Spain's 2024 record in centres for homeless people: what the +57.5% measures

Abstract

INE, Spain's statistics office, reported a mean of 34,145 occupied places a day held by adults in centres for homeless people in 2024. That is +57.5% on the 21,684 of 2022, and the highest level and steepest rise in the series INE charts from 2006. The figure is a mean of the places that responding centres report as occupied on two days. It is not a count of homeless people.

The rise.

• Centres that INE classes as specialised in immigrants account for 10,128 of the 12,461 additional occupied places, or 81.3%.
• INE does not say what these centres are. By its 2024 definition they include temporary accommodation for asylum seekers, immigrants and seasonal workers.
• In the record year INE also moved its definition of homelessness to the ETHOS typology, and for the first time asked each centre to declare its specialisation. Its release mentions neither change.

The split.

• A symmetric split gives 5,128 places to more immigrant-specialised centres (4,110–6,145 across orderings) and 5,000 to more people per such centre (3,983–6,018).
• In the other centres, occupied places rose +17.1% and their number offering accommodation +11.9%, so people per centre rose +4.7%. That average mixes −24.6% in gender-violence shelters with +7.9% in the rest.
• The share is 78–81% under the three variants that can move it.

The revision. The figure first published, 33,758 (+55.7%), was changed on 17 October 2025. The change adds 387 occupied places, all in immigrant-specialised centres; regionally only Navarre's figures changed. No note explains it.

Two comparisons.

• INE's immigrant-specialised places equal 17%, 27% and 37% of the places of the state asylum and humanitarian reception programmes in 2020, 2022 and 2024. This ratio is not a coverage rate.
• The national strategy's coverage indicator divides the survey's places, reception included, by an estimate of homeless people that contains almost no reception residents (0.4%). On the strategy's basis it rose from 70.7% to 148.1%.

What the record is. The record is a real record of occupied shelter places. Under the ETHOS typology that Spain adopts, the people in them are houseless, though no EU state recognises reception as homelessness in law. The record is not evidence of a comparable rise outside immigrant-specialised centres.

We register computable predictions for the 2026 edition.

1. Introduction

On 26 September 2025 INE published its survey of centres and services for homeless people for 2024 (ECAPSH). The headline:

«Una media de 34.145 personas mayores de 18 años se alojó diariamente en el año 2024 en centros de atención a personas sin hogar (un 57,5 % más que en 2022).»

The same day:

• RTVE wrote that «El sinhogarismo alcanza cifras récord».
• Europa Press led with the number.
• Organisations working with homeless people read it as a rise in homelessness, «con especial impacto en la población migrante» (FACIAM).

The release itself says more:

«Los centros que atendieron exclusivamente a personas inmigrantes ofrecieron una media diaria de 21.298 plazas de alojamiento (un 100,9 % más que en 2022), con una media de 18.173 plazas ocupadas (un 125,9 % más que en 2022).» «Cabe destacar que las llegadas masivas de inmigrantes a Canarias y el reparto de inmigrantes provocaron el aumento del número de plazas de alojamiento disponibles, especialmente en el periodo invernal.»

So INE separates the immigrant-specialised component and names a cause. It does not say:

• how much of the headline that component is;
• how much of the rest follows from the larger number of centres in its tables;
• that its definition of the population changed in the same edition.

This paper does that arithmetic, documents the definition change, tests how much the result depends on definitions, and shows how the figure enters the national homelessness strategy. It then registers what the next edition should show under two readings.

2. What the survey counts

A census of a directory, without imputation.

• INE: «Se investigan exhaustivamente todos los centros del directorio facilitado al INE por el Ministerio de Derechos Sociales y Agenda 2030 a través de las correspondientes consejerías de todas las comunidades autónomas, excepto el País Vasco, con referencia a 30 de junio de 2024».
• The Basque Country is in the data. Its fieldwork is done by its own statistics office: «excepto en el País Vasco, donde ha sido el Instituto Vasco de Estadística (EUSTAT) el encargado de los trabajos de campo».
• INE's standardised methodological report (IME) states «La tasa de imputación es A7 = 0%» and «Señalar que no se calibra en esta operación estadística».
• The published totals are therefore sums over the centres that answered: 79.9% or 80.0% of the directory in 2024, depending on the document. The 1,376 centres INE reports are respondents. A rise in that number mixes centres that opened, centres that entered the directory and centres that answered this time but not last time.

