For AUD traders, the essential takeaway is that China's PMI complex delivers four data points a month, not one, and each carries a different weight. Manufacturing surprises—especially from the NBS survey—typically provoke the sharpest, fastest market reaction because they offer a direct read on the heavy-industry and construction-linked demand that drives Australian commodity exports. Services and non-manufacturing prints usually burn more slowly, but they matter for the domestic demand story, above all the construction sub-index inside the NBS Non-Manufacturing PMI, which speaks directly to Chinese property activity and its knock-on effect on steel and iron ore consumption.

The private Caixin/RatingDog services survey, oriented toward consumption and smaller firms, can sometimes tell traders more about the health of Chinese household spending than either official gauge. When manufacturing and services data from both providers move in the same direction, conviction behind AUD positions builds quickly; when they diverge—particularly between the factory and services sides—the market tends to seek confirmation elsewhere before committing.

---

China publishes four PMIs a month across two providers and two sectors, meaning AUD traders who watch only the manufacturing headline are missing half the picture.

Summary

China's PMI complex has four regular components each month: the official NBS Manufacturing PMI, the NBS Non-Manufacturing PMI covering services and construction, the private Caixin/RatingDog Manufacturing PMI, and the private Caixin/RatingDog Services PMI.

Both providers also publish composite output indices that blend their manufacturing and services readings into a single gauge of overall economic activity.

The NBS surveys draw on large panels weighted toward bigger, state-owned and state-linked enterprises, while the private surveys use smaller panels skewed toward small and medium-sized, often privately owned, export- or consumer-facing firms.

All the indices share the same basic method—a diffusion index built from weighted sub-components, with 50 marking the line between expansion and contraction—though the underlying questions and weightings differ slightly by sector and provider.

The NBS Non-Manufacturing PMI's construction sub-index provides a particularly direct read on Chinese property and infrastructure activity, which feeds through to demand for Australian iron ore, coal, and other commodities.

The private services survey tends to skew toward consumer-facing activity, so it can offer a different—and sometimes earlier—signal on Chinese household demand than the state-weighted NBS services gauge.

This week's data illustrates the pattern: the NBS Manufacturing PMI for August rose to 49.8 from 49.2, while the NBS Non-Manufacturing PMI held at 49.0. The private manufacturing survey is due today, with consensus expecting a further edge higher from July's 50.9 (NBS factory beat, Caixin PMI preview).

China is unusual among major economies in publishing not one but two full sets of Purchasing Managers' Index surveys each month—one official and one private—and each of those covers both manufacturing and services. The two sets also arrive on a staggered schedule: the NBS releases land on the last day of the reporting month or the first day of the following month, while the Caixin/RatingDog prints follow a day or so later for manufacturing and around the start of the third business day for services. That sequencing means the official data usually sets the initial tone, with the private surveys acting as a confirmation or a challenge within days. Understanding how the four resulting data points relate to each other, and to the composite indices built from them, helps explain why the Australian dollar sometimes reacts sharply to one release and barely move on another.

The official side comes from China's National Bureau of Statistics (NBS), working with the China Federation of Logistics and Purchasing. It publishes a Manufacturing PMI and a separate Non-Manufacturing PMI—the latter covering both services and construction—with the two combined into a Composite PMI Output Index. The panels behind these surveys lean toward larger enterprises, including a significant share of state-owned and state-linked firms. That skew makes the NBS Non-Manufacturing PMI's construction sub-index especially informative, since it offers a fairly direct line of sight into Chinese property and infrastructure activity, a sector with an outsized effect on demand for Australian iron ore, coal, and other bulk commodities.

The private side—long published as the Caixin PMI and now branded RatingDog in partnership with S&P Global—mirrors that structure with a Manufacturing PMI, a Services PMI, and a blended composite. Its panels are smaller, generally around 500 firms, and weighted more heavily toward small and medium-sized, privately owned businesses. On the manufacturing side, that skew favors export-oriented producers; on the services side, it captures more consumer-facing activity, from retail and logistics through to travel and technology-related services. That makes the private services gauge a useful—sometimes earlier—cross-check on the state of Chinese household demand, an area the state-weighted NBS panel is less well positioned to capture cleanly.

