Did purchases by foreign non-residents fall after Spain ended its golden visa? Evidence from MIVAU's quarterly transactions, 2007–2026

Abstract

The rule. Spain's residence permits for investors ended on 3 April 2025, when Ley Orgánica 1/2025 emptied articles 63–67 of Ley 14/2013. The permits were known as the golden visa, and their most used route was buying property worth €500,000 or more. Applications filed before that date are decided under the old rules, and permits already granted keep their validity. The Prime Minister announced the end on 8 April 2024.

The data. The Ministry of Housing (MIVAU) publishes an open quarterly count of home purchases by the buyer's residence and nationality, for 52 provinces, from 2007Q1 to 2026Q2. It is built from notarial deeds. We compared three years of four quarters each:

• the year before the announcement, B (2023Q2–2024Q1);
• the year from announcement to repeal, A (2024Q2–2025Q1);
• the year after the repeal, P (2025Q2–2026Q1).

Answer. The end of the golden visa coincided with a fall in foreign non-resident purchases in Madrid and Barcelona. The tests fixed in advance do not support attributing it to the repeal.

The tests fixed in advance.

• Spain. Foreign non-resident purchases fell 9.8%.
• The six provinces where the Government placed 90% of the permits fell 9.3%, against 11.0% in the other 46. Across 19 units the contrast is null (+0.02 log points, permutation p = 0.84).
• The planned exposure, a province's pre-2024 share of foreign non-resident buyers, has the wrong sign. Provinces with more of them fell less (p = 0.13).

Madrid and Barcelona, exploratory. They were chosen as the treated group after a pilot had shown that they fell most, so the contrast below describes the data; it does not test a hypothesis.

• Size. Their foreign non-resident purchases fell 35.9% (2,651 to 1,699), against 8.5% in the rest of Spain: −30.0% relative. Valencia, also one of the six, fell 26.9%.
• Only these buyers. Other buyer groups in the two cities moved like the rest of Spain.
• Against history. The planned placebo-date p-value is 0.051, and 0.125 over spaced dates. Non-overlapping windows allow no p below about 0.17.
• The baseline year was unusually high. Madrid's 614 purchases in P match its 2019 count (609). Against 2018–2019 the two cities' share of national foreign non-resident purchases is 22% lower, not 29%.

The timing is sharp but has rivals. The fall came in 2025Q2, the first quarter after the repeal. Three things limit what that shows:

• Currency. 2025Q2 was also the quarter of the euro's largest rise of the 2007–2026 sample against the dollar (+7.7%) and the yuan (+7.1%).
• Deadlines. A drop just after a deadline cannot tell the end of a rush from a fall in demand. Portugal's 1 January 2022 deadline produced the same kind of break, followed by growth.
• The tax announcement. On 13 January 2025 the Government announced a tax on non-EU non-resident buyers. It was never enacted, but it reached the same buyers one quarter earlier. It cannot be separated from the repeal.

What is new. The press reported the rush and the fall in 2026 from notarial and registry data. What this study adds is an open, reproducible province-level analysis and a pre-registered prediction for MIVAU's release of 16 December 2026.

1. Introduction

Golden visa schemes exchange a residence permit for an investment. In southern Europe the investment was mostly housing, which made the schemes a target of housing policy. Spain announced the end of its scheme in April 2024 and ended it in April 2025. Portugal first confined its real-estate route to the interior and the islands (from 1 January 2022) and then closed it (October 2023).

The academic literature has measured what such schemes did when they were introduced:

• in Spain (Foremny, Li, Martínez-Toledano and Segú 2025);
• in Portugal (Pereira dos Santos and Strohmaier 2024);
• across the EU (Surak and Tsuzuki 2021).

The press has reported, from notarial and registry figures, a rush of non-EU buyers before Spain's deadline and a fall in non-resident purchases after it (§3). Those reports use national aggregates from sources whose terms restrict reuse.

This study. It asks the same questions with the one open, quarterly, province-level source that classifies Spanish home purchases by buyer residence and nationality: MIVAU's transactions statistics. It asks:

• did foreign non-residents' purchases fall, and where;
• was there a rush first;
• how much of the change can be separated from the other things that changed for the same buyers at the same time.

The most important of those other things is a tax on non-EU non-resident buyers, announced three months before the repeal and never enacted.

2. Legal frame (quoted as in force on 4 October 2026)

2.1 The repeal

Ley Orgánica 1/2025, de 2 de enero, de medidas en materia de eficiencia del Servicio Público de Justicia (BOE no. 3, 3 January 2025), «Disposición final vigesimoprimera»:

• «Uno. Se dejan sin contenido los artículos 63, 64, 65, 66 y 67.»
• «Dos. Se introduce una nueva disposición transitoria primera … Aquellos inversores o familiares de inversores que, con anterioridad a la fecha de entrada en vigor de esta disposición transitoria, hubieran presentado la correspondiente solicitud, podrán recibir el visado o autorización correspondiente conforme a la normativa vigente en la fecha de presentación de la solicitud.»
• «Tres. Se introduce una nueva disposición transitoria segunda … Los visados y autorizaciones para inversores que tengan validez a la fecha de la entrada en vigor de esta disposición transitoria, conservarán dicha validez durante el tiempo para el que hubieran sido expedidos. En el caso de presentarse solicitudes de renovación, se tramitarán y resolverán conforme a la normativa vigente en la fecha de concesión de la autorización inicial.»

The two transitional provisions are therefore provisions of Ley 14/2013, not of LO 1/2025.

