HousingGauge research
ZIP-code divergence and convergence: what history shows
Research run convergence-v3-2026-10-09 · model convergence-v3 · price appreciation only (not total investor return)
Findings
- FHFA repeat-sales index (annual). After unusually high divergence, the next 5 years' dispersion was lower in 73% of 791 metro-years (after low divergence: lower in 32% of 935) — consistent with dispersion reverting to its typical level from either side. Prior 5-year laggards versus leaders over the next 5 years: not distinguishable from zero: +0.2 pp a year (95% interval −0.1 to +0.6; 3507 metro-years). Low-priced, weak-performing ZIPs versus the metro: not distinguishable from zero: +0.2 pp a year (95% interval −0.0 to +0.4; 1998 metro-years). Lower-priced ZIPs within 5 miles of affluent ZIPs versus farther away: outperformed: +0.4 pp a year (95% interval +0.2 to +0.6; 1892 metro-years).
- Zillow Home Value Index (year-end). After unusually high divergence, the next 5 years' dispersion was lower in 77% of 474 metro-years (after low divergence: lower in 23% of 172) — consistent with dispersion reverting to its typical level from either side. Prior 5-year laggards versus leaders over the next 5 years: not distinguishable from zero: +0.3 pp a year (95% interval −0.3 to +1.0; 1524 metro-years). Low-priced, weak-performing ZIPs versus the metro: not distinguishable from zero: +0.3 pp a year (95% interval −0.2 to +0.8; 1512 metro-years). Lower-priced ZIPs within 5 miles of affluent ZIPs versus farther away: not distinguishable from zero: +0.2 pp a year (95% interval −0.2 to +0.6; 1444 metro-years).
- Catch-Up Potential: the predetermined rule selects ZIP codes that lagged their region, are priced below its median, sit within 5 miles of a top-quartile-price ZIP code, at a time of high regional divergence. It did not pass every validation check fixed in advance, so HousingGauge does not publish a predictive catch-up label; ZIP pages show descriptive facts only. Where the rule did well, it is a candidate to re-test as more years of data arrive, not a finding.
- FHFA repeat-sales index (annual): convergence rule +1.1 pp/yr vs the metro over 5 years (403 metro-years; interval +0.6 to +1.5); above the random-selection band; interval above zero; starts through 2010: +0.4 pp/yr, interval above zero; starts after 2010: +1.6 pp/yr, interval above zero.
- Zillow Home Value Index (year-end): convergence rule +1.3 pp/yr vs the metro over 5 years (293 metro-years; interval +0.3 to +1.9); above the random-selection band; interval above zero; starts through 2010: +0.4 pp/yr, not testable (too few start years for an interval); starts after 2010: +1.4 pp/yr, interval above zero.
- Turning points: the predetermined rule selects ZIP codes that trailed their region for 5 years (bottom quarter, at least 1 point a year behind), were still losing ground a year earlier and have kept pace or better since, while the region and the nation were rising. It did not pass every validation check fixed in advance, so HousingGauge shows turning points as descriptions of what happened, not as a signal of what comes next.
- FHFA repeat-sales index (annual): turning-point rule −0.2 pp/yr vs the region over 3 years (2499 metro-years; interval −0.3 to −0.1); not above the random-selection band (−0.1 to −0.0); interval includes or is below zero; halves 1981–2001: −0.2 pp/yr, interval below zero; 2002–2022: −0.2 pp/yr, interval includes zero.
- Zillow Home Value Index (year-end): turning-point rule +0.9 pp/yr vs the region over 3 years (954 metro-years; interval +0.2 to +1.5); above the random-selection band (−0.1 to +0.1); interval above zero; halves 2006–2014: +1.5 pp/yr, not testable; 2015–2022: +0.6 pp/yr, interval includes zero.
- Turns confirmed by new businesses: the same rule, kept only where the ZIP's count of business establishments grew faster than its region's over the latest three years published at the time. It did not pass every validation check fixed in advance, so business and income growth are shown as neighborhood facts, not as confirmation of a signal.
- FHFA repeat-sales index (annual): turn confirmed by new businesses −0.2 pp/yr vs the region over 3 years (1306 metro-years; interval −0.4 to +0.0); not above the random-selection band (−0.2 to +0.0); interval includes or is below zero; halves 1999–2010: −0.1 pp/yr, interval includes zero; 2011–2022: −0.3 pp/yr, interval below zero.
- Zillow Home Value Index (year-end): turn confirmed by new businesses +0.9 pp/yr vs the region over 3 years (709 metro-years; interval +0.3 to +1.5); above the random-selection band (−0.2 to +0.1); interval above zero; halves 2006–2014: +1.7 pp/yr, not testable; 2015–2022: +0.4 pp/yr, interval above zero.
Data and design
- FHFA repeat-sales index (annual): 8,965 ZIP codes in 97 regions, values 1975–2025; start years 1978–2025.
- Zillow Home Value Index (year-end): 9,566 ZIP codes in 97 regions, values 2000–2025; start years 2003–2025.
