What Household Debt, Bank Profitability and External Data Can and Cannot Tell Us
For investors trying to understand China, the hardest problem may not be forecasting economic growth. It may be determining what can actually be measured.
China publishes an enormous amount of economic and financial data. Its largest banks publish audited financial statements. Regulators report capital ratios, non-performing loans and other financial indicators. International institutions including the IMF and BIS publish extensive datasets on Chinese debt and financial conditions.
But the existence of a number does not automatically make that number independently verifiable.
A few years ago, I tried to analyze the balance sheets of major Chinese banks and the financial consequences of China’s property boom. Eventually, I stopped.
The problem was not a lack of numbers. The problem was that I could not become sufficiently confident that the publicly available information allowed an outside investor to measure the underlying economic risks with the precision the analysis required.
Having lived and worked in China for more than seven years, I had also experienced how the officially visible picture and conditions experienced on the ground could sometimes feel very different. Of course experience is not data I am targeting.
Nor should conversations with people inside China, however useful they may be in generating questions, be treated as evidence about an economy of more than a billion people.
So this time I want to approach the problem differently.
The purpose of this research is not to prove that Chinese financial data are false. Nor is it to assume that official Chinese data are correct simply because they have been published or audited.
The objective is narrower and, I believe, more useful: can outside investors independently verify the health of China’s financial system? And where exact verification is impossible, can we at least determine the direction in which financial risk is moving?
That distinction may be the key to understanding China. Opacity may prevent us from measuring the exact size of a financial problem.
It does not necessarily prevent us from identifying its “direction”.
1. What Does the Term “Verification” Actually Mean Used Here?
Before looking at the numbers, we need to establish what verification means. There are at least three different categories of evidence.
The first is independently observable data.
These include data produced outside China, such as Australian exports to China, Korean exports to China, international commodity flows and revenue reported by multinational corporations with significant exposure to the Chinese economy.
These numbers cannot tell us the true level of non-performing loans inside Chinese banks. But they allow us to observe parts of the Chinese economy without relying exclusively on the Chinese statistical and banking system.
The second category is directionally verifiable data.
A Chinese bank may report a particular mortgage exposure or NPL ratio that an outside investor cannot independently reconstruct. But we may still be able to ask whether the reported trend is consistent with other evidence.
For example, if housing activity remains weak, household borrowing slows, construction-related demand deteriorates and bank profitability continues falling, does a remarkably stable reported NPL ratio fit comfortably with the economic picture? That does not prove the NPL ratio is wrong. It tests whether the reported financial picture is consistent with what can be observed elsewhere.
The third category is information that cannot be independently verified with reasonable confidence.
This includes the true economic value of restructured loans, the ultimate losses embedded in LGFV debt, implicit government guarantees, the economic impairment of loans whose maturities have been extended, and the true system-wide level of distressed assets. These numbers should not be reverse-engineered into false precision. If they cannot be independently verified, the correct conclusion is simply they cannot be independently verified.
That is not a failure of research.
It is itself relevant information for an investor.
2. China’s Household Debt: We Can See the Direction More Clearly Than the Exact Level
Household leverage is one of the most important starting points because China’s property cycle was financed not only through developers and local governments, but also through household balance sheets.
The IMF’s 2025 Article IV Consultation estimates Chinese household debt at 59.4% of GDP in 2025, down from 61.4% in 2024.
The same IMF dataset shows household debt moving from around 60% of GDP earlier in the decade into the low-60% range before declining more recently.
Source: International Monetary Fund, 2025 Article IV Consultation with China
At first glance, this looks encouraging.
Household leverage relative to GDP is no longer rising rapidly. But this is where the verification problem begins.
The IMF series is useful because it standardizes the data and places China within an internationally comparable framework. It should not, however, be interpreted as a completely independent measurement of Chinese household liabilities from outside China.
Much of the underlying information ultimately comes from Chinese financial statistics.
The BIS provides another standardized series measuring credit to Chinese households and non-profit institutions serving households as a percentage of GDP.
Source: Bank for International Settlements, Credit to the Non-Financial Sector
This gives investors another internationally standardized benchmark. But the BIS is not independently auditing millions of Chinese household balance sheets either.
This means we should be precise about what these numbers can tell us.
We cannot independently prove that China’s true household-debt ratio is exactly 59.4%, 60%, or some other number.
What we can observe much more confidently is the direction.
Chinese household leverage rose substantially during the property expansion. More recently, household credit growth and household debt relative to GDP have stopped rising at their previous pace.
