Bangladesh halved its poverty rate in two decades. In 2000, nearly half the country, 48.9%, lived below the upper poverty line. By 2022, that number had fallen to 18.7%. By any measure, this is one of the most remarkable sustained poverty reductions in the developing world, achieved without oil wealth, without large-scale foreign aid dependency, and largely through the dynamism of its garment sector, remittance economy, and agricultural modernization.
But the national average conceals a fracture. When you disaggregate the data by division, by district, by the rural-urban divide, the story of convergence dissolves into one of divergence. The satellites that track nightlights, built-up area, and vegetation health see it clearly: Bangladesh's growth has been spatially concentrated, and the places it has not reached are falling further behind.
This is Part 5, the final installment of the Satellite Bangladesh series. It examines who has been left behind, where they live, and what the data says about the structural barriers that keep them there.
The national story
The headline numbers are genuinely impressive. Bangladesh's poverty headcount ratio under the upper poverty line dropped from 48.9% in 2000 to 40.0% in 2005, 31.5% in 2010, 24.3% in 2016, and 18.7% in 2022. That is a decline of more than 30 percentage points in 22 years.
The extreme poverty rate, measured under the lower poverty line, fell even faster: from 34.3% in 2000 to just 5.6% in 2022. Extreme deprivation, the kind that kills, has been reduced to a fraction of its turn-of-the-century level.
But the rural-urban decomposition reveals the first crack in the narrative. In 2000, rural poverty stood at 52.3% and urban poverty at 35.2%, a gap of 17.1 percentage points. By 2022, rural poverty had fallen to 20.5% and urban poverty to 14.7%, narrowing the gap to 5.8 points. Progress on both fronts, yes. But rural Bangladesh still harbors one in five people below the poverty line, and the pace of rural poverty reduction slowed between 2016 and 2022 compared to the preceding decade. The easy gains have been made. The remaining poverty is harder, deeper, and more spatially concentrated.
The divisions that diverge
The Multidimensional Poverty Index, which captures deprivation across health, education, and living standards simultaneously, exposes the geographic fault lines that the national headcount obscures.
The OPHI/GED Bangladesh National MPI 2025, built on MICS 2019 data, places the national multidimensional poverty headcount at 24.1%. Sylhet division leads at 37.7%, followed by Mymensingh at 35.0% and Barishal at 31.6%, while Khulna sits lowest at 15.2%. The Sylhet to Khulna gap is more than two and a half to one.
The HIES income poverty data tells a similar story. In 2022, Barishal had the highest income poverty headcount at 26.9%, followed by Rangpur at 24.8% and Mymensingh at 24.2%. At the other end, Dhaka and Khulna sit lowest, both around 6 to 8%.
The satellite-derived poverty proxy index, constructed from VIIRS nightlights, GHSL built-up area, and MODIS vegetation health, adds spatial resolution to these survey-based estimates.
At the district level, the proxy reveals pockets of deprivation scattered across multiple divisions. Kurigram registers the highest satellite poverty index at 0.534, followed by Bhola at 0.525, Manikganj at 0.524, Satkhira at 0.522, and Sunamganj at 0.522. These are districts with low nightlight intensity, sparse built-up infrastructure, and vegetation patterns consistent with subsistence agriculture rather than commercial farming. At the other end, Gazipur (0.494) has the lowest proxy poverty score, driven by industrial activity and high nightlight output, followed by Feni (0.498), Chattogram (0.498), Khagrachhari (0.498), and Joypurhat (0.499).
The range is narrow, from 0.494 to 0.534, reflecting the composite nature of the index. But within that range, the spatial pattern is notable: the highest-scoring districts span the northwest (Kurigram in Rangpur), the coast (Bhola in Barishal, Satkhira in Khulna), the northeast (Sunamganj in Sylhet), and the center (Manikganj in Dhaka). The lowest-scoring districts cluster around industrial and urban centers (Gazipur, Chattogram) and the Chattogram Hill Tracts (Khagrachhari). This is not random variation. It is a map of structural inequality.
The children who suffer most
Poverty is an abstraction until you measure its consequences in the bodies of children. The 2022 Demographic and Health Survey provides that measurement with uncomfortable precision.
Sylhet again stands apart. Per the BDHS 2022 Key Indicators Report, 31.2% of children under five in Sylhet are stunted, meaning they are too short for their age, an irreversible marker of chronic malnutrition that constrains cognitive development and lifetime earning potential. That is well above the national stunting rate and roughly twelve percentage points above Khulna's rate of about 19%. Sylhet has come down from 49.6% in 2014, a real reduction, but from a catastrophic baseline.