Occupancy is a mean of two days. Places and occupied places refer to two reference dates (14 June and 16 December in 2024), and INE publishes their mean. Since 2022 occupied places count adults only, while places count all beds:

«A partir del año 2022 se contabilizan las plazas ocupadas por personas de 18 y más años, mientras que la capacidad se refiere al número total de plazas disponibles en el centro.»

The occupancy ratio (occupied ÷ places) therefore also moves with the share of children, not only with how full centres are.

The definition changed in 2024.

• In 2020 and 2022, INE's methodology defined a homeless person by eight accommodation situations. It then added: «Se incluyen, además, los siguientes tipos de centros para asegurar la continuidad de la serie histórica y la comparabilidad internacional». The list included «Centros de inmigrantes: Centros de estancia temporal de inmigrantes (C.E.T.I.) y Centros de Acogida a Refugiados (C.A.R.)» and «Centros destinados al alojamiento de trabajadores temporeros».
• In 2024: «En esta encuesta se toma como referencia la Tipología europea de sinhogarismo y exclusión residencial (ETHOS …) contemplada en la Estrategia Nacional para la lucha contra el Sinhogarismo en España 2023-2030». Among the homelessness categories it lists: «Personas en centros de alojamiento para solicitantes de asilo e inmigrantes: Personas inmigrantes que viven en alojamientos temporales por su estatus de extranjeros o trabajadores temporeros».
• The operation was renamed «… y en exclusión residencial».
• The 2024 release does not mention the change.

"Specialisation" exists from 2020, and its source changed in 2024.

• INE's definition, the same in the 2020, 2022 and 2024 methodologies: «Distintas situaciones que el centro atiende específica y exclusivamente. Se consideran las siguientes: mujeres víctimas de violencia de género, inmigrantes y sin especialización o con otra distinta de las anteriores».
• The 2024 questionnaire asks the centre itself on its cover: «Centro especializado en la atención a migrantes». It does not say "exclusively".
• The published 2020 and 2022 questionnaires have no such item, and INE does not say how those editions classified centres.

So the comparison 2022 → 2024 crosses both a change of definition and a change of classification method. Following INE's tables, we call these centres immigrant-specialised (IMM). INE does not break them down, so how many places are asylum reception, humanitarian reception after arrival by sea, CETI, or seasonal-worker housing is unknown.

The sister survey counts people, not places.

• INE's survey of homeless people (EPSH, 2022) estimated 28,552 adults using accommodation or meal services in municipalities of more than 20,000 inhabitants. It is a sample survey.
• It drew its sample from the ECAPSH 2020 directory.
• Its scope listed CETI and CAR «para asegurar la continuidad de la serie histórica y la comparabilidad internacional». Yet only «el 0,4% en centros de ayuda al refugiado» of the people it counted had slept in refugee centres (release of 19 October 2022).

Is reception homelessness?

• Under full ETHOS, as reproduced in the Strategy (Cuadro 1) and adopted by INE in 2024, «Personas en centros de alojamiento para solicitantes de asilo e inmigrantes» are «Sin vivienda», houseless.
• The European Observatory on Homelessness (Comparative Study 14, 2025) states the limits of that view: «people in systems that are designed to provide accommodation to asylum seekers are in situations of homelessness according to ETHOS. Administratively, politically and legally no EU Member States recognises this categorisation of homelessness. ETHOS Light …, which is the basis for many homelessness counts …, does not define accommodation for immigrants as 'homelessness'».

This paper does not take a side. Spain's strategy and INE count these people as houseless, they need shelter, and the statistic is to be read for what it counts.

3. Prior work

What we searched. On 4 October 2026 we ran 53 searches, all listed in data/prior_work_search.csv:

• 29 on DuckDuckGo;
• 17 on Crossref and OpenAlex;
• 3 on GitHub;
• 4 attempts with a web-search tool whose quota had run out.

We also read the press coverage of the release, the statements of organisations, the national strategy and its progress reports, the FEANTSA overview and, after review, EOH Comparative Study 14. The press search is the weak point: no general web search engine was available.

The point was made on the day of release.

• HOGAR SÍ, 26 September 2025. It made both points this study quantifies.
  • Composition: «La inclusión de plazas correspondientes a centros que atienden a personas migrantes y solicitantes de asilo y refugio -con un incremento del 97,2 %- condiciona drásticamente el aumento del 55% en el número total de plazas que refleja la encuesta».
  • The frame: «la organización advierte sobre incoherencias en la selección de los centros en los que se recogen los datos de la encuesta que podrían distorsionar la comprensión del fenómeno del sinhogarismo en España».

