How we rank America's boom towns and declining counties.
The Boom Town Index is a composite score from 0 to 100 that identifies U.S. counties where economic strength outpaces housing cost — places where what you get outweighs what you pay. It's built on a machine learning model trained on two decades of government economic data across every U.S. county.
Unlike rankings that use arbitrary weighted averages, the BTI learns from historical data which factors actually predict county-level housing outperformance. Quality of life, job growth, affordability, and price momentum all contribute — but the model determines how much each matters, not us.
A score of 80 means the county ranks higher than 80% of all U.S. counties on projected long-term growth. It does NOT mean the county is "80% good" — it's a relative ranking.
We train two separate gradient boosted regression models. Both predict how much a county will outperform or underperform the national housing market (the residual), but over different time horizons. The models learn different lessons from the same data.
The BTI Score drives the main ranking. It predicts 5-year cumulative growth relative to national, which forces the model to weight structural fundamentals over short-term noise. Counties rank high when they offer strong quality of life, growing job markets, and affordable housing — not just because prices happened to spike recently.
| Category | Weight |
|---|---|
| Quality of Life (amenity, health, education, tax) | 39% |
| Price Momentum (3-month trend) | 25% |
| Jobs & Wages (QCEW employment, wage growth) | 13% |
| Housing Valuation (P/E ratio, rent, permits) | 10% |
| GDP Growth (BEA county GDP) | 8% |
Weights shown exclude national macro conditions (~19% of raw model importance), which affect all counties equally and don't influence relative rankings.
The 1-year forecast predicts next-year growth relative to national. Because short-term housing markets are heavily momentum-driven, this model leans on recent price trends and macro conditions. It's useful for seeing where prices are accelerating right now, but it tells you less about whether those gains are sustainable.
| Category | Weight |
|---|---|
| Price Momentum (3-month trend, YoY lags) | 61% |
| Quality of Life (amenity, health, education) | 13% |
| Housing Valuation (P/E ratio, rent, permits) | 9% |
| Jobs & Wages | 7% |
| GDP Growth | 5% |
Weights shown exclude national macro conditions (~33% of raw model importance).
We don't manually assign these weights. Both models learn from 20 years of historical data which factors matter most — and how they interact. The 5-year model independently discovers that quality of life and economic fundamentals matter more over longer horizons, while the 1-year model learns that momentum dominates short-term. The weights above are derived from the model's learned feature importances.
Both models are gradient boosted regression trees trained on every U.S. county × every year going back to 2005. For each historical data point, the model sees the county's recent economic and housing indicators, then learns to predict how much that county outperformed or underperformed the national average.
The key: both models predict the residual (county growth minus national growth), not the absolute price change. This forces them to learn what makes specific counties different — not just whether the national market is up or down.
National housing market conditions are included as a feature, allowing each model to learn how sensitive each county is to macro conditions. The residual prediction is then added to Zillow's forward-looking national forecast to produce a projected growth rate. All counties are ranked by the 5-year projection and converted to a 0–100 percentile score — the BTI Score.
Both models consider 22 economic, housing, and quality-of-life indicators for each county:
Three years of home price trajectory (Zillow ZHVI year-over-year changes), a 3-month momentum indicator (annualized short-term price change), the housing P/E ratio (home value relative to GDP per capita), and rent trends (Zillow ZORI year-over-year change).
Three years of county GDP growth (Bureau of Economic Analysis), two years of employment growth, and two years of wage growth (Bureau of Labor Statistics QCEW).
Two years of building permit trends (Census Bureau). New construction signals developer confidence in future growth, but oversupply can also depress prices.
Natural amenity score (USDA climate, topography, and water area index), effective property tax rate, educational attainment (% with bachelor's degree+), life expectancy, and uninsured rate (County Health Rankings). These structural factors capture permanent desirability and cost-of-living characteristics. They play a modest role in 1-year predictions but become the single largest driver over 5-year horizons.
Each county page shows three short-term projections based on the 1-year model:
These bracket the 1-year forecast with a simple ±1% range to account for uncertainty.
| Source | Provider | Frequency |
|---|---|---|
| Home Value Trends (ZHVI) | Zillow Research | Monthly |
| Rent Trends (ZORI) | Zillow Research | Monthly |
| County GDP (Real, Chained 2017$) | Bureau of Economic Analysis (BEA CAGDP9) | Annual |
| Employment & Wages | Bureau of Labor Statistics (QCEW) | Quarterly |
| Building Permits | U.S. Census Bureau (BPS) | Annual |
| Natural Amenities | USDA Economic Research Service | Static |
| Income, Housing, Population | U.S. Census Bureau (ACS 5-Year) | Annual |
| Property Taxes & Education | U.S. Census Bureau (ACS 5-Year) | Annual |
| Health Outcomes (Life Expectancy, Uninsured Rate) | County Health Rankings & Roadmaps | Annual |
| National Home Price Forecast | Zillow Research | Monthly |
All data is publicly available from U.S. government agencies and Zillow Research. We do not use proprietary or paywalled data sources. The model trains on data going back to 2005; current predictions use BEA GDP through 2022, BLS employment data through 2025 Q3, and Zillow data updated monthly.