Does Upzoning Actually Build Housing? A Difference-in-Differences Study of U.S. Zoning Reform Outcomes

Research Question

Since 2018, a wave of residential zoning reforms has swept across U.S. cities: Minneapolis eliminated single-family zoning in 2019, Oregon passed a statewide middle housing mandate in 2021, California legalized ADUs statewide and passed SB 9/10 in 2021, and dozens of cities have adopted ADU ordinances, missing middle legalization, or corridor upzoning. Policymakers argue that relaxing land use restrictions will increase housing supply. But will it?

This study asks: Do cities that adopt upzoning reforms actually produce more permitted housing units? And which reform types and design features predict larger supply responses?

Literature Review

The theoretical link between zoning and housing supply is well-established. Glaeser, Gyourko, and Saks (2005) documented that zoning constraints impose significant "regulatory taxes" on housing production in coastal markets, and that relaxing constraints should lower prices and increase supply. Hsieh and Moretti (2019) estimated that housing supply restrictions in high-productivity cities cost the U.S. economy approximately 2% of GDP annually in misallocated labor.

Empirically, however, the supply response to upzoning is contested. Pennington (2021) found that upzoning in San Francisco increased land values without increasing supply in the near term — a result attributed to speculative land price effects. Asquith, Mast, and Reed (2023) found that new apartment construction in low-income neighborhoods modestly reduces nearby rents, suggesting supply does reach cost-sensitive households. Freemark (2020) examined Chicago's 2013 transit-adjacent upzoning and found limited near-term construction effects, underscoring that zoning reform is necessary but not sufficient without favorable construction economics and streamlined permitting.

This study contributes by examining a broader cross-section of reform types across 35 cities, with explicit attention to treatment effect heterogeneity.

Data

The analysis uses the following sources:

• HUD Building Permit Survey: Annual and monthly permit counts by structure type (single-family, 2–4 units, 5+ units) for cities and metro areas, 2010–2025
• Local Zoning Reform Database: Hand-coded dataset of reform adoption dates, types (ADU, missing middle, corridor upzoning, citywide upzoning), and design features (setback reductions, parking minimum elimination, ministerial/by-right approval) across 35 U.S. cities
• Census Bureau / ACS: City-level population, median income, renter share, and existing housing stock characteristics
• FRED: State-level construction cost indices and 30-year mortgage rates for time-varying controls
• Zillow Research: City-level rent index (ZORI) and home value index (ZHVI) for price environment controls

The final panel covers 35 cities across 20 states, each with at least one qualifying reform adopted between 2015 and 2023 and a minimum 24-month post-reform observation window.

Methodology

The primary identification strategy is a staggered difference-in-differences (DiD) design. Cities are treated when they adopt a qualifying zoning reform; control cities are those in the same state or region that have not yet adopted a comparable reform. The outcome variable is annual permit count per 1,000 existing housing units, normalized for city size.

Given staggered adoption timing, we apply the Callaway-Sant'Anna (2021) heterogeneity-robust DiD estimator, which avoids the contamination bias associated with two-way fixed effects when treatment effects vary across cohorts. This produces cohort-specific average treatment effects (ATTs) for each adoption-year group.

We complement the DiD with an event study specification plotting the dynamic treatment effect from 3 years pre-reform to 5 years post-reform. The absence of pre-trend divergence serves as a validity check on the parallel trends assumption.

To examine heterogeneity, we interact the treatment indicator with city-level moderators: initial land costs (ZHVI level), construction cost index, baseline vacancy rate, and whether the reform included by-right permitting approval. We hypothesize that supply responses will be larger where land costs are moderate, construction economics are favorable, and permitting is streamlined.

Preliminary Findings

Descriptive event studies reveal marked heterogeneity across reform types. ADU-only reforms produce the fastest supply response: permitted ADU counts increased 40–120% within two years of adoption in cities like Seattle, San Jose, and Los Angeles. However, ADU production remains a small share of total housing need and tends to concentrate in higher-income neighborhoods. Missing middle reforms show a more modest and delayed response, with the largest effects in smaller cities where construction economics are more favorable.

Cities that combined upzoning with streamlined by-right permitting approval show permit counts approximately 25% higher in the post-reform period than cities that adopted zoning changes without permitting reform — suggesting administrative friction is a binding constraint independent of zoning legality. Cities with very high pre-reform land costs (San Francisco, NYC) show small or negligible supply responses, consistent with the hypothesis that zoning reform is insufficient when construction economics are fundamentally unfavorable.

Conclusion and Implications

The evidence is suggestive that upzoning reforms do increase permitted housing in cities where construction economics are feasible — but the effect is heterogeneous, delayed, and sensitive to complementary policy design. Zoning reform is a necessary but not sufficient condition for supply response. Cities seeking meaningful production gains should pair upzoning with by-right permitting, impact fee reductions, and infrastructure investment, rather than treating land use reform as a standalone solution.

For policymakers, the practical implication is that context matters enormously. The same reform that unlocks 1,000 units in Austin may unlock 50 in San Francisco. Understanding local construction economics, land costs, and permitting friction is essential to calibrating realistic expectations for supply response timelines.

Limitations and Future Research

The hand-coded reform database introduces subjectivity in classifying reform scope and design features. The parallel trends assumption may be violated if cities adopt reforms in response to rising rents that are themselves correlated with supply. Future research should apply synthetic control methods to construct more credible counterfactuals for high-profile reforms like Minneapolis and Oregon where clean control groups are difficult to identify. It would also be valuable to track whether permitted units are actually built and occupied within 5 years — construction may be permitted but stalled due to financing constraints or development economics that shift after permit issuance.