Producer Prices Look Like Costs. The Trade Services Line Measures Margins.

The Producer Price Index for final demand rose 0.4 percent in August 2026, seasonally adjusted. Inside the same release, published the same morning from the same survey, the index for final demand trade services fell 0.2 percent. Neither number is a revision of the other. Both describe August. They move in opposite directions because they do not measure the same kind of quantity.

The headline is an average of transaction prices and distributor margins blended into one index. Most of its components answer one question: what did a producer charge for this good or service. One whole branch of it answers a different question entirely: how much did the spread between a wholesaler’s buying price and selling price change. A spread is not a price. It can compress while every price inside it is rising, and it can widen while prices fall.

This is not a subtlety buried in a technical note. In August 2026 it produced a clean, legible contradiction on the face of the release: the index for final demand energy goods rose 4.2 percent in the month, and in the very same month margins for fuels and lubricants retailing fell 11.3 percent. Energy got more expensive. The people selling it earned a thinner spread. Both statements are true, both come from the Bureau of Labor Statistics release for August 2026, and a reader who treats the headline as a pure cost gauge will reconcile them incorrectly or not at all.

The August Release Contains Two Different Kinds of Number

Start with the structure. Final demand splits into goods and services. In August the goods side advanced 1.1 percent, driven by the 4.2 percent energy move, with foods up 0.1 percent and goods less foods and energy up 0.4 percent. The services side edged up just 0.1 percent. That single tenth conceals a wide internal dispersion across three branches, and only the second and third of them are price measures.

• Trade services, down 0.2 percent. This branch measures margins received by wholesalers and retailers, not the prices of anything they sell.
• Transportation and warehousing, up 2.3 percent. A genuine price measure, and the largest monthly move of the three branches.
• Services less trade, transportation and warehousing, unchanged. Also a price measure. Flat.

The aggregate arithmetic is what makes this legible. A goods side up 1.1 percent and a services side up 0.1 percent netted to a headline of 0.4 percent. Had the trade services branch been a price index behaving like the transportation branch rather than a margin index moving the other way, the services contribution and therefore the headline would have read differently. The composition below is worth holding in mind for the rest of this piece.

What a Trade Services Index Actually Measures

The convention is deliberate and long-standing. The BLS Handbook of Methods states the reasoning plainly: "The PPI views wholesalers and retailers as suppliers of distributive services (rather than goods) because little, if any, transformation of these goods takes place." What the trade indexes therefore track is "the average changes in gross margins received by wholesalers and retailers," where those margins "reflect the value added by these establishments for services such as marketing, storing, and displaying goods in convenient locations and making the goods easily available for customers to purchase."

The logic is sound. Counting the full retail price of a television inside the PPI would double-count the manufacturer's price, which the survey already captures upstream. Only the distributor's value added belongs in the aggregate. But the accounting choice has a consequence that survives into every monthly print: the trade services line is a difference of two prices, and differences behave unlike levels.

A margin narrows under at least four distinct conditions, and the index cannot distinguish among them. Acquisition costs may rise faster than selling prices, which is the classic squeeze. Selling prices may fall faster than acquisition costs, which is discounting into weak demand. Volumes may shift toward lower-margin product mixes. Or a duty, a freight surcharge or a wholesale energy spike may land on the acquisition side and be absorbed rather than passed on. A single monthly print of -0.2 percent is consistent with all four. The sign tells a reader that the spread compressed. It does not identify which blade of the scissors moved.

That ambiguity is the reason this line deserves separate handling rather than silent inclusion. It is also why BLS itself publishes an aggregate that strips it out, and why the gap between the two is worth measuring directly.

The Gap Between Two Published Aggregates

BLS publishes PPIFIS, final demand, and separately WPSFD49108, final demand less trade services. Both are seasonally adjusted, both cover the same reference month, and the only structural difference between them is the presence or absence of the margin branch. Differencing their monthly percent changes gives a direct, published-to-published measure of how much the margin line pulls the headline around.

In August the headline rose 0.3999 percent and final demand less trade services rose 0.5552 percent. The ex-margin aggregate ran 0.16 percentage point hotter than the headline. Over twelve months the same comparison on seasonally adjusted index levels gives 5.41 percent for final demand against 5.64 percent for final demand less trade services, a gap of 0.23 percentage point. BLS publishes the unadjusted headline at 5.4 percent for the twelve months ended in August, which rounds to the same place.

Two cautions belong with that arithmetic before the chart. First, this gap is the difference between two separately published aggregates, not a formal weighted contribution decomposition; the ex-trade index reweights the remaining components rather than simply subtracting a term. Second, the gap is not a stable drag. It changes sign often, and in both directions.

The chart makes the instability concrete. In May 2026 the headline rose 0.51 percent while final demand less trade services rose 1.32 percent, a gap of 0.81 percentage point and the widest of the eight months shown. One month later the sign reversed: in June the headline fell 0.13 percent while the ex-margin aggregate fell 0.45 percent, so the margin branch worked in the opposite direction. Across the nineteen months from February 2025 through August 2026, the ex-trade aggregate ran hotter than the headline in eleven months and cooler in eight. There is no stable wedge to memorise and subtract.

This is the practical consequence. A single month's divergence between the two aggregates carries almost no information on its own, because the margin branch is noisy in both directions. A persistent divergence across several months does carry information, because sustained margin compression means distributors are absorbing something they cannot yet pass on.

Concrete Framework

The mechanism that connects a cost shock to a falling margin index runs in a fixed order, and each stage operates on its own clock. The acquisition side reprices at the speed of invoices and customs entries. The selling side reprices at the speed of menu changes, contract renegotiations and shelf tags. The margin is the residual between them, so it absorbs the timing difference first and mechanically before any deliberate pricing decision is made.

