All Housing is Housing

Luxury apartments reduce rents for the poor too

In order to be allowed to build housing in much of America, it is required that some portion of the units sold must be “affordable” housing. This is justified on the grounds that otherwise, the only housing which will be built will be luxury apartments, unaffordable to anyone but the rich. Rich consumers will get housing, while poor residents will only end up displaced out of their homes.

This is wrong. All housing is housing. When a developer builds luxury apartments, it diverts high income consumers away from other, lower-quality housing, which in turn diverts others, until eventually it has filtered down to the lower classes. Building luxury housing is equivalent to building affordable housing. In fact, in some sense all goods are other goods, because they all trade off against a consumer’s budget constraint.

You don’t need to take my word for it – we can trace back the entire chain of people moving. Finland has extremely detailed data on where people live. Bartu, Harjun, and Saarimaa (2023) look at the previous address of people who move into newly constructed housing, and then at the address of the people who moved into the vacated housing; and so on, six rounds of moves down the line. By the fourth round, half the people able to move to better housing came from below median income. Building 100 units of market rate housing in the center of Helsinki is the equivalent of building 66 units of housing in below median income neighborhoods, and the equivalent of 31 units of housing for the bottom quintile. Mast (2021) does the same exercise for the United States, though he has less precise data and must rely upon the characteristics of the neighborhood which people move from. Housing anywhere reduces the price of housing everywhere.

Nor does building cause a local increase in rents. There are a number of studies looking at the local impacts of new construction, and they unanimously find that construction reduces local rents. Asquith, Mast and Reed (2019) use a difference-in-differences approach in 11 major US cities with data from Zillow, exploiting idiosyncratic differences in when housing was completed. They found that nearby rents decreased between 5 and 7 percent, and that poor people are more likely to migrate to the area around the new construction. Kate Pennington (2020) uses building fires as an instrumental variable for new buildings. After a building burns down, developers build back better and create more housing than was there before. She can then show that the people who live nearby are actually 17% less likely to be displaced. Building more housing reduces gentrification. Andreas Mense (2025) uses weather shocks as an instrument, on the grounds that heavy rainfall and the intensity of winter frost will both delay construction, in some cases until the next year. (Most outdoor construction stops over the winter). What I appreciate about this paper is that the two instruments are demonstrably uncorrelated with each other. You might have guessed the results already, but building more housing is beneficial to everyone.

Affordable housing requirements are not merely a harmless addition. They are a tax on new units. We are requiring that developers, in order to build, must include unprofitable units of housing. This will have no effect on total number of units built if and only if the right to build units is assigned as part of a quota, and there are substantial monopoly profits to the builder. Given the number of firms engaged in building, it seems more likely that excess profits are small, and that affordable units reduce the total number constructed.

To put it plainly, affordable housing units are equivalent to rent control, which tends to increase the total cost of rent while causing massive misallocation. Calculating this is extremely difficult, because we can’t directly observe the housing that would have been built in the absence of rent control laws, but we do have some evidence. Diamond and McQuade (2018) exploit the 1994 rent control initiative in San Francisco to estimate the effects on rent. Landlords remove units from the market, causing rents to actually increase by 5%. (NB: the general equilibrium effects part was cut from the later published version in the AER.) A comprehensive review of the literature agrees – while rent control can transfer, it does so only at a considerable cost to everyone. I did come away from researching this question thinking that there had been considerably more work on the general equilibrium effects than there actually was.

So why is filtering so hard to make clear? Well, building housing is often confounded with shifts in the demand curve. When demand increases, developers try to provide more housing, but are prevented by land use restrictions. Thus, we have both a spate of building, and a rise in income. To link the building to rising rents is just totally wrong, though. It is like looking at people in a rainstorm carrying umbrellas, and deciding that since they are a bit damp the thing to do is to get rid of umbrellas entirely.

It might be best to argue for this by pointing first toward other goods, which people are more willing to entertain. If you claimed that America is suffering from an affordable car crisis, and that to solve this we need to manufacture old, used, cars, people would look at you a bit funny. Obviously, affordable cars are just the new cars of yesteryear; and when we had a negative shock to new car production during the pandemic, the price of used cars went up despite nothing about the supply of old cars changing! Housing works exactly the same way.

Could there be a justification for affordable housing? The only plausible argument is if people have positive spillovers, and more people in an area lead to more positive spillovers. In this world, any person choosing to get more space makes the price higher for everyone else, which leads to an inefficiently high number of people living further out into the city. Requiring housing to be affordable effectively functions as a tax on living space, and forces people into smaller apartments.

There are some problems with this. First of all, we don’t actually have a cheaper way of producing low-quality housing. It all costs about the same to make. The most efficient way to make cheap housing is to make new housing 80 years ago. Second, it’s not at all clear that affordable housing restrictions have net positive spillovers at all. Autor, Palmer, and Pathak (2014) study the unexpected ending of rent control by statewide ballot in Cambridge, Massachusetts. Ending rent control both considerably increased turnover, and greatly increased the value of properties in Cambridge, including the untreated ones. This isn’t quite the same as affordable housing requirements, but is close evidence.

Considering positive spillovers makes it clear that focusing solely on the price of housing is an error. It’s conceivable for reducing the barriers to building housing to actually increase the price. A city is valuable because everyone else is there. On some margins, allowing more people to move somewhere makes it more valuable to be there, and this outweighs the response of supply. We can imagine an empty field in Kansas with a house on it. Obviously, if that house were transported to the heart of New York City, or if another New York sprung up around, rent would be higher! Yet, this is only due to the place being so much better now to live in. Reducing restrictions on building housing will always be welfare improving. That reducing building restrictions empirically reduces rents is a happy coincidence, but is not necessary for landuse restrictions to be bad.

If we want to help low-income households, the way to do it is to help low income households. We should simply give them money, not mess with production.