Entry into force. Final provision 38.1: «La presente ley entrará en vigor a los tres meses de su publicación en el Boletín oficial del Estado», that is, on 3 April 2025. The BOE's consolidated text of Ley 14/2013 confirms it for article 63: «Se deja sin contenido, con efectos de 3 de abril de 2025, por la disposición final 21.1 de la Ley Orgánica 1/2025».

Later references. The BOE lists none that touches final provision 21; the correction of errors of 11 January 2025 concerns other provisions.

2.2 What was repealed

In the wording in force before the repeal (2015), article 63.2.b granted the permit for «la adquisición de bienes inmuebles en España con una inversión de valor igual o superior a 500.000 euros por cada solicitante». Article 64 adds two details.

• The €500,000 was per applicant and free of charges. It could be spread over several properties, so a single purchase under €500,000 could qualify.
• A deed was not needed to apply. With a notarised pre-contract and a blocked deposit, an applicant got a six-month visa: «Si el extranjero no ha formalizado la compra del inmueble o inmuebles pero existe un precontrato con garantía en su cumplimiento por medio de arras …». Some golden-visa deeds can therefore fall after 3 April 2025.

2.3 The announcement

La Moncloa, 8 April 2024, from Dos Hermanas (Seville): «Vamos a eliminar la concesión de la Golden Visa que permite acceder al régimen de residencia cuando se invierte más de 500.000 euros en bienes inmuebles».

The next day the Council of Ministers took note of a report (reference of 9 April 2024):

• «Entre 2013-2023, se han concedido en España 14.576 visados, vinculados a las inversiones en inmuebles. 2020: 632 · 2021: 997 · 2022: 2.017 · 2023: 3.273 · 2024 (hasta febrero): 426»;
• «Principales nacionalidades: China, Rusia, Reino Unido, EE.UU, Ucrania, Irán, Venezuela y México»;
• «Principales provincias de inversión, que concentran el 90% de las autorizaciones: Barcelona, Madrid, Málaga, Alicante, Baleares y Valencia»;
• «Las adquisicione s de viviendas ligadas a inversiones previstas en Ley 14/2013 han llegado a representar hasta el 7,1% en Marbella y el 5,3%, en Barcelona, del total de las compraventas de vivienda anuales.»

The route to law ran through an amendment to the justice bill. According to idealista/news, the Government tabled it in July 2024, the Congress approved the bill in November 2024, and the final vote came in December 2024.

2.4 The tax that was announced and not enacted

La Moncloa, 13 January 2025: «se va a limitar la compra de viviendas por parte de extranjeros extracomunitarios no residentes. En concreto, se incrementará la carga fiscal que tendrán que pagar en caso de compra hasta el 100% del valor del inmueble, en línea con países como Dinamarca o Canadá.»

On 4 October 2026 no such rule is in the BOE. We checked three texts:

• the consolidated ITP-AJD law (last updated 21 March 2026), whose only mentions of non-residents concern the representative's tax domicile;
• the Government's housing decree-law of 29 September 2026 (Real Decreto-ley 26/2026, BOE 30 September 2026);
• Real Decreto-ley 27/2026.

RDL 26/2026 limits purchases by entities buying below 70% of appraisal value until 2028. It also allows municipal surcharges on the property tax of tourist flats, and changes the SOCIMI regime. It contains no tax on non-EU buyers.

We could not check whether a bill was tabled: the Congress's site answers 403 to our User-Agent. A tabled bill would place the threat in 2025, which matters for timing.

2.5 Regional purchase taxes

• Madrid. The transfer tax on dwellings is 6% (Decreto Legislativo 1/2010, art. 28, «se aplicará el tipo del 6 por 100»), unchanged since 1 January 2014.
• Catalonia. Decreto-ley 5/2025, de 25 de marzo, was published on 26 March 2025 and changed article 641-1 of the Catalan tax code from 27 June 2025.
  • Before: a scale of 10% up to €1,000,000 and 11% above.
  • After: 10% up to €600,000, 11% to €900,000, 12% to €1,500,000 and 13% above.
  • Large holders and buyers of whole buildings pay 20%.
  • The transfer tax applies to second-hand homes; new homes pay VAT and the stamp duty, which did not change.
  • The rise was known from late March 2025. One would expect buyers of expensive second-hand homes in Barcelona to bring purchases forward into 2025Q2, which works against a fall in that quarter.

2.6 Portugal

Two Portuguese rules ended the real-estate route of the residence permit for investment (ARI). We read both in the consolidated texts of the Procuradoria-Geral Distrital de Lisboa.

• Decreto-Lei n.º 14/2021, de 12 de fevereiro. It added to article 3 of Lei 23/2007: «4 - Os imóveis adquiridos nos termos previstos nas subalíneas iii) e iv) da alínea d) do n.º 1 que se destinem a habitação, apenas permitem o acesso ao presente regime caso se situem nas Regiões Autónomas dos Açores e da Madeira ou nos territórios do interior, identificados no anexo à Portaria n.º 208/2017, de 13 de julho.»
  • Article 4: «O presente decreto-lei entra em vigor a 1 de janeiro de 2022.»
  • It was approved by the Council of Ministers on 22 December 2020, under an authorisation in the 2020 budget law.
  • From 1 January 2022, buying a home in Lisbon, Porto or on the coast no longer gave the permit. That is the Portuguese analogue of Spain's deadline for the big cities.
• Lei n.º 56/2023, de 6 de outubro («Mais Habitação»).
  • Article 42.1: «Não são admitidos novos pedidos de autorização de residência para atividade de investimento, concedidos ao abrigo do disposto nas subalíneas i), iii) e iv) da alínea d) do n.º 1 do artigo 3.º da Lei n.º 23/2007». That closed what remained of the real-estate route, and the capital-transfer route too.
  • Article 55: «A presente lei entra em vigor no dia seguinte ao da sua publicação», that is, 7 October 2023.
  • The Council of Ministers communiqué of 16 February 2023 approved the «Mais Habitação» plan but does not mention the permits, so we date no announcement from it.

