Bottom line
The price side has turned, but broad domestic-demand confirmation is still absent. The evidence is more consistent with an upstream- and external-price-led nominal rebound than with a durable, demand-led reflation. The implied GDP deflator turned positive at +1.5% year on year in 2026Q2, from -0.1% in 2026Q1, after negative readings in every quarter from 2023Q2 to 2026Q1.[1] That is a real change after three years of deflation.
What has not happened yet is breadth. Household credit, property and core inflation have not followed prices, and the leading competing explanation, H0 (the rebound is driven by supply-side discipline and imported costs), still fits the pattern better than a demand-led recovery. Realized policy impulse remains weak in the quantitative data, while the forward policy stance has become more supportive (section 4).
Judgment: 30% analyst probability that the thesis holds at the 2027Q1 decision window. This is a discretionary judgment on a five-point grid, not a statistically calibrated model probability. Confidence low · status ACTIVE.
How to read this note
Every indicator is read through four lenses. Level is where it stands. Marginal change is which way it is moving now, which cumulative series can hide. Breadth is whether the move is shared across sectors and demand channels. Mechanism is what drives it: supply, imported costs, policy or demand. Confusing them is how a price rebound gets mistaken for a demand recovery.
1. The question and how we measure it
Is China forming a sustainable turning point in nominal growth? We measure the turn through the implied GDP deflator, which is nominal GDP growth relative to real GDP growth. Monthly price, demand, credit, property, external and market series serve as leading or confirming evidence. "Sustainable" is defined in advance: the deflator must be non-negative in both 2026Q4 and 2027Q1, and at least three of six domestic-demand tests must pass at the 2027Q1 release. The thesis is therefore falsifiable on two separate legs: price and demand.
2. Prices: the level has turned (H1)
PPI was +3.8% year on year in 2026-08, against -1.9% in 2025-12: factory-gate prices have moved from deflation to inflation within the year.[2] Industrial profits were +15.7% year on year, year to date, in 2026-08, after +0.6% for 2025 as a whole.[3]
Level and momentum tell different stories. Month-on-month PPI momentum stalled over the summer. Month-on-month PPI momentum has faded: +1.7% in 2026-04, then -0.3% in 2026-06, -0.7% in 2026-07 and +0.4% in 2026-08.[4] Cumulative series flatter the picture in the same way. Year-to-date industrial profit growth has slowed in every month since May: 18.8% in 2026-05, 18.7% in 2026-06, 17.6% in 2026-07 and 15.7% in 2026-08.[5] Because year-to-date profit growth has been falling since May, growth in the latest single month is running below the cumulative rate. The headline year-to-date figure therefore overstates current profit momentum.[6] The September survey suggests factory-gate momentum is picking up again. The PMI output price index rose to 54.0 in 2026-09 from 47.8 in 2026-07, a sign that factory-gate momentum picked up again going into the September PPI release.[7] The official PPI print on the first release after the baseline is the test. Consumer prices remain subdued: CPI was +0.8% and core CPI +1.0% year on year in 2026-08.[8]
3. Supply or demand? (H0 versus H2)
This is the central question. H0 is the leading competing explanation: the rebound is driven by supply-side discipline and imported commodity costs. The data are consistent with it, but they do not prove the mechanism.
Mechanism (H0). The largest PPI gains coincided with the oil spike: PPI moved +1.0% and +1.7% month on month in 2026-03 and 2026-04, when Brent averaged 99 USD and 103 USD.[9] Brent averaged 104 USD a barrel in 2026-05 and 84 USD in 2026-07, against 62 USD in 2025-12.[10] The PPI rise is upstream: producer-goods PPI was +5.0% year on year in 2026-08 while consumer-goods PPI was still -0.5%.[11] Copper averaged 13,543 USD a tonne in 2026-07, against 9,771 USD a year earlier.[12]
Level and breadth (H2). Property sales area was -12.1% year on year, year to date, in 2026-08: a steeper decline than the -8.7% recorded for 2025 as a whole.[13] The year-to-date decline in property sales area has widened in every month since April: -10.2% in 2026-04, -10.8% in 2026-05, -11.6% in 2026-06, -11.8% in 2026-07 and -12.1% in 2026-08.[14] Because the cumulative decline has widened every month since April, sales in the latest single months are weaker than the year-to-date rate. The change-based property test that passes compares August with the weak January-February starting point, so that pass reflects the weak early-year starting point more than a recent improvement.[15] Net new medium and long-term household loans were 1,584 in 2026-06, -1,202 in 2026-07 and -822 in 2026-08 (CNY 100m, derived from PBOC balances): mortgage books are still shrinking on balance.[16] Retail sales grew 0.4% year on year in 2026-08, against 3.4% a year earlier.[17] M1 growth was 4.1% in 2026-08, down from 5.9% in 2026-02.[18] Core CPI was +1.0% in 2026-08, no higher than the +1.2% of 2025-12: no sign yet of demand-pull inflation.[19]
Forward-looking signals. Orders send a mixed message. The manufacturing PMI new orders index was 50.5 in 2026-09, but its three-month average was 49.9: orders are hovering at the expansion line rather than expanding.[20]
Reading. So far the price turn looks supply-led. It is concentrated in producer goods, it coincides with the oil and metals shock, and it lost momentum when oil eased. That fits H0 better than a demand-led recovery.[21] Broad domestic-demand confirmation is still absent. By level, property sales, household mortgage flow, M1 and core CPI remain weak. The demand tests that do pass do so on marginal changes from weak levels, and the one forward-looking improvement, September manufacturing new orders above the expansion line, is not yet confirmed by its three-month average.[22]
4. Policy and external demand (H3, H4)
Policy has to be read in two layers. Realized impulse, measured in credit and fiscal data, remains weak. Total social financing stock growth slowed to 7.2% in 2026-08 from 8.3% in 2025-12.[23] General budget expenditure grew only 1.3% year to date in 2026-07.[24] Policy impulse is fading rather than building: credit growth is decelerating and on-budget spending is barely growing, so policy is not yet bridging the gap left by private demand.[25] The forward stance has become more supportive: a late-September State Council executive meeting called for stronger counter-cyclical adjustment and a batch of incremental measures, and at the end of September the NDRC reported that the full year's consumer trade-in funding had been allocated. These are announcements and allocations, not yet realized impulse, and they were not part of the information in the frozen baseline judgment. They are listed with their official sources and stage in the policy context on the Evidence page.
