Honest Accounting

Executive Summary

Bangladesh's banking sector non-performing loan ratio reached a record 35.73 percent of total disbursed loans at the end of September 2025, on Bangladesh Bank data published in November 2025, after Basel III reclassification ended years of understated reporting; it eased to 30.6 percent by December 2025. The Sammilito Islami Bank merger, absorbing five failed banks at a paid-up capital of Tk 35,000 crore, is the largest single banking-sector intervention in the country's modern history. Governor Ahsan H Mansur's forced departure in February 2026 and the appointment of a BGMEA official as his successor introduce a new political-economy risk to a cleanup that remains structurally incomplete.

A textile-trim importer in Narayanganj walked into the branch that used to be his Social Islami Bank in early May 2026 and asked to withdraw Tk 8 lakh from his savings account, an account he had held continuously since 2014. The branch is now signposted "Sammilito Islami Bank PLC", the state-owned entity into which his old bank and four others were merged in late 2025 under the Bank Resolution Ordinance. The counter officer ran the request through the new withdrawal regime: Tk 2 lakh as an initial release, then Tk 1 lakh every three months over a 21-month phased schedule, full access restored by the early months of 2028. There was a window for medical emergencies (up to Tk 10 lakh on documentation; unlimited for certified critical illness) and a one-time exception schedule the bank's depositor desk could review. He took the Tk 2 lakh.

This is not a story about a single bank failing. It is a story about five banks failing simultaneously, being declared failed by their regulator, being taken over, being merged, and the depositor being asked to wait two years for his own savings while the state writes off the political-loan portfolio that destroyed the franchise value of the accounts the savings sat in. The five were EXIM Bank (controlled until 2024 by Nazrul Islam Mazumder, former chairman of the Bangladesh Association of Banks), First Security Islami, Global Islami, Social Islami, and Union, the last four substantially captured by family members of Mohammed Saiful Alam, chairman of S Alam Group. Administrators arrived on November 5, 2025. The letter of intent for the merged entity was issued November 9. Sammilito Islami Bank PLC began operations on December 2, 2025, and the final licence followed in early 2026. The deposit base is paid back, in pieces, on the schedule the importer just heard at the counter.

The cleanup happened. The political-economy question now is whether the new government in office will preserve it or unwind it.

What the published numbers now say

Bangladesh's banking-sector accounts changed shape in late 2024. Bangladesh Bank, then under Governor Ahsan H Mansur, adopted Basel III loan classification rules that ended the long-standing practice by which a loan rescheduled three or four times could continue to sit on the books as performing. The reported headline non-performing loan ratio had stood at 10.11 percent in June 2023. By the end of December 2024 it was 20.20 percent, equivalent to Tk 3.45 trillion in bad credit. By March 2025 it was 24.1 percent. By the end of September 2025 it had reached a record 35.73 percent of total disbursed loans, equivalent to Tk 6.44 trillion, before easing to 30.6 percent by December 2025. The jump was driven by two forces working together. The classification change forced years of evergreened loans onto the books at once, and the same window saw genuine new defaults and the collapse of the S Alam-linked banks that were resolved in late 2025. Reclassification revealed the size of the problem; it did not, by itself, create all of it.

The decomposition by bank type shows where the bad debt actually sits. State-owned commercial banks hold about a quarter of system assets but carry an NPL ratio far above the private-bank average. Bangladesh Bank has imposed reduction targets: state-owned banks must cut their NPL ratio to 10 percent by June 2026, private banks to under 5 percent. Whether either deadline is met depends on enforcement choices that fall to the new central-bank leadership, not to the leadership that wrote the targets.

