Your income in square metres

What the European data shows

Everything in this section comes from the TU Wien study and the ESPON table behind it: 31 countries, asking prices, income after tax per adult-equivalent. The American section follows, and the two do not share a scale.

The two maps measure the same thing from two sides. For buying, the authors take a third of a municipality's average income, assume it goes into the monthly instalment of a 30-year mortgage at the country's fixed rate, and ask how many square metres the resulting loan buys at local prices, taxes and fees included.

For renting they divide the same third by the local rent per square metre. Both answers are in square metres, so places can be compared without knowing a single price.

Close to 44 % of Europeans live where that mortgage buys fewer than 50 square metres, and about 28 % where it buys 50 to 75. Put together, more than 70 % of the population lives in a place where a third of the average income buys at most 75 square metres. Fewer than 5 % live where it buys more than 150.

Renting looks much the same. More than 39 % live where a third of the average income rents fewer than 50 square metres, and about 22 % where it rents 50 to 75.

That worst rental band, under 50 m², is mostly an urban problem: of the people who live in it, more than 65 % live in cities, about 26 % in towns and suburbs and about 10 % in the countryside. Rural areas stand out for another reason: for 12 % of the rural population, the collection of listings found not a single rental offer.

Which country is cheapest depends on how big a flat you want. Small flats cost more per square metre everywhere, but how much more is a national trait, and it is enough to reorder the continent.

Rank the 31 countries by the years of average income a flat costs outright, and moving the question from 25 m² to 150 m² lifts Norway ten places and Spain nine, while Germany and Sweden each fall seven. Germany is the seventh easiest country in which to buy 25 m² and the fourteenth for 150 m²; Spain is nineteenth and tenth.

Nothing about incomes or overall price levels differs between the two rankings. Only the size did.

How far to trust this depends on where the extra cost of small flats was actually measured. In 23 of the 31 countries it comes from the country's own listings in all five size bands, from under 30 m² to over 120 m², and every country named above is among them; in the other 8 the smallest band had too few listings, and the European median is used instead.

One caution the paper is explicit about: where a band had too few listings, the authors filled it in by interpolation, so a band that carries a number was not always actually observed.

The geography is what you would expect, only sharper. The capital region of nearly every country sits in the least affordable class and pulls its neighbours with it: unaffordability spills into the regions around Paris, Berlin and Madrid despite the higher incomes there, and across borders around Switzerland and Luxembourg.

The coasts of France, Spain, Portugal, Greece and Croatia and the tourist valleys of the Alps are far less affordable than the countryside behind them. Scandinavia and the Baltic states come out markedly more affordable, save a few urban hotspots. The map shows this as dots; the rankings and the cards carry the regional figures.

At the country level the authors name Portugal, the Netherlands and Switzerland as rather unaffordable, and likewise Poland and Hungary.

There is a caveat on Poland that matters for reading the whole map: an extremely high share of Poles own their home outright, with no mortgage, so the gap between prices and incomes is not felt by the majority. It is felt by newcomers: young households, and anyone who has to move.

Renting is more affordable than buying in Germany, France, Belgium, Austria and Switzerland, but not in Portugal, Greece, Bulgaria or Ireland. In Vienna a third of the average income rents 52 m² but buys only 31; in Madrid it buys 49 and rents 38.

The authors warn against reading the two maps as a like-for-like choice: a mortgage instalment builds equity, a rent payment does not, and the security of tenure differs. Both maps describe the market from a newcomer's point of view, at today's prices; they say nothing about what current residents pay.

What the American data shows

The American layer rests on a single survey: the Census Bureau's American Community Survey, five-year estimates for 2020–2024. It gives every county four numbers: the median home value, the median household income, the median rent and the population. A median is the middle value: half are above it, half below.

That is 3,141 counties and 100 % of the population of the fifty states and the District of Columbia. It is a survey of households, not a record of sales, and every number below inherits that.

Nationally the median home costs 4.1 years of the median household income, and the median rent takes 21 % of the median monthly income.

