A research pass over every U.S. ZIP code tabulation area — correlations, what predicts yield, and six honest ZIP archetypes.
This page is generated entirely from zip_master.json — Census ACS,
Zillow ZHVI/ZORI, HUD FMR, BLS, IRS migration and FEMA National Risk Index.
No number here is hand-entered. Where a relationship is weak or null, it is
reported as weak or null.
Pairwise Pearson correlation across every numeric ZIP metric. The strongest links — both directions. Bar length is |r|.
Note: the very top positive pairs
(fema_risk_score ~ fema_eal_score, FMR ~ rent) are partly
definitional — FEMA's risk score is built from expected annual loss, and HUD
fair-market rent is itself a rent estimate. They are shown for completeness,
not as discoveries.
Gross yield = annual rent ÷ home value. OLS regression on the complete-case ZIPs, predictors standardized so coefficients compare directly (yield change per +1 standard deviation).
| Predictor | Std. coefficient | Univariate r |
|---|
The model explains roughly a quarter of ZIP-level yield variation — honest, and modest. Home value is the dominant predictor: cheap entry, not high rent, is what makes a high-yield ZIP. Most of the variance is left unexplained by these metrics.
k-means (k=6, numpy implementation) on home value, yield, income, FEMA risk, education, unemployment and population — magnitude features log-scaled, all winsorized at the 0.5/99.5 percentiles. Each name is taken from the two axes on which that cluster's centroid is most extreme. Names describe centroids, not narrative.
Cross-metric results that cut against intuition — each one a real correlation from the data.