Bank mechanism for hiding bad debt · easy-to-read version

"Extend and Pretend" - When Banks Pretend Bad Debt Doesn't Exist

A pho restaurant borrowed 1 billion to expand. Revenue dropped, 100 million interest was due and there was no money. The bank has two ways to handle it - a "proper" way to look bad on quarterly reports, a "subtle" way to keep good books. The second way is called extend and pretend - renew and pretend. The Japanese call it zombie lending. IMF called loan evergreening. This article tells that story - in everyday language, with charts - and shows how a small trick on the books can "lose a decade" for an entire economy.

1 · The Story of Mr. Tuan's Pho Restaurant

Mr. Tuan opened another pho restaurant branch. Bank loan 1 billion, interest 10%/year. After 1 year, you have to pay 100 million in interest. One day, revenue dropped because the neighborhood was digging up roads. Mr. Tuan told the credit officer: "I didn't collect enough. Please give me a few months delay."

Credit officers have two ways to handle this:

That's the whole idea. The name changes with the times but the mechanism is the same: America in the 80s called it "regulatory forbearance", Japan in the 90s called "zombie lending", Europe 2010s called "zombie credit", America 2024 calls it straight "extend and pretend", the IMF/Basel technical name is "loan evergreening" or "forbearance measures" (BIS BCBS d403).

Why is this trick dangerous - imagine it with a snowball

Each time it "extends", the debt balance increases because unpaid interest is added to the principal. 5 years later, the original 1 billion loan became 1.6 billion. Mr. Tuan had no way to pay. The bank did not dare to record it - because the loss it had to record was now twice as big:

Why do banks still choose method B?

The logic is very human: every credit officer wants this quarter to have good profits. The bank CEO also wants the stock to not fall. When bank capital is weak - that is, there is no "cushion" to record more losses - method B becomes an almost mandatory choice. Two economists Peek & Rosengren point out a paradoxical but true rule (NBER WP 9643):

Accounting logic explains this paradox: the weaker the bank, the more it must avoid recording losses - because recording losses will make capital even weaker, possibly violating the minimum capital standards of the supervisory agency. The only way out is to not let the loan officially fail - that is, continue pumping credit. Healthy businesses do not need this credit; dying businesses are the "only" customers that weak banks have to keep.

2 · Japan - Patient Zero of the World

In 1990, the Tokyo land bubble burst. Land in Ginza has decreased by 70% since its peak. Nikkei from 39,000 to 14,000 in just 2 years. A series of construction, real estate, and retail companies - borrowing capital with land as collateral - no longer have cash flow to repay debt.

Normally, banks must record losses, handle assets, and let capital and labor move to healthy businesses. Japan doesn't do that. For 10 years, Japan's big banks chose method B - huge scale. The classic study by Caballero, Hoshi & Kashyap (CHK 2008) measures this phenomenon:

The price - not just a bank loss

Most people think the biggest problem with zombie lending is hidden accounting losses that eventually have to be recognized. Yes, but that's not the biggest problem. The biggest problem is zombies oppress healthy businesses:

• Zombies hold land which should have been sold to release the mortgage - causing land prices to not go down, healthy businesses not being able to buy premises.
• Zombies hold labor but should have quit - making healthy businesses unable to hire good people.
• Zombies hold market share and are willing to sell below price to have cash flow - making healthy businesses unable to raise prices and unable to innovate.

The result: the whole economy stood still. Here's the "lost decade" - actually two decades:

3 · Europe 2010s - Same Mechanism, Same Consequences

After the Eurozone public debt crisis of 2010-2012, Southern European banks fell into the same situation as Japan 20 years ago. In July 2012, Mario Draghi (ECB) announced "whatever it takes" save the euro. Southern European interest rates decreased, banks were "reborn" indirectly. But research by Acharya and colleagues (Review of Financial Studies 2019) shows that: the majority of new capital does not flow to healthy businesses, which continues to raise zombies.