- Point in time. Each start year uses only values at or before it: prior appreciation, price levels, the region's own divergence history, lagged metro employment and the mortgage rate. Outcomes are realized values; a start year contributes to a horizon only when the full horizon has been observed (a 2022 start has no five-year outcome).
- Fixed rules. Quartiles, the high-divergence cut (≥75th percentile of the region's own prior history), "affluent" (top 25% of ZIP prices), adjacency (5 miles between ZIP centers) and regimes were set before any result was computed; nothing was fitted, so every start year is out of sample. Results are also shown for start years up to and after 2010.
- Averaging and uncertainty. Each region-year counts once. Intervals are 95% circular block bootstrap intervals over start years (block = horizon), which allow for overlapping windows and shared national shocks; with fewer than eight start years no interval is reported. A random-selection null shows what picking ZIPs at random would have produced.
- Relative vs absolute. "Excess" is a ZIP's annualized appreciation minus its region's housing-weighted benchmark over the same years. A ZIP rising 4% a year while its metro rises 8% did not catch up.
A. Does high divergence come before convergence?
Dispersion of H-year ZIP appreciation in the H years before each start year versus the H years after, for the same ZIPs. Negative change = convergence.
FHFA repeat-sales index (annual)
| Starting divergence | 1-yr: converged / mean change, pp | 3-yr: converged / mean change, pp | 5-yr: converged / mean change, pp |
|---|---|---|---|
| high | 66% / −0.65 (−1.3 to −0.2; n=622) | 74% / −0.54 (−0.8 to −0.3; n=680) | 73% / −0.37 (−0.5 to −0.2; n=791) |
| moderate | 53% / −0.03 (−0.2 to +0.2; n=1681) | 52% / −0.01 (−0.2 to +0.1; n=1562) | 50% / +0.02 (−0.1 to +0.1; n=1296) |
| low | 49% / +0.10 (−0.1 to +0.3; n=1488) | 37% / +0.19 (+0.0 to +0.4; n=1159) | 32% / +0.25 (+0.1 to +0.4; n=935) |
Zillow Home Value Index (year-end)
| Starting divergence | 1-yr: converged / mean change, pp | 3-yr: converged / mean change, pp | 5-yr: converged / mean change, pp |
|---|---|---|---|
| high | 75% / −0.97 (−1.3 to −0.7; n=527) | 73% / −0.63 (−0.8 to −0.4; n=518) | 77% / −0.69 (−0.9 to −0.5; n=474) |
| moderate | 51% / +0.15 (−0.1 to +0.4; n=756) | 54% / −0.11 (−0.3 to +0.1; n=626) | 43% / +0.11 (−0.2 to +0.4; n=393) |
| low | 32% / +0.59 (+0.4 to +0.9; n=533) | 32% / +0.44 (+0.3 to +0.6; n=283) | 23% / +0.46 (+0.3 to +0.6; n=172) |
High divergence is defined against each region's own history, so some convergence afterward is expected from regression to the mean alone; the low-divergence row is the comparison. Growth convergence is not price-level convergence: dispersion can fall while every ZIP falls together, or while gaps in price levels persist.
B. Do lagging ZIP codes later outperform?
ZIPs grouped within each region by 5-year appreciation relative to the region (Q1 = weakest). Forward excess vs the region, with 95% intervals.
FHFA repeat-sales index (annual)
| Prior group | 1-yr excess, pp/yr | 3-yr excess, pp/yr | 5-yr excess, pp/yr | 5-yr absolute, %/yr | Metro-years |
|---|---|---|---|---|---|
| Q1 laggards | +0.8 (+0.5 to +1.2) | +0.2 (+0.0 to +0.4) | +0.2 (+0.0 to +0.3) | +4.3 | 3507 |
| Q2 | −0.1 (−0.1 to +0.0) | −0.1 (−0.2 to +0.0) | −0.1 (−0.1 to +0.0) | +4.1 | 3507 |
| Q3 | −0.3 (−0.4 to −0.1) | −0.1 (−0.2 to −0.0) | −0.1 (−0.2 to −0.0) | +4.0 | 3507 |
| Q4 leaders | −0.5 (−0.9 to −0.2) | −0.1 (−0.3 to +0.2) | −0.1 (−0.3 to +0.1) | +4.1 | 3507 |
| Q1 minus Q4 | +1.3 (+0.7 to +2.0) | +0.3 (−0.2 to +0.7) | +0.2 (−0.1 to +0.6) | +0.2 | 3507 |
Zillow Home Value Index (year-end)
| Prior group | 1-yr excess, pp/yr | 3-yr excess, pp/yr | 5-yr excess, pp/yr | 5-yr absolute, %/yr | Metro-years |
|---|---|---|---|---|---|
| Q1 laggards | −0.3 (−0.6 to +0.0) | +0.1 (−0.3 to +0.5) | +0.3 (−0.0 to +0.7) | +4.4 | 1524 |
| Q2 | −0.1 (−0.3 to +0.1) | −0.1 (−0.3 to +0.2) | −0.0 (−0.2 to +0.2) | +4.0 | 1524 |
| Q3 | +0.0 (−0.1 to +0.2) | −0.1 (−0.3 to +0.1) | −0.1 (−0.4 to +0.0) | +3.9 | 1524 |
| Q4 leaders | +0.5 (+0.2 to +0.9) | +0.1 (−0.3 to +0.6) | −0.1 (−0.4 to +0.3) | +4.0 | 1524 |
| Q1 minus Q4 | −0.8 (−1.4 to −0.2) | −0.1 (−0.9 to +0.7) | +0.3 (−0.3 to +1.0) | +0.3 | 1524 |
At a one-year horizon the two sources disagree (FHFA repeat-sales index (annual): +1.3; Zillow Home Value Index (year-end): −0.8). Zillow's index is smoothed, which builds short-term momentum into consecutive years, so one-year results are the least reliable here; multi-year horizons are the comparison that matters.
