The Graduate Without a Job: Bangladesh's Joblessness Sits at the Top of the Schooling Ladder

Executive Summary. Bangladesh does not have an unemployment problem by the standards of its comparison group. It has a graduate unemployment problem. The headline rate was 3.78% of the labour force in 2025, third of six Asian economies and 0.54 points above the peer median of 3.24%. Youth unemployment, at 9.39%, was fourth of six and sat 0.21 points below the peer median of 9.59%. But among Bangladeshis with advanced education the rate was 10.85% in 2024, second of the four countries in the group reporting that series, 4.72 times the median of the three reporting peers (2.3%) and 2.87 times Bangladesh's own national rate. This is not a story about producing too many graduates: tertiary gross enrolment was 23.73%, fourth of the five countries reporting and 12.28 points below their median of 36.01%. The constraint is on the demand side. 57.35% of Bangladeshi employment is vulnerable employment, own-account or unpaid family work, second of six and 5.95 points above the peer median of 51.4%. An economy that multiplied enrolment 5.44 times since 1990 did not build the wage jobs a degree is a claim on.

Two unemployment rates, one country

Read the headline number and there is nothing to write about. Unemployment in Bangladesh was 3.78% of the labour force in 2025, third of six Asian economies. India was at 4.22%, Pakistan 5.42%, Indonesia 3.24%, Thailand 0.78%, Vietnam 1.52%. Bangladesh sits 0.54 points above the peer median of 3.24%, which is 1.17 times it. On this measure the country is unremarkable.

Now condition on education. Among Bangladeshis in the labour force with advanced education, unemployment was 10.85% in 2024. The comparison group thins out here: Pakistan and Indonesia have no observation for that year, so the ranking rests on four countries. India was at 13.47%, Vietnam at 2.297%, Thailand at 1.71%. Bangladesh is second of the four reporting, above the two Southeast Asian comparators that report by a distance no measurement convention explains away: 8.56 points above the peer median of 2.3%, which is 4.72 times the median.

The domestic comparison is starker than the cross-country one. Bangladesh's advanced-educated unemployment rate is 2.87 times the national rate. Education is supposed to run the other way. Where a graduate premium functions, more schooling buys a shorter search and a lower probability of being out of work. In Bangladesh the relationship inverts, and it inverts hard.

Source: World Bank World Development Indicators, unemployment with advanced education (% of labour force with advanced education, SL.UEM.ADVN.ZS), 2024.

Two things to keep straight before this figure is used. The graduate reading is 2024 and the national reading is 2025, so the 2.87 ratio is built from adjacent years, not the same one. And Bangladesh's own advanced-education series is thin: nine observations since 2006, which is enough to establish a level and not enough to draw a trend anyone should defend.

It is not a youth problem

The obvious first explanation is that this is the standard young-country story, a large cohort arriving faster than the economy can absorb it. The age data does not support that reading.

Youth unemployment, ages 15-24, was 9.39% in 2025, fourth of six and 0.21 points below the peer median of 9.59%, at 97.8% of it. India stood at 16.02%, Indonesia at 12.98%, Pakistan at 9.59%, Vietnam at 6.17%, Thailand at 4.53%. On the age cut, Bangladesh is mid-table and slightly better than the median comparator.

Source: World Bank WDI, unemployment, youth total (% of labour force ages 15-24, SL.UEM.1524.ZS), 2025.

The same labour market that places its young near the middle of the group fails one slice of them at more than four times the median of the peers reporting. The two populations overlap, which limits how far this can be pushed: many graduates fall inside the 15-24 band, others age out of it while still searching, and the indicator set used here holds no cross-tabulation of education by age. What the two series jointly support is narrow and enough. Being young in Bangladesh is a median condition for this group. Holding a degree is not.

It is not an oversupply of graduates either

The second obvious explanation is credential inflation, a system that produces more degrees than any economy at this income level could use. The enrolment data does not support that reading either.

Tertiary gross enrolment was 23.73% in 2024, fourth of the five countries reporting (Indonesia has no observation that year). Thailand was at 44.98%, Vietnam 37.59%, India 34.42%, Pakistan 10.86%. Bangladesh sits 12.28 points below the peer median of 36.01%, at 65.9% of it. Only Pakistan, of the five reporting, enrols a smaller share, and Pakistan does not report the graduate unemployment series at all.

