ONS decides to continue with inaccurate RPI calculation
Statistics agency chooses consistency over accuracy.
Against expectations and the recommendations of a number of influential economists, as well as its own consumer prices advisory committee, the ONS has decided not to change how the RPI – one of the two key indices measuring the rate of price inflation – is calculated.
The pressure to examine the index comes from the longstanding difference between the RPI and CPI measures of inflation. Except for the brief period in the midst of the recession, when both indices were recording deflation, RPI has consistently shown levels of inflation higher than CPI. Here's the chart for the last three years, for instance:
Some of that difference is due to the fact that the two indices measure subtly different things – RPI includes a broader measure of housing costs, for instance, and it ignores very high and low income households. But the ONS has known for a while that there is also a discrepancy caused by the different formulae used to calculate them.
The ONS began a consultation into whether and how it should eliminate this "formula effect", and has concluded that:
Use of the arithmetic formulation (known as the ‘Carli’ index formula) in the RPI is the primary source of the formula effect difference between the RPI and the CPI… This formulation does not meet current international standards.
So what's it going to do about it? Well, nothing:
The National Statistician also noted that there is significant value to users in maintaining the continuity of the existing RPI’s long time series without major change, so that it may continue to be used for long-term indexation and for index-linked gilts and bonds in accordance with user expectations.
Therefore, while the arithmetic formulation would not be chosen were ONS constructing a new price index, the National Statistician recommended that the formulae used at the elementary aggregate level in the RPI should remain unchanged.
The ONS will, however, develop a new measure of inflation, called RPIJ, which will use a different, better, mathematical formula.
The consumer prices advisory committee, a body which meets around five times a year to advise the ONS on measures of inflation, accepted that the ONS has a responsibility to make sure that there is a level of continuity in the RPI calculations which would be destroyed if there were a change to the formula. But, given the ONS also has a responsibility to compile those statistics "in line with best practice", CPAC concluded that not changing RPI would be "unsuitable".
On the other side, arguing for no change, were 332 of the 406 replies to the public consultation. The ONS said:
The large majority of responses did not address methodological issues but identified the impact that the changes implied… would have for them.
The competing requirements present a tricky path for the ONS to follow, but it does feel like it has picked the wrong option at this point. The job of the statistics agency is surely to produce accurate statistics, rather than statistics which are continually inaccurate in known ways. The fact that RPI is used to decide the value of, amongst other things, index-linked bonds is a reason for it to be correct, not for it to be artificially inflated.
As it stands, the ONS has decided to continue publishing a "measure" of inflation which has an accepted and understood upward bias of 1 per cent a year. It has done this, not because of any real statistical reasoning, but because greater accuracy would be bad for a majority of stakeholders. That seems like a bizarre abdication of its duty.