The most precisely measured finding about children in Gaza comes from a peer-reviewed satellite analysis rather than from any United Nations document: 68.2% of the Strip’s 475 education facilities sustained damage during the campaign’s first 46 days, 7 October to 22 November 2023, and 40.2% were functionally destroyed (Asi, Mills, Greenough, Kunichoff, Khan and colleagues, 2024, Conflict and Health, peer-reviewed). Public answers to the question rest almost entirely on the phrase “UN reports say,” which stands in for several bodies publishing several classes of document across three years, and the figures those documents carry are largely relayed rather than measured by them. This report names the documents, dates them, and reads the independently measured record behind them.
Key Findings
- 68.2% of Gaza’s 475 education facilities sustained damage in 46 days, and 40.2% met the study’s threshold for functional destruction, on multi-temporal radar imagery covering 7 October to 22 November 2023 (Asi, Mills, Greenough, Kunichoff, Khan and colleagues, 2024, peer-reviewed). The figures are satellite-only, carry no ground truth, and cover the campaign’s first phase alone.
- Sex- and age-disaggregated data exist for 253 of the 1,591 active displacement sites the Site Management Cluster verified in Gaza between 3 February and 19 April 2026, and data on female children for 15.5% of sites, so every demographic share in the largest field enumeration of the displaced rests on a subsample.
- The 82% water-insecurity share is a share of 342 assessed households, measured by the WASH Cluster in a phone survey conducted 2 March to 15 March 2026 whose authors state that its findings “cannot be considered representative of the population in Gaza.” It reached readers written as a share of families.
- The most rigorous published casualty estimate carries no child-alone share. The Lancet reports 16,699 of 28,257 deaths with age and sex recorded, 59.1%, as women, children under 18 and older people counted together as one composite (Jamaluddine and colleagues, 2025).
- UNICEF’s mid-year update of 17 July 2026 attaches no source, no date and no denominator to its 90% water-and-sanitation figure, and footnotes its displacement figure to “a UN estimate in 2025,” a cumulative construction relayed from an unnamed estimate of a prior year.
How many of Gaza’s schools were damaged in seven weeks?
68.2% of the Gaza Strip’s 475 education facilities sustained damage between 7 October and 22 November 2023, the first 46 days of the Israeli military campaign, in a peer-reviewed satellite analysis published in Conflict and Health (Asi, Mills, Greenough, Kunichoff, Khan and colleagues, 2024).
The Conflict and Health analysis measured that damage by multi-temporal coherent change detection applied to Sentinel-1A synthetic aperture radar imagery, which registers change in the radar signature of a building footprint between satellite passes. Of the same 475 facilities, 191, or 40.2%, met the study’s threshold for what it calls functional destruction: a damage area of 50% or more, read off the imagery.
Functional destruction in this study is a subset of damage rather than a second population, and the two may not be added. It is a threshold on an image, not a verified loss of service, and the analysis does not establish that any particular school stopped teaching. The denominators come from open-source humanitarian building inventories cross-checked against mapping data rather than from an official facility register, so the percentages move with the inventory rather than with a fixed list.
The Conflict and Health team states its own principal limit plainly: the lack of ground-truth data prevents the ability to quantify the accuracy of satellite radar-based damage assessments. The window is the first phase of the campaign alone. These are not war-total figures, and they describe nothing about any later period.
An education facility is a building where children are, and that is the whole of what the figure measures. Nothing in the evidence assembled here counts pupils, attendance, or how long a school stayed shut, and the humanitarian category for improvised classrooms, the Temporary Learning Space, is enumerated nowhere as an outcome. The most precise number available about children’s schooling in Gaza is a number about roofs.
School damage ran highest north of Wadi Gaza, at 83.4%
83.4% of Gaza governorate’s 199 education facilities sustained direct damage over the same window, against 47.1% in the three southern governorates below Wadi Gaza that the evacuation order of 13 October 2023 designated as the evacuation zone (Asi, Mills, Greenough, Kunichoff, Khan and colleagues, 2024).