3. Prior work

We searched on 4 October 2026 and logged every query in data/prior_work_search.csv. Open web search was unavailable to us (session budget exhausted). Our search went through:

• OpenAlex and Crossref;
• the Banco de España's sitemaps;
• EUR-Lex;
• one news outlet's topic page.

El País, elEconomista and Transparency International answered 403. The press items below were found by the independent reviewer of this study through one search-engine query. We know them only from the search snippets: El Periódico answers 406 to automated access, La Razón's robots.txt reserves its content against AI crawlers, the Notariado's carries «Content-Signal: ai-train=no, search=yes, ai-input=no», and El País blocks our User-Agent. We cite them, but have not read them.

The press has measured the end of the Spanish scheme.

• El País, 23 May 2026. «Los extranjeros aceleraron la compra de casas de más de medio millón de euros en los últimos meses de vida de la golden visa»; snippet: «La urgencia de los no comunitarios por el fin anunciado de las golden …». A data piece on the rush above €500,000 by non-EU buyers.
• El Periódico, 9 April 2026. «Las adquisiciones de vivienda por parte de extranjeros residentes en España crecieron un 5,3% en 2025, mientras que las de no residentes retrocedieron un 9,4%, tras la derogación de la 'golden visa'».
  • MIVAU's count of foreign non-resident purchases fell 9.5% from 2024 to 2025 (56,777 to 51,367), consistent with it.
• La Razón, 9 April 2026. «La compra de viviendas por extranjeros no comunitarios se contrae … tras la eliminación de la golden visa». A post by the newspaper cites the Notariado's statistics: non-resident buyers −15.1% in the second half of 2025.
• Colegio de Registradores, Estadística Registral Inmobiliaria, 1T-2025, press release. «El 54,6% de los adquirentes extranjeros de vivienda pertenecen a la UE»: a quarterly EU/non-EU split of foreign buyers.

Academic work measured the introduction of the schemes.

• Spain, 2013 (Foremny, Li, Martínez-Toledano and Segú 2025). With the universe of Spanish real-estate transactions, the number above the €500,000 threshold «increased by 43% more for non-EU relative to EU investors after the introduction of the programme». Non-EU investors paid a premium of €11,521 around the threshold, and prices rose 0.19% with average exposure.
• Portugal, 2012 (Pereira dos Santos and Strohmaier 2024). They find bunching at €500,000 and a «Golden Visa Premium» of about €38,000 at the threshold.
• The EU (Surak 2020; Surak and Tsuzuki 2021). Uptake compared across schemes. The real-estate impact is «trivial, with the sole exception of Greece» at national level.

On Spanish foreign demand, descriptive.

• Banco de España, Boletín Económico 2/2020. Foreign investment in housing, 2007–2019.
• Banco de España, Documento Ocasional 2433 (Lajer Baron, López Rodríguez and San Juan 2024).
  • Foreign non-residents paid in 2023 a price per square metre 70% above Spanish residents.
  • About 65% of their purchases were in the Comunitat Valenciana and Andalucía.

On the repeal.

• Legal analyses. Antràs Porta (2025, Esade Law Review), who treats the repeal together with the announced tax, and Galán Azofra (2025).
• A teaching case. SAGE (2024).
• Housing Studies (Gaspar and He 2026). On money transfers for housing investment by golden-visa holders. We read only its metadata.

What this study adds is narrow.

• An open, reproducible analysis. By province, from data under an open licence, with placebo buyer groups and placebo dates. Every number can be rebuilt by anyone with one command.
• A prediction registered before the next release.

It does not add the finding that there was a rush and a fall, which the press reported first. It cannot add an EU/non-EU split, which the notarial and registry sources have and the open data do not.

4. Data and method

4.1 The MIVAU statistics

MIVAU's «Estadística de Transacciones Inmobiliarias» (PEN 9198) counts the dwellings sold by public deed before a notary. Its source is «Documentación administrativa del Colegio General del Notariado» (INE inventory of statistical operations, no. 25003).

• Table 1.6. Purchases by the buyer's residence (resident, non-resident) and nationality (Spanish, foreign, not stated), for every province and quarter since 2007. We read the release of 1 October 2026. 2026Q2 is provisional.
• Table 1.7. Free-market purchases by value band (bands of €150,000, all buyers). It covers one first quarter per release. We have 2021Q1, 2022Q1 and 2025Q1 from the Internet Archive and 2026Q1 live.

The statistic does not give the buyer's country, so EU and non-EU buyers cannot be separated. It does not mark golden-visa purchases. The licence is CC BY 4.0.

4.2 Design, and what was fixed when

The plan was frozen on 4 October 2026, before any of its measures were computed (METHOD.md §1; its SHA-256 is in data/freeze.json). By then a scout's pilot had computed the four-quarter sums for seven territories. They showed that Madrid (−54% from the rush) and Barcelona (−45%) fell most. That knowledge splits the tests in two.