Laggards minus leaders by condition at the start (5-year lookback, 5-year horizon):
FHFA repeat-sales index (annual)
| Condition | Q1 minus Q4, pp/yr | 95% interval | Metro-years |
|---|---|---|---|
| high divergence | +0.7 | −0.0 to +1.5 | 791 |
| moderate divergence | +0.1 | −0.2 to +0.3 | 1296 |
| low divergence | −0.1 | −0.3 to +0.2 | 935 |
| boom | +0.7 | −0.2 to +1.3 | 455 |
| correction | +0.9 | +0.7 to +1.5 | 653 |
| affordability shock | +0.3 | — | 137 |
| recovery | +0.7 | +0.2 to +1.0 | 569 |
| normal | −0.3 | −0.6 to −0.0 | 1693 |
| start ≤ 2010 | +0.1 | −0.2 to +0.6 | 2537 |
| start > 2010 | +0.4 | +0.0 to +0.9 | 970 |
Zillow Home Value Index (year-end)
| Condition | Q1 minus Q4, pp/yr | 95% interval | Metro-years |
|---|---|---|---|
| high divergence | +0.5 | −0.9 to +1.7 | 474 |
| moderate divergence | +0.3 | −0.5 to +1.1 | 393 |
| low divergence | −0.1 | −0.4 to +0.1 | 172 |
| boom | −0.3 | −1.6 to +0.5 | 269 |
| correction | +0.7 | +0.5 to +1.4 | 454 |
| affordability shock | — | — | 0 |
| recovery | +1.5 | −0.2 to +2.0 | 322 |
| normal | −0.4 | −0.6 to −0.0 | 479 |
| start ≤ 2010 | +0.5 | — | 557 |
| start > 2010 | +0.3 | −0.8 to +1.3 | 967 |
C. Does starting affordability matter?
Price level relative to the region (below or above its median) crossed with prior relative appreciation (bottom or top half).
FHFA repeat-sales index (annual)
| Cohort | 1-yr excess | 3-yr excess | 5-yr excess | 5-yr, high divergence | 5-yr, low divergence |
|---|---|---|---|---|---|
| low-priced, weak prior | +0.7 (+0.4 to +1.0) | +0.2 (+0.0 to +0.5) | +0.2 (−0.0 to +0.4) | +0.4 | −0.0 |
| low-priced, strong prior | −0.5 (−0.9 to −0.2) | −0.1 (−0.5 to +0.3) | −0.1 (−0.5 to +0.3) | −0.2 | −0.2 |
| high-priced, weak prior | +0.2 (−0.1 to +0.5) | −0.1 (−0.4 to +0.3) | −0.1 (−0.5 to +0.2) | −0.1 | +0.0 |
| high-priced, strong prior | −0.5 (−0.8 to −0.2) | −0.2 (−0.5 to +0.1) | −0.2 (−0.6 to +0.2) | −0.4 | +0.1 |
Zillow Home Value Index (year-end)
| Cohort | 1-yr excess | 3-yr excess | 5-yr excess | 5-yr, high divergence | 5-yr, low divergence |
|---|---|---|---|---|---|
| low-priced, weak prior | −0.0 (−0.4 to +0.4) | +0.2 (−0.3 to +0.6) | +0.3 (−0.2 to +0.8) | +0.7 | +0.3 |
| low-priced, strong prior | +0.5 (+0.1 to +0.8) | +0.3 (−0.3 to +0.9) | +0.3 (−0.3 to +0.8) | +0.5 | +0.4 |
| high-priced, weak prior | −0.3 (−0.7 to +0.1) | −0.2 (−0.9 to +0.5) | −0.2 (−0.8 to +0.4) | −0.6 | −0.6 |
| high-priced, strong prior | −0.1 (−0.4 to +0.2) | −0.3 (−0.7 to +0.3) | −0.4 (−0.9 to +0.1) | −0.7 | −0.5 |
D. Does proximity to affluent ZIP codes matter?
Lower-priced ZIPs (below the region median) whose center is within 5 miles of a top-quartile-price ZIP, minus lower-priced ZIPs farther away.
