Eight months to build it. Everything else is paper and waiting.

The headline finding of this study is about measurement, not construction. Not one of the 38 countries examined publishes all three phases of housing delivery as a regular statistical series. Most publish one phase well, one badly, and one not at all — and which phase they choose to measure turns out to correlate with which phase they are good at.

Countries with fast permitting publish permitting statistics. New Zealand reports a median building-consent decision in 13 working days, quarterly, by authority. Denmark publishes municipal case-handling times for 30,288 cases a year. Neither publishes how long it takes to build the house. Conversely, the United States measures construction to within a tenth of a month across a million homes a year and does not collect the date a permit application was filed — anywhere, at any level of government.

The three phases, and why they must stay separate

Every figure in this study is assigned to one of three phases, and never blended without a label:

• P1 — Approval. Building-permit application filed until construction is legally authorised. Excludes land acquisition and architectural design.
• P2 — Construction. Authorised start or ground-breaking until completion or occupancy.
• P3 — Total elapsed. P1 start until move-in ready, including the idle gap between the permit being issued and a shovel entering the ground. That gap is reported separately wherever a source allows.

The separation is not pedantry. It is the whole finding. In the Netherlands — which has the best-decomposed housing data in the world, a full register count rather than a survey — the median new dwelling in 2025 spent 8.1 months waiting after its permit was granted and 13.3 months being built. Ten years earlier those numbers were 5.0 and 9.4. The waiting grew 62%; the building grew 41%. And critically, that 8.1-month wait sits entirely downstream of the Dutch statutory permit deadline, so no amount of permit-office reform touches it.

Read that chart carefully, because its most interesting row inverts the expected order. In Poland, single-family houses take twice as long to build as apartment blocks — 51.0 months against 24.8. Poland's statistical office is one of the few in the world that publishes construction duration split by who commissioned the work, and the split explains it: developer-built housing takes 24.9 months, owner-built housing takes 59.1 months. The same office publishes the same split by construction technology, and the technology barely matters. Using identical monolithic construction, self-builders take 47.8 months and developers take 24.5. The binding constraint is not the method. It is the client's cash flow.

That pattern recurs everywhere the data can see it. In the United States, an owner-built house takes 12.8 months of construction against 6.1 for a builder's spec home — a 2.1× ratio inside one country, wider than the gap between the United States and most of its comparators. In Lagos, where roughly 90% of housing is self-built and mortgage lending is about 0.5% of GDP, the median self-build takes 41.7 months and the mean takes 48.3. The gap between that median and that mean is the signature of a build financed out of savings: foundation, then blockwork, then roof, then finishes, with months of nothing in between.

1. The fastest and the slowest

A single global ranking is not available, and the reason matters: only about a dozen of the 38 countries publish a construction duration at all. The ranking below covers those that do, and states the statistic type each time, because construction duration is right-skewed and a mean will always exceed a median.

Five fastest on construction (P2)

1. Japan — about 5 months for an owner-commissioned detached house (median, derived from a 2,435-response industry survey). The explanation is unglamorous: the houses are small (122.5 m² average), the large housemakers are 45.6% factory-prefabricated, and the timber supply chain is mature. Caveat: the survey covers the big industrialised builders, not the small local firms that build most Japanese houses, so the national median is probably longer.
2. United States — 7.4 months (single-family, mean, Census Survey of Construction, 2025). Wood light-frame construction — 96% of owner-built homes use it — is fast and weather-tolerant enough to proceed year-round across most of the country. The median is about 5.9 months; the mean is dragged up by a 12% tail of homes taking 13 months or more.
3. Australia — about 8.7 months for a detached house (derived from ABS Building Activity). Brick veneer over timber frame. Note this is Australia's best number: its apartment sector is one of the worst performers in the study.
4. Canada — 9.7 months (single-detached, mean, CMHC, June 2026). Platform timber frame. Canada is the only country here whose construction phase is now faster than its pre-COVID norm, having peaked at 10.7 months in 2023.
5. Netherlands — 10.3 months (single-family, median, full register count, 2025). Masonry and concrete with a high prefabricated component.

