Eighty three million workers arrive whether or not the jobs do

Bangladesh 2036, chapter 4 of 10. Contents, sources and the five numbers.

Eighty three million workers arrive whether or not the jobs do

The demographic dividend is usually described as an opportunity. On the current numbers it is a deadline.

The labour force grew by 14.8 million between 2010 and 2024, from 58.80 million to 73.62 million. Then it went backwards: 74.79 million in 2023 to 73.62 million in 2024, a fall of 1.17 million, with employment falling 1.91 million from 70.98 million to 69.07 million in the same year. Participation for those aged 15 and over was 58.6 percent in 2010, 61.9 percent in 2022 and 58.9 percent in 2024.

Women did the retreating. Female participation was 35.7 percent in 2010, 33.5 percent in 2016, 43.7 percent in 2022, 38.7 percent in 2024 and 38.6 percent in 2025. Roughly nine million women joined the labour force between 2010 and 2022; five points of participation is about 3.2 million of them. Regional comparison sets the scale of the loss: Vietnam runs 68.9 percent, Malaysia 51.6 percent, the Philippines 50.5 percent, India 32.4 percent and Sri Lanka 31.0 percent.

The 2022 to 2024 reversal in female participation is the single largest reversible number in this piece, and it is worth more to 2036 output than any infrastructure project in the ADP.

The jobs that exist are the wrong jobs. Informal employment was 84.0 percent of total employment in 2024, 96.0 percent for women and 95.0 percent for workers aged 15 to 24. Self employment, employers and unpaid family work together are 62.1 percent of employment: 35.76 million own account workers, 4.40 million unpaid family workers and 2.73 million employers against 25.99 million employees. The wage and salaried share moved from 37.66 percent in 2010 to 38.58 percent in 2024, nine tenths of a percentage point in fourteen years. Worse, agriculture's employment share rose from 41.3 percent in 2022 to 44.7 percent in 2024, and 75.4 percent of employed women are in agriculture, up from 64.8 percent in 2010. Structural transformation ran backwards for two years.

Youth idleness is where the two statistical systems disagree most sharply. The World Bank series puts youth not in employment, education or training at 15.43 percent in 2024 and 19.72 percent for young women. The 2022 census, averaged unweighted across the 64 districts, gives 35.44 percent overall, 54.42 percent for women and 13.72 percent for men. The district spread runs from 25.44 percent in Dhaka to 42.16 percent in Lakshmipur, and female youth NEET reaches 61.35 percent in Bhola. Both numbers are real. They are measuring different things, and the country does not have an agreed answer to how many of its young people are doing nothing.

A youth NEET rate above forty percent in Lakshmipur, Brahmanbaria and Noakhali sits on top of the highest remittance receipts in the country, which is what an economy looks like when migration substitutes for a domestic labour market.

The arithmetic to 2036 is unforgiving. The UN medium variant puts the population at 196.74 million in 2036, up from 175.69 million in 2025, with median age rising from 25.65 years in 2024 to 30.02 in 2036 and fertility falling from 2.14 to 1.92. Holding the 2024 ratio of labour force to total population, the labour force reaches about 83 million by 2036. At 4 to 5 percent unemployment that requires employment of 79 to 80 million, against 69.07 million in 2024. That is 10 to 11 million net new jobs, roughly one million a year, every year, for eleven years. The old age dependency ratio rose 45 percent between 2010 and 2024, from 6.82 to 9.92 per hundred, and on its 0.22 point annual trend crosses 13 around 2038.

Annual births fall from 3.47 million in 2024 to 2.98 million in 2036 on the same UN path. The workforce of the 2050s is already smaller than the workforce of the 2030s, and the window to employ the large cohorts closes with this decade.

The AI decade arrives on the last big generation, and that generation is losing the skills to use it

Bangladesh's largest young cohorts enter work in the same decade that generative AI reprices the entry level jobs they were meant to take, and the two inputs that decide how they fare, jobs and skills, both moved the wrong way between 2022 and 2024.

Start with the jobs. Employment of people aged 15 to 24 was 10.54 million in 2016, 13.70 million in 2022 and 13.84 million in 2023, then 12.29 million in 2024. That is 1.55 million youth jobs gone in one year, out of a total employment fall of 1.91 million, so four fifths of the country's job losses in 2024 landed on people under 25. Young men lost 0.99 million jobs, from 6.91 million to 5.92 million, and young women 0.56 million, from 6.93 million to 6.37 million. The two sexes responded differently: youth unemployment was 9.63 percent in 2024, 14.13 percent for young men against 5.02 percent for young women, because young men queue and young women exit. The World Bank youth NEET series, which had fallen from 28.88 percent in 2016 to 14.96 percent in 2022, turned up to 15.43 percent in 2024 and 15.60 percent in 2025, with young women at 20.10 percent.