Exports grew 25.0% year on year in dollar terms in 2026-08, after 23.9% in 2026-07.[26] The PMI new export orders index was 50.0 in 2026-09, at the expansion line, so the export strength is not yet showing up as broad new orders.[27] Exports are a genuine support for nominal growth. But external demand cannot by itself make the turn sustainable under this thesis, which requires domestic demand confirmation.[28]
5. Global and market cross-check
The US 10-year Treasury yield averaged 4.99% in 2026-09, up from 4.14% in 2025-12, while the renminbi strengthened to 6.71 per dollar from 7.04.[29] A stronger renminbi alongside higher US yields gives the PBOC some room. A stronger currency also dampens imported inflation, which matters if the price turn depends on imported costs. The China 10-year government bond yield is the market's own read of nominal growth. It has no redistributable public source and is therefore not part of this public build.
6. Judgment
The price leg of the thesis is more likely than not to hold through 2027Q1. The factory-gate price level is well above a year earlier, so the deflator would turn negative again only if commodity prices fell back sharply.[30] The demand leg is the weak link. At the baseline, fewer demand tests passed than confirmation requires, and those that passed did so on marginal changes from weak levels. We therefore put the analyst probability that the thesis is confirmed at 30%, with low confidence. This is a judgment, not a model output. Confidence is low for three reasons: GDP history is a revised vintage rather than first prints, consensus forecasts are not available in the public build, and both oil and exports can swing sharply.
7. What would change our mind
| Release | What we will check | Moves |
|---|---|---|
| September CPI and PPI | PPI momentum, PPI versus six months ago, core CPI, consumer-goods PPI | H1, H2, H0 |
| Third-quarter GDP and September activity | implied deflator for 2026Q3, retail sales | H1, H2 |
| September credit data | household mortgage flow, M1, credit growth | H2, H3 |
Toward confirmation: core CPI turning up, consumer-goods PPI turning positive (pass-through), net mortgage lending turning positive and, on policy, incremental measures showing up as faster government-bond financing and budget spending.
Toward falsification: PPI momentum and demand failing together (the early falsification rule), or the deflator slipping back below zero once oil base effects fade.
8. Research implications (conditional)
These are conditional research implications of the macro scenarios, not recommendations and not forecasts of market levels. The system tests the macro scenarios against preregistered rules; it does not test the market implications, which follow from them.
| Upstream-led rebound fades (H0 holds) | Policy-supported stabilisation | Demand-led reflation (H2 confirms) | |
|---|---|---|---|
| China rates | Little case for pricing a durable reflation; long yields are more likely to follow growth and policy signals than inflation expectations. | The net effect on yields is ambiguous: more bond supply and a firmer growth outlook push yields up, while easier monetary tools and weak private credit pull them down. | Rising nominal growth and inflation expectations would push long yields higher and steepen the curve. |
| CNY | No domestic-reflation support; direction depends more on the US rate path and the external balance. | Easier monetary tools would lean against CNY strength, a firmer growth outlook would support it; the net direction also depends on the US rate path. | A firmer domestic growth and rate outlook would be supportive, within the limits set by the policy response to higher yields. |
| Broad equities and cyclicals | Upstream pricing power matters more than domestic demand; downstream margins are squeezed if input costs stay high while consumer prices lag. | Investment-linked and policy-supported cyclicals benefit first; broad earnings upgrades need a private-demand response. | Earnings upgrades broaden from upstream to domestic-demand sectors and cyclicals. |
| Banks | A price-only turn adds little to loan demand or net interest margins; asset quality and property exposure remain the swing factors. | Government-led credit supports volumes more than margins; property and local-government exposure remain the main risks. | Loan growth and net interest margins improve if household and corporate credit demand recover, and asset-quality pressure eases. |
| Confirming evidence | Consumer-goods PPI stays negative; core CPI and household mortgage flow stay weak; oil and metals ease and PPI momentum fades again. | Credit growth and government-bond financing stabilise; budget spending and project starts accelerate; trade-in categories lift retail sales. | Core CPI and consumer-goods PPI turn up; mortgage flow turns positive; M1 and retail sales accelerate; at least three of the six demand tests pass. |
| Falsifying evidence | Core CPI and consumer-goods PPI turn up; household credit and retail sales improve; the three-month average of new orders moves above the expansion line. | The announced measures prove small or slow to be funded; credit growth keeps decelerating; retail sales fade as trade-in funds are used up. | PPI momentum fades while the demand tests fail together; the deflator slips back below zero. |
The scenarios are not mutually exclusive in time: a policy-supported stabilisation can precede either of the other two.
Limitations
GDP and history are current revised vintages, not first prints. Release dates for history are recorded only where an official calendar confirms them; otherwise availability is taken as our retrieval time. Two diagnostics (government fund revenue, exports to the US) have no public route and are absent. The China 10-year yield has no redistributable public source and is not in this build; its licensed source (Wind) is used only in the researcher's local private research mode, and nothing derived from it is published. Policy events are curated by hand from official sources and are context only.