The Reading Test

Executive Summary. Bangladesh gets 98 percent of its primary-age children into school, yet 51 percent of them cannot read and understand a simple grade-level passage by age ten, per the World Bank's April 2024 Learning Poverty brief. The binding constraint is not access; it is learning. The country spends 2.03 percent of GDP on education (World Bank, 2024), against Vietnam's 2.89 percent and India's 4.10 percent, and the outcome gap tracks the spending gap. The Tarique Rahman BNP government's pledge to lift education spending to 5 percent of GDP would close the spending gap on paper, but at the current tax-to-GDP floor near 8 percent the money does not exist: the pledge is a revenue-reform claim, not an education claim. The falsifiable test of this argument is simple. If learning poverty does not fall when a national Class 3 and Class 5 reading assessment names underperforming schools, the diagnosis here is wrong. Until that assessment runs, the policy machinery is flying blind on its single largest economic problem.

In a government primary school in a union of Manikganj district on a humid morning in late March 2026, a Class 5 teacher named Nazma Khatun handed each of her thirty-eight students a single A4 page with a six-sentence Bangla passage about a fisherman in Barishal who lost his boat in a 1991 cyclone and rebuilt it with help from a cooperative society. She asked them to read it silently and then to answer three questions written below the passage: what year did the cyclone happen, what did the fisherman lose, and what did the cooperative society do to help him. She gave them twenty minutes.

By the seventh minute, four students had finished and were sitting quietly. By the fifteenth minute, eleven students had completed at least the first answer. By the end of the twenty minutes, twenty-two of the thirty-eight had written something on the page that demonstrated they had understood at least the first sentence. The remaining sixteen had not. Some had copied words from the passage into the answer space. Some had drawn straight lines. Three had not picked up their pencils. A few were staring at the page with the same expression that an adult might use to look at a passage in a script they cannot read.

This was not a diagnostic exam. It was Nazma's private experiment, run after morning assembly, with materials she had photocopied at her own expense from a 2019 BRAC reading-benchmark booklet she found at a second-hand book stall in Manikganj town. The system she works in does not ask her to run this test. The system she works in does not ask the question the test answers. The system she works in promotes all thirty-eight students to Class 6 in November regardless of whether they can read the passage.

The official primary net enrollment ratio is 98 percent. The official learning poverty rate, the share of ten-year-olds who cannot read and understand a simple paragraph, is 51 percent (World Bank, April 2024). Both numbers are real. Both numbers describe the same children. The country has solved a problem that was largely solved by 2010 (getting children to school) and has not solved the problem that was unsolved then and remains unsolved now (whether they learn).

The number we celebrate, the number we hide

Bangladesh's headline education statistics look impressive from the angle that publishes them. Primary net enrollment of 98 percent is among the highest in lower-middle-income countries. Gender parity in primary and secondary enrollment has been better than 1.0 (more girls than boys) for over a decade, the result of the female stipend programmes that began in the mid-1990s. Adult literacy reached 79 percent in 2022 per World Bank data, with female literacy at 76.5 percent. Education spending was 2.03 percent of GDP in 2024 according to World Bank data, on the same order as Pakistan (1.95 percent) and below Thailand (2.52 percent), Vietnam (2.89 percent), and India (4.10 percent at the consolidated central-plus-state level). Education accounts for 11.9 percent of total government expenditure.

The numbers underneath are different.

Learning poverty in Bangladesh is 51 percent, the World Bank's April 2024 figure, derived from grade-5 reading data measured in 2022. This is the share of children at the end of primary age who cannot read and understand a simple age-appropriate text. It sits below the South Asia average by 8 percentage points, but it remains worse than every East and Southeast Asian comparator. Vietnam's learning poverty is 18 percent and Thailand's is 23 percent. It is roughly level with Indonesia's 53 percent, better than India's 56 percent and Pakistan's 77 percent, and far behind the East Asian frontier. The pattern is unambiguous: Bangladesh is at the top of its South Asian peer group and at the bottom of the wider Asian one.

The dropout numbers tell a parallel story. The official primary dropout rate is approximately 18 percent. The secondary dropout rate is approximately 37 percent. By the end of secondary school, more than a third of the children who entered first grade have left the system before completing tenth grade. They leave with formal completion certificates for whatever stage they have reached but, in many cases, without the literacy and numeracy that those certificates are supposed to attest.

