Search for housing policies in any city. The AI will find recent legislation, identify sponsors, and assess whether it helps or hinders housing affordability.

City or Metro Area

Metropolitan Housing Policy Efficiency Dashboard

2011–2024  ·  U.S. Metro Areas  ·  ACS + Census Permits

---

Research Question

Which U.S. metropolitan areas show evidence that restrictive housing supply policy is causing measurably higher rent burdens and poverty rates than their economic fundamentals would predict — and by how much?

---

Data Notes

2020 is excluded from the analysis due to COVID-19. The pandemic caused severe disruptions to housing markets, permit activity, rent burden, and poverty rates that are not representative of underlying policy conditions. Including 2020 would distort the model's estimates and flag metros for pandemic effects rather than housing policy failures.

Data Sources

American Community Survey (ACS) 5-Year Estimates via NHGIS, 2011–2024. Metro-level variables include rent burden, poverty rate, median household income, population, renter household count, educational attainment, and mean commute time.

U.S. Census Building Permits Survey, 2011–2024. Annual new residential building permits at the CBSA (Core Based Statistical Area) level, used to measure housing supply activity.

---

Variable Construction

Permits per 1,000 Residents
Raw permit counts divided by population, scaled to 1,000. This normalizes supply across metro areas of different sizes, making Phoenix and Providence comparable.

Permits Lag (permits_lag2)
Permits per 1,000 lagged by 2 years. Housing takes time to build — a permit issued in 2020 typically adds supply to the market in 2021–2022. Using a 2-year lag better captures the causal relationship between permitting decisions and actual rent outcomes. This is why 2011 and 2012 cannot be analyzed (no prior permit data available).

Rent Burden
The share of renter households spending 30% or more of gross income on rent. This is the standard HUD definition of housing cost burden. Sourced directly from ACS table B25070.

Excess Rent Burden
The residual from Model 1 (Expected Rent Burden), clamped at zero from below. A positive value means the metro's actual rent burden is higher than the model predicts given its supply, income, education, and commute characteristics. This excess is interpreted as the component of rent burden attributable to policy failure rather than underlying demand fundamentals.

Excess Poverty / Residents Pushed Into Poverty
Estimated using Model 2 (Poverty Model). The model estimates how much of a metro's poverty rate is explained by its excess rent burden. Multiplying the excess poverty rate by renter household count gives an estimated headcount — the number of residents in poverty specifically due to housing cost overburden beyond what fundamentals predict.

Policy Inefficiency Score
A composite z-score index combining three standardized components:

1. Supply deficit — negative z-score of permits per 1,000 (lower supply = higher score)
2. Excess rent burden z-score — how far above average the metro's excess rent burden is
3. Excess poverty z-score — how far above average the metro's housing-driven poverty is

The three z-scores are summed. A higher score indicates worse overall policy performance. Negative scores mean the metro performs better than average on all three dimensions simultaneously.

Policy Flags
Metros scoring in the top 5% of the index are flagged RED (Highest Concern). Metros in the top 6–20% are flagged YELLOW (Moderate Concern). All others are GREEN (Lower Concern).

---

Econometric Models

Model 1 — Expected Rent Burden
Two-way fixed effects OLS (metro FE + year FE) predicting rent burden from lagged permits per 1,000, education share, and mean commute time. Metro fixed effects absorb time-invariant city characteristics (geography, land constraints, historical zoning). Year fixed effects absorb macro shocks (recessions, interest rate cycles). The residual from this model is excess rent burden.

rent_burden_it = α_i + λ_t + β₁ permits_lag2_it + β₂ bachelors_share_it + β₃ commute_time_it + ε_it

Why these controls? Education share (bachelors_plus_share) proxies for labor market demand — high-skilled cities attract workers and face higher housing demand. Mean commute time proxies for suburban substitutability — cities where workers can easily commute from far suburbs face less housing pressure in core areas. Both are included to isolate the supply effect on rent burden.

Model 2 — Poverty Model
Two-way fixed effects OLS predicting poverty rate from rent burden, education share, and commute time. This model quantifies the poverty cost of rent burden. The coefficient on rent_burden (~4.9) means a 1 percentage point increase in rent burden is associated with roughly a 4.9 percentage point increase in poverty rate, holding metro and year fixed effects constant.

Model 3 — Supply → Rent (Robustness)
A direct model of lagged permits on rent burden. Confirms that the permit lag coefficient is negative and significant — metros that build more housing in prior years have lower rent burden today, consistent with supply-demand theory.

Standard Errors
All models use standard errors clustered by CBSA code to account for serial correlation within metros over time.

---

Limitations

• Causality is suggestive, not definitive — permits are correlated with but do not fully capture zoning restrictiveness
• The 2-year permit lag is theoretically motivated but the true construction lag varies by project type and metro
• Excess rent burden may capture demand shocks not fully absorbed by year fixed effects (e.g. remote work migration post-2020)
• Poverty headcount estimates assume a linear relationship between rent burden and poverty, which may not hold at extremes