What the map adds is how unevenly that is spread: 29 % of Americans live in a county where the home costs five years of income or more, and 9.7 % live where it costs eight or more; 1.8 % live where it costs ten or more.

Where it costs most. By years of income the top of the list is Teton County, Wyoming, at 13.2: that is Jackson Hole, where housing is priced for buyers who do not earn the local income. Then Nantucket at 11.4.

Right behind them sit two of the three most densely populated counties in the country, Brooklyn at 11.3 and the Bronx at 10.9, and those two hold four million people between them, against twenty-three thousand in Teton. Manhattan is 10.5, San Francisco 9.9, Los Angeles 9.3.

The cheapest big counties are in the industrial North and along the Rio Grande: Pittsburgh's Allegheny County and Detroit's Wayne County at 2.9, Hidalgo County on the Mexican border at 2.6. A median home costs about one year of income only in a handful of thinly populated counties in Kansas and Texas.

The rent share should not be read without a warning. Only 1.2 % of Americans live in a county where the median rent takes 30 % or more of the median income, and that sounds like good news only until you notice what it measures: the middle rent divided by the middle income of all households, owners included.

Ask the question household by household instead, as the same survey does, and 51 % of American renters pay 30 % or more of their income in rent. Both numbers are correct; they answer different questions, and a county map can only draw the first.

Where the square metre stops. Everything above avoids the word "area", not by choice. The survey asks an owner what the home is worth and a tenant what the rent is; it does not ask how big the place is, and no other federal survey asks it county by county. That gap is why the three square-metre views are switched off on the US map.

Whether an American home is expensive per square metre is a question American public statistics cannot answer. This page is built on public statistics only, and where they end, the page ends too.

Five years averaged into one number, and a value that is only the owner's opinion. The five-year estimate is not a snapshot of 2024 but an average of interviews spread over 2020 to 2024, expressed in 2024 dollars. It is the only edition that covers every county, because the one-year survey publishes only places above 65,000 people.

And the value is what the owner says the home would sell for, not what anyone paid: a self-assessment, not a sale price. Neither is a defect to be corrected; both are facts to be stated before the map is read.

How firm each county's number is, is published too. Unlike the European layer, the survey ships a margin of error with every estimate, and every county card shows it. For the median county the interval is ±5.9 % of the value; for one county in ten it is wider than ±16 %.

In 187 counties, home to 0.29 % of the population, it exceeds a fifth of the estimate, and the card says so in words. Those counties stay on the map, because the survey did happen there and a blank would be worse than a wide interval that is clearly flagged.

What the Latin American layer shows

Only two cities in the region publish a price per square metre by district, and both are capitals: Lima, through its central bank, and Buenos Aires, through the city's statistics institute. That is why the block above is labelled by city and never by country.

Lima: what buyers paid. Every quarter the Banco Central de Reserva del Perú publishes the price per square metre, in US dollars, of flats sold in 12 districts of the city. In the fourth quarter of 2025 it runs from US$1,392 in La Molina to US$2,463 in Barranco.

The bank also publishes how many years of rent the price equals, and dividing one by the other gives a rent: from US$7.5 per square metre a month in La Molina to US$13.4 in Barranco.

Buenos Aires: what sellers asked. The city's statistics institute, the Instituto de Estadística y Censos de la Ciudad de Buenos Aires, publishes every quarter the average asking price per square metre in listings, by barrio, separately for each flat size.

This page takes two-room resale flats, the size that covers the most barrios: 43 of 48 in the second quarter of 2026, from US$1,437 in Villa Lugano to US$5,805 in Puerto Madero. The quarter is provisional and the institute may revise it.

The institute shows a barrio only when enough flats were on offer in the quarter; the other 5 stay grey on the map.

The two cities cannot be compared with each other. A price paid and a price asked differ by the bargaining in between; one comes from a central bank's survey of developers and agents, the other from listings; and the quarters differ, the fourth of 2025 in Lima and the second of 2026 in Buenos Aires.

That is why each map is coloured on the districts of its own city: green is the cheapest fifth, red the dearest. A green barrio in Buenos Aires does not mean cheaper than a red district in Lima.