Five slowest on construction (P2)

1. Poland — 59.1 months for owner-built housing, 51.0 for single-family overall (volume-weighted mean, GUS, 2025). Incremental cash-financed building, and a statistical definition that stops the clock at legal occupancy regardless of finish state.
2. Nigeria — 41.7 months for Lagos self-build (median, derived from a published survey distribution). A household savings rate expressed as a construction schedule.
3. South Korea — 38 months for apartments (mean, 2022), up from 30 months in 2013 and peaking at 41 in 2020. Attributed to the 52-hour working week, material costs and labour shortage. Korea has the best raw building data in the world — its SEUMTEO system holds permit, start and completion dates for every building in the country — and publishes almost nothing about duration.
4. Israel — 37.8 months per dwelling, 32.3 per building (mean, CBS, 2025). This is a recent and largely war-driven deterioration: the 2019 baseline was 29.2 and the 2006–2018 average was 26.7. The Bank of Israel is explicit that the October 2023 suspension of Palestinian workers' entry permits, which cut the industry's headcount by more than a third, produced “extended construction timelines.” Israel's largest apartment blocks now take 45.6 months of construction alone.
5. Chile — 29.9 months from permit to final reception (mean, measured over 6.6 million m² across a decade). Note this figure includes the idle gap, so it is not strictly comparable to a pure construction number.

On total elapsed time (P3)

Here the spread is not a factor of ten. It is a factor of forty, and the extremes are worth stating precisely.

Fastest: Japan at roughly 10 months from signing a building contract to completion; Australia at 11.5 months from approval to completion for a detached house; the United States at 8.8 months from permit issuance — though the American figure excludes an approval phase that nobody measures, so it is not a true P3 and should not be presented as one.

Slowest: Chile at 65.7 months from first drawing to final reception, up from 42.6 in 2018 — a 54% deterioration in seven years. Israel at roughly 46–51 months. Singapore's public housing programme at 51.6 months at its COVID peak, since recovered to a 36–48 month target. And England, where the median large housing site takes 5.2 to 6.3 years from planning application to its first completed home, and where the Letwin Review found a median build-out period of 15.5 years for the site as a whole. The honest English answer is that the house takes months, the site takes fifteen years, and which number you quote depends on whether you are the builder or the person waiting for the nine-hundredth home.

2. Where approval dominates — the more interesting finding

In several countries the paperwork is longer than the building. This is the finding worth carrying away, and it is well evidenced in five places.

England. The mean time to determine an outline planning application for major housing reached 710 days in 2024, against 284 days in 2014 — an increase of 426 days, or 150%, both as Lichfields publishes them. Only 4% of 2024 applications were decided inside the statutory 13 weeks; 64% took more than a year. Two details make this worse than it sounds. Approvals take longer than refusals (857 days against 453) — the delay is concentrated in getting to yes. And it is now roughly six months quicker on average to be determined on appeal than by the local authority, an inversion that says something stark about local planning capacity. Meanwhile England's official statistic reports that 91% of major applications were decided “within 13 weeks or the agreed time,” a figure rendered almost meaningless by the fact that 77% of those applications carry a negotiated extension.

Chile. The clearest documented case anywhere. Total elapsed time rose 703 days — 54% — in seven years. Chile has the shortest statutory permit deadline in this study (30 calendar days, 15 with an independent reviewer) and the longest observed permit stage (223 days for the building permit alone, 372 for the full pre-construction sequence). That is 7.4× the legal ceiling. The spread between municipalities in one metropolitan area is 2.7×: Renca 311 days, Conchalí 837. Chile's own national productivity commission mapped 439 permitting procedures across 71 public entities, 63 of them capable of stopping a project outright, and found four agencies holding 5,517 pending applications out of 8,881 submitted. Permit processing time is estimated at roughly 12% of a housing unit's value.

Spain. No municipality in the Community of Madrid meets the three-month legal maximum. The capital averages 11.0 months; Las Rozas averages 21.1. The developers' association that compiled the figures put it plainly: licence times are “in many cases longer than the construction of the building itself.”

Israel. Mean permit issuance of 319 days across all tracks, rising to 407 days where a relaxation or non-conforming use is involved — against a World Bank benchmark of 114 days that Israel's own State Comptroller cites. The Comptroller found that in six of eight committees audited, the 2014 licensing reform produced no real shortening at all.

Canada. Municipal planning approval for multi-unit housing averages 11.2 months nationally — independently corroborated by CMHC's own estimate of “on average, a full year” — with a 15× spread from Saskatoon's 2.0 months to Hamilton's 31.0. Add a high-rise permit-to-start gap of 9 to 15 months and Canadian apartment delivery is dominated by waiting, not building.

And the counter-examples, which are just as instructive

Two countries have driven the approval phase close to zero, by opposite routes.

Saudi Arabia outsourced the review. Its Balady platform publishes a service-level agreement of 1 to 10 days for a building permit, while the ministry's own service card says “immediately.” Both are true and the difference is the finding: substantive review — including Saudi Building Code compliance — has been moved out of government entirely, to accredited private engineering offices and technical inspection companies that carry liability, backed by mandatory latent-defects insurance. The government's residual step is a payment gate, and it is genuinely instant, because the government is no longer doing the reviewing. In 2019 the same permit took 19 days.

