The price of a house – a stupid average

The ONS published this week a new – I’d say conceptually more sound – experimental house price index. It is based on the stock of homes not the flow so tells us what’s happening to the whole market not the price of what’s just been sold.  The estimate of the average price is lower than under the old methodology – £194,000 compared to £215,000 in the old measure, about 10% lower. The stroke of the methodologist’s pen has made homes more affordable even though no prices have changed! Perhaps this is the time to reflect on the full range of house price estimates at our disposal – and, dare I say it, how meaningless the average numbers are? Depending on what you count and how you add the numbers up, the resulting averages can be wildly different, as much as £100,000 apart. 

Homes come in all shapes and sizes. A few moments on the web and you can find them for a few thousand pounds or many millions depending on size, state of repair and location. These are so different it is hardly appropriate to bundle them into a single average and expect anyone to make any sense of it. Despite this, the “average” house price is a number that is focused on by the media and many others. Owners and would-be owners pick them up and feel joy or despair. Yet the average is irrelevant as each family only engages in a very small part of the whole market.

Acadata, a publisher of one HPI, produced a useful briefing note in July 2017 that set out the array of indices. At that time they suggested the reported average property price ranged from under £200,000 to almost £300,000. The latter is 50% more than the former. Where else would we accept such a range? If we were given various estimates of the average height of UK residents ranging from 1.4m to 2.1m, a similarly wide range, we’d laugh them off.

Differences in the house price averages can arise for two broad reasons: the data collected and how the raw numbers are aggregated. 

Looking first at the aggregation methods, as the Acadata chart below shows, they explained that using the sales price data, the mode (the most common price) was £165,000, the median (the mid price) was £204,000 and the mean was £295,000. The geometric mean would be different again (and below the arithmetic mean). It is a striking example of how aggregating or presenting the same data in different ways can produce conceptually reasonable measures yet give a very different impression of the state of the market. 

Screen Shot 2018-02-20 at 23.46.53.png

Turning then to how different producers of HPIs collect different data. Broadly speaking the indexes we see might be asking prices, valuations or actual sale prices. Despite being labelled as being the price in a given month, the figure might be smoothed over several months or relate to a time period other than that month. (The IFS produced a report in 2014 on the different measures in theory and practice and so did, more recently, the Land Registry, from which the table below comes.)

Screen Shot 2018-02-21 at 16.15.36.png

The latest figures for the main measures are:

Halifax. The average price in January 2018 was £223,000, with the annual rate at 2.2%. The background to the index is set out here, and, since the index was bought by Markit, all data bar the headlines now have to be purchased.

Nationwide. Its latest figure (January 2018) is £211.000, with an annual rate of 3.2%. Its methodology and data can be found here.

Acadata. This data shows a much higher average of £301,00 (January 2018) and a modest annual fall (the first since 2012).

Rightmove. This is a survey (details here) of asking prices and the latest data relates to February (showing an average of £300,00, up from January’s £298,000).

ONS. The official data are presented here, and show an average price of £227,000, but are a month behind the other measures. In a decision that can only be designed to confuse users and reflect departmental infighting, users are directed to the Land Registry pages on for the detailed figures.

Hometrack. They produce a UK index and a city index. The December report showed a UK price of £211,000 and a city price of £251,000.

The range is pretty much unchanged since last year’s Acadata report. (There used to be separate ONS and Land Registry estimates which were merged into the existing ONS HPI two years ago.)

Not only are the average levels different, but the annual rates of change and, importantly, the cumulative growth over a period of years, are different too. It is strange that such confusion exists for a market that is so important personally, politically and economically to people and policy-makers. We need a dataset readily available to users who want to understand how house prices are changing in the area and property type that is of interest to them. Sadly there is little prospect of us getting that.



The ONS published a stock-weighted House Price Index (HPI) on 20 February. It was a welcome concept in principle and is presented as an experimental series, with data only up to 2015. The pre-existing ONS HPI was a “flow” measure, one that reflects residential properties purchased during a particular period, while the new one estimates the HPI for the stock of residential properties. The two can obviously be different as the small number of homes sold each year (less than one in 20 of the stock) might not be representative at all of the stock.

As the article explained, at December 2015 the average house price for England and Wales under the new stock measure was £194,000, while for the (old) so-called flow measure this was £215,000. (See the chart below.)

Screen Shot 2018-02-20 at 23.21.22.png


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s