No local income, so only yours. Neither city publishes an income by district, so the maps cannot say what a local resident can afford; they answer for the income you type into the calculator above.

Buying also needs a mortgage rate, and the layer has none for either city. None is borrowed from elsewhere: in both cities, buying is counted only at a rate you type yourself.

The OECD adds four countries, as a share rather than a price. Its Affordable Housing Database gives the median share of income that households with a mortgage or a rent spend on it: Chile 23.7 % in 2022, Colombia 20.5 % in 2024, Costa Rica 17.0 % in 2024, Mexico 14.4 % in 2022.

The four are not fully comparable either. Chile, Colombia and Mexico divide by income before tax, Costa Rica by income after tax, which is smaller, so on the others' terms its share would be lower still. Chile's jump to 31.8 % in 2020 comes from a survey run during the pandemic, and two years later the share is back down.

Limitations, theirs and mine

Both the data and the arithmetic can be wrong. The figures were collected by machine from several statistical offices and reworked here by one person: a wrong join, a stale file or a slip in a formula are all possible, and none of the sources has checked this page. Before you rely on a number, check it against the source, which every section links to.

In Europe, listed prices are not transaction prices. The study works from asking prices, and the gap to the price actually paid is not known at municipal level.

Where the price paid is known, it correlates well with the listing, but it follows patterns of its own. The American layer is neither: it is what owners think their homes are worth, which is not a listing and not a sale.

A place is one number, and an American county is a large place. Inside a city the difference between districts can be as large as the difference between regions, and the map cannot show it. A county is coarser still: Los Angeles County holds ten million people and one median home value.

The map also cannot shade regions: no freely reusable outlines of European regions exist at this detail, so this page fills whole countries and marks places as dots instead. Regional and state figures are still in the search, the cards and the ranking.

European local income is an estimate. The survey income is spread across municipalities by a model, not measured in each one.

In several Eastern European countries that spread varies little outside the big cities, so some intermediate regions, neither mostly urban nor mostly rural, may look poorer than they are. The authors ask that absolute values be treated with caution; the patterns are more robust than the digits.

No down payment, no loan limits. The study abstracts from both. The down-payment slider on this page is my addition: it reduces the loan and lets you see what a saved deposit is worth. The limits banks set on a loan, such as a cap on its share of the home's value or on its size relative to income, and other formal barriers to buying are still not modelled.

Your income is not adjusted to match either definition. When you type a household income, the page takes a third of the whole thing, because that is what a household actually spends, and uses the household composition only for the line comparing your income with the local average.

That comparison line is against an income per adult-equivalent in Europe and against a median household income before tax in the United States, and is labelled as such — but the square metres, and in the United States the years of income, come from your money and the local price, and that is the same arithmetic everywhere.

Timing. European listings are from 2024–2025, incomes 2024, rates mostly 2023 averages, population the 2021 census round. The American figures are five-year survey estimates covering 2020 to 2024, stated in 2024 dollars.

So the American side is not a snapshot at all but an average across five years, which in a period of fast-moving prices lags a current market. That is one more reason the two layers sit on separate maps.

The two incomes are not the same quantity. Europe's is a mean equivalised disposable income: after tax, divided by a household-size weight, so it is per adult-equivalent. America's is a median household income before tax, for everyone living together, with no equivalence scale at all.

Both are what their statistical office publishes; neither can be converted into the other without modelling tax and household composition, which would be a guess dressed as a number.

This page does not make that conversion. It keeps the layers apart and offers the honest alternative: put in your own income and both sides answer the same question.

Asking prices against owners' estimates. The European layer is built from listings — what sellers asked. The American layer is built from the survey — what owners say their home is worth.

Neither is what buyers paid, and the two differ from a sale price in different ways, so the two layers' prices are not compared anywhere on this page.

And the two prices include different things. The European figure is harmonised to carry real-estate transfer tax or VAT, land registry and notary fees — in Vienna that is 6.4 % on top of the listed price.

The American figure is the owner's estimate of the home alone: closing costs, transfer taxes where states levy them, and agent commission are all outside it.

