A Country Reorganised

Seventy-three days into the conflict, Lebanon holds roughly one million displaced people, of whom 127,000 are registered in collective shelters. The remaining nine-tenths sit in host families, rented rooms, and arrangements no agency tracks at scale. The footprint peaked at 141,700 on 16 April, dropped 30,700 in a single day when the ceasefire opened, and has partly re-accumulated since.

The geographic story matters most. The conflict concentrated its damage in a handful of southern cazas, and in those cazas the population is gone, sitting now in Beirut, Mount Lebanon, and the coastal South. The shelter system holding them is structurally at the ceiling: Beirut runs at 101 percent, the coastal South at exactly 100, and the national buffer is perhaps 10,000 places against waves that have already run 4,000 to 5,000 in a day.

The aid envelope is structurally mismatched to the displacement it nominally addresses: emergency food flows, reconstruction does not. The aid that holds the displaced alive at the shelter is operating. The aid that would reverse the displacement by reconstructing origin villages has not begun.

The official headline reads one million displaced; the shelter system holds one in ten of them. Three measurements converge within 2 percent on the in-shelter caseload: the Ministry of Social Affairs count (127,563 on 13 May), the UN weekly restatement (125,621), and the national estimate (126,900). The number outside that convergence is the one that matters: UNRWA restates the national displaced count at 1,100,000, which leaves 974,000 individuals, 88.6 percent of the displaced, in arrangements the system does not see directly.

The 88 percent informal share is not transient; it is the structural feature of Lebanese displacement absorption, the same 85-to-90-percent pattern the 2024 cycle produced. Two consequences follow. The collective shelter system does not measure displacement; it measures the residual caseload with no kinship or financial access to informal hosting. And the system is more exposed to second-order shocks than the headline suggests: a rental-market squeeze or a kinetic event in hosting cazas would push the informal caseload back into a formal system with no capacity to receive it.

The trajectory tells a story of stabilisation under stress: a climb through March, a peak coinciding with the heaviest strike week, a one-day return of unmatched scale, and a climb back under continued strike pressure. The net position carries the underlying message of the report. Outbound displacement is the floor, not the ceiling.

The country's displacement starts where its buildings were destroyed, and the buildings were destroyed where the strikes fell: 66 a day before the ceasefire, 36 after it. The bombing slowed. It never stopped.

The May 8 Core Group damage assessment found 12,188 buildings moderately to severely damaged, with Marjaayoun and Bent Jbeil accounting for 60 percent of the damage on 14 percent of the assessed stock; their worst villages carry rates of 30 to 60 percent. For displacement, the threshold matters more than the totals: a village at 5 percent damage can take its people back and rebuild around the loss; a village at 30 percent cannot. The deep south sits decisively on the second side. Sour and Saida register 2,400 damaged buildings but at rates under 1 percent, with their urban cores intact, which is why the coast absorbs its own displaced while Bent Jbeil and Marjaayoun cannot. Full cadastre-level figures are documented in SF-LBN-DMG-2026-05.

The displaced went north and west. Eight cazas absorb 80 percent of the registered caseload, peaking in Beirut: 49,520 individuals across 138 sites, against a nominal capacity of 48,872, with 43 percent of sites above capacity. Per capita, Beirut carries 142 displaced per 1,000 residents, 2.5 times the next-heaviest caza. Bent Jbeil and El Nabatieh sit at near zero on that map, because the populations that would constitute the denominator have left.

Mount Lebanon is the country's absorption layer, the only major hosting governorate with headroom: roughly 10,000 places, the formal system's entire buffer against further re-displacement, smaller than two of the three waves already produced. The coastal South has none: Saida, Sour, and Jezzine run at exactly 100 percent. Displacement also crossed the borders: the first two weeks produced 256,647 exits at Lebanon's crossings, nearly three-quarters of them Syrian nationals moving back the other way.

The displaced outside collective shelter follow a different geography, concentrated in Beirut, coastal Mount Lebanon, and the western Bekaa. The kinship system that holds those 974,000 depends on host households' physical capacity, financial capacity, and willingness to continue. None is unlimited, and the rental-market compression visible in central Beirut since April indicates the financial dimension is already under pressure.

