Public finance

The Finance Division's iBAS++ actuals put tax revenue at 6.79 percent of GDP in FY25 and total revenue at 7.85 percent [MoF iBAS 2025]. That small revenue base limits what Bangladesh can spend on services, infrastructure and financial repair. External public debt service rose to 3,371.6 million USD in FY24 [ERD 2024], adding a growing claim on the same budget. The fiscal question is how to collect more reliably, evaluate subsidies and complete worthwhile projects within a credible spending envelope.

The tax ratio has weakened

The record is a decade of regression, not stagnation. The World Bank central government series shows the tax ratio at 9.02 percent of GDP in 2012 and 8.69 percent in 2011, sliding through 8.50 percent in 2015 and 7.32 percent in 2016 to 7.00 percent in 2017, then 7.73 percent in 2018, 7.64 percent in 2019 and 7.00 percent again in 2020, before 7.64 percent in 2021, the latest year the series carries [WB WDI 2026]. The fiscal year record is worse: the iBAS++ actual for FY25 shows tax collection of 376,945 crore BDT, 6.79 percent of GDP, and total revenue of 436,054 crore BDT, 7.85 percent of GDP, while interest on sovereign debt cost 134,430 crore BDT in the same year, 35.7 percent of tax collected [MoF iBAS 2025].

Revenue excluding grants tells the same story from the wider base, 11.15 percent of GDP in 2012 against 9.54 percent in 2021 [WB WDI 2026]. Total expense was 8.32 percent of GDP in 2021 and the overall balance minus 2.82 percent of GDP [WB WDI 2026]. The 2021 recovery to 7.64 percent is one year of data, not a trend, the FY25 actual sits below it, and the NBR reform agenda of the transition government is the hypothesis it tests. NBR collection by head, VAT versus income tax versus customs, is not held as a series in the available data, so head level shares could not be confirmed and are not quoted; the NBR annual reports would settle it [NBR 2025].

Budget transparency has deteriorated

Spending quality has an independent paper trail. The Open Budget Survey transparency score fell from 58 in 2012 to 41 in 2017 and 30 in 2021, then recovered to 37 in 2023. The 2023 survey scored public participation at 11, legislative oversight at 33 and audit oversight at 45, each out of 100 [IBP Open Budget 2023]. A state that collects little and publishes less gives its own citizens and lenders a credibility discount before any economic argument starts, and chapter 12 carries the governance side of this record.

State enterprises add fiscal exposure

The state owned enterprise layer is a hidden budget. The Finance Division's consolidated accounts of the 20 major SOEs show total liabilities of 3.46 trillion BDT in FY23, up from 2.69 trillion BDT in FY22, against total assets of 4.74 trillion BDT and equity of 1.29 trillion BDT, leaving a debt to equity ratio that rose from 2.12 to 2.69 in one year [MoF Budget 2023]. The operating statement is worse than the balance sheet: sales revenue of 1.94 trillion BDT in FY23 produced operating profit of 111.2 billion BDT, but net profit fell from 18.1 billion BDT to 14.4 billion BDT, dividends paid were 7.0 billion BDT, and the 20 companies repaid 49.3 billion BDT of debt service liabilities to the Treasury [MoF Budget 2023].

A dividend of 7.0 billion BDT on assets of 4.74 trillion BDT is a return of 0.15 percent, which is the fiscal expression of the SOE governance problem in one number. The perimeter is the 20 major SOEs the Finance Division consolidates, not the whole state enterprise sector, so the true stock is larger. The power sector share of those liabilities is the capacity payment story of chapter 10; the point here is that SOE losses are contingent fiscal claims that sit outside the headline deficit the public sees.

