Rent Control

What the evidence says

The clearest and most consistent finding is that rents on controlled units sit well below market: controlled units in Cambridge, MA rented 25-40 percent below nearby non-controlled properties, and reviews of New York City's controls describe a persistent "wedge" between controlled and uncontrolled rents, with quality-adjusted rents consistently higher in the uncontrolled sector across the studies reviewed. But the price effect runs the other way at the market level: when landlords respond to rent control by shrinking the units they offer, the resulting supply loss pushes up rents for everyone else. In San Francisco, landlords subject to an expansion of rent control cut rental housing supply by 15 percent, which caused a city-wide rent increase of 5.1 percent. Liberalizing rent-controlled units produces a similar market-level effect in reverse: in high-demand submarkets such as Amsterdam, rents "quickly spiral upward" once units are freed from protection, while rent-protected units see only incremental annual increases.

The dominant supply mechanism identified across studies is conversion out of the rental stock rather than abandonment of the units themselves. Following San Francisco's rent-control expansion, treated buildings were 8 percentage points more likely to convert to condominium or tenancy-in-common ownership, the number of owner-occupants at affected properties rose 8.1 percentage points, and the number of renters in rent-controlled apartments fell 24.6 percent relative to a 1990-1994 baseline. Running the same mechanism in reverse, Cambridge's 1994 decontrol raised the share of units offered as rental housing by nearly 6 percentage points and lifted condominiums' share of the housing stock by 32 percent over the following decade, driven "in large part" by conversion of existing apartments and houses. Consistent with this, rent regulation is described as making conversion to owner-occupation more profitable than continuing to rent, and rent control is noted as a potential constraint on tearing down existing renter-occupied buildings — though one study finds it does not raise the marginal cost of constructing a new unit.

Effects on construction and investment are contested rather than settled. Cambridge's decontrol was followed by roughly a doubling of aggregate annual real permitted building expenditures, from $21 million to $45 million per year. Against that, a review of the broader empirical literature on "second-generation" rent controls (controls paired with vacancy decontrol, as in Cambridge and San Francisco) concludes that many of their claimed effects are "imperceptible" and that no consensus exists on how New York City's controls affected maintenance or construction. Complicating the simple undermaintenance story further, San Francisco's rent-control expansion was followed by a 4.6 percent per-unit increase in renovation permit filings at treated buildings — though the source flags the mechanism as unclear, since renovation could reflect either maintenance or value-maximizing upgrades ahead of a conversion sale. On maintenance specifically, the literature is described as unable to uncover significantly higher maintenance levels in the uncontrolled sector, despite the theoretical expectation that landlords facing a rent freeze cut back on housing services.

Rent control decreases tenant mobility. There is described "widespread agreement" that New York City's controls caused mobility to fall in the controlled sector, though the same review calls the broader empirical evidence for lower average mobility only "weak." A stronger, quasi-experimental estimate from San Francisco puts a number on the same effect: beneficiaries of rent control were 10-20 percentage points more likely to still be at their 1994 address in the medium-to-long term than the control group.

Landlord behavior in response to local demand shocks shows rent control reshaping which supply-reduction channel landlords use rather than eliminating supply reduction. Under San Francisco's cap (annual increases limited to 60 percent of regional CPI), controlled landlords could not fully price a positive demand shock (proximity to a new transit shuttle stop) into sitting tenants' rents, and responded by increasing market-withdrawal filings; when landlords instead faced added barriers to withdrawing units from the market, they substituted toward at-fault evictions. Large shocks pushed landlords toward supply-reducing no-fault evictions targeting one tenant per building, while smaller, more easily priced-in shocks instead produced supply-positive responses — more repair and unit-addition ("up-conversion") permit filings, with no measured rise in evictions. This response was strongest among mid-size (3-6 unit) buildings, led very small (2-unit) buildings to withdraw entirely from the controlled market, and was substantially weaker for buildings built before 1906.

