Using recent selling prices as a guide to rental values

HM Land Registry publish their price paid data for England and Wales free online, showing the full address and postcode and split into selling prices for flats, terraced houses, semi-detached and detached houses.

We last downloaded price paid data for the year 2014; deleted the ones without a postcode leaving 850,000 actual transactions; applied the magic of pivot tables; calculated the average selling price of a standard three-bed semi-detached house (as reduced by one-sixth to use Savills' lower average price) in each of 2,288 postcode districts (there are about 10,000 homes in each postcode district); ranked them in ascending order and typed in the average rent from www.home.co.uk (adjusted as mentioned above) for each 20th district up to those with average selling prices £750,000 (things get a bit hit and miss after that); and ended up with the following chart:


The co-efficient of correlation is 0.87, which is high enough for these purposes. 0.0 = no correlation, 1.0 = perfect correlation. Whether the outliers are just outliers or whether some of the underlying data is inaccurate is unknown, but this will do to illustrate the principle.

We can then deduct the annual running costs (see above) from the gross rental values to arrive at the land-only rental value/site premium and express it as a percentage of selling prices. This suggests that there are very few areas where the site premium is less than 3.5%, and these are probably due to patchy data more than anything else:


Using the last twelve months' data gives us 300 to 400 sales in each postcode district, or 100 sales in each smaller postcode sector. Obviously, there is a trade-off between using selling prices for the last twelve months and for longer periods. If you use longer periods, then you have a bigger sample size but if prices have changed a lot, then the older prices themselves will be less indicative.