We like Disraeli’s oft-quoted observation: ‘Lies, damned lies and statistics’. Statistics are a brilliant way to shed light on the way markets behave and have become an essential part of business life in these days of big data (and fake news).  Data for the housing market is no different, produced by national government, the Bank of England, mortgage lenders, industry bodies, estate agencies and many more, features on the housing market enter our news feeds on a regular basis.

In many instances data tells a similar story, however, there are instances where data appears to show completely contrasting views. Why? In simple terms it’s like trying to compare apples with pears - the source of the figures, the precise area they cover (catchment) the sample size (how many people, how many properties are included), and the calculation used on the data while seeming the same will all be different.

Let us explain. A recent headline stated in the press that Oxford is the city that has experienced the greatest fall in house prices over the last year while the dataloft inform data for Oxford indicates house price growth in the area.

The statistic:

-4.9% annual house price growth across Oxford

(Source: UK HPI, January 2018 versus January 2017)

The statistic:

3% annual house price growth across Oxford

(Source: Dataloft Inform, based on a rolling 12 months data)

Question: Which is right?

Answer:  They both are

 

The first figure comes from the UK House Price Index, the official statistic for monthly and annual house price change across the UK and the only national dataset for regional and Local Authority house price inflation. The calculation of the monthly UK House Price Index, is performed by the Office for National Statistics. It’s based on sale prices from HM Land Registry for England and Wales, Registers of Scotland and HMRC SDLT data for the Northern Ireland Residential property prices index along with data from the Council of Mortgage Lenders and Valuation Office Agency[1]. The catchment area  for this statistic is the local authority district of Oxford.

In January 2018, the average price of the ‘basket’ properties across Oxford was £392,792, in December 2017 the figure for the ‘basket’ of properties was £412,322, a difference of -4.9%.

The sample ‘basket’ of properties available to calculate the Oxford average prices each month is, in statistical terms, relatively small (around 105 pcm) but would normally be sufficient. However, in January 2018 less than half this number of sales is currently recorded, and this will undoubtedly be enough to skew the inflation figure.

The second figure is quoted for the Oxford catchment by dataloftInform. It too uses data from the Land Registry. However, the figure on inform is based on the average over the past twelve months versus the previous twelve months, i.e. the average price of a property sold February 2017 to January 2018 versus February 2016 to January 2017. It is also based on a geomean average.  In simple terms an average which indicates a central tendency and thus reduces the impact of outliers in the data. Finally it is based on the dataloftinform catchment of Oxford. Inform catchments are based on housing market areas and not local authority districts. In effect it will be a slightly different area and thus the basket of property sales it uses will be different to those used by the UKHPI.  

So to answer the question, which is right: both -4.9% and 3% are technically correct. The important question is realistically, which one feels right? To us at the moment, with annual house price growth across the UK standing at 4.9%, it feels more likely to be 3%. 

As research analysts, it is our responsibility to interpret and explain the different answers and to use statistics to illustrate our reading of the market. Often the most important commitment is to be consistent in the way you report statistics – to report the same stat from the same source over time so that we are comparing like with like.  That is what we do on inform, we quote the average prices and value growth based on a twelve month rolling time period every time.

[1] The data is mix-adjusted (for types of property, re-sale and new build) and is based on a hedonic regression analysis that takes a ‘fixed’ basket of properties sold each month. The report is time lagged by two months to reflect the 2–6 week between completion of a sale and registration at the Land Registry. That is a well-respected and sound methodology for reporting percentage change in prices.

 

 

X

Book demo today

The best property players
rely on inform local property market analysis