The Financial Times today ran a story that the ONS has admitted errors in its measurement of the telecoms sector. It seems that the ONS has effectively been focussed on output of the telecom sector as based on turnover of the providers, and making a price assumption of the goods and services they sell. On this methodology, the ONS shows prices of telecoms were flat between 2010 and 2015, and turnover of the companies fell slightly, implying the real output of the industry was down 4% over the period. However, this approach fails to recognise huge changes in the quality of the goods and services sold. If prices stayed flat, then the telecom sector would have to be recognised as a productivity leader, similar to computers and computer components manufacturers, over the same period and for the same reasons.
However, if you try to adjust for the vastly improved quality of the hardware (which is surely seen to have improved faster than prices have risen) on the one hand, and on the data speeds and coverage on the other, then we have been in a productivity boom in this area for a number of years. And on top of that, the packages of data and texts and so on that people are now getting for a not dissimilar amount to before show that prices have plummeted on the services side. Putting this together, growth has been underestimated, productivity has been underestimated, and prices have been overstated.
Firstly to give the ONS their dues, this is remarkably similar to the issue that happened in the US in March 2017 when the statistician recognised the dominance of the ‘unlimited data package’ in the US (rather than pay as you go per unit of data up until then), and made a hedonic adjustment to telecoms services prices that bluntly was the single largest catalyst of the low US inflation story in 2017. Yes, larger in influence on US CPI than wages, which are indirect in their effect on CPI, and which were essentially flat in 2017. This will no longer be a negative drag when we get the April inflation data in the US, and is one of a plethora of reasons that may well account for the strong performance in recent months in US breakevens.
We are all used to seeing significant revisions to growth data, and particularly productivity data, and often over long and distant periods of the past. But CPI has never been backwardly revised since its introduction in 1996. It also raises a number of questions, such as what would happen to contracts that are linked to CPI (eg, pensions, pay), and would social security have to be reimbursed by people who were overpaid relative to revised, lower, CPI? These are surely too politically unpalatable to even consider? But if the changes were made from the point of revision to the data, there would then be a potentially large discontinuity between past CPI and present and future CPI. Would the Bank of England, for instance, at the point of revision, have to loosen policy aggressively so as to get new inflation up to 2%? But, in spite of these consequences of improving the data measurement, change for the better must occur. The question, for others, and for another time, is: how?
What might these changes to the inflation history of telecoms do to CPI? For this, as in many instances in the past, I have to turn gratefully as ever to Alan Clarke of Scotiabank, who pointed out the story to me this morning, and who had already run his own very rough numbers. Telephone equipment and services has a weight of 2.5% in CPI, so assuming 50% of that is services, and assuming that prices could be 35% to 90% lower in 2010-15 (as the FT article discusses), CPI could be 9bps to 23bps lower each year.
It may never happen, and if it does is likely not to be until at least 2019. Furthermore, the revisions are likely to only be made to CPI, as the ONS insists that RPI is ‘never revised’. This is important because holders of index-linked gilts earn RPI, not CPI, and the wedge (the amount RPI is from CPI, it can be lower as well as higher!) is assumed to be between 0.75% and 1%. If RPI isn’t changed, the wedge will have to be revised up to 0.85%-1.2%, and breakevens will move up, meaning linkers will outperform conventional gilts. RPI is already subject to criticism for being too high and antiquated, and this will not help matters. So, a slow burner, and one to watch. Mind you, we should recognise the difficulty in measuring this stuff. And perhaps if the quality and quantity adjustments were made properly to my shrinking soups and dry sandwiches around the City of London at the same time, inflation wouldn’t fall, it would actually rise!