The scale of the problem becomes clearer against regional peers. Measured on the same World Bank Financial Soundness Indicators basis (latest available, around 2022 to 2023), Pakistan reported an NPL ratio of 6.6 percent, Thailand 2.8 percent, Vietnam 2.3 percent, Indonesia 2.0 percent, and India 1.7 percent. Against the Basel III-honest Bangladesh figure of 24.1 percent in March 2025, the gap is not marginal: Bangladesh is operating at roughly fourteen times India's ratio and more than ten times Vietnam's, on those FSI figures. Bangladesh Bank's own end-September 2025 reading of 35.73 percent placed the country highest in the world, above Ukraine's wartime 26 percent. The depositor who accepted a phased withdrawal schedule and the importer who needs a trade-finance line that prices in counterparty risk are both priced on the honest number, not the reported one.

The external-sector picture has improved more sharply than most observers expected. Foreign exchange reserves had fallen to roughly USD 21.4 billion at the end of 2024. As of May 10, 2026, gross reserves stand at USD 34.14 billion and the BPM6-standard measure at USD 29.47 billion, both held up by sustained remittances (record USD 26.9 billion in 2024 and higher in the first months of 2026) and the temporary respite from import compression. Four-and-a-half months of import cover is not generous but it is not the crisis-margin position it was eighteen months ago.

The story in those two paragraphs is the operational reform under the post-August 2024 interim government. Five banks declared failed. NPL classification honest. Reserves restored. Sammilito Islami Bank PLC capitalised at Tk 35,000 crore in paid-up capital (Tk 20,000 crore from the state, Tk 15,000 crore from depositor and institutional balances converted to shares, two years of interest accrual scrapped, depositors losing roughly Tk 15,000 crore in interest income across the merged entities). It is the largest single banking-sector intervention in the country's modern history.

Then the architect was fired

On February 16, 2026, the central bank's own Officers' Welfare Council publicly denounced the governor. By February 25 a gazette notification appointed Md Mostaqur Rahman, managing director and chief executive of Hera Sweaters Limited and standing-committee chairman of the Bangladesh Garment Manufacturers and Exporters Association, as the new Bangladesh Bank governor for a four-year term. He had served on the BNP's election steering committee during the parliamentary vote earlier that month. The Daily Star editorialised that the dismissal was "shocking". The Asia Times headline read "Why Bangladesh just fired the man who saved its economy". This was the first time in the country's history that a businessman, with current ties to the export industry the central bank regulates, has held the office.

In his first weeks the new governor made public commitments that were almost the opposite of what his appointment had been read to signal. The Sammilito Islami Bank merger would continue, he said, with no rollback of the resolution framework. He would not bow to political pressure. He would protect Shariah board independence at Islamic banks. He floated an eleven-point reform agenda. He proposed lowering interest rates to revive private credit growth. He instructed the system to introduce service-level agreements on application-decision timelines.

The reading inside the banking community is that the Sammilito merger is now too far advanced to reverse without a depositor crisis the new government cannot afford to trigger. The harder question is what happens at the next decision: which surviving private bank gets resolution next, which captured exposure gets written down, which BGMEA-member borrower gets a debt forbearance the regulator's previous leadership would have refused. The architecture is intact; the discretion has changed hands.

The depositor in Narayanganj does not see this argument at the counter. What he sees is that his Tk 8 lakh will return in pieces over twenty-one months, that he is owed two years of interest he will not receive, and that the bank he held an account at for eleven years has been replaced by a state-owned entity with a different signpost.

The cost paid, the cost still to pay

What the post-2024 banking reform cost the system is now partly calculable, partly still modelled. Against a nominal GDP of Tk 55.5 trillion in FY25, the Sammilito merger's Tk 35,000 crore in paid-up capital is roughly 0.6 percent of GDP. The forgone depositor interest of approximately Tk 15,000 crore is another 0.3 percent. The system-wide NPL stock of Tk 6.44 trillion is itself about 11.6 percent of GDP. How much of that converts into a capital hole depends on the provisioning coverage assumed: at a standard coverage shortfall on the order of 30 to 35 percent of the recognised bad stock, the system-wide additional provisioning and write-down requirement is roughly Tk 1.9 to 2.2 trillion, or about 3.5 to 4 percent of GDP, if every loan now classified non-performing is provisioned to Basel III minimums. That figure is a modelled estimate, not a published audit total. The fiscal carry of state-owned bank recapitalisation through bond injection (the path Bangladesh has used in prior cycles) is on top of that.