Conflict and Health recorded North Gaza between the two, at 79.3% of its 87 education facilities. In the southern zone, 89 facilities showed direct damage and 24, or 12.7%, met the functional-destruction threshold, against 40.2% for the Strip as a whole.
The Conflict and Health analysis also reports a global spatial-autocorrelation statistic, whose result the authors describe as indicating under a 1% likelihood “that this amount of clustering would occur by random chance, regardless of the degree of damage.” That is a statement about how detected damage is distributed across the map. The conclusion the author team draws from it, an allegation that international humanitarian law was violated, is the team’s own inference and appears in the paper as such; it is a legal argument rather than a measurement, and the two are kept apart here.
The southern figure is the one a reader is most likely to want and least likely to be given. The three governorates below Wadi Gaza are the area the 13 October order named as the destination, and close to half of their education facilities registered damage inside the same 46 days. That is a distribution of detected damage and nothing more: the study measures where damage appears, not who moved, not why, and not whether any instruction was followed.
The facility inventory the paper works from covers all five governorates and holds 97 health, 475 education and 152 water facilities. Every share in this section is a share of the education column of that inventory.
Which body measured the figures these reports carry?
The 90% water-and-sanitation figure in the mid-year humanitarian update UNICEF published on 17 July 2026 carries no source, no date and no denominator anywhere in the document, which is a three-page gap analysis and funding appeal for children in Gaza rather than a statistical release.
UNICEF states the same document’s displacement figure in its own footnote: “An estimated 1.9 million out of the total 2.1 million people in the Gaza Strip have been displaced, many of them several times (UN estimate in 2025).” That is a cumulative construction, counting people displaced at any point rather than people displaced now, relayed from an unnamed estimate of a prior year and carried forward into a 2026 document.
A gap analysis exists to state unmet need and ask for money, and reading one as a statistical release is where the provenance chain breaks. The figures at the front of such a document are context, gathered from elsewhere to frame an appeal. What UNICEF publishes as its own measurement is the shortfall in its own response, which is what the document was written to report.
Two other kinds of body sit behind the phrase. A commission of inquiry reported in June 2026, and a United Nations human rights office published on Gaza through 2026; each holds a position on contested questions, each is dated, and this report carries no figure from either. The compiled case material about children shot in areas described as safe is set out separately, in the Institute’s report on the compiled shooting cases, which records what that record contains and what it lacks.
So the answer to the heading, for the document most often invoked, is that the publisher measured none of the three context figures it prints. Two of the three name no measuring body at all. The third does, and that body is the subject of the next section.
Where does the 82% water-insecurity figure come from?
82% is a share of 342 assessed households. The WASH Cluster measured it in a phone survey conducted between 2 March and 15 March 2026, and its authors state that the findings “cannot be considered representative of the population in Gaza.” It reached readers written as a share of families.
UNICEF reports the finding this way: “An estimated 82 per cent of families are water insecure, and up to 70 per cent cannot access 6 litres per person per day.” Families in Gaza number in the hundreds of thousands. Assessed households in this instrument number 342, contacted by telephone across a fortnight, and the distance between those two bases is the whole of the claim.
The measuring body is named in the record and is not the relayer, which is why this report attributes the figure to the cluster that ran the survey. That is not a technicality about attribution. A reader who knows the figure is a cluster’s telephone sample can weigh it; a reader who receives it as a share of families in Gaza cannot, because the sentence as relayed implies a population measurement the survey’s own authors decline to claim.
The second half of the relayed sentence carries the same shape. Up to 70% is a ceiling rather than a point estimate, and the 6 litres per person per day it is measured against is a threshold the sentence states without saying whose it is or where it comes from, so the threshold is doing work the sentence does not explain.
Neither figure is a figure about children. A household is a unit containing adults and children in unrecorded proportion, and the WASH Cluster publishes no child-level water share from this instrument at all. For a question specifically about children in Gaza, the most-quoted water figure in circulation measures households, on a sample of 342, in a survey that says it does not represent the population.
What did the field enumeration record inside the displacement sites?
