House prices are attracting a lot of attention at the moment with speculation around whether we may see a housing market crash amidst a sluggish economy. Here at the ONS, we have also seen some debate about how average house prices, such as those published in our UK House Price Index (HPI) are calculated as well as some queries around the timeliness of our HPI data. Chris Jenkins explains the detailed calculations that take place in order that we can produce meaningful and comparable measures of change across the UK.
The change in prices of the goods and services we purchase has never been more topical, particularly given the recent rises in the cost of living. At the ONS, we produce a range of indicators to measure these changes in prices.
Measuring price change is not as straightforward as you may think. The key principle that underpins all our price indices is the need to ‘price to constant quality’, which I discussed in a recent blog. Basically, this means we only want to capture genuine inflationary pressure in our price indices and strip out any change in quality.
For consumer inflation, this is done by pricing ‘like for like’, so we pick a product at the start of the year, such as an 800g own brand loaf of white bread, and price the exact same product each month. However, when we think about housing, this becomes much more difficult.
Firstly, house price indices depend on prices collected from property that has been sold recently, and it is very unlikely that the same property is traded in consecutive periods. Secondly, unlike many other products (such as white bread) no two properties are identical. Two houses on the same street might differ in size, number of rooms or state of repair, all of which might impact the sale price.
Additionally, it is unlikely that the properties with the same features are traded in the same locations in each period, resulting in changes in the composition of sales from period to period. As both the features of the property and its location can impact prices, using basic average sale prices could result in a very unstable price index from one period to the next.
For example, consider how the average price would change if we were comparing a 4-bedroom detached house sold in London in one period (London being the most expensive region of the UK) with a 1-bedroom flat sold in Wales in another period. We would see a large price change driven by differences in what has been sold and where, rather than genuine inflationary pressure.
To ensure the UK HPI reflects pure price change, and not the change in composition and quality of houses sold each month we apply three methods:
- Hedonic regression.
- Mix adjustment and;
- Weighting.
Hedonic regression is a well-established and internationally recognised method for the calculation of house price indices. It is a statistical modelling technique that basically treats a property as a bundle of price determining characteristics, such as the location, size, type of property etc. Using data on the latest property transactions, a hedonic regression will estimate an average price for a property that accounts for differences in each of these characteristics, effectively stripping out any quality changes to give us a modelled price we can use in our calculations. Once we have these modelled prices, we then apply mix-adjustment and weighting.
Mix adjustment means we divide the property market into similar groupings, or ‘strata’. Each stratum will contain, as far as possible, properties with similar characteristics and locations in an attempt to compare like-for-like each period. So, for example, a stratum could be all semi-detached properties in a specific local authority.
Once we have our average prices for each stratum, we then combine them, or aggregate them, to produce our published indices such as by local authority, region or for the UK. We combine strata using ‘weights’, which is a common process used in the calculation of price indices. A weight is used to reflect the relative importance of a component in a price index. If one stratum has considerably more transactions than another, then we would expect that to have a relatively larger influence on the aggregated price. For the UK HPI, we use 12 months’ worth of the previous year’s transactions to calculate updated weights at the start of each year and this ensures our HPI remains reflective of the latest trends in the property market.
There are several private sector house price indices that are published alongside the UK HPI, but all will use a similar approach to ensure any change in composition or quality is removed from the calculation process. The difference in these measures is explained in detail on the UK HPI pages. Taken together, these indices alongside the UK HPI provide a comprehensive account of the UK property market.
The index published by Rightmove measures changes in the asking price of advertised property, so this provides an early indication of how house prices are starting to change. The indices published by Halifax and Nationwide reflect the price of a property being sold at the mortgage approval stage, so again these provide an early indication of changes in property price and in particular what is happening in the mortgage market. Finally, the UK HPI, and the index published by LSL Acadata measure changes in the final transaction price, so these are less timely but ultimately provide a final and full picture of inflationary pressure across the whole property market.
Christopher Jenkins is Assistant Deputy Director for Prices Division at the ONS