We share the observations, not the framing «vinculado a la migración y no al sinhogarismo» (§2).

• RTVE, same day: «El crecimiento se debe, en gran medida, al fuerte aumento en la atención a personas migrantes». The sentence follows the number of centres («La red de centros se expandió hasta los 1.375 establecimientos»), so it is about centres. RTVE also wrote that the figure «evidencia una presión sin precedentes sobre la red asistencial».
• INE named a cause in each recent release.
  • 2024: the Canary Islands (§1).
  • 2022: «Cabe destacar que el levantamiento de las restricciones de aforo impuestas por la COVID-19 y la afluencia de personas desplazadas por el conflicto en Ucrania provocaron el aumento del número de plazas de alojamiento disponibles, especialmente en el periodo invernal».
• The Strategy warned in 2023: «una parte del incremento de las plazas contabilizadas por esa encuesta podría también deberse a la mejora en cuanto a la capacidad de detección e identificación por parte del INE de los centros que prestan este tipo de servicios».

Readings of the figure as a count of homeless people.

• Accem (Europa Press, 24 November 2025) used the revised figure. It wrote that «crece el número de personas migrantes o solicitantes de protección internacional que acude a recursos para personas sin hogar, suponiendo el 52%, según el INE». INE's 52.0% is the share of centres oriented to «Inmigración/Solicitud de protección internacional» among the centres that declare any specific orientation (437 of 841), not a share of people.
• The Ministry of Social Rights, in its 2024 progress report on the strategy (cover dated June 2026), wrote that the survey «estimaba en 34.145 el número de personas sin hogar en España» and that «Entre 2020 y 2024, el número total de plazas en centros de alojamiento colectivo casi se duplica ante un contexto de alta demanda social».
• Two leads we could not read under our access rules.
  • A Government written answer to the Congress of October 2025 on the rise. congreso.es refuses robots.txt.
  • A Cáritas campaign of October 2025. EFE's robots.txt disallows our User-Agent.
• FEANTSA and the Fondation pour le Logement, eleventh overview (September 2026): Spain is among six countries that «have no new data since the previous edition of this report». They did not use the 2024 survey.

Research.

• Pleace, Filipovič Hrast, O'Sullivan, Allen and Benjaminsen (European Observatory on Homelessness, Non-EU Migrant Homelessness, Comparative Studies 14, December 2025).
  • It discusses whether reception is homelessness (quoted in §2).
  • Its Spain section notes that the reception system «provides over 34 000 places, up from approximately 1 920 in 2015», and that «widespread use of homelessness services by people seeking asylum was not reported».
  • It does not analyse the 2024 survey.
• Meyer, Wyse and Williams (2026, Journal of Public Economics; NBER working paper 2025) analyse «an unprecedented 43 percent increase in the number of people residing in homeless shelters in the United States between 2022 and 2024».
  • They estimate «that new migrant inflows were the primary driver of this dramatic rise, explaining about 59–62 percent of the two-year increase in sheltered homelessness».
  • They ask for care with «population shifts, shelter capacity, and measurement practices when interpreting changes in official homeless counts over time».
• On counting:
  • in Ireland, Davret (2026) notes that the official figure excludes asylum seekers;
  • the Observatory studied asylum seekers and homelessness services in 2016;
  • a 2025 review of homelessness measurement in Spain (Gallén-Granell, Botija and Galán Sanantonio) points to «methodological inconsistency»;
  • Marbán Gallego and Rodríguez Cabrero (2025) cite the 2022 edition: «Some 20% of these accommodation centres catered exclusively to immigrants».

GitHub. No repository for this survey.

What is new here, narrowly:

• the quantification of the 2024 rise by segment, with the split into centres and people per centre and its range across orderings;
• the documentation that INE's definition and classification method changed in the record year;
• where the post-publication change sits (one region; all of it immigrant-specialised);
• the arithmetic against the reception system;
• the population mismatch inside the strategy's coverage indicator;
• predictions for the 2026 edition, registered before it.

The observations themselves are HOGAR SÍ's and INE's. The method has been applied in the United States. The conceptual question is the Observatory's.

4. Data and method

Data.