Debt service is rising faster than the stock

The flow is the story, not the stock. Total external debt service ran 1,592.5 million USD in FY19, 2,017.0 million USD in FY22, 2,670.2 million USD in FY23 and 3,371.6 million USD in FY24, an increase of 112 percent over five years, split in FY24 between 2,021.8 million USD of principal and 1,349.8 million USD of interest [ERD 2024]. The interest line alone nearly tripled in two years, from 492.1 million USD in FY22 to 935.7 million USD in FY23 and 1,349.8 million USD in FY24, as the zero interest era of LDC concessional lending gives way to semi concessional terms [ERD 2024].

Debt service to revenue and grants reached 12.76 percent in FY23 from 8.3 percent in FY18 [ERD 2023]. Against that flow the stock stays moderate, 15.5 percent of GDP for external debt in FY23 [ERD 2023], and the IMF projects general government gross debt at 41.8 percent of GDP for 2026, from 42.0 percent in 2025 [IMF WEO 2026]. The domestic side is short maturity and bank held, which chapter 06 traces through the banking system's government securities exposure. Official inflows remain substantial, commitments of 10,720 million USD in FY24 against disbursements of 9,892.0 million USD, on par with the 9,440 to 10,180 million USD annual commitment range of FY20 to FY22 [ERD 2024].

After graduation in November 2026, chapter 02 documents the expected shift in the terms of these flows, and the fiscal meaning is a higher average interest rate on the same volume.

Tax exemptions, subsidies and spending choices

The revenue decline has three mutually reinforcing causes. First, structure: the economy's surplus is concentrated in land, informal trade and garments, all three historically under taxed, while the VAT, the workhorse tax, suffers a compliance gap that NBR's own reform documents acknowledge in qualitative terms; the precise VAT gap figures could not be confirmed from the cited sources and are not quoted. Second, administration: the transition government split the revenue board into separate policy and management divisions and pushed the online return agenda, but one year of ratio recovery from 7.00 to 7.64 percent of GDP [WB WDI 2026] is a start, not a regime change.

Third, politics: exemptions, statutory regulatory orders and reduced rates are the budget's favours economy, and every budget speech promises their reduction while the SRO count behaves otherwise; the count itself could not be confirmed from the cited sources. The distributional result is regressive at the margins: the taxes easiest to collect, VAT at the border and on formal transactions, fall on the visible economy, while the gains from land, property and capital income sit largely outside the net, a structure that links this chapter to the inequality record in chapter 01 and the governance record in chapter 12.

The subsidy bill is where world prices meet the budget. Fossil fuel subsidies, the SDG 12.c.1 measure of consumption and production support, jumped from 0.38 percent of GDP in 2020 and 1.98 percent in 2021 to 6.41 percent of GDP in 2022, then fell back to 1.78 percent in 2023 as administered prices were adjusted upward, and to 1.10 percent in 2024 [UNSD 2026, ER_FFS_CMPT_GDP]. On the wider GFS definition, subsidies and other transfers absorbed 34.12 percent of central government expense in 2021 [WB WDI 2026]. The line item subsidy allocations of the FY26 budget, energy versus fertiliser versus food, are not held as a series in the available data and could not be confirmed; the budget documents would settle it, and chapter 10 quantifies the power sector piece from BPDB accounts.

The mechanism is asymmetric: subsidies expand passively when world prices and the exchange rate move against the state, and contract only through administered price increases that are politically expensive, which is why the 2022 spike and the 2023 correction are the decade's template.

Expenditure composition follows the revenue constraint. Interest, subsidies and the wage bill take the first claim on current spending, development spending takes what remains, and within development the two big sectors noted below crowd out health, social protection and maintenance.

Development spending and project delivery

The FY26 development programme approved by the National Economic Council on 18 May 2025 totals 230,000.00 crore BDT, financed by 144,000.00 crore from the exchequer and 86,000.00 crore in project loans and grants, and 238,695.64 crore once the 8,695.64 crore of own funds of autonomous bodies and corporations is added, across 1,113 projects in the programme proper and 1,173 including the own funded ones [Planning Commission ADP 2025, ADP at a glance table]. Transport and communication takes 58,973.39 crore BDT or 25.64 percent of the approved programme and power and energy 32,392.26 crore BDT or 14.08 percent, the two largest sectoral claims and 39.72 percent of the programme between them [Planning Commission ADP 2025, ADP at a glance table].

