About This Project
Oregon’s SB 608 (effective February 2019) caps annual rent increases at CPI + 7 percentage points on most units over 15 years old. California’s AB 1482 (effective January 2020) imposes a similar cap of CPI + 5 pp. This project estimates whether these policies changed housing permits, home prices, rents, affordability, residential mobility, and labor-market conditions — using only free, publicly available data from the Census Bureau, FHFA, and BLS, with eleven comparison states that had no statewide cap during the sample period.
Beyond statewide estimates, the project extends to building-level analysis in twelve cities. The strongest local package is New York City, where 89,100 buildings are tracked across six distinct modeling approaches using HPD housing violations and Rent Stabilized Building Lists.
Statewide Key Findings
Pooled event-study estimates for Oregon and California relative to 11 donor states. Values show the average post-treatment coefficient as a percentage of the treated pre-policy mean. Evidence quality varies by domain.
Strongest statewide outcomes: Housing supply (permits) and capitalization (FHFA HPI) have the cleanest pre-trends and most complete coverage. ACS rent and mobility outcomes are informative but limited by the missing 2020 survey year. QCEW employment and wage outcomes remain exploratory due to a short pre-period starting in 2014.
Statewide Cross-Outcome Summary
Post-treatment shift across all 12 outcome variables. Each bar shows the average post-treatment event-study coefficient as a percentage of the treated pre-policy mean.
Magnitudes are not directly comparable across domains — coverage windows, evidence quality, and pre-trend reliability differ. See States for full tables and robustness checks.
NYC Building-Level Results
Strongest local packageThe NYC extension links the official Rent Stabilized Building List (RGB, 2024) to the HPD violations database (NYC Department of Housing Preservation and Development) at the building level using borough-block-lot identifiers. HPD violations are code-enforcement records filed when inspectors find conditions like heat/hot-water failures, lead paint, pests, or structural defects. The analysis tracks 32,793 stabilized buildings against 56,307 non-stabilized controls across six modeling approaches, all since 2019.
Building Fixed-Effects Event Profile
Estimated differential in mean HPD violations per building-year between rent-stabilized and non-stabilized buildings, relative to the 2019 baseline. Building fixed effects absorb all time-invariant building characteristics. 89,100 buildings, 623,700 building-year observations. Cluster-robust standard errors.
All coefficients significant at p<0.001. The widening gap suggests stabilized buildings accumulated violations faster than non-stabilized controls, though this is design-sensitive and not a settled causal claim.
Mean Violations: Stabilized vs. Non-Stabilized Buildings
Average HPD violation count per building per year. “Stabilized” = appears on the 2024 RGB Rent Stabilized Building List. “Non-stabilized” = registered with HPD but not on the stabilized list.
NYC results are design-sensitive. Simpler models (borough-year) show larger effects; tightly matched designs (within-community-board, refined matching with 28,746 pairs) show effects near zero. The building-FE model falls between these extremes. These are not settled causal estimates.
Where the Gap Lives
2025 follow-onThe building-FE model shows a widening gap, but where does it concentrate? A second round of follow-on questions decomposes the 2025 result. These remain descriptive, design-sensitive findings—not causal claims.
2025 Monthly Violation Gap
Treated-minus-control mean violations per building by month. The gap is positive every month but compresses in June and November when control-building violations spike, likely reflecting seasonal enforcement cycles.
Narrower read: The gap concentrates in the Bronx, in buildings with recent registration activity (2025-plus bin: +5.1 gap), and in unresolved enforcement cases. The signal appears to come from a smaller set of heavily burdened stabilized buildings, not a uniform shift across all stabilized stock. These patterns narrow the interpretation but do not settle causality.
Statewide Analysis
Pooled TWFE event-study estimates for Oregon and California across 13 outcome variables from four federal data sources. Full robustness battery: placebo timing, donor-pool sensitivity, single-state models, state interactions, and quarterly timing checks.
City Extensions
Building-level analysis across twelve cities. NYC tracks 89,100 buildings through six modeling approaches. Oakland has a near-full administrative panel with 3,805 RAP case rows linked to 311 enforcement data. San Francisco, West Hollywood, and Los Angeles have upgraded inventory, linkage, and pilot packages.
Reproducibility
Every data source is free and federally published. Every download is scripted. The full pipeline — from raw data through panel construction, event-study estimation, and robustness battery — can be reproduced from scratch. See the GitHub repository for setup instructions and the complete test suite.