The Shelter Arithmetic

Executive Summary. Bangladesh's Multipurpose Disaster Shelter Project shows why resilience should be counted as a network rather than a building total. At the last reported date in 2024, the project had completed 408 new shelters against a target of 530 and rehabilitated 450 existing shelters against a target of 450. It reported 1,033,440 beneficiaries with shelter access, below the 1,207,900 closing target, and raised the share of priority shelter needs met from 60.0% in 2015 to 77.51%, short of 80.0%. The same investment delivered 2,574 classrooms and 142.16 kilometres of rural roads, because a shelter must function before, during, and after a disaster. The arithmetic should therefore include catchment population, safe access, school use, maintenance, disability access, water, sanitation, and actual evacuation. Bangladesh should manage each shelter as a service node with a verified readiness record, not as a completed structure on a project list.

Construction met one target and missed another

The project rehabilitated 450 shelters, exactly matching its 450 target. New construction reached 408 shelters against a target of 530 at the last reported date.

Source: World Bank Multipurpose Disaster Shelter Project results report, actuals at 30 November 2024 and closing targets.

The difference matters, but it should not erase what was delivered. Rehabilitation can restore capacity faster than new construction where buildings and sites already exist. New shelters require land, design, procurement, contractor performance, access works, and local acceptance. Each delay can leave a specific catchment without safe capacity.

A project total cannot reveal which people remain outside reach. Missing shelters are not interchangeable across the coast. A structure in one union does not compensate for a gap where evacuation distance, river crossings, road condition, or population concentration make access difficult. The planning unit should be the shelter catchment, not the national count.

This changes completion reporting. A shelter should be considered operational only when the building, approach route, water, sanitation, power, communications, management committee, and accessibility arrangements work together. A construction certificate is necessary. It is not sufficient evidence that people can evacuate safely.

Capacity itself needs a transparent definition. Floor area can produce a theoretical occupancy figure, but safe use depends on circulation, toilets, water storage, ventilation, privacy, and the space taken by supplies. Households may also arrive with essential belongings. A catchment register should distinguish nominal capacity from usable emergency capacity and record the assumption behind each. Otherwise a map can show every settlement covered while the nearest building cannot safely hold the population assigned to it.

Access is the outcome that counts

The project reported 1,033,440 beneficiaries with access to multipurpose shelters, compared with a closing target of 1,207,900. This outcome is closer to the public purpose than the number of buildings, but it still depends on how access is defined.

Source: World Bank Multipurpose Disaster Shelter Project results report, beneficiary-access indicator.

Physical distance is only one part of access. Warning lead time, transport, household decision-making, livestock, disability, gendered safety, privacy, water, toilets, lighting, and trust can determine whether a family uses a shelter. A structure can be geographically near and socially inaccessible.

Operational evidence helps. The report says about 190 project shelters were used during Cyclone Remal in 2024. Actual use demonstrates that at least part of the network functioned under stress. It does not prove that every designated catchment evacuated, that capacity was adequate, or that conditions were acceptable. Those questions require event-level records.

The Cyclone Preparedness Programme provides the warning and volunteer system that makes physical assets usable. Buildings without trusted warning, household communication, and evacuation support are inert capacity. The reverse is also true: volunteers cannot evacuate people safely when routes fail or the nearest structure is full. The shelter account and preparedness account need to meet at the catchment level.

Evacuation is a sequence. A warning must reach the household, be understood, be trusted, leave enough time, and offer a destination that people consider safer than remaining. Each failed transition can be measured during a drill or event. Recording only final occupancy loses the reason some households did not arrive. A readiness system should therefore collect process observations from volunteers and residents, including barriers that the building owner cannot solve alone.

Multipurpose is an operating model

The share of priority shelter needs met rose from 60.0% in 2015 to 77.51% at the last reported date, against an 80.0% target.

Source: World Bank Multipurpose Disaster Shelter Project results report, priority-needs indicator.

The project also reported 2,574 classrooms against a target of 2,940 and 142.16 kilometres of rural roads against a target of 220.0. These outputs use different units and should not be placed on one quantitative axis. Their relationship is functional: the shelter operates as a school in ordinary time, and the road determines access during an emergency.

That dual use is not a compromise. It is a financing advantage. A building used daily is more likely to be observed, maintained, and socially known than one opened only for disasters. Students, teachers, and local residents can identify roof leaks, blocked toilets, broken pumps, inaccessible entrances, or electrical failures before a cyclone tests them.