What the protected sectors crowd is measurable in the same table: health takes 18,148.14 crore BDT or 7.89 percent and agriculture 10,795.97 crore BDT or 4.69 percent, while education takes 28,557.43 crore BDT or 12.42 percent, housing and community facilities 22,776.40 crore BDT or 9.90 percent, and local government and rural development 13,472.26 crore BDT or 5.86 percent [Planning Commission ADP 2025, ADP at a glance table]. The development budget is a transport and power programme with a social sector annex, and the annex is where execution slips first when the year turns tight. The size of the programme against the outturn shows the gap: development expenditure actually executed in FY25 was 156,706 crore BDT [MoF iBAS 2025], so the FY26 allocation of 230,000 crore BDT is 46.77 percent above the previous year's spend.

The project level record shows where the money concentrates. The FY26 RADP project table in the available data records 1,356 projects with 166,900.5 crore BDT of current year allocation, of which project aid finances 60,290.6 crore BDT or 36.12 percent, and the ten largest projects alone absorb 31,201.3 crore BDT or 18.7 percent of the allocation [Planning Commission ADP 2025, fiscal year 2025-26]. A programme whose tail is 1,356 projects and whose head is ten megaprojects is a programme built for allocation, not completion: execution inside the year follows a known rhythm in which allocations thin out in the first half while procurement runs, spending concentrates in the fourth quarter, and uncompleted allocations lapse.

The IMF programme, in its ECF and EFF reviews, has made ADP execution quality and revenue mobilisation its two structural anchors; the programme's specific quantitative targets by review could not be confirmed from the cited sources, and the reviews themselves would settle it [IMF Art IV 2025].

A 2.3 trillion taka development programme

The signal is in the arithmetic, not the speech. A development programme of 230,000 crore BDT [Planning Commission ADP 2025] financed by a state whose latest measured tax ratio is 7.64 percent of GDP [WB WDI 2026] forces the gap into three channels: bank borrowing, 114,161 crore BDT net in FY25 [MoF iBAS 2025], which transmits the state's financing need into the private credit squeeze chapter 06 measures; national savings certificates, whose rates the deposit market must follow and chapter 05 measures; and external aid, 9,892.0 million USD disbursed in FY24, whose concessionality graduation will erode [ERD 2024].

The SOE channel adds a fourth, quieter source of fiscal risk: companies carrying 3.46 trillion BDT of liabilities that paid their owners 7.0 billion BDT in dividends in FY23 [MoF Budget 2023]. The headline revenue and deficit figures of the FY26 budget itself are not held as a series in the available data and could not be confirmed; the budget documents would settle it, and the series above are the structure the budget documents must reconcile with.

Fiscal scenarios depend on revenue

The fiscal arithmetic the chapter 15 scenarios assume runs as follows. The reform scenario assumes the tax ratio reaches 10 percent of GDP by around FY31, the minimum consistent with financing the interest bill, the post reform subsidy bill, the climate adaptation plan in chapter 11 and a development programme that maintains infrastructure without new SOE arrears; the baseline scenario assumes the ratio drifts to 9 percent and the stall scenario assumes it stays near 8. The external debt service flow, already 3,371.6 million USD in FY24 [ERD 2024], is the projection with the widest confidence band because it depends on the post graduation borrowing terms chapter 02 documents, and the ERD commitment and disbursement series are the leading indicators.