The chart below isolates one component of that estimate: BDPolicyLab's bank-by-bank capital-shortfall model for the six state-owned commercial banks, which sums to roughly Tk 555 billion, about 1 percent of GDP. These are model estimates, not reported figures, and they cover only the state-owned subset; the 3.5-to-4-percent system-wide number above includes the captured private banks and specialized banks on top.

The cost still to pay falls into three categories. First, the residual NPL clean-up at the remaining captured private banks and the six state-owned commercial banks. Bringing the state-owned banks to the 10 percent NPL target by June 2026 implies roughly Tk 1 trillion of provisioning and write-down beyond what has already happened, on the same modelled basis. Second, the conversion of the Sammilito depositor liability into a sustainable balance sheet, which will require ongoing capital support and the absorption of remaining bad assets into an asset-management vehicle. Third, the prosecution and asset recovery against the named beneficiaries of the original political-loan portfolio, which will determine whether the recapitalisation is a transfer from taxpayers to looters or a transfer from looters back to the public balance sheet.

The third track is the one that determines whether the bank cleanup compounds into the rest of the reform agenda or stops at recapitalisation. Without asset recovery, the precedent is set that capture has no cost beyond the loss of operational control of the franchise. With asset recovery, the precedent is that the state can take the assets back, which materially changes the calculation for the next would-be captor.

The Tarique Rahman government has spoken about asset recovery in general terms in its 180-day priority plan. It has not yet named the first beneficiary against whom a recovery case will be filed. The window in which prosecution is politically possible is narrowing as new business relationships form.

The connected balance sheet

The bank capital hole is connected to the energy-sector capital hole. The Bangladesh Power Development Board ran an aggregate revenue shortfall of about Tk 556.6 billion in FY25, a loss of roughly Tk 5 per kilowatt-hour on the electricity it sold, according to IEEFA's reading of the BPDB accounts. The government covered the bulk of it with a record subsidy injection of about Tk 386 billion, leaving a residual after-subsidy loss on the order of Tk 170 billion for the year. Capacity charges paid to private power producers, including for plants that were not dispatched, remain one of the largest single cost lines: the bill for the twelve months to June 2024 was about Tk 409 billion, and BDPolicyLab's FY25 energy model puts it in a similar range. The new government has signalled that it will renegotiate these contracts. The unpaid bill to India's Adani Power for electricity from the Godda plant in Jharkhand peaked near USD 900 million during 2024-25 before large repayments in mid-2025, and was reported back around USD 850 to 900 million in arrears by late 2025.

The energy-sector capital problem is therefore on the same balance sheet as the banking-sector capital problem. The IPP capacity charges flow through state-owned bank loan portfolios. The unpaid Adani liabilities sit on BPDB's balance sheet, which the state stands behind. The decisions Bangladesh Bank's new leadership makes on private-credit policy interact with the decisions the new power-sector leadership makes on IPP renegotiation, which interact with the fiscal envelope set by the next budget the Tarique Rahman government tables. None of these can be optimised separately.

What honest accounting looks like, from here

A continuation of the post-2024 cleanup that does not roll back the resolution framework requires three things from the new governor and the new finance ministry, none of which are technically novel and all of which depend on political will.

First, complete the publication of bank-level forensic audit findings for every commercial bank with a non-performing loan ratio above twenty percent on Basel III classification. The Sammilito merger was preceded by these audits at the five failed banks; the same exercise is necessary at the surviving captured private banks and at the six state-owned commercials. Without the underlying numbers in the public domain, the recapitalisation discussion at the next budget is uninformed and the political-economy pressure to soften the terms is unconstrained.

Second, name the asset-recovery cases. The S Alam group is the obvious starting point. The political signal of the first filed case will determine whether the prosecution track is real or theatrical. The new BB Governor has the statutory authority to refer cases; the finance ministry and the law ministry have the prosecutorial authority. The 180-day window is roughly half-gone.