About 1.5 million residents were living in approximately 246,000 tents or tent-like shelters, an average of about six people per tent against a standard emergency tent designed for five, in the Site Management Cluster’s verification of every displacement site on its master list, with fieldwork from 3 February to 19 April 2026.
The Site Management Cluster reported 1,591 sites as active out of 1,856 targeted, holding an estimated 1,708,822 individuals across 354,550 households at the data extraction of 16 May 2026. Twelve partner organisations deployed 165 trained enumerators across 116 neighbourhoods to produce it, which makes this the largest primary field enumeration of Gaza’s displaced population on the record.
The shelter findings underneath that headline are the specific ones. About 59,000 shelters held more than eight people. 7,700 households, roughly 38,500 individuals, were sleeping outdoors with no shelter at all, and they were present in 19% of the sites surveyed. 136,000 households, about 680,000 people and 38% of all households in sites, had received no shelter assistance in the six months before the survey.
The cluster records 83% of the 1,591 sites as makeshift, meaning informal, open-air and tented. About 11% are collective centres in schools and public buildings, and scattered sites of fewer than ten households make up the remainder. That 11% is where this section meets the first: the buildings sheltering displaced families in the collective-centre category are in substantial part the same class of building the damage analysis counted, which is why a school-damage share and a shelter share describe one physical stock from two directions.
Makeshift here is a site type rather than a shelter type, and its counterparts are a collective centre and a scattered site, never a permanent building. The population figures are estimates throughout, and the assessment presents itself as a baseline rather than a point-in-time snapshot, with the earliest data about four months old at publication.
Age was recorded at 253 of the 1,591 displacement sites
Sex- and age-disaggregated data exist for 253 of the 1,591 active displacement sites the Site Management Cluster verified in Gaza, and data on female children for 15.5% of sites, so every demographic share in the largest field enumeration of the Strip’s displaced rests on a subsample.
The household count is derived rather than enumerated. Where disaggregated data were unavailable, the cluster applies a standard household size of five, so no per-household figure in the assessment is a count of households, and any figure expressed per household inherits that assumption rather than measuring it.
The Site Management Cluster states what the site count is, in its own words. The 1,591 “should be understood as a social count: A measure of how many organised displaced communities were identified and reached rather than a census of discrete physical locations.” A single interview per site also means the figures for persons with disabilities and for female-headed households are underestimates, which the assessment says of itself.
Two figures for displacement circulate and they may not be subtracted from one another. The roughly 1.7 million counted in sites is a count of people in sites at a stated date; the 1.9 million relayed elsewhere is cumulative, covering people displaced at any point since October 2023, many of them repeatedly. No publisher reconciles the two, and a residual produced by subtracting one from the other would belong to whoever performed the subtraction.
What that leaves, for a question about children, is this. The Site Management Cluster recorded the ages of the people it found at fewer than one site in six, and the ages of girls at fewer than one in six again. The enumeration is real, the enumerators were trained, the coverage was near-total by site, and the demographic answer still rests on a subsample the cluster names precisely.
Does any casualty estimate publish a child-alone share?
59.1% of the deaths for which age and sex were recorded, 16,699 of 28,257, were women, children and older people counted as one composite category, in a peer-reviewed capture-recapture analysis published in The Lancet (Jamaluddine and colleagues, 2025). No child-alone share appears anywhere in it.
The Lancet analysis reports three groups in that one number: women, children under 18, and people aged 65 and over. Nothing in the paper separates the middle category, so the figure a reader wants, the share of the dead who were children, cannot be recovered from it by any arithmetic the paper supports.
The denominator matters as much as the composite. 28,257 is the count of deaths for which age and sex data were available, and it is not the study’s estimate of total mortality. That estimate is 64,260 deaths due to traumatic injury between 7 October 2023 and 30 June 2024, with a 95% confidence interval running from 55,298 to 78,525. Applying 59.1% to 64,260 produces a figure that appears nowhere in the paper, built by multiplying a share from one base against a total from another.
The estimate is also narrower than the word casualty suggests. It covers traumatic injury only, and the authors state that it excludes deaths from healthcare disruption, food insecurity, water and sanitation failure and disease, which is the category most likely to bear on the very young.
