The way providers of house price indices seasonally adjust their figures is a farce or, seasonally adjusted, a comedy.
Because in many months the seasonal adjustment skews the real figures so much the result is that the comment generated is often misleading.
For example, anyone who has seen some of the very bearish comments on house prices this morning following publication of Halifax’s May house price index announced yesterday could be forgiven for not realising that in the first five months of this year the real, i.e. non seasonally adjusted, Halifax house price index has actually recorded a small rise of 0.5%. In the same period the Nationwide real house price index recorded a rise of 3.1%.
The table at the bottom provides damming evidence going back to the beginning of 2009 of the farce of seasonally adjusting house prices. The figures shown in the table are taken from the latest monthly reports, i.e. for May, of Halifax and Nationwide, together with archive figures from their web sites.
Most house prices indices measure something different, and reflect different timescales, but because the Nationwide and Halifax indices are broadly designed to measure the same thing over the same timeframe, and hence are reasonably comparable, they are the best indices to use for comparison purposes.
In addition to the real figures the table shows both the seasonally adjusted change originally reported and the latest revised figure. Very often, particularly with Nationwide, many of the seasonally adjusted figures are changed, or to use the technical term revised, each month, whereas the real figures do not change.
In some cases the changes are small but in others material. Often a particular month’s figure will be adjusted in different directions in different months and so even when the net change shown in the table is small sometimes the totality of the changes will have been much larger.
The biggest real monthly rise recorded by Halifax since January 2009 was 2.4% in May 2009 and the biggest real monthly fall was 3.1% in September 2010. In both these months the impact of the seasonal adjustment was to increase the size of the rise and fall respectively. So much for a system which is designed to smooth out price changes.
Looking at Nationwide I have taken two months as an example to indicate the significance of revisions, which inevitably generate no media coverage – February 2011 and March 2010. The February 2011 seasonally adjusted house price index was originally reported as an increase of 0.3% but just one month later in the March report the increase had more than doubled to 0.7%. By April the figure had been revised back down to +0.4% and in May back up to +0.5%. The real change that month was nil.
For a longer term example March 2010 is interesting; the seasonally adjusted change was originally reported as +0.7% but the following month increased to +1.0%, where it remained unchanged for three months. In July it was downgraded to an increase of 0.9%, a figure which was unchanged for six months.
Then from January 2011 it was changed each month, down to +0.8% in January, +0.7% in February and then a big change in March, down to +0.3%, which was obviously a balancing figure as on an annual basis the seasonally adjusted change must always agree with the real change. Even over a year later this figure is still being revised – it was increased to + 0.4% in April and then in May reduced back to +0.3%.
Can there be any value in an index which initially reports a monthly increase of 0.7%, but with this increase subsequently revised upwards only a month later by nearly 50% and then a whole year later revised downwards to less than 50% of its original figure, and by 70% of its peak figure? Bearing in mind the importance attributed to these statistics by the Treasury, Bank of England and others, such revisions bring the original seasonally adjusted figures into disrepute.
The real figures for each month, on the other hand, do not change because they are real and so by definition nothing happens subsequently to require a revision – assuming no error in the original calculation.
Because on any annual basis the seasonally adjusted price change must always, by definition, agree with the real price change it is not unusual for there to be a significant change in the seasonally adjusted figure in the anniversary month, with both Nationwide and Halifax. This change results from a balancing figure being needed to get the right answer.
As shown in the last column in the table there have been 12 months, 41% of the total since January 2009 when the seasonal adjustment used by the two lenders has varied by at least 1%, with the difference being at least 1.5% on 8, 27% of those occasions.
There have also been 10 months, 34%, those highlighted in the last column, when Nationwide’s seasonal adjustment, based on the latest revised figures, moved the real figure in one direction and Halifax’s the opposite way.
Based on past form we can expect this nonsense to happen again next month. If index providers can’t even agree on the direction of any seasonal adjustment, let alone broadly the scale of it, that in itself is a pretty damning indictment of the whole shoddy seasonal adjustment concept, at least as far as house prices are concerned.
Whilst it is true that activity in the housing market tends to be seasonal, this doesn’t automatically mean that prices are impacted by the seasons. At the times of the year when activity is normally greatest both more buyers and more sellers tend to be active.
Whether buyers or sellers predominate, both at times of greater or lesser activity, is much more likely to be driven by other factors, such as interest rates, mortgage availability and consumer confidence.
To help restore confidence in all UK house price statistics, not just those from Nationwide and Halifax, I suggest all providers should as an absolute minimum give at least as much prominence in their press releases to the real figures as they do to the seasonally adjusted ones.
This would make it much easier for policy makers and commentators to interpret the figures sensibly without needing to do further research.
The government’s current review of house price statistics should make this a key recommendation.