So the European price already includes the costs of the purchase and the American one does not, and no attempt is made here to add them to the American one, because the rates differ by state and by deal and estimating them would be a guess.

The American home value is an opinion, not a price. The survey asks owners what they think their home would sell for. That is not a transaction price and not an asking price, and nothing on this page pretends otherwise.

It is the only county-level measure of housing value the United States publishes, and it is used here under its own name.

American rent is gross and per home, not net and per square metre. Median gross rent includes the utilities a tenant pays, and it is the rent of the home in the middle of the range, whatever that home's size.

It cannot be divided into square metres, and comparing it with the European rent per square metre would be comparing two different things — so the American rent share is computed on the county's own median rental and labelled that way.

The American rent share is a ratio of medians, not a count of burdened households. The map says the median rent takes 30 % or more of the median income in counties holding 1.2 % of Americans.

Asked household by household, the same survey says 51 % of renters pay 30 % or more. The second is the number that describes people; the first is the only one that can be drawn on a county map. The gap between them comes from how unevenly rents and incomes are spread among households, and a single number per county cannot show that spread.

There are no American square metres at all. Floor area is not collected by the ACS or by any other federal source at county level, so the three square-metre views are switched off on the US map. Better a switched-off button than a map of guesses.

Beyond Europe and the United States, only two cities. Lima and Buenos Aires publish a price per square metre by district, and they sit on maps of their own, outside the rankings and the comparison.

Adding a country would take a price per unit of area published by its statistics office or central bank, and an empty map with an honest label beats a filled one with an invented number.

Where the numbers come from

Every number on this page comes from a public file that anyone can download without a key, an account or a payment. Here is the whole provenance in one table; the paragraphs after it are the parts that needed a decision.

Read the price and income rows together and the reason the layers stay apart is visible at a glance: one column is asking prices against after-tax income per adult-equivalent, the other is owners' estimates of whole homes against before-tax income per household. Both are defensible measurements. Neither converts into the other.

The paper's data availability statement says the data is available at reasonable request to the authors. No table is attached to the paper; the supplementary files are the two maps as vector PDFs. The numbers on this page therefore came from the next best place, and the following is exactly what happened.

The ESPON map service. The HOUSE4ALL project publishes an interactive StoryMap on the ESPON ArcGIS portal. Its maps are drawn from three public feature services on the programme's GIS server.

One of them, HOUSE4ALL_data_at_LAU_level, carries a row per municipality with the 2024 mean equivalised income, the corrected sale price and rent per square metre, the national mortgage rate and the two affordability indicators of the paper, sa_m2 and ra_m2.

On 14 September 2026 all 89,601 rows were read through the service's query interface, 2,000 at a time, with the centroid of each municipality. That table, trimmed to the fields used here, is published next to this page as house4all-lau-2024.csv (10.7 MB).

The formula was checked, not assumed. For all 78,758 municipalities with sales data, a third of the monthly income, turned into the loan it can carry over 30 years at the row's rate, divided by the corrected price, reproduces the service's sa_m2 to within one part in a billion; the rent indicator likewise.

So the calculator uses the study's arithmetic, not an approximation of it, and with the defaults left alone it shows the study's own numbers.

The corrected price is the listing price, weighted by how long each listing stayed online, with transaction costs added: in Vienna it is 6.4 % above the raw listing price, which is about what Austrian transfer tax, registration and notary fees add up to.

What is taken from the service, and what is not in it. The five size-class prices are in the service, but the published indicators use the overall corrected price, and so does this page.

The degree of urbanisation and the coastal flag come from Eurostat's LAU 2021 correspondence table, which covers the EU-27 only; Swiss, Norwegian and Icelandic municipalities appear as "not classified" in the chart above. Overseas territories, the Canaries, the Azores and Madeira are outside the frame but inside the search.

A third is not 33 %. The paper says a third of income, and the first version of this page read that as 33 % because the slider works in whole percent.

But 33 % is 1 % smaller than a third: 33 against 33.33, a difference of one hundredth of the value. That was enough to put every figure on the page 1 % below the study's own, while the page claimed to reproduce it exactly.

