Use the form below to login to your account. If you have problems contact the helpdesk.
Enter your email address and we will send you a password reset link or need more help?
This paper sets out an overview of published house price indices and a unique analysis of the key measures of house price index performance. It concludes that the most accurate way to track asset value across a portfolio is not indexation but via a specific valuation for each property.
How businesses use indices
Changes in residential property values have a direct impact on strategy and operational decisions for businesses with residential portfolios. National and regional house price indices are published by a wide range of organisations. Businesses use these indices for business planning, financial contracts, restating values across portfolios and as benchmarks of performance. But do users, often boards and executive teams, really understand how suitable different indices are for these purposes and the level of reliance to place upon them?
Too much focus on index ‘inputs’
There have been many articles on house price indices and their respective strengths and weaknesses. These tend to focus solely on index ‘inputs’ such as the volumes and source of data, market coverage, the headline methodology and the general story each tells. There is very little analysis on the performance of the resulting ‘outputs’ and the suitability of these indices for different business applications.
Importance of data volumes is over-stated
Big is not necessarily best when it comes to house price indices. Practical experience within Hometrack from the recent development of a new set of monthly city level indices, as well as benchmarking all published indices against the underlying market, has revealed that index performance has more to do with the index model algorithms and the process for creating the index than it does headline data volumes or type of index methodology used.
House price index methodologies
There are two primary methodologies for creating house price indices. The first set follow a ‘repeat sales’ approach, using pairs of price points for individual properties as the basis in constructing an index. This approach is widely used internationally and by the Hometrack and Land Registry indices.
The other major UK indices, including Halifax, Nationwide and ONS all use a multi-variate regression, or so-called ‘hedonic’ approach. This approach is data heavy, taking all the pricing inputs over a period and calculating the value of all the property attributes in the sample. The pricing of the attributes are used to calculate the price of a virtual, ‘standardised’ home which provides the basis for the index.
Black box models mask important differentials
Data sources and headline methodologies are relatively easy things to understand. The real differentials between indices tend to reflect the specific processes and algorithms used to construct each series. Different approaches to how indices are weighted or mix adjusted have a material impact on performance - and this is before the impacts of seasonal adjustment and any index smoothing.
Benchmarking index performance
It is possible to benchmark the performance of national and regional house price indices - and to use this analysis to draw conclusions on their suitability for different applications. Most comparisons of indices is focused on the differential in the average price or the headline rate of growth. Figure 1 above shows the average price over time by index while figure 2 below shows the spread in headline year on year house price growth for the same six index series.
Figure 2: Spread of UK year on year growth
Running a ‘matched pairs’ analysis enables a more robust measure of index performance. This is achieved by taking ‘pairs’ of registered sales prices for individual properties and then applying each individual index to the first price point to index forward and predict the second price point. The variance between the price suggested by the index, and the actual price at the second price point, is used to calculate two measures – index bias and index accuracy. Applied to millions of sales pairs over a 20 year period enables a detailed analysis of index performance over time.
Indexation bias a key metric
The bias of an index is the level to which it might systematically under or over-state the strength of the underlying market. The ideal index is the one that tracks as close to a zero bias over time.
Index bias is important for those using indices to re-state house prices across a portfolio as well as in contracts for land and other applications. Over time index inaccuracies can compound and when the sale completes, or assets within the portfolio are sold, so discrepancies with the indexed value will arise creating unexpected losses or gains.
Figure 3 plots the index bias for six national, UK house price indices over time. The chart shows that at a UK level published indices currently vary from over stating the strength of the market by 6% to understating price growth by 5%. The lines show how well the different index models use the data available to them to reflect price changes across the actual market. The charts shows a very marked reversal in bias in the Halifax index in the last 3 years of c12%. When run at a regional level, the pattern is broadly consistent across different series.
Do users really understand the suitability of different indices and the level of reliance to place upon them?
Index performance declines outside South East
Index accuracy measures how closely an index predicts the second price point in the matched pair’s analysis across a sample of sales. It is typically expressed as the percentage of cases that fall within 10% or 20% of the actual price. All regional indices have a broadly similar performance (with some exceptions). Figure 4 shows the spread of accuracy of all regional house indices between 2003 and 2015.
The chart shows that all South East house price indices predict the second price point to within 20% of the actual price in 83% of cases. This varies over time from 72% to 90%. Average accuracy falls to 66% in the North West i.e. the index is within 20% in only two thirds of cases. These results reflect variations in underlying housing markets in terms of homogeneity and transaction volumes.
The analysis highlights the inherent risk in deploying a regional index to adjust prices across a portfolio of individual assets.
Implications for business users
This analysis shows that national and regional house price indices published by the various providers, including Hometrack, are useful tools for identifying the general direction of travel in house prices. All exhibit a broadly similar level of accuracy but bias varies more widely over time.
However, deploying these series to benchmark local markets or re-value specific assets can be misleading as the user is assuming the local market is performing in line with the regional average.
Figure 5 plots the distribution of growth across London and the rest of the UK excluding London using Hometrack’s localised index series. It shows that the majority of housing in London is experiencing localised price inflation of between 7.5% and 12.5% whereas published regional indices put average growth in London as high as 17% and 10% for the UK.
Given this, users should consider how much reliance they should be placing on a high level snapshot of house price inflation at national or regional level. Tracking asset prices across a relatively illiquid market with a lack of homogeneity will always be complex. Businesses should look to focus on more localised or asset level analysis of price inflation.Figure 5: Distribution of price growth within regions
Hometrack indices and automated valuation model
Hometrack run monthly house price indices at national, regional and city level, all available here. The business runs thousands of localised indices which form an integral part of our products and services.
The most accurate way to track asset value across a portfolio is not indexation but via a specific valuation for each property. Hometrack’s automated valuation model is used to value many millions of homes each year, replacing indexation for use in provisioning and capital modelling alongside other applications.
Get articles like this straight to your inbox - subscribe to Hometrack newsletters using the link at the top of the page.
In the wake of the Brexit vote all eyes are on market data that can provide some sense of direction for what the impact on the housing and mortgage markets will be.
The housing market has been at the centre of a storm in recent years. From rent cuts for housing associations, to extending the Help to Buy scheme, and tax changes aimed at curbing investor demand, all types of organisations have had to react to a plethora of policy changes.
Brace yourself for a brave new world in the mortgage market, because it’s here. Anyone in our industry that has been following the ceaseless march of technology won’t have failed to notice some significant announcements this year.
The decision to leave the European Union means the near term prospects for the UK housing market now look very uncertain.