The teacher input is below comparable benchmarks. The pupil-teacher ratio in primary schools is approximately 30 students per teacher (2018 data; current data suggests modest improvement to around 28). Vietnam's primary PTR is 20. Thailand's is 17. Indonesia's is 17. India's is 33. The South Asian comparison is a wash; the East Asian comparison is a different country. Trained-teacher share in primary is approximately 78 percent, which means roughly one in five primary teachers does not have formal pedagogical training.

The tertiary outcome is worse. Tertiary gross enrollment is 22 percent, low by lower-middle-income standards. Among tertiary graduates, official graduate unemployment is approximately 12 percent. The Human Capital Index measured by the World Bank put Bangladesh at 0.46 in the 2020 vintage (boys 0.45, girls 0.48), meaning a child born today will achieve 46 percent of the productivity she would achieve under complete education and full health. Vietnam scored 0.69. Indonesia 0.54. India 0.49.

This is a system that has solved access and has not solved quality. The autopass episodes during the COVID-19 years and the SSC and HSC examination credibility crises of the same window made it worse rather than better. The reading test that Nazma ran in her Class 5 in Manikganj is the symptom of a structural failure that the policy machinery has been unwilling to acknowledge.

Where the spending goes, and where it does not

The 2.03 percent of GDP spent on education breaks down predictably. About 70 percent of the budget goes to salaries. Within salaries, the largest line is primary teacher pay, then secondary teacher pay, then administrative salaries. Capital expenditure on school construction, classroom equipment, libraries, science lab equipment, and digital infrastructure is a residual category. The non-salary recurrent budget per school per year, the line that funds chalk, exam paper, teacher training, and basic operational costs, is among the lowest per-pupil expenditures in the lower-middle-income comparison set. Primary education expenditure per child in Bangladesh is USD 223 (PPP), which the World Bank's April 2024 brief puts 81 percent below the South Asia average and 73 percent below the lower-middle-income average.

The structural problem this creates is straightforward. Bangladeshi primary teachers are paid modestly by labour-market standards: a Grade 12 government primary teacher starts at approximately the same gross salary as an entry-level garments-factory line supervisor. The implicit signal is that the labour market values the supervisor's skills more than the teacher's. Predictably, recruitment into the teaching profession has not attracted the upper bands of the tertiary graduate pool, except in the sense that those who cannot find higher-paying employment elsewhere fall back into the teaching pipeline. The patronage hiring channel for assistant teachers, where local political networks rather than competitive examinations have determined recruitment in many districts during the 2010s, made the selection problem worse.

The result is a teacher population in which the modal primary teacher is competent and dedicated but is working with a class size, a curriculum, and a non-salary budget that do not let her teach effectively. Underneath the competent core sits a measurable gap: roughly one primary teacher in five lacks formal pedagogical training (the trained-teacher share is about 78 percent), and the districts where patronage hiring ran hottest in the 2010s are where that gap concentrates. The reading test in Manikganj is what happens when this population is asked to deliver foundational literacy to thirty-eight students at a time with the materials and the training that the system has historically provided.

The reforms that have been studied to death

The basic reform set is not controversial inside the education policy community. It has been written into successive Education Policy documents and ESPSE (Education Sector Programme) frameworks for fifteen years. The reason the documents have not become the system is the political economy of teacher recruitment, the political economy of the madrasah education stream, and the budget constraint inherited from the eight-percent tax floor.

The set, in priority order for the next five years, is the following.

First, annual standardised learning assessment at Class 3 and Class 5, results published at school level, with school-level remediation plans triggered by below-threshold results. Owner: the Directorate of Primary Education, with the National Curriculum and Textbook Board on instrument design. This is the single highest-yield reform because it forces the data into the system. India's ASER (Annual Status of Education Report) and Pratham's reading benchmarks are the model; Bangladesh has run smaller equivalents through BRAC and a2i but never at universal scale. The cost of running a national Class 3 and Class 5 reading and arithmetic benchmark is modest, probably under Tk 200 crore annually. The success signal is concrete and falsifiable: the published share of Class 5 students reading at grade level rises year on year in the districts that started worst. The political cost is that the first results will be embarrassing and will name underperforming districts and schools. That is precisely why the reform has not happened, and precisely why it must go first.

Second, teacher recruitment through competitive examination at the BCS-equivalent grade, with district-level recruitment quotas based on demand, replacing the local patronage hiring that has filled assistant-teacher positions in many districts. Owner: the Ministry of Primary and Mass Education with the Public Service Commission. This is the political-economy keystone. The pre-2024 government's teacher recruitment process had been substantially undermined by patronage; the post-2024 reform window allows this to be rebuilt. The 2025 BCS-recruitment teacher batch has begun to reverse the pattern in selected districts. The success signal is the share of new assistant-teacher posts filled through open examination rather than discretionary channels, reported district by district. Universalising this is twelve to twenty-four months of administrative work and ten years of cohort replacement before the teacher population is meaningfully changed.

