Consumers of UK house price information are spoilt for choice, or are they?

The wide range of options for understanding national or regional house price movements does not always bring clarity to the true position of UK house prices. The long-established traditional indices used in credit risk – those provided by Halifax and Nationwide – have been joined more recently by measures from the UK HPI index produced by the Office for National Statistics (ONS) and Rightmove. Automated Valuation Models (AVM) like the one produced by 4most offer another avenue to valuing property. What are the key differences between the best-known measures of house prices? And what are the main advantages and disadvantages?

Summary table of HPI characteristics

 
 

Because Rightmove prices are based on asking prices, this information is available almost immediately. While the index may capture the trend adequately, there are clear drawbacks. The average time to sale is currently 82 days in London and most properties are likely to be sold below the asking price. However, the sample is large, something that is not true of the Nationwide and Halifax indices which are both based on mortgage approvals (but not Buy-to-Let). Again, the price could change between approval and completion, but these indices are quite timely; the Nationwide index is typically published on the first working day of the following month.

The most accurate view of actual prices paid is obtained from Land Registry records – these form the basis of the ONS UK HPI index and 4most’s AVM and regional forecasts. This has many advantages. For one, it includes cash buyers which has been a growing part of the market in recent years. In London, the prices paid by cash buyers have been falling faster than those using mortgages. This may speak to prices behaving differently in certain market segments, but it emphasises that those wishing to more accurately value mortgage portfolios may want to look at more granular developments in the market.

 
 
 
 

Although the indices paint a picture of a slowing housing market, there is no consensus on how much the market is easing. Some of this variability will reflect methodological differences. Although the ONS index is based on observing pairs of sales over time, the Nationwide and Halifax indices represent the price of a house with a typical set of characteristics. The Nationwide update this view regularly but the Halifax outlook is based on what was typical in 1983.

The ONS measure is now the Bank of England’s preferred measure because it covers the entire market and is less volatile than the others. The latter issue – which reflects the use of 1983 characteristics to define house prices – has made movements in the Halifax index difficult to interpret and recent months have seen wild swings in monthly prices. This volatility is very unwelcome for those organisations using the index to value their portfolios and is likely to be the main reason driving the decision by Lloyds Banking Group and Markit (the index owner) to redevelop it. The results of that redevelopment are now available to subscribers to the index, whilst the old methodology will be run in parallel until at least 1 July 2020.

Changes of this type can have big implications for banks given the myriad of uses of house prices. For example, the restatement of the Halifax index will rewrite history and what we might think the downturn used to calculate capital requirements looked like. If the revisited Halifax index shows a say 20% fall in prices in the 1990s recession rather than the 12.7% fall in the current index, that could have implications on the amount of capital firms would have to hold in the event of another downturn.

 
 

Clearly there are a variety of options for firms that use house price indices. How will price information be used in the future - our view is that analysis will become more granular. The ‘old’ indices describe the big picture but by using the underlying data, it should be possible to look more closely at a granular level, both regionally and by property type. The 4most Automated Valuation Model uses the full Land Registry dataset and is complemented by locational (distance to stations, schools etc) and regional (census and net migration) statistics.

Benefits of using the land registry data are many. It is publicly available and updated monthly with a 2-month lag. This allows for an up to date view of a given property’s current value as well as the latest metrics of how property is performing in different regions. Regardless of which method is used to value a property, the price paid will implicitly incorporate the characteristics of the house and the region, making it a powerful tool in calculating an accurate value when harnessed correctly, and one that should perform consistently through time. Where having a garage, number of bedrooms or closeness to tube stations may vary and become more or less important over time, the price will always capture market sentiment towards whatever properties the house and region have.

To evaluate how a region is performing, 4most’s AVM looks at all sales in the region. Where a house has been sold twice then the growth in price from one sale to the next can in most cases be attributed to price growth in the region. By aggregating this information, a regional view can be extracted which can be used to create an up to date value of any given house in the region. Evaluating 2018, our model displayed the below best and worst performing Local Authorities that year, proving, as expected, that the removal of the toll on the Severn Bridge has had a positive impact on house prices in South Wales.

 
 

In our view, the key attributes for the next generation of house price indexation/valuation tools for use in assessing property value risk in a portfolio are:

  1. The index should prioritise comprehensive use of available actual sales data over using asking prices or valuations which are only marginally more current but contain material other biases

  2. The indexation mechanism should be reproducible and open to the user based on public data to support full independent validation, given this is likely to be a material model for many mortgage lenders in capital and provisioning models

  3. The estimates should be as granular as feasible given the data volumes – using modelling techniques to interpolate by region, property type and other attributes will always yield better quality and less biased estimates than simple averages

  4. The indexation technique should not be biased towards a basket of average properties in the market but reflect the mix of the portfolio being considered.

The 4most AVM achieves all of these aims and provides an accurate and open mechanism for providing property valuations.

For further information on 4most AVM, please contact:

Keith Church, Head of Economics: keith.church@4-most.co.uk or Rob McDowell, MD: rob.mcdowell@4-most.co.uk