If Benjamin Disraeli were alive today, living in an average £153,000 house in Norfolk rather than Hughenden Manor in Buckinghamshire, and holding bank, telecoms and media shares alongside his Anglo-Mexican and South African mining speculations, he might be tempted to reorder his classification of “lies, damned lies, and statistics”. Because the latest market data seem anything but mendacious.
Over the 11 trading days to July 27, the FTSE 100 index rose to a higher closing price every day, gaining 11.1 per cent – driven by HSBC, Lloyds, Vodafone, and Pearson (owner of the Financial Times). And, over the 31 days of July, the Nationwide house price index recorded higher average selling prices than in April, May or June – with three-month gains averaging 2.6 per cent on the previous period, or 6.5 per cent in East Anglia.
These are historically significant data series. Only twice in the 25-year history of the FTSE 100 has it registered 11 consecutive days of gains – in April 1997 and in December 2003 – and it was on the way to 6,000 on both occasions. This week’s record-equalling performance was more impressive, though – the previous 11-day gains were just 7.1 and 4.1 per cent respectively. And not since February 2007 has the Nationwide index seen such a positive rate of change over consecutive three-
month periods – causing Nationwide chief economist Martin Gahbauer to suggest “prices could end the year slightly higher than where they started”.
But are these data series statistically significant? There is certainly evidence to suggest the numbers are not anomalous.
UK equity prices appear to have responded to better-than-expected second-quarter earnings. HSBC calculates that almost three quarters of companies reporting in the past two weeks have beaten consensus forecasts, and RBS has a model for earnings growth that gives a year-end fair value for the FTSE of 4,700.
UK house prices have been stabilising on even the most cautious of measures. Land Registry figures – which record completed transactions rather than mortgage approvals – showed house prices rising in June for the first time in almost 18 months. Across the country, the monthly increase was only 0.1 per cent, but London prices were up 2 per cent. Estate agent Knight Frank concurred, saying demand from overseas buyers and a shortage of supply in exclusive neighbourhoods had resulted in a fourth month of higher agreed prices for £1m-plus London properties. Nor is it just a London phenomenon – Skipton Building Society said sales backed by its mortgages were up 23 per cent in the first six months of 2009, compared with the same period last year.
So, statistically, the data points may be reliable. But what about the number of data points – the “sample size”?
As 17th century mathematician Jacob Bernoulli put it in Ars Conjectandi (“The Art of Conjecture”): “Even the stupidest man knows for sure that the more observations that are taken, the less the danger will be of straying from the mark.”
How wide of the mark, then, might these market observations be?
Investors would certainly be a little stupid to think that the FTSE rally has been measured on a normal sample of trades. Over the 11 days of gains, daily volumes on the Stock Exchange averaged less than 1bn, compared with 2.3bn-2.5bn a year ago. As market watcher David Buik, of BGC Partners, put it: “Volumes have been absolutely rubbish, 800m-900m – it just goes to show the brittleness of the rally”. It’s insufficient data for Manoj Ladwa of ETX Capital, who this week warned: “A lot of the improvement in corporate performance has been down to cost cutting, stock market volumes are still low, confidence is fragile.”
Homeowners would be even more stupid to believe those house price rises were based on anything like a meaningful sample size. Over the periods covered by the indices, sale volumes were down by around 75 per cent, compared with the 10-year average – and down by 50 per cent on last summer. As former statistics software vendor Nick Hopkinson, now of Property Portfolio Rescue, put it: “Such an illiquid housing market makes looking at monthly price trends statistically meaningless.” It’s not sufficient data for Paul Samter of the Council of Mortgage Lenders, who this week admitted: “The outlook is still sluggish.”
Any statistic can lie if the sample size is small enough. Take house price inflation at Hughenden Manor. Technically, it is 1.1 per cent a year. But there are only two available data points: a Domesday Book valuation of £10 in 1086, and Disraeli’s purchase price of £35,000 in 1847 – which he could only afford with a 70 per cent loan-to-value mortgage from Lord Henry Bentinck!
