The ONS puts consistency first
Britain's Office for National Statistics has decided, when it comes to inflation, it's better to be consistent than to be right.
It is not going to seriously change the way that Britain's oldest inflation index is calculated, even though that goes against the recommendation of its own Advisory Committee and many international bodies such as the International Monetary Fund.
Those in the financial markets who had bet on a different outcome will be disappointed. So too will the Chancellor - who stood to save billions a year in debt interest from the change.
Economists and statisticians who think the Retail Prices Index (RPI) is seriously flawed won't be happy either. Many of them thought a change was long overdue.
But, anyone with a pension or investment that is still linked to the RPI will be deeply relieved. There are nearly £300bn-worth of index-linked government bonds, or gilts. Countless other investments and pensions are linked to it as well.
If you're dependent on an index-linked annuity, the change being contemplated by the ONS could have cut its value by 15% over 20 years, with the cost widening every year.
Coming on top of the other savings that the government has been making, that would have been a serious blow. Even if many economists would say that the change had made the index more accurate.
Carli v Jevons
If we were creating the Retail Prices Index from scratch, the National Statistician, Jil Matheson, admits that we would not design it the way it is now.
The Carli formula - which is used to generate nearly 30% of the RPI - has some strengths, but is not used by any other serious statistical body in the world because it tends systematically to overstate inflation, especially when prices for a certain good are highly variable.
If the prices of clothes rise very sharply, then come back down to where they were, you might think that an inflation index should show no overall change. That is what the Jevons method - used in the Consumer Prices Index (CPI) and around the world - would show.
But the Carli formula would say there had been some inflation - even though prices are back to where they were at the start.
The Jevons (geometric) method also takes greater account of the fact that consumers tend to switch to other products if one becomes more expensive. Whether that makes it better than the Carli depends on the good you're talking about, but many statisticians consider it another plus.
So much for the statistical niceties. What difference does it make?
By itself, this difference in formula has tended to make the RPI 0.9 percentage points higher than the CPI, on average, since 2010. (That was when they changed the way they collected price data for clothes. Before that, the formula effect led to a difference of around half a percentage point.)
Remember, that's independent of any difference between them that is caused by the fact that more housing costs are included in the RPI.
That sometimes makes it lower than the CPI and sometimes makes it higher. You might remember that RPI inflation rate went negative in 2009, when interest rates were falling sharply, whereas the CPI measure never went below 1%.
The point about the formula effect is that it systematically pushes the RPI higher, relative to the other index, when inflation is positive.
So, time for a change, many people thought. But in weighing this decision, the ONS also had to weigh the interests of the millions of people who hold investments and pensions that are linked to the RPI.
In effect, Ms Matheson has decided to put them - and the historical integrity of the index - before anything else.
As I said, it will be a shock for the City: the shorter-term index-linked bonds, which have suffered in the past few months as a result of the consultation, will probably go up today. But some inside the ONS will not be surprised.
I served briefly on the Consumer Prices Advisory Committee, set up a few years ago to advise on such changes, along with other outsiders like the FT Economics Editor and representatives of the Bank of England and the CBI.
I left before they got into discussing this issue. But even from the few meetings I attended, it was clear that people at the ONS take the historical consistency of the RPI very seriously indeed.
I often have to explain why GDP and other economic statistics have been revised down - or up. You might have noticed that that never happens with the CPI or RPI. There is just too much money riding on them.
Once released, the inflation numbers stand for all time, even if it turns out someone has made a mistake. The RPI does not get revised. And looking ahead, the ONS has decided it's not going to be seriously updated, either.
Update: 11.28 GMT
The FT's Economics Editor, Chris Giles, has been much more pointed in his coverage of the RPI decision. (As I mentioned earlier, he still serves on the Consumer Prices Advisory Committee, which favoured a change in the RPI formula).
He says he "cannot think of another occasion when national statisticians have discovered a major conceptual error and decided to do nothing about it".
He also disputes my suggestion that the RPI has been historically consistent, pointing out there have been many changes over the years, like the introduction of mortgage interest payments, which make it very different today from when it started in the late 1940s.
This paper from the ONS outlines the history of these changes, some of them fairly significant.
Looking at the list, it's hard to see a change that systematically affected the value of the entire index, in a single direction, as much as dumping the Carli formula would have done.
Still, you might say the size of the change doesn't matter. The point is that they have made changes before, in order to make the index more accurate, even at the cost of making it less consistent over time.
If the ONS believed this change would also improve the quality of the index - why not go ahead?
In looking for an answer, it's hard to escape the conclusion that the size of the change did matter, a lot. The National Statistician looked at the sheer scale of the financial impact, and the public response to the consultation (more than 80% opposed the change, almost always for financial reasons), and just couldn't stomach it.
Some will find the decision welcome, and understandable - especially those who think the RPI and the CPI are both a poor reflection of the inflation faced by pensioners.
As Robert Peston has pointed out, holders of indexed gilts could argue that they paid for the inflation protection offered by the traditional RPI, and they are legally entitled to get it.
Others will say the ONS chickened out.
But critics of this decision are right to emphasise one point. It's not just the Chancellor that loses out - it's all of us, in our capacity as taxpayers.
If, as the ONS itself accepts, the RPI has an "upward bias", then we as taxpayers have been over-compensating investors for the impact of inflation, to the tune of billions of pounds a year.
Thanks to this decision, we're going to carry on overcompensating them.
It's worth pointing out that any young person with a student loan will also lose out, because their official loans are indexed to the RPI. Rightly or wrongly, the ONS has put the interests of pensioners and bondholders first.