Measured by corporate capital ratio stuck in zombies 2013 (OECD):

BIS measures the same phenomenon in 14 developed countries over the long term:

4 · China 2021-2026 - Repeat Now, Larger Scale

The two cases, Japan and Europe, are in the past. China is the present - and the largest iteration in history of the same pattern.

After China's real estate bubble burst in 2021 - detailed macro developments in one separate post - China's financial system faces the same question that Japan had to answer 30 years ago: record losses immediately, or extend and pretend? Atlantic Council (July 2025) wrote directly:

Two parallel fronts. Both use the same formula.

Front 1: LGFV - loan extension for 25 years

LGFV (Local Government Financing Vehicle) is a company owned by the Chinese local government, used to borrow money for infrastructure projects - avoiding the official debt ceiling. When real estate collapsed, local land sales collapsed (which accounted for 30-40% of local revenue), LGFV had no cash flow to pay debt.

Unbelievably large scale:

• IMF FSAP 4/2025: LGFV outstanding debt at the end of 2023 ~RMB 60.4 trillion (~USD 8.4T, ~47% of China's GDP), estimated end of 2024 ~ USD 9T (IMF CR 25/100).
• Goldman Sachs bottom-up from >4,000 LGFV: ~USD 12.1 trillion late 2024.
• For comparison: China's total GDP ~17 trillion USD. That means LGFV debt is equal to ~50-70% of GDP.

The policy response from July 2023: ICBC + CCB starts issuing LGFV term loans 25 years (compared to 10 years which was the old standard), some loans have no interest or principal payments for the first 4 years - but interest still accrues. This is an almost "permanent" extension. (SCMP).

In November 2024, Beijing announced the package RMB 10 trillion "swap" - convert LGFV underground debt into official local government debt at lower interest rates. Important features: No haircut, no central bailout, no structural reform. Carnegie calls this "a RMB 10 trillion accounting exercise" (Carnegie). Fitch: swap package only covers approx 25% of true hidden debt.

"Successful" progress on paper: from March 2023 → September 2025, the number of LGFVs decreased by 71%, financial debt balance decreased by 62%. But Caixin investigation discovered: >70% of "disappeared" LGFVs were actually just renamed or merged, rather than repaid debt (Caixin).

Front 2: Property developers - 4 ways out, same result

China's biggest real estate developers all follow the same path - extend, extend again, then restructure with a big haircut:

In early 2026, Chinese regulators allow banks extend loans for "whitelist" projects by up to 5 more years compared to the old deadline end-2026. RMB 4 trillion credits are in this category (Yicai). That means rules extend again - push the maturity wall to 2031.

The hardest evidence: the rate of zombie firms has surpassed Japan

In December 2025, the Dallas Fed announced shocking data on the ratio of zombie firms' assets to Chinese listed companies:

The Dallas Fed mechanism pointed out exactly the same as Peek-Rosengren in Japan 30 years ago:

IMF warned, but was denied data

IMF Article IV China 2024 and FSAP 4/2025 write directly (very unusual for a report on China):

Notably: IMF criticized the Chinese agency refused to provide bank-level exposure data to LGFV - that is, the IMF itself cannot see the real level of stress. More: the rate of "loss loans" (principal + irrecoverable interest - NPL's worst bucket) has increased ~40% of total NPL.

On the official books, Big 4's NPL is still around 1.3% (ICBC 1.31%, CCB 1.31%, ABC 1.27% at the end of 2025). But CCB's real estate NPL is 5.64% - 4 times the total number (Seafarer). The gap between "gross" NPL and "stressed core industry" NPL is a classic sign of focused evergreening.

In short, China is the biggest test case in evergreening history:

• Absolute scale: ~4 times more than Japan when the bubble burst.
• The zombie share in real estate has surpassed Japan's peak - first time seeing it.
• State-owned structure making the exit rate even slower (no market pressure, no independent inspectors with full access).
• Lender of last resort is unclear with LGFV - PBOC can inject liquidity but there is no official insolvency mechanism.