• INE tables of the 2012–2024 editions: 50 files, locked by SHA-256.
• INE's releases, methodologies, questionnaires and IME.
• The 2024 release as first published (Internet Archive, 26 September 2025, 13:25 UTC).
• The Ministry of Inclusion's report of 20 June 2025 on the international-protection system, with annual places of the asylum reception system (SAPI) and the humanitarian programme (PAH), 2015–2025, as republished on La Moncloa.
• Eurostat asylum applicants.
• Arrivals as reported by AIDA from the Ministry of the Interior.

The Ministries of Inclusion and of the Interior answered our requests for robots.txt with 403, so we used no file from their sites. Details are in METHOD.md.

Segments. IMM is «Especializado en inmigrantes». GBV is «Especializado en mujeres víctimas de violencia de género». OTH is «Sin especialización/Otra especialización». The core is GBV + OTH.

The split. For each segment, occupied places O = C · o, where C is the number of centres offering accommodation and o is occupied places per centre.

• The headline is the symmetric split, the Shapley value for two factors: ΔO = ΔC · (o₀ + o₁)/2 + Δo · (C₀ + C₁)/2.
• We also report the two one-at-a-time orderings: centres first, ΔC · o₀ + Δo · C₁, and per centre first, Δo · C₀ + ΔC · o₁.
• What "centres" means. It is centres in INE's tables. The split books an entering or leaving centre at the average size of its segment, so a frame change can land in either term. The centres term is not an upper bound on the frame effect, and the per-centre term is not "existing centres filling up". The 2025 revision is a worked example (§5.4).

5. Results

5.1 The composition of the rise

Mean daily occupied places, adults. Segments add up to the total within INE's rounding.

Immigrant-specialised centres hold 53.2% of occupied places in 2024, against 37.1% in 2022 and 30.3% in 2020. The 2020 share counts all ages.

5.2 Centres and people per centre

Immigrant-specialised centres became both more numerous and larger. How the 10,128 divides between the two depends on the ordering by about 1,000 places either way.

The core.

• Occupied places rose +17.1% and the number of centres offering accommodation +11.9%. People per centre rose +4.7% in two years.
• That average hides a mix effect. Gender-violence shelters grew in number (84 → 100) but held fewer people each (13.8 → 10.4, −24.6%). The other centres rose from 21.19 to 22.87 per centre (+7.9%).
• Split segment by segment and then summed, the core gives 1,603 for more centres and 731 for more people per centre.

Across the variants we computed, the per-centre change in the core runs from +3.3% to +11.9%:

The claim that prompted this study. It compared the core's +17.1% with the +17.1% growth in all centres in the tables (1,175 to 1,376). The numbers are right, but they are unlike quantities:

• all centres include immigrant-specialised ones, which grew from 235 to 360;
• core centres of any kind grew +8.1%;
• core centres offering accommodation grew +11.9%.

Like for like, the core grew faster than its centres.

The long series. From 2012 to 2022, occupied places per centre offering accommodation stayed between 22.0 and 24.3, while centres in the tables grew from 736 to 1,175. In 2024 the figure is 30.5. The core alone moves from 20.0 (2020) to 20.3 (2022) to 21.2 (2024).

5.3 Sensitivity

Only three variants can move the IMM share. The others leave it unchanged by construction:

The previous edition, 2020 → 2022 (+22.0%), was also mostly IMM: 68.2%. That comparison crosses the adults-only rule, and the break is large in IMM: occupancy 99.2% in 2020 (all ages) against 75.9% in 2022.

What the classification itself could do is untested. The variant that matters most cannot be computed from published tables: a different assignment of centres to IMM.

The snapshots. INE does not split them by specialisation. Nationally, occupied places rose +48.6% between the June dates and +65.8% between the December dates.

• In the Canary Islands they went from 1,598 (June 2022) and 1,565 (December 2022) to 2,692 (June
  1. and 5,001 (December 2024).
• In 2024 the islands account for 40.5% of the national gap between June and December. That is the winter peak INE describes.
• Without the Canary Islands, the national rise is +50.7%. The islands' mean went from 1,582 to 3,847, 18.2% of the national change. INE's tables count five immigrant-specialised centres there in both years. INE does not publish occupied places by region and specialisation, so the islands' rise cannot be assigned to segments.

5.4 The change after publication

The release of 26 September 2025 (Internet Archive capture) said:

• «Una media de 33.758 personas mayores de 18 años se alojó diariamente en el año 2024 en centros de atención a personas sin hogar (un 55,7% más que en 2022)»;
• immigrant-specialised centres, «20.911 plazas de alojamiento (un 97,2% más que en 2022), con una media de 17.786 plazas ocupadas (un 121,1% más que en 2022)»;
• 1,375 centres, 1,119 with accommodation.

