Recently Shelter came to us asking for data on house prices across the UK to help them with some research in support of campaign on housing affordability.

This is a challenge we’re well suited to address, in fact a large fraction of the ScraperWiki team have scraped property price data for our own purposes. Usually though we just scrape a local area, using the Zoopla API, but Shelter wanted the whole country. It would be possible to do the whole country by this route but rate-limiting would mean it took a few days. So we spoke nicely to Zoopla who generously lifted the rate-limiting for us, they were also very helpfully in responding to our questions about their API.

The raw data for this job amounted to 2 gigabytes, 34 pieces of information for each of 500,000 properties for sale in the UK in August 2013. The data tell us about the location, the sale price, the property details, the estate agent details and the price history of each property.

As usual in these situations we fired up Tableau to get a look at the data, Tableau is well-suited to this type of database-table shaped data and is responsive for this number of lines of data.

What sort of properties are we looking at here?

We can find out this information from the “property type” field, shown in the chart below which counts the number of properties for sale in each property type category. The most common category is “Detached”, followed by “Flat”.

We can also look at the number of bedrooms. Unsurprisingly the number of bedrooms peaks at about 3 but with significant numbers of properties with 4, 5 and 6 bedrooms. Beyond that there are various guest houses, investment properties, parcels of land for sale with nominal numbers of bedrooms culminating in a 150 bedroom “property” which actually sounds like a village.

What about prices?

This is where things get really interesting. Below is a chart of the number of properties for sale in each price £25k price “bin”, for example the bin marked 475k contains all of the houses priced between £475k and £499,950 – the next bin being labelled 500k containing houses priced from £500k to £525k. We can see that the plot here is jagged, the numbers of properties for sale in each bin does not vary smoothly as the price increases, it jumps up and down. In fact this effect is quite regular, for houses priced over £500k there are fewest for sale at the round numbers £500k, £600k etc most for sale at £575k, £675k and so forth.

But this doesn’t just effect the super-wealthy – if we zoom into the lower priced region, making our price bins only £1k there is a similar effect with prices ending 4,5 and 9,0 more frequent than those ending 1, 2, 3 or 6, 7, 8. This is all the psychology of pricing.

Distribution of prices around the country?

We can get a biased view of the population distribution simply by plotting all the property for sale locations. ‘Biased’ because, at the very least, varying economic conditions around the country will bias the number of properties for sale.

This looks about right, there are voids in the areas of the country which are sparsely populated such as Scotland, Wales, the Peak District and the Lake District.

Finally, we can look at how prices vary around the country – the map below shows the average house price in a region defined by the “outcode” – the first group of letters in a UK postcode. The colour of the points indicates the average price – darkest blue for the lowest average price (£40k) and darkest red for the highest average price (£500k). The size of the dots shows how many properties are for sale in that area.

I’m grateful to be living in the relatively inexpensive North West of England!

There’s plenty more things to look at in this data, for example – the frequency of street names around the UK and the words used by estate agents to describe properties but that is for another day.

That’s what we found – what would you do?