Reading Pan Gongsheng’s “Understanding the Profound Changes in China’s Financial Structure”
On September 16, 2026, People’s Bank of China Governor Pan Gongsheng published an article in Qiushi titled “Deepening Understanding of Changes in China’s Financial Structure and Improving the Adaptability of Finance to the Real Economy.” If one only checks the numbers, the article is easy to read straight through. Most of the data are real, and many individual observations are reasonable: lending to property and local-government financing platforms is declining, government bond financing is rising, bank loans account for a smaller share of aggregate financing to the real economy, and more financial resources are flowing to technology, green development, and small businesses. These changes are real.[1]
But true data establish only that something happened. They do not automatically tell us why it happened, and they cannot by themselves prove the policy conclusions that follow. The first sentence may be true and the next sentence may also be plausible; the question is whether the first actually supports the second. The most interesting parts of Pan’s article are therefore the places where the word “therefore” carries more weight than the evidence beneath it.
1. Direct financing has increased — but whose financing has increased?
Pan writes that by the end of June 2026, direct financing had reached roughly one-third of the outstanding stock of aggregate financing to the real economy.[1] The number is not the issue. The question is what “more direct financing” means once government bonds, corporate bonds, and equity financing are placed in the same category.
According to the People’s Bank of China’s financial statistics for the first half of 2026, corporate bonds accounted for 7.8% of the outstanding stock at the end of June, up only 0.1 percentage point from a year earlier. Domestic equity financing of non-financial enterprises accounted for 2.7%, down 0.1 percentage point. Government bonds accounted for 21.9%, up 1.3 percentage points.[2] On this stock measure, nearly all of the year-on-year increase in the direct-financing share came from government bonds, while the combined share of corporate bonds and equity was essentially unchanged.
That distinction matters. When local debt that was previously carried through bank-linked financing platforms is replaced by special-purpose government bonds, the statistics will naturally show fewer loans and more bonds. But that is not the same economic event as a technology company replacing bank borrowing with corporate bonds, equity issuance, or venture capital. One is a change in the government’s financing channel; the other is a change in the financing structure of firms.
There is nothing wrong with putting both under the statistical heading of “direct financing.” But if the same shift is then taken as evidence that corporate financing is moving decisively from banks to capital markets, the two need to be separated. Pan himself acknowledges that heavier government bond issuance, swaps of hidden local debt into special-purpose bonds, and larger bank write-offs all raise the bond share and lower the loan share. He then states that “even after excluding these factors, the change in China’s financing structure remains significant.”[1]
That sentence should be the beginning of the demonstration, not the end of it. If the argument depends on what remains after those factors are excluded, the post-exclusion structure should be shown. Otherwise, the true statement “direct financing has increased” is being asked to prove more than it can. Who is doing the financing determines what the number means.
2. Fewer loans do not tell us by themselves why loans are slowing
The article gives several reasons for slower loan growth: the stock of finance is already very large; lending linked to property and local financing platforms is falling; the economy is shifting from finance-intensive sectors toward technology-intensive sectors; and bonds and equity are taking over part of the role once played by bank credit. These changes are then placed within a broader story of a transition in China’s stage of development, leading to the conclusion that the old pace of loan growth is both difficult and unnecessary to maintain.[1]
That explanation may be right, but it still needs to be identified. Slower lending can also reflect lower expected returns on investment, weaker collateral values after a property adjustment, heavy existing debt, lower bank risk appetite, greater caution among firms, or weak aggregate demand. These forces can coexist with industrial upgrading.
The first fact we truly know is simply that loan growth has slowed. Whether it slowed because an upgraded economy needs less bank credit, because firms lack attractive investment opportunities, or because some firms want credit but cannot obtain it are different questions.
If industrial upgrading and substitution toward other financing channels are the main mechanism, we should at least see alternative forms of capital gradually taking shape rather than only bank lending retreating. Conversely, if equity, venture capital, and corporate bonds fail to provide sufficient replacement funding, if non-property lending remains weak after old-sector contraction is absorbed, and if investment and profits do not improve, then the claim that slower lending mainly reflects healthy structural upgrading needs to be reconsidered.