• Fixed independently of the pilot. Neither of these tests singles out a province by its outcome.
  • The exposure the study brief required: the pre-period share of foreign non-residents in all purchases.
  • The six provinces the Council of Ministers named before the repeal.
  • The share of homes over €600,000.
  • The out-of-sample prediction.
• Chosen after the pilot. The treated group «Madrid and Barcelona». Every test built on it is exploratory, because the group was picked because it fell most:
  • its window contrasts and placebo dates;
  • its triple difference;
  • its binary exposure.

The rest of the design:

• Windows. B, A and P, four quarters each, so seasonality cancels.
• Measures. The rush (A/B), the fall net of the rush (P/B, primary) and the fall from the rush (P/A).
• Placebo groups. Within each province: foreign residents, Spanish non-residents and Spanish residents.
• Placebo dates. 58 earlier announcement quarters, 2008Q1–2022Q2.
• Across provinces, 19 units. The 18 provinces with at least 100 foreign non-resident purchases in B, and the other 34 pooled.
  • Event study.
  • Window differences-in-differences.
  • Inference by permuting the exposure across units.
• Added after the freeze, and labelled as such (METHOD.md §9):
  • the history of the two cities' share;
  • placebo dates without pandemic windows;
  • the one-quarter break;
  • Portugal;
  • after the independent review: non-overlapping placebo dates, the euro's quarterly moves, differenced currency models, and Portugal's 2022 deadline.

5. Results

5.1 The tests fixed in advance

None of the planned geographic tests supports a repeal effect.

• The six provinces named by the Government fell 9.3%, a little less than the other 46 (−11.0%). In the primary, weighted specification their contrast is +0.02 log points (p = 0.84).
  • The unweighted one is −0.15 (p = 0.025). It gives each province the same weight, so Madrid, Barcelona and Valencia count as much as Alicante or Málaga.
  • The six did rise more in the anticipation year (+9.7% against +1.1%).
• The share exposure has the wrong sign. Provinces where foreign non-residents are a large share of buyers fell less: Alicante (a third of buyers), Málaga and Baleares (a quarter). Madrid (0.8%) and Barcelona (2.0%) fell more. In the event study the share exposure's mean post coefficient is also positive (+0.53).
• The share of homes over €600,000 does not predict the fall. Baleares and Málaga have the most such homes and fell little.

5.2 Madrid and Barcelona (exploratory)

What the numbers show.

• The fall. Foreign non-residents bought 35.9% fewer homes in the two cities in P than in B. The rest of Spain fell 8.5%. Relative to the rest the fall is −30.0%: Madrid −30.4%, Barcelona −29.7%.
• In purchases. Had the two cities followed the rest, they would have had about 727 more foreign non-resident purchases in P.
• Valencia, the third city among the six, fell 26.9%, without a prior rush.
• The cross-sectional tests built on this group are not tests. The two cities are the two largest falls among the 19 units (binary-exposure p = 0.0058, the smallest attainable). The triple difference against foreign residents is −0.51 (p = 0.021). Both test a group chosen because it fell most. They are reported as descriptions, not as evidence.

Other buyer groups in the two cities moved roughly like the rest of Spain. The fall is in foreign non-residents, the group that could seek the permit. Two qualifications:

• Madrid's foreign residents. They had their own smaller rise and fall: +17.3% from B to A, then −17.1% from A to P (−16.7% relative to the rest). Pooling with Barcelona's foreign residents (+15.6% relative from B to P) hides it.
• Spanish non-residents are not a clean placebo. Ley 20/2022, additional provision 8, opened Spanish nationality to people «nacidos fuera de España de padre o madre, abuelo o abuela, que originariamente hubieran sido españoles» and lost it through exile, among others. Buyers who used it would move from «foreign» to «Spanish» non-resident.
  • In the two cities Spanish non-residents rose by 111 purchases, small against the fall of 952 foreign non-resident purchases.
  • We did not quantify it.

5.3 Against history

The windows are 12 quarters long, so consecutive placebo dates share most of their data.

• Non-overlapping windows. There are only five fully non-overlapping placebo dates (2010Q2, 2013Q2, 2016Q2, 2019Q2 and 2022Q2), with relative falls of +5.3%, +38.5%, −9.4%, −19.0% and +60.2%. The real −30.0% is below all five, which is the smallest p attainable (1/6 ≈ 0.17).
• The pool is not stationary. In 2007–2011 Madrid had 8–50 foreign non-resident purchases a quarter, against about 250 in 2023. Early placebo dates rest on very small counts.
• The pandemic exclusion was decided after seeing that the five most negative placebo values come from windows covering 2020–2021. Its p = 0.024 is post hoc.

The planned evidence against history is p = 0.051. It is weaker on any reading that respects the overlap.

The relative rush (+17.7%) is unremarkable against the same dates (p = 0.34; 0.31 over the spaced subset).

5.4 The baseline year

The size of the fall depends on the reference year.

