Algorithms That Decide Your Rent in Spain and the United States

Algorithms that decide rents are revenue management programs that recommend to large landlords the price of every flat every day, and their effect is that, without anyone deciding to raise rents, they all rise at once. The renewal letter arrives with almost ten percent more, the tenant asks why and the answer is calm: the system says so. There is no person behind that figure but a program that spent the night reviewing thousands of leases, empty units and renewals in the same area and returned a number; the landlord does not negotiate because, strictly speaking, he decided nothing. The question is who answers when the one responsible is a model.

Revenue management: maximise the total, even with flats left empty

The mechanism works like the airlines'. Large owners hand a technology company data that is not public, actual signed rents, discounts, lease lengths and occupancy building by building; the program pools that information from owners who compete with each other and returns a recommended price for every unit. The key is that the model does not try to fill every home but to maximise total revenue, and it often recommends leaving flats empty for a few weeks rather than cutting the rent, because a cut contaminates the price of the whole building. That detail changes the nature of the market: in theory each owner decides alone and competes for the tenant, and that balance disappears when everyone consults the same oracle fed with everyone's data.

A cartel without minutes: tacit coordination and the comparison with redlining

Nobody meets, nobody signs a pact or keeps minutes, and yet the outcome looks like a classic cartel; lawyers call it tacit coordination through an intermediary, a company that competes with nobody but sits at the centre and synchronises everyone, replacing the illegal agreement with an apparently technical recommendation. The uncomfortable comparison is with redlining: for decades, official maps painted in red the Black and immigrant neighbourhoods of American cities where mortgages should not be granted, an exclusion signed and stamped that could therefore be exposed and taken to court. The current version carries no signature. Cathy O'Neil warned in Weapons of Math Destruction (2016) that such systems inherit the prejudices of the data that trained them and hand them back as calculation.

United States: the lawsuit against RealPage and the city bans

The most advanced case is in the United States. In 2024 the Department of Justice, together with several states, sued RealPage, the leading company in the sector, for reducing competition between landlords in rent setting, and also sued large owners who used its program; some have reached financial settlements without admitting fault. A White House report estimated that algorithmic price setting cost tenants around 3.8 billion dollars in 2023, and cities such as San Francisco, Philadelphia and Minneapolis have banned setting rents on a program's recommendation; the company has sued back, arguing that banning a recommendation amounts to censoring information. The battle is no longer about planning but legal and algorithmic.

Spain: portals, automated valuations and the public reference index

In Spain the phenomenon is blurrier and less documented. There is no equivalent court case, but tenant unions argue that the big property portals and automated valuation tools publish price estimates that owners adopt as a minimum reference: the whole market looks into the same mirror, and the mirror always suggests raising the price. The administration has answered with an algorithm of its own, the official reference price index of the 2023 housing law, which caps what corporate landlords may charge in areas declared under strain, so that the only remaining argument is about which model rules.

The counterargument deserves attention: its defenders hold that these programs do not create scarcity, they only measure it, and that if a city lacks homes the price would climb anyway; the software would be a thermometer, not the fever. One can reply that a thermometer that recommends leaving flats empty rather than cutting the rent does not measure the market but manufactures it, and that real scarcity does not justify coordinating those who should compete. The lesson is that responsibility does not disappear by being delegated to a model: whoever hands over data and applies the recommendation decides, and housing policy has to be able to audit the algorithm, ban coordination and return to negotiation between landlord and tenant what a program has turned into a number.