Cross-national comparison and social-housing-specific studies round out the picture but rest on weaker designs. Countries with tightly regulated rents and well-regulated landlords, such as Germany and Switzerland, are associated with large private rental sectors serving low- and middle-income households with no measured negative effect on overall supply — an associational, cross-country comparison rather than a causal estimate. Within regulated systems, turnover still lets landlords raise rents substantially for new tenants even where sitting-tenant rents are constrained. In Argentina, a 2020 law strengthening rental regulation was followed by 45 percent of landlords exiting the rental market, and the same regulatory regime's effect on real rent growth over 1914-2017 was statistically significant only in the short, high-inflation periods following the two World Wars, with no measured linear effect across the full period. In social housing, modeled rent-control strategies narrow the affordability gap between new and long-tenured tenants, but the same modeling finds only modest overall affordability gains and concludes rent price alone is "too crude an instrument" to substantially improve renters' financial situations.

Where the evidence splits

Whether rent control meaningfully suppresses construction and investment is unresolved on this page. Autor's quasi-experimental estimate finds Cambridge's 1994 decontrol roughly doubled annual permitted building expenditures ($21 million to $45 million), consistent with rent control having suppressed investment beforehand. Arnott's review of the broader empirical literature, by contrast, concludes that many claimed effects of these same "second-generation" controls are "imperceptible" and that no consensus exists on how New York City's controls affected construction. The claims do not explain this gap directly, but Arnott's assessment is explicitly a review of prior empirical literature ("my reading of the empirical evidence") rather than a new quasi-experimental estimate, while Autor's and Diamond's numbers come from natural-experiment designs (Cambridge's decontrol, San Francisco's control expansion) — a difference in method the sources establish even if they don't cross-reference each other.

Maintenance and renovation behavior under rent control also point in different directions. Theory and the Arnott review predict landlords "cut back on the quantity of housing services provided" under a rent freeze, and the same review reports the empirical literature has been unable to confirm significantly higher maintenance in the uncontrolled sector. Diamond's San Francisco estimate instead finds a 4.6 percent per-unit increase in renovation permit filings at rent-control-treated buildings — the opposite of an undermaintenance signal. The source itself flags the mechanism as unresolved, noting the increase could reflect landlord value-maximization ahead of a conversion sale rather than improved upkeep for existing tenants, so the split is not reconciled by the sources on this page.

The relationship between regulation strength and rental-sector size also splits by level of analysis. Hochstenbach's cross-national comparison associates tightly regulated rents and well-regulated landlords (Germany, Switzerland) with large, stable private rental sectors and no measured negative effect on supply. Diamond's and Autor's causal, building-level estimates instead find rent control reduces the rental stock at treated properties through condo/TIC conversion (8 percentage points in San Francisco) and owner-occupancy conversion (32 percent condo-stock increase in Cambridge after decontrol). The difference is one of scope and design: Hochstenbach's claim is a national-level associational comparison across different regulatory regimes, while Diamond's and Autor's are causal, within-city estimates of specific rent-control policy changes — the sources do not otherwise reconcile the two findings.

Thresholds and dose-response

Where a legal cap on renewal-rent increases (25 percent over a two-year cycle, under Turkey's July 2022-June 2024 policy) fell below the rate landlords could charge new tenants, the gap widened sharply with inflation: cumulative new-contract rent growth exceeded the permitted two-year renewal path by 59 log points in Istanbul and by 104-108 log points in the Black Sea strata.

Regulatory strength and novelty change the sign of the measured effect on rent growth: rental market regulations showed a statistically significant negative effect on real rent growth only in the short periods following World War I and World War II, "when regulations were novel and particularly strong"; across the full 1914-2017 period in a linear model, none of the regulation indices were statistically significant.

Whether a demand shock can be priced into a capped rent changes the direction of landlords' supply response: a small price shock that can be largely priced into rents produced supply-positive responses (more unit-addition and repair permit filings, no increase in evictions), while a large price shock that could not be fully priced in instead produced supply-reducing no-fault evictions and market-withdrawal filings.

Open empirical questions

• Whether the 4.6% renovation-permit increase at San Francisco's rent-control-treated buildings reflects genuine maintenance improvement or value-maximizing renovation ahead of a conversion sale is not established.
• Whether "second-generation" rent control (paired with vacancy decontrol) has different long-run construction and maintenance effects than the "first-generation" controls Arnott describes as "almost certainly" harmful is not directly measured on this page.
• What numeric size of demand shock separates a "small" shock (priced into rents, supply-positive landlord response) from a "large" shock (not priced in, supply-reducing eviction response) is not quantified.
• Whether the modest affordability gains modeled for social-housing rent-control strategies would hold under actual implementation, rather than in a forecasting model, is untested.