A Roof Is Not a House

Executive Summary. Bangladesh's Population and Housing Census material tables describe 40,257,678 households in 2022. The headline looks reassuring: 74.762% had metal-sheet roofs and 23.618% had cement, concrete, or tile roofs. But a roof is only one surface. The same census shows 42.0% of households with metal-sheet walls, 6.84% with mud walls, and 47.992% with soil, sand, or mud floors. Those combinations can provide possession and privacy without providing thermal comfort, flood resistance, safe sanitation, ventilation, tenure security, or affordable access to work. Housing policy that counts units or roof materials will therefore overstate adequacy. Bangladesh needs a housing-quality account that treats the dwelling as a system, links census materials to services and hazards, and directs upgrading toward the weakest component rather than replacing entire settlements.

The roof creates a comforting statistic

The national roof table is dominated by corrugated metal. 74.762% of households had metal-sheet roofs, while 23.618% had cement, concrete, or tile roofs. Only 1.211% used light natural or improvised materials in the combined census category. The residual categories together accounted for 0.409%.

Source: Bangladesh Bureau of Statistics, Population and Housing Census 2022 district roof-material table; national shares re-derived by summing district counts.

This is a story of reach. Metal sheet is modular, tradable, repairable, and available across the country. It converts cash into enclosure without the capital, engineering, or land certainty required by concrete construction. For a rural family expanding one room at a time, those properties matter. A housing policy that dismisses the material as merely temporary would misunderstand why households choose it.

But the statistic cannot bear the weight usually placed on it. A roof material does not say whether the structure is anchored against wind, elevated above floodwater, insulated from solar gain, sealed against rain, or ventilated at night. It does not reveal crowding, ownership, rent burden, water, sanitation, drainage, electricity reliability, or travel time to work. A metal roof can cover a secure and comfortable home. It can also turn a single room into a heat trap.

The first policy correction is conceptual. Housing adequacy is not the presence of a durable-looking top surface. It is the performance of a connected system under ordinary life and predictable shocks.

That system also changes through the day. A roof that sheds monsoon rain may store solar heat and release it into the room after sunset. A wall that provides privacy may leave little cross-ventilation. A raised and washable floor may protect health even when the surrounding lane floods. These are interactions, which is why a single-material score can misdirect subsidy. Field assessment should record the failure households actually experience, such as dangerous indoor heat, repeated damp, wind damage, unsafe wiring, or inaccessible sanitation, and then match the improvement to that failure.

The walls reveal a more divided stock

The wall table breaks the reassuring roof story. 46.66% of households had cement, concrete, brick, or terracotta walls. Another 42.0% had metal-sheet walls. Mud walls accounted for 6.84%, and wood, bamboo, mat, palm, or trunk walls for 3.458%. Other wall categories together made up 1.042%.

Source: Bangladesh Bureau of Statistics, Population and Housing Census 2022 district wall-material table; national shares re-derived by summing district counts.

These categories do not form a simple ladder from bad to good. Thick earthen walls can moderate temperature better than thin metal. Bamboo can be repairable and appropriate where supply chains and skills support it. Masonry can fail in earthquakes when it is unreinforced. Metal can survive rain while creating dangerous indoor heat. Material names are therefore signals, not verdicts.

The useful policy question is how each dwelling performs in its hazard context. In cyclone-prone areas, connections among roof, wall, and foundation matter. In flood zones, elevation, sacrificial materials, drainage, and safe access matter. In dense cities, fire separation, ventilation, heat, sanitation, and legal access for emergency services matter. A national category can guide investigation, but it cannot substitute for a performance assessment.

This distinction also changes the treatment of informal settlements. Demolition responds to legal status and land value, not necessarily to the household's most urgent housing defect. If the structural frame is serviceable but drainage and sanitation are not, infrastructure upgrading can deliver more welfare with less displacement. If the site is physically unsafe, relocation must preserve livelihoods and social networks rather than count only replacement units.

The floor reverses the impression

The floor table offers the sharpest correction. 47.992% of households had soil, sand, or mud floors. 44.75% had cement, concrete, brick, or terracotta floors. Tile or mosaic accounted for 5.311%, wood or bamboo for 1.827%, and other materials for 0.12%.

Source: Bangladesh Bureau of Statistics, Population and Housing Census 2022 district floor-material table; national shares re-derived by summing district counts.