The results will be clear in 5-10 years - and will shape the global economy, not just China. Especially for Vietnam: the prolonged China shock means low export demand, careful withdrawal of FDI capital, and deflationary pressure spreading throughout the region.

Why can China's deflation spread to Vietnam and East Asia?

When China's domestic demand is weak (due to balance sheet recession), Chinese factories continue to produce - because the capacity has already been invested, stopping production will result in even more losses. Surplus goods must be exported. This is the "exporting deflation" mechanism - exporting deflation. There are 5 main transmission channels:

• Commodity price channel: China accounts for 30-50% of global demand for steel, cement, copper, and coal. Weak Chinese demand → world commodity prices fall → raw material exporting countries (Indonesia, Malaysia, Australia) earn less. Vietnam imported cheap Chinese steel → domestic steel enterprises (Hoa Phat, Hoa Sen) had to cut prices accordingly, profit margins thinned.
• Manufacturing competition channel: China has surplus capacity in solar panels, EVs, textiles, and household electrical appliances - selling them to the world at low prices. Enterprises in the same industry in Vietnam, Thailand, and Indonesia face direct competition - either cutting prices (reducing profits), or losing orders (closing).
• Exchange rate channel: Yuan is weak compared to USD → other Asian currencies (VND, THB, IDR, MYR) must decrease accordingly to maintain export competitiveness. When the VND depreciates, imported goods become more expensive, but at the same time cheap Chinese goods still flood in - the result is "imported deflation" for competitive products, inflation is only for non-competitive products.
• FDI channel: Chinese capital previously flowed strongly into ASEAN (industrial parks, real estate, M&A). When Chinese enterprises have to preserve capital back home, FDI flows from China to Vietnam/Thailand/Indo slow down. This is good news in terms of reducing dependence, but has short-term implications for growth.
• Tourism and consumption channel: Chinese tourists accounted for 30-40% of intra-East Asian tourism before COVID. When Chinese consumption is weak, the Thai, Singaporean, and Vietnamese tourism industries lose revenue - especially high-spending customers who come to buy branded goods and real estate.

This is not theoretical - it has happened. China's PPI (factory price) is negative 40+ consecutive months from 2023-2026. During the same period, Vietnam's PPI was also negative or near 0% in many quarters, despite high nominal GDP growth. Japan during its "lost decade" period also exported deflation to East Asia - leading to the 1997-1998 Asian financial crisis (through the yen carry trade exchange rate channel). The mechanism now runs backwards: China is the epicenter, smaller surrounding countries absorb the shock.

The worst currency backdrop: China exports deflation, America exports inflation - a "double squeeze"

If only one side exports price pressure - only deflation from China, or only inflation from the US - the open economy still has a way to adapt. The 2024-2026 issue is two shocks coming in parallel from the two poles, neither of which cancels out the other.

US side - exports inflation through three channels:

• Budget deficit + high public debt: The US runs a deficit of ~7% of GDP, public debt exceeds 130% of GDP. To finance, the Treasury issues large amounts of bonds → the supply of USD bonds is high → to absorb, either yields must rise, or the Fed must print. Both export effects to the world.
• Trump Tariffs 2.0 (2025-2026): A 10-25% tariff on imported goods pushes up global consumer goods prices - because the supply chain has to be restructured, everyone has to pay more. Vietnamese goods exported to the US are more expensive or are replaced by production in lower tax countries.
• Strong USD + USD pricing commodity: The Fed keeps rates higher than expected because of persistent inflation → USD is strong → Vietnam's imported goods are now more expensive in USD than in VND. Oil is still ~$80-90/barrel despite weak Chinese demand, due to strong USD and US military spending.