3. The new economy may be light on bank loans, but not on capital
Pan points to an interesting contrast. In traditional sectors, the ratio of medium- and long-term loan balances to sectoral value added is generally high: 1.7 for real estate and 3.3 for transportation. In new-growth sectors it is much lower, with information services at roughly 0.1. The article uses this to describe the financing needs of the new economy as becoming “lighter.”[1]
The article clearly understands that technology firms need a different kind of money. It notes that such firms have high R&D spending, high uncertainty, and long investment cycles; that early-stage firms rely heavily on private equity and venture capital; and that later-stage firms can use loans, bonds, and equity.[1] The issue, then, is not whether the article recognizes diversified financing. The issue is whether lower bank-loan intensity is enough to justify the broader label of “lighter financing needs.”
Semiconductors, artificial intelligence, and biomedicine are not cheap industries. Some require enormous, long-duration, high-risk upfront investment. They may be light on bank loans without being light on capital. The change is not that they “need less money,” but that they need a different kind of money — more equity and long-term risk capital that can absorb failure.
A successful structural transition therefore cannot be judged only by how much bank lending falls. We also need to see how much equity, venture capital, and corporate bond financing grows on the other side. If bank lending falls by 100 and other capital rises by 120, the transition may indeed be successful. If bank lending falls by 100 and other capital rises by only 20, the result may be a shortfall in capital formation rather than simple “structural optimization.”
4. A better composition does not answer whether the total is sufficient
The article notes that the share of new lending going to property and infrastructure has fallen from more than 60% to around 10%, while the share going to the “five priority areas of finance” has risen to more than 70%. It then adds that this structural change “may not show up in loan growth, but can improve the efficiency of fund use.”[1]
The first half of that sentence already recognizes that composition and aggregate volume are different questions. The difficulty lies in the second half: why should a change in destination, by itself, prove that capital is being used more efficiently?
Consider a simple example. Suppose new lending was once 100, of which 20 went to technology. Now new lending is only 20, of which 14 goes to technology. The technology share has risen from 20% to 70%, but the absolute amount received by technology has fallen from 20 to 14. A better ratio does not tell us whether the total amount is sufficient.
Nor can the fact that money goes to technology, green projects, or small firms by itself establish higher efficiency. Efficiency must ultimately be observed in outcomes: profits, cash flow, productivity, technological progress, or perhaps new low-return investment. Where the money goes is an allocation question; what the money produces is an efficiency question. The same ratio cannot prove both.
5. “Effective financing demand is fully met” is difficult to verify without a clear test
The article states that China’s financing market has shifted from a supply constraint to a demand constraint, that “effective financing demand can be fully met,” and that financing conditions remain relatively accommodative.[1] The immediate question is: what counts as “effective” financing demand?
Suppose a firm wants to borrow and a bank rejects the application. Is that evidence of unmet financing demand, or evidence that the demand was not “effective” in the first place? Without a criterion for effectiveness that is independent of the lending outcome, the definition risks becoming circular: approved demand is effective, while rejected demand can be reclassified as ineffective.
This does not mean that effective financing demand is impossible to measure. A central bank can examine loan-demand surveys, credit utilization, financing costs, and other indicators. The problem is that this article does not tell the reader what standard supports the claim that demand is “fully met,” nor what evidence would require that claim to be revised.
A proposition without boundaries can easily accommodate every outcome. And a theory that can explain every outcome often ends up explaining very little.
6. Revitalizing existing loans is not automatically the same as making new loans
One sentence in the article deserves particular scrutiny: “For supporting economic growth, revitalizing inefficient existing loans and making new loans are essentially the same.”[1]
Improving the use of existing financial resources is obviously important. If the sentence means only that both actions can ultimately support the real economy, the point is understandable. But if it is taken to mean that they have the same macroeconomic effects on credit creation, bank capital, and new purchasing power, then the specific transaction matters.