Three decisions decide which path holds. The NBR reform decision, whether the online return, electronic payment and data matching agenda becomes binding on the large taxpayer base, is the highest leverage fiscal act available in the window the IMF programme targets, before FY30. The subsidy pricing decision, whether fuel and fertiliser prices track cost with targeted cash transfers replacing blanket subsidy, decides whether the subsidy bill compounds or stabilises, and chapter 10 carries the energy side. The SOE decision, whether power sector arrears are recognised and restructured, decides how much of the debt stock is real, and chapters 10 and 06 carry the sector and banking sides.

Risks and opportunities

Risks. First, the revenue trap: if the tax ratio stalls below 8 percent of GDP, the state finances the decade with arrears and accounting, and the indicators to monitor are the next observations of the World Bank tax series [WB WDI 2026] and the NBR outturn against target in the annual reports [NBR 2025]. Second, hidden liability crystallisation: if SOE arrears, chiefly the power sector's, are suddenly recognised as government obligations, the debt ratio jumps without any new borrowing, and the indicator to monitor is the consolidated SOE liabilities series, already up 29 percent in one year to 3.46 trillion BDT [MoF Budget 2023].

Third, an aid terms shock: if post graduation borrowing terms shift faster than projected, debt service to revenue, at 12.76 percent in FY23 [ERD 2023], rises past the low twenties by the end of the decade, and the indicator to monitor is the average interest rate on new commitments in the ERD annual tables.

Upside. First, the formalisation dividend: the payment and digital trail documented in chapter 13 makes the informal economy taxable at lower administrative cost than at any point in NBR's history. Second, the transparency lever: an Open Budget score that fell from 58 in 2012 to 37 in 2023 [IBP Open Budget 2023] has fallen far enough that even a partial recovery is a cheap, credible signal to lenders and rating agencies. Third, ADP quality: cutting the project count from 1,113 [Planning Commission ADP 2025] toward a few hundred completable projects raises execution without one additional taka of allocation. Fourth, the asset lever: the SOE portfolio holds 4.74 trillion BDT of assets against dividends of 7.0 billion BDT [MoF Budget 2023], and closing even part of that gap funds the social sector annex of the ADP outright.

Testing the policy case

The Finance Division and NBR should assess revenue reform against collection, compliance costs and the services funded. A rising tax ratio with growing arrears or stalled priority projects would weaken the claim that revenue alone was the binding constraint; expenditure management and political allocation would require separate action.

What to watch

The thresholds below are author-proposed monitoring markers, not official targets or estimated policy effects. Interpret them alongside the source dates and definitions.

1. Tax revenue to GDP. Latest cited value 7.64 percent of GDP in calendar 2021, the latest observation in the World Bank series [WB WDI 2026], and 6.79 percent in FY25 on the Finance Division's iBAS++ actuals [MoF iBAS 2025]. Threshold: a sustained print above 9 percent would be consistent with the reform scenario; a slip back under 7 percent would be consistent with the stall.
2. External debt service. Latest cited value 3,371.6 million USD in FY24 [ERD 2024]. Threshold: a rise past 5,000 million USD a year before post graduation terms are locked is the terms of trade shock for the budget.
3. SOE consolidated liabilities. Latest cited value 3.46 trillion BDT in FY23 [MoF Budget 2023]. Threshold: growth above nominal GDP growth for two consecutive years is the arrears regime the reform must prevent.
4. ADP size and project count. Latest cited value 230,000.00 crore BDT approved and 1,113 projects in FY26 [Planning Commission ADP 2025]; the available dataset's ADP total allocation series reads 255,267.44 crore, which is the project level sum of the FY2023-24 programme book and not an FY26 figure, and is not used. Threshold: a budget that cuts the project count below 1,000 while holding the allocation signals execution reform; a larger ADP with more projects signals the old regime.
5. Budget transparency. Latest cited value 37 out of 100 in 2023 [IBP Open Budget 2023]. Threshold: a score above 50 in the next survey wave is the credibility regime change.