A "double whammy" for Vietnam and the open economies of East Asia:

• Selling businesses cannot increase prices (competing with Chinese products) But input costs increase (USD + US commodity) → corporate profit margins are thinning - especially heavy for export industrial manufacturing.
• Exchange rate stuck in both directions: Keeping VND stable requires high interest rates (killing domestic growth + increasing NPL pressure), while devaluing VND makes imported goods more expensive (import inflation, crushing people). There is no "beautiful" exit.
• People: CPI increased due to imported gasoline and food while wages were stiff (businesses with thin margins could not increase wages). Real purchasing power decreases.
• Bank: It is more difficult for businesses to repay debt → NPL increases → Evergreening pressure increases. This is a closed feedback loop - and is why this article about evergreening is directly related to today's exchange rate and inflation story.

A similar "double squeeze" situation happened with Latin America in the 1970s and 80s ("double squeeze stagflation" - OPEC oil price increased + strong USD due to Fed Volcker), and with ASEAN in late 1996 before the 1997 crisis (CNY devalued ~34% in 1994 + strong USD + export orders gradually disappeared to China). Common denominator: Domestic enterprises are forced to squeeze margins + domestic budget is forced to tax + domestic banks are forced to increase NPL - all three at the same time.

Signals to monitor: the difference between imported CPI and domestic CPI (if expanded = imported inflation main), profit margin of industrial manufacturing enterprises (If it falls continuously = Chinese competition forces it), Stage 2 loan rate in bank financial statements (official NPL early warning), VND vs USD interest rate difference (If you have to loosen it = accept exchange rate instability to keep growth).

5 · Why Do Regulators Sometimes Allow It?

The right question is not "why do banks evergreen loans?" (answered: for quarterly profit). The important question is: "Why does the supervisory agency - the institution assigned to fight this practice - sometimes officially allow it?" There are three reasons, from good to bad.

Good reason - temporary shock indeed

Imagine COVID in March 2020. Millions of businesses temporarily had no revenue because of lockdown - but still an intrinsically healthy business. If you force the bank to record bad debt immediately, the bank will tighten credit, and healthy businesses that only need to hold out for 6 months will actually go bankrupt. Forbearance 6-12 months is helping healthy businesses through temporary stress.

This is the official logic endorsed by the IMF, BIS, and FSB in March 2020 - June 2021. Not wrong in principle.

Neutral reason - avoid bank run

If Bank X reports NPL jumping from 2% to 8% in one quarter, depositors may panic and withdraw funds, even if Bank X is actually liquid. The SVB case in March 2023 showed that modern bank runs can unfold in hours through phones. Forbearance helps "smooth" information over time.

This logic is correct - but it is also easily abused to justify permanent concealment.

Bad reason - putting stress on the successor

Recording large-scale NPLs has political costs: state-owned bank profits decrease → the budget reduces bank tax revenue → the bank CEO is replaced → may have to inject capital from the budget. Every link is "painful" for someone in office.

Forbearance allowed kick the can down the road. The problem becomes that of the successor. This is the mechanism that killed the Japanese economy for 20 years.

Central finding

Regardless of the reason, a 2020 IMF study (Bergant & Kockerols) concludes bluntly:

The last sentence is important: banks spend capital to raise zombies and have no more capital to lend to startups, healthy SMEs, and first-time home buyers. This is the mechanism Japan has been stuck in for 20 years.

6 · COVID - The Largest Forbearance in History

COVID created a global experiment. Most countries applied widespread forbearance in 2020-2021. The difference was how they exited. Most of the EU and US ended forbearance 2021-2023, reported NPLs increased in 1-2 years later - and the system digests it. Some areas kept it longer, the consequences continue today:

• EU: By June 2020, moratoria had been granted for EUR 871 billion (~6% of total credit). Ends September 2021. Stage 2 loans (early warning) increase to 9.5% mid-2022, higher than COVID peak (EBA).
• US - commercial real estate (CRE): When the Fed raises interest rates in 2022-2023, regional banks often extended CRE loan maturities instead of recognizing NPLs. The New York Fed report directly called it that "extend and pretend": 27% of US bank capital is stuck in CRE extend-and-pretend - This is the core reason the market is still worried about US regional banks (Fed NY SR 1130).