INE's results page now reads «Datos revisados y modificados con fecha 17/10/2025», and the release «Actualizada el 22 de octubre de 2025».

What changed between the two vintages. The differences include:

• one centre and one centre with accommodation;
• 387 places and 387 occupied places, at both reference dates (June places 37,114 → 37,501; December 41,887 → 42,274);
• meals, staff and the women's share.

All of the place change is in immigrant-specialised centres. We infer this: IMM places and occupied places each rose by 387, and the IMM share of centres from 26.1% (359 of 1,375) to 26.2% (360 of 1,376).

Where it happened. In the release's regional table only Navarre changed: 403 → 790 mean places, occupancy 75.7% → 87.6%. INE shows the six changed cells of that table (national total and Navarre) in red.

What INE says about revisions. The release, in both vintages: «Los datos publicados hoy son definitivos». The IME, §17:

«Esta política general fija los criterios que se deben seguir para los diferentes tipos de revisiones: rutinarias …; revisiones mayores, por cambios metodológicos o de fuentes básicas de referencia de la estadística; y revisiones extraordinarias (por ejemplo, las debidas a un error en estadísticas ya publicadas). No existe revisión de datos.» «Los datos se publican cuando son definitivos, no están sujetos a revisión.»

The change fits the extraordinary type. Our finding is narrow: the published figure changed after release, and INE published no note on what changed or why.

The change moved the headline from +55.7% to +57.5%: 387 places, 3.1% of the national change. We did not find press coverage of it. Accem's statement of November 2025 used the revised figure without comment.

A worked example for the split. The change adds one centre with accommodation and 387 occupied places to IMM: a pure frame change. The symmetric split books 100 of those places as "more centres" and 287 (74%) as "more people per centre". This is why §4 does not read the terms as frame and pressure.

Other inconsistencies:

• INE's chart in the 2024 release gives 16,334 places and 14,038 occupied for 2012; its table 9743 gives 15,638 and 13,991.
• For 2016 places the chart has 19,224 and table 25830 has 19,124.
• The IME gives the 2024 response rate as 79.9%, unit non-response as 20.4%, and «La tasa de no respuesta es 18,2%».

5.5 Against the reception system: arithmetic only

The Ministry of Inclusion's report of June 2025 gives one value a year for places in the asylum reception system (SAPI) and in the humanitarian programme for arrivals by sea and at Ceuta and Melilla (PAH). Only the 2024 values are dated in the text, to 31 December.

AIDA's 2025 update revises the 2024 arrivals to 64,019.

Growth by interval:

In 2020–2022 INE's component moved with SAPI, the period of the displacement from Ukraine. In 2022–2024 it moved with PAH and arrivals. Each matches INE's own explanation for that edition.

Why the ratio is not a coverage rate:

• We cannot show that INE's immigrant-specialised centres are a subset of SAPI and PAH. The category includes seasonal-worker housing, the CETI of Ceuta and Melilla may or may not sit inside PAH, and regional, municipal or private shelters may answer as immigrant-specialised.
• The Ministry's figures are stocks on one day; INE's are a mean of two days.
• With three editions this is description, not a model.

A rising ratio is consistent with the survey reaching more of the reception system. It is equally consistent with the category picking up places outside it.

For scale: the 2024 IMM growth is in places (+100.9%). The headline is in occupied places, which in IMM rose +125.9%.

5.6 The strategy's coverage indicator

The Strategy 2023–2030 defines a «Tasa de cobertura de la red de plazas de alojamiento»:

• baseline «70,7%» (ECPSH 2020);
• «alcanza el 85%» by 2028 and «alcanza el 90% o más» by 2030;
• footnote 35: «Estimación realizada a partir del número total de plazas existentes a 31 de diciembre de 2021 (disponibles en la ECPSH, 2020), sobre el total de personas en situación de sinhogarismo identificadas en la ENPSH 2022».

The 70.7% is the 20,191 places of 15 December 2020 over 28,552 people.

The problem is not that reception residents are counted; it is that they are counted on one side only.

• The numerator is the survey's places, which include immigrant-specialised places: 53.2% of occupied places in 2024.
• The denominator is the 2022 survey of people. Of the people it estimated, «el 0,4% en centros de ayuda al refugiado».
• So the ratio divides capacity that includes reception by a population that almost excludes it.