Dubai is automating it. On 9 August 2026 — eight days before this study was compiled — Dubai Municipality announced an AI system that issues villa building permits automatically, verifying plot requirements, checking architectural designs against the Dubai Building Code, assessing MEP systems and validating structural safety. Full rollout is scheduled 26 weeks out, around February 2027. Abu Dhabi reports its own platform has cut residential villa permit processing 57%.

Japan offers the closest thing to a natural experiment. Since 1999 its building-confirmation checks — a conformance verification, not a discretionary approval — can be performed either by municipal officials or by competing private certification bodies. On the routine wooden two-storey class, under the identical statute and the identical 35-day ceiling, private bodies average 10 days where public officials average 35. On that class the variance is administrative throughput, not rule ambiguity. The honest qualifier is that the advantage is not universal: on the class that requires a full structural conformity review, the ordering has just reversed — in the current financial year private bodies are averaging 38 days against the public officials’ 34, having been 35 against 62 the year before. Private review wins decisively on routine work and is roughly at parity on hard work. And Japan is still only at about 56% electronic submission: the rules are already largely deterministic, the pipeline is not yet automated.

3. Comparability warnings

Several of the numbers above cannot be placed in one column without qualification. These are the material ones.

“Completion” means different things

• Ireland is the most objective: completion is the date an ESB Networks domestic electricity connection is energised — one national dataset, one organisation, every dwelling in the state.
• New Zealand is the strictest and the most distorted: the Code Compliance Certificate is a formal legal statement, but if the owner has not applied by the two-year anniversary of consent, the authority must decide anyway. That piles artificial mass at exactly 24 months, which is why the widely-quoted “two years to build a New Zealand house” is an artefact of an administrative backstop rather than a construction measure.
• England has no certificate of occupancy at all. Completion is an inspector's signature on a building-control return, covering approximately 80% of housebuilding activity, and the government warns it should be treated as a leading indicator rather than a count.
• Australia is builder-declared via a sample survey: a building is complete when it “can fulfil its intended function.”
• Poland stops the clock at official commencement of use regardless of the building's state of finish — so a house occupied unfinished counts as complete.
• Singapore issues two certificates: the Temporary Occupation Permit (the keys moment) and the later Certificate of Statutory Completion. Conflating them is the commonest error in Singapore comparisons.
• China separates completion filing from delivery entirely. Under the presale system, when a buyer takes possession is governed by the sale contract, not by any completion event — and since 2021, frequently never.

Mean where others report median

Construction duration is right-skewed, so this matters more than it sounds. Germany, Poland, Israel, Canada, Australia and the United States publish means only for construction duration. The Netherlands is the only country in this study that publishes a construction-duration median — and it publishes both statistics plus the full quantile distribution, which makes the gap visible: a 2025 median of 23.4 months against a mean of 27.3, with a p90 of 47.3. A comparison that puts a US mean beside a Dutch median overstates US duration by roughly two months before anything real is measured. New Zealand does publish medians, but only for the consent-processing phase; its construction figure is a model assumption and is not a median of anything. Austria is the sharpest case: Statistik Austria computes a median construction duration, uses it internally as an imputation parameter, and has never published the value.

Countries counting only permitted formal construction

This is the largest single distortion in the dataset, and it is concentrated in the developing-country figures. Permit-based statistics only see housing that goes through the permit system. Where most housing does not, the permitting indicator describes a minority of production — and a country can have fast permits precisely because most housing bypasses them.

• Nigeria: roughly 80–90% of housing is self-built and outside the permit system. Every Lagos permit figure in this study describes about a tenth of Lagos housing. Slum share: 49.0% of the urban population.
• Kenya: 50.8% slum share — over half the urban population. Annual demand of about 250,000 units against roughly 50,000 produced. The 200,000-unit gap is met, to the extent it is met, by construction that never enters the permit system.
• Mexico: 57.3% of owner-occupied dwellings were self-built or self-produced (INEGI, 2020), rising to 62.8% of the total stock on a 2022 measure, and ranging from 31.6% in Nuevo León to 86.8% in Oaxaca. The peer-reviewed literature puts self-build duration at over ten years. Note the irony: Mexico has the fastest formal permitting in Latin America and the largest measured informal sector.
• Philippines: 36.6% slum share. The 22-procedure permit chain is a formal-sector artefact.
• India: the largest documented housing-completion stream in the country is a rural programme that sits outside municipal permitting entirely — 29.5 million owner-built houses. About 75% of documented subsidised completions are owner-built.
• Turkey: a single 2018 zoning amnesty legalised 3,599,867 structures against a twenty-year permit total of 14.5 million dwellings.
• Brazil, South Africa, Indonesia, Argentina all carry the same problem at 17–25%.