7 · East Asia Post-COVID - Four Different Ways Out

In the region, there are four forbearance exit paths worth examining – and Vietnam is one of them.

Vietnam - a continuous 5-year lineage

The State Bank of Vietnam has no break between forbearance periods from March 2020 to present - a flow through 6 circulars:

The gap between two NPL numbers

This is where the accounting language starts to get vague. The SBV reports two different metrics, and the gap between them indicates how much stress is "in the waiting room":

There is one notable observation: The announced NPL ratio decreased while the absolute NPL outstanding balance continued to increase - because credit growth is too fast (17.87% in 2025, highest in a decade). Moody's calls this condition "unseasoned risks" - the risk has not yet matured. New loans have not had enough time to reveal their true quality; The denominator (total credit) swells faster than the numerator (recorded NPL) - the ratio decreases but risk increases (Moody's).

Outside voices

The most honest way to understand the status quo is to read the parties with no direct political interest:

World Bank also tracks the real estate segment - accounting for 24% of Vietnam's total credit end of 2025 and is the industry most sensitive to evergreening: real estate NPL increases from 1.7% (2022) → 2.73% (2023) → 3.7% (7/2024) (WB Taking Stock 8/2024).

SCB - case study of extreme evergreening

The SCB case is not a typical example of the system - it is an outlier with fraudulent elements - but it illustrates clearly How far can the evergreening mechanism go when supervision is bypassed?. 91.5% of SCB shares are controlled by one owner through a proxy. The ecosystem creates 2,500+ fake loans in 10 years. The loan is not paid due → new loan is issued to a ghost company in the same ecosystem → money is used to pay off the old loan → "performing" books for nearly a decade. When it broke open: NPL SCB is 97% total credit, the State Bank must inject liquidity ~VND 657,000 billion (~USD 26 billion), proposed to be paid in installments over 15 years (CNBC).

8 · Two Ways Out of Crisis - Sweden Vs Japan

The two countries experienced the same banking crisis in the early 1990s, of similar scale. The results are different:

The difference is not the size of the bubble (both are huge), but rather recording and processing speed:

• Sweden: Independent audit of the entire system in 6-12 months → announcing real losses → establishing "bad banks" (Securum, Retriva) to dispose of collateral quickly → temporarily nationalizing 2 large banks → reselling shares 5 years later. Final cost: ~2% GDP.
• Japan: allowing banks to value their own assets → implicit forbearance → bad assets kept at the original bank for 10+ years → until 1998-1999, there was the first major bailout package. Final cost: ~10-20% GDP, no significant recovery.

9 · Six Signs to Watch for

For every banking system - not just any country - here are six indicators that people and investors can check for themselves from bank financial statements (quarterly/annual financial statements) or central bank reports:

10 · Conclusion - Three Things to Bring

First: evergreening is not a separate behavior, but one incentive structure. When bank capital is weak + supervisory pressure is low + market interest rates are low, the system will naturally slide into evergreening - regardless of country, regardless of era. Japan 1990s, US S&L 1980s, Europe 2010s, US CRE 2022-2024, China 2021-present - same pattern. This is not a "personal fault". an equilibrium that any banking system can fall into.

Second: the biggest price is not the accounting loss, but productivity growth. This is the most important discovery in the literature of the past 20 years. Zombies keep resources away from healthy businesses. An economy that is 15-30% zombie can never grow to its potential - even if everything else macro is right. This is an impact channel that is rarely seen, but is the most severe in the long term.

Third: escape velocity is much more important than the magnitude of the shock. Sweden 1992 (quick escape, ~2% of GDP, recovery) vs Japan 1990 (hidden for 10 years, ~20% of GDP, "lost decades"): the difference is not the size of the bubble (both are large) but the recording speed. Each month forbearance lasts = a month of accumulated hidden debt + a month of zombies oppressing healthy businesses + a month of widespread misallocation.