A household can therefore appear durable from above and remain exposed underfoot. Floors shape damp, cleaning, contamination, insects, mobility, and recovery after inundation. They also reveal the incremental logic of household investment. A family may buy sheets for the roof first because rain protection is immediate, then walls, then a finished floor when resources allow. The sequence is rational even when the final dwelling remains inadequate.

That logic argues for targeted finance. A generic mortgage product is irrelevant to a household that needs a raised plinth, a stronger roof connection, a washable floor, a toilet upgrade, or a safer electrical connection. Small, verified improvement loans and grants can follow the actual sequence of construction. Technical assistance should be delivered with the finance, because a stronger component can transfer stress to a weaker one.

The census tables are especially valuable because they cover the national stock rather than only formal transactions. But they also show the limit of a material inventory. A floor category does not measure dampness or elevation. A wall category does not measure reinforcement. A roof category does not measure fastening. Bangladesh needs to keep the census categories while adding a small set of performance questions that can be repeated.

Repeatability matters more than an encyclopedic survey. A short housing module can ask whether rain enters, whether floodwater enters, whether the main room can be cooled safely, whether the toilet remains usable, and whether residents expect to remain. Those answers can be tested against a sample of physical inspections. The state would then know not only what dwellings are made from, but which observable combinations predict failure. That evidence can guide a cheaper and more precise upgrading programme than a national catalogue of preferred materials.

The strongest counterargument is that materials are a crude proxy

The strongest objection is that material categories are too crude for a national housing argument. The census does not identify workmanship, maintenance, household income, local climate, tenure, plot safety, or service quality. Adding district counts can produce a clean national percentage while hiding large geographic variation. It would be wrong to label every mud floor deprived or every concrete house adequate.

That objection is decisive against a ranking of households by material alone. It is not an argument for ignoring the tables. Materials are observable entry points into a broader assessment. They show where a roof-only narrative fails, where large groups may benefit from incremental upgrading, and which combinations deserve local inspection.

The right response is to join, not discard. Census material fields can be linked at aggregate level to drinking water, toilet use, electricity source, household size, location, and hazard layers. Privacy must be protected, and the joins should remain at a geography large enough to prevent identification. The result would not declare individual homes safe. It would locate concentrations of probable need and make field verification more efficient.

There is another caution. A national housing programme can damage households by turning a measurement tool into an eviction tool. Residents will avoid honest enumeration if a weak-material label threatens removal. The data contract should explicitly separate service eligibility from tenure enforcement. Trust is part of measurement quality.

What would change this conclusion

A clear test would change the argument: if a joined national dataset shows that roof material predicts thermal comfort, flood resilience, service access, and tenure security with little added value from walls, floors, location, or services, then roof-led classification may be sufficient. Until that evidence exists, the three census surfaces should be treated as complementary, not hierarchical.

First, publish a housing-quality matrix from existing census tables. Cross-tabulate roof, wall, floor, water, toilet, electricity, household size, and urban-rural location at safe aggregate geography. Owner: Bangladesh Bureau of Statistics. Success signal: a documented public table in which national totals reconcile exactly across material modules and suppression rules protect small cells.

Second, create an incremental upgrading standard. Define tested packages for roof fastening, insulation, ventilation, raised floors, drainage, toilets, electrical safety, and fire access, adapted to major hazard zones. Owner: housing and public-works authorities with local governments and engineering institutions. Success signal: every publicly financed upgrade records the original defect, the chosen package, inspection result, and household retention.

Third, finance the weakest component without forcing displacement. Offer small grants or loans tied to verified upgrades, with relocation reserved for sites where risk cannot be reduced in place. Owner: housing-finance institutions and local governments. Success signal: follow-up surveys show improved indoor comfort, dry floors, safe services, and lower disaster damage without a rise in involuntary moves.

Bangladesh has already counted the surfaces. The next step is to measure whether they work together. A roof keeps rain off a room. A house keeps a household safe, healthy, connected, and able to remain.

Sources

1. Bangladesh Bureau of Statistics. Population and Housing Census archive. Retrieved 2026-08-23: https://bbs.gov.bd/pages/static-pages/6922e073933eb65569e27220
2. Humanitarian Data Exchange. Bangladesh Population and Housing Census district tables, source BBS. Retrieved 2026-08-23: https://data.humdata.org/dataset/bangladesh-population-and-housing-census