House prices – lies, damn lies and seasonal adjustments
It’s not often I cut and paste a press release, but as this one is from super-mortgage nerd (meant in the most affectionate way) Ray Boulger, and is a really nice piece digging up the gravel of differences in a subject many are interested in, I couldn’t resist it.
This is a beautiful piece of research going into ridiculous but crucial detail on house prices; a subject we all care about. And as it’s not my work, I should be careful to credit Ray, who is the senior technical director of www.charcol.co.uk mortgage brokers.
LIES, DAMN LIES AND SEASONALLY ADJUSTED HOUSE PRICE STATISTICS
– Halifax & Nationwide Seasonally Adjusted February House Price Indices Differ by 1.2%
– Halifax & Nationwide Real February House Price Indices Differ by 0.1%
Ray Boulger comments on the disagreements between Halifax and Nationwide on the movement in house prices, highlighted by today’s release from Halifax of its February house price index.
Most of the house price indices published measure different things and furthermore figures for any given month in reality cover activity for different periods. For example The Land Registry figures have the largest sample and hence are the most comprehensive, but they are also the most historic because they are based on when details of completed sales are reported to the Land Registry. Near the other end of the timescale the Nationwide and Halifax figures are based on mortgage offers issued by those lenders and so are more timely, but also include figures for some transactions which never actually complete.
“For these reasons there are good reasons why different indices might record different figures for the same month but the Nationwide and Halifax indices are the exception because they are broadly designed to measure the same thing on nearly the same timeframe. This begs the question as to why they often vary so much, particularly on a monthly basis.
“One issue hampering all house price index compilers is the current historically low level of transactions. Unless a provider has significantly increased their market share this has reduced sample sizes, which makes it more challenging to make the necessary mix adjustments to reflect any changes in the type of property purchased from month to month. Nationwide has roughly maintained its market share in the current much smaller market, but Halifax’s market share has fallen and this means there is a double whammy effect for Halifax in terms of having to base its figures on a smaller share of a smaller market.
“Notwithstanding these challenges a major reason for differences between monthly figures from these two lenders is that the figures they announce in their press releases are seasonally adjusted. When I first started looking at reasons for the differences in these two indices I expected to find that it was primarily due to differences in the real, i.e. non seasonally adjusted, figures but what actually quickly became apparent was that differences in the seasonal adjustments used by each lender were often a major factor in the figures they announced showing such large discrepancies.
“The real, i.e. non seasonally adjusted, figures for average house prices are helpfully provided by Nationwide in its press release and so it is a simple matter for anyone with a calculator to work out the real change on a monthly, or other period, basis. Halifax, very unhelpfully, makes absolutely no reference to the real figures in its press release and so one has to know where to look on its web site to dig them out.
“The February figures from each lender provide a good example of how the seasonal adjustments are seriously distorting the way house price changes are reported. Halifax reported a seasonally adjusted fall of 0.9% for February but the real change was + 0.1%, whereas Nationwide reported a seasonally adjusted increase of 0.3% but its real figure was unchanged on the month!
“Thus the real house price change in February recorded by both lenders only differed by 0.1%, well within the impact of sampling differences, whereas the seasonal adjustments applied by the lenders differed by a whopping 1.3%, with Nationwide increasing the real figure by 0.3% and Halifax reducing it by 1.0%. This makes nonsense of the seasonal adjustments and hence the reported figures. Trying to work out seasonal adjustments is an art, not a science and, taking the February figures as an example, applying seasonal adjustments to the real figures turns what were virtually identical figures into significantly different ones.
“Furthermore, whilst the real house price figures are normally never adjusted, each month’s new press release revises many of the seasonally figures for the preceding 12 months, with sometimes the largest adjustment being a balancing figure for the same month a year ago because the one time when both figures normally have to agree is on any 12 month period! These adjustments are of course never highlighted by the lenders or reported.
“It would be understandable if there were differences in the real figures from Nationwide and Halifax as they obviously use different raw data to calculate those figures. However, such wide differences in the seasonal adjustments, even to the extent of one lender increasing the real figures and the other reducing them, just demonstrates what a farce the seasonal adjustments are,.
“Nationwide uses the US Bureau of the Census X12, whereas Halifax uses the US Bureau of the Census X11, for their seasonal adjustment calculations. The fact these two methods of adjustment produce such different results suggests that either one or both of them are flawed or one of them is not suitable for the purpose for which one of the lenders is using it.
“Many factors affect house prices, some more than the seasons. Changes in interest rates and consumer confidence and significant changes in mortgage criteria and availability all have a critical input. The various house price indices would be much more useful if the focus of the reporting was on the real figures, although of course the companies providing the information must be free to interpret the figures as they wish and provide whatever additional information they consider appropriate, including seasonally adjusted figures, however discredited these may be.
“If necessary the Government could take the lead here, in the interest of increasing clarity in these economically important monthly statistics. The review of the various house price indices currently being carried out by the Government provides an ideal opportunity to assess how presentation of the data can be improved. Provided sufficient historical data was included in the monthly press releases it would be a simple matter for any competent commentator to look at changes in the figures between the same months of the previous year as well as current trends and make their own judgement on how important the seasonal influence is.
“If investors, encouraged by regulators, had not blindly relied on flawed assessments on credit quality by the rating agencies but instead made their own judgement of the quality of so called triple AAA securities the credit crunch might well have been much less severe. Because the movement in house prices is such an important factor influencing many UK policy makers it is vital everyone concerned has good access to the real figures, which then allows individuals to formulate their own view as to how important seasonal influences are.”