• Against 2018–2019. The two cities' share in P is 22% lower (Madrid −13%, Barcelona −26%). Against B it is 29% lower.
• Madrid. Its B share (1.78%) was the highest it had ever had. Its 614 purchases in P are about its 2019 count (609). Most of Madrid's fall is a return from an unusually high 2023–2024.
• Barcelona. Its B share was close to its 2019 share (3.15%), and its P share is below every year from 2014 to 2019. Barcelona's fall does not depend on the reference year in the same way.
• The event study (two cities, weighted).
  • Their deviations from B were −0.28 to −0.53 log points in 2020Q3–2022Q4.
  • The post-repeal coefficients are −0.22, −0.41, −0.45 and −0.35 in 2025Q2–2026Q1.
  • The post-repeal level is inside the range of the pre-period deviations, so parallel trends do not hold.
• The rise into B coincides with the rise in real-estate golden visas, from 997 in 2021 to 2,017 in 2022 and 3,273 in 2023. If that rise was golden-visa demand, the fall measures its end. If it was a post-pandemic recovery, the fall overstates the repeal.

5.5 Timing, and what a sharp break can and cannot show

**Quarter by quarter (data/core_quarterly.csv).**

• Madrid. About 250 foreign non-resident purchases a quarter in 2023. Then 325–342 a quarter from 2024Q2 to 2025Q1, 176 in 2025Q2, and 135–154 since.
• Barcelona. 376–505 a quarter from 2023 to 2025Q1, then 310 in 2025Q2, then 247–270.

The change from 2025Q1 to 2025Q2. The two cities' level relative to the rest fell −0.558 log points (−42.7%). That is the largest first-to-second-quarter fall in 2007–2025, and the largest of the 76 quarter-to-quarter changes since 2007. This test was added after the freeze.

Two other things happened in that quarter, or explain such a drop:

1. The euro. From 2025Q1 to 2025Q2 it rose 7.7% against the dollar (1.0523 to 1.1338) and 7.1% against the yuan (7.655 to 8.197). These are its largest quarterly rises of the 2007–2026 sample for both currencies. The next largest are +6.6% (USD, 2017Q3) and +5.2% (CNY, 2022Q4). Buyers paying in dollars or yuan saw Spanish homes become about 7% dearer in one quarter.
2. The end of a rush. A deadline makes buyers bring purchases forward and then stop. A sharp drop just after it is what a pure rush would produce, even with no lasting fall. The break cannot tell «the rush ended» from «demand fell».
   • Portugal shows the pattern (§5.8): a spike before its 1 January 2022 deadline, a break of the same size, then growth.
   • The lasting part of the Spanish change is the P-against-B contrast (§5.3–§5.4), not the break.

5.6 Other things that changed

The general market did not fall.

• In the two cities, total purchases rose 14.1% from B to P, against 15.6% elsewhere.
• Purchases by Spanish residents rose 15.6% against 20.1%.

Regional taxes.

• Madrid did not change its 6% rate.
• Catalonia's new scale applies from 27 June 2025, to second-hand homes over €600,000 bought by anyone. Barcelona's fall began in 2025Q2, before it, and in a quarter when expensive purchases should, if anything, have been brought forward.
• The rise may have weighed on Barcelona's high-value purchases from 2025Q3. We cannot measure that by residence.

Currency: it cannot be bounded.

• From B to P the euro rose 6.9% against the dollar, 5.8% against the yuan and 0.2% against sterling.
• In levels. We regressed the two cities' relative log level on an equal-weighted index of the three rates, 2010Q1–2024Q1, with quarter-of-year dummies.
  • The elasticity is −2.57, implying a −10.1% relative change from B to P.
  • The residuals are strongly autocorrelated (Durbin–Watson 0.49), the sign of a near-spurious regression between two trending series. Its standard error is invalid.
• In differences. The same relation in four-quarter differences gives +0.07 (SE 0.45), and in one-quarter differences with seasonal dummies +0.33 (SE 0.82). Both imply roughly no currency effect (+0.3% and +1.4%).
• The data cannot say whether the currency explains none, some or much of the fall.
• The coast is no control for the currency. Its non-EU buyers are probably mostly British, and sterling was flat. A dollar or yuan shock would hit hardest wherever dollar- and yuan-linked buyers concentrate. The Council of Ministers' list of golden-visa nationalities (China, the United States, Venezuela, Mexico…) suggests that is the two cities, but we have no data by nationality and province.

Price segment. We compared all free-market purchases in 2025Q1 (anticipation) and 2026Q1 (post).

• Purchases of €600,000 or more. Madrid −0.4% (2,909 to 2,897), Barcelona +1.4%, the rest of Spain +6.8%.
• Purchases under €450,000. Madrid −11.2%, Barcelona −8.0%, the rest −7.2%.
• Madrid, within the territory. The high-value segment did 10.8 points better than the low one, against 14.0 points in the rest. That is a gap of about 3 points, about 90 purchases. Madrid lost 200 foreign non-resident purchases between those quarters (335 to 135), so at most part of the loss shows up in the high-value segment.
• Barcelona. The loss was 244 (502 to 258). Had those been €600,000 or more, its high-value count (936) would have fallen about 26%; it rose.
• These are single quarters, for all buyers, with no 2023 or 2024 first quarter to compare. They neither support nor rule out a golden-visa channel.

The 100% tax announcement.

• It was made on 13 January 2025, in the quarter before the repeal, and aimed at the same non-EU non-residents. Quarterly data cannot separate a reaction to it from a reaction to the repeal.
• Foreign non-resident purchases in the two cities did not fall in 2025Q1, the quarter of the announcement: 335 in Madrid and 502 in Barcelona, against 342 and 482 in 2024Q4.
• The geography does not discriminate either. The two cities' foreign non-resident buyers are probably more often non-EU than the coast's: the coast's are mostly European, and the golden-visa nationalities were mostly non-EU. Any non-EU-specific shock would then produce a larger percentage fall in the two cities, with or without a golden-visa channel. That includes the tax threat and the dollar. We cannot check it without nationality data.

