Spurious accuracy seduces journalists time and time again

An insidious abuse of statistics is spurious accuracy. With this, numbers that have considerable margins of error in their construction are reproduced to fractions of a decimal point or to the nearest pound.
 
A prime example is a story in Monday's Guardian  reporting a study by the Nationwide building society. It purports to show that parents pay an "extra £8,670" for a home in the catchment area of their top local primary school. The story was repeated in The Daily Telegraph today. Here's The Guardian's take:
 
                 
 
A moment's thought would show that this is a ridiculously precise figure. To produce it you have to calculate the extra a house fetches in the catchment area of the school. Estimates of average house price are necessarily imprecise. The price of each individual house depends on many characteristics of that house, so estimates depend on whether the sample of those sold accurately reflects the characteristics of the stock as a whole. So they cannot be accurate. This difficulty is magnified when you are looking at a subset of houses for a particular local area.
 
So to conclude, as the report does, that the proximity adds 3.3 per cent to the value of a house, then to multiply that by the average house price to get the £8,670  is to multiply together two figures with substantial margins of error. To then report this as £8,670 is a plain error, as egregious as an error in calculation. Even worse, similar calculations are reported region by region. At regional level, samples are smaller and therefore the statistical error potentially greater.
 
There is also an interesting issue of causality. It may be that the schools getting top results are doing so because they are situated in areas populated by parents of a higher educational attainment and therefore more likely to produce achieving offspring.
 
So why do such distorted reports appear? Because false precision creates an impression of accuracy. A report that good schools are adding around £9,000 to house prices, though less unlikely to be false, would lack impact. PR companies and PR departments understand this perfectly well - which does not stop less savvy journalists falling for the trick time and time again.