The 630 collective sites are the absorption layer of last resort, and they are operating at the structural limit: 65 percent at 90-plus percent utilisation, 77 sites above capacity. Beirut is the most over-stressed system in the country, 101 percent and stable there for three weeks. The capital is full. A re-displacement of even 5,000 from the deep south would have nowhere to go inside the South governorate: the overflow goes north to Mount Lebanon, into over-density compression, or into host families that may themselves be at capacity.

The formal system holds 124,000; UNRWA's parallel camp system holds 1,100; everything else, 974,000 individuals, sits in arrangements the response does not track at the household level. The operationally consequential decisions about the displaced majority, income support, child protection, education continuity, food security, take place outside the formal system entirely.

The sheltered population is younger than the resident demography: children under 18 carry 37 percent of the age-coded caseload against a resident share near 31, because families with school-age children evacuated at higher rates while working-age men stayed at origin for property protection. The female share, 52 percent, is uniform across regions. The substantive variation runs along age: rural hosting in the North and Akkar draws child shares above 40 percent, while the urban South sits at 31, with operational consequences the aggregate conceals, child-heavy caseloads in small hosting systems and large caseloads with no headroom for child-specific space.

Two dimensions sit outside the data feeds: sect is not recorded at registration, and pre-conflict income is not captured at all. The geography speaks to both. A predominantly Shia population from rural agricultural cazas now sits in mixed and majority-Sunni or Druze cazas, a minority by sect and by class in most of the cazas hosting it. The class dimension carries the harder consequence: agricultural livelihoods do not transfer to urban hosting, and the income channels sustaining displaced households, remittances, savings, informal aid, each carry a finite horizon. That horizon is the structural variable that determines how long the current configuration can hold.

The ceasefire produced the single most consequential one-day movement in the record, 30,700 returns in 24 hours, and it tested the proposition that displacement was conditional on the active strike envelope. The proposition held in some places and failed in others: Baalbek-El Hermel emptied 40 percent in a week, the coast the same, while El Nabatieh and Bent Jbeil registered essentially zero. One variable structures the pattern, the survivability of the origin cadastre. Return is a function of physical reoccupation feasibility, not ceasefire political status.

Re-displacement followed: +5,400 on 20 April, +4,500 on 1 May, +5,000 on 4 May, each at a 24-to-72-hour lag behind an identifiable strike wave. The cumulative 16,000 re-displaced hardened the composition: the returnees were households whose villages remained reoccupiable, and the population that came back to the shelters concentrates the worst-damaged cadastres. The aggregate is steady. The structure underneath it is not the same caseload it was in March.

Over the next 30 to 90 days, the most likely path holds the caseload at 125,000 to 130,000 under compressed kinetic activity. A lower likelihood branch has a second ceasefire extension returning 5,000 to 10,000 out of the Bekaa and the coast; its mirror has strike re-intensification pushing 10,000 to 20,000 into the Mount Lebanon headroom. The lowest likelihood branch, reduction toward pre-conflict levels, requires a sustained 90-day ceasefire and reconstruction at scale, conditions the data does not show developing.

Since 2 March, $2,404 million has been committed or paid; $501 million has actually been paid. The aggregate disguises the pattern that matters. Emergency food flows at near-complete paid share: the displaced are eating. Reconstruction relief is the inverse, $657 million committed and $1.7 million paid, with water and sanitation at 4.5 percent on the same curve. These are the sectors whose investment would reverse displacement. They are not flowing, and the gap is disbursement, not commitment: rebuilding the deep-south cohort costs an estimated $300 to $600 million, inside the already-pledged envelope.

The donor pattern mirrors the sectoral one: multilateral reconstruction commitments accumulate without disbursement (UNDP, $1,370 million committed, $3 million paid) while smaller emergency disbursements actually move (WFP, $81 million of $136 million). The aid record anticipates the structural condition this file describes. The displacement system is durable not despite the aid envelope but in part because of how it is structured: short-cycle response keeps the displaced alive in place, and long-cycle reconstruction that would reverse the displacement has not begun.