The Lancet study documents three input lists: a hospital morgue record holding 22,368 identified decedents, a rolling online mortality survey launched on 1 January 2024 holding 5,817, and manually scraped social media obituary pages. Two of the three are self-reported, and the first belongs to the health ministry of a party to the war, reported here as that ministry’s own record. The estimate is independent academic work built on those inputs, and it publishes no figure for children alone.
Why do published shares of women and children killed differ?
9,692 unidentified decedents sat in the Gaza Ministry of Health’s 30 June 2024 update, 26% of its cumulative total, and were excluded from the hospital list the analysis worked from, along with 930 people reported missing from its survey list (Jamaluddine and colleagues, 2025).
That is the mechanical answer. The Gaza Ministry of Health reported 37,877 decedents as of that date, of whom 28,185 were identified, and age and sex were available for 28,257 deaths in the analysis. So four different denominators are in circulation for the same event, and each produces a different share: all reported deaths, identified deaths, deaths with age and sex recorded, and a modelled estimate of total traumatic-injury mortality.
The study’s under-reporting figure carries the same trap. Its 41% is a share of its own 64,260 estimate, never a multiplier on the ministry’s 37,877. The gap between the two is 26,383 deaths, which is 41% of 64,260 and 70% of 37,877, so a reader who applies the percentage to the ministry figure arrives at a number the paper does not contain.
One further absence explains more of the divergence than any of the above. The Gaza Ministry of Health publishes no combatant and noncombatant breakdown at all, and the Costs of War compilation at Brown University records that absence (Crawford, 2025, university research project, not peer-reviewed). Every civilian share in circulation is therefore somebody’s construction from other inputs rather than a category the registry itself keeps.
Set together, the published shares differ because the bodies producing them count different populations over different windows on different definitions, and because the composite categories they publish were not built to answer a question about children. The distinctions between the counting bodies are set out at greater length in the Institute’s report on what each body counts.
How did the schools finding change as it left the journal?
42.1% of Gaza’s 152 water facilities sustained damage in the same 46 days, and the announcement of the study wrote “more than half of healthcare, education, and water facilities,” a sentence that holds only of the pooled total and not of water taken alone (Asi, Mills, Greenough, Kunichoff, Khan and colleagues, 2024).
The Harvard FXB Center published that announcement on 9 April 2024. The arithmetic behind it is one division: pooled across the three classes, 447 of 724 facilities showed damage, which is 61.7% and above half. Taken separately the three shares are 60.8% of health facilities, 68.2% of education facilities and 42.1% of water facilities, and the third does not clear half by any reading.
The Conflict and Health Discussion states it in the authors’ own words: over half of health and education facilities, and over a third of the water facilities.
A parallel limit sits on the health column. The paper’s denominator is 97 health facilities, a category that includes clinics and primary-care centres, against 36 hospitals in the Strip. So 60.8% is a share of health facilities of every kind and is never a share of hospitals, however often the two are written as though they were one.
This is the same mechanism the earlier sections describe, running once more and in a place a reader can check in a single division. A telephone survey of 342 households became a share of families. A cumulative estimate of people ever displaced became a description of people displaced now. Three separately reported shares became one sentence about more than half.
None of these relays invents a number. Each takes a real figure produced by a named body and widens the thing it is a figure about, and the widening is what travels, because the widened version is the shorter sentence.
Methodology
This report assembles published figures about children in Gaza and traces each one to the body that produced it. Every figure is labelled by source type and dated, and where a document relays a figure rather than measuring it, the measuring body is named instead of the relayer.
Peer-reviewed research supplies the two measured spines. Conflict and Health published a spatial analysis of Sentinel-1A radar imagery over 97 health, 475 education and 152 water facilities across all five governorates for the 46 days from 7 October to 22 November 2023; it is satellite-only, has no ground truth, and its authors state that this prevents quantifying the accuracy of the damage assessment. The Lancet published a three-list capture-recapture estimate of traumatic-injury mortality from 7 October 2023 to 30 June 2024, with a point estimate of 64,260 and a 95% confidence interval of 55,298 to 78,525.