The crucial question is what that change means.Is it healthy deleveraging?
Or does it reflect weak mortgage demand, falling confidence in housing and reduced household willingness to borrow?
Those are very different economic outcomes. A falling debt ratio is not automatically bullish.
In a balance-sheet recession, deleveraging can be the consequence of declining risk appetite rather than improving confidence.
For financial markets, the direction of household behavior may therefore matter more than the headline debt percentage itself.
3. Bank Profitability May Tell Us More Than the Headline NPL Ratio
If I had to choose one set of indicators to examine alongside NPLs, it would be bank profitability.
The reason is straightforward. A weak loan can sometimes be refinanced.
But carrying low-return or economically impaired assets for long periods usually has consequences somewhere in the financial system.
Those consequences can appear through weaker net interest margins, lower returns on assets, lower returns on equity, higher provisions, reduced capital generation or weaker future lending capacity.
This makes the direction of Chinese bank profitability extremely important.
The IMF’s 2025 Financial System Stability Assessment reaches a similar conclusion. It describes financial stability risks as elevated and identifies the property downturn and highly leveraged LGFVs as important risks. It also notes declining bank profitability and warns that loss-deferral practices can reduce transparency.
Source: International Monetary Fund, Financial System Stability Assessment — China
The individual banks show a particularly interesting pattern.
ICBC reported an NPL ratio of 1.31% for 2025, down from 1.34% in 2024. But its return on average total assets declined from 0.78% to 0.72%, while return on weighted average equity fell from 9.88% to 9.45%. Its provision coverage ratio also declined slightly from 214.91% to 213.60%.
At the same time, ICBC reported impairment losses on assets of RMB134.9 billion in 2025, up 6.5% from RMB126.7 billion in 2024, even as operating income increased by 1.9%.
Source: Industrial and Commercial Bank of China
China Construction Bank presents a similar picture. CCB reported an NPL ratio of 1.31% for 2025. Its return on assets was 0.79%, return on equity was 10.04%, and net interest margin was 1.34%.
Source: China Construction Bank, 2025 Annual Report
Bank of China provides another example.
BOC reported a 2025 net interest margin of 1.26%, return on average assets of 0.70% and return on equity of 8.94%. Its reported NPL ratio at the end of 2025 was 1.23%, down slightly from the beginning of the year.
Source: Bank of China, 2025 Annual Results
None of this proves that the NPL numbers are wrong. There are legitimate reasons why Chinese bank margins have declined. Interest rates have fallen. Banks have been encouraged to support the real economy. Mortgage rates have been reduced.Competition for deposits and lending has affected spreads. Policy objectives can therefore compress profitability even without a deterioration in asset quality.
But this is precisely why several variables must be analyzed together.
If reported asset quality remains stable while profitability weakens, property stress persists and significant amounts of debt require refinancing or restructuring, investors should ask whether headline NPL ratios fully capture the economic cost of the balance-sheet adjustment.
That is a question, not an accusation.
4. The Reported NPL Ratio Is the Number We Should Be Most Careful With
China’s headline banking-system NPL ratios remain low by conventional international standards.
But the IMF’s Financial System Stability Assessment explicitly cautions against relying too heavily on the headline number.
The IMF notes that NPL ratios have remained broadly stable below 2% for several years while balance sheets remain opaque and public disclosure regarding NPL inflows and loss-deferral practices is weak.
It specifically identifies pre-emptive transactions, regulatory forbearance and disposals of NPLs to asset-management companies as areas that can affect the timing and transparency of loss recognition.
The IMF also notes that sale prices to asset-management companies and their eventual recovery rates are not publicly available, making the effectiveness and pricing of distressed-asset resolution difficult to assess externally.
Source: IMF Financial System Stability Assessment
This is an extremely important distinction.
A reported NPL ratio measures loans recognized as non-performing within the relevant classification framework. It does not necessarily measure every loan that may be economically impaired.
Consider a borrower that cannot repay a loan according to its original schedule. The loan could default and become non-performing.
Government support could change the borrower’s immediate repayment capacity. The original economic problem may therefore continue even without an immediate conventional default.
This leads to one of the central questions of this research: If a loan is refinanced or its maturity is extended, has the underlying credit risk actually disappeared? - Not necessarily.
This means an outside investor should read a reported NPL ratio carefully.
If a bank reports an NPL ratio of 1.3%, we can verify that the bank reports 1.3% of its relevant loans as non-performing under its reporting framework. We cannot necessarily verify that only 1.3% of its loans are economically impaired.
