The most expensive statistical mistake in history

The most expensive statistical mistake in history

What would you say if I told you that, due to a bizarre statistical anomaly UK taxpayers — you and me — were unnecessarily being charged £2bn more than we should be? You’d be pretty shocked, I imagine — particularly if you heard that these costs were almost certainly avoidable.

And yet that is precisely what has happened as a result of an arcane decision made by the Office for National Statistics a couple of years ago. It was then that it concluded that the way the Retail Price Index, until then one of the main official measures of inflation, was not being calculated properly. Due to the way it calculated the change in prices (using an archaic formula called Carli, named after an 18th century Italian count (yes, really)) RPI had a nasty habit of persistently overstating price changes.

In fact, it emerged that RPI was, on average, 0.5% to 0.9% higher than it should have been. The ONS has since published a version of RPI based on a more sensible formula (known as Jevons). You can see the difference between the two of them here:

By the way, it’s worth emphasising that the difference between the two measures above is purely down to methodology, not due what they purport to measure. Don’t confuse this with the debate over the comparison between RPI and entirely separate measures of inflation, such as the Consumer Price Index.

So what? Well, it happens that RPI is used in a number of important formulas that determine prices and charges in the economy. Although most sensible people (including the Bank of England) have since ditched RPI in favour of the CPI, it is still used to calculate increases in train fares, in some pension formulas and to work out how high the interest rate should be on index-linked gilts. Which brings us to that £2bn.

Because RPI is higher than it should, in reality, be, the Government is paying out considerably more in interest to holders of those index-linked government bonds (essentially, inflation-proofed bonds) than it should. At the moment that disparity — between the actual cost of paying that interest, and the cost if the interest was calculated using that RPIJ index above, is £2bn a year.

The ONS could have avoided this. It could have merely changed the way it calculated the RPI. However, having taken advice (and presumably out of fear of lawsuits from bondholders) it suddenly decided that it should not do so. RPI would be left as it was. There would merely be a new index, RPIJ, which few people know or pay attention to, since unlike RPI or CPI it isn’t actually used for anything.

What’s even more inexcusable is that the Government, or to be precise the Debt Management Office, continues to issue index-linked bonds linked to RPI. Indeed, so many of these bonds are there floating around the financial system that British taxpayers are presently due to keep paying out on them all the way through til 2066. Unless the Treasury does something about this — and soon — the bill will only increase exponentially.

No wonder the holders of those gilts (including many big investors and pension funds) were so delighted when the ONS took the decision not to change the calculation of the index a couple of years ago. Even they were expecting the ONS to do the right thing and change the way it calculated RPI. When the ONS surprised everyone by saying it wouldn’t touch RPI, the index-linked bond market enjoyed the biggest one-day rally since 1987 (note that a fall in yields, shown below, equates to a rise in prices).

Ironically, the £2bn we annually pay out to such investors is about the same amount the Government squeezes off the financial system through its annual tax on the banks. So perhaps it’s all just swings and roundabouts.


To confuse things further, Paul Johnson, the IFS head who has been carrying out a review into the inflation numbers, also says the ONS should consider overhauling its main inflation index, the CPI, so that it includes housing costs. However (yes, this story gets more and more peculiar) CPIH, the measure the ONS produces that supposedly includes housing costs, is suspect. Like RPI, it has been stripped of its status as a national statistic (although in its case only temporarily).

Argh. What a mess. And, unfortunately for all of us, it’s a mess with expensive consequences. Expensive for the rail users who get charged over the odds because of the dodginess of RPI. And expensive for the rest of us because this dodgy statistic means we pay out more than we should to bondholders.

Though it often gets it in the neck for this and that mistake, the ONS does a largely brilliant job in collecting and curating Britain’s statistics. It remains credible and highly-respected around the world. But this episode reflects rather badly on both it and the Treasury. Let’s hope they confront it soon.


In the meantime, if you want to read more on the topic (I thoroughly understand if, having read this far, you don’t) do check out the Johnson Review. Lots of interesting findings in there, including that we should investigate using more online methods to improve inflation collection (which ties neatly with this report we made over Christmas). And that inflation is usually a lot higher for lower income households. See the chart below, which shows how inflation compares for the highest and lowest earnings fifths of the population.