What has changed is the speed. Bangladeshi tertiary enrolment was 4.36% gross in 1990 and 23.73% in 2024: a rise of 19.37 points, 5.44 times the 1990 level, or 444.3% in thirty-four years.

Source: World Bank WDI, school enrollment, tertiary (% gross, SE.TER.ENRR), Bangladesh against comparators, 1990-2024.

Put the two facts together and the shape of the problem changes. Bangladesh is not over-educating its population relative to the group. It is educating it fast, from a very low base, into a labour market whose formal wage sector did not expand on the same schedule. The relevant ratio is not graduates to population, where Bangladesh is behind, but graduates to the jobs that require them, which no indicator in this set measures directly.

The cross-country pattern should be stated at exactly its evidentiary weight. Vietnam pairs tertiary enrolment of 37.59% gross with graduate unemployment of 2.297%; Thailand pairs 44.98% with 1.71%; India pairs 34.42% with 13.47%. Among the four reporting both series, the two South Asian economies pair higher graduate unemployment with lower enrolment than the two Southeast Asian ones. That is four data points. It describes a contrast and identifies no cause, and the obvious confounder is industrial composition: Vietnam and Thailand host large formal manufacturing and export-electronics sectors that hire technical graduates at scale. This indicator set carries no sectoral employment breakdown, so that explanation is named, not tested.

What the labour market offers instead

If the wage sector is the missing piece, the composition of employment should show it, and it does.

57.35% of Bangladeshi employment was vulnerable employment in 2025, the statistical category covering own-account workers and contributing family workers. That is second of six, 5.95 points above the peer median of 51.4% and 1.12 times it. Only India, at 71.58%, is higher. Pakistan was at 55.51%, Vietnam 51.4%, Indonesia 49.96%, Thailand 47.98%.

More than half of Bangladeshi work therefore sits outside an employment contract. That is the realistic alternative facing a graduate who stops searching: not a lower-paid salaried job, but a shop, a plot, or unpaid work in a family enterprise. A degree is a claim on a wage relationship, and where most employment is not a wage relationship, the claim has nothing to attach to.

Participation reinforces the point and complicates the measurement. Labour force participation was 58.8% of the population aged 15 and over in 2025, fourth of six, 7.88 points below the peer median of 66.68% and at 88.2% of it. Vietnam reached 72.78%, Indonesia 67.97%, Thailand 66.68%, India 55.66%, Pakistan 52.34%. An unemployment rate is measured against the labour force, so people who abandon the search leave the numerator and the denominator together, and the 3.78% headline rests on a narrower base than Vietnam's or Thailand's. That cuts against reading it as evidence of a healthy market. It cannot be pushed further here: participation in this group turns heavily on female participation and on how many young people are in full-time study, and neither split is in the indicators used.

For scale, this is a country of 173.6 million people with a US$450 billion economy that grew 4.22% in 2024. The rate of growth is not the binding constraint. Its composition appears to be.

The part of the number that is not a failure

One mechanism deserves to be stated in its own favour, because it is real and it is routinely ignored.

Unemployment requires the ability to be unemployed. Where the fallback is own-account work at 57.35% of employment, a graduate from a household that can absorb the cost may rationally spend a year or two queuing for a formal or public post rather than accepting work that wastes the credential. On that reading, part of the 10.85% is a queue for scarce good jobs rather than idle capacity, and the rate rises with family means as much as with job scarcity. It predicts a rate concentrated among first-time entrants from households above subsistence, falling sharply with search duration.

The indicators used here cannot test any of it: no duration of unemployment, no public versus private split, no household income, no field of study. The queue hypothesis is plausible and unfalsified, which is a reason for humility about the 10.85% rather than grounds to dismiss it. Taken at its strongest it still describes a market where the expected wait for a job matching a degree is long enough to justify years of measured joblessness, a demand shortfall stated politely.

What would change this conclusion

A specific, checkable test. If, by the 2030 WDI vintage, unemployment among the advanced-educated has fallen below 5% while tertiary enrolment keeps rising toward the current peer median of 36.01%, the argument here is wrong: the wage sector will have caught up with the education system without the composition of employment changing much. If instead the graduate rate is still near 10.85% and still close to 2.87 times the national rate, a decade of enrolment expansion will have been poured into the same wall.

Three moves, with owners and signals.

