The mystery of the £39 orange
Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.
In January 2021, an Office for National Statistics agent surveyed a shop in the Yorkshire and the Humber region, and recorded an orange on sale for £39.
It was unbelievable. The eye-popping price reading for “Orange-class 1-each” was designated a class 2 validity code by the ONS — “rejected by user”.
Thankfully, many other price readings were deemed less outlandish, and statisticians eventually used 507 different oranges — priced between £1.29 and 14p — to generate an average price for that month: 37p.
Had the £39 orange been allowed to stand, it would have represented a max observed price that was roughly 278 the minimum, and, as a percentage of the mean, the most expensive max price observed since at least September 2019 — calling into question the observation system, the shop in question’s business model, and/or the agent’s typing skills.
It would not, however, have represented the biggest min/max spread observed over that period.
That title belongs to ‘Wall hanging mirror upto 1.5sq’. In March this year, ONS agents found the cheapest such mirror at £1.99, and the most expensive at £3,695.
This time, the gap was accepted. The £1.99 mirror was given a validity code 4, “accepted by user”, and a special status indicator “N”, for “non-comparable”. The £3,695 mirror received code 3, “validated”, and no indicator.
With more log scale abuse:
There’s a somewhat serious point here: the way the ONS’s inflation basket operates is really weird if a £1.99 mirror and a £3,695 one (1,857 times the price) can make it into the same category.
Readers will recall that we’ve been down this road before (eventually culminating in a mental breakdown). But with some help from mainFT’s (outgoing) Max Harlow, we’ve been able to splice together a load of the ONS’s observed price data sets and actually look at the effect in aggregate.
And, in what is perhaps not a surprise, the ‘small caged mammal’ effect — ie, an apparent lack of clarity over exactly what is being observed — is, well, rife.
Picking a few egregious examples:
— “Floor rug”, November 2021. Min price £4, max price £1,740.
— “Bookcase/freestanding shelving unit”. May 2023. Min price £15, max price £7,481.
— “Women’s formal jacket”, February 2022. Min price 69p, max price £950.
Now, you might well be sat on your yacht, yelling at your laptop: “Wow great observations Alphaville, but median averages exist for a reason!”. And you’d be right — there’s no reason to assume these outliers messed up the eventual readings.
But we have enough data to make a wacky chart and we’ll be damned if your negativity is going to stop us.
Quick preamble: the following chart contains the max price as a multiple of the minimum for every item in the CPI basket since September, on a timeline. In our view, this is a vaguely appropriate proxy for the smallcagedmammalness of a given item.
But it is important to note that these ranges do still include all non-zero prices observed, regardless of whether the ONS subsequently deemed the valid. So, for instance, our friend the £39 orange will be present in the January 2021 orange prices. Errors of data entry will be just as likely to show up as errors of judgment. Don’t @ us.
Good luck finding anything useful in here:
Further reading
— Orange juice crisis prompts search for alternative fruits (FT)
Comments