The inflation data with a missing month

Headline inflation eased to 2.7 per cent year-on-year in November, the Bureau of Labor Statistics reported. Economists who went through the underlying data found a problem with no real precedent. The index had been built with a whole month of price data missing.

October, copied from September

During the 43-day government shutdown in October, federal field staff were barred by law from collecting retail prices, so the BLS had no price data for the month. The agency's statisticians filled the gap by carrying September's prices forward into October, which smoothed out month-to-month swings. When collection resumed in November, the method understated how fast prices of durable goods and services had actually been rising.

The release itself showed the strain. The bureau published no October index for most items and reported changes over the two months from September to November instead of the usual monthly figures. Core inflation, which excludes food and energy, came in at 2.6 per cent, its lowest since early 2021, a result that surprised forecasters who had expected tariffs to keep goods prices climbing. Collection also resumed only partway through November, so a larger share of the month's prices were gathered in the discount-heavy weeks around Black Friday.

Where the bias sits

The largest distortion is in housing, which makes up about a third of the index. The BLS measures rents by dividing its sample of housing units into six groups and pricing each group once every six months, then converting the six-month change into a monthly rate. The group due in October was never priced, and the carry-forward effectively recorded no change in those rents. Because owners' equivalent rent, the imputed cost of owner-occupied housing, is built from the same rent data, the gap flows into the index's single biggest component.

The bias should unwind on a known timetable. The housing units skipped in October are due to be priced again six months later, in April, and their rent increases will then show up in full. That will lift measured shelter inflation in the spring, just as the months after the shutdown drop out of comparisons. The effect on other items is harder to trace, because goods and services prices are sampled more often and the missing month's changes may have been partly captured in November's two-month comparison.

A statistical agency under strain

The shutdown hit an agency already stretched. The BLS suspended price collection in several smaller cities earlier this year because of staffing limits, relying more heavily on imputation from other areas. In August the president dismissed the bureau's commissioner after a weak jobs report, an unusual step that raised questions about political pressure on official statistics. None of that implies November's figures were manipulated, but it means markets approach each technical adjustment with less goodwill than before.

The closest precedent is the pandemic. In the spring of 2020 lockdowns stopped statisticians in many countries, including the BLS, from collecting prices for large parts of the basket, and agencies filled the gaps with imputations. The distortions then were mostly in categories such as travel and restaurant meals, where prices were changing unpredictably. This time the gap covers every category for a full month, which makes the effect broader if shallower.

Disinflation on paper

The carry-forward produced a downward trend in inflation that the economy had not delivered. Retailers had been passing on tariffs and supply-chain surcharges while nobody was counting.

The consequences reach beyond headlines. The Federal Reserve's preferred inflation measure, the personal consumption expenditures price index, draws on many of the same price data, so its November reading inherits part of the problem. Treasury inflation-protected securities are indexed to the consumer price index, so holders are credited with less inflation than they should have been until the catch-up arrives. And any wage settlement or contract linked to the November figure is set slightly too low.

The distortion will also echo a year from now. Annual comparisons for next October and November will be measured against a base that was itself built partly on carried-forward prices, so if the spring catch-up lifts the level of the index, the year-on-year rate next autumn will look a little higher than the underlying pace for purely arithmetic reasons. Statisticians call this a base effect, and it means the shutdown will leave marks on the data long after the missing month is forgotten.

The case for real disinflation

Not every economist thinks the slowdown is an artefact. Market rents for new leases have been flat or falling in many cities for more than a year as a wave of apartment completions came onto the market, and official shelter inflation has been drifting down for some time as those new leases filter into the index. Weaker hiring has also cooled wage growth in services. On that view, the carry-forward exaggerated a real trend rather than inventing a false one.

That argument is correct about direction and incomplete about size. Even economists who accept the rent story generally agree that the October gap pulled the November reading below where a complete survey would have put it. The question is by how much, and the answer will only become clear in the spring, when the skipped housing units are priced again and the missing month's goods price changes work through.

What to watch

The December and January consumer price reports will show whether the downward drift continues once collection is back to normal. Shelter inflation in the spring releases, especially the April figures published in May, will reveal how much was missed. The Fed will also be reading its own business surveys on pricing plans, which were not interrupted by the shutdown and offer an independent check on the official data.

November's 2.7 per cent is a best guess patched over a hole in the data. It flatters the inflation picture, and the price rises missed during the shutdown will probably turn up in the figures over the next few months. The clearest catch-up should come in the spring housing data, which is likely to push measured inflation back up just as the Fed is deciding whether to cut again.