FHFA repeat-sales index (annual)
| Condition | 1-yr, pp/yr | 3-yr, pp/yr | 5-yr, pp/yr | Metro-years (5-yr) |
|---|---|---|---|---|
| all | +0.3 (+0.1 to +0.5) | +0.4 (+0.1 to +0.6) | +0.4 (+0.2 to +0.6) | 1892 |
| rising metro prices | +0.3 (+0.1 to +0.6) | +0.3 (+0.0 to +0.7) | +0.3 (+0.1 to +0.6) | 1449 |
| falling metro prices | +0.3 (−0.1 to +0.7) | +0.5 (+0.3 to +0.7) | +0.6 (+0.1 to +0.8) | 443 |
| stronger job growth | +0.4 (+0.2 to +0.8) | +0.4 (+0.1 to +0.8) | +0.4 (+0.1 to +0.8) | 899 |
| weaker job growth | +0.2 (−0.1 to +0.4) | +0.3 (+0.0 to +0.5) | +0.4 (+0.2 to +0.5) | 990 |
| high divergence | +0.5 (+0.1 to +0.8) | +0.4 (+0.1 to +0.7) | +0.4 (+0.2 to +0.7) | 623 |
Zillow Home Value Index (year-end)
| Condition | 1-yr, pp/yr | 3-yr, pp/yr | 5-yr, pp/yr | Metro-years (5-yr) |
|---|---|---|---|---|
| all | +0.1 (−0.2 to +0.5) | +0.2 (−0.3 to +0.6) | +0.2 (−0.2 to +0.6) | 1444 |
| rising metro prices | +0.4 (+0.0 to +0.7) | +0.3 (−0.2 to +0.8) | +0.2 (−0.3 to +0.8) | 1014 |
| falling metro prices | −0.5 (−0.8 to −0.1) | −0.2 (−0.5 to +0.4) | +0.1 (−0.2 to +0.5) | 430 |
| stronger job growth | +0.2 (−0.1 to +0.6) | +0.3 (−0.1 to +0.8) | +0.3 (−0.1 to +0.8) | 714 |
| weaker job growth | +0.0 (−0.3 to +0.4) | +0.0 (−0.4 to +0.4) | +0.1 (−0.3 to +0.5) | 727 |
| high divergence | +0.4 (−0.1 to +0.9) | +0.4 (+0.1 to +0.8) | +0.3 (+0.1 to +0.6) | 459 |
Rental demand, supply constraints and income growth could not be tested point-in-time with the data available for this run (ZIP rents start in 2015; ZIP permits and annual ZIP incomes are not available).
E. When does catch-up fail?
Share of prior laggards (Q1) still trailing their region over the horizon, by condition at the start.
FHFA repeat-sales index (annual)
| Laggards | 1-yr still trailing | 3-yr still trailing | 5-yr still trailing | Metro-years (5-yr) |
|---|---|---|---|---|
| all laggards | 49% | 51% | 50% | 3507 |
| metro correction at start | 42% | 43% | 42% | 653 |
| no correction at start | 50% | 52% | 52% | 2854 |
| lag deepened in the last year | 46% | 50% | 50% | 3498 |
| lag narrowed in the last year | 55% | 53% | 52% | 3216 |
| low-priced | 44% | 45% | 45% | 1947 |
| high-priced | 48% | 52% | 52% | 1911 |
Zillow Home Value Index (year-end)
| Laggards | 1-yr still trailing | 3-yr still trailing | 5-yr still trailing | Metro-years (5-yr) |
|---|---|---|---|---|
| all laggards | 55% | 51% | 47% | 1524 |
| metro correction at start | 51% | 46% | 43% | 454 |
| no correction at start | 57% | 52% | 49% | 1070 |
| lag deepened in the last year | 60% | 54% | 50% | 1521 |
| lag narrowed in the last year | 41% | 39% | 38% | 1346 |
| low-priced | 52% | 46% | 43% | 1455 |
| high-priced | 58% | 57% | 58% | 1409 |
Population decline, vacancy, insurance costs and infrastructure could not be tested point-in-time with this run's data; they remain open questions, not findings.
T. Turning points: a long lag, then a turn
Each ZIP's relative price (its value divided by a chain-linked, housing-weighted index of its region's ZIPs) is classified at every year end with the same code the site uses today. "Lagged": bottom quarter of the region on relative growth over the 5 years ending a year earlier, and at least 1 point a year behind. "Turning up": lagged, relative price still falling over the year before, flat or rising over the last year. The rule adds rising regional and national prices (12-month growth above zero). Outcomes are the annualized change in relative price over the next 1, 3 and 5 years (pp a year vs the region).