Third, lock in central-bank independence in statute, not in the personal undertakings of a serving governor. The new governor's verbal commitment not to bow to political pressure is welcome; the institutional protection that survives his successor is what matters. The 2026 reform window is the moment to amend the Bangladesh Bank Order to make the operational independence of the rule-enforcement function irreversible, not contingent.

The depositor in Narayanganj will receive his Tk 8 lakh by the early months of 2028. Whether the bank he hands his next deposit to is one he can trust depends on what is decided in the next six months. The infrastructure of the cleanup exists. The question is whether the new political leadership chooses to preserve, extend, or quietly hollow it out.

The reform that the next reform sits on

The banking cleanup, costed on a modelled basis, sits at roughly 3.5 to 4 percent of GDP across recapitalisation, asset write-down, and the residual fiscal carry of the resolution framework. That is roughly the same magnitude as the entire tax-expenditure leakage discussed in the first piece in this series. The two reforms are interlinked: the state cannot finance the banking cleanup without the tax revenue, and the productive economy cannot absorb the capital deployed without a banking system that intermediates honestly.

The five banks that failed did so because the regulator was not allowed to enforce its own rules during the 2010 to 2024 period. The five could be merged because the rules were briefly enforced in late 2024 and through 2025. Whether the next round of failed banks gets the same treatment is the question the new BB Governor has not yet been asked to decide on. When he is asked, the depositor in Narayanganj, the importer in Mirpur, and the boro farmer in Sunamganj will be on the answer.

Sources

• Sammilito Islami Bank final licence: thedailystar.net/business/news/sammilito-islami-bank-gets-final-licence-4047446
• Five merging banks Bangladesh Bank preliminary licence: tbsnews.net/economy/banking/sammilito-islami-bank-gets-preliminary-license-bangladesh-bank-1281276
• Sammilito Islami Bank Tk 35,000 crore paid-up capital, five banks, operations from December 2, 2025: en.prothomalo.com/business/local/mvn48uemhn
• Bangladesh NPL 35.73% (Tk 6.44 trillion, end-September 2025) and 20.2% (Tk 3.45 trillion, December 2024), Bangladesh Bank data, with peer comparison: en.bonikbarta.com/business/6ZAHEFUvFGC0bkfL
• Bangladesh NPL ratio eased to 30.6% in December 2025: ceicdata.com/en/indicator/bangladesh/non-performing-loans-ratio
• Peer NPL ratios (World Bank Financial Soundness Indicators, latest available): data.worldbank.org/indicator/FB.AST.NPER.ZS
• Ahsan H Mansur ouster, February 2026: thedailystar.net/news/bangladesh/news/shock-ouster-bb-governor-4115256
• Mostaqur Rahman appointment: tbsnews.net/bangladesh/process-underway-appoint-new-bb-governor-replacing-ahsan-h-mansur-sources-1370886
• New BB Governor reform pledge: dhakatribune.com/business/banks/404486
• Forex reserves USD 34.14bn, May 10, 2026: bssnews.net/news-flash/380959
• Bangladesh nominal GDP FY25 (Tk 55.5 trillion, current prices, BBS): tbsnews.net/economy/bangladeshs-gdp-growth-revised-down-349-fy25-1371596
• BPDB FY25 revenue shortfall (Tk 556.6 billion), government subsidy (Tk 386 billion), and FY24 capacity charge (Tk 409 billion), IEEFA analysis of BPDB accounts: ieefa.org/articles/bangladesh-power-development-board-can-save-us12-billion-annually-through-key-electricity
• Adani Power arrears (peak near USD 900 million; mid-2025 repayments): tbsnews.net/bangladesh/infrastructure/adani-pushes-payment-roadmap-pdb-arrears-mount-845m-1038346
• Bank-level state-owned capital-shortfall figures and the 3.5-to-4-percent-of-GDP system-wide provisioning estimate are BDPolicyLab banking-sector model outputs, not published audit totals.