The inflation drop that came from a rulebook

America's inflation figures for August looked like good news. The personal consumption expenditures price index, the measure the Federal Reserve targets, rose 3.4 per cent in the year to August, well below the 3.7 per cent economists had pencilled in. The core index, which strips out food and energy, rose 3.0 per cent against forecasts of 3.3 per cent. Futures traders cut the odds of a Fed rate rise in October to about 35 per cent, from around half the day before. The bond market did not join the celebration. The yield on the 10-year Treasury note rose by more than four basis points on Wednesday, to about 5.3 per cent.

There is a reason for the split. Most of the improvement in the headline numbers came from a change in how they are calculated, not from a change in how fast prices are rising.

A new ruler, applied backwards

The Bureau of Economic Analysis, which compiles the figures, used Wednesday's release to change its methods for three categories of spending: portfolio management and investment advice, legal services, and computer software and accessories. The changes were applied to the data all the way back to the start of 2021, so the history moved along with the latest month.

Each fix has a technical rationale. For investment advice, the bureau now estimates the quantity of services consumed from employment in the industry, which should stop swings in asset prices from showing up as swings in the price of advice. For legal services, it replaced unpublished consumer-price figures supplied by the Bureau of Labor Statistics, which the bureau said did not meet the statistical agency's own quality standards and had moved erratically, with an index built from producer prices. For software, it widened the set of prices it tracks to include video games and web hosting.

The combined effect was large. July's core inflation rate, first reported at about 3.35 per cent, was revised down to about 2.98 per cent. Headline inflation for July fell from 3.70 per cent to 3.36 per cent. The monthly rise in core prices for July was cut roughly in half, from 0.25 per cent to 0.12 per cent.

Like for like, nothing fell

Once the revisions are taken into account, August looks rather different. The core rate of 3.0 per cent is essentially the same as July's revised 2.98 per cent. The monthly increase in core prices, about 0.25 per cent, was double July's revised pace. Annualised, a month like August works out at roughly 3 per cent, half as much again as the Fed's 2 per cent target.

The forecasts were the other part of the story. The consensus forecast of 3.3 per cent matched July's unrevised rate, which suggests many economists were still working with the old method. Measuring the outcome with the new ruler produced a "miss" that says more about the statistics than about the economy. Analysts at RBC made this point before the release, warning clients not to read a lower-than-expected number as a sign of disinflation ahead.

The Fed was not taken by surprise either. Its own economists had estimated in advance how much the software change alone would lower measured inflation. A revision that everyone at the central bank saw coming is a weak reason for the central bank to change course.

Where the pressure still is

Beneath the totals, the mix is uncomfortable. Prices of core services, which reflect wages and rents and tend to move slowly, rose 3.35 per cent over the year. Core goods prices rose about 2 per cent. Energy prices were up 16.8 per cent from a year earlier, as tension with Iran over shipping through the Strait of Hormuz has kept oil expensive. Energy is excluded from the core measure, but it feeds into transport, airfares and the inflation that households expect, which the University of Michigan's survey showed rising sharply in September.

Fed officials sounded patient rather than relieved. John Williams, president of the New York Fed, said he saw no urgency to raise rates again and suggested one more increase late in the year, a hint that December is more likely than October. Michael Barr, a Fed governor, observed that in the past 20 months only two months of data had been consistent with 2 per cent core inflation, and pointed to energy prices and heavy investment in artificial intelligence as reasons for caution.

Why long yields rose on soft data

The move in Treasuries makes sense in this light. The two-year yield, which tracks expectations for the Fed's next few decisions, was little changed at about 4.9 per cent. The 10-year yield rose. A market that thought inflation had genuinely broken would have pushed long yields down. Instead investors read the data as a slightly gentler path for short-term rates over a level of inflation that has not moved, and a Fed less inclined to fight it hard. That is a recipe for a steeper yield curve, with more of the risk priced into the long end.

It is the same pattern this column described on Tuesday. The rise in long yields reflects a term premium, the compensation investors demand for holding long bonds, as much as it reflects the Fed. Large deficits, heavy Treasury issuance and an oil shock that has no clear end all argue for a higher premium. A change in how legal fees are deflated does nothing to any of them.

The case for taking the numbers at face value

None of this means the new figures are wrong. The old legal-services data were, by the statisticians' own account, of poor quality, and a software price index that leaves out a large chunk of what people buy is not much of an index. If the revised series is the more accurate one, then inflation since 2021 has been somewhat lower than reported, and the Fed has been fighting a slightly smaller problem than it thought. On that reading, the three-month annualised rate of core inflation, which on the revised data is about 2 per cent, is a genuine sign that pressure is easing, and the Fed can afford to wait.

The weakness in that argument is timing. A lower level of inflation across five years of history does not mean inflation is falling now. The revisions lowered the starting point and the end point together. What matters for policy is the direction from here, and August's monthly reading pointed the wrong way.

What to watch

The September jobs report, due later this morning, is the next test. Strong hiring and firm wage growth would revive the odds of an October rise quickly, whatever the methodology. The consumer price index for September, out in mid-October, is calculated by a different agency and was not touched by this week's changes, so it offers a cleaner read on whether prices are still accelerating. Oil and the talks with Tehran remain the largest single source of risk to both measures.

Wednesday's report changed the measuring tape, not the thing being measured. Core inflation is running at about 3 per cent on the new basis, much as it was on the old one once the history is restated. Until the monthly figures start to fall on a like-for-like basis, the Fed is likely to keep its bias towards tightening, and the bond market is likely to keep asking to be paid for the risk.