Building brings rents down: Austin, Auckland, Minneapolis and twenty years of evidence

Austin, Asheville, Minneapolis, Auckland, Helsinki: wherever the housing stock has grown markedly faster than demand, rents have fallen or stopped rising. Studies measuring the effect of new construction on neighbouring rents have pointed the same way for twenty years. Density is not the variable that matters; what counts is the ratio between what gets built and what the population needs. The lesson applies to Luxembourg, where 4,294 hectares of land already zoned for building are waiting.

“Building expensive new homes does not lower rents”: the intuition that persists

New buildings let for high rents, attract affluent households and pull the neighbourhood's prices upward; building more would amount to producing luxury without changing anything for everyone else. Economists have given this intuition a name: supply skepticism, the doubt that more supply helps. It holds a grain of truth at the scale of a street, where new housing is generally dearer than its neighbours. It is false at the scale of a market. The household that moves into a new flat frees up the home it occupied, the one that takes it over frees up another, and within a few years the chain works its way down to modest neighbourhoods. When vacant homes pile up, landlords lower their asking prices, new or not. American cities have just demonstrated this in real time.

Austin: a housing stock growing 6 to 10% a year, rents down by a fifth

Austin, Texas, is the best-documented case, with two measures of 2025 deliveries. Northmarq, using CoStar data, counts 30,002 apartments delivered in the metropolitan area in 2025, a record, or roughly 8.7% of the stock. RealPage, over the twelve months to the end of September 2025, counts 20,898, or 6.4% of a stock of 345,300 multifamily units; annual growth of the stock had peaked at 10% at the end of 2024. The gap comes down to scope and time window, not to the trend.

Rents followed, downward. According to RealPage, they were falling 7.4% year on year in October 2025, and 14.6% for class C units, the most modest. Apartment List measures a 19.9% drop for a two-bedroom apartment between the August 2022 peak and October 2025, from $1,725 to $1,382 a month. The vacancy rate rose from 3.96% to 9.71%.

−19.9%

Developers responded by halting new starts; deliveries expected for 2026 fall, depending on the source, to between 4,600 and 12,000 units. In the second quarter of 2026, Northmarq finds Austin rents rising again after a three-year slide. The mechanism works both ways: when supply overflows, rents fall; when it dries up, they climb again.

Asheville: a national record for deliveries, rents down despite the hurricane

Asheville, North Carolina, is a mid-sized city that made the same bet. According to RealPage, 3,549 multifamily units were expected there in 2025, or 13.1% of the existing stock: the highest growth rate of any American market tracked, against a national average of 2.4%. No law capped rents there; the city granted permits and legalised duplexes across a large part of its territory.

Rents there have fallen year on year without interruption since November 2023: −4.2% over one year in 2025 according to Apartments.com, with vacancy projected by RealPage above 14% through 2026. One qualification is needed: Hurricane Helene, in September 2024, hit the region hard and weakened demand. But the trend was under way a year before the storm.

Minneapolis: two readings of the same gap

Minneapolis is the most-cited American case, and the most disputed. The “Minneapolis 2040” plan, adopted in 2018, abolished single-family-only zoning and parking minimums. According to the Pew Charitable Trusts (January 2024), the housing stock there grew 12% between 2017 and 2022 while rents rose 1%; in the rest of Minnesota, the stock grew 4% and rents 14%.

The Federal Reserve Bank of Minneapolis published a counter-reading in August 2025. Over 2019-2023, the median rent rose 25.7% in Minneapolis against 28.9% in a panel of comparable cities: a gap of 3.2 points. More than 18,000 multifamily units were permitted between 2018 and 2022, a record, but only 225 duplexes or triplexes, the form the reform targeted, were built between 2020 and 2024. Above all, the Fed attributes most of the gap to a negative demand shock in 2020, when vacancy jumped to 9.5%. Minneapolis remains a case that favours the thesis, not proof on its own.

Auckland: the only experiment measured over eight years

Auckland provides the best-measured case. In 2016, the Unitary Plan allowed more height and density on roughly three quarters of the city's residential land. Ryan Greenaway-McGrevy and Peter Phillips (Journal of Urban Economics, 2023) estimate that the reform produced 21,808 additional permits in five years, about 4% of the stock; a method using comparable cities puts it at 43,500 homes in six years, about 9%.