1. Publish a standing graduate labour-market series. The 10.85% cannot be acted on because nothing decomposes it: not by field of study, not by institution type, not by search duration. Owner: Bangladesh Bureau of Statistics, alongside the Labour Force Survey, with the University Grants Commission supplying the institution frame. Success signal: an annual published table splitting graduate unemployment by field and by months of search, which makes the difference between a queue and a surplus an observable rather than an argument.
2. Stop treating enrolment as the output of the tertiary system. Enrolment rose 5.44 times since 1990, and no comparable series tracks what happened to those graduates afterwards. Owner: Ministry of Education, with the University Grants Commission. Success signal: the ratio of advanced-educated unemployment to national unemployment falling below 2 across two consecutive WDI vintages, from 2.87 today.
3. Treat formal wage-job creation as the binding constraint it appears to be. A graduate policy that does not move the composition of employment cannot work, because the jobs a credential is written for sit outside the 57.35% of employment that is vulnerable. Owner: Ministry of Labour and Employment, with the investment promotion agencies on the demand side. Success signal: vulnerable employment falling below the current peer median of 51.4% in the WDI series.

The counterargument

The strongest objection is a measurement objection and it is serious. The advanced-education series is reported by four of the six countries in 2024, so the peer median rests on three observations, one of which is itself the median. "Advanced education" is a harmonised international category whose mapping onto national qualifications differs, and unemployment is defined against a job-search reference period that national surveys implement differently. On top of that, India's rate of 13.47% is above Bangladesh's, so the pattern is regional rather than Bangladeshi.

Three answers, the last of which is a concession.

First, the thinness of the peer group cuts both ways. Three comparators is a weak base for a median, but the two that report are the group's large formal-manufacturing economies, and they come in at 2.297% and 1.71%. The Bangladeshi figure is not marginally outside that range. A definitional gap capable of producing a fourfold difference would be large enough to invalidate the series for every user, not only this one.

Second, India having more of the problem does not make the Bangladeshi level tolerable. It makes it a shared regional failure with a shared cost, and it removes the comfortable reading that this is a transitional artefact of one country's expansion. Two large South Asian economies with different education systems and different industrial policies arriving at the same inversion is the observation, not the excuse.

Third, the concession. Nothing here identifies a cause. Six countries observed in a single year for most indicators is a comparison, not a sample, and every association is consistent with stories the data cannot separate: composition of output, quality of tertiary provision, the ability of households to finance a search, and the survey conventions of six different statistical offices. What the data does establish is a level and a shape. Bangladeshi unemployment is ordinary at the headline, ordinary among the young, and extraordinary among the educated, and the distance between those three readings is what needs explaining.

---

Data sources: World Bank World Development Indicators, retrieved from the BDPolicyLab data lake, 2026-08-10. Comparators are India, Pakistan, Indonesia, Thailand and Vietnam; "peer median" is the median of the comparators reporting in the stated year, excluding Bangladesh. Years differ by indicator because coverage does: unemployment among the advanced-educated and tertiary enrolment are 2024, the national unemployment rate, youth unemployment, labour force participation and vulnerable employment are 2025. Unemployment among the advanced-educated reports for four of the six countries in 2024 and tertiary enrolment for five; every comparison drawn from those two series is stated on that basis. The labour series reach WDI from ILOSTAT, harmonised across national labour force surveys of differing design, and the harmonisation is ILO's rather than this essay's.

Sources

1. World Bank WDI, Unemployment with advanced education (% of total labor force with advanced education), SL.UEM.ADVN.ZS: https://data.worldbank.org/indicator/SL.UEM.ADVN.ZS
2. World Bank WDI, Unemployment, total (% of total labor force), SL.UEM.TOTL.ZS: https://data.worldbank.org/indicator/SL.UEM.TOTL.ZS
3. World Bank WDI, Unemployment, youth total (% of labor force ages 15-24), SL.UEM.1524.ZS: https://data.worldbank.org/indicator/SL.UEM.1524.ZS
4. World Bank WDI, School enrollment, tertiary (% gross), SE.TER.ENRR: https://data.worldbank.org/indicator/SE.TER.ENRR
5. World Bank WDI, Labor force participation rate, total (% of total population ages 15+), SL.TLF.CACT.ZS: https://data.worldbank.org/indicator/SL.TLF.CACT.ZS
6. World Bank WDI, Vulnerable employment, total (% of total employment), SL.EMP.VULN.ZS: https://data.worldbank.org/indicator/SL.EMP.VULN.ZS