The 12 percentage-point gap in stunting between Sylhet and Khulna is not a health statistic. It is a prediction about the next generation's capacity to escape poverty. A child stunted today in Sylhet will, on average, earn less, learn less, and produce less over their lifetime than a child in Khulna or Dhaka.
What the satellites cannot see
The satellites that produce nightlight composites, built-up area estimates, and vegetation indices are measuring proxies, not poverty itself. They cannot detect whether a child was breastfed exclusively for the first six months. They cannot determine whether a school in Kurigram has a functioning toilet. They cannot assess whether a household in Sunamganj has access to clean drinking water or whether the nearest health facility is staffed.
What the satellite data can do, and what this series has attempted to demonstrate, is identify spatial patterns of deprivation at a resolution and frequency that household surveys cannot match. The DHS happens every five to seven years. HIES happens every four to six years. Satellites observe every day. The proxy is imperfect, but the coverage is total and the cadence is continuous.
The real gap is not in the data. It is in the institutional response. Bangladesh has some of the best poverty and health survey infrastructure in the developing world. It produces data that many richer countries cannot match. The question is whether the spatial patterns that emerge from that data, the same patterns visible from orbit, are being translated into spatially targeted policy interventions.
What policy must do
The satellite and survey evidence converges on three priorities.
First, concentrate social protection spending in the northwest corridor and Sylhet. The poverty proxy map, the MPI data, and the stunting figures all point to the same geographies. Kurigram, Bhola, Manikganj, Satkhira, and Sunamganj, the five highest-scoring districts on the satellite poverty proxy, span multiple divisions, and the broader Rangpur and Sylhet divisions are where the remaining poverty is most entrenched. Bangladesh's social safety net programs, including the Vulnerable Group Development and Vulnerable Group Feeding schemes, should be weighted more heavily toward these areas. The current allocation formulas, which distribute resources partially on the basis of population rather than need, dilute spending in exactly the places where concentration would have the highest marginal returns.
Second, launch a division-level nutrition emergency response in Sylhet. A stunting rate of 31.2% is not a development challenge. It is a crisis. Sylhet's 2022 stunting rate is roughly ten percentage points above the best-performing divisions, and the per-child cost in lost cognitive development and lifetime earnings is enormous. The response should be modeled on the community-based management of acute malnutrition (CMAM) approach that has succeeded in similar settings: direct supplementary feeding, micronutrient supplementation, behavior change communication on infant and young child feeding practices, and integration with the existing community clinic network.
Third, use the satellite poverty proxy as a continuous monitoring layer between household surveys. The five-to-seven-year gap between DHS rounds and the four-to-six-year gap between HIES rounds leaves policymakers flying blind between survey years. A satellite-derived poverty index, updated annually, cannot replace survey-based measurement but can flag districts where conditions are deteriorating before the next survey confirms it. The technical infrastructure exists. The policy infrastructure to act on it does not.
Five chapters. Forty years of satellite data. The evidence is unambiguous: Bangladesh has made extraordinary progress, visible in retreating flood footprints turned to farmland, in nightlights spreading across a once-dark landscape, in forests stabilizing after decades of loss. But the same evidence shows that this progress has been unevenly distributed, that coastal exposure is structural and worsening, and that millions of Bangladeshis, disproportionately in the north and northeast, live in conditions that the national averages do not reflect. The satellites have done their part. They have watched, measured, and recorded. What happens next is not a question of data. It is a question of political will.
Data Sources
- Poverty headcount (HIES): Bangladesh Household Income and Expenditure Survey, BBS. Years: 2000, 2005, 2010, 2016, 2022. Output:
bd_gis/outputs/poverty/hies_national_hcr_timeseries.csv. - Multidimensional Poverty Index (MPI): OPHI/GED Bangladesh National MPI 2025 report, built on MICS 2019 data. Division-level headcount ratios.
- Satellite poverty proxy: Composite of VIIRS nightlights (2020), GHSL built-up area (2020), MODIS NDVI (2020). 64 districts. Output:
bd_gis/outputs/poverty/poverty_district_ranking.csv. - Child nutrition (BDHS 2022): Bangladesh Demographic and Health Survey 2022 (ninth round), Key Indicators Report. Division-level stunting, wasting, underweight.
- HIES division-level poverty: BBS HIES 2022, division-level headcount ratios; Barishal replaced Rangpur as the highest-poverty division.