Primary field enumeration by humanitarian coordination clusters supplies the shelter and water figures. The Site Management Cluster ran its displacement-site verification from 3 February to 19 April 2026, extracting data on 16 May 2026; its population figures are estimates, its household counts apply a standard size of five where disaggregated data were unavailable, and it describes the 1,591 sites as a social count rather than a census of locations. The WASH Cluster measured its March 2026 light-touch monitoring on 342 assessed households by telephone, and states that it cannot be considered representative of the population in Gaza.
An official statistical record supplies the population base. The Palestinian Central Bureau of Statistics counted 1,899,291 residents of the Gaza Strip in its 2017 census, the most recent enumeration, so every later population total for the territory is a projection off that measured base.
Two documents are read here for provenance rather than for their figures: the mid-year humanitarian update UNICEF published on 17 July 2026, a gap analysis and funding appeal, and the announcement the Harvard FXB Center published on 9 April 2024 relaying the education-damage study. Costs of War at Brown University is used for one recorded absence, and it is not peer-reviewed.
Three limits bind everything above. No independent enumeration of Gaza’s population or its dead has been possible since the war began, so every count is a registry, a projection, a satellite reading or a modelled estimate. The shares here have different bases, and each is stated with its own: 68.2% is of education facilities, 82% is of assessed households, 59.1% is of deaths with age and sex recorded. And no source assembled here publishes a child-alone count of deaths or injuries in Gaza; that absence is recorded rather than filled.
Conclusion
So what do these reports find about children in the Gaza Strip? Read one by one, with their dates and denominators attached, they find less than the phrase promises and more than a summary can carry. The strongest measured findings are a satellite reading of 475 school buildings over 46 days, a field enumeration of 1,591 displacement sites in which age was recorded at 253 of them, and a mortality estimate whose only demographic figure puts women, children and older people into a single composite on a partial denominator.
Around those, the record thins fast. The most-relayed water figure was measured on 342 households by telephone and states in its own methodology that it does not represent the population. The most-quoted displacement figure is cumulative, relayed, and a year older than the document carrying it. The most-quoted damage sentence pools three facility classes to clear a threshold that one of them does not clear alone. Every one of those figures is real, every one has a named producer, and every one describes something narrower than the sentence it arrives in.
What is missing is the simplest thing anyone would ask for. There is no register in this evidence base that counts Gaza’s children as children: not their deaths, not their injuries, not their schooling, not their ages inside the sites they live in. The composite exists because the categories were built for something else, and the subsample exists because an enumerator with one interview per site cannot ask everything.
A figure that has been widened once travels better than the figure it came from, because the widened version is shorter and answers the question people actually asked. Households become families, sites become the displaced, three shares become more than half. If the public record about the youngest people in a war is assembled almost entirely out of numbers that were widened on the way, what exactly is it that the public knows?
Frequently Asked Questions
Which satellite method produced the education-damage figures?
Multi-temporal coherent change detection applied to Copernicus Sentinel-1A synthetic aperture radar imagery, overlaid on facility locations drawn from open-source building footprints (Asi, Mills, Greenough, Kunichoff, Khan and colleagues, 2024). Radar registers change in a building footprint between satellite passes, so it detects structural change rather than confirming what happened inside a building.
Is the 60.8% health-facility figure a share of Gaza’s hospitals?
No. The denominator is 97 health facilities including clinics and primary-care centres, against 36 hospitals in the Strip (Asi, Mills, Greenough, Kunichoff, Khan and colleagues, 2024). A share of health facilities of every kind and a share of hospitals are different quantities, and the study reports only the first.
Why can the two displacement figures not be subtracted from one another?
The roughly 1.7 million counted by the Site Management Cluster is a count of people in displacement sites at a stated date; the 1.9 million UNICEF relays is cumulative, covering people displaced at any point and many of them more than once. No publisher reconciles them, so a difference between them would be the subtractor’s own number.
What does the Site Management Cluster mean by a makeshift site?
An informal, open-air, tented site, which is one of three site types alongside collective centres in schools and public buildings and scattered sites of fewer than ten households. It describes the site, not the shelter, and 83% of the 1,591 verified sites fall into it.
