The Industry That Did Not Arrive: Output Without Jobs

Executive Summary. Bangladesh produces like a mid-tier industrialiser and hires like a country that never started. Manufacturing value added was 22.44% of GDP in 2025, third of six Asian economies and above the peer median of 19.07%, up from 13.24% in 1990, a rise of 69.51% on that base. Industrial employment was 18.03% of all jobs, last of the six, against a peer median of 25.26%. Agriculture still holds 44.26% of Bangladeshi workers, the highest share of the six and 1.55 times the peer median of 28.56%. The country puts 2.45 people on a farm for every one it puts in industry. The farm share has fallen a long way, from 68.02% in 1995, a drop of 23.76 points, but that labour did not walk into factories: industry's entire share today is 18.03%. It went into services, at 37.71% still second lowest of the six, and above all into work without an employer. Self-employment is 61.04% of employment and vulnerable employment 57.35%, both second highest of the six behind India.

The two rankings that do not fit

Bangladesh holds the top and the bottom position in the same table.

Of the six Asian economies compared here, Bangladesh has the highest share of employment in agriculture, 44.26% in 2025. India follows at 41.63%, then Pakistan at 36.23%, Thailand at 28.56%, Indonesia at 27.30% and Vietnam at 25.04%. The peer median is 28.56%, so the Bangladeshi farm share runs 1.55 times the median, 15.7 points above it.

On industry the same six invert. Bangladesh employs 18.03% of its workforce in industry, the lowest of the six. Vietnam is highest at 34.83%, then India at 25.82%, Pakistan at 25.26%, Indonesia at 22.53% and Thailand at 22.24%. The peer median is 25.26%, so Bangladesh sits 7.22 points under it, at 0.71 of the median level.

Source: World Bank World Development Indicators, employment in agriculture (SL.AGR.EMPL.ZS), Bangladesh and comparators, 2025.

Two rankings, one country. Bangladesh employs 2.45 farm workers for every industrial worker. Because it holds the highest agricultural share and the lowest industrial share of the six, no other country in the group can have a wider distance between the two. Vietnam is the only one of the six that has crossed over: 34.83% of its workers are in industry against 25.04% in agriculture.

The transition that half happened

The farm share did move, and it moved hard. Bangladeshi agricultural employment fell from 68.02% in 1995 to 44.26% in 2025, a decline of 23.76 points, or 34.93% of the 1995 level. More than two of every three workers were on the land; now fewer than one in two are.

Those 23.76 points went somewhere, and arithmetic constrains where. Industry's entire share today is 18.03%, so even on the impossible assumption that Bangladesh had no industrial workers at all in 1995, industry could have taken at most 18.03 of the 23.76 points released by agriculture. It did not start from zero, so it took less.

The rest went to services, which now employ 37.71% of the workforce. That is not a triumph either: the services share is the second lowest of the six, ahead only of India's 32.56% and 2.42 points below the peer median of 40.13%. Indonesia is at 50.17% and Thailand at 49.21%. Bangladesh has run a structural transformation in which labour left the farm and landed in the two sectors where its shares rank last and second last of the six.

Source: World Bank WDI, employment in industry (SL.IND.EMPL.ZS), Bangladesh and comparators, 2025.

What replaced the farm job

Sector labels say where work happens. They say nothing about what kind of contract it runs on, and on that second question the picture gets sharper.

Self-employment accounts for 61.04% of Bangladeshi employment, second highest of the six behind India's 74.90%, and 7.69 points above the peer median of 53.35%. Vulnerable employment, which counts own-account workers and contributing family workers together, is 57.35%, again second highest behind India's 71.58%, and 5.95 points above the peer median of 51.40%.

Thailand is the useful contrast. It still has 28.56% of its workers in agriculture, a larger farm share than Indonesia's 27.30%, yet it records the lowest vulnerable employment of the six at 47.98% and the lowest self-employment at 50.70%. A large farm sector does not by itself determine the contract structure of the rest of the economy.

The mechanism is not subtle. An economy that moves people out of unpaid family farm work and into own-account retail, transport and repair has changed the sector code on the job without changing its terms. No employer, no payroll, no contributory coverage, no severance, and income that stops on the day the worker stops. Wage employment in a factory is not automatically good work, but it is the form in which labour regulation, contributory pensions and enforceable hours can exist at all. At 61.04% self-employment, most Bangladeshi workers are outside the employment relationship those instruments are written around.

Output climbed, hiring did not follow

The output ledger reads like a success story.

Manufacturing value added was 22.44% of GDP in 2025, third of the six behind Vietnam's 24.53% and Thailand's 23.74%, and 3.38 points above the peer median of 19.07%. India, at 13.47%, and Pakistan, at 13.02%, are far below. Over the long run Bangladeshi manufacturing rose from 13.24% of GDP in 1990 to 22.44%, a gain of 9.20 points, 1.70 times the 1990 share.

Industry as a whole, which adds construction, mining and utilities to manufacturing, was 34.03% of GDP, also third of the six, behind Indonesia's 38.71% and Vietnam's 37.65%, and 2.95 points above the peer median of 31.07%.

Source: World Bank WDI, manufacturing value added (NV.IND.MANF.ZS), Bangladesh against comparators, 1990-2025.

Now set the two ledgers side by side. On output Bangladesh ranks third of six in manufacturing and third of six in industry overall. On employment it ranks last of six in industry. A country can hold those positions simultaneously in only one way: its industrial value added is generated by a comparatively small slice of its labour force, and the rest of that labour force is somewhere with lower output per worker.

Nothing in this data says why. High value added per industrial worker can come from genuine capital deepening, which is what a country wants. It can also come from a narrow export band whose scale is capped by something other than technology, from national accounts that record informal manufacturing work under other sectors, or from output prices rather than volumes. These seven indicators cannot separate those explanations, and this essay does not pretend to.

