On a clear night, the data came back beautiful: China’s August manufacturing PMI printed at 49.8, up 0.6 points from July, published by the National Bureau of Statistics on August 31. The headline number still sits below the 50 line that separates expansion from contraction, so the first reflex is caution. I will confess the 49.8 gave me a moment of disappointment before I looked inside the components. But a PMI reading is never one number — it is a spectrum, and the spectrum this month carries a signal that has been missing for five months.
The signal no one should miss: orders finally overtook production
For the first time in five months, the new orders index, at 50.6, moved above the production index, at 50.4. New orders rose 2.1 points in a single month, production rose 0.5. Let me say that again slowly, the way I would read a redshift measurement: demand moved first, and production followed. That ordering matters. In the months before this, production had been running ahead of orders — factories kept making things because prior order books and restocking rhythms carried them forward, not because new demand was arriving at the door. An economy can hum like that for a while, but it is a humming that has no fuel. This month, the fuel gauge moved.
New export orders climbed back above 50, to 50.1, returning to expansion for the first time in a long stretch. I spent an evening turning that number over. Exports had been the soft spot all year, the point in the spectrum where the telescope picked up the most noise. A reading of 50.1 is not a bright flare; it is a faint, repeatable signal. But after months of watching that band go dark, I will take the measurement and record it carefully. I should be careful with the export number, though — one month at 50.1 does not rewrite the trade picture, and external demand can move on a coin toss. What it does is reopen a door that had been shut: for the first time in months, the foreign line is no longer dragging the composite down.
How to read a reading below 50
Here is where I have to hold myself honest. A PMI below 50 is still a contraction reading, and I should not dress it up as anything else. The composite barometer is 49.8 — below the line. But the composition of the month matters more than the single value, the same way a spectrum tells you more about a star than its apparent magnitude does. Sixteen of twenty-one industries reported month-over-month improvements. That is roughly three-quarters of the surveyed field moving in the same direction. A composite number can be lifted by a few heavy sectors; breadth is harder to fake. Sixteen of twenty-one is not unanimity — five industries still moved the other way — but it tells me the improvement is not confined to one corner of the factory floor.
Let me check my own arithmetic here. The bureau’s sub-indices: production 50.4, new orders 50.6, new export orders 50.1. That is a coherent picture — all three sitting just above 50, with new orders the strongest of the three. That configuration is the classic signature of a demand-led recovery beginning at the margin. It is not a boom. It is a turn.
The shadow in the spectrum: a widening cost squeeze
Now the shadow. The index for major raw material purchase prices jumped 3.4 points to 56.6, returning to the high end of recent years. The factory gate output price index rose 2.6 points to 50.4, back into expansion. I sat with those two numbers for a while, because the gap between them is the real story. Purchase prices at 56.6, output prices at 50.4 — the spread has widened to 6.2 points.
That spread is the squeeze, and it lands on the midstream. For a manufacturer sitting between raw material suppliers and downstream buyers, a 6.2-point gap means costs are rising faster than selling prices. Some of that can be absorbed, some cannot. When input prices climb toward recent highs while output prices barely clear 50, the margin of the firm in the middle gets compressed. This is the part of the reading that tempers my enthusiasm. The demand signal is real; the cost pressure is also real, and they are not moving in harmony.
Size matters: large firms recovered, small firms did not
The spectrum also carries a size dimension. Large enterprise PMI rose 1.1 points to 50.6, back into expansion territory. Medium enterprises slipped to 49.4, down 0.3 points. Small enterprises rose 0.5 points to 47.9 but remain firmly in contraction. Three size classes, three different worlds.
The divergence also has a calendar logic. Large firms typically lead a cycle because they hold the inventory and the contracts; small firms follow once demand has been confirmed. So the honest reading of this month is that the large-firm expansion is the leading edge, and the small-firm contraction is the lagging edge. If the turn is real, the lag should narrow over the next two readings. If it does not, then the recovery is a large-firm phenomenon and the base of the economy is still waiting for its turn.