Now the skills. The UNESCO measure behind Sustainable Development Goal 4.4.1 asks whether a person has used a basic arithmetic formula in a spreadsheet, which is the floor of the office work that AI tools are built on top of. Among 15 to 24 year olds the share fell from 34.6 percent in 2021 to 23.1 percent in 2023, young men from 37.7 to 25.2 percent and young women from 31.9 to 21.2 percent. Adults aged 25 to 74 moved the other way, from 9.3 to 14.1 percent. The cohort that will be 27 to 36 years old in 2036, the prime age core of that year's labour force, is measured with a third less of this skill than the cohort two years ahead of it.

Four fifths of the 2024 employment fall landed on people under 25, and the digital skill of that age group fell by a third in two years, which is the worst possible starting position for the cohort that AI meets first.

The export rung the cohort would climb is the rung that is shrinking. Computer and information services earned 740.89 million USD in FY25, the series high in dollars, but as a share of all service exports the line has gone the wrong way for a decade: 17.61 percent in 2016, 8.79 percent in 2025. Pakistan went the other way, from 8.84 percent in 2012 to 46.06 percent in 2025, and India has held near 48 percent throughout. The Pakistan comparison is worth stating carefully. In 2016, the year Bangladesh's digital export share peaked, the two countries had the same income per head, 4,579 international dollars against Pakistan's 4,631. By 2025 Bangladesh had pulled far ahead, 10,154 against 6,573. Bangladesh won the income race and lost the digital export race to the country it started level with. The local data pull carries only the share for Pakistan and not the dollar level, so how much of that rise is digital exports growing and how much is other service exports shrinking cannot be separated here. The work behind the Bangladeshi line is concentrated in data entry and low end outsourcing, which is the segment generative AI compresses first, and the country never built the layer above it while the building was cheap.

Two countries that started the decade at the same income per head ended it with one of them holding a digital export sector and the other holding a shrinking share of a smaller one.

The infrastructure beneath that line is thin. Secure internet servers rose from 228.71 per million people in 2021 to 527.06 in 2024, fixed broadband reached 8.09 subscriptions per 100 people in 2024, and residents filed 74 patent applications in 2021 against 373 by non residents. Because 95 percent of workers aged 15 to 24 are informal, the tier of screen based formal jobs that AI displaces directly is small. The generational effect runs through a different door: the entry rung of the formal economy, which is the rung the export numbers say is shrinking.

The schooling pipeline that would supply the layer above is not filling either. Gross secondary enrolment reached 67.04 percent in 2022 and has fallen in each year since, to 64.30 percent in 2024. Tertiary enrolment is flat at 23.73 percent, having moved less than a point since 2021. The cohort now leaving school is the largest the country will ever send into work, and it is leaving school at a slightly lower rate than the cohort before it, with a measured digital skill a third lower.

The state's answer is a budget line that has spent nothing. The revised development programme for FY26 carries a Department of Youth Development project for skill development of IT literate youth through artificial intelligence technology, with a total cost of 46.36 crore BDT, an FY26 allocation of 13.25 crore and cumulative expenditure of zero. The ICT Division's 18 projects total 2,286.25 crore. The youth AI project is 0.02 percent of the 230,000 crore programme. No source held for this piece publishes a national figure for AI compute capacity, and no exposure estimate for Bangladeshi jobs is used here because none exists on a verifiable basis.

The generational arithmetic makes this a one time decision. The working age share of the population was 65.69 percent in 2025 and still rising, the total dependency ratio 52.24 and still falling, and the birth cohorts behind the current one are smaller every year. Annual births fall from 3.47 million in 2024 to 2.98 million in 2036 on the UN medium variant, so the cohort that is 15 to 24 today is the last large one the country will send into work. Its members turn 27 to 36 in 2036, which is the prime age core of the labour force in the year every scenario in this piece is priced. Whatever that cohort can do with the tools of the AI decade is what the 2036 economy can do, and the state is currently handing it a measured skill decline, a falling secondary enrolment rate, a shrinking digital export sector, and a skills project that has spent nothing.

---

Previous: One product, one market, one date

Next: The human capital bill is unpaid, and health is where the arrears sit

Bangladesh 2036, the series. Contents and the five numbers · 1 · 2 · 3 · 4 · 5 · 6 · 7 · 8 · 9 · 10

Sources for every figure in this chapter are listed on the contents page and in the series source ledger.