater insight on the dynamics of property prices.
The UK market is dominated by Hometrack which among others is used by TSB and Foundation Home Loans. Short-term property lenders Precise and Shawbrook now use AVMs for both remortgages and purchase mortgages; in 2017, Precise increased the maximum property value for an AVM appraisal from £0.5m to £1m.
AVMs currently account for around a third of purchase mortgages. The UK’s top (and many smaller) banks regularly use AVMs for portfolio valuation with one AVM provider currently responsible for half of all UK mortgage portfolio valuations. AVMs are also commonly used as a forensic tool.
Valuation: Art or Science?
With bigger data sets; superfast computers and smarter self-learning algorithms, the art of property valuation is evolving. Has it turned into a science?
Property markets are not a simple supply and demand equation. Buffeted by winds bringing unexpected gifts – market volatility, Brexit, Panamagate, price shocks, economic sanctions – they are messy and present a multitude of risks that must be managed and mitigated.
This is easier to do around the middle of the market where there are large data sets of comparable properties to work with, LTV’s are low and publicly available data includes purchase prices, marketing prices and surveyor opinions. So AVMs work well in densely packed urban areas characterised by conventional property types.
It is much harder when LTV’s are high; there are fewer comparable properties such as happens in rural areas; or the area is dominated by unusual or unorthodox property types. Then the science of an AVM is best complemented by the art of the appraiser’s instinct.
A Niche Opportunity: HouseMartin
A recent study revealed the considerable time and resource challenges faced by small and medium-sized estate agents operating without access to an AVM. In the absence of granular data, historical prices, and broad trends, estate agents and indeed their customers generally use a crude proxy – desk research. Estate agents use simple valuation tools provided by banks and large real estate companies such as zoopla, mouseprice, rightmove to generate price estimates. These are cross-referenced with property prices for listings on the latter’s websites and used to generate valuation estimates for their customers.
Often, estate agents’ estimates are contested by their customers especially those with considerable local knowledge of education, health, transport services and amenities in their area.
Strasys recognised that an effective, customer-focused platform needs to be able to marry these discrete pieces of information – aggregate historical price data, socio-economic market trends and other commercial indicators – with customer data and local area knowledge.
We have developed a proprietary machine algorithm that estimates current property values by combining customer-generated local area knowledge, publicly available government data on historical (number of years?) house valuation and sales prices.
The beta release provides fast and accurate – within 3% of sold market prices – predictions for property prices in England and Wales. This will improve as the data available to the algorithm grows and its learning advances. It also gives 24 month forecast at the property level alongside relevant market data for buyers and sellers.
Although it is aimed primarily at UK estate agencies, the application can be equally relevant to mortgage lenders and large property investors looking for a more holistic approach to property valuation. It can also be used by surveyors as a ‘first port of call’ in the due diligence process for a complex valuation.
In the refinancing sector, HouseMartin can be a fast and simple property valuation solution for both homeowners and investors looking to realise value or take advantage of changes in interest rates to plan their strategy – be it selling, remortgaging or investing.
The current version of HouseMartin is ready to go to market with scalable infrastructure.
Going forward, we have a defined product development roadmap to improve the quality of prediction so that the algorithm can parse and contextualise current socio-economic indicators (education, health, transport services); changes to greenbelt and land-use policy; as well as future developments that could impact property value such as Brexit, immigration trends, social housing, and regulatory changes to pensions, inheritance laws and care funding mimicking hedonic models- more efficiently.
Further functionality, such as dynamic forecasting and reporting capability, will be added to enable lenders to use it as a tool to monitor LTVs over the term of a loan which could reduce the need for revaluations – and also serve as a trigger for revaluations.
We are also developing a Rental Valuation Model (RVM) for the rental sector which is being developed to serve the needs of both landlords and tenants.
The time from application to conversion depends on a whole host of disparate factors and can therefore vary significantly across applications. Future versions will include a ‘Time to Sell’ capability to enable all parties to the transaction to get an estimate of how long a particular property sale might take.
Having worked in North America, Europe and the Middle East, Sami comes to Strasys with over 15 years of strategic and operational expertise in business transformation, customer experience and product development. At Strasys he leads our digital transformation practice helping organisations take advantage of the AI revolution.
- Downie M L and Robson G (2008) ‘Automated valuation models: an international perspective’.
- Mortage Strategy (2007). ‘Fitch ratings reports increase in AVM use’.
- Standard and Poors (2011). ‘U.K. RMBS Methodology And Assumptions’.
- Wikipedia (2018). ‘Automated valuation model’.
- EMF-ECBC (2017). ‘Study on the valuation of the property for lending purposes’.
- Mortgage Strategy (2017). ‘Head 2 Head: Are AVMs the most reliable method of valuing’.
- Todd S (2017). ‘What are the benefits of using AVMs?’
- Davidson S (2016). ‘Lenders likely to rely on AVMs increasingly’.
- Gaylor R, Debashree S, Pugh R, King S (2015). ‘Best Practice Validation and Comparison for Automated Valuation Models’.
- European Mortgage Federation (2016). ‘EMF and EAA map features and applications of automated valuation models’.
- Matysiak G (2017). ‘Automated Valuation Models (AVMs): a brave new world?’