← Studies Suggest ⚖️ Policy

Rent Control Raised Rents. San Francisco's 1994 Expansion Cut Rental Supply 15% and Pushed Citywide Rents Up 5.1%

A quasi-experiment covering nearly every adult in San Francisco found rent control kept incumbent tenants housed while landlords pulled 15% of rental supply off the market, lifting citywide rents 5.1%. The authors' verdict: the landlord response essentially erased the tenants' gains.

By Marcus Reeves, Urban Policy - September 19, 2026

Ask a roomful of renters whether the government should cap rent increases and most hands go up. The logic feels airtight: rents are soaring, landlords hold the leverage, and a legal ceiling keeps people in their homes. It is one of America's most popular housing policies, and one of the most dreaded by economists, who have warned for fifty years that capping prices shrinks supply.

In 2019, three Stanford economists measured what happens when a city expands rent control, using a quirk of San Francisco history as a natural experiment. Their answer pleased nobody completely: rent control worked exactly as advertised for the tenants it covered, then landlords responded in ways that wiped out the gains for the city as a whole.

A Natural Experiment in San Francisco

The experiment began with a 1994 ballot initiative. The city's rent control law dated to 1979 and originally exempted small multifamily buildings of four units or fewer. The 1994 measure erased that exemption for buildings constructed before 1980. Overnight, tenants in older duplexes, triplexes, and fourplexes gained protection, while tenants in otherwise identical post-1980 buildings did not. That sharp cutoff handed researchers something close to a randomized trial, with treatment and control groups separated by a construction date neither tenants nor landlords chose.

The authors combined address-level migration records for nearly every adult in the city with parcel-by-parcel assessor data, tracking both groups for years. The design is a textbook difference-in-differences comparison, published in the American Economic Review and since cited hundreds of times.

It Worked, Until the Landlords Responded

On the tenant side, rent control delivered. Beneficiaries were 10 to 20 percent more likely to still be living at their 1994 address years later than the control group, especially older households and longtime residents. Mobility fell 20 percent and displacement declined. The authors valued the benefit at $2,300 to $6,600 per person annually. For a tenant who would otherwise have been priced out, that is not an abstraction. It is the difference between staying and leaving.

Then the landlords moved, cutting rental supply 15 percent by selling to owner-occupants, converting rentals to condos, and redeveloping. Some reinvested in renovations, which attracted higher-income tenants in an already stratifying city. The supply drain fed a citywide effect the authors' model puts at a 5.1 percent rent increase across San Francisco, including for renters the law never covered.

The welfare losses from the landlord response were almost as large as the benefits to the fortunate tenants, essentially erasing the gains. A policy designed to make housing cheaper made it more expensive for everyone except the incumbents it shielded.

The Math Nobody Ran

The paper never translates the 5.1 percent into a household budget, so let's do it with the assumptions stated plainly. Take a renter paying $2,500 a month, a plausible figure. A 5.1 percent increase is $127.50 a month, or $1,530 a year, paid by a renter the policy was never designed to help. Set that against the $2,300 to $6,600 a year flowing to each covered tenant, and the shape of the transfer comes into focus: concentrated gains for incumbents, diffuse costs spread across every future renter in the city, including the low-income newcomers the policy's supporters most want to protect.

Second: 15 percent of the treated rental supply vanished into owner-occupancy. Those units left the rental market young workers, immigrants, and anyone not ready to buy must use, and ownership conversions are a one-way door: once a rental becomes a condo, it almost never converts back.

The Strongest Counterargument

The defense of rent control deserves its full weight, because the study itself supplies most of it. First, it worked for the people it was meant to protect: displacement fell, and tenants gained thousands of dollars a year in housing security while keeping their homes. Telling a family that avoided eviction that the general-equilibrium welfare effects net to zero is technically correct and humanly beside the point.

Second, the landlord backlash ran through specific escape hatches. The Ellis Act lets owners exit the rental business entirely, and condo conversions offered a profitable off-ramp. A stricter law closing those exits would produce a smaller supply response. The 15 percent is a property of this law's design, not a law of physics. St. Paul, Minnesota learned the design lesson the hard way. Its 2021 ordinance capped increases at 3 percent with no new-construction exemption; developers walked away before the ink dried, and the city amended the law within a year.

Third, timing flatters the skeptics. The 1994 expansion landed just as the dot-com boom began detonating local rents. Difference-in-differences is built for exactly this problem, since the control group faced the same boom, but spillovers cut the other way: treated landlords' condo conversions raised rents for the control group too, distorting the comparison.

Fourth, the welfare arithmetic depends on whose welfare counts double. The model treats a dollar to a landlord and a dollar to a struggling tenant as equals. If you believe a dollar matters more to the tenant, the same numbers can justify the policy. And the authors' own preferred fix concedes the point. They argue tenant protections should be funded as government social insurance rather than an unfunded landlord mandate, keeping the stability gains without the supply damage.

What We Didn't Prove

Several limits deserve naming. This is one city, and San Francisco's housing supply was already among America's most constrained. The landlord response might be smaller where building is easier. The 5.1 percent citywide rent effect comes from a structural model with assumptions, softer than the directly measured 15 percent supply cut. The study measures staying put and dollars, not well-being, so it cannot say whether the tenants who stayed were better off in any broader sense. And it says nothing about the vacancy-decontrol choice that decides what happens to rents when a tenant leaves.

The Bottom Line

Rent control did exactly what its supporters promised for the tenants it covered, and exactly what its critics predicted for the housing market around them. Fifteen percent of the affected rental supply left the market while rents rose 5.1 percent citywide. The gains to incumbents were, in the authors' accounting, nearly erased by the costs imposed on everyone else. The uncomfortable implication is not that tenant protections are worthless, but that price caps without a supply plan tax the very people, future renters, the policy claims to serve.

What You Can Do

If you rent in a controlled unit, learn which protections actually cover you: exemption rules, Ellis Act eviction history, and condo-conversion ordinances are public records. If you vote on housing measures, ask the supply question first. Does the proposal exempt new construction? St. Paul's failure to do so killed projects before they started. Does it pair caps with funding, or just bill landlords and hope they absorb it? For policymakers, the evidence points toward targeted help: rental vouchers and direct subsidies reach low-income tenants without giving landlords a reason to shrink the rental stock. And if you are apartment hunting in a rent-controlled city, expect the misallocation economists document: incumbents stay put for decades, so available units skew toward the uncontrolled segment, where the 5.1 percent lives.