Third, the madrasah education quality integration. Bangladesh has approximately 3.5 million students in formal Alia and Qawmi madrasahs (BANBEIS, 2023). The political reality is that madrasah education is not going away and any policy that pretends it will is fantasy. The achievable reform is the parallel quality reform: a baseline curriculum that includes Bangla, mathematics, science, English, and basic civic competencies, with examination at the same Primary Education Completion and SSC equivalents that the secular stream uses. Owner: the Ministry of Education in negotiation with the Bangladesh Madrasah Education Board and the Qawmi board. The Qawmi recognition framework agreed in principle in 2017 was a precedent. Operationalising it now, with budget support tied to curriculum compliance, is the next step. The success signal is the share of madrasah-stream candidates sitting and passing the common foundational examinations. The political cost is concentrated on the madrasah-board leadership; the benefit is concentrated on the future employment prospects of the 3.5 million students currently completing certificates that the labour market does not value.

Fourth, TVET expansion physically co-located with industrial parks. Current TVET enrollment is approximately 14 percent of the relevant age cohort. The target in successive policy documents has been 30 percent by the early 2030s, in line with the upper-middle-income Asian benchmark. Owner: the Directorate of Technical Education with BIDA and the Bangladesh Economic Zones Authority. The implementation gap has been that TVET institutions have been built in administrative locations rather than at industrial demand-points. The next-decade build should put new polytechnics and skills institutions inside or adjacent to the SME industrial parks in Mymensingh, Faridpur, Khulna, Cumilla, and the Mongla-Khulna economic zone, with curricula co-designed with the resident industry. The success signal is the placement rate of graduates into firms in the co-located zone within six months. Vietnam's vocational education boom, which underwrote much of its electronics-and-textile labour pipeline in the 2010s, used exactly this co-location pattern. Bangladesh has not.

Fifth, secondary-school examination credibility restoration. The autopass episodes between 2020 and 2023 and the question-paper leaks of the same window destroyed public confidence in the SSC and HSC examinations. Owner: the education boards under the Ministry of Education. The Yunus interim administration in 2025 began the technical fixes: encrypted question-paper distribution, biometric exam-centre attendance, third-party invigilation in selected districts. The Tarique Rahman BNP government, sworn in on February 17, 2026, has inherited that pilot and now has to make it universal and sustain it for a full secondary cohort cycle (roughly six years). The success signal is a measured decline in confirmed leak and irregularity incidents per examination cycle, published. Only then do the SSC and HSC certificates again carry the labour-market signalling value they had in the mid-2010s.

The BNP government's published education agenda is more ambitious than the 180-day priority plan signals. The campaign pledge was to raise education spending to 5 percent of GDP, more than double the current 2.03 percent. The Fifth Primary Education Development Programme (PEDP-5) begins July 1, 2026, with a new assessment system for Classes III to V that combines 30 marks of continuous assessment with 70 marks of summative testing including mandatory oral or practical components. The continuous-assessment design is the right architectural choice for the learning-poverty problem this piece describes. The 5 percent of GDP funding pledge is consistent with the Vietnam comparison; it is also conditional on revenue reform, because at a tax-to-GDP floor near 8 percent the funding mathematics does not clear regardless of how the pledge is worded.

What it would cost, what it would buy

Bangladesh's nominal GDP was Tk 55.5 lakh crore (Tk 55,52,753 crore) in FY2024-25 (BBS provisional). The full reform set, costed at the upper end, requires roughly 0.6 to 0.8 percent of GDP in incremental annual recurrent spending, plus a one-off capital programme of about 1 percent of GDP spread over five years. The two are distinct lines and must be read separately. On the FY2024-25 base, the recurrent line is Tk 33,000 to 44,000 crore per year (0.6 to 0.8 percent of GDP), and the capital line is about Tk 55,000 crore in total (1 percent of GDP) spread across the half-decade, roughly Tk 11,000 crore per year. Only the recurrent line raises the standing education-to-GDP ratio: adding 0.6 to 0.8 percent of GDP takes that ratio from 2.03 percent to between 2.6 and 2.8 percent, closer to the Vietnam comparison and well below the Indian consolidated level. The one-off capital outlay does not enter that ratio because it is not a permanent recurring claim.

