Measurement matters
After three consecutive shockers, the US consumer price index for April behaved itself last week. Core CPI inflation in the month was 0.3 per cent, bang in line with expectations.
If that is replicated every month, the US inflation rate would be pushing close to 4 per cent, so April’s slowdown is not sufficiently good to give the Federal Reserve confidence to cut interest rates soon — but the relief that the figures were not worse was palpable.
Within the CPI, owners’ equivalent rent — a proxy for the housing costs of owner occupiers assuming they have to rent their homes — rose 5.8 per cent in the year to April and contributed 1.9 percentage points of the 3.6 per cent rise in annual core CPI inflation.
The big beasts of US economics are noticeably unhappy about this phenomenon. Nobel prizewinner Paul Krugman complains that owners’ equivalent rent is “an imputed cost of housing that nobody actually pays and is very much a lagging indicator”. With US inflation back to target excluding the measure, he thinks the US economy is fine and poor sentiment data just reflects partisan vibes.
In contrast, in a February paper, former Treasury secretary Lawrence Summers came to the opposite conclusion. True, owners’ equivalent rent is not the way people think about housing costs, he argued. But if you look at interest payments and include them in inflation measures (as they did in the US before 1983), the pain imparted by the Fed fully explains people’s misery.
This can be summarised in two words: measurement matters.
I could leave it there or adjudicate between Krugman and Summers, but instead I am going to offer an alternative. They both want an inflation measure that reflects what people pay. And the UK’s Office for National Statistics has been developing exactly that for some years — it calls the result a household costs index.
The price data is exactly the same as the ONS uses for its headline CPIH measure (which includes owner occupied housing costs), but its calculation of inflation from the prices it collects each month differs significantly. First, it uses mortgage interest paid for owner occupiers and, second, it gives every household’s expenditure pattern the same weight in the overall index. Normally, inflation measures are weighted by the amount people spend, giving rich people a much higher weight; this is called democratic weighting.
With some small quibbles, I like the path the ONS is travelling along. The UK will end up with two inflation measures. One is theoretically sound and relevant for economic measurement (CPIH), and will use expenditure weights because spending matters for the overall economy. A separate index (HCI) seeks to measure the lived experience of households.
The chart below compares the two, showing the HCI has been higher during this inflation episode, first because food prices rose sharply and these receive a higher weight (democratic weighting) and second because mortgage rates went up.
If the Bank of England starts cutting rates soon, the inflation measures will cross over, with HCI inflation falling below CPIH.
The really neat thing about using democratic weights, however, is that it makes it easy for the ONS to show inflation at different parts of the income distribution and for different groups.
As I said, measurement matters, and generally households with mortgages have the highest inflation — that also means richer households and those with children. Pensioners, who often own their homes outright, have the lowest inflation rate, although that does not make them happy (as this clip from the past week demonstrates). See the chart for more data.
Fast food
Another measurement innovation to come out of the UK in the past week has been a significant step towards having a real-time index of the prices of goods in supermarkets.
Using web scraping and a large language model to categorise products, Richard Davies and Finn McEvoy are creating a daily food inflation figure based on the same definitions as the official measure from the ONS.
Instead of sending price collectors out to stores once a month, as the ONS and most other statistical agencies still do, their algorithms collect 100,000 prices a day from seven of the UK’s largest supermarkets comprising 80 per cent of all grocery sales.
Soon they will be able to publish a daily grocery inflation measure and examine the pricing policies of supermarkets, such as whether they are withdrawing their cheapest lines. When there are sudden changes, such as Brexit, new taxes on groceries (for example raising alcohol duties) or natural disasters, they will be able to provide the Bank of England and others with immediate information on the effects on prices.
The big question is whether the results are accurate. Davies told me that the large language model they used (ChatGPT 4) was initially quite poor at categorising products properly. “It took ages to get the LLM working right,” he said.
Now the results are pretty good, as you can see in the chart below comparing the auto CPI results with the official data for monthly price changes since last September.