INE does not split December places by specialisation, so IMM can be removed only from the mean.

On our calculation, the 2022 edition already put the indicator above its 2030 goal. That edition was published on 26 September 2023, eleven weeks after the strategy was approved on 11 July 2023. The Ministry's progress reports do not publish the indicator. Without immigrant-specialised places it rises steadily but stays below the 2028 goal of 85%.

A consistent indicator would put the same population on both sides. One option: places outside immigrant-specialised centres over people outside reception, or reception places over reception residents.

5.7 Regions (exploratory)

INE publishes occupied places by region, and centres by region and specialisation. It does not publish occupied places by region and specialisation.

Across 19 regions:

• the 2022–2024 change in occupied places correlates weakly with the change in IMM centres (Pearson 0.28, Spearman 0.18);
• it correlates more with the change in other centres (0.51, 0.55).

The largest rises:

• Canary Islands, +2,265;
• Catalonia, +1,819, where IMM centres went from 5 to 48;
• Madrid, +1,769;
• Murcia, +1,399.

Counts of centres are a poor proxy for capacity when one centre can hold hundreds of people. We draw no conclusions here.

6. Discussion

What the record is.

• More shelter places were occupied on INE's two reference days of 2024 than in any edition INE charts since 2006.
• Most of the increase is in immigrant-specialised centres: temporary accommodation for asylum seekers, other immigrants and seasonal workers, in proportions INE does not publish.
• Under the definition Spain has adopted, the people in them are houseless. Whether or not one accepts that classification, they are people needing shelter.

What the record is not.

• It is not a count of homeless people. It is a mean of occupied places on two days, so it misses people sleeping rough and turnover between the days.
• It is not evidence that homelessness outside immigrant-specialised centres rose by anything like 57.5%. Outside those centres, occupied places rose +17.1%: more centres in the tables, plus +4.7% per centre (+7.9% in non-specialised centres).

That residual can be real growth. The data cannot separate a centre that opened from one that entered the directory or answered for the first time, and some of the rise per centre can be the same kind of frame change (§5.4).

What INE says, and how the figure was read.

• INE separated the immigrant-specialised figures and named a cause. It did not give their share of the headline.
• It did not mention that its definition moved to ETHOS and that centres now classify themselves.
• It did not explain the post-publication change beyond a notice and six red cells.
• The Ministry's progress report treats the occupied places as the number of homeless people.
• The strategy's coverage indicator, as defined, passes its 2030 goal because its two sides cover different populations.

What would help.

• A headline that gives the immigrant-specialised component next to the total.
• Occupied places by specialisation at each reference date and by region.
• The number of directory units and respondents by specialisation.
• A breakdown of the immigrant category (asylum, humanitarian, seasonal work).
• A dated note on what each correction changed.

INE already holds all of this.

7. Predictions for 2026

PROTOCOL.md was registered by commit 844d5a3 on 4 October 2026. Its §12 adds dated clarifications made after review and before any 2026 data exist. The 2026 edition is expected around late September 2027, an inference from past release dates; INE's 2027 calendar, as updated on 25 September 2026, lists a single release and no ECAPSH date.

The quantities.

• q1: the log change in core occupied places per centre offering accommodation. 2024 baseline: 21.21; the band for reading A is 19.19 to 23.44.
• q1_oth: the same for the OTH segment alone, now co-primary.
• q2: the change in core occupancy. 2024 baseline: 85.9%.

The verdicts.

• Per-centre pressure (reading B, "real growth") means q1 and q1_oth above +0.10 and q2 above +4 points.
• No per-centre pressure means all three inside ±0.10, ±0.10 and ±4.
• Reading A, "frame and composition", is supported only with positive evidence: no per-centre pressure and the immigrant-specialised component moving with the reception system's places (P4, on occupied places and on places). Without P4, the verdict is "no per-centre pressure; frame and growth through new centres not separable".
• Mixed outcomes are "undetermined", stated with the values.
• If INE changes the specialisation item, the scope, the adults-only rule or the reference dates again, the verdict is "not scored".

The rules are code (scripts/score_2026.py). In sample (2022 → 2024, not blind) they give q1 +0.046, q1_oth +0.076 and q2 +1.1 points.

Context. Arrivals fell from 63,970 in 2024 to 36,775 in 2025. SAPI places rose to 34,062 by June 2025 and PAH places fell to 24,532.