FHFA repeat-sales index (annual) (5-year pent-up window)
| Selection | 1-yr vs region, pp/yr | 3-yr vs region, pp/yr | 5-yr vs region, pp/yr | 3-yr absolute, %/yr | Metro-years | |
|---|---|---|---|---|---|---|
| turning-point rule | −0.85 (−1.0 to −0.7) | −0.20 (−0.3 to −0.1) | −0.07 (−0.1 to +0.0) | +5.4 | 2499 | |
| turning-point rule | first half (1981–2001) | −0.68 (−0.8 to −0.5) | −0.25 (−0.3 to −0.2) | −0.14 (−0.2 to −0.1) | +4.7 | 1160 |
| turning-point rule | second half (2002–2022) | −0.99 (−1.2 to −0.7) | −0.17 (−0.4 to +0.1) | −0.01 (−0.1 to +0.1) | +5.9 | 1339 |
| random selection (null) | −0.01 (−0.1 to +0.1) | −0.06 (−0.1 to −0.0) | −0.07 (−0.1 to −0.0) | — | 2508 | |
| turning up | −0.76 (−1.0 to −0.5) | −0.19 (−0.3 to −0.0) | −0.08 (−0.1 to −0.0) | +4.1 | 3180 | |
| turning up | region beating the nation | −0.78 (−1.0 to −0.5) | −0.09 (−0.3 to +0.2) | −0.01 (−0.1 to +0.1) | +5.3 | 1455 |
| turning up | divergence settled after a peak | −0.58 (−0.8 to −0.3) | −0.16 (−0.3 to +0.0) | −0.06 (−0.2 to +0.0) | +3.6 | 1301 |
| turning up | region or nation falling | −0.41 (−1.1 to +0.4) | −0.12 (−0.5 to +0.4) | −0.11 (−0.2 to +0.1) | −0.4 | 681 |
| turning up | stronger half | −0.88 (−1.1 to −0.6) | −0.19 (−0.3 to −0.0) | −0.05 (−0.1 to +0.0) | +4.4 | 2866 |
| turning up | weaker half | −0.33 (−0.6 to −0.0) | −0.17 (−0.3 to +0.0) | −0.18 (−0.3 to −0.1) | +3.0 | 1550 |
| lagged | −0.05 (−0.2 to +0.1) | −0.02 (−0.2 to +0.2) | +0.02 (−0.1 to +0.1) | +4.3 | 3506 | |
| lagged, still falling behind | +1.18 (+0.8 to +1.6) | +0.29 (+0.1 to +0.5) | +0.23 (+0.0 to +0.4) | +5.0 | 2821 | |
| lagged, gap narrowing | +0.50 (+0.2 to +0.8) | +0.08 (−0.1 to +0.3) | +0.07 (−0.1 to +0.3) | +4.6 | 2603 | |
| catching up (turned over a year earlier) | −0.81 (−1.1 to −0.5) | +0.03 (−0.2 to +0.2) | +0.07 (−0.1 to +0.3) | +4.4 | 1951 | |
| turned up without a long lag | −0.64 (−0.8 to −0.5) | −0.24 (−0.4 to −0.1) | −0.19 (−0.3 to −0.1) | +4.1 | 3567 | |
| leaders turning down | +0.81 (+0.6 to +1.0) | +0.08 (−0.1 to +0.2) | −0.03 (−0.2 to +0.2) | +4.3 | 3193 |
With a 10-year pent-up window: rule +0.06 (−0.1 to +0.2; 1777 metro-years); all turning up +0.01 (−0.2 to +0.2; 2260).
Zillow Home Value Index (year-end) (5-year pent-up window)
| Selection | 1-yr vs region, pp/yr | 3-yr vs region, pp/yr | 5-yr vs region, pp/yr | 3-yr absolute, %/yr | Metro-years | |
|---|---|---|---|---|---|---|
| turning-point rule | +0.88 (+0.5 to +1.3) | +0.87 (+0.2 to +1.5) | +1.04 (+0.4 to +1.7) | +7.1 | 954 | |
| turning-point rule | first half (2006–2014) | +1.47 (—) | +1.55 (—) | +1.57 (—) | +5.8 | 301 |
| turning-point rule | second half (2015–2022) | +0.56 (+0.2 to +1.1) | +0.56 (−0.0 to +1.2) | +0.83 (—) | +7.8 | 653 |
| random selection (null) | +0.02 (−0.1 to +0.1) | −0.01 (−0.1 to +0.1) | −0.01 (−0.1 to +0.1) | — | 954 | |
| turning up | +0.76 (+0.3 to +1.2) | +0.72 (+0.3 to +1.2) | +0.78 (+0.3 to +1.4) | +4.7 | 1391 | |
| turning up | region beating the nation | +0.48 (+0.0 to +1.1) | +0.50 (−0.1 to +1.3) | +0.60 (−0.0 to +1.6) | +4.8 | 706 |
| turning up | divergence settled after a peak | +0.83 (+0.5 to +1.3) | +0.73 (+0.1 to +1.4) | +0.90 (+0.2 to +1.5) | +6.1 | 434 |
| turning up | region or nation falling | +0.47 (−0.4 to +1.7) | +0.39 (+0.0 to +1.0) | +0.33 (+0.3 to +0.6) | −0.6 | 437 |
| turning up | stronger half | +0.80 (+0.3 to +1.3) | +0.80 (+0.3 to +1.3) | +0.88 (+0.2 to +1.6) | +5.3 | 1209 |
| turning up | weaker half | +0.60 (+0.2 to +1.0) | +0.35 (+0.1 to +0.6) | +0.30 (+0.2 to +0.4) | +2.1 | 666 |
| lagged | +0.10 (−0.3 to +0.5) | +0.33 (−0.1 to +0.8) | +0.49 (−0.1 to +1.2) | +4.4 | 1605 | |
| lagged, still falling behind | −0.69 (−1.1 to −0.2) | −0.16 (−0.7 to +0.4) | +0.18 (−0.4 to +0.8) | +4.2 | 1306 | |
| lagged, gap narrowing | −0.61 (−1.1 to −0.2) | −0.05 (−0.5 to +0.4) | +0.30 (−0.1 to +0.8) | +4.4 | 1325 | |
| catching up (turned over a year earlier) | +1.39 (+1.0 to +1.8) | +1.20 (+0.6 to +1.8) | +0.99 (+0.1 to +1.8) | +5.1 | 1051 | |
| turned up without a long lag | +0.42 (+0.2 to +0.7) | +0.19 (+0.0 to +0.4) | +0.11 (−0.0 to +0.2) | +4.2 | 1590 | |
| leaders turning down | −0.60 (−1.2 to −0.1) | −0.67 (−1.1 to −0.3) | −0.64 (−0.8 to −0.4) | +3.2 | 1386 |
With a 10-year pent-up window: rule +1.20 (+0.5 to +1.8; 761 metro-years); all turning up +1.13 (+0.5 to +1.8; 850).
