China's H1 2026 Logistics Value Tops RMB 181 Trillion, Outpacing GDP Growth

The Analysis

The China Federation of Logistics and Purchasing published its first-half 2026 figures on 29 July, and before anyone does anything with them, the nature of the document matters. Statistical releases are not rules. Nothing in this release creates an obligation, changes a threshold, or moves an effective date. What it does is establish a set of published reference points that other people will start quoting at you: in tenders, in internal forecasts, in policy consultations, in a dispute about whether the market was strong or weak in a given quarter. That is the reason to read it carefully rather than skim it. A statistic that nobody relies on is harmless. A statistic that ends up written into a contract clause is not.

The two numbers that will travel farthest are the headline and the gap. Total social logistics value came to RMB 181.1 trillion for the first half, up 5.1 per cent year on year, which the release puts 0.4 percentage points above GDP growth over the same period. Logistics industry revenue was RMB 7.2 trillion, up 4.7 per cent. The logistics prosperity index averaged 49.9 per cent, up 0.3 points from the first-quarter average. Every one of those is a different instrument measuring a different thing, and I have watched them get swapped for one another in board papers for years.

My own reaction, reading it on the day, was less about the 5.1 and more about the shape of the year inside it. The release gives 6.2 per cent for the first quarter and 4.4 per cent for the second. That is a steep deceleration inside a single half-year, and it is the kind of thing that gets averaged away in a headline number and then surprises people when the quarterly series catches up with them. Also note that 6.2 and 4.4 do not average to 5.1. The half-year rate is not a simple mean of the two quarterly rates, which is worth remembering before anyone in your organisation tries to rebuild a quarterly series out of the half-year figure and presents it as fact.

Start with what the headline actually measures, because this is where most misreadings begin. Total social logistics value is a measure of the value of goods moved through the logistics system over the period. Strictly speaking, it is not the size of the logistics market, it is not logistics spending, and it is not logistics cost. The same good can be counted more than once as it passes through successive stages, so the figure is designed to track the flow of goods through the economy, not to price the services that move them.

If you want something closer to an industry revenue measure, the release gives you a separate line for it: logistics industry revenue of RMB 7.2 trillion, up 4.7 per cent. Those two numbers differ by orders of magnitude and they move for different reasons. Putting 181.1 trillion into a slide about how much logistics costs is an error I have seen more than once, and it is the kind of error that gets quoted back at you later.

On the 0.4 percentage points. The release says logistics value grew 0.4 points faster than GDP over the same period. Don't rush to a conclusion about what that means. A gap of four tenths of a percentage point is small enough that it sits close to the range where revision matters, and it is a comparison between two aggregates compiled by different methods. What I would take from it is narrow and defensible: logistics activity did not lag the broader economy in the first half of 2026, and it did not lead it by much either. If someone builds an investment case, a budget or a contract clause on the strength of four tenths of a point, the premise is thinner than the argument built on it.

Here is the exception, and it is the number most likely to be read backwards. The logistics prosperity index averaged 49.9 per cent, which the release describes as up 0.3 points from the first-quarter average. That arithmetic puts the first-quarter average at 49.6. The index is constructed around a neutral line at 50, with readings above it indicating expansion relative to the previous period and readings below it indicating contraction. A reading of 49.9 is below the line. What the release shows is therefore not a return to expansion. It shows contraction continuing at a slower pace. Recovering by three tenths of a point while remaining under 50 is a materially different statement from having crossed back over it, and the two get conflated constantly. If your planning documents describe the first half as a recovery in industry conditions, go and check which of those two things you actually meant.

On the quarterly slope, which is where I would put most of my attention. Six point two in the first quarter and four point four in the second is a fall of 1.8 percentage points inside six months. On my time scale, quarters matter as evidence and years matter as trend, and one half-year does not establish a trend on its own. But it does establish that the half-year number is not a stable description of either quarter inside it. Any contract, forecast or internal target set against a single half-year growth rate inherits an averaging error before it starts. The practical fix is unglamorous: work off the quarterly series where the release gives it to you, and say explicitly which period you are anchoring to.

The structural detail is where this release earns its keep. High-tech manufacturing logistics demand rose 13.3 per cent and digital product manufacturing 12.3 per cent. Against that, crude oil import volume fell 11.4 per cent and steel 11.3 per cent. Meanwhile semiconductor manufacturing equipment imports rose 20.4 per cent by volume and integrated circuits 8.1 per cent. Read together, that is not a market moving in one direction at one speed. It is two markets moving in opposite directions at double-digit rates, and the distance between them is over twenty percentage points. For anyone trying to plan off an industry average, the average is close to useless here. The structural split is the signal; the 5.1 is not.