Deep time has a way of settling arguments about what these divergences mean. Large firms have pricing power, financing access, and order books that small firms do not. When a recovery is led by large enterprises while small ones sit three points below the line, the recovery is real but uneven — it has not yet reached the base of the pyramid where most employment sits. I try not to overread one month. But the size split is a reminder that a barometer measures pressure at one altitude; the weather below is always more varied.
What the September test will tell us
Now the question everyone with a stake in this reading asks next: will the September peak season deliver? The manufacturing calendar has its own seasons, and autumn is the busy one — the stretch of the year when orders historically pile up. This year the September-to-October window carries an extra weight, because the August reading handed us a hypothesis: demand is beginning to lead production.
The honest framing is that August gives us one month of the pattern, not proof of it. If the September reading confirms — if new orders hold above production, if the export line stays above 50, if the breadth of improving industries holds at fifteen or sixteen — then the turn becomes a trend. If September reverts to the old configuration, production running ahead of orders, then the August blip was exactly that, a blip, and the underlying stagnation reasserts itself. I will not claim to know which it will be. The long view is the only honest view, and the long view says we are one month into a possible turn, not at the end of one.
The September test has a second layer worth naming. Peak-season orders do not materialize by themselves; they are a function of domestic policy support holding up and overseas demand not collapsing. Part of the August improvement may be mechanical — order backlogs from earlier in the year now being executed — which still shows up in the data as demand, and still gives factories a reason to run. The question for September is whether fresh bookings arrive to replace the carry-over, or whether the books thin out again. That is the difference between a handoff and a stall.
Let me be explicit about what I am watching next. Three readings, three coordinates: the new orders index versus production; the purchase-to-output price spread; and the small enterprise index. The first tells me whether demand leadership persists. The second tells me whether the midstream squeeze is tightening or easing — if the spread narrows on a strong order month, the recovery has real legs; if it widens while orders weaken, we are looking at stagflationary pressure in miniature. The third tells me whether the recovery reaches down. All three are published monthly, all three are verifiable, and all three were present in the August release.
A note on method, and on wonder
A last observation about the composite itself: 49.8 is closer to the line than it looks in print. The distance from contraction to expansion, in PMI terms, is the width of a single percentage point, and movements of two or three points in one month are not rare. That is precisely why I read the components instead of the headline. A composite at 49.8 with demand leading is a different animal from a composite at 49.8 with inventories piling up. Same number, opposite weather. The components are the only way to tell them apart, and this month they point one way while the headline points another.
On sources, I have stayed strictly inside the official release. The National Bureau of Statistics published the August PMI on August 31, and every figure I cite — the 49.8 composite, the 50.6 new orders, the 50.4 production, the 50.1 export orders, the 50.6 large-firm reading, the 49.4 medium, the 47.9 small, the 56.6 input prices, the 50.4 output prices, the sixteen-of-twenty-one industry count — comes from that release and the macro note built directly on it. The 6.2-point spread is subtraction, not a new source. No number that I could not trace has been allowed into the field of view.
There is a temptation, in months like this, to let the numbers tell you what you already want to hear. I caught myself doing it — reading 49.8 and wanting to call it a recovery. Then I re-checked the components and the size split, and I had to correct myself: the reading is a mixed spectrum, a dawn rather than a full light. That correction is part of the discipline. Evidence keeps the awe honest; without the correction, the measurement becomes a mirror of desire instead of a record of the sky.
The long view on a single month
Measured wonder, in the end, is what I am practicing here. The August reading is genuinely interesting — a demand-led turn at the margin, after five months of the opposite configuration, is not noise. It deserves to be recorded and it deserves to be watched. But the astronomer’s habit is to take the observation and file it against a longer series, not to build a cathedral on one night’s data. The September and October readings will decide what the August signal was worth.
If I had to compress this month into a single line for the logbook, it would be this: orders have begun to lead production again, but costs are still squeezing the middle, and the small firms have not yet felt the turn. That is a picture of a fragile improvement, not a durable one. It is a dawn you can measure, not a morning you can bank on. The long view settles the argument eventually; for now, the sky says the turn is possible, and the numbers say we are watching.