The sources disagree. Over 3 years the rule returned −0.20 pp a year with FHFA repeat-sales index (annual) and +0.87 pp a year with Zillow Home Value Index (year-end). Zillow's index is smoothed across months and neighboring areas, which carries a ZIP's recent direction forward and makes turns look persistent; FHFA's index is unsmoothed but noisier, and a noisy year tends to reverse the next. A rule that holds in one source and not the other is not a reliable signal, which is why both are required.
Zillow's index is smoothed, which carries short-term momentum from one year into the next; one-year Zillow results overstate persistence, so the test uses three years and FHFA's unsmoothed repeat-sales index as the cross-check. The ordering used for lists ("stronger half") is tested here too; if the stronger half did no better, the ordering is presentation, not evidence.
F. Neighborhood fundamentals: new businesses and rising incomes
Business establishments (Census ZIP Codes Business Patterns, from 1994; restaurants, cafés and bars from 1998) and incomes (IRS statistics by ZIP: adjusted gross income per tax return and the share of returns over $100,000, from tax year 2011). At each year end a ZIP's 3-year growth is compared with its region's, using only years already published: business data 2 years old, IRS data 3 years old. Minimums: 20 establishments (5 restaurants and bars, 500 tax returns) at the start of the window. Outcomes are changes in relative price, as in T.
FHFA repeat-sales index (annual) (5-year pent-up window)
| Selection | 1-yr vs region, pp/yr | 3-yr vs region, pp/yr | 5-yr vs region, pp/yr | 3-yr absolute, %/yr | Metro-years | |
|---|---|---|---|---|---|---|
| turn confirmed by new businesses | −0.99 (−1.2 to −0.7) | −0.22 (−0.4 to +0.0) | −0.03 (−0.1 to +0.1) | +5.8 | 1306 | |
| turn confirmed by new businesses | first half (1999–2010) | −0.74 (−1.1 to −0.5) | −0.09 (−0.4 to +0.3) | −0.05 (−0.2 to +0.0) | +4.0 | 529 |
| turn confirmed by new businesses | second half (2011–2022) | −1.13 (−1.4 to −0.8) | −0.31 (−0.5 to −0.2) | −0.01 (−0.1 to +0.1) | +7.1 | 777 |
| random selection (null) | −0.03 (−0.2 to +0.1) | −0.08 (−0.2 to +0.0) | −0.08 (−0.1 to −0.0) | — | 1321 | |
| turn confirmed by businesses and incomes | −1.21 (−1.9 to −0.5) | −0.53 (—) | −0.24 (—) | +7.5 | 352 | |
| turning-point rule, business data available | −0.90 (−1.1 to −0.7) | −0.15 (−0.3 to +0.1) | +0.00 (−0.1 to +0.1) | +6.0 | 1560 | |
| turning-point rule, businesses growing slower | −0.95 (−1.2 to −0.6) | −0.05 (−0.3 to +0.3) | +0.10 (+0.0 to +0.2) | +6.1 | 1341 | |
| lagged, businesses growing faster | −0.09 (−0.4 to +0.2) | +0.02 (−0.2 to +0.2) | +0.09 (−0.0 to +0.2) | +4.6 | 2109 | |
| lagged, businesses growing slower | +0.07 (−0.2 to +0.4) | +0.12 (−0.1 to +0.4) | +0.15 (+0.0 to +0.3) | +4.7 | 2126 | |
| lagged, restaurants and bars growing faster | +0.10 (−0.2 to +0.4) | +0.17 (−0.1 to +0.4) | +0.24 (+0.1 to +0.4) | +4.4 | 1694 | |
| lagged, incomes growing faster | −0.42 (−1.0 to +0.2) | −0.30 (—) | −0.24 (—) | +8.1 | 547 | |
| lagged, high earners rising faster | −0.45 (−1.0 to +0.1) | −0.38 (—) | −0.33 (—) | +8.0 | 554 | |
| lagged, businesses and incomes both faster | −0.35 (−0.9 to +0.4) | −0.28 (—) | −0.21 (—) | +8.0 | 487 | |
| all ZIPs, businesses growing faster | −0.05 (−0.1 to +0.0) | −0.13 (−0.2 to −0.0) | −0.15 (−0.2 to −0.1) | +4.4 | 2320 | |
| all ZIPs, businesses growing slower | +0.06 (−0.0 to +0.1) | −0.01 (−0.1 to +0.1) | −0.02 (−0.2 to +0.1) | +4.5 | 2320 | |