Other dated measures that reached investors in the two cities in 2024–2025, and that we did not verify against official texts:

• Barcelona's plan not to renew tourist-flat licences by 2028 (announced June 2024), and court rulings on Catalonia's tourist-flat rules;
• Catalonia's 2025 regulation of seasonal lets;
• Madrid city's «Plan Reside» on tourist flats (2025);
• the national register of short-term lets (Real Decreto 1312/2024).

All of them apply to residents as well, who did not fall.

5.7 EU and non-EU buyers

MIVAU does not separate them, and neither does the INE's ETDP. The Notariado and the Colegio de Registradores publish national EU/non-EU aggregates in press releases (§3). Their sites restrict reuse or block us, and our access rules kept us from using them.

The only open source with that split, for a comparable rule, is Portugal's.

5.8 Portugal

Statistics Portugal publishes transactions by the buyer's tax domicile at national level, from 2019Q1: Portugal, the EU, and «Restantes países». The last includes Portuguese emigrants and, since Brexit, British buyers. Regional cells are not available.

The 1 January 2022 deadline. That day DL 14/2021 removed Lisbon, Porto and the coast from the residential route.

• The rush. In 2021Q4 non-EU-domiciled purchases spiked to 1,845 from 1,142. Against EU buyers, that was a rise of 0.38 log points.
• The break. In 2022Q1 they fell back to 1,121: −0.49 log points against EU buyers. That is the largest fall in the 29 quarter-to-quarter changes since 2019, larger than the pandemic's −0.44 in 2020Q3, and about the size of Spain's 2025Q2 break (−0.56).
• Then they grew. Over 2022 as a whole, non-EU purchases were 12.5% above 2021 (EU purchases 27.5% above, so −11.8% relative), and they rose again in 2023.
• The Portuguese break was mostly a rush ending, followed by a return to trend. A break of the same size in Spain is therefore not, by itself, evidence of a lasting fall.

The October 2023 law closed what remained of the route. In the year after it (2023Q4–2024Q3), non-EU-domiciled purchases were 1.2% above 2022, and EU-domiciled ones 16.7% below. By then the route no longer reached Lisbon or Porto, so this second event is a weak comparison.

5.9 Size against the permits

• The Council of Ministers' count is 3,273 real-estate visas in 2023, 90% of them in six provinces. The source says «visados … vinculados a las inversiones en inmuebles». That the count includes family members is our inference, not stated in the text.
• The two cities' deficit is 727 foreign non-resident purchases a year in P, and the excess was 497 in A.
• Order of magnitude. If golden-visa buyers came with one or two family members on average, a few hundred permit-driven purchases a year in the two cities would match these numbers. We cannot check it.
• A parliamentary answer, as quoted by Europa Press (idealista/news, 25 November 2024): 780 visas in January–October 2024, with an average investment of €657,204, 573 of them after the announcement. It is not comparable with the 2023 figure, because consular visas and authorisations granted in Spain are different counts.

5.10 Pre-registration

MIVAU publishes 2026Q3 on 16 December 2026 (calendar of the Programa anual 2026 of the Plan Estadístico Nacional).

• What is predicted. PREDICTIONS.md gives each unit's share of Spain's foreign non-resident purchases in that quarter, from a model that assumes the post-repeal level persists.
  • Madrid: 1.25% (0.98–1.65%, 90% interval).
  • Barcelona: 2.13% (1.51–2.92%).
• The deviations of 4 October 2026. They were added after the independent review and before the release.
  • The rebound test was respecified (above the upper 90% bound).
  • The near-certain test 2b was replaced with a further-fall test (below the lower bound).
  • The counts «test», which had no criterion, was removed.
  • The scored values were frozen in data/predictions_2026Q3_frozen.csv, whose SHA-256 is in data/freeze.json with the plan's.

6. Discussion

What the data show. Purchases by foreign non-residents fell after Spain's golden visa ended. They fell 9.8% in Spain, and by about a third in Madrid and Barcelona. The fall in the two cities:

• came in the quarter after the repeal, after a rise that began in the quarter of the announcement;
• is absent from other buyer groups in the same cities;
• is shared, in part, by Valencia.

What the tests fixed in advance say. They do not support attributing the fall to the repeal.

• The six provinces where the permits concentrated did not fall more than the rest.
• The exposure the study was asked to use has the wrong sign.
• The contrast that does stand out, Madrid and Barcelona, was chosen after seeing that it stood out.

What else fits the two cities' fall.

1. The tax announcement of 13 January 2025 aimed at the same buyers, and cannot be separated.
2. The euro's record rise against the dollar and the yuan came in the very quarter of the fall. It would bite hardest where dollar-linked buyers concentrate, probably these two cities.
3. The end of a rush. A deadline produces a sharp drop with no lasting effect, as Portugal's did in 2022.
4. The baseline. The year before the announcement was unusually high, especially for Madrid, whose post-repeal count equals its 2019 count.

What the study does not say. It does not say the repeal had no effect: a golden-visa channel would produce the same pattern in the two cities. It says that the open data cannot tell that channel apart from the others.

The pre-registered prediction tests one thing the data can still decide: whether the two cities' new level persists, rebounds or falls further. The press, with notarial and registry microdata that include the buyer's nationality, is better placed to separate non-EU from EU buyers.