What this comparison cannot settle

Four limits on what any of this proves.

The employment shares are modelled ILO estimates carried in the WDI, not headcounts from a fresh national survey. They are fitted to a common international definition, which is exactly what makes six countries comparable on one axis and what makes any single figure softer than a survey number. Bangladesh's own labour force survey runs on a different frame, and the two should not be spliced.

Counting rules move the farm share. How much unpaid family labour a survey captures, particularly women's work on household plots, changes the agricultural number directly, and because these are shares of a fixed total, it changes every other sector's number too. Countries differ in that practice, and the differences do not cancel.

Emigration is invisible here. Workers who leave the country appear in none of these shares. Two economies with identical domestic industries will report different employment structures if one sends a larger share of its working-age population abroad, and this data cannot net that out.

And six countries in one year is a comparison, not a causal design. Nothing here shows that a small industrial workforce caused high self-employment, or that a high farm share caused either. The six differ in arable land per person, energy prices, trade preferences, education stock, exchange rate history and the age of their industrial base, and none of that is in this pack. The pattern is a fact about the six; the explanation is not.

The stake, though, is not in dispute. This is an economy of 173.6 million people producing US$450 billion of output in 2024 and growing at 4.2% that year. Employment shares at that scale are not statistics. They are the structure of most people's working lives.

What would change this conclusion

The claim is that Bangladesh's industrial expansion has been an output event rather than an employment event, and that labour leaving agriculture is landing in own-account work rather than in wage jobs. Two readings would falsify it. If the industrial employment share rises above the current peer median of 25.26% while the agricultural share falls below Pakistan's current 36.23%, then industry is absorbing labour at scale and the argument here is wrong. If instead industry sits near 18% while vulnerable employment stays above 57%, the pattern has hardened rather than passed.

1. Measure the industrial job, not just the industrial firm. The shares that anchor this essay are modelled, and a target aimed at the industrial share needs a measured one. Owner: Bangladesh Bureau of Statistics, on the labour force survey cycle. Success signal: a published quarterly employment series cross-cut by sector and status in employment (wage employee, own-account, contributing family worker) with a lag under two quarters, so 18.03% can be checked against a headcount rather than a model.
2. Condition industrial support on the employment it creates, not the output it reports. Bonded warehousing, cash assistance, tax holidays and priority energy allocation are currently defended by export value. Owner: Ministry of Industries with the National Board of Revenue. Success signal: an industrial employment share above Thailand's current 22.24% in the WDI series, which on today's peer values would take Bangladesh off the bottom of the six.
3. Treat own-account work as the destination it actually is. With self-employment at 61.04% and vulnerable employment at 57.35%, the worker leaving a farm is joining a category that industrial policy does not address and labour law barely reaches. Owner: Ministry of Labour and Employment. Success signal: vulnerable employment falling below the current peer median of 51.40%, which would put Bangladesh under the middle of this group on job security instead of 5.95 points above it.

The counterargument

The strongest objection concedes the numbers and reverses the sign. Bangladesh industrialised in a narrow, capital-light band and that band did what it was asked to do. Manufacturing reached 22.44% of GDP, third of six and above the peer median, from 13.24% in 1990. Output per industrial worker is high because Bangladeshi industry is efficient. Asking for a larger industrial employment share is asking for a worse industry.

Two replies.

First, "high productivity" describes the gap rather than explaining it. A small industrial workforce producing 22.44% of GDP is what you see when industry is small and good. It is also what you see when industry is small and enclosed, hiring at the rate its own product mix requires and no faster, while 44.26% of the workforce stays on the land because there is nowhere else with an employer. The output data alone cannot tell those apart, which is precisely why the employment ranking is the more informative of the two.

Second, the objection optimises the wrong variable. Households do not consume the manufacturing share of GDP; they consume the job structure, and this job structure ranks highest of six in farm employment, last of six in industrial employment, and second highest of six in both self-employment and vulnerable employment. An industrial sector that produces well without hiring broadly is a good sector sitting inside an incomplete transformation. The second is the harder problem, and it is the one the output figures are least equipped to reveal.

---

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 five comparators, excluding Bangladesh, and every comparison rests on all six countries reporting. Employment and value added shares are 2025, the latest year in the series; 1995 (agricultural employment) and 1990 (manufacturing value added) are the baselines carried in the fact pack. Population, GDP and growth are 2024. Employment series are modelled ILO estimates.

Sources

1. World Bank WDI, Employment in agriculture (% of total employment), SL.AGR.EMPL.ZS: https://data.worldbank.org/indicator/SL.AGR.EMPL.ZS
2. World Bank WDI, Employment in industry (% of total employment), SL.IND.EMPL.ZS: https://data.worldbank.org/indicator/SL.IND.EMPL.ZS
3. World Bank WDI, Employment in services (% of total employment), SL.SRV.EMPL.ZS: https://data.worldbank.org/indicator/SL.SRV.EMPL.ZS
4. World Bank WDI, Manufacturing, value added (% of GDP), NV.IND.MANF.ZS: https://data.worldbank.org/indicator/NV.IND.MANF.ZS
5. World Bank WDI, Industry (including construction), value added (% of GDP), NV.IND.TOTL.ZS: https://data.worldbank.org/indicator/NV.IND.TOTL.ZS
6. World Bank WDI, Vulnerable employment, total (% of total employment), SL.EMP.VULN.ZS: https://data.worldbank.org/indicator/SL.EMP.VULN.ZS
7. World Bank WDI, Self-employed, total (% of total employment), SL.EMP.SELF.ZS: https://data.worldbank.org/indicator/SL.EMP.SELF.ZS