8. Limitations

• No unit-level data. The split uses published aggregates. "Centres" are respondents in the tables. Frame changes can sit in either term, and the split between terms depends on the ordering.
• The definition and the classification changed in 2024 (ETHOS; self-declared specialisation with a questionnaire wording that omits "exclusively"). INE does not break the immigrant category down, so we cannot say who is in it.
• The adults-only rule of 2022 breaks comparisons with 2020 and earlier. It also makes the occupancy ratio sensitive to the share of children.
• Two snapshot days a year, with the December one weighted by the Canary Islands' winter peak.
• Reception data are not survey data. We compare directions and a ratio, with three points, and do not claim that one set contains the other.
• No access to the Ministries of Inclusion and the Interior (403 on robots.txt). Reception places come from a Ministry report republished on La Moncloa. Arrivals come from AIDA and are typed from it, not locked.
• The press search ran without a general web search engine. Two leads (Congress, Cáritas) are unread.
• Regions are exploratory.

9. Data, code and licences

• Code (scripts/, tests/): Apache-2.0. Our data and text: CC BY 4.0.
• INE data are reused under INE's terms: «Se prohíbe expresamente desnaturalizar el sentido de la información. Debe citarse la fuente …». Source: INE website, www.ine.es; 2024 tables last modified 17 October 2025.
• Reception places: «Origen de los datos: sitio web de lamoncloa.gob.es. Ministerio de la Presidencia.»
• Asylum applicants: Eurostat.
• data/ holds only aggregates we derived. scripts/run.sh rebuilds them from the 54 locked inputs. scripts/fetch.py downloads those again and checks their SHA-256.

10. Automation and review

AI agents (Claude models) did the work on 4 October 2026:

• retrieved the INE tables and documents, the Ministry report, Eurostat data and the press items;
• wrote the protocol and the code;
• ran the analysis and the prior-work searches;
• drafted the text.

A first agent started the downloads and was stopped. Two helper agents stalled. Their downloads and logs were kept and the work was finished by one agent in sequence.

An independent AI agent then reviewed the study adversarially. It recomputed every number from the raw tables with its own parser and found no arithmetic error. It re-fetched INE's documents and checked the prior work. Its findings, all applied:

• the methodological report had been quoted out of context;
• the 2024 definition change had been missed;
• "largely asylum and humanitarian reception" was unsupported;
• the split depends on the ordering, and its centres term is not an upper bound on frame effects;
• the core average hides a mix effect;
• "part of the reception system" was not shown;
• the coverage argument contradicted the paper's own premise;
• the Observatory's 2025 study was missing;
• the protocol's scoring favoured one reading and had a rule that could not be computed.

Corrections to the protocol are dated in its §12; nothing above it was changed. The study coordinator committed the protocol (844d5a3) and commits the code and frozen values listed in its §12 (D2) before publication. The reviewer has not yet re-read this version.

11. Competing interests

None known. EasyByte Hub S. Coop. Mad., which runs EasyxLab, has no commercial relationship with INE, the Ministry of Social Rights, the Ministry of Inclusion, any organisation that runs shelters or reception centres, or any organisation that campaigns on homelessness or migration.

12. Ethics

• Aggregates only. We used national and regional tables published by INE and national figures published by the Ministry. No unit-level data, no names of centres, no data on individuals. Navarre is named because INE published the changed regional figures; we add nothing about any centre there.
• Polite access. We fetched public pages and files only, with a declared User-Agent, one request at a time and at least 1.2 seconds between requests to a host. We read robots.txt first and did not use sites that refused it.
• No contact. We contacted no one.
• What this study does not say. It does not minimise homelessness, and it does not attribute homelessness, or its rise, to asylum seekers or migrants. People in reception centres are people needing shelter, and under the definition Spain has adopted they are houseless. A record driven by reception capacity is still a record of people needing shelter. The study examines what the statistic measures and how it has been read.