The corridor figures sit underneath that split and they are worth having on hand. National railway freight volume was 2.015 billion tonnes for the half, up 1.8 per cent. China-Europe freight train departures reached 11,178 in the period, up 20 per cent, reaching 26 countries and 236 cities, described in the release as a record for the comparable period. Civil aviation cargo and mail volume was 5.073 million tonnes, up 6.0 per cent, with international cargo and mail at 2.32 million tonnes, up 13.9 per cent. Note the pattern: the modes exposed to cross-border and high-value traffic are growing far faster than the domestic bulk modes. Rail freight at 1.8 per cent and international air freight at 13.9 per cent are describing two different economies.

On effective dates and applicable preconditions, and this is the part I care about most. None of the figures above is the basis for applying any rule. Every instrument you operate under has its own precondition, its own scope and its own commencement date, and none of them is triggered or switched off by a statistical release. Let me be concrete about the trap, because the import numbers invite it. Crude oil import volume down 11.4 per cent and steel down 11.3 per cent do not relax anything.

Classification obligations, licensing requirements, inspection and quarantine treatment, and controlled-item screening attach to the goods and to the transaction, not to whether the national total went up or down. A fall in national import volume of a commodity does not change the precondition for any single shipment of that commodity. I have seen volume declines cited as evidence that enforcement posture would ease. That inference has no basis in the way these instruments are written, and I would not let it into a memo.

There is a mirror-image version of the same error on the growth side. Semiconductor manufacturing equipment imports up 20.4 per cent by volume and integrated circuits up 8.1 per cent describe a flow that is getting denser, not a flow that has been cleared. If anything, higher volumes in categories that attract classification sensitivity and export-control screening mean the evidentiary burden on individual shipments goes up rather than down, because there is more of the thing that has to be documented correctly. The applicable precondition for any relief or facilitation you might claim is set in the instrument that grants it. The statistics tell you nothing about whether you meet it.

On evidence and burden of proof. The party asserting a fact generally carries the burden of proving it, and in this area the assertion is usually yours. If you claim that your costs rose in line with the industry, that your sector contracted, that a corridor was congested, or that a delivery window was commercially impossible, you are making a claim about your own circumstances. An industry average is not evidence of your circumstances. It is context, and context has a habit of being admitted and then weighed at very little. What carries weight is your own contemporaneous record: your own volumes by month, your own rates by lane, your own dwell times, your own exception logs. The release published on 29 July can sit beside that record as corroboration of the general environment. It cannot stand in its place.

That distinction becomes concrete the moment a statistic ends up inside a contract. Suppose a framework agreement adjusts rates by reference to the logistics prosperity index. Now the index is no longer commentary, it is an operative term, and three questions immediately become live. Which reading, the monthly print or the half-year average, which came to 49.9. Which publisher and which release, because more than one body publishes indices in this space. And whether the figure is the initial publication or the revised figure, because initial publications get revised. If the clause does not answer all three, then on a dispute each side will produce a different number from the same named index, and the tribunal will be left to decide which reading the parties meant. That is an evidentiary problem created entirely by drafting, and it is cheap to prevent and expensive to litigate.

On records. I am not going to tell you how to make an entry or how to file anything; that is not my lane. What I will describe is the shape of a record set that lets you answer a question in 2028 about the first half of 2026 without reconstructing it from memory. Keep a monthly volume and value ledger by product family, kept on the same family definitions as your external reporting so the two can be reconciled rather than argued about. Keep a mode-level ledger separating rail, road, air and China-Europe train movements, with the corridor and the operator named, because the corridor figures are where the volatility lives.

Keep an import-structure record for categories that attract classification or control sensitivity, naming the supplier, the classification relied on and the basis for it. Keep carrier and intermediary qualification files with the dates they were checked, not the dates they were issued. Keep the exception log: what went wrong, when it was known, what it cost, and who decided the response. And keep a one-page caliber note recording how each figure in your own reporting was derived, updated whenever the derivation changes. That last one is the document that makes the rest of the file legible to a stranger.

A calculation, with the assumptions stated. Assume a company incurred RMB 60 million of logistics cost in the first half, of which 40 per cent, or RMB 24 million, relates to high-tech manufacturing product families and 60 per cent, or RMB 36 million, relates to bulk and raw-material-linked families. Assume the first group moves in line with the 13.3 per cent logistics demand growth in the release, with unit rates unchanged, so it becomes 24 million times 1.133, which is RMB 27.192 million, a rise of RMB 3.192 million. Assume the second group falls in line with the 11.4 per cent decline in crude oil import volume, and assume unit rates ease 1 per cent on softer demand, so it becomes 36 million times 0.886 times 0.99, which is RMB 31.577 million, a fall of RMB 4.423 million.

