A very interesting article in yesterday’s Financial Times by its Economics Editor, Chris Giles, sheds more light on the RPI vs CPI arguments (Straight Statistics, passim).

He points out that in August clothing and footwear measured by RPI had increased in price by 6.3 per cent over August 2009, while measured by CPI it had fallen by 1.7 per cent. That’s an absurd discrepancy.
 
It arises from the different methods of calculating average rise in prices. RPI uses the arithmetic mean, in two ways (explained here in the earlier posting).One of these gives most weight to the most expensive product, one to products that have risen most in price. The CPI uses a geometric mean which gives less weight to these two classes of product.
 
In goods with a huge price range, from Primark at one end to Bond Street at the other, this makes a big difference. In routine goods such as food or fuel where prices vary much less from outlet to outlet, it makes a much smaller difference.  
 
To make matters worse, ONS relaxed guidelines early this year for its researchers who check prices in shops. Instead of comparing items that were absolutely identical month to month, they were given more flexibility. This increased the variability of prices, especially for items such as clothing or footwear, which in turn exaggerated the difference between arithmetic and geometric means.
 
That helps to explain how it happened, but hardly justifies it. ONS urgently needs to sort this muddle out and provide a single inflation  measure that does the job as well as possible.