Those are different statements. The second cannot currently be established independently with sufficient confidence.
5. Property Is Where the Verification Problem Becomes Most Important
China’s property market sits at the center of the financial-system question.
Housing connects households, property developers, banks, local governments, construction, commodities and consumer confidence.
If property weakens substantially, it is expected that effects appear across multiple areas of the economy. This gives us an opportunity to conduct external consistency checks.
We do not have to rely exclusively on China’s own property-investment or construction statistics.
Australia provides one important external observation point.
China remained Australia’s largest export destination in 2025, with Australian exports to China totaling A$195.6 billion.
More detailed Australian trade statistics allow researchers to examine commodity flows to China rather than relying only on Chinese import data. This is particularly useful for commodities such as iron ore that are heavily connected to Chinese steel production.
But an important qualification is necessary. Iron ore is not a pure property indicator.
Chinese steel is also consumed by infrastructure, manufacturing, machinery, automobiles and export industries. Strong iron-ore imports can therefore coexist with weak residential construction.
That limitation makes the research more difficult, but it does not make the data useless. The correct question is: does the direction of externally observable construction and industrial demand support or contradict the stabilization suggested by China’s domestic indicators?
Korean trade statistics provide another external observation point.
The Korea Customs Service reported that exports to China increased 46.8% year over year in January 2026, after declining during several months in mid-2025 and returning to growth late in the year.
Korea’s overall semiconductor exports increased 102.5% year over year in the same month.
Source: Korea Customs Service, January 2026 Trade Statistics
But again, this does not prove that China’s property market has recovered.
The semiconductor cycle is being driven by factors including AI infrastructure and electronics demand that can be very different from domestic Chinese housing demand.
That distinction is essential. China can have a weak property economy and a strong export-manufacturing or technology economy at the same time.
This is precisely why headline GDP alone is insufficient for financial-system analysis.
6. China May Be Stronger Externally While Remaining Weak Internally
This may be one of the most important observations emerging from the evidence.
The IMF estimates that China’s real GDP grew 5.0% in 2025.
But the same assessment describes private domestic demand as lackluster and identifies the unresolved property adjustment as a major risk.
At the same time, China’s current-account surplus increased to an estimated 3.3% of GDP in 2025, supported by strong exports.
Source: IMF 2025 Article IV Consultation
These developments are not necessarily contradictory.
China can simultaneously experience strong exports, competitive manufacturing, weak property demand, cautious households, falling bank margins and continued pressure from local-government-related debt.
That combination may actually be one reason China’s financial condition is so difficult to read.
External manufacturing strength can support headline growth even while domestic balance-sheet problems remain unresolved.
For investors, strong exports should therefore not automatically be interpreted as evidence that the property and banking adjustment has ended. Likewise, weak property should not automatically be interpreted as evidence that China’s entire economy is collapsing.
It may be more useful to think of China as containing two increasingly different economic engines.
One is a globally competitive manufacturing and export economy. The other is a highly leveraged domestic property, household and local-government economy working through the consequences of an earlier investment model.
China’s financial system connects them.
7. What We Can Verify — and What We Cannot
At this stage, publicly available information does not appear sufficient for an outside investor to calculate the true level of impaired assets in China’s financial system with confidence.
We cannot independently establish the true system-wide NPL ratio.
We cannot independently determine the ultimate economic losses embedded in LGFV debt.
We cannot fully reconstruct the economic quality of every refinanced, extended or restructured exposure.
We cannot reliably quantify every implicit government guarantee.
It is not easy to find out from public information alone whether every economically distressed loan is being recognized at the same speed it would be under a more market-driven resolution system.
Those are real limitations. But the direction is less opaque than the precise level.
Several important trends can be observed. Household debt is no longer rising relative to GDP at the pace seen during the property boom.
That may represent deleveraging, but it may also reflect weaker credit demand. Bank profitability has weakened materially. Net interest margins are under pressure. ROA and ROE have declined at several major banks.
The IMF identifies declining bank profitability as a financial-stability concern. At the same time, reported NPL ratios have remained remarkably stable.
The IMF continues to identify the property sector and LGFV exposures as material risks and explicitly warns about loss-deferral practices and inadequate transparency.
None of this gives us the true size of hidden credit losses. But it tells us the direction in which we should look.
The inability to verify the level of risk does not mean the direction is unknowable.
8. The LGFV Problem Makes a Western “Lehman Moment” an Imperfect Analogy
This brings us to the most dramatic question - could China have a Lehman moment?
