The RPI may not measure `real' inflation. It doesn't really matter

the problems of price indices

One of the complications the Bank of England's Monetary Policy Committee faces as it ponders, yesterday and today, what to do with interest rates, is the question of how good a guide the published inflation figures are to "true" inflation. For there is quite a vogue for arguing that the published retail price index, even on the target measure excluding mortgage interest payments, shows inflation to be higher than it really is.

This is in principle distinct from the argument that the economy has entered a new era of permanently low inflation because of a revolutionary improvement in productivity thanks to new computer technology. As Gavyn Davies demonstrated in his column earlier this week, this is a questionable assertion, as low inflation in recent years can be fully explained by low growth.

But in practice, the "new era" school of thought shores up its optimism with the idea that inflation is even lower than it appears to be, because of upward biases in the price indices. The link is that one of these biases in price measurement is the omission of rapidly falling computer prices and quality improvements.

The notion of serious mis-measurement stems from the report last year of the Boskin Commission in the US. Its economists concluded that true US inflation might be as much as a full percentage point or more below the official figure. There were several reasons for this.

One was that the index was not constructed using the lower prices charged in new kinds of retail outlets, mainly discount warehouse clubs. Nor did it include new products, like computers and other electronic goods, whose prices were falling. Nor did it take account of quality improvements that delivered better value for the same price. It missed the fact that people switch away from goods whose prices are rising too rapidly - for example, they buy chicken if fish becomes too dear. In addition, the commission criticised the formula used to construct the US consumer price index.

These conclusions proved controversial, and the US has not decided to implement them all. Meanwhile, the Office for National Statistics here has just published an assessment of how far the Boskin points apply to our Retail Price Index. The conclusion is: not very much. The RPI is based on a different formula. The UK does not have many discount clubs.

Just as important, the goods included in the RPI and the weights attached to them are updated every year on the basis of a survey of family spending patterns carried out the previous year, whereas the basket of goods in the US CPI has been updated only once a decade.

This year's RPI basket is based on 1995/96 spending patterns. The ONS has looked back at how different measured inflation would have been if the updating had been even faster. During recent years they found it to be only 0.06 to 0.07 percentage points, less than half the bias in the US figures the Boskin Commission attributed to this source.

This will not necessarily satisfy those who think the RPI overstates inflation and therefore makes the Bank reach for the interest rate trigger too early. For the index excludes some of the goods which are seeing the fastest price falls and biggest quality improvements - computers. The reason is that measuring their price and quality has simply been too difficult.

However, the ONS has started publishing a separate price index designed to be compatible with how inflation is measured in the rest of the European Union. This figure, the "harmonised index of consumer prices" or HICP, includes computers and, almost as troublesome because of their rapidly improving quality, new cars. As the chart shows, for the duration of its short existence, inflation measured by the HICP has been significantly lower than inflation measured by the RPI.

But does this make the case that inflation is "really" low and the Bank of England has nothing to worry about? Measures of price changes are needed for different purposes. To uprate social security benefits, for instance, an index which includes computers would probably be inappropriate, as the poorest families and pensioners buy very few of them. The Bank's interest in inflation is as an indicator of whether or not the economy is growing at a sustainable pace. Slow and steady inflation is essential as a solid platform for growth and jobs.

In a sense, therefore, it does not matter exactly which measure of inflation the Bank uses, for all tend to show the same broad trends. For month- to-month monitoring it is better to use figures that the statisticians can construct fairly promptly. The Government's target RPI measure excludes mortgage payments for the special reason that raising interest rates to help reduce inflation actually increases the headline RPI via this channel. The Bank itself would prefer also to exclude tax-related price changes on the grounds that these contain no information about the state of the economy.

But, broadly speaking, these three - the RPI, RPIX and RPIY - tend to show the same trends. The broader GDP deflator shows lower inflation because it includes import prices. The narrower producer price series shows lower inflation because it excludes retail margins. But any one of these would be suitable as a target measure. The key decision is setting the level of the target; and there is no convincing evidence that RPI growth of 2.5 per cent is incompatible with steady, sustainable growth and employment.

The Bank cannot entirely ignore the "new era" arguments. Technological change is making it harder to understand which prices matter. Should the ONS be measuring the price of books bought at a discount over the Internet? How can a conventional price index take account of the fact that a lot of computer software is free?

These will become more important issues over time. But there is nothing here that changes the kind of calculations the Monetary Policy Committee should be making this morning - nothing to persuade its members to relax about inflation.