“Revitalizing existing loans” may mean loan transfers, securitization, debt restructuring, disposal of non-performing assets, or repayment of an old loan followed by a new loan to another firm. These operations do not have identical effects on bank balance sheets, corporate liabilities, or newly created purchasing power.
To see in what sense the two are “the same,” there is no need for a larger abstraction. Follow the balance sheet: whose asset was the loan before? Whose asset is it now? Who reduced a liability? Who received new financing? Did the bank actually regain lending capacity? Once the accounts are reconstructed, the broad phrase “revitalizing the stock” breaks into several quite different economic actions.
7. The deeper risk is a policy self-validation loop
There is a deeper issue in the article. In recent years China has adjusted the property sector, addressed local-government debt, dealt with banking risks, and used a wide range of structural financial policies. Pan does not deny that policy itself has changed the financial structure; the article explicitly mentions special-purpose bond swaps, loan write-offs, risk resolution, and financial reform.[1]
Precisely because policy helped produce the observed outcome, using that outcome to justify continuing in the same direction requires a separation between changes generated by the economy itself and changes generated by policy. The point is not that the policies are necessarily wrong. The point is that policymakers are changing the economy they are observing.
Observed outcomes reflect more than an economic law. They also reflect policy interventions, external shocks, and the responses of households, firms, and financial institutions. Otherwise, policy changes the data; the changed data are described as structural improvement; and the structural improvement then returns as evidence for continuing the same type of policy. That is how a self-validating loop can form.
This is not a uniquely Chinese problem, nor a uniquely central-bank problem. Any institution with policy power can encounter it. The stronger the policy instrument, the more important the counterfactual becomes: what would have happened without the policy? What would have happened under a different policy? Without those questions, policy outcomes can too easily become evidence for the policy itself.
8. What matters most is to leave a condition for failure
This is why Pan’s article is useful for thinking about verifiable reasoning. Verification does not end with checking whether the data are correct. That is only the beginning. After reading a conclusion, we should still ask: why this explanation? What other mechanisms could produce the same result? And finally, what future evidence would force us to revise the explanation?
Take the proposition that slower but higher-quality loan growth may become a new normal.[1] That can be a legitimate judgment. But it should have a failure condition. If loan growth remains weak for years while private investment keeps falling, corporate profits deteriorate, risk capital fails to fill the gap, innovative firms remain short of long-term financing, and nominal growth stays weak, should we still call the outcome “slower but higher-quality” credit growth?
If the answer is always yes, the concept has no boundary. If the answer is no, the boundary should be stated in advance. A good theory does not guarantee that it will always be right. It leaves the future a way to prove it wrong.
Conclusion: the problem is often not in the facts, but between the facts
Pan Gongsheng’s article contains a large amount of valuable and largely verifiable financial data. China’s financial structure is indeed changing, and loan growth alone is increasingly inadequate as a measure of financial support to the economy. Those points can be retained.
But moving from “the financial structure has changed” to “the change is mainly the result of economic upgrading,” then from “economic upgrading” to “slower lending is a healthy new normal,” and finally to “the current policy direction has therefore been validated” requires several additional steps.
That is why, when reading an article like this, I pay particular attention to one ordinary word: therefore. Remove it once. The first sentence may be true. The second may also be true. Then ask: did the first sentence actually prove the second?
That is often where serious verification begins.
References
[1] Pan Gongsheng. “Deepening Understanding of Changes in China’s Financial Structure and Improving the Adaptability of Finance to the Real Economy” [《深刻认识中国金融结构变迁 提升金融服务实体经济适配性》]. Qiushi, No. 18, September 16, 2026. https://www.qstheory.cn/20260915/1c41fc49b4b44db4b2550ddd41f40315/c.html
[2] People’s Bank of China. “Financial Statistics Report for the First Half of 2026” [《2026年上半年金融统计数据报告》]. July 15, 2026. https://www.pbc.gov.cn/goutongjiaoliu/113456/113469/2026071512340454869/index.html