Daily use can also create conflict. School furniture, records, and routines must shift during evacuation. Management authority can be unclear. Privacy and protection needs change when classrooms become crowded temporary accommodation. A multipurpose standard must specify the transition from school to shelter and back, including who decides, who holds keys, where supplies are stored, and how learning resumes.

Maintenance finance is the quiet part of the arithmetic. Capital projects can count structures at completion. Local institutions inherit pumps, ramps, solar systems, toilets, roofs, and roads that deteriorate on different cycles. A national dashboard should show readiness and repair responsibility long after project closure.

School use can make maintenance visible, but it can also shift costs to education. A head teacher should not have to choose between classroom repair and emergency equipment from an ordinary school budget. The asset register should separate routine school maintenance from resilience-specific systems and assign funding for both. It should also record whether repairs preserve accessibility and safety features rather than replace them with cheaper components that pass a visual inspection but fail during evacuation.

The strongest counterargument is that incomplete output can still be success

The strongest objection is that a project can improve resilience substantially without meeting every target. It rehabilitated the planned existing stock, completed hundreds of new shelters, provided access to more than a million beneficiaries, and supported daily education. Delays in part of the portfolio should not turn a large delivery record into a failure narrative.

That objection is right. The reported results do not support calling the project a failure. They support a more precise conclusion: substantial delivery coexists with identifiable coverage gaps. Honest evaluation should preserve both facts.

The problem arises when aggregate success hides the geography of the shortfall. A target miss concentrated in a high-risk or hard-to-reach area can matter more than the national percentage suggests. Conversely, a lower construction count may have less effect if siting and capacity reach more people than assumed. Only catchment-level data can resolve that.

The appropriate accountability is therefore not punishment for every missed output. It is a public completion and readiness plan for the people still outside adequate access. Project closure should transfer unresolved items to named domestic owners rather than make them disappear from reporting.

That transfer should include documents, warranties, maintenance schedules, land records, and unresolved contractor claims. Local managers cannot maintain an asset whose design and responsibility are unclear. A closing report should identify which institution accepts each shelter, road segment, and classroom, what condition it accepted, and which defects remain. The record protects both the community and the receiving agency from an unfunded handover.

What would change this conclusion

A checkable test would alter the argument: if catchment analysis shows that the unfinished new shelters do not leave any high-risk population beyond safe evacuation time or capacity, and if existing facilities meet readiness standards, then the construction shortfall would be less consequential than the raw target gap suggests. The analysis must publish routes, capacity, population assumptions, and accessibility criteria.

First, publish a shelter-catchment register. For every shelter, record capacity, catchment population, route condition, travel constraints, accessibility, water, sanitation, power, communications, and readiness date. Owner: disaster-management authorities with local government and the Cyclone Preparedness Programme. Success signal: every high-risk coastal settlement maps to a verified operational shelter and route, or to a dated gap-closure plan.

Second, finish or re-plan the missing capacity transparently. Review incomplete sites, determine whether completion, replacement, expansion, or transport support best closes each gap, and publish the decision. Owner: Local Government Engineering Department. Success signal: the gap between 408 completed new shelters and the 530 target is converted into site-level resolutions with budgets and accountable deadlines.

Third, test the network before the hazard. Run evacuation drills that include schools, volunteers, people with disabilities, women, children, older residents, and livestock arrangements, then repair failures. Owner: local disaster committees and shelter managers. Success signal: drill records show warning receipt, route clearance, opening time, usable services, occupancy, and corrective action for every participating catchment.

Shelter arithmetic begins with structures but ends with people arriving safely. Bangladesh's next gain will come from making the whole network visible, including the unfinished parts.

The network view also makes comparison fairer. A manager responsible for an older, heavily used shelter should not appear less effective than one receiving a new structure. Readiness, access, maintenance, and actual event performance provide the common standard.

Sources

1. World Bank. Multipurpose Disaster Shelter Project implementation results report. Retrieved 2026-08-23: https://documents1.worldbank.org/curated/en/099121724085041563/pdf/P1464641a9b4400ff1b6a21b729b52e5c2e.pdf
2. Cyclone Preparedness Programme. Official portal. Retrieved 2026-08-23: https://cpp.gov.bd/pages/static-pages/6922e077933eb65569e2734e