This is large but not impossible, and the binding constraint is allocation discipline rather than aggregate fiscal space, conditional on the revenue side reforming. The recurrent increment is under one percentage point of GDP. The question is not whether the macroeconomy can absorb it; it is whether the budget is willing to reallocate toward a payoff that arrives over a generation rather than an electoral cycle.

What the spending buys is a generation. The fifteen-year payback horizon on education investment is long enough that the political incentive structure ordinarily defeats it. The argument for acting now, in the post-2024 reform window, is that this is the moment when the political economy of patronage hiring has been substantially disrupted, the testing-credibility crisis has created public demand for visible quality reform, and the demographic dividend window (closing roughly by 2035 as fertility falls below replacement and the working-age share peaks) leaves the country with at most one full cohort of school-leavers before human-capital productivity has to substitute for labour-force quantity in driving income growth.

The strongest objection to this sequence is that assessment without remediation capacity simply publishes failure. If a national Class 3 and Class 5 benchmark names the worst schools but the system has no trained teachers, no remediation budget, and no recruitment pipeline to send them, the data shames without fixing, and the reform collapses into a league table. That objection is correct, and it is why the assessment, the recruitment reform, and the non-salary remediation budget must move together rather than in sequence: the test identifies the schools, the recruitment reform supplies the teachers, the recurrent increment funds the remediation. Run the assessment alone and the diagnosis here still holds, but the cure does not.

Nazma's reading test in Manikganj is not a small story. It is the country's main economic problem at the granular level. The thirty-eight students who could not write coherent answers will be the median Bangladeshi worker in 2040. Whether they can read a simple Bangla passage by age ten shapes whether they will be operating a CNC machine or a sewing needle by age twenty-five. The country has the systems and the budget headroom to make the right answer happen. The reform window is now.

Sources

• World Bank Bangladesh Learning Poverty Brief, April 2024 (learning poverty 51%, HCI 0.46, per-child primary spend USD 223 PPP): documents1.worldbank.org/curated/en/099090524113159222/pdf/P17920914f1c370071b32417af49f33a784.pdf
• World Bank India Learning Poverty Brief, April 2024 (56%): documents1.worldbank.org/curated/en/099090524113131044/pdf/P17920918dbd990091900117c6f4b92d182.pdf
• World Bank Indonesia Learning Poverty Brief (53%): documents1.worldbank.org/curated/en/099082924151529593/pdf/P179209-8b1c2fc9-3312-46f6-abb4-e1a5d85e5dd2.pdf
• World Bank Thailand Learning Poverty Brief, April 2024 (23%): documents1.worldbank.org/curated/en/099090524093027254/pdf/P17920914982d600a1850d186714631dcb2.pdf
• World Bank Pakistan Learning Poverty Brief, June 2022 (77%): documents1.worldbank.org/curated/en/099812207212211713/pdf/IDU0e0c38ddc0f77b04fda0a7ad0e2f4235d517a.pdf
• World Bank Vietnam Learning Poverty Brief, June 2022 (18%): documents1.worldbank.org/curated/en/099017207152211608/pdf/IDU042cf1562001b20471708cf008277da618cc9.pdf
• World Bank Learning Poverty indicator (database): data.worldbank.org/indicator/SE.LPV.PRIM
• World Bank Human Capital Index 2020: worldbank.org/en/publication/human-capital
• Bangladesh nominal GDP FY2024-25, Tk 55,52,753 crore (BBS provisional, May 2025): bssnews.net/news-flash/277329
• BANBEIS Bangladesh Education Statistics 2023: banbeis.gov.bd
• PEDP-5 launch July 1, 2026: banglamirrornews.com/2026/04/15/fifth-primary-education-development-programme-to-begin-1-july-pm-tells-js/
• PM Tarique Rahman 5% of GDP education pledge: thedailystar.net/news/governance/news/pm-reiterates-pledge-raise-education-spending-5-gdp-4146561
• BNP 12-point education reform agenda: en.bonikbarta.com/bangladesh/GPcxVn9cLO2VOz08
• Class III-V new 30/70 assessment system: bssnews.net/news/349887
• BDPolicyLab data: bdpolicy.db series 66 (literacy rate), 67 (school enrollment), 97 (education expenditure % GDP), 98 (female literacy), 222 (pupil-teacher ratio primary), 223 (education spending % govt), 224 (children out of school), 227 (trained teachers primary); peer series 470-474 (education expenditure), 1093-1097 (pupil-teacher ratio)