Even though the data doesn’t match the ONS every month, the patterns are very similar and differences tend to iron out. When the annual data is available in the summer, it can be expected to track that of the ONS, but appear up to eight weeks earlier. In Germany, the Bundesbank is clear that this type of data helps its short-term inflation forecasts.
Price data is not the only application for such AI-assisted research. The origin of imported produce sold in UK supermarkets has also been categorised by Davies and McEvoy. Anyone with a long memory will recall that Brexiters told us in 2016 that the UK was shackled to a corpse of the EU economy and geography did not matter any more. It was obviously wrong then and, as the map below shows, it is still wrong today.
Web scraping is a great method for fast and cheap collection of price data. But the holy grail of inflation measurement is the use of the stores’ own scanner data, which can show prices and the quantities of products people are buying.
This means statisticians and central banks can weight their inflation measures by what people are actually buying, not what they were spending a few years ago. (The US currently uses 2022 weights in its CPI, for example.)
Australia led the world here and I caught up with Professor Kevin Fox of the University of New South Wales, who has helped the Australian Bureau of Statistics and other countries in using such data.
He told me that one of the problems of these techniques can be “chain drift”. This was evident in the Netherlands when it was examining the techniques. The quantities of product bought in one month form the weights in the next period. It found that when washing powder went on sale, quantities could shoot up by 1,000 per cent. The problem was that the high weight would be used the next month when the product went back to its normal price, vastly exaggerating inflation. The indices could be “explosive”, Fox said.
Complex calculations using multilateral index number methods (don’t ask), can resolve these problems and the use of scanner data for measuring grocery inflation is now taking the world by storm, but not in the US.
The UK will introduce scanner data for groceries into the official CPI in March next year and will publish early estimates of the likely effects towards the end of 2024.
Awkward adjustments
On a parochial level, the ONS in the UK has a policy of avoiding revising its price indices and has not changed the CPI since its launch in 1996, although it has twice revised CPIH (including owner-occupied housing costs) in 2015 and 2017.
There is a very strong case that it should have revised CPIH again this year.
In March, it fundamentally revised the collection of private rents, which are used extensively to proxy owner-occupied housing costs on a rental equivalence basis in CPIH. The chart below shows the new measure of rents compared with the old one.
There is no doubt that the new methods are an improvement, using a much more detailed and more comprehensive data set and giving more weight to rents in cities where private rented accommodation tends to be found.
But the effects are large. Between the start of 2018 and 2024, CPIH inflation was 1.6 percentage points higher (26.0 per cent increase vs 24.4 per cent increase) using the new rent index, but this will never get into the data that is released. The required CPI revisions are much smaller at 0.3 percentage points, but this does mean social security and pensions, for example, have not been uprated by the right amount.
This is a tricky call. Everyone should want statistical agencies to collect the best data possible, so no one should criticise the ONS for updating its collection techniques. But we might also want to revise our view of the past.
What I’ve been reading and watching
If you want more on measurement, read this great column by Alan Beattie highlighting how difficult it is to assess the scale of industrial subsidies
The European Central Bank is worried about the financial stability consequences of loose fiscal policy. This is a difficult issue as debt levels are rising in the main because servicing costs are high
If you want to read more about the April US CPI, Rob Armstrong gets it right in saying these were good, but not great, figures
Nigeria has a new central bank governor. Olayemi Cardoso, a former Citigroup executive, gives an appropriately orthodox message to the FT as he battles high inflation
Want to know about US government debt? This is the topic of Soumaya Keynes’ new podcast, The Economics Show, out yesterday and every Monday
A chart that matters
Financial markets just cannot decide. After spending most of 2024 downgrading their expectations of rate cuts, they are beginning to rise again, especially for the US and UK.
The FT’s public policy editor Peter Foster is relaunching his Brexit newsletter with a refreshed remit. “The State of Britain” will tackle everything from skills, planning reform and devolution to post-Brexit regulation, foreign direct investment and trade. Premium FT subscribers can sign up here.
Letter in response to this article:
A democratic measure of household income growth / From Martin Weale, Professor of Economics, King’s College London, London WC2, UK
Comments