Turning that mechanism into a reading procedure takes five steps.

1. Separate the two aggregates before forming a view. Read PPIFIS and WPSFD49108 side by side rather than reading the headline alone. The difference between their monthly changes is a gap between two published aggregates rather than a weighted contribution, and in August it was 0.16 percentage point with the ex-margin aggregate on top.
2. Check the sign against the goods side. When goods prices rise and the trade services line falls in the same month, the combination is consistent with absorption at the distribution layer. When both rise together, it is consistent with pass-through plus markup. The August pairing of energy at plus 4.2 percent and fuels retailing margins at minus 11.3 percent sits in the first case.
3. Require persistence before drawing an inference. One month of divergence is inside the normal range shown in the chart above. Three consecutive months of the same-signed gap is a different object, because the gap has otherwise favoured the ex-margin aggregate in eleven of the last nineteen months and the headline in the other eight.
4. Use the ex-foods-energy-trade-services aggregate as the arbiter. WPSFD49116 strips volatile goods and the margin branch at once. It rose 0.3 percent in August and 4.7 percent over twelve months on an unadjusted basis, against 5.4 percent for the headline. When those two measures diverge sharply, the divergence is located in energy, foods or margins rather than in underlying service prices.
5. Date the vintage. Every figure above is a first or near-first print. PPI indexes are recalculated after publication, and the arithmetic in steps one through four should be repeated on the revised vintage rather than assumed to hold.

Step five is not a formality. The BLS calculation methodology describes indexes that "undergo five iterative updates from first issuance through final posting," with final indexes posted "4 months after original publication." Separately, "each calendar year, BLS recalculates the previous five years of PPI seasonal factors based on the most recent eight years of price movements." Both mechanisms bear directly on a margin line: a residual between two larger numbers is proportionally more sensitive to revisions in either of them than a price level is, and the monthly changes charted above are seasonally adjusted values that the annual factor update can move.

Where This Doesn't Apply

Three objections deserve to be stated rather than buried.

The margin line is not noise, and stripping it is not automatically the better read. Distribution is real economic activity, and the value added by wholesalers and retailers belongs in a measure of what the economy produces. An analyst who habitually quotes only the ex-trade aggregate is discarding information about the profitability of an entire sector. The argument here is for reading the two side by side, not for replacing one with the other.

A single month's margin move usually has a mundane cause. The August fuels and lubricants figure is the clearest illustration in the release, but retail fuel margins are mechanically volatile: pump prices adjust on a different cadence from wholesale rack prices, so a sharp move in wholesale fuel costs shows up in the margin with no policy content whatsoever. Reading a duty into that specific number would be an error. The release identifies the margin move; it does not attribute it.

The gap measured here is an approximation. Differencing the monthly changes of PPIFIS and WPSFD49108 is not identical to computing the weighted contribution of trade services to the headline, because the ex-trade aggregate redistributes weight across the components that remain. The two quantities are close and move together, and the differencing approach has the advantage of using only published series, but the figures quoted above are gaps between aggregates and should be described that way rather than as contributions.

There is a fourth limitation worth flagging for anyone extending this to trade policy. A margin index shows absorption in aggregate; it cannot say whether a domestic importer, a foreign exporter or a final customer bore a given cost. That allocation question needs customs and import-price data rather than producer-price data, and those series have their own definitional traps. The import price indexes, for example, exclude duties by construction, which is why a 2026 swing in the value of imports from China reads as a quantity change rather than a price change. The collections side has a matching problem, where refunds running ahead of collections make a single customs line readable two ways. And the statutory authority behind a duty is a separate variable again from its rate, which is why the tariff statute has to be separated from the tariff rate before inflation data can be read against it. Four datasets, four different definitions of the same transaction.

What to Watch Next Week

The next PPI reference month is September 2026, and the observable checks are narrow enough to state in advance.

• The sign of the trade services line against the goods side. If final demand goods rise again while trade services falls for a second consecutive month, the persistence test in step three of the framework moves from one month to two. If trade services turns positive while goods hold, the August print resolves as ordinary fuel-margin volatility.
• The gap between PPIFIS and WPSFD49108. August's gap was 0.16 percentage point with the ex-margin aggregate higher. A September gap of similar sign and size would be the first evidence of something other than noise; a reversal toward the June pattern, where the headline fell 0.13 percent against 0.45 percent for the ex-margin aggregate, would not.
• Whether the fuels and lubricants retailing margin mean-reverts. A minus 11.3 percent monthly move in a retail fuel margin is large, and whether it reverses is an observable question rather than an assumption. A second consecutive double-digit decline would be the more unusual outcome.
• The spread between headline and WPSFD49116. Twelve-month readings of 5.4 percent and 4.7 percent leave a 0.7 percentage point spread, which is the coverage difference between the two aggregates: foods, energy and trade services. Watch whether that spread narrows because the headline cools or because the core-core measure accelerates. Those two paths have opposite implications.
• Revisions to June and July. Under the four-month schedule, earlier months in the chart above are still moving. Any conclusion drawn from the May to June sign flip should be re-tested once those vintages settle.

None of this resolves into a forecast, and it is not meant to. The claim is narrower and more durable than a forecast: a reader who knows which line in the release measures a spread rather than a price will not have to explain away the contradiction between energy at plus 4.2 percent and fuel retailing margins at minus 11.3 percent, because there is no contradiction to explain. The two numbers were answering different questions all along.