The effect on rents is measured in a second paper by Greenaway-McGrevy (Economic Inquiry, 2025): eight years after the reform, Auckland rents are about 23% lower than they would have been without it. Criticism exists: the economist Cameron Murray disputes that the reform accelerated construction and attributes the rise in permits to the cycle. The debate is about the size of the effect, not its sign.

−23%

Helsinki and Tokyo: the boom is over, the oversupply remains

Helsinki went through a construction boom in the late 2010s. In the first quarter of 2026, open-market rents in Greater Helsinki were down 1.3% year on year according to the manager INNA, and Nordea notes that “the rental oversupply is being absorbed slowly”.

Tokyo deserves a correction. The formula “Tokyo builds massively and its rents have been stable for thirty years” circulates widely. The first half is true: the Greater Tokyo area recorded 286,709 housing starts in fiscal 2023, out of 801,760 across Japan (Ministry of Land, relayed by Patience Realty). The second half no longer holds: Savills measures a 5.2% year-on-year rise in rents in the 23 central wards in 2025. Tokyo remains affordable for a conurbation of its size, because it let people build; but “stable” is a shortcut to abandon, for want of a long series to support it.

City	Stock growth	Rents
Austin (metro area)	+6.4% in 12 months (RealPage), up to +10%/year at end-2024	−7.4% year on year (Oct. 2025); −19.9% since August 2022
Asheville	+13.1% in 2025 (RealPage)	−4.2% year on year (2025)
Minneapolis	+12% in 2017-2022 (Pew)	+1% (Pew); −3.2 pts vs comparable cities (Fed, 2019-2023)
Auckland	+4% to +9% attributable to the 2016 reform	−23% vs counterfactual after 8 years
Greater Helsinki	Boom 2016-2021, retreat since	−1.3% year on year (Q1 2026)
Counter-example
San Francisco	705 homes permitted in 2024, ~1,400 in 2025	+12% in 2025; +23.1% year on year (August 2026)

Density describes a state, not a dynamic

The objection is that Austin is a sprawling city, with no comparison to a dense European one. That confuses two things: density measures the state of the stock at a given moment; rents respond to its change relative to demand.

San Francisco demonstrates it in the negative. Dense, at around 7,000 inhabitants per square kilometre, it posts some of the highest rents in the world. It permitted 705 apartments in 2024 and about 1,400 in 2025; 2,677 homes were completed there in 2025 against 5,440 in 2020, according to Apartment List. With the artificial intelligence boom bringing back jobs and salaries, rents are up 12% over 2025 and 23.1% year on year in August 2026: demand leaping against a frozen supply, and density changes nothing.

The inner suburbs of Paris illustrate the same point: municipalities often above 7,000 inhabitants per square kilometre, prices among the highest in Europe, and local planning rules that most often limit height to three or four storeys. Density there is a ceiling already reached, not a reserve. The argument strikes me as convincing, but it is a reading, not a measurement.

Low density does not create affordability; the freedom to build does.

Twenty years of studies: what new housing does to neighbouring rents

City cases show correlations; academic studies isolate the cause. When a new building opens, what happens to the rents of existing homes a few hundred metres away? They fall.

Kate Pennington (2021) used San Francisco's fire-damaged plots, rebuilt larger independently of the neighbourhood's dynamics: within a 500-metre radius, rents fall 2.3% and resident departures 17.1%. Brian Asquith, Evan Mast and Davin Reed (Review of Economics and Statistics, 2023) tracked large new buildings in low-income neighbourhoods across eleven American cities: neighbouring rents fall 5 to 7%. Xiaodi Li (Journal of Economic Geography, 2022) finds the effect in New York.

Evan Mast (Journal of Urban Economics, 2023) followed 52,000 residents of new buildings in twelve cities to reconstruct moving chains: 100 new market-rate homes free up, within two to five years, about 70 homes in neighbourhoods with below-median income. Bratu, Harjunen and Saarimaa (2023) find the mechanism in Helsinki using exhaustive registers. Vicki Been, Ingrid Gould Ellen and Katherine O'Regan reviewed this literature in “Supply Skepticism Revisited” (Housing Policy Debate, 2025): new supply reduces or slows rents at the regional scale, often locally too, and nothing shows that it increases displacement. The doubt was put on trial; it did not survive.