| all ZIPs, incomes growing faster | −0.16 (−0.4 to +0.1) | −0.28 (—) | −0.33 (—) | +8.0 | 582 | |
| all ZIPs, incomes growing slower | +0.12 (+0.0 to +0.2) | +0.09 (—) | +0.11 (—) | +8.4 | 581 |
Zillow Home Value Index (year-end) (5-year pent-up window)
| Selection | 1-yr vs region, pp/yr | 3-yr vs region, pp/yr | 5-yr vs region, pp/yr | 3-yr absolute, %/yr | Metro-years | |
|---|---|---|---|---|---|---|
| turn confirmed by new businesses | +0.80 (+0.4 to +1.3) | +0.86 (+0.3 to +1.5) | +1.11 (+0.5 to +1.7) | +7.0 | 709 | |
| turn confirmed by new businesses | first half (2006–2014) | +1.45 (—) | +1.70 (—) | +1.78 (—) | +6.0 | 231 |
| turn confirmed by new businesses | second half (2015–2022) | +0.45 (+0.2 to +0.8) | +0.45 (+0.1 to +0.9) | +0.81 (—) | +7.4 | 478 |
| random selection (null) | −0.02 (−0.2 to +0.1) | −0.02 (−0.2 to +0.1) | −0.01 (−0.2 to +0.1) | — | 709 | |
| turn confirmed by businesses and incomes | +0.38 (−0.1 to +0.9) | +0.34 (—) | +0.55 (—) | +7.2 | 251 | |
| turning-point rule, business data available | +0.90 (+0.5 to +1.3) | +0.90 (+0.3 to +1.5) | +1.06 (+0.4 to +1.7) | +7.1 | 942 | |
| turning-point rule, businesses growing slower | +0.94 (+0.6 to +1.5) | +0.96 (+0.3 to +1.5) | +1.08 (+0.4 to +1.7) | +7.1 | 820 | |
| lagged, businesses growing faster | +0.12 (−0.3 to +0.6) | +0.40 (+0.0 to +0.9) | +0.57 (+0.2 to +1.1) | +4.5 | 1506 | |
| lagged, businesses growing slower | +0.08 (−0.4 to +0.5) | +0.28 (−0.3 to +0.8) | +0.45 (−0.3 to +1.3) | +4.4 | 1551 | |
| lagged, restaurants and bars growing faster | +0.06 (−0.4 to +0.5) | +0.39 (−0.0 to +1.0) | +0.57 (+0.1 to +1.2) | +4.5 | 1485 | |
| lagged, incomes growing faster | −0.43 (−0.7 to −0.1) | −0.32 (—) | −0.20 (—) | +7.2 | 547 | |
| lagged, high earners rising faster | −0.44 (−0.8 to −0.1) | −0.30 (—) | −0.22 (—) | +7.3 | 561 | |
| lagged, businesses and incomes both faster | −0.49 (−1.0 to +0.0) | −0.27 (—) | −0.12 (—) | +7.3 | 506 | |
| all ZIPs, businesses growing faster | +0.04 (−0.1 to +0.2) | −0.03 (−0.2 to +0.2) | −0.07 (−0.3 to +0.2) | +3.9 | 1620 | |
| all ZIPs, businesses growing slower | +0.05 (−0.1 to +0.2) | −0.01 (−0.2 to +0.2) | −0.01 (−0.2 to +0.2) | +4.0 | 1620 | |
| all ZIPs, incomes growing faster | −0.10 (−0.4 to +0.2) | −0.24 (—) | −0.28 (—) | +7.3 | 582 | |
| all ZIPs, incomes growing slower | +0.11 (−0.0 to +0.3) | +0.14 (—) | +0.12 (—) | +7.8 | 581 |
Business growth did not sharpen the turn. Over 3 years, turns where establishments grew faster than the region did no better than turns where they grew slower (FHFA repeat-sales index (annual): −0.22 vs −0.05 pp a year; Zillow Home Value Index (year-end): +0.86 vs +0.96 pp a year).
Rising incomes came too late. Lagging ZIP codes whose income per tax return had been growing faster than their region's went on to trail it over 3 years (FHFA repeat-sales index (annual): −0.30; Zillow Home Value Index (year-end): −0.32 pp a year; too few start years for intervals). One reading, not tested here: by the time tax data are published, about three years on, the change is already in prices.
Business counts are by USPS ZIP code and are matched to ZIP Code Tabulation Areas; PO-box ZIPs drop out. IRS income is per tax return, not per household, and the $100,000 line is nominal; comparing each ZIP with its own region removes the national drift. Income results start in 2014 (tax year 2011 plus the publication lag), so they cover few start years and no full housing cycle.