7. Limitations

• Selection. The treated group was chosen after a pilot. Only the six-province, share and high-value exposures and the prediction are independent of it.
• Outcome. Foreign non-residents include EU citizens, who never needed the permit. Residence and nationality are as recorded in the deed.
• Baseline. B was a high year (§5.4), and parallel trends do not hold in 2020–2022.
• Inference.
  • Placebo windows overlap; non-overlapping windows allow no p below about 0.17.
  • The placebo pool's early years rest on very small counts.
  • With 19 units, clustered standard errors are unreliable, and we do not rely on them.
• Currency. It cannot be bounded (§5.6). The levels regression is near-spurious.
• The tax announcement cannot be separated from the repeal. Whether a bill was tabled is unknown to us.
• Price segment. All buyers, single first quarters.
• Permits. No official count by province or year after February 2024 could be read. Inclusión answers 403, and we did not bypass it.
• Portugal. National level only; tax domicile, not nationality.
• Other local measures in Madrid and Barcelona were listed but not verified (§5.6).
• Prior work. No open web search was possible. The press items in §3 are known from search snippets, not read.
• Provisional data. 2026Q2 is provisional. Earlier releases show small revisions, mostly in the last quarter.
• Post hoc analyses. Several analyses were added after the freeze, and two after the review. They are labelled. Only the 2026Q3 prediction is blind.

8. Data, code and licences

• Code: Apache-2.0. Our data and text: CC BY 4.0.
• data/ holds only aggregates: counts by province, quarter and buyer group, and every measure.
• Sources and credits:
  • MIVAU, «Transacciones inmobiliarias», CC BY 4.0;
  • European Central Bank, reference exchange rates;
  • Statistics Portugal (INE), indicators 0012785 and 0012786;
  • legal texts from the BOE, La Moncloa, the Government of Portugal and the Procuradoria-Geral Distrital de Lisboa.
• Everything runs with Python and xlrd:
  • bash scripts/run.sh fetch, then bash scripts/run.sh;
  • after the December release, bash scripts/run.sh fetch && bash scripts/run.sh evaluate.
• Code and data: https://github.com/easybytehub/easyxlab (studies/s21-golden-visa-end).

9. Automation and review

This study was carried out by AI agents (Claude, Anthropic) under the direction of EasyByte's researchers.

How the work was done.

• Earlier work. A scout agent proposed the question and the pilot. A first study agent downloaded the MIVAU workbooks, wrote the parser and verified the legal texts. It was stopped.
• The study agent that finished the work:
  • recomputed the pilot;
  • found the official release date;
  • froze and ran the plan;
  • wrote the code, tests and text.
• Three research sub-agents were started and then stopped by a session watchdog. Their partial downloads were reused only after the study agent re-read them.
• Manual steps. The analysis has none except reading the release calendar, an image-based PDF, by OCR and on screen.
• What the check covers. scripts/check_headlines.py recomputes the abstract's and the main tables' numbers from data/, and the two freeze hashes. Other figures in the text were rechecked by hand by the reviewer.

The independent review. An independent agent reviewed the study adversarially on 4 October

1. It re-parsed the raw workbook with its own code, re-ran the package, and read the legal and

Portuguese texts. It found no arithmetic or date error, and eight major problems:

1. the treated group was chosen after the pilot;
2. the placebo p-values overstated the evidence;
3. the break coincides with a record currency move;
4. the currency «bound» came from a near-spurious regression;
5. the Portuguese comparison used the wrong event;
6. the press had already measured the rush and the fall;
7. the pre-registration's scored file was not frozen, and two of its tests were weak;
8. the abstract's framing of the baseline was off.

This version applies the fixes («Fixes applied» in the internal review file). It changes the headline: from «a third, only there» to «coincided, not supported by the tests fixed in advance».

No partner of EasyByte has read the study yet.

10. Competing interests

EasyxLab is the research lab of EasyByte Hub S. Coop. Mad., a software company. EasyByte has no client in real estate, investment migration or residence services, and no stake in the outcome. No one funded this study.

11. Ethics

• Every source is aggregate. No personal data were collected.
• We respected robots.txt wherever it can work as a reservation of text and data mining, and never bypassed an access control. Hosts that blocked our User-Agent were left alone, including the ministry that publishes the permit statistics. We did not read archived copies of them instead.
• The Notariado's statistics portal forbids reproduction, and its site reserves its content against AI input. Registradores' site blocks us. We cite only what the press reported from them.

How to cite

EasyxLab (2026). Did purchases by foreign non-residents fall after Spain ended its golden visa? Evidence from MIVAU's quarterly transactions, 2007–2026. Study S21. EasyByte Hub S. Coop. Mad. https://github.com/easybytehub/easyxlab