One caution on the slum-share indicator: it measures housing deprivation (durability, crowding, water, sanitation, tenure), not permit status, so it is a proxy and not a substitute. It also produces artefacts at the top of the income distribution — Ireland registers 8.5%, which is a definitional quirk of overcrowding and tenure measures in a high-income housing system, not a finding about Irish informality.

Two indicators that get misused constantly

The World Bank's Doing Business “Dealing with Construction Permits” series is the only dataset that ever published a comparable permitting duration across ~190 economies, and it appears in the dataset here for all 38 countries because nothing else covers them all. Three things must always be said with it. It was discontinued in September 2021 and its final data is from May 2019. It measures a two-storey, 1,300.6 m² general-storage warehouse, not a dwelling. And its “time” includes obtaining water and sewerage connections and registering the completed building at the property registry — it is a whole-lifecycle compliance clock, not a permit clock.

On the integrity question, the record is narrower than commonly reported. The WilmerHale investigation named China, Saudi Arabia, the United Arab Emirates and Azerbaijan. Chile does not appear in the report at all. And the manipulated indicators were Starting a Business, Getting Credit and Paying Taxes — construction permits was not among them. The 2019 permit figures can be cited without an integrity asterisk, provided the discontinuation and the warehouse caveat are both disclosed.

Its successor, B-READY, does publish a time in days — but for a five-storey, 1,800 m² commercial office building, and it currently covers only 17 of these 38 economies. Nineteen more arrive in the 2026 edition and two in 2027. Do not place the two side by side as a 2019-to-2025 trend: they measure different buildings with different scopes, and the apparent US movement from 81 to 97 days is an artefact of the instrument.

4. Methodology note: what would close the biggest gaps

Most of the missing data is not missing. It is unpublished, and in several cases the agency holding it has already computed it.

1. The United States should add one field. The Census Bureau's Building Permits Survey already has the reporting relationship with roughly 20,000 permit-issuing jurisdictions. Adding an application-date field would produce the first national P1 series in American history. At present the only evidence ranges from two days to over three years depending on the jurisdiction, and the best national estimate is an academic survey last run in 2018.
2. Four agencies should publish a join they already hold. Dubai Land Department's Projects dataset publishes start date, end date, completion date and completion percentage per project, with CSV export — everything needed for a median construction duration by asset type, missing only the aggregate. Caixa Econômica Federal holds unit-level contracting-to-delivery dates for over 8.4 million Minha Casa Minha Vida homes as a funds-release control. Korea's SEUMTEO system holds permit, start and completion dates for every building in the country. Mexico's RUV tracks individual dwelling units through construction verification. None publishes a duration.
3. Three agencies should release a number they have already calculated. Statistik Austria computes a median construction duration as an imputation parameter. Statistics Finland runs a per-building construction-time model whose output feeds the volume index and is never released. CBS Netherlands produced its phase decomposition as a commissioned custom table — the same commission could add the client-type split.
4. Spain and France could compute a true cohort duration from existing filings. Spain's ministry collects both the visado de dirección de obra at the start of works and the certificación de fin de obra at the end, for the same work by the same technician, and publishes only counts. France's Sitadel database records the application date, the authorisation date, the construction-opening declaration and the completion declaration — four dates that would yield genuine medians for P1, the idle gap and P2, split by detached, grouped and collective housing.
5. England should publish a duration instead of a compliance percentage. MHCLG holds the decision date for every planning application in the country. It publishes only percentages against statutory and renegotiated targets. Scotland already does it properly — mean, median, and applications with and without processing agreements reported separately — and the presentation should simply be adopted.
6. Every country should publish the idle gap. It is the phase nobody measures and, on the evidence here, the one growing fastest. The Netherlands is the only country with a clean published series (5.0 months in 2015, 8.1 in 2025). Germany's can only be derived, at roughly 9.2 months. Israel publishes it as a survival curve rather than a duration. It is derivable in nearly every country from two dates the building register already holds.

Where a figure below is our own derivation rather than a published number, it is marked [ESTIMATE] and the arithmetic is shown in the notes column. Where sources conflict, both are carried in the dataset with a note on which we trust and why. Where a country does not support a segment, the row reads “not available” and names the agency that would hold it. Nothing is interpolated.