How to cite

EasyxLab (2026). Spain's 2024 record in centres for homeless people: what the +57.5% measures. Study S22. EasyByte Hub S. Coop. Mad. https://github.com/easybytehub/easyxlab

References

Official statistics and documents:

• INE. Encuesta de centros y servicios de atención a las personas sin hogar. Año 2024. Nota de prensa, 26 September 2025, updated 22 October 2025. https://www.ine.es/dyngs/Prensa/es/ECAPSH2024.htm. First version: Internet Archive capture of 26 September 2025, 13:25:24 UTC.
• INE. ECAPSH results, editions 2012–2024 (INEbase tables; ids in scripts/locked_inputs.json); methodologies 2012–2024 and questionnaires 2020–2024 (/daco/daco42/epsh/); Informe metodológico estandarizado, operation 30469; press releases of 26 September 2019, 29 September 2021 and 26 September 2023.
• INE. Encuesta sobre las personas sin hogar 2022: methodology (epshper_22.pdf) and press release of 19 October 2022.
• INE. Calendario 2027, last updated 25 September 2026. https://www.ine.es/dynt3/Calendario/calenHTML.htm?q=2027
• Ministerio de Derechos Sociales, Consumo y Agenda 2030. Estrategia Nacional para la lucha contra el Sinhogarismo en España 2023-2030; Plan Operativo 2023-2024; Informe de Progreso 2023; Informe de progresos 2024 (June 2026). https://www.dsca.gob.es
• Ministerio de Inclusión, Seguridad Social y Migraciones. Informe Sistema de Protección Internacional español, 20 June 2025, and the report of 20 June 2024, published on lamoncloa.gob.es.
• Eurostat. Asylum applicants by type, citizenship, age and sex, annual aggregated data (migr_asyappctza), updated 3 September 2026.
• ECRE. AIDA country reports: Spain, updates 2020–2025. https://asylumineurope.org/reports/country/spain/

Research:

• Pleace, N., Filipovič Hrast, M., O'Sullivan, E., Allen, M., and Benjaminsen, L. (2025). Non-EU Migrant Homelessness. EOH Comparative Studies on Homelessness 14. European Observatory on Homelessness, Brussels, December 2025. https://www.feantsa.org/files/Observatory/Comparative-Studies/14/Feantsa-Studies-14.pdf
• Meyer, B. D., Wyse, A., and Williams, D. (2026). Asylum seekers and the rise in homelessness. Journal of Public Economics, 105577. https://doi.org/10.1016/j.jpubeco.2026.105577 (NBER Working Paper 33655, 2025).
• Davret, J. (2026). The Politics of Counting Homelessness: The Case of Ireland. Housing Policy Debate. https://doi.org/10.1080/10511482.2025.2599125
• Gallén-Granell, E., Botija, M., and Galán Sanantonio, A. (2025). Más que una foto fija: revisión sistemática integrativa sobre la medición del sinhogarismo en España. Revista iberoamericana de estudios de desarrollo. https://doi.org/10.26754/ojs_ried/ijds.11169
• Marbán Gallego, V., and Rodríguez Cabrero, G. (2025). The fight against homelessness and its links with anti-poverty policies in Spain (OpenAlex W7132082154).
• Baptista, I., Benjaminsen, L., Busch-Geertsema, V., and Pleace, N. (2016). Asylum Seekers, Refugees and Homelessness: The Humanitarian Crisis and the Homelessness Sector in Europe. European Observatory on Homelessness.
• Fondation pour le Logement and FEANTSA (2026). Eleventh Overview of Housing Exclusion in Europe.

Press and organisations:

• HOGAR SÍ, «El sistema de atención al sinhogarismo sigue sin dar una respuesta eficaz: datos sobredimensionados, pocas plazas en vivienda y mismas carencias estructurales», 26 September 2025. https://www.hogarsi.org/encuesta-ine-centros-2024/
• RTVE, «El número de personas sin hogar atendidas aumenta un 55% y una de cada cuatro son mujeres», 26 September 2025.
• Europa Press, «Una media de 33.758 personas al día durmieron en centros para personas sin hogar en 2024, un 55% más que en 2022», 26 September 2025.
• Red FACIAM, «El sinhogarismo en España crece con especial impacto en la población migrante y en las mujeres», 26 September 2025.
• Europa Press, «Accem alerta de un incremento del 'sinhogarismo': En 10 meses se atendieron el mismo número de personas que en 2024», 24 November 2025.

Cite this study

EasyxLab (2026). Spain's 2024 record in centres for homeless people: what the +57.5% measures. Study S22. EasyByte Hub S. Coop. Mad. https://github.com/easybytehub/easyxlab/tree/main/studies/s22-homeless-shelter-record

@techreport{easyxlab_s22,
  title       = {Spain's 2024 record in centres for homeless people: what the +57.5% measures},
  author      = {{EasyxLab}},
  institution = {EasyByte Hub S. Coop. Mad.},
  number      = {S22},
  year        = {2026},
  url         = {https://github.com/easybytehub/easyxlab/tree/main/studies/s22-homeless-shelter-record}
}