Strategy comparison
Predetermined selections at each start year (5-year lookback). Absolute and metro-relative annualized appreciation; volatility is the standard deviation of metro-year outcomes; "down years" is the share of metro-year selections whose price fell. Appreciation only: no rent, costs, taxes or financing.
FHFA repeat-sales index (annual)
| Strategy | 5-yr absolute, %/yr | 5-yr vs metro, pp/yr | 95% interval | Volatility | Down | Metro-years | No outcome data |
|---|---|---|---|---|---|---|---|
| metro benchmark | +4.2 | 0.00 | 0.0 to 0.0 | 0.00 | 16% | 3507 | — |
| cheapest quartile | +4.3 | +0.14 | −0.3 to +0.7 | 1.16 | 24% | 2010 | 2% |
| strongest prior quartile | +4.1 | −0.06 | −0.3 to +0.1 | 1.02 | 16% | 3507 | 1% |
| weakest prior quartile | +4.3 | +0.16 | +0.0 to +0.3 | 0.88 | 15% | 3507 | 2% |
| laggards near affluent ZIPs | +4.3 | +0.25 | −0.0 to +0.6 | 1.44 | 21% | 1747 | 2% |
| convergence rule | +6.4 | +1.09 | +0.6 to +1.5 | 1.91 | 15% | 403 | 2% |
| random selection (null) | — | −0.02 | −0.0 to −0.0 | 0.01 | — | 3507 | — |
Zillow Home Value Index (year-end)
| Strategy | 5-yr absolute, %/yr | 5-yr vs metro, pp/yr | 95% interval | Volatility | Down | Metro-years | No outcome data |
|---|---|---|---|---|---|---|---|
| metro benchmark | +4.1 | 0.00 | 0.0 to 0.0 | 0.00 | 23% | 1524 | — |
| cheapest quartile | +4.7 | +0.64 | −0.4 to +1.6 | 1.85 | 25% | 1524 | 0% |
| strongest prior quartile | +4.0 | −0.06 | −0.4 to +0.3 | 1.37 | 22% | 1524 | 0% |
| weakest prior quartile | +4.4 | +0.29 | −0.0 to +0.7 | 1.34 | 25% | 1524 | 0% |
| laggards near affluent ZIPs | +4.4 | +0.28 | −0.2 to +0.8 | 1.95 | 24% | 1386 | 0% |
| convergence rule | +7.7 | +1.26 | +0.3 to +1.9 | 2.35 | 3% | 293 | 0% |
| random selection (null) | — | +0.01 | −0.0 to +0.0 | 0.02 | — | 1524 | — |
How much of the U.S. do 10,000 ZIP codes cover?
| Measure | Share in the 10,000 most populous ZIP codes |
|---|---|
| Population | 85% |
| Adjusted gross income (IRS, 2022) | 87% |
| Value of owner-occupied homes (ACS) | 87% |
| Housing units | 83% |
| Renter-occupied units | 90% |
- Ranked by ACS 2020–2024 population across 33,772 ZIP Code Tabulation Areas (Puerto Rico included in the Census totals).
- IRS adjusted gross income is matched by ZIP code to ZCTAs; the IRS omits ZIPs with very few returns, so its national total is slightly understated.
- Owner-occupied home value is the ACS aggregate of owners' own estimates. Income, home value and household wealth are different things; none of these shares measures total household wealth.
Limitations
- House-price appreciation is not investor return: rent, vacancy, operating and transaction costs, taxes and financing are not included.
- Zillow's ZHVI is smoothed and its history reflects Zillow's current methodology; FHFA's ZIP5 index is developmental and includes appraisals. Both sources revise history, so a rerun can differ; each run records the SHA-256 of its input files.
- ZIP Code Tabulation Areas approximate USPS ZIP codes; ZIPs are assigned to the metro of the county holding most of their land. Housing-unit weights come from the latest ACS, a small look-ahead for weighted statistics (equal-weighted results avoid it).
- Many metro-years overlap; intervals account for this with block bootstrapping over start years, but regional booms and busts are shared, so results are less independent than the counts suggest.
Reproducibility
Run convergence-v3-2026-10-09 (convergence-v3), computed October 9, 2026. Every rule and threshold is stored with the run; inputs are identified by release and file hash below. Re-running the same code on the same files reproduces these numbers exactly.
| census_acs_zcta | ACS 5-year 2024 and 2014 | API |
| census_gazetteer_zcta_2024 | 2024 | 9c2630c5b2415a2e… |
| census_zbp_zip | ZIP Codes Business Patterns 1994–2023 | API |
| census_zcta_county_2020 | 2020 | 3ed41278d637dc24… |
| fhfa_hpi_zip5 | through 2025 | 3d3da1123aea4222… |
| irs_soi_zip | IRS SOI ZIP data, tax years 2011–2022 | API |
| nber_cbsa_county_2023 | 2023 | a444e048c52fa5c9… |
| zillow_zhvi_zip | Zillow ZHVI through 2026-08 | 9dcd2793d97e5f60… |
| zillow_zori_zip | Zillow ZHVI through 2026-08 | a60963f429dbacf9… |
Research describes what happened historically. It is not a forecast, investment advice or an offer of any kind. Methodology · Market divergence dashboard