All four headline indices are built the same way: as diffusion indices assembled from weighted sub-components such as output, new orders, employment, and prices, with a reading above 50 signaling expansion and below 50 signaling contraction. Because the method is shared, the manufacturing pair and the services pair tend to move broadly together over time, since they ultimately measure the same economy. Divergences are nevertheless common and often informative in their own right—for instance, when infrastructure spending lifts state-linked manufacturers faster than it reaches smaller private firms, or when export orders strengthen even as domestic consumption remains soft.

For traders using the Australian dollar as a liquid proxy for China exposure, all four data points carry information, but not equally. Manufacturing surprises, particularly from the NBS survey, tend to generate the fastest and sharpest reaction given how closely they track the commodity-intensive, capital-heavy side of the Chinese economy. Non-manufacturing and services releases are usually slower burners for AUD, but the construction detail within the NBS Non-Manufacturing PMI deserves particular attention given its bearing on the property sector, while a strong or weak private services print can shift the broader read on Chinese consumption trends. In practice, the PMIs also land earlier than China's hard data—monthly industrial output, retail sales, and fixed-asset investment figures, which typically follow around mid-month—so the surveys are often the first structured read on the economy each month, with the harder figures arriving afterward to confirm or contradict them.

This week has offered a live illustration of that interplay: the NBS Manufacturing PMI improved to 49.8 in August from 49.2 even as the NBS Non-Manufacturing PMI held flat at 49.0, underlining that a pickup on the factory floor has not yet been matched by strength in services and construction. Today's private manufacturing survey—expected to edge further above July's 50.9—will add another piece to the picture, and traders would do well to weigh it, and its services counterpart when released, alongside the official data rather than in isolation.

---

What is a diffusion index?

A diffusion index is a way of turning a survey of qualitative opinions—"did things get better, worse, or stay the same?"—into a single number that can be tracked over time. PMIs are the best-known example, but the technique dates back much further, to the 1940s and 50s, when the concept was developed to summarize business-cycle surveys at a time when actual production or sales figures were not yet available on a timely basis.

How it's built

Each month, purchasing managers are asked whether a given indicator—output, new orders, employment, and so on—has increased, decreased, or stayed the same compared with the prior month. The index is then calculated as:

Diffusion index = (% reporting an increase) + (0.5 × % reporting no change)

Respondents reporting "no change" receive half weight because they neither add to nor subtract from the direction of travel; they pull the index toward the neutral midpoint rather than counting as a full vote either way.

Why 50 is the line

If every single respondent reported an increase, the index would read 100. If everyone reported a decrease, it would read 0. If results were evenly split between more, fewer, and the same—or if literally nothing changed for anyone—the index sits at 50. That is why 50 is treated as the boundary between expansion and contraction: above it, more firms are seeing improvement than deterioration; below it, the reverse.

What it captures, and what it doesn't

The key point is that a diffusion index measures the breadth of change, not its size. A PMI of 55 means more firms are seeing growth than not, but it says nothing about how much output has actually grown: a small increase reported by 60% of firms produces the same lift to the index as a large increase reported by 60% of firms. That is part of why PMIs are best read as sentiment and momentum indicators rather than as a proxy for GDP growth itself, and why they are valued for coming out early each month—well before harder production or trade data—even though they survey opinion rather than actual output.

Why this matters for reading NBS versus Caixin/RatingDog

Because composite PMIs are simply weighted blends of several component diffusion indexes—output, new orders, employment, delivery times, and inventories—a headline move can be driven by very different things underneath. Two PMIs can print the same headline number for entirely different reasons: one because more firms saw modestly stronger orders, another because fewer firms saw sharply weaker output. That is exactly the kind of detail that explains why the sub-indices, not just the headline, are worth watching when NBS and Caixin/RatingDog readings diverge.