Sources used

[WB WDI 2026] World Bank World Development Indicators via bdpolicy.db and the indicators snapshot updated 2026-09-04, series: GC.TAX.TOTL.GD.ZS, GC.REV.XGRT.GD.ZS, GC.XPN.TOTL.GD.ZS, GC.NLD.TOTL.GD.ZS, GC.XPN.TRFT.ZS, central government basis, calendar years.

[ERD 2023] Economic Relations Division, Ministry of Finance, external debt and debt service series via bdpolicy.db, series: erd_foreign_debt_gdp_pct, erd_debt_service_revenue_pct.

[ERD 2024] Economic Relations Division, Ministry of Finance, debt service and aid series via bdpolicy.db, series: erd_debt_service_total_usd_mn, erd_debt_service_principal_usd_mn, erd_debt_service_interest_usd_mn, erd_foreign_aid_commitment_usd_mn, erd_foreign_aid_disbursement_usd_mn.

[MoF Budget 2023] Finance Division consolidated SOE evaluation, SPFMS Monitoring Cell via bdpolicy.db, series: soe_consolidated_total_liabilities_bdt_m, soe_consolidated_total_assets_bdt_m, soe_consolidated_total_equity_bdt_m, soe_consolidated_debt_equity_ratio, soe_consolidated_sales_revenue_bdt_m, soe_consolidated_operating_profit_bdt_m, soe_consolidated_net_profit_bdt_m, soe_consolidated_dividend_paid_bdt_m, soe_consolidated_dsl_paid_bdt_m.

[Planning Commission ADP 2025] Annual Development Programme 2025-2026, Programming Division, Bangladesh Planning Commission, June 2025: the ADP at a glance sector table and the ministry wise allocation table, in crore BDT, with every share taken on the 230,000.00 crore approved programme, , accessed 6 September 2026. The same tag covers the revised FY2025-26 project table held in the available data as bdpolicy.db adp_projects, 1,356 projects, allocations and project aid in lakh BDT converted to crore, a later and separate vintage that is named as the revised programme wherever it is used.

[IBP Open Budget 2023] International Budget Partnership, Open Budget Survey 2023, Bangladesh country results: transparency 37, participation 11, legislative oversight 33 and audit oversight 45, each out of 100. The survey assesses material through 31 December 2022. Country results, checked 7 September 2026.

[IMF WEO 2026] IMF World Economic Outlook database via bdpolicy.db indicators snapshot updated 2026-09-04, series: imf_gross_debt_pct_gdp, general government gross debt percent of GDP, 2025 value and 2026 projection, confirmed against the IMF DataMapper GGXWDG_NGDP series for BGD on 2026-09-06.

[MoF iBAS 2025] Finance Division iBAS++ fiscal year actuals via bdpolicy.db, FY25 (July 2024 to June 2025), series: mof_actual_tax_revenue, mof_actual_revenue_total, mof_tax_gdp_ratio_actual, mof_revenue_gdp_ratio_actual, mof_sovereign_debt_interest_service, mof_actual_development_expenditure, mof_bank_borrowing_net_deficit, ratios on the 2015-16 GDP base.

[IMF Art IV 2025] IMF Article IV and ECF/EFF review documents for Bangladesh, programme structural anchors, resolving source for review level quantitative targets.

[UNSD 2026] UN Statistics Division, SDG Global Database, indicator 12.c.1 fossil fuel subsidies as a proportion of GDP, series ER_FFS_CMPT_GDP, pulled from the UN SDG API (unstats.un.org/sdgapi, Bangladesh geoAreaCode 50) on 6 September 2026, calendar years, source line IEA fossil fuel subsidies database, OECD.Stat and IMF Energy Subsidies Template. An earlier draft carried 0.44, 2.32, 7.39 and 2.07 percent for 2020 to 2023 from the bdfacts snapshot; those values do not match the API and are not used, the chapter 41 correction applying here too.

[NBR 2025] National Board of Revenue annual reports, resolving source for collection by head and outturn against target.