References

• Ley Orgánica 1/2025, de 2 de enero, de medidas en materia de eficiencia del Servicio Público de Justicia. BOE-A-2025-76. https://www.boe.es/buscar/act.php?id=BOE-A-2025-76
• Ley 14/2013, de 27 de septiembre, de apoyo a los emprendedores y su internacionalización (consolidated; and the wording of 29 July 2015). BOE-A-2013-10074. https://www.boe.es/buscar/act.php?id=BOE-A-2013-10074
• Ley 20/2022, de 19 de octubre, de Memoria Democrática, additional provision 8. BOE-A-2022-17099.
• Real Decreto Legislativo 1/1993 (Texto refundido de la Ley del ITP-AJD), consolidated. BOE-A-1993-25359.
• Real Decreto-ley 26/2026, de 29 de septiembre. BOE-A-2026-20266.
• Decreto Legislativo 1/2024, de 12 de marzo (Código tributario de Catalunya, libro sexto), consolidated, art. 641-1; Decreto-ley 5/2025, de 25 de marzo. BOE-A-2024-6951; BOE-A-2025-10270.
• Decreto Legislativo 1/2010, de 21 de octubre, de la Comunidad de Madrid, art. 28. BOCM-m-2010-90068.
• La Moncloa. «Sánchez anuncia que el Gobierno eliminará la 'Golden Visa'…», 8 April 2024; Referencia del Consejo de Ministros, 9 April 2024; «Sánchez … foro vivienda», 13 January 2025; Referencia del Consejo de Ministros, 29 September 2026. www.lamoncloa.gob.es
• INE. Calendario de difusión, Programa anual 2026 del Plan Estadístico Nacional 2025–2028. https://www.ine.es/normativa/leyes/plan/plan_2025-2028/calendario2026.pdf ; Inventario de Operaciones Estadísticas, 25003.
• Ministerio de Vivienda y Agenda Urbana. Transacciones inmobiliarias (tables 1.6, 1.7). https://apps.fomento.gob.es/BoletinOnline2/ ; datos.gob.es, e05233601.
• Decreto-Lei n.º 14/2021, de 12 de fevereiro (Portugal). Consolidated text, Procuradoria-Geral Distrital de Lisboa: https://www.pgdlisboa.pt/leis/lei_mostra_articulado.php?nid=3378&tabela=leis
• Lei n.º 56/2023, de 6 de outubro (Portugal). Same source, nid=3709.
• Instituto Nacional de Estatística (Portugal). Indicators 0012785, 0012786.
• Governo de Portugal. Comunicado do Conselho de Ministros de 16 de fevereiro de 2023.
• European Commission (2019). Investor Citizenship and Residence Schemes in the European Union. COM(2019) 12 final.
• El País (23 May 2026). «Los extranjeros aceleraron la compra de casas de más de medio millón de euros en los últimos meses de vida de la golden visa». Known from a search snippet.
• El Periódico (9 April 2026). Purchases by non-residents in 2025, Notariado figures. Known from a search snippet.
• La Razón (9 April 2026). «La compra de viviendas por extranjeros no comunitarios se contrae …». Known from a search snippet.
• Colegio de Registradores. Estadística Registral Inmobiliaria, 1T-2025, press release. Known from a search snippet.
• Foremny, D., Li, Z., Martínez-Toledano, C. and Segú, M. (2025). Golden Visas and Real Estate Markets. SSRN 5276559 / 5347820. https://doi.org/10.2139/ssrn.5276559
• Pereira dos Santos, J. and Strohmaier, K. (2024). All that Glitters? Golden Visas and Real Estate. SSRN 4761334. https://doi.org/10.2139/ssrn.4761334
• Surak, K. (2020). Who wants to buy a visa? Comparing the uptake of residence by investment programs in the European Union. Journal of Contemporary European Studies. https://doi.org/10.1080/14782804.2020.1839742
• Surak, K. and Tsuzuki, Y. (2021). Are golden visas a golden opportunity? Journal of Ethnic and Migration Studies. https://doi.org/10.1080/1369183x.2021.1915755
• Holleran, M. (2019). Buying Up the Semi-Periphery: Spain's Economy of 'Golden Visas'. Ethnos. https://doi.org/10.1080/00141844.2019.1687548
• Gaspar, S. and He, Y. (2026). International money transfers for housing investment: the case of golden visas. Housing Studies. https://doi.org/10.1080/02673037.2026.2724837
• Antràs Porta, M. (2025). Las medidas adoptadas para la limitación de la inversión extranjera en el mercado inmobiliario español. Esade Law Review. https://doi.org/10.56269/elr/2025.6/map
• Galán Azofra, M. O. (2025). Derogación del visado de residencia para inversores en España, «golden visa»: LO 1/2025. https://doi.org/10.69592/979-13-7011-288-2-cap-21
• Banco de España (2020). La inversión extranjera en el mercado inmobiliario residencial español entre 2007 y 2019. Boletín Económico 2/2020.
• Lajer Baron, A., López Rodríguez, D. and San Juan, L. (2024). El mercado de la vivienda residencial en España: evolución reciente y comparación internacional. Banco de España, Documentos Ocasionales 2433. https://doi.org/10.53479/37873
• SAGE Business Cases (2024). The Downfall of the Golden Visa: Why Spain Cut Its Lucrative Program. https://doi.org/10.4135/9781071971031
• idealista/news (2024). «El Gobierno ha concedido 573 'golden visa' desde que anunció su eliminación», 25 November 2024.

Cite this study

EasyxLab (2026). Did purchases by foreign non-residents fall after Spain ended its golden visa? Evidence from MIVAU's quarterly transactions, 2007–2026. Study S21. EasyByte Hub S. Coop. Mad. https://github.com/easybytehub/easyxlab/tree/main/studies/s21-golden-visa-end

@techreport{easyxlab_s21,
  title       = {Did purchases by foreign non-residents fall after Spain ended its golden visa? Evidence from MIVAU's quarterly transactions, 2007–2026},
  author      = {{EasyxLab}},
  institution = {EasyByte Hub S. Coop. Mad.},
  number      = {S21},
  year        = {2026},
  url         = {https://github.com/easybytehub/easyxlab/tree/main/studies/s21-golden-visa-end}
}