Study	Setting	Measured effect
Pennington (2021)	San Francisco, fire-damaged plots	Rents −2.3% within 500 m; departures −17.1%
Asquith, Mast, Reed (REStat 2023)	11 US cities, low-income neighbourhoods	Neighbouring rents −5 to −7%
Li (J. Econ. Geography 2022)	New York	+10% stock within 150 m → rents −1%
Mast (JUE 2023)	12 US cities, 52,000 residents	100 new homes → ~70 freed up in modest neighbourhoods
Bratu, Harjunen, Saarimaa (JUE 2023)	Helsinki, exhaustive registers	Moving chains reaching down to low incomes
Been, Ellen, O'Regan (HPD 2025)	Literature review	Supply reduces or slows rents; no additional displacement

The counter-model: capping rents without building

Faced with rising rents, the most popular response is not to build but to cap. Its effect has been measured too. Rebecca Diamond, Tim McQuade and Franklin Qian (American Economic Review, 2019) studied the 1994 extension of rent control in San Francisco: affected landlords cut their rental supply by 15% and the mobility of protected tenants fell by 20%. Sitting tenants gained; the next ones found fewer homes, at higher prices.

Paris offers a more nuanced result. The Atelier parisien d'urbanisme (APUR, the city's planning agency, 2025) estimates that the rent cap introduced in 2019 kept rents about 5% below their counterfactual level, 1,632 euros a month against 1,717 over July 2024-June 2025, with no lasting reduction in the number of homes let between 2018 and 2025. Compliance among listings reaches 70% in Paris and 56% in Seine-Saint-Denis. Rent control therefore moderates at the margin, when it is respected, and creates no housing: it redistributes a shortage, it does not resolve it.

Luxembourg is following the same path with bill 8184, which would lower the rent ceiling on new leases from 5% to 3.5% of invested capital. The text is still under discussion; the Observatoire de l'habitat published three studies on 17 April 2026 to frame its effects. Whatever one thinks of it, the measure will not produce a single square metre. The question of volume remains wide open, and it is land hoarding, far more than the level of rents, that governs it.

And Luxembourg?

The country had 690,959 inhabitants on 1 January 2026 according to STATEC, up 1.3% over one year. In the absence of a recent official count of the housing stock, dividing the population by households of 2.4 people on average gives an order of magnitude of 280,000 to 300,000 homes. It is an estimate, and it is enough for the reasoning.

A 10% increase in the stock, the order of magnitude of Austin or Minneapolis, therefore represents 28,000 to 30,000 homes. The theoretical need is 6,000 homes a year according to the Observatoire de l'habitat, cited by Le Quotidien, against an average output of 4,072 a year between 2004 and 2022. In other words, +10% of the stock is five years at the pace of need, seven at the historical pace, and nearly nine at the pace of the 3,404 homes permitted in 2025, the fourth year of decline according to STATEC's Note de conjoncture 1-2026 (economic outlook). Land, for its part, is not lacking: Note 32 from the Observatoire (July 2023) records 4,294 hectares already zoned for housing, a potential of 161,500 homes according to Antoine Paccoud (LISER). I detailed what this land is and where it lies in a previous article.

161,500

Meanwhile, the rental market is heading the opposite way. According to the Observatoire de l'habitat's analysis report no. 25 (June 2026), advertised rents for apartments rose 4.4% year on year in the first quarter of 2026, accelerating, against inflation of 1.6%. The number of rental listings grew 20% in a year without slowing prices: the flow of listings reflects tenant turnover, not the size of the stock.

What I see in Immovalu's data points the same way. As of 8 September 2026, the site tracks 3,066 rental listings, with a median asking rent of 2,215 € a month and 31.7 €/m². These are medians of listings, higher than the portals' averages (1,647 euros according to Switchr.lu and Immotop.lu for 2026): they measure what a new tenant pays today, not existing leases. They can be compared municipality by municipality on the towns page, and rent set against purchase price on Buy or rent.

Density describes the state of the market, not its dynamic

The cities gathered here have only one thing in common: their stock grew much faster than demand for several years, and rents gave way. The counter-examples, San Francisco and the inner suburbs of Paris, are dense cities that stopped building. Density tells you about the state of a market; it says nothing about its dynamic, which depends on a single ratio: homes delivered relative to households arriving. In Luxembourg, that ratio has been below one for twenty years, while buildable land exists and needs have been quantified. The housing crisis is not a geographical fate. It is a choice of volume, and the cities that made the other choice show the result.