It is possible but it may be the wrong framework through which to understand Chinese financial risk.
China’s financial architecture is fundamentally different from the market-based financial system in which Lehman Brothers failed in 2008.
China’s system connects state-owned banks, local governments, LGFVs, state-owned enterprises and the central government. Capital controls also reduce some of the channels through which a conventional external funding run can spread. The state therefore has considerably more influence over who recognizes a financial loss, when it is recognized and where that loss ultimately resides.
The IMF notes that weak Chinese financial institutions are often intervened in before an official determination of non-viability, with the pervasive presence of the state used to distribute and absorb losses.
This changes the potential crisis mechanism.
A Chinese adjustment may instead have more opportunities to follow a path such as:
Loss → refinancing → maturity extension → restructuring → debt swap → state-bank absorption → recapitalization
The second process may reduce the probability of a sudden Lehman-style failure. But it does not make the original economic loss disappear.
It changes where the loss sits, who carries it and how quickly it is recognized.
9. Can China Prevent a Banking Crisis While Still Suffering the Economic Consequences of One?
This may ultimately be more important than asking whether China will experience a Lehman moment.
There are possibilities that the Chinese governments prevent its major banks from failing, roll over LGFV debts, or extended, or more cash injection to those state-owned banks if necessary.
There may never be a single event comparable to Lehman Brothers but economic costs can still accumulate.
Banks holding low-return or impaired assets may generate weaker returns. Lower profitability reduces the ability of banks to generate capital internally. Banks may become less willing to extend credit to riskier private businesses. Capital can remain tied to old projects rather than moving toward more productive uses.
Households may save more and consume less if property values remain weak.
Local governments facing weaker finances may reduce expenditure.
The economy could therefore suffer some of the economic consequences normally associated with a banking crisis without experiencing a visible wave of major bank failures.
This is not merely theoretical. The IMF warns that continued reliance on temporary or partial solutions can worsen capital misallocation, mispricing of risk and the gradual accumulation of what it describes as veiled losses with macroeconomic consequences.
That observation gets very close to the central issue. The absence of a banking crisis is not evidence that the losses have disappeared.
10. What China Looks Like From Outside China
The next step is therefore not to search for a single “true” Chinese banking number. It is to compare two different pictures.
One is the China described by official Chinese statistics and bank financial statements. The other is the China that can be observed from outside - Australian commodity exports, Korean exports, energy imports, shipping flows, multinational companies’ revenues on China, international tourism and travel activities.
None of these indicators is sufficient on its own but together they offer directional verification.
If Chinese property and household activity are genuinely stabilizing, some externally observable indicators should eventually move in a consistent direction.
If bank asset quality is genuinely improving, profitability, credit transmission and related financial indicators should eventually show stabilization as well.
If the official and externally observable pictures repeatedly diverge, that divergence itself becomes important information.
Conclusion: Opacity Is Also a Financial Variable
The starting question of this research was simple: can we really verify the China’s financial system?
The honest answer is Not Sure.
Some of the most important financial numbers cannot be independently reconstructed by an outside investor. The true level of impaired bank assets cannot currently be established with sufficient confidence.
The ultimate economic losses associated with LGFVs cannot be measured precisely from publicly available information.
Refinancing, restructuring and maturity extensions make conventional default and NPL statistics harder to interpret.
And international datasets improve comparability without completely solving the underlying verification problem.
But that does not mean we know nothing.
We can observe the direction of household leverage and declining bank profitability. We can compare reported NPL stability with movements in margins, returns, provisions and impairment charges.
We can also catch parts of China’s economy through data generated in Australia, Korea and other trading partners.
We can examine whether those external observations are consistent with the economic picture presented by official Chinese statistics and the country’s major banks.
That may be the most realistic standard investors can apply.
The key question is therefore not whether every Chinese financial number is true or false.
It should be;
What can be verified, what can only be inferred, and what remains fundamentally unknowable? If systemic financial risk cannot be independently measured, should opacity itself carry a risk premium?
China does not necessarily need a Lehman moment to suffer the economic consequences normally associated with a banking crisis. A financial system capable of preventing sudden institutional failure may also be capable of stretching the recognition and absorption of losses across many years.
That could make China’s financial system more stable than the most pessimistic forecasts suggest. But it could also make the adjustment slower, less visible and substantially harder to measure.
The absence of a crisis does not prove the absence of losses. And the inability to know their exact size does not mean we cannot see the direction in which the system is moving.
That may ultimately be the most important signal of all.