The default is now exactly one third and the slider reads "a third"; move it and it becomes whole percent and stops being the study's setting, which the line under the calculator says.

Checked against the service: the published affordable metres are reproduced to within 8 parts in ten billion on 78,758 municipalities for buying and 5 parts in a billion on 47,154 for renting.

Small flats cost more per square metre. The service carries prices for five size bands as well as the overall average, and the difference is not small: across the whole sample a square metre in a flat under 30 m² costs 28 % more to buy and 38 % more to rent than the average, while one over 120 m² costs 6 % and 17 % less.

The premium is a national trait more than a local one, and it varies enormously, from 13 % in Germany to 83 % in Spain. Wherever you name a size, this page now uses the band that size falls into, so the answer to "I want 25 m²" is no longer a straight-line scaling of the answer for 100 m².

It changes the order of countries: at 25 m² Germany is easier than Spain, at 150 m² Spain is easier than Germany. The headline square metres to buy and to rent still use the overall price, because that is the study's own arithmetic.

Why the map is dots and not shaded regions. No freely reusable outlines of European regions exist at this detail, so the map fills whole countries with Natural Earth's public-domain shapes and marks each place as a dot.

The region names come from Eurostat's statistical code list, which is free to reuse for any purpose. The projection is Lambert azimuthal equal-area, the family the ESPON maps use.

Where the page and the paper differ. The population shares computed from the service (41 % under 50 m² to buy, 37 % to rent) are a few points below the paper's 44 % and 39 %.

The service is the project's data release; the paper's figure 3 was drawn from the authors' working dataset, and the two may differ in population weights, in income year or in the treatment of municipalities without data. The paper's figures are quoted as the paper's; the chart shows what the public table gives.

One source for the whole American layer. The American Community Survey gives every county a median home value, a median household income, a median gross rent and a population, in one release, as a public-domain federal work.

Every row is keyed by the federal FIPS county code, so counties are joined exactly and never by name.

How the tables are read. The Census Bureau's own API requires a key, so the same public-domain tables are read through the open API of Census Reporter, a non-profit that states plainly: "Data on Census Reporter comes from the US Census Bureau and is not copyrighted."

Their server asks callers to identify themselves, so the extractor sends this page's own name and address as its user agent rather than pretending to be a browser.

Connecticut, again, and from the other side. The state replaced its eight counties with nine planning regions, and this time it is the Census Bureau that has moved on and Natural Earth that has not: the survey publishes the nine, the map file still carries the eight.

Rather than drop the state or borrow a neighbouring point, the nine centroids are computed here from the federal TIGER 2024 boundaries — an area-weighted polygon centroid with interior rings subtracted — and each was checked to fall inside the state.

Alaska needed nothing this time: the survey already uses Chugach and Copper River, the units that replaced Valdez-Cordova in 2019.

State and national rows are asked for, not assembled. A population-weighted mean of county medians is not a median, and the first version of this page built its state and national rows that way, from counties.

The ACS publishes every level, so California's row is California's own median and the national row is the national one. The page says 4.1 years of income for the median American home because the Census Bureau's two national medians say so, not because 3,141 counties were averaged into it.

Three counties lack a number, and 78 municipios were left out on purpose. De Baca County in New Mexico has no published median household income; Kenedy and Loving Counties in Texas — populations in the hundreds — have neither a home value nor a rent.

All three are in the gap file with their reason; any new gap stops the build. Puerto Rico's 78 municipios are in the same download but excluded: they come from the separate Puerto Rico Community Survey, and the Census Bureau's United States totals do not include them, and admitting them would change what the country row means.

The size premium stops at the Atlantic. The European data measures how much more a square metre costs in a small flat, country by country, and this page uses it wherever you name a size.

There is no such measurement for the United States, and there is no American price per square metre for it to act on: sizes you name apply only to the European map.

Exchange rates for the income field are the European Central Bank's euro reference rates of 11 September 2026, fixed in the page so that a shared link means the same amount next month.