“House prices surge by 68%”, Daily Express, 27/09/11

“House prices are 'treading water'”, The Guardian, 29/09/11

“House prices fall at fastest rate ever across England and Wales”, Daily Mail, 29/09/11

These three, apparently incompatible and entirely contradictory headlines all appeared in the national press within the space of three days. How is this possible? Are at least two lying? Or are they all just talking at cross purposes?

Analysis

Looking at these three articles together, Full Fact can isolate two separate potential reasons for the huge differences in these headline claims. One is that they all use data produced by different organisations, the other is that they all look at different time spans.

Lets look at the issue of the different datasets used first, and examine its significance. In these three articles there are three different datasets used; the Express uses the Halifax House Price Index, the Mail uses the Land Registry House Price Index, and the Guardian mainly uses the Nationwide House Price Index but also makes use of the Land Registry figures.

There are a few differences in the make-up of these indexes that explain why although they all purport to measure the same thing, the figures they report differ to some extent.

The Nationwide index is constructed as a measure of the price of a representative house. It is not a simple average, instead they apply a procedure known as “mix-adjusting” which is meant to ensure that the different proportion of different types of houses (types obviously vary on average in price) in a particular month or quarter does not effect the estimate of the overall price change.

The data used excludes certain non-representative houses such as council houses and houses sold not for private occupation. The data is based on Nationwide (mortgage) lending.

The Halifax index is much the same as the Nationwide one. Differences between the two come from differences in their definition of a representative house and the way in which they apply mix-adjusting.

The Land Registry index, on the other hand, is based on a dataset which contains a complete record of all residential property transactions. It claims to have a larger sample size than any other statistical measure available. It does not mix-adjust in the same way, but instead uses something called the “Repeat Sales Regression Method”. This involves measuring the change in price over repeat sales of the same property.

Although there are differences in the construction of these indexes, and they do consistently produce slightly different figures, they do not seem to differ in terms of long term trends. This is best illustrated in a graph.

This graph does not measure house prices over time directly, but instead shows how the different indexes estimate the annual per cent change in house prices each month (i.e. each month is shown as the per cent change in house price against the same month a year earlier).

Although there appears to be a slight time lag with the Land Registry figures, by and large the three sets of numbers have followed similar trends – particularly in more recent years.

It seems then, that the variations in the headlines we have looked out is much more to do with the time period looked at than the dataset used.

To underline this, it is worth looking at how each index measures the percentage change in overall house prices for each time period specified. (n.b. The Express' headline refers to specifically terrace houses. Full Fact can confirm that it correctly reports the Halifax figure for this. For comparative purposes we will now refer to overall house price changes for this period).

What we see then, is pretty much what we'd expect to see given our understanding that these indexes fairly closely map the same long-term trends. The shorter the period that we look over, the greater the variation in the figures reported by the different indexes (as a proportion of the percentage changes quoted).

The figures suggest that the smaller the time period being analysed, the more care that should be taken to look at a range of sources.

Conclusion

What this emphasises is the importance of placing short-term fluctuations in the context of long-term trends. The wildly different claims made by these different newspapers on this matter has highlighted the extent to which statistics can be manipulated in support of the story you want to tell.

The apparently irreconcilable differences in these three headlines is not due to falsehoods but to focus. In particular it is the differences in time period focused on and not the different datasets used that accounts for the extremity of the range of the claims.