RPI is silly, but not completely crazy

Chris Giles, the FT's economics editor, has recently been waging a war on the retail prices index (RPI)—Britain's venerable price statistic used to set rail prices, student loans interest, and repayment of some gilts. I'm a fan of Chris, but I think he's gone a bit far: yes, RPI is a bad index, but no, it's not necessarily unfair and wrongheaded in the way he describes.

Nowadays, the official measure of inflation is the consumer prices index (CPI), which, unlike RPI, is designated a national statistic. It's what the Bank of England uses for the flexible inflation target its monetary policy is based around and it differs from RPI by using a much better aggregation method, a bigger and more broad-based sample, and excluding housing.

It's not a judgement call: it's just a better index, which is why more or less everything has switched over. But some things aren't. For some of them it's because they date backwards. For example, the government has long sold RPI-linked gilts—there are £407bn outstanding according to Giles—from which we impute the TIPS market forecast of inflation. For others, it's less obvious why they do.

Now Giles has one very good point. In 2010 the RPI formula was changed to measure certain goods (especially clothes) more wrongly. This inflates the index. This means that repayments to pre-2010 RPI-linked-gilt-holders are higher than they would otherwise have been. Whenever this move was expected—or if unexpected, announced—this was a handout to pre-2010 holders. But after that point, it's all priced in. Everyone knows the index will overestimate inflation, and everyone knows by about how much (any error benefits the govt as much as the investors). Yes, we shouldn't have done it, and maybe we should even claw this money back—but it was a once-off error. Market pricing means it doesn't compound.

But I don't follow his other points at all. Yes, RPI adds some arbitrary amount onto "true" inflation, so post-2012 RPI-linked student loan interest rates are higher than they would be with CPI. But the interest rate on these student loans is entirely arbitrary anyway. Given their repayment rates (around 55%) and repayment schedules, the government is clearly subsidising their true cost to an astonishing degree. The RPI link is a semi subtle way of getting a small portion of that back. Like how "money illusion" means unexpected inflation is good during slumps.

The same is true of rail fares. It's good that RPI hides a little bit of extra increase in real fares. Economising on scarce resources through prices is a good thing, and we currently subsidise rail somewhat too much. If we do it through explicit price increases, people might bear a larger psychological burden when we (slightly) reduce how much the government pays for people's rail travel.

Switching to the CPI doesn't magic up money. In both cases it just makes the government pay more, and the users of the service less. Does Giles really think that the baseline is inflation plus the arbitrary number they've currently set by fiat, rather than inflation plus that arbitrary number, plus the arbitrary chunk of measurement error? It's hard to see why.

This isn't to say that we shouldn't switch away from RPI. It's a bad stat, and if Chris is right about the legality of doing so, then it sounds like we could quite easily switch, eventually, without the large reputation costs that go with seeming like we're reneging on obligations. But let's not use motivated reasoning to get there. And is it really necessary to use language like "fleecing", or blame the ONS, who almost certainly are not the ones making the final judgement call?