The net change is minus RMB 1.231 million on a base of 60 million, about minus 2.1 per cent. Now notice what that produces. Industry revenue grew 4.7 per cent over the same period and this company's logistics cost fell 2.1 per cent. Same economy, opposite signs, and neither number is wrong. The average simply does not describe this company.

Invert it, because the threshold is more useful than the example. Let the share of the high-tech group be x. Total cost is flat when 0.133 times x equals 0.1229 times (1 minus x), where 0.1229 is the combined effect of the 11.4 per cent volume fall and the 1 per cent rate easing. Solving gives x of about 48 per cent. Below roughly 48 per cent exposure to the fast-growing families, this company's logistics cost falls on these assumptions. Above it, the same company's cost rises. That is the number I would want on a planning document, because it converts a set of national growth rates into a decision boundary specific to one product mix. Change the assumptions and the boundary moves, which is precisely why the assumptions have to be written down next to the answer.

A second calculation, on the index, and it is a drafting problem rather than a costing one. The half-year average is 49.9 against a neutral line of 50, a distance of one tenth of a point. Assume, as an illustration, that the initial monthly prints of this index are subject to revision of up to half a point. On that assumption a trigger set at exactly 50 sits inside the revision band, and a clause built on it will fire or fail to fire on noise rather than on conditions. Moving the trigger to 48.5 on the downside and 51.5 on the upside puts roughly a point and a half between the neutral line and the trigger, which on the stated assumption is outside the revision band.

The quantified drafting target I would set is that every index-linked trigger carries a stated source, a stated reading, a statement on whether revised figures apply, and a band of at least one full point between the neutral line and the trigger. I would also want a fallback for the case where the index is discontinued or rebased, because indices get rebased and clauses rarely say what happens then.

What I would actually do, on a quarterly and annual scale. Before the third-quarter close, I would want the caliber note written and signed: which of the published figures your internal reporting uses, what each one measures, and what it does not measure. Owner: whoever signs the management report, not whoever builds the spreadsheet. By the fourth quarter, I would want the product-family ledger reconciled once against external reporting, with any difference explained in writing rather than plugged. The target I would set is zero unexplained reconciling items above one per cent of the reported figure.

Before the year closes, I would want every contract clause that references an external index reviewed against the four drafting elements above, with a target of one hundred per cent of such clauses carrying a named source and a stated revision rule. And on an annual cycle, I would re-derive the exposure threshold in the first calculation from the year's own mix, because a boundary computed on last year's product mix is a boundary about last year.

On alternatives and the traps around them. Changing the corridor is the obvious response to a release showing China-Europe train departures up 20 per cent into 26 countries and 236 cities, and I am not going to advise you whether to do it, because that is a commercial judgement and not mine to make. What I will say is what it does to your file. A different corridor changes the custody chain, the documents that evidence each leg, the governing liability regime for each segment, and the set of transit jurisdictions whose rules can touch the goods.

Each of those produces records you do not currently keep, and the evidentiary value of the old file does not transfer to the new route. The same caution applies to switching sourcing to chase the semiconductor equipment and integrated circuit growth, or away from the crude oil and steel decline. A new origin changes the classification analysis, the origin evidence, and the supplier qualification file. None of that is captured by a national growth rate.

The traps, then. Quoting 181.1 trillion as the size of the logistics market, when it is a goods-flow measure and the revenue line is 7.2 trillion. Treating the 49.9 prosperity reading as a return to expansion when it sits below the neutral line, and treating the 0.3 point recovery as a crossing when the first-quarter average was 49.6. Rebuilding a quarterly series from a half-year rate that is not the mean of the quarterly rates, and then presenting the reconstruction as published fact. Citing declining import volumes as though they relaxed a licensing, classification or screening obligation, which they do not. Using an index as an operative contract term without naming the source, the reading or the revision rule. And the quiet one, which does the most damage over time: letting internal figures drift away from external ones without a written caliber note, so that two departments answer the same question with two different numbers and neither can prove which is right.

I will leave it there. The figures were published on 29 July 2026 by the China Federation of Logistics and Purchasing, and I have not added a single number to them that is not either in the release or explicitly labelled as an assumption. What I have tried to do is separate what the release establishes from what people will assume it establishes. Those are different things, and the gap between them is where most of the avoidable trouble in this industry gets made.

↑ Back to the key points

— By Dr. Ingrid Voss