The disappearing new build premium in UK housing

As part of their long-running house price index Nationwide publish data on average prices for new, ‘modern’ and ‘older’ properties at UK level. According to this data the price premium on new/modern homes versus older ones has disappeared over time.

2016.02.07 Nationwide new build premium 1952-2015 UK HP Since 1952 Chart 6

Some of this is probably due to objective quality trends – i.e. the low-quality older homes being demolished and more investment being put into the remaining ones – but I think some is a symptom of our decades-long refusal / inability to build enough new housing.

In functioning housing markets new homes should be pricier due to better-than-average locations and build quality (remember that EPC statistics show new builds are on average slightly larger and much more energy efficient than the existing stock).

A consistent new build price premium would also be a sign of filtering at work, with higher income households buying new homes and older ones filtering down to lower incomes. This filtering process is a natural feature of well-supplied housing markets and is critically important for improvements in overall affordable and quality.

For comparison, in the US (where they build a lot more homes than we do) new builds command a premium of around a third.


But because in the UK we block new supply in high-demand areas the price of old homes is pushed up while new builds are diverted to worse locations. This erodes the new build premium and cuts off the filtering mechanism. It’s a pretty clear signal that we’re doing housing policy wrong.


Analysis of London housing sales data

The Land Registry has recently started publishing data on individual residential property sales, and the two charts below use the data to compare housing markets in London and the rest of England and Wales.

The first chart shows the distribution of prices (up to £1m) in the capital and elsewhere in 2012, broken down by the tenure type (leasehold and freehold, roughly corresponding to flats and houses) and whether or not the property was newly built at the time of sale. Average prices are higher in London across all categories. You can also see the effects of stamp duty in the kinks of the price distribution curves just before the threshold values of £125,000, £250,000 and £500,000.

House price distributions, London and the rest of England and Wales, 2012


The second chart plots the total value of new build and non-new build residential property sales in 2012 by local authority area, with London boroughs in cyan and other local authorities in pink. In the vast majority of cases the value of new build sales is far lower than the value of older home sales. London boroughs account for the top four new build markets, Tower Hamlets (in the top left) leading the way with new build sales that account for around 30% of total sales value in 2012. Kensington and Chelsea (bottom right) is at the other end of the spectrum, with very few new build sales (because it doesn’t really build any new housing) but an incredibly valuable resale market. Westminster (top right) combines both a lot of resale and lot of new build sales.

Total value of new and old housing sales by local authority, 2012


The data was analysed in R and the charts produced using the ggplot2 package among others.

London Housing Market Report

I have converted the GLA’s London Housing Market Report from a quarterly document (last edition here) to a live resource on the London Datastore. It uses the Google Charts service via the googlevis R package, which allows you to create HTML straight from your own data rather than having to upload it to Google’s servers. While the use of Google Charts started out as a labour-saving device and hopefully will be over time, setting this up involved a lot of fiddling with Javascript to get the charts to display okay on the Datastore’s Drupal system. I’m grateful to Scott Day in the GLA’s GIS team for help with this aspect.

The great advantage of using Google Charts are that its layout and colour defaults are very good and where changes are required they are pretty easy to implement. I used to agonise excessively over the choice of colours in charts like this one but I think the Google version looks fine.

Quarterly change in average London house prices - comparison of indices

Most of the report consists of fairly standard presentations of market data. But one thing I would like to draw people’s attention to is the presentation of changes in average rents as published by the Valuation Office Agency. We show these changes at Inner London, Outer London and London-wide level to show that the London figure can be significantly distorted by shifts in the VOA dataset, notably the exclusion of a large number of Housing Benefit cases in Outer London over the last year which has skewed the balance of the dataset towards Inner London. So the average rent for a one bedroom property in London seems to have risen 13% even though the average in Inner London rose by 8% and the average in Outer London by ‘only’ 6%! I’m not sure these features of the VOA dataset are widely understood but lots of people are using their data to calculate changes anyway.

Change in median private rents in London, March 2012 to March 2013

Net housing supply by hexagon

This map, which didn’t quite make it into Housing in London 2012, shows total net conventional housing completions between 2007/08 and 2010/11 in London, aggregated using an artificial hexagonal grid in Quantum GIS. This type of grid can be useful for aggregating point data, though it still suffers from the usual problems of shaded maps in that variation within the classes is lost.

Net conventional housing supply in London, 2007/08 to 2010/11

The data comes from the London Development Database maintained by the GLA and comprising data provided by the London boroughs. My colleagues in the GLA GIS team have made an excellent interactive map for exploring LDD data, which you can find here.

The colour of London’s commute

This map shows the broad mix of transport modes Londoners use to get to work, according to the 2011 Census. It uses RGB space to show the share of car/van/taxi/motorbike (in red), cycling/walking (in green) and public transport (in blue).



A couple of notes / caveats: it is based on place of residence rather than place of work, so the central area is green rather than blue because city centre residents tend to walk while city centre workers tend to get public transport. And it shows the main mode of transport so will tend to exclude short journey stages such as walking to or from public transport.

The map was made in R, and was inspired by James Cheshire’s maps of voting patterns.