Does the Lancet estimate include deaths from hunger and disease?
No. The Lancet estimates deaths due to traumatic injury only, and its authors state that it excludes deaths arising from healthcare disruption, food insecurity, water and sanitation failure and disease (Jamaluddine and colleagues, 2025). Its point estimate is 64,260, with a 95% confidence interval of 55,298 to 78,525.
Does the Gaza Ministry of Health publish a combatant and noncombatant breakdown?
It publishes none, and Costs of War at Brown University records the absence (Crawford, 2025, not peer-reviewed). Every civilian share in public circulation is therefore built from other inputs, and the ministry’s own registry is a party’s own count of deaths reported through health facilities.
What population base do shares about Gaza rest on?
The Palestinian Central Bureau of Statistics counted 1,899,291 residents of the Gaza Strip on a reference date of 1 December 2017, in the most recent census conducted in the territory. Every population total published since is a projection off that measured base, adjusted for war-related mortality and out-migration rather than enumerated.
Is there a published count of school attendance in Gaza during the war?
None appears in the evidence assembled here. The damage study counts buildings, the displacement assessment counts sites and shelters, and the humanitarian category for improvised classrooms, the Temporary Learning Space, is not enumerated as an outcome anywhere in that record. The absence is recorded rather than estimated.
Sources
- Asi, Y., Mills, D., Greenough, P. G., Kunichoff, D., Khan, S., and colleagues, 2024. “Nowhere and no one is safe”: spatial analysis of damage to critical civilian infrastructure in the Gaza Strip during the first phase of the Israeli military campaign, 7 October to 22 November 2023. Conflict and Health, 18, article 24. DOI 10.1186/s13031-024-00580-x. Peer-reviewed, open access.
- Crawford, N. C., 2025. The Human Toll of the Gaza War: Direct and Indirect Death from 7 October 2023 to 3 October 2025. Costs of War project, Watson School of International and Public Affairs, Brown University, 7 October 2025. costsofwar.watson.brown.edu. University research project report, not peer-reviewed.
- Harvard FXB Center for Health and Human Rights, 2024. Announcement of 9 April 2024 relaying the Conflict and Health spatial analysis of damage to civilian infrastructure in the Gaza Strip. Research-centre announcement, cited here as the relayer of a peer-reviewed finding and not as a source of measurement.
- Jamaluddine, Z., Abukmail, H., Aly, S., Campbell, O. M. R., Checchi, F., 2025. Traumatic injury mortality in the Gaza Strip from Oct 7, 2023, to June 30, 2024: a capture-recapture analysis. The Lancet, 405(10477), pp. 469 to 477. DOI 10.1016/S0140-6736(24)02678-3. Peer-reviewed.
- Palestinian Central Bureau of Statistics, 2018. Population, Housing and Establishments Census 2017, final results. Ramallah. pcbs.gov.ps. Official record, national census.
- Palestinian Ministry of Health, 2024. Cumulative casualty figures for the Gaza Strip as of 30 June 2024, comprising 37,877 reported decedents of whom 28,185 were identified, as reported in Jamaluddine and colleagues (2025). A party’s own registry count, not independently verified.
- Site Management Cluster in the Occupied Palestinian Territory, 2026. Gaza Displacement Site Verification Assessment: Baseline Findings for May 2026 Humanitarian Planning and Coordination, published 22 June 2026. cccmcluster.org. Humanitarian coordination body, primary field enumeration.
- UNICEF State of Palestine, 2026. Humanitarian Situation Update: Analysis of the Response to the Needs of Children at Mid-year 2026, 17 July 2026. unicef.org/sop. Agency situation update and funding appeal; the context figures it prints are relayed from other bodies and are read here for provenance.
- WASH Cluster, occupied Palestinian territory, 2026. Light Touch Monitoring, Gaza, March 2026. Phone survey of 342 assessed households conducted 2 March to 15 March 2026; its authors state the findings cannot be considered representative of the population in Gaza. Humanitarian coordination cluster, primary assessment, recorded within the UNICEF mid-year update that relays it.