For Vietnam – and more broadly post-COVID East Asia – the question is not "is there evergreening?" (every system has to a certain extent) but "is the exit mechanism ready?":

• SBV did not extend Circular 02 in 2025 - a good signal of change according to IMF recommendations.
• But TT53 (Yagi) extends the mechanism of "holding debt groups" through another vehicle until the end of 2025, with the last payment period until 2027.
• The Law on Credit Institutions 2024 has "legalized" part of Resolution 42 on NPL handling - important but The right to seize collateral has not been fully restored which the IMF considers a necessary condition.
• New draft decree: banks with NPL ≥3% are required to sell to VAMC - a formalization step, but VAMC is a "parking lot" if there is no real liquidity to handle.

Vietnam has two favorable factors that Japan in 1990 and Europe in 2010 did not have: very high nominal GDP growth (≥8%) helps dilute the NPL numerator quickly; and credit growth is very fast helps dilute the denominator. These two factors can mask stress for several years. But both have limits: credit-based GDP growth is unsustainable; Fast credit growth accumulates immature risks.

The final lesson - and why this is worth reading for investors, depositors, and policy watchers alike - is extend and pretend works until it doesn't. Each reasonable and time-limited forbearance is a good tool; Each unlimited wave is one step towards the trap. Japan needed 20 years to understand this. Europe needs 10 years. China is in the middle of the process – and may redefine the maximum scale evergreening can push. East Asia is in the early stages of a similar test - the results will be clear in the next 3-5 years, not in 3-5 months.

A global tail risk - when all central banks are stuck after the COVID pump

Starting context - COVID synchronized injection. In March 2020, most major central banks reacted in the same way: cutting interest rates to zero and purchasing assets (QE) on an unprecedented scale. Fed balance sheet from $4.2T (early 2020) to $9T (peak 2022). ECB from €4.7T to €8.8T. BoJ continues to maintain more than 130% of GDP. PBOC injects indirectly through the state-owned banking system. In 18 months, the total assets of major central banks increased by ~$10 trillion - equivalent to ~10% of global GDP.

Why they cannot exit Theory: after COVID, when the economy returns to normal, the central bank must normalize - attract liquidity, sell assets, and bring rates to neutral levels. The reality is different. Evergreening in the financial system (US CRE, Chinese LGFV, Southern European public debt, East Asian real estate) makes the central bank not dare to absorb liquidity quickly - because doing so would trigger a wave of mass defaults of "zombies" who are being kept alive by low interest rates. We also don't dare lower rates back to zero - because inflation is not low enough. Stuck between the two sides: ballooning balance sheet, moderately soft rates, no clear exit.

Why together → the risk is much greater than alone. In previous cycles, when a central bank was weak (e.g. BoJ in the 90s), capital flowed to a country with a strong central bank (Fed). Confidence in the overall fiat system is not challenged - just one particular currency is weak, others remain safe havens. Now all major central banks are stuck → there is no fiat "safe haven" left → the market can question the entire fiat system instead of just choosing the stronger one.

Two outcomes can occur in parallel or in series:

• Gold/crypto/hard assets absorb capital withdrawn from fiat - "loss of monetary credibility" scenario. Gold from ~$2,000/oz (2023) to ~$5,000 (mid-2026) - up 150% in 3 years even though real interest rates are still positive in most countries - could be the first signal the market is pricing in this risk.
• Government bond yields rise vertically When the market realizes that the central bank cannot keep rates low forever - the "bond vigilantes" return. UK gilt crisis 9/2022 (30Y yield increased from 3.7% to 5.1% in 4 days, almost destroying the UK pension industry) is a small version. When it happens at the scale of the US Treasury or German Bund, the consequences are many times greater - because this is the reference asset of the entire global financial system.

The level and timing are difficult to predict. It could also have been avoided if central banks had coordinated slow, disciplined normalization, and the government had pulled down the deficit. But the characteristics of crises of this type: When the time comes, people will say they should have seen it coming. At that point, national "accounting tips" meet their global limits - and Japan's warning becomes the world's warning.